Basic Functions of Language

BASIC FUNCTIONS OF LANGUAGE,
READING AND
READING DISABILITY
NEUROPSYCHOLOGY AND COGNITION
VOLUME 20
Series Editor:
R. Malatesha Joshi, College of Education, Texas A&M University, U.S.A.
Advisory Board:
Torleiv Hoien, Center for Dyslexia Research, Norway
George Hynd, University of Georgia, U.S.A.
C.K. Leong, University of Saskatchewan, Canada
John Marshall, University of Oxford, U.K.
Gabriele Miceli, Universita Cattolica del Sacro Cuore, Italy
Loraine Obler, City University ofNew York, U.S.A.
Pieter Reitsma, Paedologisch Instituut Amsterdam, The Netherlands
The purpose of the Neuropsychology and Cognition series is to bring out volumes
that promote understanding in topics relating brain and behavior. It is intended for
use by both clinicians and research scientists in the fields of neuropsychology,
cognitive psychology, psycholinguistics, speech and hearing, as well as education.
Examples of topics to be covered in this series would relate to memory, language
acquisition and breakdown, reading, attention, developing and aging brain. By
addressing the theoretical, empirical, and applied aspects of brain-behavior
relationships, this series will try to present the information in the files of
neuropsychology and cognition in a coherent manner.
The titles published in this series are listed at the end of this volume.
BASIC FUNCTIONS OF LANGUAGE,
READING AND
READING DISABILITY
edited by
Evelin Witruk
University of Leipzig, Germany
Angela D. Friederici
Max-Planck Institute of Cognitive Neuroscience,
Leipzig, Germany
Thomas Lachmann
University of Leipzig, Germany
SPRINGER SCIENCE+BUSINESS MEDIA, LLC

Library of Congress Cataloging-in-Publication Data
Basic functions of language, reading and reading disability / edited by Evelin Witruk,
Angela D. Friederici, Thomas Lachmann.
p. cm.—(Neuropsychology and cognition)
Includes bibliographical references and index.
ISBN 978-1-4613-5350-8 ISBN 978-1-4615-1011-6 (eBook)
DOI 10.1007/978-1-4615-1011-6
1. Psycholinguistics. 2. Language acquisition. 3. Comprehension. 4. Reading,
Psychology of. 5. Reading disability. I. Witruk, Evelin. II. Friederici, Angela D. III.
Lachmann, Thomas. IV. Series.
P37 .B337 2002
401'.9—dc21 2002016231
Copyright © 2002 by Springer Science+Business Media New York

Originally published by Kluwer Academic Publishers in 2002
Softcover reprint of the hardcover 1st edition 2002
A l l rights reserved. No part of this work may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical,
photocopying, microfilming, recording, or otherwise, without the written
permission from the Publisher, with the exception of any material supplied
specifically for the purpose of being entered and executed on a computer system,
for exclusive use by the purchaser of the work.
Printed on acid-free paper.

TABLE OF CONTENTS
PREFACE IX
INTRODUCTION
INTRODUCTION TO BASIC FUNCTIONS OF LANGUAGE, 3

READING, AND READING DISABILITY
T. Lachmann, A. D. Friederici & E. Witruk
FROM LANGUAGE TO READING AND READING DISABILITY: 9

Cognitive Functions and their Neural Basis
A. D. Friederici & T. Lachmann
BASIC FUNCTIONS OF LANGUAGE ACQUISITION

AND LANGUAGE COMPREHENSION
CONTEXT EFFECTS ON LEXICAL PROCESSING DURING 25

AUDITORY SENTENCE COMPREHENSION
On the Time-Course and Neurological Bases of a Basic
Comprehension Process
D. Swinney and T. Love
MEASURING THE NEURAL DYNAMICS OF LANGUAGE 41

COMPREHENSION PROCESSES
J. D. Saddy & P. beim Graben
A MODEL OF LEARNING SYNTACTIC COMPREHENSION 61

FOR NATURAL AND ARTIFICIAL GRAMMARS
P. F Dominey
THE ACQUISITION OF VERB PLACEMENT IN GERMAN: 79

ANew Look
J. Weissenborn
MERGE AS A BASIC MECHANISM OF LANGUAGE: 105

Evidence from Language Acquisition
S. M Powers
BASIC FUNCTIONS OF READING
COGNITIVE PROCESSES AND EYE MOVEMENTS DURING 121

READING
M S. Starr, G. Kambe, B. Miller & K. Rayner
THE ROLE OF ATTENTION AND SPATIAL SELECTION 137

IN FLUENT READING
R. Radach, A. Inhoff & D. Heller
THE EFFECT OF FOCUSING ON A SENTENCE IN JAPANESE 155

READING SPAN TEST
M Osaka & N Osaka
BASIC FUNCTIONS OF READING DISABILITY
READING DISABILITY AS A DEFICIT IN FUNCTIONAL 165

COORDINATION
T. Lachmann
THE NEUROBIOLOGY OF READING DIFFICULTIES 199

J. Stein
A SENSORY-LINGUISTIC APPROACH TO NORMAL AND 213

IMPAIRED READING DEVELOPMENT
J. B. Talcott & c. Witton
ANATOMY OF THE TEMPORAL PROCESSING DEFICIT IN 241
DEVELOPMENTAL DYSLEXIA
A. M Galaburda
READING DISABILITIES IN A LANGUAGE WITH TRANSPARENT 251

ORTHOGRAPHY
J. E. Jimenez Gonzalez
DYSLEXIA, THE CEREBELLUM AND PHONOLOGICAL SKILL 265

A. J. Fawcett
WORKING MEMORY IN DYSLEXIC CHILDREN: 281

How General is the Deficit?
E. Witruk, C. S.-H Ho & u. Schuster
VI
KINDERGARTEN PHONOLOGICAL AWARENESS AND RAPID 299
SERIAL NAMING AS PREDICTORS OF GRADE 2 READING AND
SPELLING
J. K. Uhry
PHONOLOGICAL RECODING PROBLEMS IN CHILDREN WITH 315

SEVERE CONGENITAL SPEECH IMPAIRMENTS:
The Importance of Production Speech
A. Dahlgren Sandberg
SUBLEXICAL AND LEXICAL PROCESSING OF YOUNG ADULTS 329

WITH LEARNING DISABILITIES AND ATTENTION DEFICIT:
Hyperactivity Disorder
N Gregg, C. Coleman, R. Stennett, M Davis, K. Nie/sen, D. Knight &
C. Hoy
THE MISMATCH NEGATIVITY AS AN INDEX OF AUDITORY 359

DYSFUNCTION IN DYSLEXIA
T. Kujala
INDEX 371
VII
PREFACE
The present book contains selected contributions from the international conference
"Basic Mechanisms of Language and Language Disorders". This conference was
held in Leipzig in September 1999, and was organized by the Department of
Psychology at the University of Leipzig in collaboration with the Max-Planck-
Institute of Cognitive Neuroscience, and the Max-Planck-Institute for Evolutionary
Anthropology. It was held to commemorate the 120th anniversary of the foundation
of the world's first institute of experimental psychology by Wilhelm Wundt in
Leipzig.
This edition examines new results from different fields of psychology and neuro-
psychology of language, reading, and reading disability. The presented book focuses
on the following main topics:
• BASIC FUNCTIONS OF LANGUAGE ACQUISITION AND

LANGUAGE COMPREHENSION
• BASIC FUNCTIONS OF READING
• BASIC FUNCTIONS OF READING DISABILITY
The title "Basic Functions of Language, Reading and Reading disability" expresses
the interdisciplinary character of the book. It aims not only at bringing together
different theoretical approaches, but also at connecting these approaches with
applied work. Since it is necessary to understand basic functions of language and
reading in order to understand reading disability, the present book strives to foster a
scientific exchange, and to promote the emergence of synergy effects between the
different fields.
The editors wish to warmly thank Katja Brendler from the University of Leipzig for
the strenuous task of formatting the chapters, as well as for her organizational help
in editing this volume. Many thanks are also owed to the anonymous reviewers
committed by Kluwer Academic Publishers for their constructive remarks to the
chapters.
Evelin Witruk
Angela D. Friederici
Thomas Lachmann
Leipzig, January 2002
IX
SECTION I
INTRODUCTION
T. LACHMANN, A. D. FRIEDERICI, & E. WITRUK
INTRODUCTION TO BASIC FUNCTIONS OF

LANGUAGE, READING, AND READING
DISABILITY.
The past decade has seen enormous growth in our knowledge of the basic functions
of language processing as well as their disorders. Theoretical, behavioral and
neurophysiological, as well as neuroanatomical studies have helped to advance our
knowledge in the field. On the one hand, the development of different
methodologies and their application to the investigation of normal and impaired
language processing has lead to new insights into the underlying mechanisms. On
the other hand, it appears that novel data have forced the reconceptualization of
some of the existing views.
The selection of the chapters in the present book was intended to provide a broad
overview of the most relevant, sometimes controversial, issues and theories
discussed in the field of language and language disorders. A particular emphasis is
given to reading and reading disabilities since this research domain has received
considerable attention during the recent years. Papers on lexical and syntactic
processes in the adult and the developing language system provide complementary
information necessary to create an adequate picture of the basic mechanisms of
language. The picture we present in this book is by no means complete. However, it
sketches the most important aspects of the system under investigation.
1. THE STRUCTURE OF THE BOOK
The book consists of four sections. The paper by Friederici and Lachmann in the
introduction section aims to introduce in the topic of the book and seeks to point out
that reading comprehension is the end product of the coordination of a number of
subfunctions which involve the mediation from visual input to phonological
representation and processes of language comprehension modulated by aspects of
working memory and attention. It is argued that to understand the phenomenon of
reading disability, it is necessary to understand the nature of these sub functions and
their coordination in normal reading (see also Rayner, 1993). In general, the
language system is seen as primary, as reading optimally builds on spoken language
(Perfetti & Sandak, 2000). Following this line of argumentation, the second section
presents papers on basic functions of language acquisition and language
comprehension. The third section deals with principle aspects of normal reading,
E. Witruk, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading,

and Reading Disability, 3--7.
@ 2002 Kluwer Academic Publishers.
4 T. LACHMANN, A. D. FRIEDERICI, & E. WITRUK
followed by papers of the fourth and biggest section, which tries to cover different
theoretical approaches and empirical studies to reading disability.
2. BASIC FUNCTIONS OF LANGUAGE ACQUISITION AND LANGUAGE

COMPREHENSION
This section contains a selection of compelling papers which investigate basic

mechanisms of lexical and syntactic processes during language comprehension and
production in the adult and the learning system. The paper by Swinney and Love
provides an integrative view of lexical access and lexical integration during auditory
sentence comprehension in adults. Special emphasis is given to each hemisphere's
contribution to language processes: while the left hemisphere is considered to be
responsible for fast on-line language processes, the right hemisphere is thought to
support slower processes involving contextually guided and repair processes. The
paper by Saddy and Beim Graben investigates the processing of morphosyntactic
information during adult sentence comprehension in German using
electrophysiological measures. Applying new and innovating analysis techniques,
they are able to demonstrate the brain's response to morphosyntactic markers in
each of three time windows predefined by a current neurocognitive model of
language comprehension. The paper by Dominey applies computational modeling as
an approach to investigate the use of word order and function word to assign
thematic roles during comprehension. It is demonstrated that the inability to
comprehend non-canonical sentences is due to a failure of the application of abstract
transformational rules - be it in natural or artificial grammars. The paper by
Weissenborn presents convincing data in support of the view that basic syntax
parameters are already installed early during language acquisition. He demonstrates
that children as young as two years old are in control of the knowledge of verb
placement in German main and subordinate clauses. The paper by Powers also uses
data from language acquisition to show that particular mechanisms assumed by
linguistic theory, namely the mechanism Merge (Chomsky, 1995), is at work in
early childhood and can thus be considered as one of the most basic mechanisms of
language.
3. BASIC FUNCTIONS OF READING
The third section deals with principle aspects of reading. The first paper describes
the complex processes underlying normal reading. Starr, Kampe, Miller, and Rayner
provide a complete overview of eye movement studies during sentence reading
uncovering the main cognitive processing underlying reading performance. The two
following papers evaluate the influence of attention on reading and reading
comprehension. Radach, Inhoff and Heller provide a critical overview of different
models describing the role of visual selective attention during reading. Particular
emphasis is given to the time course of linguistic processing and oculomotor control
during the reading process. The paper by Osaka and Osaka evaluates the
mechanisms underlying the well known reading span test (Daneman & Carpenter,
INTRODUCTION 5
1980) applied in numerous studies on sentence processing. They show that a

variation in the focus of attention on a particular word can influence reading span.
4. BASIC FUNCTIONS OF READING DISABILITY
The fourth and biggest section presents a collection of theoretical and empirical
papers discussing or testing the most influential causal hypotheses of reading
disability.
4.1 Theoretical Approaches
In the paper by Lachmann, it is argued that reading disability results from a failure
in learning to optimize the coordination of the subfunctions involved in reading with
the consequence of errors or delays in integrating reading related information.
Within this multicausal model, so-called reversal errors, for instance, are
reinterpreted as a functional coordination deficit. In this respect, it is argued that
learning to read implies learning to treat graphemes as symbols instead of objects,
and thus to suppress visually symmetrical information in the representation of visual
symbols (Brendler & Lachmann, 2001). A failure in this suppression produces
ambiguous relations between visual and phonological information and disturbs the
functional coordination, and thus may cause problems in learning to read.
Stein presents his theory of dyslexia which holds that reading difficulties are the
consequence of the impaired development of magnocellular neurons in the human
brain. This impairment causes behavioral deficiencies which are not solely confined
to reading, but also to other cognitive domains.
In his paper, Galaburda advances the theory that children who fail to learn to
read suffer from a specific type of brain anomaly affecting low level auditory and/or
visual processing, as well as linguistic processing. As this brain anomaly is the result
of inappropriate neural migration to the cerebral cortex during development, the
paper discusses possible genetic and epigenetic influences acting during the period
of neuronal migration.
The chapter by Talcott and Witton presents a specific view on reading
development. It is called a sensory-linguistic approach, as it focuses on the interface
between orthographic and phonological information in the text, and the visual and
auditory skills necessary to extract higher level information from lower level
information computed by the auditory and visual systems. They hold that capacities
of the latter processes constraint reading comprehension.
The paper by Jimenez investigates reading disabilities in a language which, in
contrast to English, has a transparent orthography. A comparison of empirical
studies in English and Spanish suggests that the relevant factor for the definition of
dyslexic is a phonological rather than an orthographic one.
Fawcett advances a theory which considers a dysfunction of the cerebellum as an
underlying cause of dyslexia. The observed correlation between problems in
articulation, i.e. the timing of articulatory gestures, which may be caused by a
cerebellar deficit and problems in reading, are taken to support this theory.
6 T. LACHMANN, A. D. FRIEDERICI, & E. WITRUK
4.2 Empirical Studies
The chapter by Witruk, Ho and Schuster discusses three experiments conducted

with Chinese and German dyslexics. The conceptual differentiation of these studies
is based on the Baddeley's working memory model (Baddeley, 1995), with its
specific modalities for incoming and maintaining information, on Cowan's model
(1995), with regard to automatic versus controlled executive functions, as well as on
the neurophysiological approach of Goldman-Rakic (1998). The central issue of this
chapter is to discuss the importance and extent of assumed working memory
impairments in dyslexic children with the focusing on the generality versus the
specificity of these impairments. The results show the dependency of working
memory performance in dyslexic children on the language system, on the specific
type of modality, and on a specific kind of material.
Uhry presents a longitudinal study in English-speaking kindergarten children in
which phonological awareness and rapid serial naming is evaluated as a predictor
for reading and spelling abilities in Grade 2. Both measures, taken at Kindergarten,
were shown to make major contributions to later word-reading abilities and thus
serve as a good predictor.
The paper by Dahlgren Sandberg takes an interesting approach to investigate the
role of phonological recoding in reading. Children with severe expressive speech
impairment are examined in different verbal and non-verbal tasks. It is conducted
that the inability to articulate is an important factor in the reading and spelling
difficulties observed in these children.
Gregg, Coleman, Stennett, Davis, Nielsen, Knight and Hoyet present empirical
data on several language tests including spelling and reading in three groups of
English speaking adults, which suffer either from a learning disability or from an
attention deficit / hyperactivity disorder, or both. Particular factors which predict
reading abilities in these groups are discussed.
The paper by Kujala finally focuses on the neurophysiological parameters of
auditory dysfunction in dyslexia. A particular electrophysiological marker, i.e. the
mismatch negativity is proposed as an object index of auditory dysfunction in
dyslexia. The advantage of this measure lies in the fact that it can be applied very
early in life, that is, long before dyslexia occurs.
5. SUMMARY
The present collections of papers provide an interesting view on a human ability

which with the emergence of writing systems has become two-layered, namely
language. The primary language domain is auditory. It is through auditory input that
a native language is learned. Interestingly, the complex syntactic and lexical aspects
of language establish quite early during language acquisition and are shown to have
a relatively specific neuronal basis. Reading and writing is clearly secondary.
Learning of these skills is obviously accompanied by special difficulties. It appears
that most researchers in the field of dyslexia agree that the interface between time
based phonological processes and the visual system is the focus of difficulty,
although they might disagree on the underlying neuronal cause. It may not come as
INTRODUCTION 7
a surprise that recent brain imaging studies have demonstrated an increased

involvement of Broca's area, i.e., a brain area known to support phonological
processes, during reading comprehension compared to auditory comprehension
(e.g., Poldrack et ai., 1999).
6. REFERENCES
Baddeley, AD. (1995). Working memory. Oxford: Clarendon Press.
Brendler, K., & Lachmann, T. (2001). Letter reversals in the context of the Functional Coordination
Deficit Model. In E. Sommerfeld, R. Kompass, & T. Lachmann. Fechner Day 2001. Proceedings of
the International Society for Psychophysics (pp. 308-313). Lengerich, Berlin: Pabst.
Chomsky, N. (1995). The Minimalist Program. Cambridge: MIT Press.
Cowan, N. (1995). Attention and memory: An integrated framework. Oxford: Oxford University Press.
Daneman, M., & Carpenter, P. A (1980). Individual differences in working memory and reading. Journal
of Verbal Learning & Verbal Behavior, 19, 450-466.
Goldman-Rakic, P. S. (1998). The prefrontal landscape: Implications of functional architecture for
understanding human mentation and the central executive. In A C. Roberts, W. T. Robbins et al.
(Eds), The prefrontal cortex: Executive and cognitive functions (pp. 87-102). New York: Oxford
University Press.
Perfetti, C. A, & Sandak, R. (2000). Reading optimally builds on spoken language: Implications for deaf
readers. Journal of Deaf Studies & Deaf Education, 5, 32-50.
Poldrack, R. A, Wagner, A. D., Prull, M. W., Desmond, 1. E., Glover, G. H., & Gabrieli, 1. D. E. (1999).
Functional specialization for semantic and phonological processing in the left inferior prefrontal
cortex, Neuroimage, 10, 15-35.
Rayner, K. (1993). Visual processes in reading: Directions for research and theory. In D. M. Willows, R.
S. Kruk, & E. Corcos (Eds.), Visual Processes in reading and reading disability (pp. 475-480).
Hillsdale, N.J.: Erlbaum.
A. D. FRIEDERICI & T. LACHMANN
FROM LANGUAGE TO READING AND READING

DISABILITY:
Cognitive Functions and their Neural Basis
Abstract. Reading comprehension is seen as an end product of a number of subprocesses involving the
mediation from visual input to phonological representation, in particular, and processes of language
comprehension in general which in turn are modulated by aspects of working memory and attention. The
neural basis of each of these functionally different subprocesses constituting normal reading is presented.
Empirical findings and theories of reading disability are discussed with respect to a possible impairment
of one or more of these subprocesses.
1. INTRODUCTION
For millions of years, humans have spoken and understood language. Their ability
to read and write, however, has been established only in more recent times. From a
psycho linguistic perspective, reading must be considered as a secondary process
which apart from the visual identification of the word form relies in its consecutive
processes on the primary language system (e.g., Perfetti, 1998; Perfetti, Bell &
Delaney, 1998; Perfetti & Sandak, 2000). The language system provides the
phonological, morphological, semantic and syntactic information over which
comprehension processes operate. The processes and components that are specific to
reading as compared to auditory language comprehension are (i) the identification
of visual features relevant to define letters, (ii) the identification of a visual word
form, (iii) and transcoding from orthography to phonology. Thus, there are a
number of functionally distinct subprocesses which must be intact to guarantee
normal reading. Any research on impaired reading must keep the complexity of the
process in mind. The present book discusses the relevant processes and subprocesses
involved in reading with respect to their functional relevance and partly with respect
to their neural basis allowing the reader to locate past and present empirical findings
and theories.
In the following, we will first outline the basic functions of normal reading and
their neural basis. We will see that the process of normal reading involves the
fundamental processes known to constitute the comprehension of spoken sentences
plus those basic processes that allow the mediation from visual input to the
phonological representation on which higher order linguistic processes are built.
Moreover, it will be shown that cognitive processes outside the language domain,
namely working memory and attention affect reading processes. After laying these

and Reading Disability, 9-21.
© 2002 Kluwer Academic Publishers.
10 A. D. FRIEDERICI & T. LACHMANN
grounds we will try to specify at which of these different processing level and in
which domains, normal reading might derail.
2. BASIC FUNCTIONS OF LANGUAGE COMPREHENSION
The initial level that becomes active within the language system proper is the
phonological processing level. At this level the phonological information is
assembled and a phonological representation is built up. This phonological
representation allows access to the lexicon and to identify a particular lexical entry
that matches the phonological representation (Marslen-Wilson & Warren, 1994;
Pisoni & Luce, 1987). At the lexical level, semantic and morphological information,
as well as syntactic information (i.e., word class, argument structure), become
available. This information is the basis for the relevant processes at the sentence
level. At this level the syntactic structure is built up and thematic relations between
the different elements in the sentences are assigned to achieve a sentence
representation (Frazier, 1987; Just & Carpenter, 1992, see also Swinney & Love,
this volume; Saddy, this volume). The further integration of such a sentence
representation into the world knowledge may be viewed as the process of
understanding in its most general use of the term. Integration of a sentence
representation into prior discourse is achieved at the discourse or text level (Kintsch,
1988; Noordman & Vonk, 1999).
Each of these processes has been identified to correlate with particular brain
areas within a network supporting language comprehension. In brief, these brain
areas can be specified as follows. Phonological processes have been located at an
integrated network consisting of the posterior part of the superior temporal
gyrus/sulcus and the inferior frontal gyrus (Brodmann Area (BA) 44), (Demonet et
aI., 1992; Zatorre, Evans, Meyer, & Gjedde, 1992; Price et aI, 1992). Lexical-
semantic processes have been reported to correlate with activation in the middle
temporal gyrus and the inferior frontal gyrus (BA 45/47), (Fiez, 1997; Kapur et aI.,
1994; Thompson-Schill, D'Esposito, Aquirre, & Farah, 1997). Syntactic processes
were shown to involve the anterior part of the superior temporal gyrus and the
inferior frontal gyrus (inferior part of BA 44 and the adjacent frontal operculum),
(Caplan, Alpert, & Waters, 1998; Friederici, Meyer, & von Cramon, 2000; Just,
Carpenter, Keller, Eddy, & Thulborn, 1996; Stromswold, Caplan, Alpert, & Rauch,
1996). Discourse level processes, furthermore, involve brain areas in the frontal
median wall (Ferstl & von Cramon, 2001).
Although all these areas show their activation predominantly in the left
hemisphere, right homologue areas are reported to be simultaneously active, in
particular, at the sentential level when the input is presented auditorily. Recent
studies have shown that the increased right hemisphere activation is partly due to
prosodic information present in spoken sentences (Meyer, Alter, & Friederici, in
press). Furthermore, a comparison between auditory and visual studies at the
sentence level reveals that the left inferior frontal gyrus is always active when the
input is visual, but also, albeit less consistently when it is auditory (Michael, Keller,
Carpenter, & Just, 2001). This seems to suggest that during sentence reading
phonological processes (located in the superior part of BA 44) become active.
FROM LANGUAGE TO READING AND READING DISABILITY 11
3. BASIC FUNCTIONS OF READING
In most general terms reading can be characterized as language comprehension,

including all aspects introduced above, plus visual decoding (Everatt et aI., 1999).
However, reading cannot be defined as simply the sum, but rather as an interaction
of these components (e.g., Perfetti & Sandak, 2000). Furthermore, the visual
processes involved in reading are not identical to those of visual object recognition.
Instead, they are highly specialized to a fast and accurate desymbolization of visual
icons (Deacon, 2000). In this respect, reading can be characterized as a highly
automated desymbolization process. Different levels of the visual domain are
involved in this desymbolization process.
At the visual feature level, different features relevant for the identification of a
letter such as lines, angles and contours have to be processed to allow the activation
of letters (Massaro, 1998). At the visual word from level, the identification of visual
word forms takes place (Seymour & Evans, 1993). Dual Route Models of word
reading assume a direct route from the visual word form to the word's meaning for
high frequency words and a second route for low frequency words (Coltheart &
Rastie, 1994; Ellis, 1984; Joubert & Lecours, 2000; Samuels, LaBerge, & Bremer,
1978; see also Jimenez-Gonzalez, this volume). This latter route to the lexicon
proceeds via a grapheme-to-phoneme conversion rule system in which individual
letters are mapped onto phonological units before these are assembled into a
phonological word form. All models agree that the latter system certainly has to be
activated during pseudoword reading. However, the different models are less
univocal as to whether low frequency words are the only real words processed by
means of grapheme-to-phoneme conversion rules (Coltheart & Rastie, 1994) or
whether the identification of a word always involves the immediate co-activation of
graphemic and phonological constituents (Booth, Perfetti & MacWhinney, 1999).
The brain areas which were identified to be involved in word reading are the
following. Brain areas in the occipital lobe specified to detect line orientation and
contours are involved in the processing of the visual features of letters. Within the
occipito-temporal brain areas the fusiform gyrus is activated when reading between
words and pronounceable non-words is compared (Fiebach, Schlesewsky, &
Friederici, in press; Herbster, Mintun, Nebes, & Becker, 1997). The middle part of
the fusiform gyrus has been taken to be the locus of the visual word form area, in
which a perceptually invariant higher-order orthographic unit (Le., a visual word
form) is computed from the visual input (Cohen et aI., 2000). Note that this brain
area in the fusiform gyrus, although relevant for reading, is not reading-specific as it
is also activated during visual face recognition (Haxby, Hoffman, & Gobbini, 2000),
and more generally when experts process highly familiar objects of the category of
their expertise (Gauthier, Skudlarski, Gore, & Anderson, 2000). Fiebach et al. (in
press) proposed that this notion of a category-specific recognition component could
be extended to written words - a category for which any normal literate adult is an
expert.
The left inferior frontal gyrus and the anterior insula have been associated with
phonological processing, phonological retrieval and phonemic analysis (for review
see Fiez & Petersen, 1998). Activation of this area was not only observed during
pronuncIatIOn tasks, but also during lexical decision (Fiebach et al., in press).
Increased activation of the superior part of the left pars opercularis (BA 44) was
observed for low frequency words and for pseudo-words compared to real words
suggesting that lexical access was mediated by phonological information for the
former word types. Thus, it appears that this area can be viewed to subserve
grapheme-to-phoneme conversion. Note, however, that this area is involved in
phonological processes both during production and perception. The left middle
temporal gyrus known to be involved in auditory language processing has been
observed as part of the neural network of word reading. The combined data from
different studies suggest that activation in this area can be associated with the
activation of phonological word forms (Hagoort, Brown, & Osterhout, 1999; Price
et al., 1994) which are part of the word's lexical entry.
Price (2001) recently pointed out that the combined results from a number of
brain imaging studies indicate that there are no brain areas that are specific to
reading. Rather it appears that learning to read involves "establishing associate
connections between object processing areas in the visual cortex and speech
processing areas in temporal and frontal cortices" (Price, 2001). This result is
compatible with the notion that from an evolutionary point of view, reading is a
secondary system which is not associated with areas specific to reading, but which
rather recruits brain areas primarily subserving other functions.
4. ASPECTS OF MEMORY AND ATTENTION DURING READING
Besides these fundamental processes of reading and language comprehension,

cognitive processes outside these domains appear to be necessary to guarantee
normal reading, namely working memory capacity and attention.
Working memory has been identified as a main factor determining reading
comprehension (Just & Carpenter, 1992; MacDonald, Just, & Carpenter, 1992; but
see Waters & Caplan, 1996). This view is based on a large number of behavioral
studies showing that sentence comprehension performance decreases when the
individual working memory capacity is low, as assessed by the Reading Span Test
or when concurrent working memory capacity is high (e.g., Carpenter, Miyake, &
Just, 1994). Event-related brain potential studies support the view that syntactic
comprehension is constrained by individual working memory constraints (Friederici,
Steinhauer, Mecklinger, & Meyer, 1998) and by concurrent working memory load
(Vos, Gunter, Schriefers, & Friederici, 2001).
The underlying factor for the performance difference in high and low span
readers, however, is defined differently in these studies. While Carpenter and Just
(1992) claim that low span readers' difficulty to process syntactically ambiguous
sentences is due to an inability to keep the two possible underlying syntactic
structures active, others hold that their problem lies in the inability to commit
themselves early to the preferred structure and in order to keep the working memory
load low (Friederici et al., 1998).
Whatever the valid interpretation of the underlying processes is, the observation
is that the correlation between the individual working memory capacity and of
reading comprehension ability is high. A correlation between working memory and
reading has also been demonstrated for the level of lexical processing. It was found
that the resolution of lexical ambiguity is .influenced by working memory
(Gernsbacher & Faust, 1991; Miyake, Just, & Carpenter, 1994). As mentioned
before, two different explanations have been put forward. The first states that low
span readers are not able to keep both reading of and ambiguous word active
(Miyake et aI., 1994), while the other claims that low span readers are unable to
actively suppress the irrelevant reading.
This correlation between lexical aspects of comprehension and working memory
has not been evaluated using electrophysiological measures so far. However,
working memory for single words has been investigated in recent brain imaging
studies and so has the relation between working memory and syntactic processing.
Aspects of lexical-semantic working memory were shown to be correlated with
activation in left temporal regions and in the left inferior frontal gyrus (BA 45/47),
(Paulesu, Frith, & Frackowiak, 1993; Gabrieli, Brewer, & Poldrack, 1998; Wiggs,
Weisberg, & Martin, 1999).
Aspects of syntactic working memory also activate parts of the left inferior
frontal gyrus, namely BA 44/45 (Fiebach et aI., in press). While this region was
shown to be active when syntactic complex sentences requiring working memory
resources are processed in English (Just et aI., 1996; Stromswold et aI., 1996).
Fiebach et aI. (in press) using German were able to show that this brain area
selectively reacts to working memory constraints rather than to syntactic complexity
per se. Note, however, that these two factors are compounded in natural sentences.
Thus, the combined data indicate that working memory as tested by the Reading
Span Test (Daneman & Carpenter, 1980) influences comprehension performance
during reading (see also Osaka & Osaka, this volume). Moreover, they indicate that
adjacent, but separable brain areas in the inferior frontal and temporal cortex
support lexical and syntactic aspects of working memory.
What seems noteworthy is that working memory capacity interacts with so-
called higher levels of processing during reading, namely lexical and syntactic
processing, and not with lower level processes such as visual perception and
phoneme-to-grapheme conversion. Attention on the other hand appears to be more
likely to interact with lower as well as higher level reading processes (McCarthy &
Nobre, 1993; Otten, Rugg, & Doyle, 1993; Smid, Jakob, & Heinze, 1999).
Taken together, reading is a complex process that involves lower level
perceptual processes and central language processes which interact with aspects of
attention and working memory. These processes were discussed to differ not only
functionally, but also at the level of neuronal systems. It is not unlikely that reading
disability may have its origin in each of these different systems. Therefore, it is not
surprising that a number of different theories and models of reading disability have
been proposed.
5. READING DISABILITY RESEARCH
Current theories and models of reading disability are anything but univocal. The
understanding of reading disability depends on the understanding of the reading
process as such (Rayner, 1993), the definition of dyslexia (Toth & Siegel, 1994),
and methodological approaches (see e.g., Miles & Miles, 1999).
As discussed, reading is a complex cognitive technique which requires the
coordination of a series of subfunctions which can be characterized as visual
junctions, such as configurational (feature) and orthographic (word form) analyses,
and verbal (language) junctions, such as phonological, semantic and syntactic
coding and decoding. From this point of view, models of reading disability can be
subdivided grossly into those assuming visual deficits and those assuming language-
based (linguistic including phonological) deficits as causal factors for reading
disability (cf. Vellutino, 1977; Vellutino & Scanlon, 1998).
Early models of word-blindness (Kussmaul, 1878) and strephosymbolia (Orton,
1925), which were dominant until the late 70's, are sometimes characterized to be
(somehow old-fashioned) visual deficit models (Vellutino, 1977; Vellutino &
Scanlon, 1987, 1998). However, these models, as well as recent adaptations
(Corballis & Beale, 1993; see also Brendler & Lachmann, 2001; Lachmann, this
volume), already suppose a failure in the binding of visual and
phonological/semantic representations to cause reading problems (reversals). Thus,
the reason for a failure in reading is assumed to be a failure in the coordination of
the subfunctions involved. This deficit is supposed to lead to a faulty integration of
visual and auditory information.
The influential studies of Liberman, Shankweiler, Orlando, Harris, & Bell Berti
(1971), Vellutino (1977, 1979) and Fisher, Liberman and Shankweiler (1978)
initiated the "phonological turnaround" (see Lachmann, this volume) in the field of
reading disability research. As a result, to this day most cognitive explanations of
reading disability are based on the assumption of phonological deficits within the
language processing system (Bradley & Bryant, 1983; Snowling, 2001; Vellutino &
Scanlon, 1987; see the chapters of Uhry, Gregg et aI., Jimenez-Gonzalez, Fawcett,
Dahlgren-Sandberg and Kujala, this volume). Reading is understood as a primary
linguistic skill (Liberman, 1983).
Visual models of reading disability are, for instance, those which assume a
deficit - or better a difference - (cf. Stein, this volume) in the temporal integration of
visual information, (Stanley & Hall, 1973) or more specifically, in the coordination
between the transient and the sustained pathway (Breitmeyer & Ganz, 1976;
Livingstone & Hubel, 1987) in the lateral geniculate nucleus of the visual system
(Lovegrove, Martin & Slaghuis, 1986; Lovegrove, 1993; see the chapters of Stein,
Galaburda, and Talcott & Witton, this volume). A failure of such a coordination
averts an adequate and fast setup of visual representations. However, there are also
models which assume a general temporal integration deficit, which is not only
effective to the visual system, but also within the auditory modality (e.g., Farmer &
Klein, 1995; Tallal, 1984). On the other hand, also those models have to be counted
as purely visual, which assume reading disability as a result of deficits in visual
memory, that is, not in building up, but in operating with visual representations
(Willows, Kruk, & Corcos, 1993 for review; Witruk, Ho, & Schuster, this volume).
These differences between the models point to another important distinction that
can be made between the models of reading disability. There are models assuming
that problems occur in early processing (low level deficits), and those assuming that
problems occur in late processing (higher level deficits including memory) of either
visual or verbal information, respectively (e.g., visual memory: Willows et aI., 1993,
for a review; visual perceptual: Breitmeyer, 1993, for a review; e.g., phonetic
memory: e.g., Katz, Healy, & Shankweiler, 1983; temporal integration of phonetic
information: Tallal, 1984). For instance, Tallal (1984) argues that deficits in
phonological representations may be due to a low-level deficit in rapid processing of
auditory information. Others argue resolutely against such an explanation
(Snowling, 2001). Hulme (1988) reviews evidence against the assumption of low-
level deficits within the visual domain measured by visible persistence, contrast
sensitivity and flicker rate sensitivity in disabled readers, and the interpretation as a
difference in functioning of the transient system (see the chapters of Stein,
Galaburda, and Talcott & Witton, this volume). In some cases, the discussion about
late versus early deficits can be characterized as the question whether late
processing deficits, and finally reading by itself, is the effect of, or the cause for,
early processing deficits (Hulme, 1988; Snowling, 2001). Alone, the fact of
correlation between measures of early or late processing and reading performance is
not sufficient to decide this question. .
Furthermore, models have to be mentioned which emphasize the importance of
more general or guiding cognitive abilities for functioning reading, such as visual
attention and eye movement control (see Radach, Inhoff, & Heller and Starr,
Kampe, Miller, & Rayner, this volume), stereo vision (Stein, Richardson, & Fowler,
2000) or other - perhaps less reading specific cognitive abilities (see Witruk, Ho, &
Schuster, this volume). However, even though these models appear to be visual
approaches at first glance, differences found between normal and poor/disabled
readers, e.g., in the pattern of saccades, the number of regressions and the location
of the optimal viewing point (see Everatt et aI., 1999 for review; see the chapters of
Starr et ai. and Radach et aI., this volume) may not necessarily require an
interpretation in favor of the visual hypothesis of reading disability, but can also be
explained in terms of differences in language processing (e.g., Kennedy, 1987;
Everatt & Underwood, 1994).
A most important distinction of models to be made is that of monocausal versus
multicausal explanations of reading disability. Orton's (1925) traditional concept
(strephosymbolia), for instance, is a monocausal theory, that is, reversal errors are
defined as the cardinal symptom of an abnormal cerebral dominance and a failure in
suppressing mirror image representations during the integration of visual and
phonological information. The aforementioned studies of Liberman et aI., (1971)
and Fisher et ai. (1978) do not doubt the significance of reversal errors in reading
disability. Rather, they dispute the role of reversals as the cardinal symptom and the
monocausality of the theory of strephosymbolia. More recent explanations of
reversal errors, such as that of the concept of symmetry generalization (Brendler &
Lachmann, 2001; Lachmann, this volume; see also Lachmann & van Leeuwen, in
press}, are an integrated part of a multicausal model.
The discussion about multi - vs. monocausality is strongly related to the
discussion about the definition of reading disability (or dyslexia). The so-called
discrepancy definition of dyslexia is based on a significant discrepancy between the
observed reading achievement and the reading achievement expected by the
students general intelligence (see e.g., Jimenez-Gonzales, this volume). This
symptom-based definition (symptom principle, cf. Tonnessen, 1995) is especially
criticized by those who define reading disability as caused (causality principle, cf.
Tonnessen, 1995) by a purely phonological deficit (Miles, 1991; Siegel, 1998;
Snowling, 1998,2001; Toth & Siegel, 1994; Wimmer, Landerl, & Frith, 1999). In
contrast, most models which concede visual differences in disabled readers do not
doubt possible phonological deficits in all, or a subgroup of disabled readers
(Becker, Lachmann, & Elliott, 2001; Doering, Trites, Patel, & Fiedorowicz, 1981;
Satz & Morris 1981; Slaghuis, Twell, & Kingston, 1996;). Most influential in this
respect was Boder's (1968, cited in Boder, 1971) distinction between dyseidetics
(10% of the sample) and dysphonetics (67% of the sample). Dyseidetics were
characterized to have problems in processing visual word gestalts and to stick to
phonological strategies even for highly frequent words. Dysphonetics were observed
to fail in adequate use of phonological patterns of words and to prefer visual
strategies even for unfamiliar words. However, almost a quarter of the sample
(23%) showed a mixed pattern of deficits and could not be classified within this
dichotomy. A body of evidence for dyslexia subtypes has been collected since then
whose significance, however, is called into question by others. Snowling (2001), for
instance, argues that differences in non-word reading are simply due to the severity
of the phonological deficit, and that parameters of visual processing (perceptual and
memory) may predict reading ability only rarely and if then because of its role as
sources of compensation of the causal phonological deficit (see also Stanovich,
1994; Stanovich, Siegel, Gottardo, Chiappe, & Sidhu, 1997). Others argue that a
failure in learning to read may be caused by problems in any of the sub functions
(including attention and guiding cognitive abilities) and their interaction and thus,
that reading disability or dyslexia labels a variety of problems (syndrome) in reading
(Rayner, 1993; Lachmann, this volume).
In summary, the field of reading disability research seems to be quite
heterogeneous. There is some contradicting experimental evidence, and certainly
contradicting interpretations of the same experimental results and, most importantly,
there is no consistency about the definition of reading disability or about whether
distinct groups of normal and reading disabled children even exist (Bryant & Impey,
1986; Stanovich, 1994; see also Jimenez-Gonzales, this volume). However, this is
not surprising given that the process of reading includes many subprocesses each
separable functionally and neurophysiologically.
Different approaches have been used to understand the phenomenon of reading
disability abetting the heterogeneity of the explanations and theories. Educational
scientists and school teachers have concentrated on individual as well as general
patterns of reading problems in order to develop special teaching programs (in the
tradition of Gates, 1936). Neurologists and cognitive neuroscientists focused on
neurophysiological differences between normal and disabled readers (see e.g., the
chapters of Kujala and Fawcett, this volume). Cognitive psychologists using
experimental procedures tried to test the separate functions which are assumed to
cause the failure in reading. However, for an ultimate understanding of reading
disability, all these different approaches are important. The bridge between the
different research domains must be built to foster an integrative theory of reading
disability. The present book, which brings together researchers from the
neurosciences, from language acquisition, from adult language processing and
reading with those investigating reading disabilities lays a first foundation for such a
bridge.
6. AFFILIATIONS
Dr. Angela D. Friederici

Max Planck Institute of Cognitive Neuroscience
PO. Box 500 355
D-04303 Leipzig, Germany
e-mail: angelafr@cns.mpg.de
Dr. Thomas Lachmann
University of Leipzig
Seeburgstrasse /4-20
D-04103 Leipzig, Germany
e-mail: lachmann@uni-Ieipzig.de.
7. REFERENCES
Becker, C., Lachmann, T., & Elliott, M. (2001). Evidence for impaired integration-segmentation
processes and effects of ociIlatory synchrony on stimulus coding in dyslexics. In E. Sommerfeld, R.
Kompass, & T. Lachmann (Eds.), Fechner Day 2001. Proceedings of the International Society for
Psychophysics (pp. 273-278). Lengerich, Berlin: Pabst.
Boder, E. (1971). Developmental dyslexia: Prevailing diagnostic concepts and a new diagnostic
approach. In H. Myklebust (Ed.), Progress in learning disabilities (pp. 293-321). New York: Grune
& Stratton.
Booth, J. R., Perfetti, C. A., & MacWhinney, B. (1999). Quick, automatic, and general activation of
orthographic and phonological representations in young readers. Developmental Psychology, 35, 3-
19.
Bradley, L., & Bryant, P. E. (1983). C..ategorizing sounds and learning to read - A causal connection.
Nature, 301, 419-521.
Breitmeyer, B. G. (1993). The roles of sustained (P) and transient (M) channels in reading and reading
disability. In S. F. Wright, & R. Groner (Eds.), Facets of dyslexia and its remediation. Studies in
visual information processing (pp. 13-31). Amsterdam: North-Holland/Elsevier.
Breitmeyer, B. G., & Ganz, L. (1976). Implications of sustained and transient channels for theories of
visual pattern masking, saccadic suppression, and information processing. Psychological Review, 83,
1-36.
Brendler, K., & Lachmann, T. (2001). Letter reversals in the context of the Functional Coordination
Deficit Model. In E. Sommerfeld, R. Kompass, & T. Lachmann (Eds.), Fechner Day 2001.
Proceedings of the International Society for Psychophysics (pp. 308-313). Lengerich, Berlin: Pabst.
Bryant, P. E., & Impey, L. (1986). The similarities between normal readers and developmental and
acquired dyslexics. Cognition, 24, 121-137.
Caplan, D., Alpert, N., & Waters, G. (1998). Effects of syntactic structure and propositional number on
patterns of regional blood flow. Journal of Cognitive Neuroscience, 10, 541-552.
Carpenter, P. A, Miyake, A, & Just, M. A (1994). Working memory contraints in comprehension:

Evidence from individual differences, aphasia, and aging. In M. A Gernsbacher (Ed.), Handbook of
psycholinguistics (pp. 1075-1122). San Diego: Academic Press.
Cohen, D., Dehaene, S., Naccache, L., Lehericy, S., Dehaene-Lambertz, G., Henaff, M.-A, & Michel, F.
(2000). The visual word form area. Spatial and temporal characterization of an initial stage of
reading in normal subjects and posterior split-brain patients. Brain, 123, 291-307.
Coltheart, M., & Rastle, K. (1994). Serial processing in reading aloud: Evidence for dual-route models of
reading. Journal of Experimental Psychology: Human Perception & Performance, 20, 1197-1211.
Corballis, M. c., & Beale, I. L. (1993). Orton revisited: Dyslexia, laterality, and left-right confudion. In
D. Willows, R. S. Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities
(pp. 57-73). Hillsdale, NJ.: Erlbaum.
of Verbal Learning and Verbal Behavior, 19, 450-466.
Deacon, T. W. (2000). Evolutionary perspectives on language and brain plasticity. Journal of
Communication Disorders, 33, 273-291.
Demonet, J. F., Chollet, F., Ramsay, S., Cardebat, D., Nespoulous, J. L., Wise, R., Rascol, A, &
Frackowiak, R. (1992). The anatomy of phonological and semantic processing in normal subjects.
Brain, 115, 1753-1768.
Doehring, D. G., Trites, R. L., Patel, P. G., & Fiedorowicz, C. A M. (1981). Reading disabilities: The
interaction of reading, language and neuropsychological deficits. New York: Academic Press.
Ellis, A W. (1984). Reading, writing and dyslexia: A cognitive analysis. Hillsdale, NJ: Erlbaum.
Everatt, 1., McCorquidale, 8., Smith, J., Culverwell, F. Wilks, A Evans, D., Kay, M, & Baker, D.
(1999). Association between reading ability and visual processes. In J. Everatt (Ed.), Reading and
Dyslexia (pp. 1-39). London: Routledge.
Everatt, J., & Underwood, G. (1994). Individual differences in reading subprocesses. Language and
Speech, 37, 283-297.
Farmer, M. E., & Klein, R. M. (1995). The evidence for a temporal processing deficit linked to dyslexia:
A review. Psychnomic Bulletin & Review, 2, 460-493.
Ferstl, E. c., & von Cramon, D. Y. (2001). The locus of coherence and cohesion in text comprehension:
An event~related fMRI study. Cognitive Brain Research, 11, 325-340.
Fiebach, C. J., Schlesewsky, M., & Friederici, A D. (in press). An ERP investigation of syntactic
working memory during the processing of German WH-questions. Journal of Memory and
Language.
Fiez, J. A (1997). Phonology, semantics, and the role of the left inferior prefrontal cortex. Human Brain
Mapping,5,79-83.
Fiez J. A, & Petersen S. E. (1998). Neuroimaging studies of word reading. Proceedings of the National
Academy of Sciences of the U.S.A., 95, 914-921.
Fisher, F. W., Liberman, I. Y., & Shankweiler, D. (1978). Reading Reversals and Developmental
Dyslexia: A Further Study. Cortex, 14, 496-510.
Frazier, L. (1987). Sentence processing: A tutorial review. In M. Coltheart (Ed.), Attention and
performance XII (pp. 559-586). Hillsdale, NJ: Erlbaum.
Friederici, A D., Meyer, M., & von Cramon, D. Y. (2000). Auditory language comprehension: An event-
related fMRI study on the processing of syntactic and lexical information. Brain and Language, 74, 289-
300.
Friederici, A. D., Steinhauer, K., Mecklinger, A, & Meyer, M. (1998). Working memory constraints on
syntactic ambiguity resolution as revealed by electrical brain responses. Biological Psychology, 47, 193-
221.
Gabrieli,1. D. E., Brewer, 1. 8., & Poldrack, R. A (1998). Images of medial temporal lobe functions in
human learning and memory. Neurobiology of Learning & Memory, 70, 275-283.
Gates, A I. (1936). The improvement of reading: A program of diagnostic and remedial methods. New
York: Macmillan.
Gauthier, I., Skudlarski, P., Gore, J. c., & Anderson, A W. (2000). Expertise for cars and birds recruits
brain areas involved in face recognition. Nature Neuroscience, 3, 191-197.
Gernsbacher, M. A, & Faust, M. (1991). The role of suppression in sentence comprehension. In G. 8.
Simpson (Ed), Understanding word and sentence. Advances in psychology, No. 77. (pp. 97-128).
Amsterdam: North-Holland.
Hagoort, P., Brown, c., & Osterhout, L. (1999). The neurocognition of syntactic processing. In C. Brown, &
P. Hagoort (Eds.), Neurocognition ofLanguage (pp. 273-316). Oxford: Oxford University Press.
Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. (2000). Trends in Cognitive Neurosciences, 4, 223-232.
Herbster, A. N., Mintun, M. A., Nebes, R. D. Becker, J. T. (1997). Regional cerebral blood flow during
word and nonword reading. Human Brain Mapping, 5, 84-92.
Hulme, C. (1988). The Implausibility of Low-Level Visual Deficits as a Cause of Children's Reading
Difficulties. Cognitive Neuropsychology, 5, 3, 369-374.
Joubert, S. A., & Lecours, A. R. (2000). The role of sublexical graphemic processing in reading. Brain &
Language, 72, 1-13.
Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehension: Individual differences in
working memory. Psychological Review, 99, 122-149.
Just, M. A., Carpenter, P. A., Keller, T. A., Eddy, W. F., & Thulborn, K. R. (1996). Brain activation
modulated by sentence comprehension. Science, 274, 114-116.
Kapur, S., Craik, F. I. M., Tulving, E., Wilson, A. A., Houle, S., & Brown, G. (1994). Neuroanatomical
correlates of encoding in episodic memory: Levels of processing effect. Proceedings of the National
Academy of Sciences of the USA, 91, 2008-2011.
Katz, R. 8., Healy, A. F., & Shankweiler, D. (1983). Phonetic coding and order memory in relation to
reading proficiency: A comparison of short-tern memory for temporal and spatial order information.
Applied Psycholinguistics, 4, 229-250.
Kennedy, A. (1987). Eye movements, reading skill and spatial code. In J. Beech, & A. Colley (Eds.),
Cognitive approaches to reading (pp. 102-142). Chichester: Wiley.
Kintsch, W. (1988). The role of knowledge in discourse comprehension: A construction integration
model. Psychological Review, 95, 163-182.
Kussmaul, A. (1878). Word-deafness and word-blindness. In H. v. Ziemssen (Ed.), Cyclopaedia of the
practice of medicine. London: Sampson Row, Maston, Searle & Rivingston.
Lachmann, T., & van Leeuwen, C. (in press). Memory-guided Inference in Same-Different Comparison
Tasks. In C. Kaernbach, E. Schroeger, & H. J. Miiller (Eds.), Psychophysics beyond sensation.
Hilldale, NJ: Erlbaum.
Liberman, I. Y. (1983). A language-orientated view of reading and its disabilities. In H. Myklebust (Ed.),
Progress in learning disabilities (pp 81-101). New York: Grune & Stratton.
Liberman, I. Y., Shankweiler, D., Orlando, c., Harris, K. S., & Bell Berti, F. (1971). Letter Confusions
and Reversals of Sequence in the Beginning Reader: Implications for Orton's Theory of
Developmental Dyslexia. Cortex, 7, 127-142.
Livingstone, M. S., & Hubel, D. H. (1987). Psychophysical evidence for separate channels for the
perception of form, color, movement, and depth. Journal of Neuroscience, 7, 3416-3468.
Lovegrove, W. J. (1993). Do dyslexics have a visual deficit? In S. F. Wright, & R. Groner (Eds.), Facets
of dyslexia and its remediation. Studies in visual information processing (pp. 33-49). Amsterdam:
North-Holland/Elsevier.
Lovegrove, W., Martin, F., & Slaghuis, W. (1986). A Theoretical and Experimental Case for a Visual
Deficit in Specific Reading Disability. Cognitive Neuropsychology, 3, 2, 225-267.
MacDonald, M. c., Just, M. A., & Carpenter, P. A. (1992). Working memory constraints on the
processing of syntactic ambiguity. Cognitive Psychology, 24, 56-98.
Marslen-Wilson, W., & Warren, P. (1994). Levels of perceptual representation and process in lexical
access: Words, phonemes, and features. Psychological Review, 101, 653-675.
Massaro, D. W. (1998). Models for reading letters and words. In D. Scarborough, & S. Sternberg (Eds.),
An invitation to cognitive psychology. Methods, models and conceptual issues (pp. 301-364).
Cambridge: MIT Press.
McCarthy, G., & Nobre, A. C. (1993). Modulation of semantic processing by spatial selective attention.
Electroencephalography & Clinical Neurophysiology: Evoked Potentials, 88, 210-219.
Meyer, M., Alter, K, & Friederici, A. D. (in press). Towards the cerebral substrates of sentence-level
syntactic and prosodic processing. Speech and Communication.
Michael, E. 8., Keller, T. A., Carpenter, P. A., & Just, M. A. (2001). fMRI investigation of sentence
comprehension by eye and by ear: Modality fingerprints on cognitive processes. Human Brain
Mapping, 13, 239-252.
Miles, E. (1991). Visual dyslexia! auditory dyslexia: Is this a valuable distinction to make? In M.
Snowling, & M. Thompson (Eds.), Dyslexia: Integrating theory and practice (pp. 195-203). London:
Whurr.
Miles, T. R., & Miles, E. (1999). Dyslexia: A hundred years on. Buckingham: Open University Press.
Miyake, A, Just, M. A, & Carpenter, P. A (1994). Working memory constraints on the resolution of
lexical ambiguity: Maintaining multiple interpretations in neutral contexts. Journal of Memory &
Language, 33, 175-202.
Noordman, L. G. M., & Vonk, W. (1999). Discourse comprehension. In A D. Friederici (Ed.), Language
comprehension: A biological perspective, 2nd Edition (pp. 230-263). Berlin! Heidelberg: Springer
Verlag.
Orton, S. T. (1925). "Word-Blindness" in School Children. Archives of Neurology and Psychiatry, 14, 5,
581-615.
Otten, L. 1., Rugg, M. D., & Doyle, M. C. (1993). Modulation of event-related potentials by word
repetition: The role of visual selective attention. Psychophysiology, 30, 559-571.
Paulesu, E., Frith, C. D., & Frackowiak, R. S. (1993). The neural correlates of the verbal component of
working memory. Nature, 362, 342-345.
Perfetti, C. A (1998). Comprehending written language: a blueprint of the reader. In C. M. Brown, & P.
Hagoort (Eds.), The neurocognition of language (pp. 167-208). Oxford: University Press.
Perfetti, C. A, Bell, L. c., & Delaney, S. M. (1998). Automatic (pre lexical) phonetic activation in silent
word reading: Evidence from backward masking. Journal of Memory & Language, 27, 59-70.
Perfetti, C. A, & Sandak, R. (2000). Reading optimally builds on spoken language: Implications for deaf
readers. Journal of Deaf Studies & Deaf Education, 5, 32-50.
Pisoni, D. B., & Luce, P. A (1987). Acoustic - phonetic representations in word recognition. Cognition,
25,21-52.
Price, A (2001). Anatomical and Cognitive Models of Reading: From Dejerine to Functional Imaging.
Paper presented at the Neural Basis of Language Conference, Groningen, July 9 - 11,2001.
Price, C. J., Wise, R. J., Ransay, S., Friston, K., Howard, D., Patterson, K., & Frackowiak, R. S. J.
(1992). Regional response within the human auditory cortex when listening to words. Neuroscience
Letters, 146, 179-182.
Price, C. J., Wise R. J., Watson, 1. D., Patterson, K., Howard, D., & Frackowiak, R. S. (1994). Brain
activity during reading: The effects of exposure duration and task. Brain, 117, 1255-1269.
Rayner, K. (1993). Visual porocessing in reading: Directions for research and theory. In D. Willows, R.
S. Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp. 475-480).
Hillsdale, NJ.: Erlbaum.
Samuels, S. 1., LaBerge, D., & Bremer, C. D. (1978). Units of word recognition: Evidence for
developmental changes. Journal of Verbal Learning & Verbal Behavior, 17, 715-720.
Satz, P., & Morris, R. (1981). Learning disability subtypes: A review. In F. 1. Pirozzolo, & M. C.
Wittrock (Eds.), Neuropsychological and cognitive processes in reading (pp. 109-141). New York:
Academic Press.
Seymour, P. H. K., & Evans, H. M. (1993). The visual (orthographic) processor and developmental
dyslexia. In D. M. Willows, R. S. Kruk, & E. Corcos (Eds), Visual processes in reading and reading
disabilities (pp. 347-376). Hillsdale, NJ: Erlbaum.
Siegel, L. S. (1998). Phonological processing deficits and reading disabilities. In J. L. Metsala, & L. C.
Ehri (Eds.), Word recognition in beginning literacy (pp. 141-160). Mahwah, NJ: Erlbaum.
Slaghuis, W. L., Twell, A 1., & Kingston, K. R. (1996). Visual and language processing disorders are
concurrent in dyslexia and continue into adulthood. Cortex, 32, 413-438.
Smid, H. G. O. M., Jakob, A, & Heinze H. 1. (1999). An event-related brain potential study of visual
selective attention to conjunctions of color and shape. Psychophysiology, 36, 264-279.
Snowling, M. (1998). Dyslexia as a phonological deficit: Evidence and implications. Child Psychology &
Psychiatry Review, 3, 4-11.
Snowling, M. (2001). From language to reading and dyslexia. Dyslexia, 7, 37-46.
Stanley, G., & Hall, R. (1973). Short-term information processing in dyslexics. Child Development, 44,
841-844.
Stanovich K. E. (1994). Annotation: Does Dyslexia Exist?J. ChildPsychol. Psychiat.,35, 4, 579-595.
Stanovich, K. E., Siegel, L. S., Gottardo, A, Chiappe, P., & Sidhu, R. (1997). Subtypes of developmental
dyslexia: Differences in phonological and orthographic coding. In B. A Blachman (Ed.),
Foundations of reading acquisition and dyslexia: Implications for early intervention (pp. 115-141).
Mahwah, N.J.: Lawrence Erlbaum Associates.
Stein, 1. F., Richardson, A 1., & Fowler, M. S. (2000). Monocular occlusion can improve binocular
control and reading in dyslexics. Brain, 123, 164-170.
Stromswold, K., Caplan, D., Alpert, N., & Rauch, S. (19%). Localization of syntactic comprehension by
positron emission tomography. Brain and Language, 52, 452-473.
Tallal, P. (1984). Temporal or phonetic processing deficit in dyslexia? That is the question. Applied
Psycholinguistics, 5, 167-169.
Thompson-Schill, S. L., D'Esposito, M., Aguirre, G. K., & Farah, M. 1. (1997). Role of left inferior
prefrontal cortex in retrieval of semantic knowledge: A reevaluation. Proceedings of the National
Academy of Sciences of the USA, 94, 14792-14797.
Tonnessen, F. E. (1995). On defining "dyslexia." Scandinavian Journal of Educational Research, 39,
139-156.
Toth, G., & Siegel, L. S. (1994). A critical evaluation of the IQ-achievement discrepancy based
definition of dyslexia. In K. P. van den Bos, & L. S. Siegel (Eds), Current directions in dyslexia
research (pp. 45-70). Lisse, NL: Swets & Zeitlinger.
Vellutino, F. R. (1977). Alternative conceptualizations of dyslexia: Evidence in support of a verbal-
deficit hypotheses. Harvard Educational Review, 47, 334-354.
Veil uti no, F. R. (1979). Dyslexia: Theory and research. Cambridge, MA: MIT Press.
Vellutino, F. R., & Scanlon, D. M. (1987). Phonological coding, phonological awareness, and reading
ability: Evidence from a longitudinal and experimental study. Merrill-Palmer Quarterly, 33, 321-363
Vellutino, F. R., & Scanlon, D. M. (1998). Research in the study of reading disability: What have we
learned in the past four decades? Manuscript of an invited paper presented at the annual conference
of the American Educational Research Association, San Diego.
Vos, S. H., Gunter, T. c., Schriefers, H., & Friederici, A. D. (2001). Syntactic parsing and working
memory: The effects of syntactic complexity, reading span, and concurrent load. Language and
Cognitive Processes, 16, 65-103.
Waters G. S., & Caplan D. (1996). The measurement of verbal working memory capacity and its relation
to reading comprehension. Quarterly Journal of Experimental Psychology, 49, 51-79.
Wiggs, C. L., Weisberg, 1., & Martin, A. (1999). Neural correlates of semantic and episodic memory
retrieval. Neuropsychologia, 37, 103-118.
Willows, D. M., Kruk, R. S., & Corcos, E. (1993). Are there differences between disabled and normal
readers in their processing of visual information? In D. M Willows, R. S. Kruk, & E. Corcos (Eds.),
Visual processes in reading and reading disabilities (pp. 265-287). Hillsdale, N1.: Erlbaum.
Wimmer, H., Landerl, K., & Frith, U. (1999). Learning to read German: Normal and impaired
acquisition. In M. Harris, & G. Hatano (Eds), Learning to read and write: A cross-linguistic
perspective. Cambridge studies in cognitive and perceptual development (pp. 34-50). New York:
Cambridge University Press.
Zatorre, R. J., Evans, A. C., Meyer, E., & Gjedde, A. (1992). Lateralization of phonetic and pitch
discrimination in speech processing. Science, 256, 846-849.
SECTION II
BASIC FUNCTIONS OF LANGUAGE

ACQUISITION AND
LANGUAGE COMPREHENSION
D. SWINNEY & T. LOVE
CONTEXT EFFECTS ON LEXICAL PROCESSING

DURING AUDITORY SENTENCE COMPREHENSION
On the Time-Course and Neurological Bases of a Basic Comprehension
Process
Abstract. This paper presents an integrated view of the effects of context upon lexical access and lexical
integration during sentence comprehension. The review incorporates evidence from both standard
psycholinguistic and neuro-cognitive approaches. Along with this integrated overview, new hemisphere-
specific processing evidence concerning context and lexical processing is presented. The evidence is
taken to support a "modes of processing" perspective in the examination of sentence comprehension.
1. INTRODUCTION
Understanding the nature of lexical representation and processing constitutes one of
the foundational issues in the study of sentence comprehension. A vast literature,
spanning decades of research, has been produced on topics and issues related to
lexical issues, resulting in a number of well-established findings but an even larger
set of conflicting evidence and theoretical claims. The goal of this chapter is to
present an integrated view of lexical access, lexical integration, and the time-course
of the effect of context upon these processes. We propose to accomplish this with
the aid of two specific levers (involving relevant new data): First, an examination of
the neurological underpinnings of these processes and second, consideration of the
issue of "modes" of processing. Ultimately, we believe that these considerations
allow for an integrated view of lexical processing in service of sentence
comprehension. In what follows, we first present some parameters that limit the
domain of the field to be covered in this chapter, followed then by our examination
of lexical processing and context effects.
2. SCOPE OF REVIEW / ANALYSIS
There are three tenets which guide the approach taken and the choice of evidence
examined in this exposition. The first is that language processing in general is
something that can be accomplished in a number of ways. Moreover, central to any
understanding of the nature of the processes underlying language is a clear, detailed
definition of the "type" of language situation - the parameters of language
processing - that are under focal investigation. It follows from this that lexical
access and integration can potentially be accomplished via varying processes in
different situations. Thus, it is absolutely critical to carefully specify the parameters
and conditions focused on in any set of claims about the nature of lexical processing.

26 D. SWINNEY & T. LOVE
The second tenet is that any deep understanding of sentence/language

comprehension can only be developed via examination of evidence which reflects
moment-by-moment details of the (sub)processes involved in ongoing
comprehension, evidence typically referred to as "on-line" reflections of processing.
We also believe, of course, that such processes must be examined in conjunction
with broader-scope (off-line) approaches in order to gain an integrated picture of all
levels of processing; however, it is the on-line details that will largely differentiate
among most current theoretical accounts of contextual processing and lexical access.
Finally, the third tenet guiding our approach is that it is only in the conjunction
of traditional on-line behavioral processing evidence combined with relevant
evidence from modern cognitive neuroscience (e.g., studies of focal brain-damaged
populations, brain imaging, etc.) that the necessary leverage will be obtained to gain
more definitive answers about the classic, central theoretical issues in the field.
In line with the above, this chapter will focus on the study of lexical processing
during auditory comprehension. The processes underlying listening and of reading
diverge and differ at numerous critical stages and detailing the differences and
similarities between the two is considerably beyond the scope and length parameters
of this chapter. In addition to focusing on the auditory domain, this chapter will also
focus on comprehension of lexical information as examined within the context of
ongoing sentence comprehension. While the processing of words outside of
sentence/discourse contexts (i.e., in isolation, in pairs, etc.) holds obvious
relationships to that found within standard sentence/discourse comprehension, the
comprehension process is sensitive to critical processing parameters which are
simply unavailable (and hence not utilized) in the processing of words outside of
discourse/sentence settings. The link between the mechanisms underlying lexical
processing as found in sentences and lexical processing as found in other settings
awaits a far larger understanding of goals and parameters of cognitive processing in
general.
3. THE EFFECTS OF CONTEXT ON LEXICAL ACCESS
Basic Issues and Evidence. The fundamental issue that has formed the basis of
debate in the lexical representation / lexical processing field has concerned the
manner in which lexical information is made available to the ongoing
comprehension process. While this has often been framed as an issue of "Modularity
vs. Interactivity" of information processing during comprehension (e.g., Fodor,
1983; Swinney, 1979; Tyler & Marslen-Wilson, 1982), each of these terms have
come to "cover" for a (surprisingly varied) number of claims about processing. We
will thus avoid use of these specific terms at first here, focusing instead on
specifically defined issues. We will be concerned with the question of how (and
when) contextual material which occurs prior to a particular word in a sentence
constrains the amount or type of lexical information about that word that is made
available to ongoing comprehension. This is the aspect of the modularity-
interactivity debate that is concerned with whether or not prior context has the
ability to limit access to information "stored with" a lexical item. 1 We will be
concerned with the time-course of availability of (all or part of) such lexical
information for further processing during ongoing comprehension. Ultimately, we
CONTEXT EFFECfS AND AUDITORY SENTENCE COMPREHENSION 27
will also be concerned with the degree to which specific lexical processes are
susceptible to effects of expectations and predictions based on world knowledge and
prior context. Alternatively, we will explore the degree to which these certain lexical
processes are fundamentally form-directed operations.
Our analysis in this chapter focuses on the processing of lexically ambiguous
words, words in which the phonological form connects to more than one meaning.
Such homophoniclhomonymic lexical elements have traditionally constituted the
major testing ground for the issues concerning context effects and lexical
processing, as they provide individuable information (the distinct "meanings" of the
ambiguity) which can be separately addressed via contextual material. This, allows
for empirical tests of when and how context may come to constrain the access to
such information (See, e.g., work by Foss, 1998; Foss, Starkey, & Bias, 1988).
Studies examining the processing of lexical ambiguities during auditory sentence
comprehension, in the absence of a biasing prior context, have nearly uniformly,
demonstrated that all meanings associated with the word form are momentarily
accessed and made available for further processing (Onifer & Swinney, 1981;
Picoult & Johnson, 1992; Prather & Swinney, 1977; Seidenberg, Tanenhaus,
Leiman, & Bienkowski, 1982; Simpson, 1981; Swinney, 1979; Swinney & Prather,
1989; Tanenhaus, Leiman, & Seidenberg, 1979, among others). Thus, when no prior
biasing constraints from world knowledge, lexical associates, plausibility, discourse
context, etc. are present, access to all information stored with a lexical entry is made
available to sentence processing for a short period. Note that there is standardly the
inherent effect of relative frequency which may determine the order of availability
of various meanings of lexical ambiguities; however, such frequency effects are
neither constraining nor precluding effects (i.e., the less frequent meanings are still
made available for further processing; see more on this, below).
Studies examining the processing of lexical ambiguities during auditory sentence
comprehension in the presence of prior context are slightly more varied in the
interpretation of their findings. The vast majority of such work has regularly and
repeatedly demonstrated, across a large range of "prior contexts", that context does
not restrict immediate or initial access to lexical information. This is nearly
universally true in research which has employed temporally-sensitive tasks which
have been demonstrated to have only a minimum of demand-characteristics that
force interactions with the lexical access process itself. Thus, for example,
unrestricted, exhaustive initial access of meanings for lexical ambiguities has been
found even in the presence of prior contexts which place strong and definitive
constraints on their interpretation in terms of: 1) Syntactic category (e.g., Prather
and Swinney, 1977; Tanenhaus et aI., 1979), 2) Semantic-associative contexts (e.g.,
Love & Swinney, 1996; Miyake, Just & Carpenter, 1994; Onifer & Swinney, 1981;
Picoult & Johnson, 1992; Seidenberg et aI., 1982; Simpson, 1981; Swinney, 1979;
among others), 3) Highly restrictive semantic-associative sentential contexts (e.g.,
Swinney, 1991), and 4) Discourse contexts (e.g., Swinney, 1982). This effect also
holds, regardless of whether the biasing context appears earlier in the same sentence
as the ambiguous word, or in a prior sentence in a larger discourse (e.g., Swinney,
1979). Similarly, it has been shown that a patient population which has well-defined
and well-known inabilities in utilizing context - chronic schizophrenics -
demonstrates their "context problems" only at a point "downstream" from initial
access of all the meanings of a lexical ambiguity, not at the point of access (Onifer,
1980). This, therefore, supports the view that the locus of "prior context" effects on
lexical ambiguity is at a point following initial access of all information stored with
that entry. Finally, we note that this same pattern of exhaustive, form-driven,
context-independent access has also been demonstrated in on-line studies of
processing in pre-school age children (Love, Swinney, Bagdasaryan, & Prather,
1999; Swinney & Prather, 1989).
This relatively large range of evidence has come from a number of different
tasks - cross-modal lexical priming, auditory ERPs, immediate judgment tasks, etc.
- which have examined context and ambiguity processing in normal, fluent speech.
In all, then, studies of fluent auditory language processing have repeatedly
demonstrated that whenever a phonetic form of a word is encountered there is
immediate and unconstrained access to all information for that word, even in the
face of a wide range of strongly constraining, prior-occurring, contexts. There is
evidence, however, that as early as 200-300 msecs following initial access, biasing
contexts may begin to have constraining effects.2. Our examination of the
neurological underpinnings of lexical ambiguity access, resolution and context
effects begins with these established findings.
In what follows immediately below we briefly present a summary of a small
portion of a recent study that replicates the basic findings discussed above. We
present it in some detail so as to allow for both a specific set of findings to carry
through in further discussion, and a specific example to consider throughout our
consideration of these effects.
Love and Swinney (1996) present an examination of effects of context and
structural processes on lexical ambiguity access and processing. The following is a
summary of the methodology employed in this study:
3.1 Methodology
A Cross-Modal Lexical Priming (CMLP) task was employed, usin~ a matched-
probe configuration (Swinney, Onifer, Prather, & Hirshkowitz, 1979) . Participants
in the part of the study to be described here were 51 non-neurologically-involved
college students, who heard sentences such as:
The professor insisted that the exam be completed in ink, so Jimmy

used the new pen* that his mother-in-law recently purchased because
the multiple colors allowed for more creativity.
There were 40 such experimental sentences (along with 40 structurally similar

"control" sentences and 10 practice sentences). All sentences were recorded (in
pseudo-random order on counterbalanced scripts) by a male speaker at the rate of
approximately 5 syllables per second. The asterisk (*) indicates position at which
the experimental and control visual probes appeared during the sentence; only a
single sentence and single probe at a single position was presented to anyone
participant; conditions were counterbalanced across a series of scripts and subject
groupS4. Participants "named" each probe word that appeared and RT to voice-onset
of these "naming" responses was recorded. The biasing contexts in the experimental
materials were created followed the criteria employed by Tabossi (1988) to establish
CONTEXT EFFECTS AND AUDITORY SENTENCE COMPREHENSION 29
a bias toward a strong "aspect" of the a priori more frequent meaning of the
ambiguity. (under these criteria, a minimum of 75% of 12 judges agreed on intended
aspect of meaning of the ambiguous word in the sentence, etc.). In a separate pretest,
the "related" visual probes were created by first obtaining associations to the
ambiguous word. Uniformly agreed-upon associates related to each interpretation of
the ambiguity were chosen as the "related" probes, and these were matched with
"unrelated control" probes (based on a-priori reaction times taken in an isolated
word "naming" task, performed with over 50 subjects from yet another pretest) Thus
(to follow the above example) associated probes related to each of the meanings of
the ambiguous word "PEN" were chosen (e.g., "PENCIL" and "JAIL"), along with
reaction-time-matched control words. Experimental and control probes were also
equated for both "goodness-of-fit" to the sentence as heard up to the point the probe
appeared (this involved a rating scale ranking of goodness of fit of the probe to the
"preceding" sentence fragment) and for "relatedness-of-probe-to-sentence" (again, a
rating scale). 5
3.2 Results
The critical findings were: A main effect for Probe Type (related vs. control)
(F(1,47) = 14.844, p=.OOl), which did not interact significantly with the Ambiguity-
Meaning factor (Primary vs. Secondary). Planned a priori comparisons performed
on the related vs. control probes for each of the Ambiguity Meanings demonstrated a
significant priming effect for both the Primary Meaning - PENCIL (t50=2.242,
p=.015) and for the Secondary Meaning - JAIL (t50 = 1.805, p=.038)
The mean reaction times for "control" and "related" probes for both primary and
secondary interpretations of the lexical ambiguity) can be seen in the following:
Ambiguity Meaning: Primary (PENCIL) Related probe: 521 msec

Control probe: 533 msec
Secondary (JAIL) Related probe: 529 msec
Control probe: 537 msec
This study thus replicated the long established finding of contextually

independent, form-driven initial access for lexical information. 6• 7
We now turn to considerations of the neuro-biological mechanisms that might
underlie both the contextually-independent, form-driven access and subsequent
"meaning resolution" evidence seen in this and many similar on-line behavioral
studies of the effects of context upon the processing of lexical information during
sentence comprehension.
4. NEUROLOGICAL BASIS OF LEXICAL PROCESSING
4.1 Lesion Evidence
Evidence from the processing of lexical ambiguities in patients with focal lesions
provides vehicle via which an understanding of aspects of the behavioral evidence
derived from non-neurologically involved populations (and discussed above) can be
refined and extended. In particular, we note that evidence concerning the processing
of lexical ambiguities in auditory sentences by patients who have Broca's aphasia
provides an interesting contrast to those for non-Iesioned patients (and to those with
lesions in other brain areas).
Broca's aphasia, is standardly associated with lesions in and around the lower
portion of the left frontal lobe of the cortex (more particularly, the opercular and
triangular portions of the inferior frontal gyrus, including the foot of the third frontal
convolution and extending into subcortical white matter8). Difficulty in both the
articulation and production of speech, accompanied by subtle difficulties in
comprehension accompany the syndrome diagnosis. Patients with damage in and
around these areas typically produce, at best, labored speech, which is poorly
articulated and telegraphic in nature (typically involving omission of many
"function" or "closed-class" words) and display comprehension problems with
complex syntactic structures, at both on-line and off-line levels of analysis.
Swinney, Zurif and Nicol (1989) presented a population of Broca's aphasics,
non-impaired age-matched control subjects, and Wernicke's aphasics (where
damage is in the left temporal lobe of the cortex - in areas considerably removed
from the damage found in Broca's aphasia) with auditory sentences containing
lexical ambiguities. These ambiguities were presented in sentential contexts that
were biased either toward the a priori primary interpretation of the ambiguity, or the
a priori secondary interpretation of the ambiguity. In a CMLP study designed
similarly to that presented just above (except that a "lexical decision" task rather
than a "naming" task was utilized in this study, and the sentences were considerably
shorter in overall length), it was found that the non-impaired control population
demonstrated immediate access for both meanings of the ambiguous word (as shown
by significant (p<0.05) priming scores for probes related to each meaning)
independently of the presence of a biasing sentential context. Similarly, the
Wernicke's aphasic population displayed the same pattern of effects (significant
priming for words related to each meaning of the ambiguity, independent of
contextual bias) Thus, brain damage per se (i.e., brain damage at just any randomly
chosen location) does not change these characteristics of lexical access. Importantly,
however, (and in contrast to the findings just reported), the patients classified as
Broca's aphasics demonstrated a different pattern of results. For the Broca's
aphasics, only the primary (a-priori most frequent) meaning of the ambiguous word
was found to be significantly primed (p<.05) immediately after that word was heard
in a sentence, regardless of the direction of the bias of the prior context in those
sentences. At least two conclusions can be drawn from these results. First, brain
damage - either in Broca's or in Wernicke's areas - does not change the
contextually-independent nature of lexical access; contextual information did not act
to limit initial access in either population (or in the age-matched control). Second,
persons with damage to Broca's area apparently have initial access to only the most
frequent interpretation of ambiguous words. This has been interpreted as indicating
that certain lexical information (and certainly that which is associated with a less-
frequent interpretation) may have temporally protracted (slower-than-normal) "rise
time". The finding that certain types of lexical information may have a slow "rise-
time" is not an unusual finding. Work with both very young children (Swinney &
Prather, 1989) and dyslexics (Swinney, 1982) have found evidence suggesting that
certain aspects of lexical information may be slower than others to have effects on
on-going sentence integration. Note, however, that this finding is not at all in
conflict with the conclusion that context does not limit initial access. The issue there
is whether context prevents access to contextually-unrelated information for words -
i.e., that it by-passes or constrains form-driven access. The fact that some
information is slower than other information to be made available, usable, (or even
detectable), should not be surprising, and is certainly separate from considerations of
the initial effects of context on lexical information access. Importantly, however,
these findings suggest that the cortical (and sub-cortical) regions of the left
hemisphere associated with Broca's aphasia may have a role in the time-course of
making aspects of stored lexical information available to more integrative, ongoing,
sentence processing.
Several related studies have supported this interpretation of the role of Broca's
area in lexical processing. For example, Swaab, Brown, & Hagoort, (1997)
employed an ERP methodology to examine the processing of lexical ambiguities in
Broca's aphasics in auditorily presented sentences. In this study the status of
activation of meanings for the sentence final ambiguous word was inferred from the
amplitude of the N400 to related targets that were presented at two different inter-
stimulus intervals. The ERP evidence was interpreted to indicate that Broca's
aphasics, in contrast to elderly controls, were not successful at selecting an
appropriate meaning based on context for the ambiguity immediately (the short lSI
condition), but that at the long lSI the patients were able to do so. While there are
always concerns over interpreting data obtained from material presented in sentence-
final and trial-final position (strategic end-of-sentence wrap-up effects, etc.), these
data at least fit with a view that contextual integration of lexical material may be
protracted in these patients, again suggesting that the neural substrate underlying
Broca's aphasia may be critical for normal rapid lexical access and integration.
More directly to the point, Prather, Zurif, Love and Brownell (1997) investigated
the "protracted lexical activation" hypothesis concerning the neural region
subsumed by Broca's area in a study of the time course of lexical activation and
priming. They employed a continuous-list priming paradigm, in which words to
which subjects make lexical decisions appear in a continuous (non-paired,
sequential) list. Hidden within this list, however, are some sequentially-contiguous
words which constitute semantic prime-target pairs). This technique is one of the
few ways in which non-sentential word priming effects can be studied without
special pair-wise "control" strategies arising due to methodology. In the study, they
manipulated the temporal delay (ranging from 200-2100 milliseconds) between
successive words in the list. In a single-case study, a classic Broca's aphasic subject
demonstrated reliable automatic semantic priming only at the relatively long lSI of
1500 msec. (This was in stark contrast to non-neurologically involved elderly
control subjects and the Wernicke's aphasic also profiled in this report, who prime at
relatively short inter-stimulus intervals - typically beginning at 500 msec.). Thus,
this Broca's aphasic retained the ability to access lexical information when allowed
sufficient time to do so.
Thus, based on lesion evidence, it appears that the (left-hemisphere) neural
substrate underlying Broca's aphasia (Broca's area) may be critically involved in the
well-established effects found in the on-line sentence processing literature for "non-
neurologically involved" listeners: namely, the rapid/immediate access of the all
(including the less frequent) of the interpretations of an ambiguous word
independently of context. In addition, there is some evidence that damage in and

around this area may also be involved in subsequent resolution of these "meanings"
to a single "contextually appropriate" interpretation, again, possibly related to the
use of this substrate in promoting and regulating the rapid availability of lexical
information to ongoing comprehension processes. Damage to the cortical (and
perhaps subcortical) region in and around Broca's area appears to undermine both
the ability to make all information stored with a word "immediately" available for
further sentence processing9 and to also disrupt the subsequent integration of those
word meanings with context. In order to further consider the neurological
underpinnings of this latter issue, we tum briefly to a description research on the
effects of brain lesions in subcortical tissue with connections to cortical tissue,
including Broca's area.
A growing body of work has developed concerning the role of the subcortical
tissue in and around the basal ganglia (but not including the thalamus) (i.e., the
caudate nucleus, putamen, internal capsule; often termed striato-capsular area).
Much of this research, while not primarily examining auditory comprehension, has
come to suggest that this neural substrate in the left hemisphere has a neuro-
regulatory role in lexical-semantic processes, particularly with regard to "controlled"
or "attentional" forms of dealing with lexical information (e.g., Copland, Chenery,
& Murdoch, 2000; Crosson, 1985). Of particular relevance is recent work by
Copland (2000), in which he used a CMLP paradigm to investigate the processing of
lexical ambiguities in a biased sentential context in patients with left-hemisphere
(language-dominant hemisphere) non-thalamic subcortical lesions. He found two
important results relevant to our concerns. First, the left-hemisphere subcortical
lesioned patients failed to maintain a selective facilitation for (failed to demonstrate
continued significant priming for) the contextually appropriate meaning of the
ambiguity, thus implicating this brain region in support of attention-based control,
maintenance and integration of lexical-semantic sentential information based on
sentential context. Second, he demonstrated that patients with these same lesions
failed to utilize discourse-level information to select relevant meanings for words. It
is worth noting in this regard that dopamine dysfunction in this same general brain
region (basal ganglia and cingulate) has been implicated in abnormalities associated
with semantic associative activation (Fuentes & Santiago, 1999; Posner & Raichle,
1994). This, of course, underscores the work reported above by Onifer (1980) on
CMLP and the locus of context effects in sentence processing in individuals with
chronic schizophrenia.
In all, it appears that the striato-capsular subcortical region of the language-
dominant hemisphere, with afferentinput from (among other cortical areas) Broca's
area, has a major role in integrating and maintaining activated lexical information
during auditory sentence comprehension. Thus, while Broca's area may be involved
in making lexically activated material available in a timely manner to ongoing
sentential integration processes, the stria to-capsular region appears involved in
integration and maintenance of contextually relevant interpretations.
4.2 Contributions of the Left and Right Hemispheres in Processing Lexical

Ambiguities
There is a very small but relevant body of research examinIng the individual
(independent) contributions of the left (LH) and right (RH) cerebral hemispheres in
non-neurologically involved individuals to lexical processing and context effects
during sentence processing. This literature makes it clear than it is not only the
language-dominant (typically, left) hemisphere which has a role to play in the
activation and integration of lexical information in service of auditory sentence
understanding. In a study involving non-brain-damaged subjects, Faust & Chiarello
(1998) investigated hemisphere asymmetries in processing lexical ambiguity within
a sentence context. Sentences containing sentence-final ambiguous words (biased
toward a single meaning) were presented, followed by a hemi-field lateralized
"lexical-decision" target word, which was related to either the contextually relevant
or contextually incongruent meaning of the ambiguous word. They found that right-
visual-field-presented contextually congruent targets were primed, while RVF
incongruent targets were not. In contrast, in the left visual field both the congruent
and non-congruent targets were primed, regardless of sentence context. Again,
employing an end of sentence-ambiguity hemifield-target priming paradigm, Titone
(1998) also found evidence consistent with the view that there is differential
sensitivity in the cerebral hemispheres to meaning salience and context. Both of
these studies, however, employed end-of-sentence ambiguity targets, something that
has been called into question due to end-of-sentence wrap-up effects in other studies
of sentence processing (see, e.g., Balogh, Zurif, Prather, Swinney, & Finkel, 1998).
Love, Bouck, Hald, Hickok, & Swinney (in prep) have recently conducted a
study involving 66 native English right handed individuals which also explored
possible hemispheric asymmetries in lexical ambiguity resolution in auditory
sentence processing employing a CMLP experiment with divided field presentation
of visual priming target probes. In this study lexically ambiguous words were
embedded in (not at the end of) auditory sentences which contained strong a-priori
contextually biasing material (As in the example given previously in this paper
concerning the ambiguity "PEN"). Visual lexical decision targets related to the
primary and secondary meanings of the ambiguity were presented to either the LH
or RH. These words were presented in normalleft-to-right word format, to the left or
right extra-foveal area. They were presented such that the innermost (most nasal)
letter of the word presented to each visual field fell at least 2° from center. Thus,
only one (L or R) visual cortex initially received and processed this visual
information. These words were presented either immediately after hearing the
lexical ambiguity in the sentence, or 750 msecs later (in both cases, during ongoing
processing of the sentence).10 Analysis of these data demonstrate that in the LH,
priming for both interpretations of a lexical ambiguity is significant at the point
immediately after the ambiguity was processed. However, at this same immediate
test point, only priming for the contextually-relevant (which also was the most
frequent) interpretation of the ambiguity is demonstrated in the RH. However, when
tested at the longer (750 msec.) lSI, the LH demonstrates priming only for the (more
frequent) contextually relevant interpretation, while the RH demonstrates priming
for both interpretations.
Overall, this body of work is thus quite in accord in the conclusion that the LH is
involved in the initial rapid access of ALL interpretations of a lexical item, even in
the face of prior biasing context. This, of course fits well with the findings from the
lesion evidence, presented above. Relatedly, it appears that the LH is involved in the
rapid post-access selection (resolution), maintenance, and integration of contextually
relevant meanings of words. The RH appears to have the capacity for slowly
activating and maintaining the less-frequent (and in these cases, contextually-
irrelevant) interpretations of word senses - something that may come into play, for
example, in re-processing the sentence when the wrong interpretation of the
ambiguous word was chosen early in the sentence.
It is worth noting that exists a separate literature on brain lesioned individuals
that supports this view of the role of the RH in lexical processing. Tompkins,
Baumgaertner, Lehman and Fossett (1997) conducted a study with RH Damaged
individuals involving auditorily presented sentences with sentence-final lexical
ambiguities. They employed an interference task involving the presentation of visual
targets one second following the end of the sentence. In this, they report that the RH
Damaged individuals (as opposed to non-impaired controls) demonstrated difficulty
in suppressing the contextually inappropriate meaning of the ambiguities. The
authors argue that this lends support to the role of the right hemisphere in
maintaining alternate the (secondary) interpretations of lexical ambiguities.
5. SUMMARY
A clear story concerning the nature and time-course of context effects on lexical
processing during auditory sentence comprehension only emerges via the integrated
examination of evidence from studies of lesion patients, studies involving
hemispheric isolation, and the general behavioral on-line processing studies (which
were performed independently from any brain-basis concerns). It appears from all of
this that the left and right hemispheres work together to produce the findings that
populate each of those literatures in isolation - findings that in isolation have often
to be at variance.
Overall, it can be seen that context does not place prior constraints on lexical
access during auditory sentence processing. All information associated with an
auditory lexical form is made available to ongoing sentence processing when it is
first encountered in an auditorily presented sentence (and at any time that it is re-
encountered during the same discourse, see Endnote 7). Contexts (of every type thus
far tested) have their effects only following this initial exhaustive access process.
The role of the Left Hemisphere in this process appears to be that which underlies
this initial form-driven exhaustive lexical access and the subsequent rapid post-
access effects of context upon this accessed material during ongoing auditory
sentence comprehension. Both studies of the LH and RH in isolation and of the
processing found in patients with focal brain lesions, supports the view that,
immediately upon "hearing" an ambiguity, there is rapid and exhaustive access of all
meanings of the ambiguity made available in the Left Hemisphere, independently of
prior contexts. It appears from lesion evidence that the brain tissue in and around
Broca's area (anterior, frontal lobe, cortical areas of the LH) is deeply involved in
the exhaustive, form-driven, fast-acting aspect of lexical access that has been so
often demonstrated in studies of "normal" auditory processing involving lexical

ambiguities. Subcortical areas (nonthalamic basal gangular regions) in support of
this cortical region appear critically involved in the rapid, contextually-directed
selection and maintenance of a single interpretation of lexical information, an
interpretation that is necessary for ongoing sentence interpretation. When these LH
areas of the brain are undamaged, and no re-processing of information is necessary,
it appears that these subcortical regions are primarily responsible for the course of
information processing that typically determines auditory sentence comprehension
(in so far as lexical processing and contextual integration are concerned).
However, because of the very rapid nature of the process which eventuates in
choice of a single lexical meaning for a word (and apparent suppression of other
meanings ll ) in the left hemisphere during the processing of auditory sentences, two
things hold. First, the left hemisphere typically produces an integrated interpretation
of material before the RH does. As such, this LH interpretation is used in all further
discourse processing (NB: this would follow from most models of the resolution of
competing solutions, such as in horse-race models; the LH interpretation is the
"winner"). Second, when "repairs" in sentence processing are required (e.g. as when
re-interpretation is required if an ultimately "incorrect" meaning of an ambiguity is
initially "chosen"), it appears that it is only the RH which maintains "alternative"
interpretations which allow recovery of correct meanings 12.
Similarly, when the LH is damaged (as in Broca's aphasia, etc.) lexical-
contextual processing must (and apparently does) rely to some extent on lexical
processing from the RH. Thus, overall, the RH appears to play an important
potential role in auditory sentence interpretation, and, in particular, in the
interpretation of context effects during lexical processing. Thus, there are different
potential "modes" of processing that may cause different processes to be called into
play in such auditory lexical processing.
The RH appears to make "all" meanings of ambiguous words available, but only
after they are already available in the LH (which may account for the apparent "slow
rise time" of secondary meanings found for ambiguities in Broca's aphasics; if not
for the RH, it is possible that, in some cases, these less frequent meanings might not
"arise" at all). However, as suggested above, this effect is not reflected in the
processing required in most "standard" on-line studies of normal sentence
processing. Thus, it appears that while the RH actively processes lexical (and
perhaps sentential) information, it may have little to no contribution to rapid first-
pass analysis of language in sentence processing conditions involving non-brain-
damaged individuals. 13 However, re-processing and re-interpreting is a common
part of language comprehension - for reasons ranging from those as mundane as
"lack of attention" or "conflicting contexts" to those as profound as "brain-damage".
Thus, it seems the RH will often have a role to play in lexicaVsentence processing.
And thus, while the above exposition in this paper appears to capture much of what
we know and have evidence for concerning lexical access and context effects rapid,
first-pass auditory sentence interpretation processes, it is only part of the larger
story. An understanding of all of the ways in which language processing may take
place is critical to any overall model of comprehension. And these clearly often go
well beyond initial, unrepaired, rapid first-pass analysis. Which brings us to:
6. A FINAL CONSIDERATION TOWARD UNDERSTANDING LANGUAGE

COMPREHENSION: THE CONCEPT OF MODES OF PROCESSING
There is, of course, some research in existence that has not been interpreted to
support the integrated story we have just presented. Evidence exists which, for
example, has been claimed to demonstrate prior-constraint (predictive) effects of
context upon lexical access. In nearly all cases, when the initial tenets which we
briefly presented to begin this chapter are carefully considered (e.g., no intention to
account for processing in different (i.e., visual) modalities; or, no intention to
account for processing evidence derived from isolated word studies; or, no intention
to be concerned with evidence obtained via tasks that have demand characteristics
that force integration of context in the very measure obtained 14; etc.) the critical
reasons that "different" interpretations are sometimes given can typically be
discerned. However, it is a Herculean (and likely endless) task - and certainly not
one possible in the constraints of this chapter - to report on each such study, and
explain each basis for interpretive difference.
It has become clear to us in investigating these various accounts carefully that
some seemingly subtle processing differences can make an enormous difference in
terms of the form that data and apparent interpretations take. As they say - the devil
is in the details. Rather than attempting to explain each detail here, we will make a
brief presentation concerning what we have come to understand as the critical
characteristic(s) that need to be considered in all such enterprises.
Primarily, in order to make a coherent story of any of the sentence processing
literature (whether aimed at lexical, contextual, structural or discourse processes) the
field (language processing) must adopt the theoretic of a "modes of processing"
account of language. What is implied in this is that there are many ways in which we
accomplish language processing, and without precise understanding of ALL of the
characteristics of EACH such "way" we are doomed to continue contrasting and
arguing over interpretation of data that are as deeply different as are apples and
pianos (merely employing the metaphoric contrast of apples and oranges doesn't
accurately portray the diversity to be found in the language processing field). The
concept of "mode" obviously covers a lot of ground. We have dealt some key
characteristics of different modes earlier in this chapter (modality, whether in
sentence context or not, etc.) However, a number of exceedingly subtle differences
in processing situation are also important. For example, work in our lab over the past
4-5 years has demonstrated that a variable as simple as the rate of information input
in comprehension vastly changes both the "demand" characteristics of several
standard tasks typically used to study lexical and structural processing as well as the
way in which the comprehension process itself is performed. And, this involves
variation of "rates" of processing throughout a range standardly considered
"normal". Similarly, modulations in processing as slight as adding regular and
consistent noise to the speech signal, appears to change (in marked ways) the
underlying on-line manner in which processing is typically performed on speech
stimuli in sentences.
Thus, it should not come as any surprise that experimental conditions can be
found in which it appears that context can be used to "predict", and thus limit,
selection of lexical information. Simply put, humans do have the cognitive capacity
to predict things. For example, when subjects are encouraged by experimental task
demands to anticipate goals, outcomes, or upcoming events (either explicitly or

implicitly), they will and can make such predictions. A careful perusal of the
language processing literature will discover a large array of conditions which
provide such task demands.
However, the overwhelming majority of studies examining auditory
sentence/discourse comprehension in which the task demands do not encourage
prediction (and where the experimental task is sufficiently sensitive to detect subtle
on-line processes) have clearly revealed that lexical access is a form-driven, context-
independent process. Rather, prior contexts have been shown to only have effects
following initial access to stored lexical information. Critically, it is precisely this
normal (rate and task), fluent "Mode of Processing" that we are attempting to model
in our studies of language comprehension. In such a "mode", the details of the
processing of lexical and contextual material in sentences, as well as their
neurological underpinnings, appear consistent and robustly replicable.
There are, as noted above, alternative "modes of processing" that may be
engaged in certain language processing conditions. Whether the robust findings
reported here for the Fluent Auditory Mode of processing will also account for the
processing employed in some of these other situations remain an empirical issue.
But, it is only by carefully detailing the modes of processing investigated in each
and every case that will we discover the answer.
7. ACKNOWLEDGEMENTS
The authors gratefully acknowledge support from the Max Planck Institute for
Cognitive Neuroscience for supporting the writing of this chapter and NIH grants
DC 02984 and DC 03681 for the supporting the research presented herein. We also
wish to thank Vikki Bouck for her contributions to projects reported in this chapter.
8. AFFILIATIONS
Dr. David Swinney

Dr. Tracy Love
Department of Psychology
University of California, San Diego
La Jolla, California
e-mail: dswinney@ucsd.edu
9. NOTES
I NB: This is not a question concerning whether prior context ultimately constrains the lexical
information used in a final interpretation of a sentence; the answer to that question is transparently
obvious-it clearly does so. This is, rather, a question concerning WHEN it does so.
2 There are, of course, some studies that have reported findings in disagreement with this general
summary; these will be discussed below under the heading "modes of processing"
J See Nicol, Fodor, and Swinney, 1994; Swinney, Prather, and Love (2000; endnote i), Swinney, Nicol,
Love, & Hald (in press, 1997) for discussion of various design-types, and their relative advantages and
disadvantages, in CMLP studies.
4 We present here only results for the first of several probe positions tested in this study; the later probe
positions examined an entirely different aspect of processing - canonical structural re-ordering - which
is beyond the scope of the current review.)
5 These controls were in response to issues originally raised in papers by McKoon and Ratcliff (1996) and
McKoon, Ratcliff, and Ward (1994), which, however, have since been demonstrated to be unfounded
concerns ( see Nicol, Swinney, Love, & Hald, 1997).
6 Note that the absolute numerical value of priming in these types of studies is, as always, uninterpretable;
the absolute size of priming to probes is influenced by all items in the sentence (including the context
itselt), and hence priming for the ,contextually related' probe is often numerically larger than that for
the probe to the ,contextually unchosen' meaning. Thus, absolute priming size is irrelevant and
uninterpretable in these studies (here, at least, size doesn't matter!). It is only as an existence proof (the
existence or lack-thereof of a priming effect for probes related to each interpretation) that has
interpretable substance. (See Swinney, 1979, 1981, 1982 for discussions of this issue).
7 Even after a "contectually relevant" meaning of a lexical ambiguity has been chosen during sentence
processing, when the system is presented again with the homophone, ALL meanings are again
momentarily accessed (Love & Swinney, 1998).
8 See Swinney, 1999; Dronkers, 1996; Friederici, 1995,among many other, for more details and
discussion concerning the precise substrates typically associated with Broca's aphasia)
9 Note that this seeming immediacy of availability of "all meanings" of an ambiguous word for non-
neurologically involved individuals may simply be a "relative immediacy". That is, it may be that we
currently have no task with sufficient sensitivity to detect very small differences in the time-course of
activation of meanings of an ambiguous word that might always exist, even in "normal" populations. It
may thus be that damage to Broca's area merely exaggerates existing differences that are too subtle to
detect in persons without damage to that neural area. This, however, does not lessen the conclusions
concerning the impact and role that this neurological substrate plays in allowing rapid availability of all
meanings of a word to ongoing sentence processing operations.
10 These timings were chosen based on earlier visual hemi-retinal priming paradigm studies of visual
(isolated) word processing by Burgess and Simpson (1988) which had demonstrate that the LH
provides activation of multiple interpretations (primary and secondary meanings) of ambiguous words
immediately upon viewing the word. However, by 750 msecs later, only the primary (more frequent)
interpretation of the ambiguity was actively maintained (primed). In contrast, the RH appeared to
initially only have access to the more frequent interpretation of an ambiguous word, and "exhaustive"
availability of both meanings of an ambiguous word (as measured via priming) at longer temporal
delays (750 msecs.). Thus, in this instance, both the visual and the auditory studies vector to the same
findings.
II See, for example, works by Gernsbacher and Faust(1991) or Friederici, Steinhauer, Meckiinger, &
Meyer (1998) for discussions of and evidence for such possible mechanisms.
12 Such need for reprocessing can be seen, for example, in classic psycholinguistic demonstration
materials such as: "The old man couldn't find his glasses in the dimly lit room, but finally found them
in the corner, filled with wine".
IJ For more on first-pass processing issues, see, e.g., Friederici (1995) or Hahne and Friederici (1999)
14 See, for example, discussions of such tasks in Swinney, Prather, and Love (2000) and in Swinney,
Nicol, Love, and Hald (in press; 1997)
10. REFERENCES
Balogh, 1., Zurif, E., Prather, P., Swinney, D., & Finkel, L. (1998). Gap-filling and end-of-sentence
effects in real-time language processing: Implications for modeling sentence comprehension in
aphasia. Brain & Language, 6, 169-182.
Burgess, C., & G. Simpson. (1988). Hemispheric processing of ambiguous words. Brain & Language, 33,
86-104.
Copland, D. (2000). A real-time examination of lexical ambiguity resolution following lesions of the
dominant nonthalamic subcortex. Ph.D Dissertation, University of Queensland.
Copland, D., Chenery, H., & Murdoch, B. (2000). Processing lexical ambiguities in word triplets:
Evidence of lexical-semantic deficits following dominant nonthalamic subcortical lesions.
Neuropsychology, 14, 379-390.
Crosson, B. (1985). Aubcortical functions in language: A working model. Brain and Language, 25, 257-
292.
CONTEXT EFFECTS AND AUDITORY SENTENCE COMPREHENSION 39
Dronkers, N. F. (1996). A new brain region for coordinating speech articulation. Nature, 384, 159-161.
Faust, M., & Chiarello, C. (1998). Sentence context and lexical ambiguity resolution by the two
hemispheres. Neuropsychologia, 36, 827-835.
Fodor, J. A (1983). The modularity of mind. Cambridge, MA: MIT Press.
Foss, D. J. (1998). Two strands of scholarship on language comprehension: Phoneme monitoring and
discourse context. Journal of Psycho/inguistic Research, 27, 191-201.
Foss, D. J., Starkey, P., & Bias, R. (1988). Sentence comprehension processes in the Pre-schooler. In R.
N. Campbell & P. T. Smith (Eds.), Recent advances in the psychology of language: Formal and
theoretical approaches. New York: Plenum.
Fuentes, L., & Santiago, E. (1999) Spatial and semantic inhibitory processing in schizophrenia.
Friederici, A (1995). The time-course of syntactic activation during language processing: a model based
on neuropsychological and neurophysiological data. Brain and Language, 50, 259-281.
Friederici, A, Steinhauer, K. Mecklinger, A, & Meyer, M. (1998). Working memory constraints on
syntactic ambiguity resolution as revealed by electrical brain responses. Biological Psychology, 47,
193-221.
Gernsbacher, M., & Faust, M. (1991) The role of suppression in sentence comprehension. In Simpson
(Ed), Understanding word and sentence (pp. 97-128). Elsevier: North Holland. Amsterdam
Hahne, A, & Friederici, A (1999). Electrophysiological evidence for two steps in syntactic analysis:
early automatic and late controlled processes. Journal of Cognitive Neuroscience, 11, 194-205.
Love, T., & Swinney, D. (1996). Coreference processing and levels of analysis in object-relative
constructions: Demonstration of antecedent reactivation with the cross-modal priming paradigm.
Journal ofPsycholinguistic Research, 25, 5-24.
Love, T., Swinney, D., Bagdasaryan, S., & Prather, P. (1999). Real-time processing of lexical ambiguities
by pre-school children. Paper presented at the 12'h annual CUNY Conference on Human Sentence
Processing: New York.
McKoon, G., & Ratcliff, R. (1996). Sentential context and on-line lexical decision tasks. Journal of
Experimental Psychology: Language, Memory and Cognition, 20, 1239-1243.
McKoon, G., Ratcliff, R., & Ward, G. (1994). Testing theories of language processing: an empirical
investigation of the on-line lexical decision task. Journal of Experimental Psychology: Language,
Memory and Cognition, 20, 1219-1228.
Miyake, A, Just, M. A, & Carpenter, P. A (1994). Working memory constraints on the resolution of
lexical ambiguity: Maintaining multiple interpretations in neutral contexts. Journal of Memory &
Language, 33, 175-202.
Nicol, 1., Swinney, D., Love, T., & Hald, L. (1997). Examination of sentence processing with continuous
vs. interrupted presentation paradigms. Center for Human Information Processing Technical Report
No. 97-3. University of California, San Diego.
Onifer, W. M. (1980). Associative intrusions in schizophrenic language. Tufts U. Dissertation Abstracts
International, Sep, v41 (n3-B):1122.
Onifer, W., & Swinney, D. (1981) Accessing lexical ambiguities during sentence comprehension: Effects
of frequency-of-meaning and contextual bias. Memory and Cognition, 9, 225-236.
Picouit, J., & Johnson, M. (1992). Controlling for homophone polarity and prime-target relatedness in the
cross-modal lexical decision task. Bulletin of the Psychonomic Society, 30, 15-18.
Posner, M., & Raichle, M (1994). Images of mind. New York: Scientific American Library.
Prather, P., & Swinney, D. (1977). Some effects of syntactic context upon lexical access. Paper presented
at a meeting of the American Psychological Association, San Francisco, California, August 26.
Prather, P. A, Zurif, E., Love, T., & Brownell, H. (1997). Speed of lexical activation in nonfluent Broca's
aphasia and fluent Wernicke's aphasia. Brain & Language, 59, 391-411.
Seidenberg, M. S., Tanenhaus M. K., Leiman, 1. M., & Bienkowski, M. (1982). Automatic access of the
meaning of ambiguous words in context: some limitations to knowledge-based processing. Cognitive
Psychology, 14, 489-537.
Simpson, G. B. (1981). Meaning dominance and semantic context in the processing of lexical ambiguity.
Journal of Verbal Learning and Verbal Behavior, 20, 120-136.
Swaab, T., Brown, C., & Hagoort, P. (1997). Spoken sentence comprehension in aphasia: Event-related
potential evidence for a lexical integration deficit. Journal of Cognitive Neuroscience, 9, 39-66.
Swinney, D. (1979). Lexical access during sentence comprehension: (Re)consideration of context effects.
Swinney, D. (1982). The structure and time-course of information interaction during speech
comprehension: Lexical segmentation, access, and interpretation. In J. Mehler, E. T. C. Walker, & M.
Garrett (Eds.), Perspectives on mental representation (pp. 151-166). Hillsdale, N.J.: Erlbaum
Associates Publishers.
Swinney, D. (1991). The resolution of indeterminacy during language comprehension: Perspectives on
modularity in lexical, structural and pragmatic process. In G. B. Simpson (Ed.), Understanding word
and sentence (pp. 367-385). Amsterdam, Netherlands: North-Holland.
Swinney, D., Nicol, I., Love, T., & Hald, L. (in press) Methodological issues in the on-line study of
language processing. In R. Schwartz (Ed.), Childhood language disorders. Hillsdale, N. Y.: Erlbaum.
Swinney, D., Onifer, W., Prather, P., & Hirshkowitz, M. (1979). Semantic facilitation across sensory
modalities in the processing of individual words and sentences. Memory and Cognition, 7, 159-165.
Swinney, D., & Prather, P. (1989). On the comprehension of lexical ambiguity by young children:
investigations into the development of mental modularity. In D. Gorfein (Ed.), Resolving semantic
ambiguity (pp. 225-238). New York: Springer-Verlag.
Swinney, D., Prather, P., & Love, T. (2000). The time-course of lexical access and the role of context:
Converging evidence from normal and aphasic processing. In Y. Grodzinsky, L. P. Shapiro, & D. A.
Swinney (Eds.), Language and the brain: Representation and processing (pp. 273-294). New York:
Academic Press.
Swinney, D., Zurif, E., & Nicol, I. (1989). The effects of focal brain damage on sentence processing: an
examination of the neurological organization of a mental module. Journal of Cognitive Neuroscience,
1,25-37.
Tabossi, P. (1988). Accessing lexical ambiguity in different types of sentential contexts. Journal of
Memory & Language, 27, 324-340.
Tanenhaus, M. K., Leiman, 1. M., & Seidenberg, M. S. (1979). Evidence for multiple stages in the
processing of ambiguous words in syntactic contexts. Journal of Verbal Learning and Verbal
Behavior, 18, 427-440.
Titone, D. (1998). Hemispheric differences in context sensitivity during lexical ambiguity resolution.
Brain & Language, 65, 361-394.
Tompkins, c., Baumgaertner, A., Lehman, M. T., & Fossett, T. (1997). Suppression and discourse
comprehension in right brain-damaged adults: A preliminary report. Aphasiology, 11, 505-519.
Tyler, L., & Marslen-Wilson, W. (1982). Speech comprehension processes. In 1. E. Mehler, T. C. Walker,
& M. Garrett (Eds.), Perspectives on mental representations (pp. 169-182). Hillsdale, N.J.:
Erlbaum.
J. D. SADDY & P. BEIM GRABEN
MEASURING THE NEURAL DYNAMICS OF

LANGUAGE COMPREHENSION PROCESSES
Abstract. Recordings of electrical activity generated in the brain in response to specific stimuli now
provide an important source of information about the temporal and topographical distribution of language
processing events in the brain. Many of our theories of language processing are based upon reaction time
studies and use the new brain based findings sparingly. One reason for this is that much of the results
from brain recordings are seen to be primarily recapitulations of reaction time findings. A new approach
to analyzing evoked electroencephalographic data, Contrastive Signal Coherence (CSC) assesses the
coherence of the evoked signal using the techniques of non-linear systems analysis. This determination of
signal quality, combined with the measure of signal quantity as calculated in the standard voltage
averaged approach to Evoked Brain Potentials (ERPs), provides a richer reflection of the dynamic
properties of language related brain potentials and significantly enhances our view of cognitive cortical
events. In this paper, we present the results of an ERP study in which we combined the traditional voltage
averaged analysis and CSC to address two competing accounts of Case ambiguity resolution, a data
driven, expectation based account following Schlesewsky (1997) and an information driven Diagnosis
and Repair account following Fodor and Inoue (1994). Both approaches are considered in the context of
Friederici's (1995) three stage description of the ERP correlates of language processing. The basic
findings provide evidence in favor of the three stage model and motivate an alternative account of Case
ambiguity resolution.
1. CASE AMBIGUITY RESOLUTION

It has been noted that, on a variety of behavioral measures, subjects show differing
responses to unexpected resolution of Case ambiguity dependent upon the source of
the disambiguating information. One source is research on processing of German
sentences containing Case ambiguous initial NPs (see also Sekerina (in press) for
Russian). In the following examples the initial NP is ambiguous between
Nominative and Accusative Case. There are two ways such an ambiguity can be
resolved: via contrast with a morphologically Case marked second NP as illustrated
in A or via number agreement or disagreement between NPl and the verb, as
illustrated in C. Example B below represents the unmarked condition in which the
initial NP is the nominative argument of the sentence. All three sentences are
unmarked and well formed in German. Examples such as those in A will be referred
to as the Case condition. Examples such as those in B will be referred to as the
Control condition and examples such as those in C will be referred to as the Number
condition.
1.1 Matrix Constructions
A. Welche Frau besuchte der Richter

Which woman [sng] (Amb) visited[sng] thejudge[sng] (Nom)

@ 2002 Kluwer Academis Publishers.
42 1. D. SADDY & P. BEIM GRABEN
B. Welche Frau besuchte den Richter

Which woman[sngj (Amb) visited[sngj the judge[sngj (Acc)
C. Welche Frau besuchten die Richter
Which woman[sngj (Amb) visit[plj the judges[plj (amb)
A series of self paced reading and reaction time experiments examining such
constructions (Hopf, Bayer, Bader, & Meng, 1998; Meng & Bader, 1996; Meng,
1998; Schlesewsky, 1997) showed a very stable reanalysis effect indicated by
prolonged reaction times in the number condition, but no reliable reanalysis effect
was found for the Case condition.
The first result is consistent with the well established preference to interpret an
initial NP as subject. If NP1 is taken to be nominative then it should agree with the
verb. A longer reading time or reaction time associated with a verb that does not
agree in number with a Case ambiguous initial NP is consistent with this strategy.
The second result is surprising. If NP1 is taken to be nominative, the occurrence of
overt nominative marking on the second NP should cause some processing
difficulty. This apparent lack of reaction to unexpected Case information has given
rise to a rich discussion on the general nature of reanalysis and ambiguity resolution.
Either the second Nominative is not noticed or discarded during language
processing, or the reanalysis is so easy that it imposes no measurable cost. Both
approaches have been proposed and both make differing predictions regarding the
resolution of similar Case ambiguities in other constructions.
Expectation based or data based processing as proposed in Schlesewsky (1997)
and Schlesewsky, Fanselow, Saddy, and beim Graben (1998) takes a cue validity
approach to explaining the basic reaction time results. The strong reanalysis effect
associated with the Number condition is accounted for as follows. The initial NP is
strategically assumed to be nominative. The first available information that could
bear on this decision is number agreement on the verb. If the verb does not agree in
number with the initial NP, then the sentence is ungrammatical at this point and
must be reanalyzed. The lack of an effect in the Case condition is explained by the
late occurrence of the cue. The initial NP is strategically assumed to be nominative.
In the Case condition the verb does have the same number marking as the initial NP
thereby providing indirect support for its nominative status. The occurrence of a
second nominatively Case marked NP leads to indeterminacy. The fact that the verb
agrees in number with the initial NP leads the parser to discount the overt Case
marking on the second NP. Two observations favor this approach: (1) The marking
of number agreement in German is not ambiguous. The fact that the Case system in
German contains systematic ambiguities l means that Case information is a weaker
information source than number information. Thus the parser may be expected to
value number information over Case information for ambiguity resolution. (2) Other
studies (Meng & Bader, 1996) have shown that subjects are more sensitive to
number marking than to Case marking in grammaticality judgements. Taken
together these observations support the suggestion that Case is somehow a weaker
information type than number in the constructions examined.
Alternatively, an information based or diagnosis and repair account (Fodor &
Inoue, 1994, 1999a, 1999b) offers a modular approach. Number agreement and Case
MEASURING THE DYNAMICS OF LANGUAGE PROCESSES 43
are two distinct information systems employed in language processing. Given the
assumption of modularity, it should be more difficult to recognize and resolve
parsing conflicts that cross modules than it is to recognize and resolve parsing
conflicts that arise within only one module. Fodor & Inoue's account of the Number
condition argues that the number agreement information carried by the verb
provides a costly diagnosis of the Case strategically assigned to the initial NP.
Furthermore, the number information provides no direct information regarding
repair. The strategic assumption is that the initial NP is the subject of the upcoming
predicate. The fact that the verb does not agree with the initial NP provides the
information that the initial NP is not the subject of predication but it does provide or
point to an alternative. The relative cost of using number information to resolve a
Case ambiguity is reflected in the reaction time. The Case condition contrasts with
the Number condition in both diagnosis and repair. The overt Nominative Case
marking on the second NP bears directly on the Case of the first NP hence the
diagnosis takes place within the same information system. Furthermore, the overtly
Nominative second NP provides the repair. It is the new subject of predication. The
lack of a reaction time effect associated with the Case condition reflects the
comparative ease of the ambiguity resolution. We show below that evoked response
potential recordings provide additional evidence relevant to the debate.
1.2 Interpreting Evoked Cortical Potentials
Recording the continuous voltage time series generated by the brain in response to
varying but highly constrained stimuli now provides an important source of
information about the time course of language processing and its functional
fractionation. Friederici (1995, 1999) proposes a three-phase model for interpreting
the behavioral markers associated with sentence processing. The model is based
upon the systematic observations found in the voltage averaged responses in a wide
range of ERP investigations as well as tMRI and MEG studies and provides an
articulated guide to interpreting brain based responses in sentence processing. In
Friederici's model, brain behavior associated with sentence processing detected in
early time periods (120-200 ms) corresponds to a first pass parse. Brain behavior
detected during the period ranging between 300 and 500 ms corresponds to semantic
and thematic integration and diagnosis, while brain behavior detected in the later
time periods, 500 to 900 ms, reflects structural repair or attempted repair processes.
This model describes three fundamental steps necessary to sentence processing and
connects them to particular time windows and voltage polarities. We interpret the
model as providing a working guide to when to expect contrastive effects in the
evoked time series. The results of the study described here provide new support for
this model.
The recorded continuous EEG is a multivariate time series that is stored as a
matrix containing the voltage time series associated with each of the recording
electrodes arrayed across the scalp plus a channel containing trigger codes which
mark the stimuli events in time. Mter filtering and artefact rejection the continuous
EEG is split into epochs according to the trigger codes. Each epoch is a short time
series consisting of the pre and post stimulus interval recorded to a single trial.
These epochs are assembled into ensembles according to the contrasting
44 1. D. SADDY & P. BEIM GRABEN
experimental conditions. These procedures are common to the two analytic

approaches employed here. The treatment of the data ensembles under the voltage
averaging approach, which yields the standard ERP component analyses, and the
CSC approach diverge significantly from this point on.
1.3 Voltage Averaged Markers (ERPs)
The voltage averaging approach characterizes the collected time series data in
quantitative terms. What is the polarity of the voltages with respect to a reference
electrode? How big is the amplitude of these deflections? When do these deflections
take place? What is their topographic distribution? Four main markers of language
processing have been identified in the literature; the ELAN, LAN, N400 and P600.
They are identified by a nomenclature which refers to their polarity, post stimulus
latency and topographic distribution. The early left anterior negativity or ELAN
occurs between 120 and 220 ms with either a left or a bilateral anterior distribution.
It is associated with strong phrase structure violations (Friederici, Pfeifer, & Hahne,
1993; Hahne, 1998; Neville, Nicol, Barss, Forster, & Garrett, 1991). The left
anterior negativity or LAN has similar topography and polarity to the ELAN but
occurs with a latency typically between 300 and 500 ms. The LAN is found in
response to agreement violations (Coulson, King, & Kutas, 1998; Gunter, Stowe, &
Mulder, 1997; Kutas & Hillyard, 1983; Osterhout & Mobley, 1995) and has been
noted for Case violations on pronouns in English (Coulson et aI., 1998). The N400 is
a negativity with a latency peaking typically around 400 ms and having a centro-
parietal bilateral distribution often with a slight right hemisphere focus. The N400
reflects the cost of semantic or thematic integration (Frisch, 2000, Kutas & Hillyard
1980a, b, 1983). The P600 or syntactic positive shift is a positivity occurring
between 600 and 900 ms with a centro-parietal distribution and is associated with
syntactic reanalysis (see Friederici, 1999; Friederici, Hahne, & Meckiinger, 1996;
Hagoort, Brown, & Groothusen, 1993; Hahne & Friederici, 1999; Osterhout, 1997;
Osterhout, McLaughlin, & Bersick, 1997)
The voltage averaging analysis is based on a number of assumptions. The central
assumption underlying the averaging is that what is recorded is the evoked signal
embedded in noise. Each ERP epoch is assumed to be a realization of a stochastic
process which contains an invariant evoked signal. The evoked signal is assumed to
be uncorrelated with the surrounding EEG activity. It is further assumed that there is
no evoked signal in the pre-stimulus interval. Each recorded epoch is assumed to be
a stationary and ergodic process. These assumptions are necessary in order to
motivate the baseline correction calculation and subsequent signal averaging at the
core of the traditional analysis. Baseline correction exploits the assumption that the
pre-stimulus interval is uncorrelated with the evoked signal. The pre-stimulus
interval therefore is representative of only the EEG noise in which the evoked signal
is embedded. The time averages of the pre-stimulus intervals are subtracted from
their corresponding epochs. What remains is presumed to be the signal evoked in
response to the experimental stimuli. The evoked signal itself is very weak
compared to the background EEG. In order to further improve the signal to noise
ratio many trials per condition and per subject and many subjects per experiment are
averaged together (typically 30 trials per condition and 16-20 subjects per
experiment in sentence processing paradigms). The resulting averaged voltage time

series are plotted and compared statistically to reveal systematic contrastive
behavior generated in response to the experimental conditions. There are two central
weaknesses associated with this approach. The first is that the basic assumptions
outlined above are known to be false 2• The second is that the averaging after
baseline correction is affected by trial to trial variability in amplitude and onset
latency. The consequence of this is that the traditional analysis necessarily
underestimates contrastive behavior in the recorded signal (see beim Graben 2000
for a critical discussion of ERP analytic techniques). As a consequence of the
inherent analytic weakness the standard ERP analyses provide a conservative but
reliable picture of when contrastive cortical behavior might be found. Thus the
standard voltage averaged analyses provides the basic framework from which other
analytic approaches can be interpreted and judged.
2. CYLINDER ENTROPIES AND CONTRASTIVE SIGNAL COHERENCE

(CSC)
We know that the brain behaves as a complex integrated system. Despite this, most
of our techniques for analyzing the behavior we can measure in the cortex either
assume the opposite or are insensitive to the complexity of the system. Previous
attempts to apply the techniques of non-linear system analysis to electroence
phalographic (or Magnetoencephalographic) recordings have shared one central flaw
with the conventional subtractive techniques, that is the assumption of stationarity of
the system or ergodicity in the data stream. beim Graben, Liebscher and Saddy
(2000) develop a technique which does not have this shortcoming. This approach
which identifies changes in signal coherence in response to experimental stimuli,
rests upon the calculation of Cylinder Entropies (beim Graben, 2000; beim Graben,
Saddy, Schlesewsky, & Kurths, 2000). This non-linear technique provides a measure
of the change in information entropy associated with the signals generated by the
processing system over time. More precisely, the Cylinder Entropies provide a
measure of changes in phase or state space volume. Episodes of Contrastive Signal
Coherence (CSC) as identified by the calculation of Running Cylinder Entropies
reveal temporal and topographic coherency analogues of the ERP components in the
signal as well as revealing contrastive signal behavior that cannot be identified with
ERP components. In addition the technique provides a significant signal to noise
improvement. Thus the computation of CSC can enhance the ERP based view of
cortical behavior.
The idea of applying the techniques of non-linear system analysis to brain
recordings is not new (see Albano et aI., 1986; Babloyantz, Salazar, & Nicolis,
1985; Basar, Flohr, Haken, & Mandell, 1983; Bullmore et aI., 1994; Freeman, 1990;
Gallez & Babloyantz, 1991; Layne, Mayer-Kress, & Holzfuss, 1986; Lutzenberger,
Elbert, Birbaumer, Ray, & Schupp, 1992; Pritchard & Duke, 1992). Most
approaches have used the estimation of correlation dimension as a method for
determining the change in computational complexity of a given brain process. This
and related approaches have proved to have limited applications because the
calculation of correlation dimensions requires the measurement of a stationary
system (see Molnar, Skinner, Csepe, Winkler, & Karmos, 1995 for another approach
46 J. D. SADDY & P. BEIM GRABEN
that avoids the assumption of stationarity). Unfortunately, ERP recordings are

anything but stationary. One important property of the calculation of Cylinder
Entropies is that it does not require or assume ergodicity of the system being
analyzed (see beim Graben, 2000; beim Graben, Liebscher et a!., 2000; Saddy &
beim Graben, 1999; Saddy, beim Graben, & Schlesewsky, 1999; for additional
exposition).
The measure of signal coherence calculated in this technique is a reflection of the
overall system dynamics associated with an ensemble of recording epochs for a
given experimental condition. The technique for capturing the system dynamics is
the Symbolic encoding of the raw data streams, the voltage time series associated
with each electrode. A Cylinder is the collection of symbolically encoded time series
recorded to all trials of a given condition at a given time point. Two encodings have
been developed to date; an encoding of the half wave inflection points of the voltage
time series for a given temporal window (beim Graben et al. in preparation), and a
static encoding, with respect to the median value of the voltage time series (beim
Graben, Saddy et aI., 2000). Both encoding techniques convert the voltage time
series into a series of symbols (zeros and ones) representing global (coarse grained)
dynamic characteristics of the system underlying the signal. Thus the actual voltage
values, amplitudes and polarities for each recording epoch are discarded. However
the data stream is converted into a form from which Cylinder sets can be calculated
for a given ensemble of epochs for each electrode. From these ensembles, the
Shannon and other Reyni entropies for given Cylinder sets is computed (Shannon &
Weaver, 1949; Reyni, 1970). A distribution of Cylinder sets has low entropy if it
contains predominantly one of the encoding symbols. A distribution of Cylinder sets
has maximum entropy when there is an equal number of each of the encoding
symbols. These information entropies are a measure of the variation in the signal
properties. As a signal loses entropy it becomes more coherent and stable. As a
signal gains entropy it loses coherence. Thus in response to a particular stimuli, a
period of entropy decrease is a period of evoked signal coherence (ESC), a period of
entropy increase is a period of evoked signal decoherence (ESD).
By looking at the Cylinders in sequence one can calculate the running Cylinder
entropy for the entire ensemble of recording epochs. A common sample rate in the
digitisation of the evoked cortical potentials is 250 Hz., that is, a voltage
measurement is taken every 4 milliseconds. The calculation of Cylinder Entropies
for such a data set could provide a characterization of the entropy dynamics for the
signal at a given electrode site over time in 4 ms slices. More than one time slice
may be considered at a time (for half wave encoding an minimum window of two
time samples is required). Given a 250 Hz sample rate, broadening the window of
observation from a width of one sample time to two sample times allows us to
investigate the dynamics in a running window of 8 milliseconds, 3 sample points
represents a window of 12 ms and so forth. Increasing the size of the window allows
for latency effects in the dynamics. By looking across electrode sites one can
observe shifts in signal coherence over topography as well as time. Two statistical
methods are applied to the coherence data. First, a comparison of the actual signal
coherence against a suite of surrogate data sets. This statistic determines the
likelihood that a given signal could have been produced by chance. Second, a X2 test
is used to contrast the distribution of symbols in ensembles associated with

contrasting experimental conditions.
The interpretation of the coherence contrast should be approached with some
caution. It is tempting to over interpret these results as the vocabulary of the analytic
technique speaks of information carrying capacity etc. Event-related signal
coherence or entropy decrease in a region of the time series means that, in response
to a particular stimulus, the system behavior revealed in the signal is much more
restricted in the experimental condition than in the control condition. This describes
a reduction of the phase or state space of the system. Contrastive signal decoherence
or increased entropy in a region of the time series means that in response to a
particular stimulus the system behavior revealed in the signal is much less restricted
in the experimental condition than in the control condition. This describes an
expansion of the phase space of the system. The phase (or state space) of a non-
linear dynamical system is the set of all possible states of the system. Since our
experimental manipulations are linguistic in nature we can assume that the system
behavior revealed by our analysis of the ERPs is, at least in part, system behavior
that underlies language processing3• In beim Graben, Liebscher et al. (2000) we
developed a dynamical model of parsing in which states in the parser are mapped on
to Cylinder sets. A reduction in Cylinder entropies or ESC describes a period in
which the system trajectories or behaviors are constrained to pass through a reduced
phase space volume, a dynamic bottleneck. Here we interpret such ESCs as marking
a period of increased common processing, shared information properties and/or
successful information integration. An ESD or increase in Cylinder entropies
describes a period in which the system trajectories or behaviors pass through an
increased phase space volume. Here we interpret ESDs as marking a period in which
information properties associated with new regions of the state space are taken up,
either in response to new or contrastive information types or difficult or impossible
information integration.
2.1 The Experiment

A paradigm containing the Number and Case conditions described above was
prepared. There were 30 trials per condition. The sentences were presented visually,
word by word, with each word appearing for 400ms followed by a 400ms inter-
stimulus interval. Recordings were taken from 16 subjects using a 24 electrode
array. In the number condition recordings are reported from the verb. In the Case
condition recordings are reported from the overtly Case marked determiner of the
second NP.
In presenting the results here we will show conventional average ERP analyses
and CSC analyses together for purposes of discussion. We will display cortical
behavior for 3 time regions; around 200 ms., around 400 ms and around 600 ms.
These time regions were chosen to correlate with the time regions described in
Friederici's three stage model4• The in the discussion that follows we will refer
frequently to the three stage model. However, since we are looking at a new type of
analytic data we feel free to interpret her model very liberally with respect to time
windows.
In the following figures we will present idealized cranial maps (anterior is up).
There are three rows in each figure. The first row presents the differences of the
averaged voltages between the experimental condition and the control condition.
The second row presents the Entropy or Signal coherence differences between the
experimental condition and the control condition. The third row presents the
significance of the Entropy or Signal coherence differences5• The maps show only
selected relevant time periods. In the phase transition significance maps which show
log probability, black indicates at least a p< .001 confidence levd.
The Number Condition
Welche Frau besuchten die Richter
Which woman[sng] (Amb) visit[pl] the judges[pl] (amb)
As we noted earlier, reaction time results for the experimental conditions
investigated here showed a reanalysis effect only for the Number condition. We
present the recording to the verb in the Number condition first. In Figure 1 we see
the 600 ms post stimulus recordings generated to the verb in the number condition:
sentences in which the initial Case ambiguous NP is singular but the verb is marked
plural. What we see in voltage recordings is a strong centro- anterior positivity in the
600 ms time region 7. This is the P600 that is understood to reflect syntactic
reanalysis8, the expected voltage marker consistent with the reaction time findings.
6()() m$ window
Number mjSln"lCh min", rontrol Vohuge d<ffennc6
Number mis_tell mInus CIHltrol Pllase W/lsltIon rigni,[lCOncts
Figure 1. Number mismatch in matrix constructions around 600ms.
The entropy/ coherence maps for this period show a corresponding significant phase
transition realized as an entropy decrease or evoked signal coherence (ESC).
The fact that a significant voltage contrast is associated with a significant change
in signal coherence follows from the general properties of voltage averaging. In
order to get a significant contrast between a control and an experimental condition,
the voltage average for the experimental condition must be relatively stable but
distinct with respect to the control condition. Since signal stability is one aspect of
coherence, the standard voltage averaging technique can be viewed as a method for
identifying domains containing CSC in the time series. However contrastive signal
coherence can and does occur in regions in which no average voltage contrast
occurs. If two sets of voltage time series contain a high level of variance, the process
of averaging may eliminate any distinction between the two even if there is a
systematic contrastive pattern in the signal coherence. Cylinder Entropies can
distinguish contrastive but turbulent regions in the voltage record.
Figure 2 provides an examination of the 200 ms time period recorded to the verb
in the number mismatch condition. This time period corresponds to the first pass
parse region in Friederici's model. Here we find a significant entropy increase or
evoked signal decoherence (ESD) with an anterior and parietal right focus at around
200 ms. There is no significant corresponding voltage effect.
Number mU"...~Io minus ClNIIrol VohPg. difl.,..nc..
0'"...
NUmHr millftoich mln"s CIHlII'OI Etltropy dllle,..nc ..
-"
",
Number misntoich milt". control PIuJ.~ tronsiliqn signijictlllCes
In Figure 3 we examine the 400 ms time window. Here we find an anterior right
parietal entropy increase or ESD with no corresponding significant average voltage
marker.
The finding of a significant Coherence effect in the 400 ms time window is
consistent with the notion that the P600 can be decomposed into at least two
contributing components. Friederici et al. in press applied the technique of principle
components analysis to the P600 generated to object initial complements and object
relative clauses. Their analysis revealed that the P600 was composed of two
principle subcomponents; one associated with an early positivity around 350 ms the
other associated with the later positivity around 600 ms. The Number condition
requires a revision from a subject initial construction to an object initial
construction. In the above recording we see that there is an anterior positivity in the
voltage record however, in contrast to the ESD, this contrast did not reach statistical
significance.
Thus we do find significant contrasts in the ERP recording to the Number
condition. The voltage averaged analysis reveals a strong positivity 600 ms post
stimulus consistent with previous reaction time evidence of reanalysis triggered by
the verb. The evoked signal coherence contrasts show significant contrasts in all
three time regions consistent with Friederici's model. The existence or absence of
50 J. D. SADOY & P. BEIM GRABEN
significant effects are summarized in Table 1 below. * indicates no significant

contrast. For the current discussion topographical information is not relevant and is
omitted.
-3.00 flV .... 3.00
Numb.r mismtJleh minus COlIIroI VoII<Jg. dif!n.,,",
-0.'0 triJopy~ , .,.0.10
~ i~~
~~~
..Q,OO - Iog(p), .5.00.
Numl>er mis..... ,ela minu> ctRlrrol Pho5O 1nI1I.1tion signijicollce.
Table 1. Voltage Averaged and Signal Coherency Contrasts in the Number Condition
Post Stimulus Time Region in ms 200 400 600
Number Condition Voltage Average * * positivity
Number Condition Signal Coherence ESD ESD ESC
2.1.1 The Case Condition
Welche Frau besuchte der Richter

Which woman {sng] (Amb) visited{sng] the judge{sng] (Nom)
We turn now to the recordings to the second Determiner in the Case condition: the
verb agrees with the Case ambiguous NPI but NP2 is overtly marked Nominative.
Recall that in reaction time studies no effect was found in this condition. Similarly,
no significant differences were found in the voltage averaged records for the entire
recording epoch. The voltage observations are consistent with the reaction time
results and support the notion that there is no reanalysis or a weak reanalysis
triggered by the Case conflict. There are however, significant contrasts in signal
coherence. Figure 4 shows the early time periods.
The recordings reveal an early right and left parietal entropy increase or signal
decoherence around 220 ms.
200 IllS window
- 300 f'" _3.00
OUt mis_felt minn confnJI Volfu~ dif/.rencu
j l ip
-0. •0 ......". dIII . .o.,O
~ , OO -iog(pl, .e,oo
D.
Cas' mi."",,,,,, minus COllf1'ol PM •• fr'on$lticm s/gn/fiCtUIco.
Figure 4. Case mismatch in matrix constructions around 200ms.
This entropy drop is followed by a left anterior entropy increase and centro-parietal
and right anterior decrease around 400 ms, Figure 5. This is the time period
corresponding to integration and diagnosis in Friederici's model.
400 rns window
II ! "
- :1 .00 ',It/ .3.DO
Co.. mh"",felt minus conf1'ol VoI"'te dIff.rencu
-O.la .glopr cIIf. +0.10
•
Ca•• 1IIU""'''''' ",Iltus eOlltml£nlmPJ difl.rene..
<lI,OO -iog(P), 03,00

. :1 • a . :1 0"
C;o •• ""''''''tell nunus COfIfrOIl',""SO fr'onJlfWR "gn!/UUllctJ
In Figure 6 we see a marked right anterior entropy increase or signal decoherence

around 650 ms. This is the period associated with reanalysis in Friederici's model.
There is a second peak in decoherence around 800 ms which is not shown.
52 1. D. SADDY & P. BElM GRABEN
6OOm. ",indo",
I L1 1·
-3.00 ~v .. 3.OD
Ca •• muma/ch minus COlIIroI V"lIage dJff.nllcts
II I
-G.1D M'ircpr cIfI. +D.1D
Casl mul/ltllCh millUS ClNltroi Entropy dl/Jenncu
_. et)'
",
,
0. .~
' 1
,r
, ~
•
. . ..
•
Cou mUllll2tch minus COlI"'" 1'1UISe tl'GnslllOil sJPUlC4nCeI
The Case condition, summarized in Table 2, looks quite different in light of the
signal coherence analyses. Although we find no significant voltage averaged
contrasts, we do find significant signal coherence contrasts in all three time regions
discussed in Friederici's model.
Table 2. Voltage Averaged and Signal Coherency Contrasts in the Case Condition
Post Stimulus Time Region in ms 200 400 600
Case Condition Voltage Average * * *
Case Condition Signal Coherence ESC ESD ESD
3. DISCUSSION
The processing task being exploited here is a type of ambiguity resolution paradigm
that sets strategic processing decisions against grammatically marked information.
Our aim is to evaluate the expectation based and diagnosis and repair accounts in the
face of new observations provided by brain recordings. We noted earlier that the
only voltage averaged component found in the experiment, a P600 in the Number
condition is consistent with previous reaction time findings. The presence of a P600
demonstrates that the response to a verb that does not agree with the initial
ambiguous NP is similar to the response to an outright ungrammaticality or garden
path configuration. This tells us that the strategic assignment of the status of the
initial ambiguous NP as subject is quite strong. This is interesting as in German
object initial and subject initial WH questions have a similar frequency of
occurrence. This suggests that the strategy is not driven by frequency information
but rather by grammatical considerations as proposed in De Vicenzi (1991).
The P600 contributes little to deciding between the two approaches under
consideration. Both approaches predict a strong effect on the verb in the Number
condition. However, the two approaches differ in their hypotheses regarding the
nature of the processing involved in the resolution of Case ambiguity. The CSC
effects allow us to evaluate these hypotheses in some detail.
Unlike the voltage averaged analysis, but as might be predicted by the three
stage model, the CSC analysis finds significant effects for both conditions in all
three time windows. The signal coherence data provides clear evidence that the Case
is not ignored or discounted in the Case condition. It evokes strong cortical
responses. In fact, the pattern of responses for the two conditions is nearly
complimentary. The pattern also suggests that the processing of the Case condition
is not "simpler" than the processing of the Number condition. Thus we fail to find
strong support for either approach. This global contrast highlights the relationship
between reaction time and ERP results. The fundamental observation underlying the
processing accounts is a lack of effect in the Case condition. The ERP record is
expected to provide a more fine grained view of the reaction time effect. With the
addition of coherency measures we also gain a view of processing associated with
the "no effect" condition.
The CSC behavior is summarized in Table 3, below.
Table 3. Signal Coherency Contrast Comparison Between Number and Ccase Conditions
Post Stimulus Time in ms 200 400 600
Number Condition ESD ESD ESC
Case Condition ESC ESD ESD
In the early time period (200 ms) we find signal coherence in the Case condition but
signal decoherence in the Number condition. The ESD in the Number condition
reveals that in response to a verb that does not agree in number with the initial
ambiguous NP the system responds with much more variability in comparison with
the response to a verb that does agree. The ESC in the Case condition reveals that in
response to a Nominatively Case marked Det as opposed to an Accusatively marked
Det the system responds by restricting its phase space behavior. Neither the
diagnosis and repair nor the expectation driven approach makes any particular
claims regarding the initial stages of the Case reanalysis. The ESD found in the
Number condition could be consistent with both models. The system is driven
towards decoherence either because of a failure to meet expectation or because of
the incompatibility between number and case processes. However the ESC in the
Case condition is more compatible with diagnosis and repair. If the cue validity of
the Nominatively Case marked second DP is weak, as claimed by the expectation
based approach, then there should be a weak contrast between the Accusatively
marked Det in the control condition and the Nominatively marked Det in the Case
condition9• On the other hand, the reanalysis and repair model predicts that the Case
information is handled within the Case processing system, which is consistent with
the ESC, since the coherence of the signal suggests a concentration of processing
within a reduced phase space.
Both the Number condition and the Case condition show ESD around 400ms. In
the three stage model, voltage effects between 300 and 500 ms are taken to be
markers of semantic and thematic integration and is the time region that corresponds
to diagnosis. After the initial parse, semantic, thematic and other relational
information 10 must be mapped into a structural sketch provided by the first pass
parse. The ESD found in this time region shows that the phase space volume
increases in response to the lack of informational fit in the experimental conditions.
The 600 ms time period is when structural reanalysis takes place in the three stage
model. In this period the CSC shows different behavior for the two conditions. The
Number condition shows an ESC indicating a phase space reduction. This is
consistent with a successful revision of the initial parse and supports both the
expectation and the diagnosis and repair models as they both predict a successful
reanalysis of the properties of the initial NP in the Number condition. The ESD for
the Case condition in this time region suggests that the Case ambiguity is not
resolved. The system explores a larger phase space than in the control condition.
This is consistent with the results of comprehension tasks associated with the
reaction time studies where it was found that accuracy was significantly higher for
the Number condition than for the Case condition. This behavior supports the
expectation model but is inconsistent with the diagnosis and repair approach which
predicts that the Case ambiguity is quickly resolved.
3.1 A System Dynamics Model

The voltage averaged markers reveal contrastive brain behavior by way of voltage
polarity, amplitude and latency contrasts. As was pointed out the absence of an
averaged voltage marker does not necessarily mean that there is no contrastive effect
since weaknesses associated with the averaging technique underestimate contrastive
effects. The signal coherence contrasts offer an alternative characterization of the
time course of processing in terms of shared or divergent system properties. The
combination of the two techniques provides a basis for reconsidering the approaches
to ambiguity resolution as represented in the two models discussed here. When we
consider the large picture, we see that the diagnosis and repair model is more
consistent with the behaviour found in the early time periods while the expectation
based model is more consistent with the later time behaviour. Such a picture is not
surprising if one considers the motivating assumptions of the two approaches. The
diagnosis and repair model is at essence a modularity based approach while the
expectation model relies on a notion of saliency relativized to a given construction,
an inherently non-modular approach. Modularity asserts that at the outset cognitive
processes are encapsulated ll . We suggest that the early and late CSC effects found
in the current results can follow from the general properties of a process based view
of modularity. In the approach presented here, the interpretation of the CSC reveals
the system dynamics while the three stage model provides a functional interpretation
of the dynamic contrasts.
3.2 Accounting for the 200 ms Effects

In the early time periods of an event-related paradigm like that employed here the
stimuli either matches the ongoing processing or not. If the parsing process that is
underway at the point that the stimulus appears entails or subsumes the processing
of the stimulus then we would expect an increase in signal coherence (ESC). This
reflects the additional load on the ongoing process; a "more of the same" effect
hence a reduction in phase space volume. If the parsing process that is underway at
the point that the stimulus appears does not entail or subsume the processing of the
stimulus then we would expect a decrease in signal coherence (ESD) indicating an
increase in phase space volume. This reflects the initiation of a new process or
processes.
In the Number condition the ongoing process is the strategic assignment of
Nominative Case or subject status to the initial ambiguous DP. The processing of the
verbal information determines the predicational properties of the verb in addition to
number. While the postulation of a subject should anticipate a predicator, the
analytic properties of the verb are not directly entailed by the ongoing strategic
operation. Consequently the processing of the verbal information cannot be directly
integrated into the ongoing processing and must be processed independently. This is
reflected in the 200 ms ESD.
In the Case condition, the ongoing process is the establishment of the predicate
argument structure determined by the verb. The processing of the information
carried by the Det is directly entailed by this operation. Thus the initial processing of
the DP can be integrated as part of the ongoing process. This is reflected in the 200
msESC.

In the 400 ms time region, the lexically based information derived from the
processing of the stimulus is available for integration and information structure is
building. During this period thematic and other relational properties are determined
and aspects of sentence level interpretation, for example plausibility, become
available.
The basic nature of this stage in sentence processing predicts that the phase space
volume will expand. Information from the ongoing processes are used to build,
augment and update the interpretational parse. Integration difficulty due to
incompatible information types is expected to generate an ESD. It is also possible
that integration difficulty may arise within an ongoing process. Integration
difficulties that take place within a single process will give rise to an ESC since the
exploration and evaluation will be maintained within the ongoing process 12•
In the Number condition integration is impeded by the failure of number
agreement. In the Case condition integration is impeded by a clash between the
strategic and the morphologically marked Cases. Thus we predict a 400 ms ESD for
both conditions.
56 J. D. SADDY & P. HElM GRABEN

In the later time periods the results of the modular processing are available and
modularity effects are expected to vanish. After integration has been attempted, the
relative strength of the information types bearing on the reanalysis can be expected
to playa role, as claimed by the expectation approach. Indeed, one would not expect
such relational information be available until after integration has been attempted in
the 300 - 500 ms time period.
In the Number condition, the unambiguous number marking on the verb plus the
fact that there is no other candidate available results in a successful reanalysis. On
the basis of the available evidence no other options are considered. This is reflected
in the 600 ms ESC. In the Case condition, there continues to be conflicting evidence.
The verb agrees in number with both available arguments, both arguments are
analyzed as Nominative and Case may be a weak cue. This is reflected in the ESD 13 •
This revised account takes a functional approach to modularity and relates the
overall system behavior to the match between the stimulus item and status of the
ongoing parse. Under this view a module is a region in the phase space of the
system. CSC provides a reflection of changes in phase space volumes. By taking
account of the change in modularization of processing over the local time windows
the revised model predicts contrastive behaviour in the early time periods and
sensitivity to strength of disambiguating cues in the later time periods. A variety of
interesting predictions follow from the revision. The characterization of the later
time periods as being sensitive to relative strengths of cues predicts that the effects
in the 600 ms region are sensitive to particular types of context effects. This claim
finds some support in Gunter et aI. (1997) which finds both syntactic and semantic
influences on the P600. In addition, the revised model predicts that positional or
order effects should differentially affect the early and late components of Case
ambiguity resolution. Preliminary results from investigations of similar ambiguity
resolution in embedded constructions as well as investigations into simple positional
effects are so far consistent with this claim (Saddy & beim Graben, 1999; Saddy et
aI., 1999)
3.5 Conclusion
We have demonstrated that signal coherence measures provide a more sensitive test
of contrastive episodes in evoked brain potential recordings. However while
coherence measures can identify more contrastive behaviour, the functional
interpretation of such behaviour in terms of language processing requires the
addition of the voltage averaged markers. Coherence measures provide only the
information that the signal is more or less coherent over time. We have presented
here an interpretation of the changes in coherence motivated by system dynamic
properties revealed by the coherence calculations but this interpretation is not
specific to language processing dynamics; rather it is a characterization of changes
in phase space volume in cognitive processing terms. The CSC offers no equivalent
to the polarity and amplitude information inherent in the voltage averaged approach.
A considerable volume of research has now identified reliable correlations between
voltage averaged markers and particular aspects of language processing. For
example, we have a fairly general consensus regarding the interpretation of the
N400 as marking some form of semantic integration. The interpretation of an ESD

or ESC in the 400 ms window in response to linguistic stimuli depends upon our
understanding of the conditions under which an N400 might or might not be elicited.
The demonstration of an ESD or ESC in a 400 ms window in response to a condition
which is predicted to generate an N400 but for which no N400 is found would
contribute to both our understanding of the N400 and the nature of the linguistic
process being investigated.
While the present experimental results are equivocal with respect to
discriminating between an expectation based model and a diagnosis and repair
model, the combination of voltage averaging and signal coherence assessment does
provide clearly interpretable behaviour in all three time regions predicted by the
three stage model for both experimental conditions. The P600 for the Number
condition demonstrates that number and Case play different roles in the resolution of
Case ambiguity. The CSC associated with the early time region suggest a modular
system dynamic consistent with the diagnosis and repair model while the CSC
associated with the late time region is more consistent with the expectation model.
However, we have gained a rich view of the time course of the ambiguity resolution
process. The ESD effects in the 400 ms time region for both Number and Case
conditions offer new support for the notion that integration takes place during this
period. Furthermore, the combination of the 400 ms ESD and the 600 ms ESC in the
Number condition provides new support for the notion that the P600 is a
combination of early and late effects. The contrastive behaviour in the early, 200 ms,
time region provides evidence that both number and Case information is retrieved
early in the parse and that the language processor can respond quickly to the number
and Case information. In addition, the articulated view of the ambiguity resolution
investigated here has allowed us to propose a revised account of Case ambiguity
resolution which matches particular aspects of the expectation and diagnosis and
repair models to specific signal coherence and voltage averaged patterns in the ERP.
4. AFFILIATIONS
Dr. Douglas Saddy
Institut fUr Linguistik and Zentrums fUr Dynamik komplexer Systeme
Universitiit Potsdam
PF601553
14415 Potsdam
e-mail: saddy@ling.uni-potsdam.de
Dr. Peter beim Graben

Institut fUr Linguistik and Zentrums fUr Dynamik komplexer Systeme
PF601553
14415 Potsdam
e-mail: peter@ling.uni-potsdam.de
58 1. D. SADDY & P. BEiM GRABEN
5. NOTE
1 Nominative-Accusative ambiguities are can be found in Feminine and Neuter noun forms. There is no
Case ambiguity with Masculine nouns
2 The assumption that ERP record a deterministic signal embedded in noise describes a system that is
non-stationary by definition. If it were not, there could be no averaged voltage deflections over time
(see beim Graben, Saddy et aI., 2(00).
3 The CSC approach to ERP data is similar to some ideas of Kelso and Basar. They describe brain activity
in terms of synergetics where experimental manipulations can be linked to control parameters of a
system that undergoes phase transitions at critical parameter values (Kelso, 1999; Basar et aI., 1983).
4 Additional CSC effects were found in the 50-100 ms time window for both conditions and an 800-850
ms CSC effect was found in the Case condition. We do not discuss these contrasts here.
S No significances are plotted for the voltage averaged maps because no significant contrasts were found
other than the P600 in the number condition.
" There are three types of phase transitions with respect to the entropy measurements. A transition from
lesser to greater entropy, a transition from greater to lesser and a transition to complementary
configurations. In this latter transition the global entropy of the system does not change but the system
moves into its mirror image. This represents the most significant contrast possible.
7 We have plotted difference values. In all cases the control (non-ambiguous) conditions are subtracted
from the experimental conditions. In this case, since the anterior region is black, indicating a positive
difference value, the voltage values in the experimental conditions had to be more positive than those
recorded to the control condition.
S We find this anterior P600 standardly in response to German V2 number violations in interrogative
constructions.
9 This is under the assumption that Case marking in general is weak.
10 For example, polarity licensing, scope effects and other discontinuous dependencies, also scalar
attributes and telicity.

11 The process underlying sentence comprehension can be conceived of as non-modular at later time
periods as we see a range of information interaction at longer post-stimulus latencies.
12 The contrast is between "John drank rocks", where "rocks" violates the selectional criteria of "drink",
and .... context ... John was tall", where the context determines that "tall" is inappropriate.
13 We have treated "der" as unambiguously Nominative but this is not the case. In the paradigm presented
here "der" is only employed as a Nominative determiner. Further the other possible interpretations of
"der" (Genitive and Dative) are not compatible with the verbs employed. Nevertheless, it is possible
that the continued ESD in the Case condition is at least partly attributable to this inherent ambiguity.
This is consistent with the current analysis.
6. REFERENCES
Albano, A. M., Abraham, N. B., Guzman, G. c., Tarroja, M. F. H., Bandy, D. K., Gioggia, R. S., Rapp, P.
E., ZimJllerman, I. D., Greenbaum, N. N., & Bashore, T. R. (1986). Lasers and brains: Complex
systems with low-dimensional attractors. In G. Mayer-Kress (Ed.), Dimensions and entropies in
chaotic systems (pp. 231-240). Berlin: Springer.
Babloyantz, A., Salazar, J. M., & Nicolis, C. (1985). Evidence of chaotic dynamics of brain activity
during the sleep cycle. Physics Letters, 111A, 152-156.
Basar, E., Flohr, H., Haken, H., & Mandell, A. J. (Eds.). (1983). Synergetics of the brain. Berlin:
Springer.
beim Graben, P. (2000). Symbolische Dynamik ereigniskorrelierte Gehirnpotentiale in der
Sprachverabeitung. Unpublished Doctoral Dissertation, University of Potsdam, Germany.
beim Graben, P., Liebscher, T., & Saddy, 1. D. (2000). Parsing ambiguous context-free languages by
dynamical systems: Disambiguation and phase transitions in neural networks with evidence from
event-related brain potentials. In K. Jokinen, D. Heylen, & A. Njiholt (Eds.), Learning to behave,
Proc. CELE-Twente Workshop on Language Technology, TWLT 18 (pp.1l9-135). Universiteit
Twente.
beim Graben, P., Saddy, J. D., Schlesewsky, M., & Kurths, J. (2000). Symbolic dynamics of event-related
brain potentials. Physical Review E, 62, 5518-5533.
Bullmore, E. T., Brammer, M. J., Bourlon, P., Alarcon, G., Polkey, C. E., Elwes, R., & Binnie, C. D.
(1994). Fractal analysis of electroencephalographic signals intracerebrally recorded during 35
epileptic seizures: evaluation of a new method for synoptic visualization of ictal events.
Electroencephalography and Clinical Neurophysiology, 91, 337-345.
Coulson, S., King, J., & Kutas, M. (1998). Expect the unexpected: Event-related brain responses to
mophosyntactic violations. Language and Cognitive Processes, 13, 21-58.
De Vicenzi, M. (1991). Syntactic parsing strategies in Italian. Dordrecht: K1uwer.
Fodor, J. D., & Inoue, A. (1994). The diagnosis and cure of garden paths. Journal of Psycholinguistic
Research, 23, 407-434.
Fodor, J. D., & Inoue, A. (l999a). Attach anyway. In J. D. Fodor, & F. Ferreira (Eds.), Reanalysis in
sentence processing (pp. 101-141). Dordrecht: K1uwer.
Fodor, J. D., & Inoue, A. (1999b). Syntactic features in reanalysis: Positive and negative symptoms.
Paper presented at the 12th CUNY Conference on Human Sentence Processing, New York.
Freeman, W. J. (1990). Searching for signal and noise in the chaos of brain waves. In S. Krasnser (Ed.),
The ubiquity of chaos (pp. 47-55). American Association for the Advancement of Science.
Friederici, A. D. (1995). The time course of syntactic activation during language processing: A model
based on neuropsychological and neurophysiological data. Brain and Language, 50, 259-281.
Friederici, A. D. (1999). The Neurobiology of language processing. In A. D.Friederici (Ed.), Language
comprehension: A biological perspective (pp. 263-301). Berlin, Heidelberg: Springer.
Friederici, A. D., Hahne, A., & Mecldinger, A. (19%). Temporal structure of syntactic parsing: Early and
late event-related potential effects. Journal of Experimental Psychology: Learning, Memory and
Cognition, 22, 1219-1248.
Friederici, A. D., Mecldinger, A., Spencer, K M., Steinhauer, K, & Donchin, E. (in press). Syntactic
parsing preferences and their on-line revisions: A spatio-temporal analysis of event-related brain
potentials. Cognitive Brain Research.
Friederici, A. D., Pfeifer, E., & Hahne, A. (1993). Event-related brain potentials during natural speech
processing: Effects of semantic, morphological and syntactic violations. Cognitive Brain Research, 1,
183-192.
Frisch, S. (2000). Verb-Argument-Struktur, Kasus und thematische Interpretation beim Sprachverstehen.
MPI Series in Cognitive Neuroscience, 12.
Gallez, D., & Babloyantz, A. (1991). Predictability of human EEG: a dynamical approach. Biological
Cybernetics, 64, 381-391.
Gunter,T. C., Stowe, L. A., & Mulder, G. (1997). When syntax meets semantics. Psychophysiology, 34,
660-676.
Hagoort, P., Brown,C., & Groothusen, J. (1993). The syntactic positive shift (SPS) as an ERP measure of
syntactic processing. Language and Cognitive Processes, 8, 439-483.
Hahne, A. (1998). Charakteristika syntaktischer und semantischer Prozesse bei der auditiven
Sprachverarbeitung. MPI Series in Cognitive Neuroscience, 1.
Hahne, A., & Friederici, A. D. (1999). Electrophysiological evidence for two steps in syntactic analysis:
Early automatic and late controlled processes. Journal of Cognitive Neuroscience, 11, 194-205.
Hopf, J., Bayer, J., Bader, M., & Meng, M. (1998). Event-related brain potentials and case information in
syntactic ambiguities. Journal of Cognitive Neuroscience, 10, 264-280.
Kelso, J. A. S. (1999). Dynamic patterns. The self-organization of brain and behaviour. Cambridge, MA:
MIT Press.
Kutas, M., & Hillyard, S. A. (198Oa). Reading senseless sentences: Brain potentials reflect semantic
incongruity. Science, 207, 203-205.
Kutas, M., & Hillyard, S. A. (1980b). Event-related potentials to semantically inappropriate surprisingly
large words. Biological Psychology, 11, 99-116.
Kutas, M., & Hillyard, S. A. (1983). Event-related potentials to grammatical errors and semantic
anomalies. Memory and Cognition, 11, 539-550.
Layne, S. P., Mayer-Kress, G., & Holzfuss, J. (1986). Problems associated with dimensional analysis of
electroencephalogram data. In G. Mayer-Kress (Ed.), Dimensions and entropies in chaotic systems
(pp. 246-256). Berlin: Springer.
Lutzenberger, W., Elbert, T., Birbaumer, N., Ray, W. J., & Schupp, H. (1992). The scalp distribution of
the fractal dimension of the EEG and its variation with mental tasks. Brain Topography, 5, 27-33.
Meng, M. (1998). Grammatik und Sprachverarbeitung: Psycholinguistische Untersuchungen zur
Berechnung syntaktischer Strukturen. Unpublished doctoral dissertation. University of Jena. Jena.
Germany.
60 1. D. 5ADDY & P. BEIM GRABEN
Meng, M., & Bader, M. (1996). Case and agreement in parsing subject-object ambiguities. Architectures
and Mechanisms for Language Processing (AMLaP), Turin.
Molnar, M, Skinner, 1. E., Csepe, V., Winkler, I., & Karmos, G. (1995). Correlation dimension changes
accompanying the occurrence of the mismatch negativity and the P3 event-related potential
component. Electroencephalography and Clinical Neurophysiology, 95, 118-126.
Neville, H. 1., Nicol, J., Barss, A, Forster, K., & Garrett, M. F. (1991). Syntactically based sentence
processing classes: Evidence from event-related potentials. Journal of Cognitive Neuroscience, 6,
233-244.
Osterhout, L. (1997). On the brain responses to syntactic anomalies: Manipulations of word position and
word class reveal individual differences. Brain and Language, 59, 494-552.
Osterhout, L., Mclaughlin, J., & Bersick, M. (1997). Event-related brain potentials and human language.
Trends in Cognitive Sciences, I, 203-209.
Osterhout, L., & Mobley, L. A (1995). Event-related brain potentials elicited by failure to agree. Journal
of Memory and Language, 34, 739-773.
Pritchard, W. S., & Duke, D. W. (1992). Dimensional analysis of no-task human EEG using the
Gassberger-Procaccia method. Psychophysiology, 29, 182-191.
Reyni, A (1970). Probability theory. Amsterdam: North-Holland.
Saddy, J. D., & beim Graben, P. (1999). The dynamical properties of category recognition. Paper
presented to European Neurological Society, Arcachon, France.
Saddy, J.D., beim Graben, P., & Schlesewsky, M. (1999). Cortical dynamics of language processes.
Paper presented at the European Conference on Cognitive Science 1999. Instituto di Psicologia
Consiglio delle Ricerche.
Schlesewsky, M. (1997). Kasusphiinomene in der Sprachverarbeitung. Dissertation, Potsdam University,
Germany.
Schlesewsky, M., Fanselow, G., Saddy, D., & beim Graben, P. (1998). Some arguments against the
existence of "Invisible but theoretically predicted reanalysis effects. Poster presented at the CUNY
Conference 98. Rutgers.
Sekerina, I. (in press). The Late closure principle: Evidence from on-line procesing of ambiguous russian
sentences. In Proceedings of the 2nd European Conference on Formal Description of Slavic
Languages.
Shannon, C. E., & Weaver, W. (1949, reprint 1963). The mathematical theory of communication. Urbana:
University of Illinois Press.
P. F. DOMINEY
A MODEL OF LEARNING SYNTACTIC

COMPREHENSION FOR NATURAL AND ARTIFICIAL
GRAMMARS
Abstract. One important function of syntactic analysis of speech is the assignment of thematic roles to
noun phrases. Thematic roles of noun-phrases in canonical sentences are assigned in "default" orderings
(e.g. Agent Object Recipient in English for the canonical sentence "John gave the ball to Mary."). This
ordering is transformed in non-canonical sentences (e.g. "The ball was given to Mary by John."), whose
thematic role assignment is guided, in part, by function items (e.g. prepositions "to" and "by").
Agrammatic patients are impaired in the syntactic comprehension of non-canonical sentences. It is not clear
whether this is due to a failure to process function items, a failure to assign thematic roles, or both.
We have recently studied artificial grammar learning in a recurrent network model in which serial surface
structure, and abstract transformational rules are represented by separate systems. We now examine the
behavior of this dual system model in syntactic comprehension of canonical and non-canonical sentences.
Function items are represented by the "surface" system and guide the application of transformational rules
that are represented by the "abstract" system. After learning the structural regularities of the target
language, the model predicts that failure in either of these systems, i.e. in the representation of function
items or in the representation of syntactic knowledge will impair syntactic comprehension. The model also
predicts that the expression of these impairments should not be restricted to performance with natural
grammars but should also be manifest in artificial grammar conditions. With respect to this second
prediction, we have recently confirmed that agrammatic patients indeed fail to process non-canonically
ordered sentences both in natural and artificial grammars. In the artificial grammar sequences, this failure is
purely related to application of the transformation, as no function items are involved in the task. The
relevance of these results to current linguistic theory will be discussed.
1. INTRODUCTION
Thematic role assignment is a central function of syntactic analysis in sentence

comprehension. The research objective in this chapter is to attempt to outline how this
capability could be implemented in a neurophysiologically feasible neural network
model that learns the structural regularities of the target language. We have recently
developed a neural network model that simulates the 0-8 month infant's sensitivity to
serial (Saffran, Aslin, & Newport, 1996), temporal/rhythmic (Nazzi, Bertoncini, &
Mehler, 1998) and abstract structure (Marcus, Vijayan, Bandi Rao, & Vishton, 1999)
of language (Dominey & Ramus, 2000). A key aspect of this model is that the infant's
sensitivity to serial and temporal structure in language can be simulated by a
"temporal recurrent network" (TRN) that relies on recurrent connections to encode
context, and an associative memory that binds context representations to behavioral
responses, while sensitivity to abstract rule-like structure requires the additional
representation capabilities.
E. Witruk, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading, and
Reading Disability, 61-77.
62 P. F. DOMINEY
Such additional representational capabilities were initially developed to simulate

human performance in the learning and transfer of abstract structure in sensorimotor
sequence learning (Dominey, Lelekov, Ventre-Dominey, & Jeannerod, 1998). The
two sequences ABCBAC and DEFEDF can be said to have different surface structures
(Le. serial orders), since they are made up of different elements. The share, however,
the same abstract structure 123213. Humans can learn this kind of abstract structure
and demonstrate transfer of this knowledge to new sequence that are constructed from
the same abstract structure (isomorphic sequences) e.g. RZPZRP, HDIDHI etc.
We demonstrated that while our TRN model can learn surface structure, it fails to
learn abstract structure because it is incapable of representing the structural relations
between repeating elements. In order to do so, we were required to add a continuously
updated short term memory of the 7+/- 2 previous elements. Each new element in a
sequence is compared with the contents of the STM, thus allowing detection and
representation of abstract structure. For example, the abstract structure of sequence
ABCBAC is represented as u,u,u,n-2,n-4,n-3, where "u" indicates "unpredictable, or
no repetition", and "n-2" indicates "repetition of the element 2 places behind," etc.
The thus updated model for abstract structure was able to simulate adult performance
in serial reaction time tasks that required abstract structure learning (Dominey et al.,
1998). Interestingly, the same model also simulates babies' sensitivity to the
difference between auditory sentences based on the abstract structures AAB and ARB,
as demonstrated by Marcus et al. (1999).
The current research tests the hypothesis that by allowing the models for surface
and abstract structure to interact, we can develop a capability to simulate adult
performance in thematic role assignment with dissociable processing of closed class
items by the surface model, and open class items by the abstract model, as described
below.
2. THEMATIC ROLE ASSIGNMENT
A central function in syntactic analysis or syntactic comprehension is the assignment

of thematic roles to noun phrases in sentences. In a simplified manner, we can
consider that in languages like French and English, there is a default or canonical
order in which thematic roles are assigned (e.g. "Agent Object Recipient" in English
for the canonical sentence "John gave the ball to Mary.") However, in non-canonical
sentences (e.g. "The ball was given to Mary by John."), this ordering is transformed,
and thematic role assignment is guided, in part, by function items (e.g. closed class
words including prepositions "to" and "by", grammatical morphemes, etc.).
The ability to assign thematic roles has been quantified in different clinical
tests used to asses agrammatism in aphasic patients suffering from lesions of the left
cortical hemisphere. A well know version developed by Caplan, Baker, and Dehaut
(1985) consists of 9 sentence types of varying syntactic complexity, five canonical
and four non-canonical, and will serve as our target problem (see Table 1). Five
sentences of each type are used for a total of 45 sentences. Sentences are real aloud to
the patients in a pseudo-random order, and after each sentence, the subject should
indicate by pointing at photographs "who did what to whom", indicating in canonical
order the agent, object and recipient. Interestingly, a rather significant subgroup of
LEARNING SYNTACTIC COMPREHENSION 63
these patients with left-hemisphere lesions demonstrate a deficit in thematic role

assignment that is highly selective for non-canonical sentences. We note that the
sentences are constructed so that no semantic interpretation can contribute to the role
assignment, which must proceed entirely as guided by syntactic function items.
Table 1. Nine Sentence Types as Specified in Caplan et al. (1985). Noncanonical Word
Ordering Indicated by *.
Active (A) The elephant hit the monkey.
Passive (P) The elephant was hit by the

Two place verb monkey. (*)
sentences Cleft-Subject (CS) It was the elephant that hit the
monkey
Cleft-Object (CO) It was the elephant that the monkey
hit. (*)
Dative (D) The elephant gave the monkey to
Three-place verb the rabbit.
sentences Dative pass (DP) The elephant was given to the
monkey by the rabbit. (*)
Conjoined (C) The elephant hit the monkey and
hugged the rabbit.
Sentences with two Subj-Obj ReI (SO) The elephant that the monkey hit
verbs hugged the rabbit. (*)
Obj-Subj ReI (OS) The elephant hit the monkey that
hugged the rabbit.
In order to realize such a task, a system should first be capable of distinguishing

function words (or morphemes) from content words. Numerous behavioral and event-
related brain potential studies indicate that indeed, adults process function and content
words in a dissociated manner. The system must also be able to store the content
words in a working memory, and then to access this memory in a non-standard order
(i.e. different from the input order) guided by the function items. This capability to
access the content words in the canonical order, guided by the function words,
provides the basis for the assignment of thematic roles.
The dual-process model that we developed to simulate human performance in
learning surface and abstract structure appears ideally suited for this task. Function
items are represented by the "surface" system and guide the application of
transformational rules that are represented by the "abstract" system. We can
demonstrate that after training on a supervised version of the Caplan task in which the
correct thematic role assignment is provided, the model can then perform the standard
unsupervised task correctly, including the generalization to new sentences. The model
can be rendered agrammatic either by disruption of the representation of the function
items, or by disruption of the selection of nouns from the working memory based on
the representation of the function words. Thus, the model predicts that failure in either
64 P. F. DOMINEY
of these systems, i.e. in the representation of function items or in the representation of

syntactic knowledge will impair syntactic comprehension.
The elephant was transferred

to the monkey by the dog.
Open Class
Stream (Content):
Stream (Function): elephant, dog,
the, was, to, by monkey
Canonically ordered nouns:

agent(dog), object(clephant), rccipient(monkey)
Figure 1. Syntactic comprehension model architecture.
3. OVERVIEW OF MODELING APPROACH
The model architecture is presented in Figure 1, and technical details are presented in
the Appendix. In Figure 1, each of the labeled elements corresponds to a 5x5 array of
leaky integrator neurons. A central premise of the model is that open and closed class
items are processed by dissociable streams as suggested by studies of
psycholinguistics, aphasia and language related ERPs (e.g. Friederici, 1985; Neville et
al. 1993; Osterhout, 1997; Pulvermiiller, 1995; Brown, Hagoort, & ter Keurs, 1999).
This is reflected in the dual processing streams for closed and open class words. As
the input sentence is read in, the recurrent network State encodes the ordered history
of closed class words, and the short term memory for transformation (STMt) stored
the open class words in the order of their arrival. Based on the syntactic context
defined by the sequence of closed class words, State units activate Modulation units to
access the contents of the STMt to retrieve the open class elements in their canonical
order.
The model is constructed from ensembles of neuron-like units, each of which
simulates the membrane potential of a neuron (or population of neurons) and
generates an analog signal of firing/discharge rate. The model acquires the capacity to
perform the syntactic comprehension task by learning that results in the modification
LEARNING SYNTACfIC COMPREHENSION 65
of connection strengths between different neurons. The desired behavior is to

reproduce human-like performance in Caplan's syntactic comprehension task in
which sentences are presented, one word at a time to the model which then responds
by emitting the nouns in their canonical order when presented with the remaining
unassigned nouns.
The 9 sentence types from Caplan et a1.'s (1985) syntactic comprehension task are
re-coded so that each word corresponds to one element in the 25 element input array
of the model. The words are categorized into closed and open classes. Closed class
words enter a processing stream directed towards the recurrent network (State) which
thus maintains an ongoirig context of the sequence of closed class words presented for
the sentence currently being processed. The open class words enter a processing
stream by which they are stored in a working or short term memory. This continues
until the final word of the current sentence is processed.
3.1 Training
During the training phase, a given sentence is presented, one word at a time in a
temporal sequence as just described. Once the sentence is presented, the model is then
provided with the open class nouns, one at a time, in their canonical order. In the
Caplan protocol, this is equivalent to the experimenter herself pointing to the nouns in
their canonical order after reading the sentence, i.e. showing the subject the correct
responses. For canonical input sentences, the open class words are thus simply re-
presented in the same order as in the initial input. For non-canonical input sentences,
the nouns are transformed into their canonical order. During this learning phase, the
model learns to associate the different patterns of closed class words (encoded in the
neural activity of the State layer), with the appropriate (re)ordering of open class
words.
For an example sentence, "The elephant was transferred to the monkey by the
dog," after the sentence is presented, the model is then provided with the first
canonically ordered thematic role, i.e. the agent which in this case is "dog." This
matches with the contents of the 3rd element of the STM. A learning signal strengthens
connections between the current set of active neurons in State (whose activity is due
to the sequence of function words that was provided) and a modulatory neuron that
will modulate that contents of the third element of the STM into the output. The result
of such learning is that when the same type of sentence is presented, the active State
units will activate the appropriate Modulation unit, directing the contents of the 3rd
STM element to the output, thus providing the agent response.
As the supervised training example proceeds, the next canonically ordered
element, the object "elephant" is provided. This matches with the first element of the
STM, and the new State activity (which has been modified due to the match
recognition) becomes associated with the modulation of the contents of the 1"1 STM
element into the output, and so on for the recipient. The result of this training, for all 9
sentence types, is that activity in the closed class stream (encoded in the evolving
activity of State during sentence presentation) allows the selection of elements from
the STM in the appropriate learned canonical order. While clearly oversimplified
from a linguistic perspective, the resulting representations are sufficient to allow the
66 P. F. DOMINEY
model to successfully perform the syntactic comprehension task. This point will be
further addressed in the discussion.
3.2 Testing and Correction
After this type of training, we then proceed to a testing period. In the testing period,
instead of providing the correct response for agent, object, and recipient, we force the
model to make a choice between the remaining nouns that have not yet been assigned.
For the sample sentence "The elephant was transferred to the monkey by the dog",
after the sentence is presented, we then present "elephant, monkey, dog"
simultaneously, and the model must choose, based on knowledge acquired in the
training described above, the noun among these three that corresponds to the agent, or
"dog". Correct responses indicate that learning has been effective, and no connection
strengths are modified. Errors result in a weakening of the connections from State to
Modulation, thus reducing the probability of the erroneous response being repeated.
Details of the simulation of processing this sentence are illustrated in Figure 2.
4. SIMULATION METHODS
A popUlation of model "subjects" was created by using the model architecture

described in Fig. 1, and starting with different seed values of a random number
generator used to assign synaptic weights to the recurrent connections. This
population underwent a training procedure where each of the first 6 sentence types
(two-argument and three-argument verb sentences) was presented approximately 10
times, in the training mode described above. That is, after the sentence was "read" to
the subject, the subject was then presented with the noun phrases, one after another, in
their canonical order. After this training phase, subjects were exposed to a testing
period. During this period, after a sentence was read, the nouns that had not yet been
assigned a thematic role were presented simultaneously and the model was forced to
choose the correct one, with error-driven learning. During this testing period, each of
the six sentence types was presented approximately 75 times.
Model subjects that completed this training and demonstrated no errors at the end
of the training were then exposed to all nine sentence types in the same testing with
error-driven learning, with approximately 60 exposures for each of the nine sentence
types. This pre training with a subset of the 9 sentence types follows the spirit of
Elman (1993) in "starting small," and allows us to isolate potential good learners.
The model subjects that successfully completed this training, i.e. those that were
able to process all nine sentence types with no errors, were then used to generate a
population of aphasic subjects by the addition of processing noise into the recurrent
state system, or the closed class stream.
Figure 2 illustrates the activity of model neurons during a sentence comprehension
trial. Reading from left to right, the activation of these units reflects the presentation
of the sentence to the model. As soon as closed class words are presented, activity in
State (not shown) begins to drive the Modulation neurons, based on the learning-
modified State-Modulation synapses. When elephant, monkey and dog are all
activated, this is the "go signal" for the model to respond with the agent. At this point,
the Modulation neuron Mod3 that directs the contents of STM3 to Output is more
active than ModI or Mod2 (813, 789, 380 respectively), so the contents of STM3 (or
"dog") is directed towards the output to specify the agent, as seen in row N. The
remaining nouns are then presented and the model should choose the object.
..
-I I-
--_...-
~
A f"1 11
B r-1
C
o .. 11
" ,.,
-
or
E
-..,
F ,---, ~
G 11 I\....rLI"'\...
H
I
_.
I .... •
-"'" -r vv
- A
,---,
'-L
J
1'-
-_. ~
- r--L
K
L
!-
- ---- --- ""L.
lL.-
L
M
1--......-
1\
N
Figure 2. Simulation example - comprehension of Dative Passive (noncanonical) sentence

"The elephant was transferred to the monkey by the dog. " Each row represents that activation
of a given unit (neuron) in the model over time. Rows A -E correspond to the closed-class words
(the, was, transfer-ed, to, by), and rows E-G correspond to the open class words (elephant,
monkey, dog). Rows I-K represent the activation of the three Modulation units that direct the
contents of the first three STM elements, respectively, towards the output. Their activation is
derived from inputs from State based on learning. Rows L-N represent the model's output
responses corresponding to the three open-class units for elephant, monkey and dog. In the
simulation they should be activated in the order agent, object, recipient for three-argument
sentences, or "dog," elephant," "monkey" in this case. See text.
The activity of the three Modulation units changes, and now ModI is more active than
Mod2 or Mod3 (1677, 1369 and 705 respectively). The contents of STMI "elephant"
is thus directed to Output to specify the object. Finally, the remaining unassigned
noun is presented and chosen to specify the recipient. Note that as required, Mod2 is
more active than ModI and Mod3 (1449; 1172 and 705 respectively). Thus, via the
activation of appropriate Modulation neurons by State, the 3 nouns are re-ordered in
their canonical order, successfully analyzing the target sentence.
68 P. F. DOMINEY
5. RESULTS AND COMPARISON WITH HUMAN AGRAMMATIC

PERFORMANCE
5.1 "Normal" Model Subjects
A set of 20 model subjects were isolated that were capable, after the training and
testing described above, of processing all 6 two- and three-place verb sentences with
no errors. Of these 20, 4 went on to successful completion of the 9 sentence types, and
define the population of subjects studied in the noise-based lesions. Of these 4, 2
made the transition from 6 to 9 sentences with no additional training. That is, the
capability to solve the 6 sentence configuration transferred directly to the nine
sentence configuration. The remaining 2 subjects went on to master the 9 sentence
version with no more than 40 exposures to each sentence type.
It should be clearly stated that for this population that successfully mastered the
nine sentence types, the learning produced a generalization capability that perfectly
transfers to new sentences that follow one of the nine forms but use new nouns, or old
nouns in a different order. That several distinct initial states of the model architecture
all converged on perfect performance of the nine sentence-type task indicates that the
solution is a replicable and robust property of the model architecture.
5.2 Baseline of Human Agrammatic Performance
Before examining the simulated agrammatic performance, we first establish a baseline

for human agrammatic performance. Figure 3 illustrates the averaged performance for
9 agrammatic patients (Dominey & Lelekov, 2000; Lelekov, Franck, Dominey, &
Georgieff, 2000) on the nine sentence types. This performance is quite comparable to
that described by Caplan et al. (1985) and will provide a baseline for evaluation of the
model's capability to simulate agrammatic performance. It is clear in Figure 3 that for
each of the three subgroups of sentences, two-argument, three argument and two verb,
that the patients process the canonical forms reasonably well, but demonstrate a much
higher error rate for the corresponding non-canonical forms.
These observations were confirmed by repeated measures ANOVA comparing
error rates for canonical and noncanonical sentences of the different argument types.
Differences in errors for canonical vs. non-canonical two-argument verb sentences
were significant (F(1,8) = 40.5, P < 0.0005), the three-argument verb sentences
= =
(F(1,8) 33.6, p < 0.005), and the two verb sentences (F(2, 16) 18.4, p < 0.0005).
5.3 "Lesioned" Model Subjects
Given this human performance, we now consider the simulation of agrammatic

performance. Ten different levels of noise were introduced into the State network for
each of the four models, thus yielding a population of 40 lesions. Figure 4 illustrates
the performance of the model subjects when processing noise is injected into the
closed class stream. Noise is injected into the closed class stream, to perturb the
representation in State, in order to simulate the degraded mode of operation in lesion
conditions. The regularity that noncanonical forms are more impaired than canonical
forms is preserved. An increase in the three-argument, and two verb sentence error
rates is observed, however, as well as an increase for all forms, including the active, in
variance with the observations of agrammatic performance in Figure 2.
Human Agrammatism
4 tF==.;;f:.;.:====""'···I·c.............
o
-I
-2~~--~----~--~----~--~----~--~----~~
Active Passive Cleft- Cleft- Dative Dative- Conjoin Sub-Obj Obj-Subj
subject Object Passive Relative
Sentence Type
Figure 3. Performance of9 agrammatic subjects on the 9 sentence-type syntactic

comprehension task of Caplan et al. 1985 (from Lelekov et. al. 2000 and Dominey & Lelekov
2000). For each of the three sentence types, patients perform worse on noncanonicalforms.
They also display an increase in errors with increasing complexity in terms of number of
arguments and number of verbs.
Syntactic Comprehenesion Model with Noise Lesion

6,5
6,0
5,5
5,0
~ 4,5
~ 4,0
3,5
3,0
2,5
Active Passive Cleft- Cleft- Dative Dative- Conjoin Sub-Obj Obj-Subj
subject Object Passive Relative
Sentence Type
Figure 4. Syntactic comprehension model with noise lesion.
Note that for each of the three sentence categories (two-, three-argument and two
verb sentences) that the simulated agrammatism adheres to the pattern that
noncanonical forms are more severely effected than canonical forms. A major
70 P. F. DOMINEY
discrepancy with the human data is seen, however, in the overall increase in error rate
for the three-place and the two-verb sentences with respect to the two-place verbs.
Even worse, we note that errors are being made now even for the active, two-place
verb sentences.
This helps us to understand this more pronounced deficit for the three-place verb, and
two verb sentences. Likewise, the canonical ordering occurs with greater frequency
than the different noncanonical orderings, and thus is learned in a more robust manner
that makes it less sensitive to the degraded processing conditions induced by noise.
Similarly, in the intact system, there is a form of competition between canonical
and non canonical forms, since either may be expected to occur in the input. In the
noise lesion case, the ability to use the function words in order to correctly decide on a
canonical or noncanonical interpretation is perturbed, yielding the situation where
noncanonical forms are more severely affected, but canonical forms are affected as
well. In other words, in this degraded mode of operation, the canonicallnoncanonical
distinction - as revealed by closed class cues - blurs, and both suffer. A potentially
useful adaptation to this situation might be to bias all interpretation towards the
canonical forms. This would theoretically reduce errors for canonical forms, while
increasing errors for non canonical forms (which are already high). In the model, this
adaptation can be realized by simply adding a bias to the activity of the modulation
unit that directs the contents of the first STM element into the output: In other words,
by adding a bias towards selecting the first appearing noun as the agent.
Lesioned Model with Adaptation
Active Passive Cleft· Cleft· Dative Dative- Conjoin Sub-Obj Obj·Subj

Subject Object Passive Relative
Sentence Type
Figure 5. Syntactic comprehension model with noise lesion and adaptation. The
Modulation neuron that corresponds to STMl is given a constant positive bias, effectively
increasing the probability that the first noun will be assigned the agent role. This yields
improved performance on the canonical forms, with a performance profile that approaches that
of agrammatic patients as seen in Figure 3.
Figure 5 illustrates the lesioned population performance with this bias in place. We
now see a profile that is much more similar to that of human performance in Figure 2.
In particular, for each of the three sentence types, the canonical error rate is
significantly lower than noncanonicai. These observations were confirmed by
repeated measures ANOVA comparing error rates for canonical and noncanonical
sentences of the different argument types. Differences in errors for canonical vs. non-
canonical two-argument verb sentences were significant (F(1,39) = 215, P < 0.0001),
the three-argument verb sentences (F(1,39) = 136, p < 0.0001), and the two verb
sentences (F(2, 78) = 92, P < 0.0001). This performance is highly correlated with the
performance of our agrammatic patients displayed in Figure 3 (patient = -.71 + .95 *
model; r2 = 0.89).
6. PREDICTION FOR ARTIFICIAL GRAMMARS
Interestingly, for the proposed model, the transformation processing described above
is not necessarily restricted to that in syntax of natural languages, and could also be
expressed in artificial grammar tasks (Dominey et aI., 1998). The model thus predicts
that the syntactic comprehension deficit seen in agrammatic aphasia reflects an
underlying impairment in processing transformations of abstract sequential structure
that is not specific to natural language processing.
To test this hypothesis we studied the ability of agrammatic aphasic subjects to
process surface and abstract sequential structure in non-linguistic sequences. The two
sequences ABCBAC and DEFEDF have different surface structures, but share the
abstract structure 123213. In this abstract structure the second triplet (213) is a
transformation of the first (123). Thus 123213 is considered a complex abstract
structure, whereas 123123 is a simple abstract structure since no transformation is
involved between the first and second triplet.
Seven aphasic subjects were required to learn simple and complex abstract
structures and then to classify letter strings as corresponding, or not, to the learned
target abstract structures. Subjects' performance in this task was compared to their
performance for simple and complex syntactic comprehension. Performance scores
for linguistic and non-linguistic impairments are significantly correlated (r2= 0.86, p =
0.003). Looking more closely at canonical vs. non-canonical processing for linguistic
comprehension and letter-sequence classification tasks, agrammatic patients correctly
process canonical but not non-canonical structure as depicted in Figure 6.
These results suggests that the transformation processing capabilities required for
the abstract structure task are the same as those used in syntactic comprehension. This
is consistent with the hypothesis that syntactic comprehension deficits reflect an
impairment in performing serial order transformations on abstract structure that is not
restricted to natural language.
7. DISCUSSION
These results provide new insight into the neural computation underlying thematic
role assignment from two complementary perspectives. First, from the perspective of
intact processing, they demonstrate the feasibility of an approach by which closed and
open class items are processed in separate streams, with open class items entering into
a random-access working, short term, memory. This memory is accessed in a specific
72 P. F. DOMINEY
order, that may be transformed with respect to the initial input order, based on cues
provided by closed class items in the separate closed class stream.
A. Non-Linguistic B. Linguistic
1.3
1.1
0.9
~
~
C1.l
u
c: 0.7
Canonical
~
C':l
§ 0.5
~
Canonical
~C1.l 0.3
:l.,.
0.1
Non- Non-
-0.1 Canonical
Canonical
-0.3
Figure 6. Selective impairment in aphasics for non-canonical structure in linguistic and

non-linguistic tasks. Mean performance values as percentage correct * 100, with standard
error (boxes), and standard deviation (whiskers). A. For the non-linguistic abstract structure
classification task, patients performed significantly better for canonical (92%) vs. non-
canonical (45%). B. Likewise, for the linguistic syntactic comprehension task, patients
performed significantly better for canonical (74%) vs. non-canonical (34%) task.
From the related perspective of degraded operation due to lesion (or noise), these
results provide one explanation for the more selective perturbation of processing of
noncanonical forms. In the 9 sentence types, 5 share the same canonical structure
while the remaining 4 use a set of different non-canonical forms. These forms are
more complex, involving transformations, and they are also less frequent, and are thus
less represented in the learning. These frequency and complexity effects, that were not
revealed in the intact processing conditions, became apparent in the lesioned
condition. The fact canonical processing was also effected, though to a lesser extent
(Figure 4), suggests the possibility that in the post-lesion human condition, a more
general breakdown may occur, followed by the adoption of a compensatory bias
towards canonical processing, as suggested by our simulations (Figure 5).
With respect to the development of syntactic analysis capabilities, it is of great
interest that the model presented here is essentially identical (with a few parameter
changes) to a model that simulated human infant's sensitivity serial and abstract
structure of language. Thus, by training on appropriate material, an infant model is
made to possess adult processing capabilities. And indeed, the training was rather
limited, with less than 200 exposures to the different sentence types. This provides an
interesting data point in the ongoing debate concerning the nature of the "language
organ," and parameter setting. Clearly these are quite preliminary observations on this
point, and will require much more investigation including crosslinguistic simulation
studies before stronger conclusions can be drawn.
One area has been explored in this direction, however. The model presented here
is not exclusively linguistic, and the surface and abstract structures that it can process
can indeed be non-linguistic. This would predict that agrammatism would have a non-
linguistic expression, as we have recently demonstrated, with a correlation in
agrammatic patients between syntactic comprehension errors and errors in processing
abstract sequential structure as defined above (Lelekov et aI., 2000, Dominey &
Lelekov, 2000). Interestingly, the observed impairment in non-canonical abstract
structure processing occurred despite the absence of a requirement to process closed
class function items. This suggests that the agrammatic deficit in processing non-
canonical order may not be exclusively due to impaired processing of function items,
and may rather be due at least in part to an impairment in the implementation of the
required transformation.
An important criticism of the current description of the model's performance (as
pointed out by David Caplan - personal communication) is that the processing
appears to take place after the sentence has been processed, during the period where
the model should provide the nouns in their canonical order, while it is clear that in
humans, sentence processing is taking place on-line, as in the model of Haarman, Just,
and Carpenter (1997). Interestingly, part of the response to this criticism lies in the
specification of the task the model was to learn, and our approach to analyzing the
performance. In the current task, the entire sentence is read in, and then the model is
to produce the nouns in their canonical order - just as in the syntactic comprehension
task defined by Caplan et al. (1985). Alternatively, we could propose a task similar to
a self paced reading task in which the model would respond to each word as it was
presented, similar to a serial reaction time task. We know from previous studies that
for trained sequences, the model responds with reduced reaction times, and that
violations of predictable regularities in sentence structure should thus yield increased
reaction times for the offending words in the model (Dominey et aI., 1998; Dominey,
& Ramus, 2000). Thus, dependant on the training and the performance measure, we
should be capable of observing on-line processing effects.
An equally important criticism concerns the linguistic realism of the model, and
particularly the role of the verb in sentence processing. A minimal treatment of the
verb should include the identification of its category in order to establish the required
number of arguments and the argument structure. In the current implementation of the
model, this information is in fact provided by the external experimental environment.
Specifically, during the response phase, the model is interrogated until all thematic
roles have been assigned, thus the number of arguments is implicitly available. This
property derives directly from the syntactic comprehension protocol employed
(Caplan et al. 1985). Thus while a principle future priority is to treat the verb in a
more realistic fashion, the model successfully simulates human performance in this
protocol.
In conclusion, the current observations allow us to propose a neural network
model of infant linguistic capability that, given appropriate training, simulates the
combination of simple functions that allow the realization of a reduced form of
syntactic analysis. The model explains data on normal and aphasic performance, and
also predicts performance of infants and aphasics in a novel artificial grammar
learning task. Likewise, the predictions made by the model are not unique to auditory
processing, and in fact are applicable to written material as well. While clearly a long
way from a complete model of syntactic analysis, this simplified model provides a
74 P. F. DOMINEY
point of departure for discussions of how syntactic analysis might and might not be
implemented.
8. APPENDIX: SYNTACTIC COMPREHENSION MODEL SPECIFICATION
The model is implemented in Neural simulation Language (http://www-hpb.usc.edu).

Words are presented in the 25 element (5x5) Input array. Elements 0-12 are reserved
for closed class words (one per element) and are directed to the closed class stream.
Elements 13-24 are reserved for open class words and are directed towards the open
class stream.
8.1 Recurrent State Representation for the Closed Class Stream
Equations 1.1 and 1.2 describe how State is influenced by closed class words from
Input, recurrent inputs from StateD, and recognition of the transformed open class
words from Recognition. In (1.1) the leaky integrator, sO, corresponding to the
membrane potential or internal activation of State is described. In (1.2) the output
activity level of State is generated as a sigmoid function, fO, of s(t). The term t is the
time, ,M is the simulation time step, 't is the time constant. As 't increases with respect
to ,M, the charge and discharge times for the leaky integrator increase. The At is 5
milliseconds. For Equations 1 - 2, the time constants are 10 milliseconds, except for
Equation 2.1 which has 5 time constants that are 100, 600, 1100, 1600 and 2100
milliseconds.
Sj (t + tlt) Sj (t) + -;-flt(""\' w~s Input (t) + "\'" wtSStateD (t) + "\' w:s Re cog (t)1(1.1)
tlt) =(1--;- j j n
f:1 f:1 f:1 j
State(t) = f(s(t)) (1.2)
The connections wIS , wSS and wRS define the projections from units in Input,
StateD, and Recognition to State. These connections are one-to-all, and are mixed
excitatory and inhibitory, and do not change with learning. This mix of excitatory and
inhibitory connections ensures that the State network does not become saturated by
excitatory inputs, and also provides a source of diversity in coding the conjunctions
and disjunctions of input, output and previous state information. Recurrent input to
State originates from the layer StateD. StateD (Equation 2.1 and 2.2) receives input
from State, and its 25 leaky integrator neurons have a distribution of time constants
from 100 to 2100 ms (20 to 420 simulation time steps), while State units have time
constants of 10 ms (2 simulation time steps).
sdj(t + flt)=(1- ~)Sdj(t)+ ~ (Statej(t)) (2.1)
StateD = f(sd(t)) (2.2)

This distribution of time constants in StateD yields a range of temporal sensitivity

similar to that provided by using a distribution of temporal delays. Such recurrent
architectures, typified by the SRN (Elman, 1990, 1993) have been demonstrated to
have significant sequence learning capabilities. The recurrent State network thus
maintains a representation of the context of closed class words that will be used to
reorder the open class words in their canonical order.
8.2 Open Class Stream
In order to represent and execute serial order transformations a system must (1) store
the elements to be transformed in an accessible manner, and (2) provide a reliable
method to access these elements in an order that may vary from the initial input order.
The model of Figure 1 realizes the first requirement with a continuously updated short
term memory (STM) of the open class elements (nouns). Each time an open class
element (noun) enters the open class stream, the STM is updated, as described in
Equation 3 so that STM always contains the open class elements of the current
sentence in the order they were presented. Each of the 5 STM elements is thus a 5x5
array. Future implementations will treat open class verbs in a more realistic manner,
including identification their category and associated argument structure. This
information is currently implicitly provided during the response phase, as the model is
interrogated until all thematic roles have been specified.
for the ith noun, STM(i) = Input (3)
8.3 Learning
During the supervised training, after a sentence is read (and STM is thus filled with
the nouns in their order of arrival), in the response phase, the model is provided with
the canonically ordered nouns one at a time. In the subsequent testing, all of the nouns
to be assigned thematic roles are presented at once and the model must choose the
correct one. Thus, during the training, the model must learn the ordered
relations/transformations between the nouns as they appear in the sentence, and the
same nouns presented in canonical order. Thus, to detect if the current response
(generated in response the presentation of the single noun) is a repetition of one of the
stored nouns, it is compared with each of the nouns encoded n STM. The result is
stored in a 6-element vector called Recognition. Each Recognition element i, for 1 ~ i
~ 5, is either zero if Out is different from STM(i) or 1 if they are the same, as
described in Equation 4. If no match is detected in STM for a given response in Out,
Recognition(O) is set to 1, indicating that a unique (u) response has occurred.
Recognitioni = STM(i) * Out (4)
The Recognition vector plays an important role, both in providing information to

State as described in Equation 1.1, and also in the learning procedure itself. State
76 P. F. DOMINEY
encodes the ordered history of closed class words, and the open class words that have
been so far assigned thematic roles.
To exploit this information in the State in order to access the STM and (re)order
the stored nouns in the canonical order, the model must selectively take the contents
of the STM, and direct this STM content to Out. In order to achieve this, a new
learning rule is developed. As stated, during the supervised training, after the sentence
is presented, the canonically ordered nouns are then provided. Each time a repetition
is recognized between one of the nouns and a noun stored in the STM (Recognition,
Eqn (4», connections are strengthened between the active context (State) units and
units in a five-element vector, Modulation (described below), that modulate the
contents of the matched STM element to the output. The result is that this State
pattern becomes increasingly associated with the occurrence of the matched element
such that, after learning, this association can be used to select the correct noun from
those presented during the testing phase. This learning rules is described in Equation
5.
(5)
The goal of this learning is to allow State to modulate the contents of specific,
predicted STM elements into Out, so that the open class elements are retrieved in their
canonical order. To permit this, a 5-element vector, Modulation, is introduced such
that for i = 1 to 5, if Modulationi is non-zero, then the contents of STM(i) is
modulated or directed to Out. This explains the Modulation term in Equation 1.1.
Based on the learning in Equation 5, State now directs this modulation of the STM
contents into Out via State's influence on Modulation, as described in Equation 6.
Modulation i = t W;JM State j (t) (6)

f-'l
After training on sequence "The elephant was given to the monkey by the rabbit. -
rabbit, elephant, monkey", when the model is exposed to a new isomorphic sequence
"The giraffe was given to the lion by the tiger" the closed class sequence "the _ was
_en to the _ by the _" will evoke a pattern of activation of State units that will drive
the Modulation unit corresponding to the STM(3) element, directing its contents
"tiger" to the output, leading to the correct selection of the agent when all three nouns
are presented. The changing State activity will allow the same selection to be made
then for the object and recipient. Errors will invoke the learning rule with a negative
learning rate, thus reducing the probability of a repetition of the same error.
9. AFFILIATIONS
Dr. Peter Ford Dominey

Institut des Sciences Cognitives
CNRS UMR 5015
67, Boulevard Pinel
69675 BRON Cedex
France
e-mail: dominey@isc.cnrs.fr
10. REFERENCES
Brown, C. M., Hagoort, P., & ter Keurs, M. (1999). Electrophysiological signatures of visual lexical
processing: Open- and closed-class words. Journal of Cognitive Neuroscience, 11, 3, 261-281.
Caplan, D., Baker, c., & Dehaut, F. (1985). Syntactic determinants of sentence comprehension in aphasia.
Cognition, 21, 117-175.
Dominey, P. F., & Lelekov, T. (2000). Non-linguistic transformation processing in agrammatic aphasia.
Comment on Grodzinsky. Beh and Brain Sciences, 23, 1, 30.
Dominey, P. F., Lelekov, T., Ventre-Dominey, 1., & Jeannerod, M. (1998). Dissociable processes for
learning the surface and abstract structure sensorimotor sequences. Journal of Cognitive Neuroscience,
10, 6, 734-751.
Dominey, P. F., & Ramus, F. (2000). Neural network processing of natural language: I. Sensitivity to serial,
temporal and abstract structure of language in the infant. Language and Cognitive Processes, 15, 1, 87-
127.
Elman, J. L. (1990). Finding structure in time. Cognitive Science, 14, 179-211.
Elman, J. L. (1993). Learning and development in neural networks: the importance of starting small.
Cognition,48,71-99.
Friederici, A D. (1985). Levels of processing and vocabulary types: evidence from on-line comprehension
in normals and agrammatics. Cognition, 19, 133-166.
Haarman, H. 1., Just, M. A, & Carpenter, P. A (1997). Aphasic sentence comprehension as a resource
deficit: A computational approach. Brain and Language, 59, 76-120.
Lelekov, T., Franck, N., Dominey, P. F., & Georgieff, N. (2000) Cognitive sequence processing and
syntactic comprehension in schizophrenia. Neuroreport, 14, 2145-2149.
Marcus, G. F., Vijayan, S., Bandi Rao, S., & Vishton, P. M. (1999). Rule learning by seven-month-old
infants. Science, 283, 5398,77-80.
Nazzi, T., Bertoncini, 1., & Mehler, J. (1998). Language discrimination by newborns: Towards an
understanding of the role of rhythm. Journal of ExpPsych. Human Percept & Perform, 24, 3, 1-11.
Neville, H. 1., Mills, D. L., & Lawson, D. S. (1993). Fractionating language: Different neural subsystems
with different sensitive periods. Cerebral Cortex, 2, 244-258.
Osterhout, L. (1997). On the brain response to syntactic anomalies: Manipulation of word position and
word class reveal individual differences. Brain and Language, 59, 494-522.
Pulvermiiller, F. (1995). Agrammatism: behavioral description and neurobiological explanation. J Cog
Neuroscience, 7, 2, 165-181.
Saffran, J. R., Aslin, R. N., & Newport, E. L. (1996). Statistical learning by 8-month-old infants. Science,
274, 1926-1928.
J. WEISSENBORN
THE ACQUISITION OF VERB PLACEMENT IN

GERMAN: A NEW LOOK
Abstract. After an overview of the V2 phenomenon and the explanations which have been given for it in
theoretical linguistics, the developmental data and different accounts for them are discussed pointing out
various problems. Based on findings from experiments with 2 to 6 years old children using the head turn
preference paradigm and a sentence repetition task a new approach is proposed arguing for a very early
access to the critical parametric information, and explaining the developmental facts as resulting from the
interaction of grammatical and processing constraints.
1. INTRODUCTION
The position of the finite and non-finite verb in main and embedded clauses is a
central locus of cross-linguistic variation. The issue of how to derive this parameter
of typological variation from more general properties of a given language has been
the major concern of much influential work in linguistic theory in the recent years
(e.g. Chomsky, 1986, 1993, 1995; Pollock, 1989). The question how the child
acquires this crucial aspect of the parametric structure of the target language has
equally played a central role in the acquisition research of the last years (e.g. Meisel,
1992). Much of this latter work draws heavily on the linguistic analyses in order to
explain how this aspect of adult linguistic knowledge is acquired by the child. Under
the assumption that the way how acquisition proceeds is not independent from the
structure of the knowledge which has to be acquired acquisition research may help
to decide which theoretical account may be considered to most adequately represent
the linguistic knowledge of the adult.
This paper will focus on a special case of verb placement, namely the acquisition
of "Verb-Second" (henceforth V2) in German which has been and still is in the
center of an ongoing debate. The structure of the paper is the following: First, we
will give a short overview of the V2 phenomenon and the accounts that have been
given for it in theoretical linguistics. Second, we will introduce the developmental
data on which our subsequent discussion will be based. Third, we will sketch out
different approaches including our own to language acquisition which underlie the
discussion of the acquisition of V2. Forth, we will present and discuss different
accounts of the developmental data concluding with our own proposals.
1.1 The Verb-Second Phenomenon

The issue of "verb placement" concerns the regularities that underlie the position of
verbs in the clause. It is widely assumed that these regularities are the result of the
interaction of properties like finiteness and agreement which may be

80 J. WEISSENBORN
morphologically instantiated by the verb. "Finiteness" is understood as a complex

concept involving tense, mood and illocution (for an illuminating discussion see
Klein, 1998). The V2 phenomenon concerns a particular case of verb placement
namely the observation that in root clauses exactly one constituent can precede the
finite verb. This property holds in a general way, i.e. for declarative and
interrogative clauses, in a number of West Germanic languages like Dutch and
German. In other languages like English or French it holds only for a subset of root
clauses, e.g. interrogatives, whereas in declarative sentences more than one
constituent may precede the finite verb as shown in Sentence 1 and Sentence 2:
(1) John he likes
(2) Jean, ill'aime bien.

John, he himClilic likes well
Thus, in contrast to generalized V2 languages, English and French are considered

residual V2 languages. Within a Universal Grammar framework these typological
differences have been related to differences in the position of the finite verb in the
underlying syntactic representation. That is, assuming X-bar Theory and two levels
of functional projections, IP and CP above VP, the finite verb could be either located
in 1° or, for Verb-second, in Co. 1
Different proposals have been made to account for these language particular
positions of the verb. In a Minimalist framework (Chomsky, 1995), they are
explained in terms of strong vs. weak verbal features, e.g. [+ Tense], [+AGR],
associated with the functional heads 1°, or Co. The overt movement of the finite verb
to CO in Dutch and German is supposed to be forced by the presence of strong verbal
features in Co. Depending on the language, these may also be spelled out by
complementizers.2 This leads to the well known root clause/non-root clause
asymmetry for the placement of the finite verb in German and Dutch: the finite verb
moves to CO in matrix clauses but has to stay clause-final in embedded clauses with
complementizers.
To summarize, the underlying structure for German declarative, interrogative
and embedded clause would thus be as shown in Sentence 3 a-c:
(3 a) [cp Hansi [c liestj [Ip t; ein Buch tj]]]

Hans reads a book
(3 b) [cp Wasi [c liestj [II' Hans ti tj]]]?

what reads Hans
(3 c) [cp [c dass [Jp Hans ein Buch liest]]]

that Hans a book reads
Since the seminal work of Bierwisch (1963) and Den Besten (1989) verb placement
in German has been extensively discussed in the literature (for an overview see
THE ACQUISITION OF VERB PLACEMENT IN GERMAN 81
Muller & Penner,1996). One major issue in this debate, relevant to the discussion of
the acquisition of V2, concerns the question whether all main clauses in German
have the same underlying representation or not. Thus contrary to the assumption that
in root clauses all verbs move to CO (e.g. Schwartz & Vikner, 1996) it has been
proposed (e.g. Reis, 1985; Travis, 1984; Zwart, 1997) that this holds only for a
subset of root clauses namely non-subject initial clauses and wh-questions but not
for subject initial clauses. In these clauses the verb is supposed to move only to 1° as
shown in Sentence 4:
(4) [IpHansj [Iliestj [vp tj einBuch tj]]]

Hans reads a book
One problem with this proposal is how to explain why structures like in Sentence 1
and Sentence 2 are. not allowed in German and Dutch. There must be a difference
between the IP of English and French on the one hand and of German and Dutch on
the other hand which blocks topicalization of objects or adjuncts.
Another problem with the proposal that subject-initial clauses should be
analyzed as IPs arises from the clause final position of the finite verb in embedded
clauses with complementizers. If one assumes that in embedded clauses the verb
also stays in 1°, one would have to pose two kinds of IPs for German: a head-initial
IP in matrix clauses vs. a head-final IP in embedded clauses. There are at least two
proposals in the literature which constitute potential solutions to this paradox. One is
Haider's (1993) suggestion that in German there is no convincing evidence for a
head-final IP projection in embedded clauses and that consequently there is only one
functional projection dominating VP which can be the result of matching several
functional projections, e.g. CONFL (see also Platzack, 1994). From this it follows
that VO must be a possible host for finite verbs. The other way to cope with this
paradox is to assume with Zwart (1997), following Kayne (1993) that there is an
universal base-structure namely (S)VO. This means that the (S)OV structure in
Dutch and German embedded clauses are not supposed to reflect the base structure
as first proposed by Bierwisch (1963) and Den Besten (1989) but has to be
considered as derived by object movement to a preverbal position.
In the following we will leave aside the question whether subject initial main
clauses differ syntactically from non-subject initial clauses and will assume that the
finite verb occupies the same position, i.e. Co, in both structures.
1.2 Acquisition Data and Linguistic Theory
Given the different accounts for verb placement in adult Dutch and German the
question arises whether the acquisition data can help us to decide between them.
Obviously the language learning the child must be able to derive at least all language
particular regularities of the target from the input. We thus assume following among
others Lebeaux (2000), Penner (1994b), Rizzi (1994), Roeper (1996), Weissenborn
(1994), that the proposed structures for the adult language should be compatible
with the acquisition data in order to be considered as potentially empirically
adequate.
82 1. WEISSENBORN
1.3 The Developmental Problem
Given the above characterization of the V2 phenomenon, the developmental issue

can be formulated as follows: First, the German child has to find out that the
language to be learned is a genuine V2 language, i.e. has obligatory V2 (V-to-C)
movement. The occurrence of non-subject sentences alone cannot tell the child that
V-to-C is obligatory in all finite matrix clauses: as we have seen above, subject
initial clauses at the first sight are ambiguous between an IP and a CP interpretation.
Second, the child has to find out that in embedded clauses introduced by a
complementizer the verb has to stay in the final position.
2. THE DEVELOPMENTAL DATA
In this section we will briefly present the data which underlie most of the discussion
on the acquisition of verb movement in German.
2.1 The Data Base
There exists a number of longitudinal corpora the classical one being the one
presented in Stern and Stern (1928) (see also Clahsen, 1991; Kaltenbacher, 1990;
Miller, 1976; Tracy, 1991). Nevertheless the overall data situation is still
unsatisfactory because given the limited number of corpora it is difficult to draw
firm conclusions from the absence or the scarcity of particular types of
developmental data. Certain developmental phenomena may be so rare and so fast
that they may have escaped documentation in the available data base. Thus, in the
absence of a near to complete, i.e. day by day, documentation of the acquisition
process for at least some children up to the age of three years, our conclusions must
stay more speculative than they ought to be. 3
2.2 Descriptive Picture of Overall Development

In the following overview we will only mention the main phenomena related to the
question of the development of verb placement.
Verbs and verb particles (e.g. rein "into", weg "away") occur from the one word
stage on (e.g. Bennis, den Dikken, Jordens, Powers, & Weissenborn, 1995; Penner,
Wymann, & Dietz, 1998). The first verbs are main verbs. The copula, modals and
auxiliaries emerge slightly later (e.g. Behrens, 1993; Kaltenbacher, 1990; Mills,
1985).
Main verbs are mainly non-finite in the beginning. Modals and auxiliaries first
occur only as finite forms. From the two-word stage on finite and non-finite verbs
when combined with an object or an adverb occur predominantly in the target
position: initially for finite verbs, finally for non-finite verb, e.g. geht nicht "doesn't
work" vs. nicht gehen "not work". An asymmetry can be observed between finite
and non-finite verbs with respect to position errors: there are almost none for the
latter, and about up to 10% for the former. Interestingly, erroneous finite verb-final
constructions are found predominantly in children with language disorders (Penner,
Wymann & Dietz, 1998; Schaner-Wolles, 1994).
Non-finite utterances like the ones in Sentence 5 and Sentence 6 are initially
predominant with up to 90% of the utterances containing a verb (e.g. Behrens, 1993;
Weissenborn, 1990):
(5) schuh ausziehn S (1;11;13) 4

shoe take off
(6) maxe auch (mu)sik mache(n) S (1;11;14)

Max also music make
The number of clauses with non-finite verbs decreases gradually over time without
disappearing completely (e.g. Lasser, 1997; Weissenborn, 1990, 1994). The number
of non-finite constructions with subjects is mostly below 10% (Weissenborn, 1990).
A source of finite verb initial utterances are clauses with missing subjects. These
are structurally adult-like although pragmatically they may be inadequate (e.g.
Weissenborn, 1992).
Similarly, the first wh-questions may occur without an overt operator resulting in
finite verb-initial utterances which look like yes-no questions (e.g. Felix, 1980;
Penner, 1994a):
(7) is das? S (2;00;05)

(what) is this
The occurrence of target-like wh-pronouns may be preceded by the appearance of

placeholders, i.e. phonologically reduced forms (e.g. Muller & Penner, 1996). A
similar development can be observed for embedded finite clauses: here too the
complementizer may be first missing resulting in deviant finite verb-end structures
(e.g. Miiller & Penner, 1996; Weissenborn, 1990):
(8) pappi sagt ( ... ) schOne hose anzieht H (2;01;18)

(= angezogen) hat
daddy says (that) (he) has put nice pants on
Here too placeholders may appear before the first overt complementizers emerge. It
should be noted that from the beginning the finite verb correctly occupies the final
position in embedded clauses even when the complementizer is missing.
(9) fenster heiss ist S (2;00;03)

window hot is
(10) dass du has(t) net die meerjungfrau B (3;00;19)

that you have not the seemaid
(Gawlitzek-Maiwald, Tracy, & Fritzenschaft, 1992)
84 J. WEISSENBORN
Occasionally other non-target-like verb placements have been observed. These are
either violations of the verb-second constraint in root clauses (Sentence 9) or
violations of the verb-end constraint in embedded clauses (Sentence 10).
Two general observations have to be made which will be discussed later. First, in
most cases we find that non-target-like structures and the corresponding target-like
structures co-occur, although in different proportions, throughout the documented
developmental period. Second, most of the non-target-like forms can occasionally
also be observed in adult speakers (Tracy, 1991; Weissenborn, 1992).
A last point concerns verbal inflection, i.e. subject verb agreement. Although not
all of the forms of the paradigm are available to the child from the beginning, verbal
inflections, when present, are mostly target-like (Behrens, 1993; Clahsen,
Eisenbeiss, & Penke, 1996; Kaltenbacher, 1990; Mills, 1985; Rohrbacher &
Vainikka, 1995; Verrips & Weissenborn, 1992). The most frequent deviation from
the target is the use of bare stems like mach "make". These forms occur in the V2
position as well as in clause final position.
3. THE DEVELOPMENT OF V2
Before we will discuss the specific proposals related to the acquisition of V2 in

German we will briefly characterize the main theoretical positions which underlie
the different developmental accounts. This discussion will allow us to evaluate these
accounts with respect to the developmental predictions which follow from them.
3.1 Theoretical Approaches to Language Acquisition

As already mentioned above the acquisition of verb placement in general and of the
acquisition of V2 in particular has been the object of considerable research during
the last years. Concerning the V2 issue, extensive discussions can be found in the
contributions to Meisel (1992), Lust, Whitman, and Kornfilt, (1994), Clahsen et al.
(1996), and in many separate papers. There are two general questions which are
obviously not specific to the acquisition of verb placement that dominate the
discussion. First, how do children's developing linguistic representations change
from the initial state to the final, i.e. the adult state? This is the issue of the
determination of developmental steps. Second, how can we account for the specific
developmental steps that we do observe?
With respect to the relation between the child's linguistic knowledge and the
adult's basically two different types of hypotheses have been proposed to account
for it:
a. The Competence Difference Hypothesis: The grammatical knowledge of
the child and the adult differ. It is assumed that the adult knowledge is
constrained by general principles and parameters of Universal Grammar
(UG) The latter can only vary along a limited number options (Chomsky,
1986).
b. The Performance Difference Hypothesis: The grammatical knowledge of
the child and the adult are basically the same. The differences result from
constraints on other components of the linguistic processing system like
perceptual restrictions or memory limitations (e.g. Bloom, 1990).
We can distinguish three versions of the Competence Difference Hypothesis:

a. The Discontinuity Hypothesis: The linguistic behavior of the child may
violate UG constraints (e.g. Felix,1984).
b. The Weak Continuity Hypothesis: The child's linguistic representations
are UG-constrained but can be inconsistent with the target parametric
system (e.g. Borer & Wexler, 1987; Hyams, 1986, Lebeaux 2000; Roeper
& de Villiers, 1992)
c. The Strong Continuity Hypothesis: The differences between the grammar
of the child and the adult stay within the parametric space of the adult
language (e.g. Pinker, 1984; Penner, 1994b; Weissenborn, 1990, 1994).5
From a learn ability point of view the complexity of the learning task decreases from
(a) through (c) because the amount of linguistic knowledge that may have to be
revised decreases. We leave aside the question whether it is at all possible for the
child to correct a wrong parametric choice given the available evidence.
With respect to the second question of how changes in the child's grammatical
knowledge are brought about we can distinguish three approaches:
(a) Changes in the child's linguistic knowledge like the availability of particular
grammatical operations are due to maturation, i.e. they emerge spontaneously
at a certain moment in development (e.g. Rizzi, 1994; Borer & Wexler, 1987).
(b) Changes in the child's linguistic knowledge are due to "lexical learning", e.g.
"The idea is that functional categories such as IP, AGRP, etc. or syntactic
features may come into the child's phrase structure representations as a
consequence of the child's learning a regular inflectional paradigm of distinct
inflectional affixes." (Clahsen et aI., 1996, p. 133).
(c) Changes are due to parameter setting based on the idea of triggering through
unambiguous trigger information which may differ in accessibility (Penner,
1994b; Penner & Weissenborn, 1996; Roeper & Weissenborn, 1990;
Weissenborn, 1990, 1992, 1994).
In the following section we will basically rely on the last approach. We assume that
the temporal course of language acquisition depends mainly on differences in the
accessibility of the input information which is needed to set a particular parameter.
The model assumes strong continuity as defined above. The observation that has
been crucial for the development of the model is that in certain syntactic domains
like for example object placement the child's linguistic behavior is almost adult-like
from the beginning (e.g. Penner & Weissenborn, 1996; Sch6nenberger, Penner, &
Weissenborn, 1997; Weissenborn, 1990). This means that the relevant parameters
must have been set extremely early, probably already during the prelinguistic phase
but no later than the one-word stage. This implies that the relevant trigger
information for setting the parameters must have been accessible to the child from
very early on.
In other syntactic domains, for example verb placement, parameter setting
apparently takes longer. We assume that in these cases the relevant trigger
information is not so easily accessible to the child than in the preceding case. We
also assume that the deviations from the target to be expected under this scenario are
strongly constrained by the specific licensing conditions of the target (Penner,
86 J. WEISSENBORN
1994b), or, in the same spirit, by local well-formedness conditions with respect to
the target (Weissenborn, 1993, 1994). That is, we assume that at any point in
development the linguistic representations of the child are compatible with the
representations of the adult in the sense that they are in a subset relation to them (cf.
also Lebeaux, 2000).
3.2 Accounting for the Developmental Data
3.2.1 Finite Clauses

The issue which has been so far dominating the discussion of the acquisition of V2
in German has been whether or not the structure underlying the child's first finite
clauses is the same as the one assumed for adult language that is a C-projection as in
Sentence 3. This discussion is exclusively based on spontaneous production data. In
3.3. it will be shown that this data has to be supplemented by data from perception in
order to obtain a more adequate picture of the child's developing linguistic
knowledge.
The view that there is no difference between the child and the adult has been
proposed by authors like Boser, Lust, Santelmann, and Whitman (1992), Poeppel
and Wexler (1993), Whitman (1994). These accounts are based on evidence like the
early occurrence of wh-questions with overt operators and the topicalization of non-
subjects (adjuncts, arguments, VPs). A mixed position is taken in Weissenborn
(1990) where wh-questions are assumed to be CPs whereas subject-initial V2
declaratives as long as non-subject initial declaratives are not produced, following
Reis (1985) and Travis (1984), are considered to project to IP.
The position that the child's syntactic representation underlying finite clauses
initially differs from the adult's one has been defended by a number of authors. Thus
some authors assume that the highest projection the finite verb initially moves to is
IP (Gawlitzek-Maiwald et aI., 1992; Rohrbacher & Vainikka, 1995;) or one of its
off-springs under a split-INFL analysis (e.g. Pollock, 1989) like TNSP (Meisel &
Miiller, 1992), or a functional projection specified for finiteness but not for
agreement like FP (e.g. Clahsen, 1991). Among the reasons for the assumption that a
C-projection is initially absent are the initial scarcity of clear cases of non-subject
topicalizations, the absence of wh-questions other than where-questions which are
assumed to be formulaic given there often prosodically reduced form (e.g. was.. ?
"where's" instead of wo ist .. ? "where is .. "), the absence of overt complementizers,
the distribution of null-subjects, and the absence of adult-like subject-verb
agreement patterns.
A phenomenon that has repeatedly been put forward as evidence for differences
between the child's grammar and the adult's are the finite verb-end constructions
mentioned above. Apparently only finite main verbs seem to occur in this position to
the exclusion of finite auxiliaries and modals (Clahsen et al. 1996). These final finite
verb constructions have lead some authors to the assumption that initially finite, i.e.
tensed verbs must raise to head-final INFL - obeying a UG principle that operators
(e.g. tense) have to bind a variable (Pollock, 1989) - but may raise only optionally to
Co, the reasons for this being that the child has not yet figured out the features of the
C-system in German which force V-to-I-to-C movement (Deprez & Pierce, 1993),
or that the C-projection may be truncated (Rizzi, 1992, 1994). Further proposals are
THE ACQUISITION OF YERB PLACEMENT IN GERMAN 87
that initially the child is undecided between a head-initial and a head-final IP-
projection (Gawlitzek-Maiwald et aI., 1992) or that agreement may not project
allowing the verb to stay in clause final position (Meisel & Miiller, 1992). Recently,
in order to account for the finite verb-final structures Clahsen et aI. (1996) proposed
that initially yO may be neutral with respect to finiteness and thus may host both
finite and non-finite verbs in contrast to the highest projection in the verbal
extension, FP, which is positively specified for finiteness.
3.2.2 Problems with the Preceding Accounts

In the following we will point out some learning theoretical and empirical
difficulties with the preceding accounts. A major difficulty with all the accounts
which assume optionality of verb raising, either from yO -to-1° (Clahsen et aI., 1996)
or from IO-to-CO (Deprez & Pierce, 1993), as well as for the switching-head-
directionality-of-IP account of Gawlitzek-Maiwald et aI. (1992), is that we observe a
completely different distribution of finite verb-final constructions in main clauses
than the one we would expect under optionality of movement, that is a random
distribution of finite verbal forms between the positions available to them. But this is
not what we find: finite verb final constructions are much less frequent than one
would expect (e.g. Clahsen et aI., 1996; Behrens, 1993; Kaltenbacher, 1990). In
addition these accounts also predict that finite modals too should occur clause finally
what, as mentioned above, they almost never do.
A further empirical problem arises for IP-accounts which assume that initially
there is only one functional projection available to host the finite verb (e.g.
Gawlitzek-Maiwald et aI., 1992; Rohrbacher & Yainikka, 1995). These accounts
cannot explain that, from the start, children's early embedded clauses are correctly
verb-final (Clahsen, 1982; Roeper, 1973; Rothweiler, 1993). That is, under a left-
headed IP analysis one would initially expect to find many more ungrammatical Y2
structures in embedded clauses introduced by a complementizer than we actually do
find (Hyams, 1996; Poeppel & Wexler, 1993; Weissenborn, 1990).
Furthermore, the sporadic occurrence of verb copying constructions as in
Sentence 11:
(11) is das is S (2;00;05)

(what) is this is
can not be explained under an account which assumes that initially there is only one
functional position available for finite verbs. But these utterances can easily be
explained if we assume that there are at least two positions for the finite verb, i.e. the
base position to the right, and a derived position to the left.
A last point concerns a problem we have already mentioned, namely why, if
German subject-initial root clauses are IPs like English root clauses, topicalization
of the type allowed in English ("John, Mary likes") is not possible. Similarly, if
children's early finite Y2 clauses were IPs on would expect argument topicalizations
to occur in the language of the child. This expectation is not borne out. There are
occasionally Y3 structures like the one in Sentence 12 to be observed.
But these type of Y3 constructions are almost exclusively introduced by a deictic
element, i.e. an adjunct, and they do occur over the whole period of data collection.
88 J. WEISSENBORN
(12) hier gummibaerchen hats S (2;00;23)

here gummy bears has it
"here are gummy bears"
The learning theoretical problem is related to the question which kind of

evidence would allow the child to retreat from her non-target like initial grammar.
For Rizzi's (1992, 1994) truncation account the question does not arise because he
assumes a maturational change. The account of Clahsen et aI. (1996) relates the
acquisition of target-like V2 the to acquisition of subject-verb agreement, i.e. the
learning of inflectional paradigms. This assumption is not compatible with the
finding that in dysgrammatic children we find a clear dissociation between the
development of agreement morphology and verb movement: there are children who
show target-like verb raising (V2) without mastery of agreement morphology and
vice versa (e.g. Hamann, Penner, & Lindner, 1999; Penner, 1994b; Schaner-Wolles,
1994). This is clear evidence for the independence of the two areas of grammar
which develop more or less simultaneously in normal children. Thus the question
remains open what could possibly induce the child to retreat from its non-target-like
grammar.
3.2.3 Non-Finite Clauses

Especially one type of early child utterances has been intensively discussed with
respect to the question whether the child's initial verb grammar differs from the one
of the adult, namely the initially very frequent infinitival sentences which are used
by the child like declarative sentences (Wexler'S "root infinitives", 1994). Under
one account these sentences have been analyzed as finite V2-clauses with a null-
modal in second position, i.e. in CO (e.g. Boser et aI., 1992; Poeppel & Wexler,
1993; Whitman, 1994). This analysis is illustrated for Sentence 13 in Sentence 14
and in Sentence 15:
(13) mone hocker haben S (2;00;05)

(Si)mone chair have
(14) Child: (cpmonei [c NULLMODALj [IP tj [vp hocker haben] tj]]]

(Whitman, 1994)
(15) Adult: [cp Simone [c will [Ip [vp den Hocker haben ]]]
Whitman (1994) extending Rizzi's wh-criterion assumes that if a modal verb stays
in a agreement relation with an overt preverbal subject it can be phonologically zero.
Basically the Null-Modal-Hypothesis (NMH) predicts that we should find all the
structures we find in normal declarative V2 clauses with an overt verb in Co, with
the constraint, that under the NMH the initial position is limited to the subject. This
is the basis for the predictions in (a-c):
(a) Only subjects should occur in initial position in infinitival sentences.

(b) More than one constituent should be allowed to occur to the left of negation or
of a focus particle ("scrambling") as shown in Sentence 16:
(16) Child: [cp Simone[c NULLMODAL[Ip das Eis j nicht[ vp tj haben ltd]]
(c) Infinitival questions should occur as shown in Sentence 17:
(17) Child: [cp Werj [cNULLMODALj hp tj [vp hocker haben tj] t'j]]]
But as shown in Sentence 18, prediction (a) is not born out:
(18) *des auch mone hol(en)6 S (2;01;21)

this also (Si)mone get
Neither are predictions (b) and (c), for which no examples are attested in the large
Simone corpus.
Instead we propose to analyze sentences like Sentence 13 as infinitival phrases
as shown in Sentence 19. Assuming Grimshaw's (1997) Extended Projection
Principle (i.e. that intermediate projections cannot be stripped) this leads to the right
predictions (Weissenborn, 1994):
(19) [vpsimOneSubj [v·hockerObj haben]]
The absence of structures like Sentence 16 is naturally accounted for if following

Penner, Tracy, and Weissenborn (2000) we assume that there is only one position
available to the left of the focus particle auch in the specifier of the focus phrase.
Under the same assumption non-subject initial structures like Sentence 18, repeated
here as Sentence 20 which are illicit under the NMH, are predicted to occur:
(20) [Focuspdesj [FOCUS auch [vp mone tj holen]]]
Furthermore infinitival questions as in Sentence 21 with the representation in

Sentence 22 are excluded given that they Grimshaw's (1997) Extended Projection
Principle: a CP must select an IP which under our analysis is not available as shown
in Sentence 23:
(21) *Wer Hocker holen?

who chair get
(22) *[cp Wer [vp Hocker holen ]]]
(23) *[CP [VP ...... ]]
We thus conclude that our analysis of infinitival utterances is correct. Notice that our
90 J. WEISSENBORN
account for the absence of infinitival wh-questions in the German child data which
seems also to hold for French (Weissenborn, 1994; see also Crisma, 1992) suggests
that child English "where it go" should also not be analyzed as an infinitival
question. It instead supports a null-DO analysis as proposed by Pesetsky (1993).
A question which arises in this context is whether the occurrences of infinitival
sentences is tied to the absence of tense-distinctions in the child's grammar
hypothesized by Wexler (1994). Behren's (1993) study of the acquisition of tense in
German shows that preterit forms as well as present perfect forms occur from early
on (e.g. at 22 months in the Simone data), and that they remain stable at the same
level of occurrence, whereas the number of root infinitives decreases steadily with
age, independently of an increase in the occurrence of preterit forms. We can thus
conclude, as also pointed out by Behrens, that the assumed connection between the
absence of tense distinctions and the occurrence of root infinitives is not born out by
the data.
Summarizing the discussion so far we can say that there is a large consensus that
verb raising takes place very early in child German but there is disagreement about
the nature of the positions the finite verb initially may occupy, and about how the
child reaches the adult knowledge. Thus crucial aspects of the development of verb
placement in German remain still unclear. An important question which has still to
be investigated is what kind of input information the child can and does in fact use
in order to derive the two main regularities of verb placement in German, i.e. V2 in
root, and V-end in non-root clauses. Another question is how to explain the specific
quantitative distribution of particular constructions in the data over time. It
especially concerns the initial preponderance of (finite and non-finite) V-end
constructions. We will discuss these questions in the following section.
3.3 A Novel Look at the Development ofV2 in German

Based on the acquisition model sketched out in 3.1 we propose that the development
of verb placement in German roughly proceeds as follows.
We assume that one of the first parameters to be set is the head-directionality
parameter for VP. Evidence for this comes from almost errorless object placement,
i.e. OV (98% correct), in infinitival clauses from the earliest possible moment on
when this regularity can be observed, i.e. from the two-word stage on (see
SchOnenberger, Penner & Weissenborn, 1997). This means that the child must have
set the corresponding parameter before this stage, possibly at the one-word stage or
even during the prelinguistic period. There is evidence which suggests that this is
achieved by the child on the basis of prosodic information in the input (see Hahle,
Weissenborn, Schmitz, & Ischebeck, 2001; Penner, Weissenborn,& Wymann, 2001)
This finding is at variance with Kayne's (1993) Universal Base Order
Hypothesis (SVO) mentioned above which by definition would constitute the initial
developmental default option. Under this hypothesis we would predict at least
gradual development for object placement and hence more errors of the type (S)V
non-finite 0 for object placement than we actually find because the child would have to
derive (S)OV from (S)VO. Recently Platzack (1996) has claimed that as predicted
by Kayne'S hypothesis there is evidence for an initial SVO stage in the acquisition
of languages like Irish and Arabic which both are VSO. But the initial German (and
Dutch) child data so far do not support the Universal Base Order Hypothesis.
How does the child succeed to set this parameter so early? We propose that the
child can set the parameter value via syntactic bootstrapping on the basis of the
root/non-root asymmetry of verb placement focusing on the information contained
in the non-root domain. This is shown in Sentence 24:
(24) /ch mag dieses Buch, obgleich du seinen Autor nicht magst.
I like this book although you do not like its author
We assume that the non-root domain which is opaque to discourse interactions

reflects the underlying structure of German (Emonds, 1976), and thus can provide
the child with the necessary unambiguous trigger information for the setting of the
V2 Parameter (Roeper & Weissenborn, 1990; Weissenborn, 1992; Penner, 1994b;
Penner & Weissenborn, 1996). 7
How could such a scenario work? Could the child unambiguously deduce from
the root clause/ non-root clause asymmetry that the finite verb in the matrix clause
has to occupy the V2 , i.e. CO position, or, in other words, that German has INFL-in-
CaMP? There is evidence that non-finite forms never show up in the V2 position.
Thus, the child must have learned from the earliest occurrence of verbal forms on
that there is a position to the left of the base position of the verb which can only host
finite verbs. Thus the distinction between finite and non-finite verbs must have been
made before the first occurrences of these forms in the speech of the children
(Verrips & Weissenborn, 1992). But the fact remains that except in the case of non-
subject initial root clauses which must project to CP the subject-initial root clauses
are ambiguous between an IP and a CP-analysis of the position of the finite verb.
Although one may wonder whether the absence of topicalizations of the type "Cake,
John likes" in German child language during the supposed IP-stage does not
constitute sufficient evidence that the child's subject initial clauses do indeed project
to CP as in the target.
But the fact mentioned above that object topicalizations and non-formulaic wh-
questions as well as overt complementizers are initially missing (e.g Weissenborn,
1990; Penner, 1994b) can be interpreted as indicating the absence of a CP-Iayer at
this point of language development.
What would then constitute unambiguous information for the child that in
German all preverbal constituents are topicalized elements? Following Penner's
(1994b) analysis for Bernese Swiss German we propose that it is the occurrence of
second person subject pronouns in preverbal position in finite root clauses that
signals to the child that German is not a residual but a generalized V2 language. The
second person pronoun du "you" must be considered an expletive element in
German given the fact that the inflectional ending -st on the finite agreeing verb
unambiguously marks a second person subject. From this follows assuming some
version of a phonological economy principle, that the subject pronoun may be
dropped or phonetically reduced. We observe that pronoun-drop and reduction of du
is only allowed in the post-verbal position and in the embedded clause. These are
arguably semantically neuter positions in which the person information provided by
the inflection of the verb is sufficient. To the contrary reduction of the second
person pronoun should not be possible in a semantically marked position like the
92 1. WEISSENBORN
topic position. And that is what we find in German: the pronoun has to be fully
realized in preverbal position in root clauses. If this kind of idiosyncratic
paradigmatic information which, as we claim, is lower on the trigger accessibility
hierarchy than information deductible from the root/non-root asymmetry, is
necessary for the child to determine that German is a generalized V2, i.e. [INFL-in-
COMP] language than we would indeed expect delayed acquisition of V2 until the
second person singular properties are acquired. This seems to be the case. The
emergence of second person subject pronouns seems to coincide with target-like
topicalization patterns, wh-questions and the appearance of overt complementizers
(see e.g. Muller & Penner 1996; for a different interpretation of the importance of
the second person inflection for the acquisition of V2 in German see Clahsen 1991,
Clahsen, Eisenbeiss, & Vainikka, 1994).
Thus, as proposed by Penner (1994b), and Muller & Penner (1996), during a
short transitional period the indeterminacy of the child with respect to the feature
characteristics of COMP is reflected on the production side in the use of
phonologically reduced place holders in the CP domain, like formulaic wh-pronouns
and reduced complementizers. On the comprehension side this indeterminacy
manifests itself in the child's difficulty to interpret wh-questions due to the
impossibility to establish interpretative operator - variable chains given the
underspecification in the CP-domain (see Roeper, 1992; Weissenborn, Roeper &
DeVilliers, 1995; for tense chains see Hyams, 1996).
Under the preceding account for the acquisition of V2 in German the occurrence
of finite V -end structures is interpreted as an interim-solution reflecting the
uncertainty of the child concerning the properties of the C-projection, and the
acquisition process focusing on non-root clauses.
Obviously the preceding account for the acquisition of Verb-second in German
relies heavily on the assumption that the child has access from very early on to the
information contained in embedded clauses. In the following we will provide
evidence from two experiments that show that this is indeed the case. These findings
are in accordance with other studies showing children's early sensitivity to central
grammatical structures of the target language well before they produce them
systematically (e.g. Gerken 2001; Gerken & McIntosh 1993; Golinkoff, Hirsh-
Pasek, & Schweisguth, 2001; Katz, Baker, & Macnamara 1974; Santelmann &
Jusczyk 1998; Shipley, Smith, & Gleitman 1969).
3.3.1 Experimental Evidence for Early Parameter Setting
Experiment 1. Evidence that German children are indeed sensitive to syntactic non-
root regularities from very early on comes from an experimental study with 26
children aged 2;7 - 6;2 years using a sentence repetition task. Of these subjects, 12
children, up to age 4;0, bear directly on the issue of early sensitivity to information
contained in embedded clauses. That is they are contained in the age bracket in
which the complementizer dass "that" is normally not used spontaneously. The test
material consisted of 64 sentences, 32 grammatical, 32 ungrammatical, half of them
containing an embedded clause introduced by a complementizer and half of them
not. Half of the embedded clauses where transitive, half of them intransitive (see
Table 1).
Table 1. Test Sentences Used in Experiment 1
Sentence transitive (n=32)

type
grammatical (n=16) ungrammatical (n=16)
+COMP Bert sagt, dass Lisa Oma hilft Bert *Oma hilft Bert sagt, dass Lisa hilft
(n=8) says that Lisa grandmother helps Oma
Bert says that Lisa helps grandmother
-COMP Bert sagt, Lisa hilft Oma * Bert sagt, Lisa Oma hilft
(n=8) Bert says Lisa helps grandmother Bert says Lisa grandmother helps
sentence type Intransitive (n=32)
grammatical (n=16) ungrammatical (n=16)
+COMP Bert sagt, daB Lisa gem hilft *Bert sagt, daB Lisa hilft gem
(n=8) Bert says that Lisa willingly helps Bert says that Lisa helps willingly
-COMP Bert sagt, Lisa hilft gem *Bert sagt, Lisa gem hilft
(n=8) Bert says Lisa helps willingly Bert says Lisa willingly helps
The matrix verb was always sagen which can take either a verb-second complement
without a complementizer or a verb-end complement with a complementizer. The
different versions of the embedded clause were chosen in order to offer the child
other possibilities than those involving the complementizer to correct the word-order
violations. Examples of the different types of possible corrections are shown in
Sentences 25:
(25) Correction types:
Ungrammatical test sentence:

*Bert sagt, Lisa Oma/ gern hilft.
Bert says Lisa grandmother/ willingly helps
Possible corrections:
a. Complementizer insertion:
Bert sagt, dass Lisa Oma/ gern hilft.
Bert says that Lisa grandmother/ willingly helps
b. Movement of the tinal verb into the second position:
Bert sagt, Lisa hilft Oma/ gern.
Bert says Lisa helps grandmother/ willingly
c. Adverb/ object deletion:
Bert sagt, Lisa hilft
Bert says Lisa helps
Ungrammatical test sentence:

*Bert sagt, dass Lisa hilft Oma.
Bert says that Lisa helps grandmother
94 J. WEISSENBORN
Possible corrections:
a. Complementizer deletion:
Bert sagt, Lisa hilft Oma/ gem.
Bert says Lisa helps grandmother/ willingly
b. Movement of the verb into the tinal position:
Bert sagt, dass Lisa Oma/ gem hilft.
Bert says that Lisa grandmother/ willingly helps
c. Adverb/ object deletion:
Bert sagt, dass Lisa hilft.
Bert says that Lisa helps
In the experiment two puppets were used, Bert from Sesame Street and a puppet
called Lisa. The children were that Bert would say something about Lisa and that
they should repeat what he said. The test sentences were presented to the children
over a tape recorder with a loud speaker.
The results were the following: Figure 1 shows that overall the sentences are less
often literally repeated the younger the children get. This effect is stronger for the
ungrammatical sentences then for the grammatical ones. This is a first indication
that the children are sensitive to grammaticality violations.
100%
80%
60%
40%
20%
0% -f--'---
6 years 5 years 4 years 2-3 years
10 grarrmatical • ungrarrmatical I
Figure 1. Literal repetitions of grammatical and ungrammatical sentences at different ages in

percent.
Turning to the distribution of the literal repetitions of ungrammatical sentences in

the two conditions, that is, with and without complementizer (see Figures 2a and 2b)
we see that in the two youngest age groups ungrammatical sentences without a
complementizer are repeated literally more often than ungrammatical sentences with
a complementizer, i.e. 62.5% vs. 56.2% in the 4 year olds, and 56.2% vs. 25% in the
2-3 year olds. We will discuss this finding below.
If our assumption is right that even the youngest children who do not yet produce
embedded clauses of the type we tested spontaneously should show adult-like
knowledge, we expect to find that the non-literal responses to ungrammatical
sentences are in fact corrections.
100% 100%
80% 80%
60% 60%
40% 40%
20% 20%
0% 0%
+ compo
"-3 years
. compo + It,oJ:r1p. . compo
4 years
10 grarrrretical • ungrarrrretical I
Figure 2a and 2b. Literal Repetitions of grammatical and ungrammatical sentences with and
without complementizer in percent.
As shown in Figure 3a and 3b this is indeed the case: The 4 year old children
correct 33,3% of the ungrammatical sentences, the 2-3 year old children almost
50%. This indicates that they must know about the relation between the presence
and the absence of a complementizer and the position of the finite verb.
100%
80%
60%
40%
20%
4 years 2·3 years

C literal resp. [] corrections Cliteral resp. Ccorrections
Dungrammatical .other l!Iungrammatical .other
Figure 3a and 3b. Types of responses (in percent) to ungrammatical test sentences in 4 and 2-
3 yearolds.
The analysis of the corrections reveals (see Figure 4) that in both age groups the
children clearly prefer corrections involving the complementizer, i.e.
complementizer insertion or deletion, to corrections which do not, i.e. verb
movement or adverb/ object deletion: 61.5% vs. 38.6% in the 4;0 olds, and 70.6%
vs. 28.4% in the 2;0-3;0 year oIds.
These findings suggest that as predicted even the youngest children must already
have detailed knowledge about the structure of the embedded clause, i.e. about the
complementizer and its effect on word order, i.e. verb placement.
96 J. WEISSENBORN
80% • deletion of corrp .

• insertion of corrp.
60%
o verb movement
40%
20%
0%
4 years 2-3 years
Figure 4. Types of corrections of ungrammatical test sentences (in percent) in 4 and 2-3 year
oids.
Experiment 2. In a follow-up experiment we wanted to answer the question whether

we could show that even younger children are sensitive to the specific
complementizer related word-order constraints in German. The method used is the
headturn preference paradigm (for a description of method see Jusczyk, 1997). In
order to make sure that children could not rely on prosodic cues to decide between
grammatical and ungrammatical sentences we run a test with low-pass filtered test-
items with adults. We asked them to decide which one of a pair of
grammatical/ungrammatical test-items sounded more normal to them. The results
(see Figure 5) indicate that their choice was random:
100,0% -r-----------------------,
50,0% +----
0,0% - l - - - -
with camp. without camp.
Figure 5. Adult preference for low-pass filtered ungrammatical sentences with and without
complementizers in percent.
This means that adult speakers do not detect any prosodic, grammaticality related
information which could influence their choice in favor of the prosodic structures
based on the grammatical test sentences.
17 children with a mean age of about 20 have been tested. The results presented
in the following are preliminary in the sense that the number of subjects is still to
low in order to consider our findings as final.
Only found a small but not significant preference in the 20 months old for the
grammatical sentences was found. This could mean that children at this age do not
yet detect any structural difference between the test items which indicates that
parameter setting has not yet taken place. A closer analysis of the data showed that
the children had a significant preference for sentences without complementizers over
sentences with complementizers (see Figure 6).
11000
10500
0------ _________ -+-with camp.
10000
9500
-Q-without
9000
~comp.
8500
8000
7500+---------------4---------------~
grammatical ungrammatical
Figure 6. Attention time in ms to grammatical and ungrammatical sentences with and without
complementizers in 20 month old children (mean age).
If these results turn out to be robust they would show that children of about 20
months of age are sensitive to the presence or the absence of the complementizer,
i.e. that they have identified it, and that the sentences with and without
complementizers must be different for them at some formal level of representation.
This would then constitute the first step in the process of parameter setting: That is
assuming further that the children have identified the verb which should be not
controversial given their productive use of verbs around 20 months they can now
engage into the distributional learning process which should eventually lead to the
recognition of the complementary distribution of the complementizer in embedded
clauses as a prerequisite to draw on the root/ non-root opposition for setting the V2-
parameter.
From a syntactic point of view we can say that our preliminary finding reflects
the problems children have to figure out the properties of the German
complementizer system (e.g. INFL-in-COMP). Thus initially they prefer structures
which can be analyzed as an IP-Projection not involving a CP-projection (26):
(26) Preferred Unpreferred
a. (IP Lisa hilft Oma) vs. (cp dass (IP Lisa hilft Oma»
Lisa helps grandma that Lisa helps grandma
b. (IPLisa Oma hilft) vs. (cp dass (IP Lisa Oma hilft»
Lisa grandma helps that Lisa grandma helps
Notice that the preferred structures are exactly those which children produce around
age 2;00 (e.g. Muller & Penner 1996; Penner, 1994b; Weissenborn, 1990, 1994).
Notice further that the preferred structure in Sentence 26b. is identical with the
complementizerless ungrammatical sentence type in experiment 1 using the sentence
repetition task 'Bert sagt, Lisa Oma hilft which was repeated most often literally in
the youngest age group. As has been argued by various authors (e.g. Clahsen et
al.,1996; Hyams, 1996; Penner, 1994b; Weissenborn, 1990) these structures may
reflect an underspecification of the properties, i.e. the feature content of COMP in
the grammar of the children at this point in development.
98 1. WEISSENBORN
Our findings about children's sensitivity to word-order violations in embedded

clauses are clearly not compatible with the assumption that the child initially has
only access to information contained in matrix clauses as proposed by the Degree-O-
Learnability Hypothesis (Lightfoot, 1994).8
Furthermore these findings clearly show that in order to evaluate children's early
syntactic knowledge we cannot exclusively rely on production data but we have to
look closer at children's perceptive linguistic capacities (see also Gerken, Landau, &
Remez, 1990; McKee, 1994)
4. CONCLUSION
As mentioned, for almost all of the "deviating" child utterances adult counterparts
have been observed. Assuming that identical child and adult utterances have the
same underlying grammatical representation, something else than grammatical
differences must be responsible for one of the main discrepancies between the
child's and the adult's linguistic behavior, i.e. the much higher frequency of
occurrence of these "deviating" structures in the child.
As we have seen above in the discussion of the developmental data, the
utterances at issue are all incomplete in some respect compared to the corresponding
full-fledged utterances while obeying well-formedness constraints like the ones
postulated in Penner and Weissenborn (1996) and Weissenborn (1994).
"Incomplete" in this context means "realizing less structure" than the corresponding
complete utterance. What we can see in young children who are at the beginning of
the language acquisition process is that they prefer to use utterances which require
less structure, like infinitival clauses as opposed to the corresponding finite clauses,
even when they are already able to produce the more complex ones.
How can we explain this preference? We would like to propose that the child's
language production procedures, especially the ones which realize the connection
between the grammar and the production system are initially less automatic than the
ones of the adult, and consequently are cognitively more costly. Evidence that
children's production routines are indeed less automatic comes from recent work on
children's speech errors. In a study on the speech errors of a child between age 2;6
and 3;6, Hohenberger and Leuninger (1997) showed that this child corrected her
speech errors significantly more often than adults do. This finding can be explained
by assuming that the child's less automatic production procedures are more easily
accessible to corrective monitoring than the highly automatic ones of adults. The
gradual decrease in children's use of structures like infinitival clauses and the
corresponding increase of finite clauses can now be accounted for as an increase in
automatization of the production procedures for finite clauses, i.e. an increase in the
availability of the structure building operations which are involved like for example
head movement. The increasing automatization of production procedures diminishes
the pressure for minimizing the output structure (see Weissenborn (1994) for a
proposal along these lines; recently Phillips (1995) and Platzack (1996) have made
similar suggestions).
If our proposal is correct we would expect that even adults under particular
production constraints which inhibit the automatic functioning of certain production
routines should fall back on less costly production procedures which obey the same
well-formedness constraints as the child's productions. Evidence that this is the case
comes from the speech production of aphasics. Thus in agrammatics we find the
same infinitival clauses as in children's language (see Kolk & Heeschen, 1992;
Weissenborn, 1994).
5. ACKNOWLEDGEMENTS
The experimental research reported in this study was partially supported by a grant
to Jiirgen Weissenborn from the Berlin-Brandenburg Academy of Science in the
framework of the working group "Rule learning and rule knowledge in biological
systems" and by the Deutsche Forschungsgemeinschaft (DFG) in the framework of
the Innovationskolleg "Formal models of cognitive complexity". We want to thank
the children and teachers of the various schools which participated in our studies,
and our students who collected the data. Special thanks go to Barbara Hohle with
who most of this research has been carried out.
6. AFFILIATIONS
Dr. Jiirgen Weissenborn

Postfach 601553
D-14415 Potsdam
e-mail: weissenb@rz.uni-potsdam.de
7. NOTES
I More specifically, in case of the adoption of one of the recent proposals for "split" representations for IP
and CP, in the head of one of their off-springs like AGRO, TO for 10 (Pollock, 1989) or like TOPO,
WHO for CO (Miiller & Sternefeld, 1992) or ForceP, TopP, FinP (Rizzi, 1994).
2 But notice that the final position of the finite verb in embedded clauses cannot follow automatically
from the V2 property as seen in Yiddisch which has V2 also in embedded clauses introduced by· a
complementizer.
J Hopefully this situation will improve soon with the addition of several longitudinal corpora to the
CHILDES data base, and the collection of new longitudinal data at the Max-Planck-Institute for
Evolutionary Anthropology in Leipzig under the direction of Michael Tomasello.
4 If not otherwise stated, the data are from the Simone Corpus (see Miller, 1976) (S), and from the
Caroline Corpus (CHILDES) (C).
~ Notice that this definition of the Strong Continuity Hypothesis is not identical with the same notion as
used by other authors under which the child's representations need not obey the parametric values of
the target (e.g. Whitman, 1994, Whitman, Lee & Lust, 1991).
6 (18) is not a possible adult structure. It obeys the LWC in that it has to be considered as an intermediary
step to the possible structure "Mone des auch holen"
7 At this point we ignore potentially conflicting input coming from bridge verbs
(sagen "say", denken "think", etc.) which allow embedding of V2 as shown below:
a. Er sagt, er hat den Mann gesehen.
he says he has the man seen
"He says he has seen the man."
b. Er sagt, dass er den Mann gesehen hat.
he says that he the man seen has
"He says that he has seen the man."
K This does not mean that we claim that children analyze syntactic structures containing embedded clauses
in the same way as adults do.
100 1. WEISSENBORN
8. REFERENCES
Behrens, H. (1993). Temporal reference in German child language. Unpublished doctoral dissertation,
Amsterdam.
Bennis, H., den Dikken, M., Jordens, P., Powers, S., & Weissenborn, 1. (1995). Picking up particles. In D.
Maclaughlin, & S. McEwen (Eds.), Proceedings of the 19th annual Boston University conference on
language development (pp.70-81). Boston: Cascadilla Press.
Bierwisch, M. (1963). Grammatik des deutschen Verbs. (Studia Grammatica II). Berlin: Akademie
Verlag.
Bloom, P. (1990). Syntactic distinctions in child language. Journal of Child Language, 17, 343-355.
Borer, H., & Wexler, K. (1987). The maturation of syntax. In T. Roeper, & E. Williams (Eds.),
Parameter setting (pp.123-172). Dordecht: Reidel.
Boser, K., Lust, B., Santelmann, I., & Whitman, 1. (1992). The syntax of CP and V2 in Early Child
German (ECG): The strong continuity hypothesis. NELS, 22, 51-66.
Chomsky, N. (1986). Knowledge of language. Its nature, origin, and use. New York: Praeger.
Chomsky, N. (1993). A minimalist program for linguistic theory. In K. Hale, & S. J. Keyser (Eds.), The
view from building 20 (pp.1-52). Cambridge, MA: MIT Press.
Chomsky, N. (1995). The minimalist program. Cambridge, MA: The MIT Press.
Clahsen, H. (1982). Spracherwerb in der Kindheit. Eine Untersuchung zur Entwicklung der Syntax bei
Kleinkindern. Tiibingen: Narr.
Clahsen, H. (1991). Constraints on parameter setting. A grammatical analysis of some acquisition stages
in German child language. Language Acquisition, 1, 361-391.
Clahsen, H., Eisenbeiss, S., & Vainikka, A. (1994). The seeds of structure. A syntactic analysis of the
acquisition of case marking. In T. Hoekstra, & 8. Schwartz (Eds.), Language acquisition studies in
generative grammar (pp. 85-119). Amsterdam: John Benjamins.
Clahsen, H., Eisenbeiss, S., & Penke, M. (1996). Lexical learning in early syntactic development. In H.
Clahsen (Ed.), Generative perspectives on language acquisition (pp. 129-160). Amsterdam: John
Benjamins.
Crisma, P. (1992). On the acquisition of WH-questions in French. Geneva Generative Papers, 0(1-2),
115-121.
Den Besten, H. (1989). Studies in westgermanic syntax. Unpublished PhD Dissertation. University of
Tilburg, Germany.
Deprez, V., & Pierce, A. (1993). Negation and functional projections in early grammar. Linguistic
Inquiry, 24, 25-67.
Emonds, 1. (1976). A transformational account to English syntax. New-York: Academic Press.
Felix, S. (1980). Cognition and language development: A German child's acquisition of question words.
In D. Nehls (Ed.), Studies in language acquisition (pp. 91-109). Berlin: Julius Gross.
Felix, S. (1984). Maturational aspects of universal grammar. In A. Davis, C. Criper, & A. Howatt (Eds.),
Interlanguage (pp.133-161). Edinburgh: Edinburgh University Press.
Gawlitzek-Maiwald I., Tracy, R., & Fritzenschaft, A. (1992). Language acquisition and competing
linguistic representations: The child as arbiter. In 1. Meisel (Ed.), The acquisition of verb placement
(pp.l39-180). Dordrecht: K1uwer.
Gerken, L. (2001). Signal to syntax: Building a bridge. In 1. Weissenborn, & 8. Hiihle (Eds),Approaches
to bootstrapping; Phonological, lexical, syntactic, and neurophysiological aspects of early language
acquisition (pp. 147-165). Amsterdam: John Benjamins Publishing Company.
Gerken, L. A., Landau, 8., & Remez, R. (1990). Function morphemes in young children's speech
perception and production. Developmental Psychology, 25, 204-216.
Gerken, L. A., & McIntosh, 8. 1. (1993). Interplay of function morphemes and prosody in early language.
Developmental Psychology, 29, 448-457.
Golinkoff, R., Hirsh-Pasek, K., & Schweisguth, M (2001). A reappraisal of young children's knowledge
of grammatical morphemes. In 1. Weissenborn, & B. Hiihle (Eds), Approaches to bootstrapping;
Phonological, lexical, syntactic, and neurophysiological aspects of early language acquisition (pp.
167-187). Amsterdam: John Benjamins Publishing Company.
Grimshaw, 1. (1997). Projection, heads and optimality. Linguistic Inquiry, 28, 373-422.
Hamann, c., Penner, Z., & Lindner, K. (1999). German impaired grammar: The clause structure revisited.
Language Acquisition, 7, 193-246.
Hiihle, 8., Weissenborn, 1., Schmitz, M., & Ischebeck, A. (2001). Discovering word order regularities:
The role of prosodic information for early parameter setting. In 1. Weissenborn, & B. Hahle (Eds.),
Approaches to bootstrapping: Phonological, lexical, syntactic, and neurophysiological aspects of
early language acquisition (pp. 251-267). Amsterdam: John Benjamins.

Hohenberger, A., & Leuninger, H. (1997). It was the hummingbird and not the titmouse. Knowledge,
lexicon, processing: Evidence from child speech errors. Ms. University of Frankfurt.
Hyams, N. (1986). Language acquisition and the role of parameters. Dordrecht: Foris.
Hyams, N. (1996). The underspecification of functional categories in early grammar. In H. Clahsen (Ed.),
Generative perspectives on language acquisition (pp 91-128). Amsterdam: John Benjamins.
Jusczyk, P. (1997). The discovery of spoken language. Cambridge, Mass.: The MIT Press.
Kaltenbacher, E. (1990). Strategien beim friihkindlichen Syntaxerwerb. Tiibingen: Narr.
Katz, B., Baker, G., & Macnamara, 1. (1974). What's in a name? A study of how children learn common
and proper names. Child Development, 45, 269-273.
Kayne, R. (1993). The antisymmetry of syntax. Cambridge, MA: The MIT Press.
Klein, W. (1998). Assertion and finiteness. In N. Dittmar, & Z. Penner (Eds.), Issues in the theory of
language acquisition. Essays in honor ofliirgen Weissenborn (pp. 225-245). Bern: Peter Lang.
Kolk, H., & Heeschen, C. (1992). Agrammatism, paragrammatism and the management of language.
Language and Cognitive Processes, 7,89-129.
Lasser, I. (1997). Finiteness in adult and child German (MPI Series in Psycholinguistics 8), Nijmegen:
MPI for Psycholinguistics.
Lebeaux, D. (2000). Language acquisition and the form of the grammar. Philadelphia: lohn Benjamins.
Lightfoot, D. (1994). Degree-O learnability In B. Lust, G. Hermon, & J. Kornfilt (Eds.), Syntactic theory
and language acquisition: Crosslinguistic perspectives. Vol. 2: Binding, dependencies, and
learnability (pp. 453-472). Hillsdale, NJ: Lawrence Erlbaum.
Lust, B., Whitman, 1., & Kornfilt, J. (Eds.). (1994). Syntactic theory and first language acquisition:
Crosslinguistic perspectives. YoU: Heads, projections, and learnability. Hillsdale, N.l.: Lawrence
Erlbaum.
McKee, C. (1994). What you see isn't always what you get. In B. Lust, B. M. Sufier, & J. Whitman
(Eds.), Syntactic theory and language acquisition: Crosslinguistic pespectives. YoU: Heads,
projections, and learnability (pp. 201-212). Hillsdale, N.J.: Lawrence Erlbaum.
Meisel, 1. M. (Ed.) (1992). The acquisition of verb placement: Functional categories and V2 phenomena
in language development. (Studies in theoretical psycholinguistics 16). Dordrecht: K1uwer Academic
Publishers.
Meisel, J. M., & Miiller, N. (1992). Finiteness and verb placement in early child grammar: Evidence from
simultaneous acquisition of French and German in bilinguals. In 1. M. Meisel (Ed.), The acquisition
of verb placement: Functional categories and V2 phenomena in language development (pp. 109-
138). Dordrecht: K1uwer Academic Publishers.
Miller, M. (1976). Zur Logik der friihkindlichen Sprachentwicklung. Stuttgart: Klett.
Mills, A. (1985). The acquisition of German. In D. Siobin (Ed.), The crosslinguistic study of language
acquisition (pp. 141-245). Hillsdale, NJ: Lawrence Erlbaum.
Miiller, N., & Penner, Z. (1996). Early subordination: the acquisition of free morphology in French,
German, and Swiss German. Linguistics, 34, 133-165.
Miiller, G., & Sternefeld, W. (1992). Improper movement and unambiguous binding. Linguistic Inquiry,
24,461-507.
Penner, Z. (1994a). Asking questions without CPs? On the acquisition of wh-questions in Bernese Swiss
German and Standard German. In T. Hoekstra, & B. Schwartz (Eds.), Language acquisition studies
in generative grammar (pp. 177-214). Amsterdam: lohn Benjamins.
Penner, Z. (1994b). Ordered parameter setting in first language acquisition. Unpublished Habilitation
Thesis. Bern: Universitiit Bern.
Penner, Z., Tracy, R., & Weissenborn, J. (2000). The acquisition of focus particles and negation. In S.
Powers, & c. Hamann (Eds.), The acquisition of scrambling and cliticization (pp. 127-164).
Dordrecht: K1uwer.
Penner, Z., & Weissenborn, J. (1996). Strong continuity, parameter setting and the trigger hierarchy. On
the acquisition of the DP in Bernese Swiss German and High German. In H. Clahsen (Ed.),
Generative perspectives on language acquisition: Empirical findings, theoretical considerations,
crosslinguistic comparisons (pp. 161-200). Amsterdam: John Benjamins.
Penner, Z., Wymann, K., & Dietz, C. (1998). From verbal particles to complex object-verb constructions
in early German. The acquisition of event structure in normally developing and language impaired
children. (Working Paper 89, Sonderforschungsbereich 471). Konstanz: University of Konstanz.
Penner, Z., Weissenborn, J., & Wymann, K. (2001). On the prosody/ lexicon interface in learning word
order. A study of normally developing and language impaired children. In 1. Weissenborn, & B.
Hiihle (Eds.), Approaches to bootstrapping: Phonological, lexical, syntactic, and neurophysiological
aspects of early language acquisition (pp. 269-294). Amsterdam: John Benjamins.
102 J. WEISSENBORN
Pesetsky, D. (1993). Some possible alternative explanations for early language phenomena. Paper
presented at the SISSA workshop, Trieste, July 1993.
Phillips, C. (1995). Syntax at age two: cross-linguistic differences. MIT Working Papers in Linguistics,
26, 325-382.
Pinker, S. (1984). Language learnability and language development. Cambridge, MA: Harvard
University Press.
Platzack, C. (1996). The initial hypothesis of syntax: A minimalist perspective on language acquisition
and attrition. In H. Clahsen (Ed.), Generative perspectives on language acquisition: Empirical
findings, theoretical considerations, crosslinguistic comparisons (pp. 369-414). Amsterdam: John
Benjamins.
Poeppel, D., & Wexler, K. (1993). The full competence hypothesis of clause structure in Early German.
Language, 69, 1-33.
Pollock, J.-Y. (1989). Verb movement, UG and the structure of IP. Linguistic Inquiry, 20, 365-424.
Reis, M. (1985). Satzeinleitende Strukturen im Deutschen. Uber COMP, Haupt- und Nebenslitze, W-
Bewegung und die Doppelkopfanalyse. In W. Abraham (Ed.), Erkliirende Syntax des Deutschen
(pp.271-31l). Tiibingen: Narr.
Rizzi, L. (1992). Early null subjects and root null subjects. Geneva Generative Papers, 1,102-114.
Rizzi, L. (1994). Some notes on linguistic theory and language development: The case of root infinitives,
Language Acquisition, 3, 371-393.
Roeper, T. (1973). Approaches to a theory of language acquisition with examples from German children.
Unpublished PhD Dissertation. Cambridge, MA: Harvard University.
Roeper. T. (1992). From the initial state to V2: Acquisition principles in action. In J. Meisel (Ed.), The
acquisition of verb placement: Functional categories and V2 phenomena in language development
(pp. 333-370). Dordrecht: Kluwer.
Roeper, T. (1996). The role of merger theory and formal features in acquisition. In H. Clahsen (Ed.),
Generative Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations,
Crosslinguistic Comparisons (pp. 415-450). Amsterdam: John Benjamins.
Roeper, T., & deVilliers, J., (1992). Ordered decisions in the acquisition of Wh-questions. In J.
Weissenborn, H. Goodluck, & T. Roeper, (Eds.), Theoretical issues in language acquisition (pp. 191-
236). Hillsdale, NJ.:Lawrence Erlbaum.
Roeper, T., & Weissenborn, 1. (1990). How to make parameters work. In L.Frazier, & J. de Villiers (Eds.),
Language processing and language acquisition (pp. 147-162). Dordrecht: Kluwer.
Rohrbacher, 8., & A. Vainikka (1995). On German verb syntax under age 2. In D. McLaughlin, & S.
McEwen (Eds.), Proceedings of the 19th annual Boston University conference on language
development (pp. 487-498). Somerville: Cascadilla Press.
Rothweiler, M. (1993). Der Erwerb von Nebensiitzen im Deutschen. Tiibingen: Niemeyer.
Santelmann, L., & Jusczyk, P. (1998). 18-months-old's sensitivity to relationships between morphemes:
interactions among relationships, frequency and processing. In A. Greenhill, M. Hughes, H.
Littlefield, & H. Walsh (Eds), Proceedings of the 22 annual Boston University conference on
language development, (Vol.2, pp. 663-674). Somerville, MA: Cascadilla Press.
Schaner-Wolles, C. (1994). Intermodular synchronization: On the role of morphology in the normal and
impaired acquisition of a verb-second language. In R. Tracy, & E. Lattey (Eds.), How tolerant is
universal grammar? Essays on language learnablity and language variation (pp. 205-224).
Tiibingen: Niemeyer.
Schiinenberger, M., Penner, Z., & Weissenborn, J. (1997). Object placement and early German grammar.
In E. Hughes, M. Hughes, & A. Greenhill (Eds.), Proceedings of the 21st annual Boston University
conference on language development, Vol. 2 (pp. 539-549). Somerville, MA: Cascadilla Press.
Schwartz, 8., & Vikner, S. (1996). The verb always leaves IP in V2 clauses. In A. Belletti, & L. Rizzi
(Eds.), Parameters and functional heads (pp.II-62). Oxford: Oxford University Press.
Shipley, E., Smith, c., & Gleitman, L. (1969). A study in the acquisition of language: Free responses to
commands. Language, 45, 322-342.
Stern, c., & Stern, W. (1928). Die Kindersprache. Reprint. Darmstadt: Wissenschaftliche
Buchgesellschaft.
Tracy, R. (1991). Sprachliche Strukturentwicklung. Linguistische und kognitionspsychologische Aspekte
einer Theorie des Erstspracherwerbs. Tiibingen: Narr.
Travis, L. (1984). Parameters and effects of word variation. Unpublished PhD Dissertation. Cambridge,
MA:MIT.
Verrips, M., & Weissenborn, J. (1992). Routes to verb placement in early German and French: The
independence of finiteness and agreement. In M. Meisel (Ed.), The acquisition of verb placement:
functional categories and V2 phenomena in language development (pp. 233-283). Dordrecht:
K1uwer.
Weissenborn, J. (1990). Functional categories and verb movement in early German. The acquisition of
German syntax reconsidered. In M. Rothweiler (Ed.), Spracherwerb und Grammatik. Linguistische
Untersuchungen zum Erwerb von Syntax und Morphologie, Linguistische Berichte, Special Issue 3,
190-224.
Weissenborn, 1. (1992). Null subjects in early grammars: Implications for parameter setting. In 1.
Weissenborn, H. Goodluck, & T. Roeper (Eds.), Theoretical issues in language acquisition (pp. 269-
300). Hillsdale, NJ: Lawrence Erlbaum.
Weissenborn, 1. (1993). Mommy's sock almost fits: Constraints on children's grammars. Plenary address
given at the 6th International congress for the study of child language, Trieste. Nijmegen: Max-
Planck-Institute for Psycholinguistics.
Weissenborn,1. (1994). Constraining the child's grammar: Local wellformedness in the development of
verb movement in German and French. In B. Lust, M. Suner, & 1. Whitman (Eds.), Syntactic theory
and language acquisition: Crosslinguistic pespectives, Vol. 1: Heads, projections, and learnability
(pp. 215-247). Hillsdale, NJ: Lawrence Erlbaum.
Weissenborn, J., Roeper, T., & de Villiers, 1. (1995). WH-acquisition in French and German: Connections
between case, wh-features and uriique triggers. In C. Jakubowicz (Ed.), Grammaire Universelle et
Acquisition du language (Recherches Iinguistiques de Vincennes 25) (pp. 125-155). Paris: Press
Universitaires de France.
Wexler, K (1994). Optional infinitives. In D. Lightfoot, & N. Hornstein (Eds.), Verb movement (pp. 305-
350). Cambridge: Cambridge University Press.
Whitman, J. (1994). A strong continuity account of the acquisition of verb-second. In B. Lust, 1.
Whitman, & J. Kornfilt (Eds.), Syntactic theory and language acquisition: Crosslinguistic
perspectives. Vol. 1: Heads, projections, and learnability (pp. 273-288). Hillsdale, NJ: Lawrence
Erlbaum.
Whitman, J., Lee K-O., & Lust, B. (1991). Continuity of the principles of universal grammar in first
language acquisition: The issue of functional categories. NELS, 21, 383-397.
Zwart, 1. (1997). Morphosyntax of verb movement. A minimalist approach to the syntax of Dutch.
Dordrecht: K1uwer.
SUSAN M. POWERS
MERGE AS A BASIC MECHANISM OF LANGUAGE:

EVIDENCE FROM LANGUAGE ACQUISITION
Abstract. One of the most basic skills of a proficient language user is the ability to combine smaller units
(e.g., words) into larger units (e.g., phrases). Therefore, a mechanism that combines such "syntactic"
objects is a good candidate for a basic component of the human computational system. Chomsky (1995)
defines a mechanism called Merge that concatenates two and only two syntactic objects (e.g.,
morphemes, words) in each application. This paper presents evidence from children's earliest productions
that Merge operates on three levels in child grammar. First, Merge operates on the word-level, fusing two
words into single lexical items. Merge also operates on the sub-word level providing a straightforward
account of the so-called "mixed" productions of bilingual children. In addition, Merge combines words
into phrases. The fact that this mechanism only concatenates two units in any single application yields an
elegant account of the observed initial period in language development in which only two-word
combinations are attested (Bowerman, 1990; Brown, 1973). Language acquisition data thus reveal a
simple concatenation operation like Merge to be one of the most basic mechanisms of language
production.
1. INTRODUCTION
In this paper, I claim that language acquisition data reveal that the simple binary
concatenation operation called Merge (as formulated by Chomsky, 1995) is a basic
function or mechanism of language. Numerous examples from children's earliest
productions show that Merge yields an interesting and completely new analysis of
children's novel word combinations. Data from monolingual children acquiring
English, Dutch, and German, and from German-French and German-English
bilinguals are presented and analyzed in terms of the operation Merge.
2. THE OPERATION MERGE
Why is an operation that concatenates syntactic objects (e.g., morphemes, words) a

necessary component of human language? Chomsky (1995, p. 226) claims that such
a mechanism is needed because a linguistic derivation (or expression) which has
reached a certain stage
can be interpreted only if it consists of a single syntactic object. Clearly then, C IIL [the
human computational system I must incl ude a second procedure that combines syntactic
objects already formed.
He defines Merge as the simplest operation which

takes a pair of syntactic objects (SO;, SO;) and replaces them by a new combined
syntactic object SOij

106 SUSAN M. POWERS
That is, when two syntactic objects (e.g., words) combine via Merge, the individual
components are no longer visible to the computational system. There is only the new
syntactic object created by Merge. This is illustrated in (1) below in which the
syntactic objects A and B are Merged together into the new syntactic object K.
(1) {A,B} -- Merge K
Though K is composed of A and B, only the syntactic object called K will enter into
further computations of the language faculty.
The introduction of the operation Merge is a radical departure from phrase
structure operations of earlier theories like Government and Binding (Chomsky,
1981). For example, X-bar structures like (2) below are comprised of three
positions: specifier, head, and complement.
(2) XP
A
specifier X'
~Pl.,"ent
head
Merge structures are strictly binary as two and only two syntactic objects can be
Merged in any single application. Returning to example (1) above, when A and B
combine via Merge, the binary branching structure in (3) below results.
(3) K
A
A B
Another radical departure from X-bar Theory is the label of structures like (3)
above. While the head of an XP like that in (2) is X, Merger structures are
asymmetric. In the words of Chomsky (1995, p. 246):
The operation Merge (A, 8) is asymmetric, projecting either A or 8, the head of the object
that projects becoming the label of the complex formed.
That is, a structure like (3) could either be headed by A as in (4a) or B as in (4b).
THE MERGE MECHANISM: EVIDENCE FROM LANUAGE ACQUISITION 107
(4a) A (4b) B
A B A B
In this way, Merge allows the head of each structure to be uniquely defined.
3. MERGE ON THREE LEVELS
3.1 Word Level Merge
Consider the examples in (5-7) below where two words have been fused into single
lexical items. 1
(5) a. can-I child English

b. awa (I-want)
c. I-can't
d. could-ya
e. would-ya
f. where-s
g. what-a(re)
(6) a. kann-ich (can-I) child German

b. hab-ich (have-I)
c. wo-(i)s(t) (where-is)
d. will-nicht (want-not)
(7) a. ik-ook (I-want) (cf. me too in adult) child Dutch

b. kan-nie (can-not),
c. mag-nie (may-not)
The first thing to note about these constructions is that the same types of
combinations appear across different child languages. Secondly, the individual
components of these amalgams do not concurrently appear in isolation. As Braine
(1976, p. 46), who noted example (Sa), importantly points out "The evidence
indicates that I has no independent morphemic status in this formula". A third
intriguing feature of these combinations is that the order of components is fixed
even when the combination is novel in meaning. For example, consider example
(6d) from Stern and Stern (1928). In this example, the child uses the combination
will-nicht "want-not" to refer to an object that he does not desire to have. Less
transparent uses of such amalgams come from child Dutch. For example, van
Kampen (p.c.) finds that one child uses "ik ook" which in adult Dutch means "I
also" or "me too" to consistently mean "I want" at a particular stage of language
development. This combination is identical in meaning to awa reported by
Bowerman (1990) in (5b).
The most interesting feature of these amalgams is that all are sequences of two
functional elements. That is, all morphemes involved would occupy adjacent
108 SUSAN M. POWERS
functional phrase structure positions in the adult grammar. This is shown in Table 1
below.
Table 1. FunctionalAmalgams
Child Child Child

English German Dutch
auxiliary + subject can-I kann-ich
CO Spec,IP
subject + auxiliary I-can't
Spec,IP 1°
wh + auxiliary where'(i)s wo'(i)s(t)
Spec, CP CO
auxiliary + neg will-nicht kannie
1° NegP
These data are compatible with postulating Merge structures in which there is
initially only one position or slot for functional elements in the phrase marker as
proposed by Powers (1998b). This analysis contrasts with building phrase structure
via successive applications of the X-bar module, which provides two phrase
structure positions per functional projection. For example, consider the proposed
transition in language development (Guilfoyle & Noonan, 1992; Radford, 1990)
between the VP stage and the IP stage depicted in (8) below.
(8) VP IP
A A
Spec V' Spec I'
A A
VP
A
Spec v'
A
yO
When the IP layer develops, both the specifier of IP [Spec, IP] and the head position
1°' enter the structure. Powers (1998b) notes a phenomenon similar to the Doubly-
filled Comp Filter of adult grammars. In early child English, elements that are
hypothesized to occupy the head position or the specifier position of functional
projections are attested but interestingly, do not occur together. That is, either a
nominative pronominal subject occurs without an accompanying auxiliary verb or an
auxiliary verb occurs without a subject. These data, and the functional amalgam data
in (5-7), can be accounted for if phrase structure builds via successive applications
of the operation Merge. Since phrase structure is built-up binarily, there is initially
only a single position for any functional morpheme as shown in (9). This single slot
can either host single functional morphemes like nominative subjects or auxiliary
verbs. This is also the position occupied by functional amalgams of nominative
subjects and auxiliary verbs like I-can't.
(9) Precursor-IP
VP
These amalgams, which typically occur much earlier in development than either
component thereof occurs, reflect the binary nature of the mechanism of Merge in
two ways. First, by their internal composition and second by their position in the
overall structure.
3.2 Sub-Word Level Merge

In addition to the functional amalgams produced by monolingual learners, bilingual
children also produce novel combinations of morphemes such as those in (10) below
from Tracy (1998) and in those in (11) from Koppe (1996).
(10) English-German Mixel

a. cleanST DU DEIN teeth? Hanna 2;9
cleans you your teeth
b. ICH HAB GEmade you much better Hanna 2;4
I have past-made you much better
(11) French-German Mixes

a. teddy reSUCHT Ivar 2;4.9
teddy again-seeks
b. ANmis DIES Ivar 2;5.7
on-put this
These combinations differ from the functional amalgams described above. For
example, in each case a functional morpheme is "mixed" together with a lexical
morpheme. Interestingly, the functional vocabulary comes from one language and
the lexical vocabulary from the other. 3 For example, in (lOa) clean and teeth are
English but each of the functional morphemes namely -st, du, and dein are German.
These data show the application of Merge at the sub-word level as it is derivational
morphology like -st, ge-, re-, and the verb particle an- that are Merged together with
nouns and verbs in (10) and (11).
These data, as well as the functional amalgams described in Section 3.1 above,
present a labeling conundrum. For example, in the Merge structure below
110 SUSAN M. POWERS
(12a)
clean - st
perhaps the verb clean should project to label this novel lexical item as it is a verb.
The case is less clear however with functional amalgams like kan-nie in (7b) from
child Dutch (Hoekstra & Jordens, 1994). This lexical item is a negative modal but
the current definition of Merge does not permit both "Mergees" to project. That is,
this lexical item could only be defined by the modal kan or the negator nie(t) not
both.
3.3 Merge on the Phrasal Level
The last level on which there is evidence that Merge applies is the phrasal level.
There is an early sub-phase in the development of combinatorial speech in which
there are strictly two-word combinations (Bowerman, 1990; Brown, 1973).
Examples of these combinations from child English are given in (12-13) below.
(12) a. more car Andrew (Braine, 1976)

b. more cereal Andrew (Braine, 1976)
c. more read Andrew (Braine, 1976)
d. more hot Andrew (Braine, 1976)
e. more sing Andrew (Braine, 1976)
f. more high Andrew (Braine, 1976)
g. more walk Andrew (Braine, 1976)
(13) a. no bed Andrew (Braine, 1976)

b. no home Andrew (Braine, 1976)
c. no mama Andrew (Braine, 1976)
d. no wet Andrew (Braine, 1976)
e. no high Andrew (Braine, 1976)
f. no fix Andrew (Braine, 1976)
g. no plug Andrew (Braine, 1976)
Typical elements appearing in these combinations include no, more, off, this, and
that. These items are both lexically and cross-linguistically specific. For example, in
child English the anaphoric negator no exclusively appears in such combinations to
the exclusion of its semantically similar counterpart not. Similarly, Dutch children
favor the anaphoric negator nee to the negator niet (see Powers, 2001).
Koppe (1996) reports similar two-word combinations from bilingual children.
(14) French-German Bilinguals

a. ENCORE kiise Ivar 1;10.30
more cheese
b. <;A <;A Sonne Ivar 1;11.17
this this sun
c. DAS musique chat Ivar 2;0.22
this music cat
d. da, BATEAU Pascal 1;6.4

there, boat
e.la, AUA Pascal 1;10.28
there hurt
Similar to the mixed utterances above, in each of these examples a noun from one
language combines with a function word from the other. Interestingly, exactly the
lexical items that are frequent in monolingual children's two-word combinations
appear in these mixed utterances as Table 2 (from Koppe, 1996) shows.
Table 2. Typical Elements Involved in Bilingual Mixes
English German French

no nee/nein non
yes ja oui
here da Iii
this dies ~a
more/again noch encore
inside darein/drin dedans
Characterizing these elements in child grammars has proved difficult. Braine (1963)
proposed that these elements comprise a grammatical class called pivots that adult
grammars lack. According to his Theory of Pivot Grammar, the child grammar
consists classes that do not, in any obvious way, overlap with adult grammatical
classes. This non-congruity was fully intentional (Bowerman, 1973, p.c.) as child
grammars were treated as separate systems, independent of adult grammars, perhaps
indefinable in the terms of adult grammar. In the child grammar, a sentence (S) is
formed by combining an initial pivot (P1) and an open class element (0) or by
combining an open class element (0) with a member of P2 (the final pivot class) as
in the sample Pivot Grammar in (16) below.
(16) S- P1 + 0 P1- all, J, no, see, more, hi, other

S- 0+P2 P2- off, by, come, there
The word order of the earliest utterances is explained by these distribution ally
defined pivot classes. For any two-word combination, the pivot will be either initial
or final. Typical pivots include more, no, see, J, and off. There appears to be no
restriction on the grammatical category of a pivot. This lack of grammatical
coherence is due to the fact the pivot and open classes were defined distribution ally
rather than in terms of adult grammatical categories (e.g., noun).
Lebeaux (1988, p. 11) suggests that the relation between Braine's pivot and open
class:
may be thought of as that between governor and governed element, or perhaps more
generally that of head to complement
My claim is that Braine's pivots are indeed heads. They are heads of structures
formed by the mechanism Merge. Unlike the functional heads I or C, which select
unique syntactic complements, the heads no and more appear with a range of
112 SUSAN M. POWERS
syntactic complements (see 13 and 14 above). These elements are also heads in the
sense that they project to define the structure that they appear in. Consider the data
in (l3), which can be accommodated by the abstract Merge structure in (17) below.
(17) more
more
Such tree structures "operate like frames or slots through which a variety of other
forms rotate (Brown, 1973, p. 172)". This is shown by the constructions with more
in (18).
(18) more
more car
cereal
cookie
fish
high
hot
In this structure, the lexical item more projects and thus heads the projection. In this
way, this Merge structure captures a defining property of these elements that were
called operators by Miller and Ervin (1964):
a few high frequency words [which] tended to be restricted to a given position and
tended to define the sentence as a whole.
In order for these elements to "define the sentence as a whole", it is crucial that the
pivot or operator projects. That is, even though these structures are formed by
Merge, they have a pre-defined head namely, the pivot. Though this restriction
appears to violate the operation Merge itself which recall, permits either A or B to
project; exactly this structure occurs when a target phrase marker expands.
According to Epstein, Thniinsson, and Zwart (1996, p. 10):
Two phrase markers are combined by expanding one (the target phrase marker) so as to
include an empty position. This takes place by adding to the target phrase marker a
projection of the target phrase marker. This projection is binary branching and has two
daughters: the target phrase marker and an empty position. The other phrase marker then
substitutes into this position.
While the expansion of target phrase markers via Merge is an operation of the adult
grammar, it is not a visible level in the derivation of a sentence. That is, there are no
intermediate grammatical representations like (17). Children's productions directly
reflect these tree structures in which a closed-class lexical item like more projects
into the syntax providing a frame for open class elements to occupy. It is perhaps
only in the child grammar that phrase markers like (17) exist as a separate level of
representation (Lebeaux, 1988, Powers, 2001).
If the earliest word combinations are indeed Merge structures as hypothesized,

another fact about early language acquisition can be explained. Recall that the first
sub-stage in word combinations consists entirely of two-word combinations. This
binarity, first observed by Brown (1973), is not naturally captured by the GB-based
theories of syntactic development developed in the 1980s and 1990s. Merge
straightforwardly explains this binarity. In fact, binary structures are expected as this
syntactic operation combines two and only two syntactic objects in each application.
4. OLD PUZZLES AND FUTURE DIRECTIONS
Merge is able to explain many of the novel words and novel word combinations that
young children produce. Another set of novel combinations found in child English
contain the pronominal it as in (19) below.
(19) a. here's a flush-it Emily (Hamburger, 1980)

(indicating a toilet handle)
b. here break it David 21 mos. (Braine, 1976)
(indicating a broken object)
c. want fix it David 22 mos. (Braine, 1976)
(indicating broken toy telephone)
d. pull it Peter 22,2 (Bloom, Light-
(touching window-shade pull) bown, & Hood, 1975)
e. screw it Peter 23,2 (Bloom et aI., 1975)
(holds out toy screwdriver)
f. need it Peter 25,0 (Bloom et aI., 1975)
(= part of toy telephone)
In all of these cases, it seems to be equivalent to the generic nominal thing. The flush
it of (19a) is a "flushing-thing" or "a thing for flushing with" namely, a toilet handle.
Merge yields an insightful account of these data. Consider the combination fix it in
(19c). When the child says fix it, he is referring to an object, namely a toy telephone
which is broken and needs to be repaired or "a thing that needs fixing". Since the
entire phrase fIX it is itself a nominal, it should be labeled as a nominal. This could
be accomplished by allowing the nominal part of the combination, namely it, to
project and define the entire structure as in (20) below.
(20) it (NP)
fix
A it
Consider example (19c) want fix it again. Merging want with the already formed
syntactic object fix it in (20) yields the new combined syntactic object in (21).
114 SUSAN M. POWERS
(21) want(... VP)
A
want it (... NP)
fix it
The entire structure is headed by the verb want which occurs with the object fix it.
This would mean something like "I want the fix it". Another possibility under Merge
is that when fix and it are initially Merged, fix not it projects, and defines the lower
part of the tree as a verbal phrase as in (22).
(22) want (... VP)
A
want fix (... VP)
fix it
In this case, this utterance would mean "I want to fix it" as both want and fix are
verbal and it represents the simple object of the verb fix.
By allowing either syntactic object that it concatenates to project, Merge permits
the same word combination to have different meanings. This is a desirable
consequence for language acquisition data as the fact that the same combinations
can have multiple meanings, noted at least since Bloom (1970), has not until now
found an adequate theoretical execution. Children's concurrent use of non-
nominative subjects as in (23b) and proto-relative clauses with it (Hamburger, 1980)
as in (23a) can also be captured by the Merge analysis.
(23 a) my (DP) (23b) did (VP)
my
A it (NP) my
A did (VP)
did
A it
A did it
In (23a), my did it which Hamburger (1980) claims is equivalent to the adult phrase
the thing that I did (see also Powers, 1996, Powers & Musolino, 1997); did and it
are Merged together and it is it which projects as did it is a nominal. The entire
structure is dominated or headed by the determiner-like element my. In (23b), did it
is a predicative phrase defined by the verb did which projects. The incorrectly case-
marked subject my is Merged into the structure but does not project as this is a
clause and thus should be a projection of or headed by the verb (see Powers, 1996;
Powers & Musolino, 1997 for more details of this analysis).
Another set of notorious child English constructions involving it also has an

interesting Merge analysis. Consider the word combinations in (23).
(24) a. fix it that. David 22 months (Braine, 1976)

b. have it egg. 10hnathan 24 months (Braine, 1976)
c. have it milk. 10hnathan 24 months (Braine, 1976)
d. have it fork. 10hnathan 24 months (Braine, 1976)
e. leave-it heel Kendall II (Bowerman, 1973)
f. horse ... see it Kendall 1 (Bowerman, 1973)
g. Mommy fix-it ear Kendall II (Bowerman, 1973)
In all of these constructions, the pronoun it is a clitic on the verb. Keyser and Roeper
(1992) analyzed the pronoun it as a base-generated verbal clitic that is co-indexed
with the object NP as in (25).
(25) V' (from Keyser & Roeper, 1992)
V CI
have it;
There are at least two possible Merge analyses of these combinations. The first, is
basically the clitic analysis as in (26) below.
(26)
have it (NP)
it
A egg
In this structure, it and egg are first merged together in a type of small clause
structure and the resulting is roughly equivalent to the one of the clitic analysis, "I
have it and it's an egg". In the alternative Merge analysis in (27) have and it are
initially Merged as in examples (20 - 23) above and then egg is Merged together
with this syntactic object.
(27)
it (NP) egg
have it
116 SUSAN M. POWERS
Although structurally similar to (25), this rather means "a have-it isa egg". Such
utterances are not surprising as children use verb+it combinations to refer to objects
(Hamburger, 1980; Powers, 1998a) at this stage. Merge offers an alternative analysis
of these intriguing constructions that is consistent with the child's grammar.
5. CONCLUSION
In this paper, I have shown how the simple binary concatenation operation Merge
can account for a range of novel constructions that children produce. While it is
obvious that Merge, an operation of the adult language, is present early on, it is not
clear whether the child's version of Merge is identical to the adult's. In particular, it
seems that children really do allow either conjunct of a Merge structure to project.
Even if it is the "wrong" conjunct i.e., it would cause the derivation to crash in the
adult grammar. Moreover, children produce novel lexical items that seem to
combine elements of both conjuncts. These uses suggest that although Merge can
capture many novel constructions, as heretofore defined, it is not flexible enough to
capture the entire range of variation attested in child grammars.
6. AFFILIATIONS
Dr. Susan M. Powers

Linguistics Department
University of Potsdam
PO Box 601553
D 14415 Potsdam
Germany
e-mail: suzipow@msn.com
7. NOTES
1 Examples 5d-g come from Peters (2001) and 7b-c from Hoekstra and lordens (1994). All the rest of the
data sources are duly noted in the text.
2 The morphemes in capital letters reflect the language that is "mixed in". In this case, the child was
speaking English and mixed in German.
3 These data are consistent with the proposal of Bradley, Garrett, and Zurif (1980) that there are two
lexicons, one containing functional and the other containing substantive lexical items.
8. REFERENCES
Bloom, L. (1970). Language development: Form and function in emerging grammars. Cambridge: MIT
Press.
Bloom, L., Lightbown, P., & Hood, L. (1975). Structure and variation in child language. Monographs of
the Society for Research in Child Development, 160, 40-42.
Bowerman, M. (1973). Early syntactic development: A crosslinguistic study with special reference to
Finnish. London: Cambridge University Press.
Bowerman, M. (1990). Mapping thematic roles onto syntactic functions: Are children helped by innate
linking rules? Linguistics, 28, 6, 1253-1289.
Bradley, D. c., Garrett, M., & Zurif, E. B. (1980). Syntactic deficits in Broca's Aphasia. In D. Caplan
(Ed.), Biological Studies of Mental Processes (pp. 269-287). Cambridge: MIT Press.
Braine, M. (1963). The ontogeny of English phrase structure: The first phase. Language, 39, 1-13.
Braine, M. (1976). Children's first word combinations. With commentary by Melissa Bowerman.
Monographs of the Society for Research in Child Development, 41, 1-97.
Brown, R. (1973). A first language: The early stages. Cambridge: Harvard University Press.
Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris.
Chomsky, N. (1995). The minimalist program. Cambridge: MIT Press.
Epstein, S. D., Thniinsson, H., & Zwart, C. J-W. (1996). Introduction. In W. Abraham, S. Epstein, H.
Thniinsson, & C. J-W Zwart (Eds.), Minimal ideas: Syntactic studies in the minimalist framework
(pp. 1-66). Amsterdam: John Benjamins.
Guilfoyle, E., & Noonan, M. (1992). Functional categories and language acquisition. Canadian Journal
of Linguistics, 37, 2, 241-272.
Hamburger, H. (1980). A deletion ahead of its time. Cognition, 8, 389-416.
Hoekstra, T., & Jordens, P. (1994). From adjunct to head. In T. Hoekstra, & B. Schwartz (Eds.),
Language acquisition studies in generative grammar (pp. 119-149). Amsterdam: John Benjamins.
Keyser, S. J., & Roeper, T. (1992). Re: The abstract c1itic hypothesis. Linguistic Inquiry, 23, 89-125.
Koppe, R. (1996). Language differentiation in bilingual children: The development of grammatical and
pragmatic competence. Linguistics, 34, 927-954.
Lebeaux, D. (1988). Language acquisition and the form of the grammar. Doctoral Dissertation,
University of Massachusetts Amherst.
Miller, W., & Ervin, S. W. (1964). The development of grammar in child language. In U. Bellugi, & R.
Brown (Eds.), The acquisition of language (pp. 9-34). Chicago: The University of Chicago Press.
Peters, A. (2001). From prosody to grammar in English: The differentiation of catenatives, modals, and
auxiliaries from a single protomorpheme, In J. Weissenborn, & B. Hohle (Eds.), Approaches to
bootstrapping: Phonological, lexical, syntactic and neurophysiological aspects of early language
acquisition. Language acquisition and language disorders (pp. 23) Amsterdam: John Benjamins.
Powers, S.. M. (1998a). Children's innovative instrument nominals. In E. Clark (Ed.), Proceedings of the
29th annual meeting of the child language research forum (pp. 201-210). Cambridge: Cambridge
University Press.
Powers, S. M. (1998b). Binarity and singularity in child grammar. In B. Hollebrandse (Ed.), New
perspectives on language acquisition. University of Massachusetts Occasional Papers in Linguistics,
22, (pp. 1-14). University of Massachusetts: GLSA.
Powers, S. M. (1996). The growth of the phrase marker: Evidence from subjects. Doctoral Dissertation,
Linguistics Department, University of Maryland College Park.
Powers, S. M. (2001). Children's semi-lexical heads. In N. Corver, & H. van Riemsdijk (Eds.), Semi-
lexical categories: The function of content words and the content of function words. Studies in
generative grammar 59 (pp. 97-126). Mouton de Gruyter.
Powers, S. M., & Musolino, J. (1997). Precursor relative clauses in the acquisition of English. In A.
Sorace, C. Heycock, & R. Shillcock (Eds.), Proceedings of the GALA '97 conference on language
acquisition (pp. 126-l31). Edinburgh: University of Edinburgh.
Radford, A. (1990). Syntactic theory and the acquisition of English syntax: The nature of early child
grammars of English. Oxford, UK: Basil Blackwell.
Roeper, T. (1996). Merger theory and formal features in acquisition. In H. Clahsen (Ed.), Generative
perspectives on language acquisition (pp. 415-450). Amsterdam: John Benjamins.
Roeper, T. (1997). Children's minimalist representations: Merge and formal features. Paper presented at
New Perspectives on Language Acquisition: Minimalism and Pragmatics, Linguistics Department,
University of Massachusetts, Amherst.
Stern, C., & Stern, W. (1928). Die Kindersprache. Leipzig: Barth.
Tracy, R. (1998). Zwei Sprachen und ein Kopf: Was [iir noise it makes. Paper presented at the Linguistics
Department, University of Potsdam, Potsdam, Germany.
SECTION III
BASIC FUNCTIONS OF READING

M. S. STARR, G. KAMBE, B. MILLER, & K. RAYNER
COGNITIVE PROCESSES AND EYE MOVEMENTS

DURING READING
Abstract. For the past century, researchers have struggled to unravel the complexities of cognitive
processes involved in reading. In this chapter, we discuss some of the remarkable progress being made in
the field of reading research, and we show how eye movement measures have become instrumental in
revealing the moment-to-moment activity of the mind during reading.
1. INTRODUCTION
In 1879, Emile Javal made the simple observation that readers' eyes do not glide
smoothly over printed text, rather they jump and fixate at different locations within
the text until they reach the end of the line (see Huey, 1968). Prior to Javal's
research, presumably many people assumed that readers moved their eyes
continuously across successive segments of text and, as a consequence of this belief,
it was held that eye movements were largely unrelated to cognitive processing
during reading. Although Javal's initial observations sparked a number of questions
regarding the role of eye movements in reading, early studies on eye movements
tended to focus upon the basic facts of oculomotor control, and few researchers were
interested in using eye movements to infer cognitive processes of the mind.
However, given the relative dearth of language theories (as well as the absence of
today's more sophisticated eye-tracking technology), researchers during this time
period discovered a remarkable number of basic facts about eye movements. Dodge
(1900), for example, noted that readers couldn't seem to extract information during
an eye-movement (a phenomenon commonly known as saccadic suppression), and
that information could only be obtained when the eyes stopped moving and fixated
at a point within the text. Other researchers examined the size of the perceptual span,
the time it takes to initiate an eye movement, and even the prevalence of word
skipping.
Despite this initial explosion of research, many years passed before researchers
really began to examine the relationship between eye movements and the activity of
the mind. Instead, as was consistent with the objectives of the behaviorist
movement, research undertaken between 1910 and 1960 generally focused upon the
observable surface aspects of eye movements (Rayner, 1978). One important factor
that altered the trend was the publication of Chomsky's (1957) seminal work on
linguistic theory. In it, he proposed that the study of language and the inner
workings of the mind are closely related to one another and that simple behaviorist
principles were insufficient to account for language acquisition and language
learning. Consequently, eye movement research in particular (and reading research
in general) experienced a radical paradigm shift in the early 1970's. Scientists began
E. Witruk, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions ofLanguage, Reading,

122 M. S. STARR, G. KAMBE, B. MILLER, & K. RAYNER
to view eye movements as a window into the cognitive processes of the mind, and
they began using eye movement measures to infer on-line, moment-to-moment
processing during reading.
In this chapter, our primary goal is to demonstrate that where readers look and
how long they look there can provide valuable insights into the nature of the mental
processes that are involved in the comprehension of printed text. We first briefly
describe some basic empirical facts about eye movements as they pertain to the
reading process, and we discuss evidence from prior studies which indicate that eye
movements can be used to infer moment-to-moment cognitive processes. The
remainder of the chapter will focus on some recent work from our lab which
illustrates the relation of eye movements to on-line cognitive processing.
2. BASIC FUNCTIONS OF EYE MOVEMENTS IN READING
As observed by Javal, although it may seem as if our eyes sweep smoothly across
lines of text as we read, in actuality, typical oculomotor activity during reading
consists of a series of saccades, wherein the eyes jump from location to location and
fixations, wherein the eyes remain relatively stable. Fluent reading consists of the
coordination of these two types of oculomotor activity, with the eyes discretely
moving and fixating so as to focus the image of a letter/word onto the retina. The
reason that saccadic eye movements are so necessary and so frequent during reading
is that our visual acuity is quite limited. The retina itself is capable of covering 240
degrees of visual angle (Llewellyn-Thomas, 1968), but high acuity vision is limited
to the fovea (comprising about 2 degrees of visual angle), which is located at the
center of the retina where neural receptors are most densely packed. Directly
adjacent to the fovea is the parafovea, which subtends an additional 10 degrees--
everything beyond the parafovea forms the periphery, where photoreceptors are less
densely packed and visual clarity diminishes rapidly. Given these limitations, as
readers, we are forced to move our eyes in order to focus the fovea onto parts of
stimuli (e.g., words) that we wish to perceive clearly.
In addition to the physiological limitations inherent in the visual system (i.e.,
acuity), the amount of information taken in during a fixation is also limited by the
range of effective vision, or the perceptual span. When we look at a page of text
either in a book or on a computer monitor, we get the subjective impression that we
are able to see many words at the same time. However, although we may be able to
discern that there are multiple lines of text or that there are many word-like objects
on the page, during normal reading for comprehension, the amount of information
we can extract from text is actually quite small. Most experiments aimed at
measuring the perceptual span have made use of eye-movement contingent display
changes (McConkie & Rayner, 1975; Rayner, 1975a) wherein the text displayed on
a computer screen is manipulated as a function of where the eyes are fixated. As the
eyes move across a line of text, letters or words are modified in foveal, parafoveal,
or peripheral locations, thus manipulating the nature and amount of information
available to the reader. Studies utilizing this paradigm have found that the perceptual
span in English is asymmetrical, extending to about 15 characters to the right of
fixation and four characters to the left (see Rayner, 1998 for a summary).
Presumably, this asymmetry is due to the fact that reading progresses from left to
COGNITIVE PROCESSES AND EYE MOVEMENTS 123
right in English, and the longer span to the right of fixation allows for visuo-spatial
attention to be directed to the next word in the text. This asymmetry is reversed in
left-directed reading such as Hebrew (see Pollatsek, Bolotsky, Well, & Rayner,
1981). Moreover, the size of the perceptual span is relatively consistent across
languages with alphabetic orthographies (Rayner, 1998).
Other studies revealed that the left and right boundaries of the perceptual span
are somewhat different in nature. Rayner, Well, and Pollatsek (1980) and Rayner,
Well, Pollatsek, and Bertera (1982), for example, manipulated the number of visible
words vs. number of visible letters available to the left and right of fixation. They
found that the primary determinant of the left boundary was the beginning of the
currently fixated word. More specifically, reading rates were solely a function of
whether the entire currently fixated word was visible-- the actual number of letters
visible to the left of fixation, by contrast, had little effect on reading rates. However,
the right boundary of the perceptual span was found to be primarily dependent on
the number of visible letters available to the right of fixation, and whether whole
word integrity was preserved had little effect on reading rates.
During normal reading, saccadic jumps serve to bring target words into foveal
focus so that they can be fixated and processed. These jumps are relatively fast,
taking only about 30 ms and, given that the speed of these saccades can reach up to
500 degrees per second, it is generally accepted that visual sensitivity is reduced
during the "blur" of an eye movement such that little or no new information is
obtained during a saccade (Dodge, 1900; Ishida & Ikeda, 1989; Wolverton & Zola,
1983). Saccade sizes typically range from less than one character space to greater
than 20 spaces, although the average eye movement is approximately 8 character
spaces in length. It is important to note that character spaces are the appropriate
metric to use in discussing saccade length since it has been demonstrated that the
distance that the eyes move is determined by letters rather than visual angle for
normal sized print (Morrison & Rayner, 1981). Saccadic eye movements in reading
are also said to be ballistic, in that once a saccade is initiated, the landing position of
the eyes generally cannot be altered.
Although the eyes typically move in the direction of the text (e.g., from left to
right in English), about 10 to 15 percent of eye movements move backward and are
termed regressions (Rayner, 1978, 1998; Rayner & Pollatsek, 1989). Most
regressions are relatively small, with the eyes moving a distance of only a few
letters. Such regressions are often made in response to comprehension difficulty (see
Rayner, 1998 for a review), but they may also occur as a result of oculomotor error
(e.g., when a saccade is too long). In addition, the eyes typically regress back to
words located on the current line of text rather than to words on previous lines
(Duffy, 1992). When regressions to previous portions of text do occur, readers are
generally very accurate at moving their eyes back to the location within the text
which caused them difficulty (Frazier & Rayner, 1982; Kennedy & Murray, 1987).
Since little or no word information is obtained during saccades, the vast majority
of lexical processing in reading takes place during fixations. Fixation durations are
variable, ranging from less than 100 ms to over 1000 ms, with a mean of about 250
ms (Rayner & Pollatsek, 1989). Of course, different measures can be taken with
respect to how long readers look at words. This is a rather complicated issue which
we will not go into here (see Rayner, 1998). The two most commonly used measures
are first fixation duration, which consists of readers' initial fixation on a word, and
gaze duration, which consists of readers' cumulated fixations (Le., refixations) on a

word before moving off the word. Throughout this chapter, for the sake of simplicity
we will typically discuss fixation time, but we'll also mention the more specific
measures (first fixation and gaze) when necessary. In general, it has been the case
that both measures are fairly consistent.
Readers tend to fixate on or near most words in text, and the majority of words
are only fixated once (Just & Carpenter, 1980). However, some words are skipped
during reading, and word-skipping tends to be a function of word length. Short
words (e.g., function words) are generally skipped about 75% of the time, while
longer words consisting of 8 or more letters are rarely skipped. When words are
fixated, however, information appears to be extracted relatively quickly. Rayner,
Inhoff, Morrison, Slowiaczek, and Bertera (1981; see also Ishida & Ikeda, 1989;
Slowiaczek & Rayner, 1987) found that most information is obtained in the first 50
ms of fixation. However, this doesn't necessarily mean that word information is only
obtained in the initial 50 ms of fixation. In a sentence reading experiment,
Blanchard, McConkie, Zola, and Wolverton (1984) briefly masked selected target
words and then replaced them with either the same or a different word (e.g., bombs
replaced by bombs or tombs). Readers were given four possible word choices, two
of which corresponded to critical targets, and were asked to select the word(s) which
were present in the sentence. When the presentation of the second possible target
(e.g., tombs) was delayed by up to 120 ms, readers could generally report both
words. Furthermore, readers could occasionally report the second target when it was
presented in the last 30 ms of fixation. Together, these results indicate that word
information may be extracted throughout fixation.
On the surface then, a relatively simple pattern of eye movements in reading
emerges. Readers generally move their eyes forward by about 7 to 9 character
spaces per saccade, and they tend to fixate upon the majority of words in text for
approximately 200 to 250 ms. If this were all there were to eye movements, we
might tend to agree with earlier researchers' contentions that eye movements only
reflect low-level oculomotor (Le., mechanical) factors. Closer examination of eye
movement patterns in reading, however, reveals a much more complex picture. Most
notably, there is a great deal of variability in both the distance that the eyes move as
well as in the duration that the eyes fixate. In addition, readers' eyes regress
backwards in text about once every 2 seconds. In the next section, we will explore
some of the factors which have been found to affect eye movement patterns in
reading, and in particular, we will discuss the cognitive processes which guide both
when and where to move the eyes.
3. COGNITIVE CONTROL OF EYE MOVEMENTS
Ultimately, the goal of most eye movement researchers is to be able to account for
the variability in both fixation duration and saccades. However, there has been
considerable debate in the past twenty years concerning the relative influences of
lower level visuomotor/oculomotor factors versus higher level cognitive factors on
eye movements in reading. Researchers who support a low-level oculomotor
account of eye movements posit that eye movements are only indirectly related to
lexical processing and that the decisions of when and where to move the eyes are
largely determined by mechanical, visuomotor factors (e.g., O'Regan, 1990, 1992).

One simple example of this is that where the eyes land in a word is largely
systematic-readers' eyes initially tend to land about halfway between the beginning
and the middle of each word, regardless of the word's content (McConkie, Kerr,
Reddix, & Zola, 1988; Rayner, 1979). In addition, it was found that the duration of a
fixation on a particular word was not correlated with either the length of the saccade
off that word or with the subsequent fixation duration on the next word in the text
(Andriessen & deVoogd, 1973; Rayner & McConkie, 1976). To some researchers,
these findings seemed to indicate that the variability in the eye movement record is
largely random.
On the other hand, researchers who support a processing approach claim that eye
movements are primarily influenced by both lexical factors and by moment-to-
moment comprehension processes. Although processing theorists don't rule out the
influence of low-level oculomotor strategies, they propose that the influence of
visuomotor factors is small relative to the influence of cognitive factors. Rayner and
Pollatsek (1981) reported a study in which the amount of text information available
to the reader was randomly varied from fixation to fixation (all other letters were
masked). They found that saccade length was a function of the number of letters
available beyond the reader's current fixation point, indicating that saccade length
was computed on-line. In the same report, Rayner and Pollatsek delayed the onset of
information available to the reader by randomly masking foveal (i.e., fixated) text
for various durations. They found that fixation durations were a function of delay,
with longer fixations being associated with longer delays. Rayner and Pollatsek
concluded that, like saccade lengths, fixation times were also computed on-line.
In the next section we will review the growing body of data which support the
view that linguistic variables have a powerful influence on eye movements.
Furthermore, we will argue that cognitive variables have profound effects on the
time spent fixating a word while also influencing where the eyes move.
3.1 When to Move the Eyes
Abundant evidence has accumulated which indicates that fixation times during
reading are closely related to the ease or difficulty of the cognitive processes
associated with comprehending a word or a segment of text. Although low-level
variables such as word length strongly influence the amount of time a reader fixates
on a word (Kliegl, Olson, & Davidson, 1982; Rayner & McConkie, 1976; Rayner,
Sereno, & Raney, 1996), fixation times have also been found to be influenced by a
variety of lexical, syntactic, and discourse variables. Even when word length is held
constant, word frequency in particular has been shown to have a profound effect on
fixation times (both first fixation and gaze duration). The time spent fixating a word
is generally longer for lower frequency words, which are less likely to be
encountered during reading, and shorter for higher frequency words, which are more
likely to be encountered during reading (Henderson & Ferreira, 1990; Inhoff &
Rayner, 1986; Kennison & Clifton, 1995; Rayner, 1977; Rayner & Duffy, 1986).
Presumably this is because higher frequency words are more easily processed than
are lower frequency words. Furthermore, there is a spillover effect for low frequency
words (Rayner & Duffy, 1986; Rayner, Sereno, Morris, Schmauder, & Clifton,
1989). When the currently fixated word is a low frequency word, cognitive
processing may be passed downstream in the text, leading to an increase in fixation
times on the next, to-be-fixated, word. An interesting corollary to the spillover effect
is that when words are encountered multiple times within a passage, fixation times
on the words decrease, particularly if they are of low frequency (Hyona & Niemi,
1990; Rayner, Raney, & Pollatsek, 1995).
The nature of a word's morphology also has a mediating effect on fixation times.
Lima (1987), for example, found that readers tend to look longer at prefixed words
than at pseudoprefixed words. More recent evidence suggests that the processing of
words is influenced by their component morphemes. The frequency of initial
morphemes is inversely related to fixation duration, with fixation times increasing
with decreasing initial morpheme frequency (Hyona & Pollatsek, 1998).
A number of variables involved in higher level text processing have also been
found to affect fixation times. When words are highly constrained by preceding text
(i.e., they are highly predictable from previous context), readers tend to fixate on
them for shorter periods of time (Balota, Pollatsek, & Rayner, 1985; Binder,
Pollatsek, & Rayner, 1999; Inhoff, 1984; Rayner & Well, 1996; Schustack, Ehrlich,
& Rayner, 1987). Context also influences fixation times on lexically ambiguous
words (e.g., river bank vs. monetary bank). When context biases one meaning of a
balanced ambiguous word - which has two equally likely meanings - gaze durations
are equivalent for ambiguous words as compared to length and frequency matched
control words. In contrast, for an unbalanced ambiguous word - which has dominant
and subordinate meanings - when context biases the subordinate (less common)
meaning, gaze durations are longer on ambiguous words than on frequency matched
controls (Duffy, Morris, & Rayner, 1988).
In another study, O'Brien, Shank, Myers, and Rayner (1988) asked participants
to read passages which contained one of three potential phrases: stabbed her with his
weapon, stabbed her with his knife, or assaulted her with his weapon. When
subsequently presented with the target word knife, readers' gaze durations on knife
were equivalent for stabbed her with his weapon as compared to stabbed her with
his knife, presumably because readers were able to infer that the weapon referred to
in the former phrase was a knife (i.e., it is unlikely that someone would be stabbed
with a gun). In contrast, for phrases such as assaulted her with his weapon,
subsequent gaze durations on the target word knife were longer.
3.2 Where to Move the Eyes

In addition to fixation time, moment-to-moment cognitive processing has also been
found to affect the decision of where to move the eyes. In general, however, it has
been demonstrated that word-length information is the primary determinant of where
the eye initially fixates in a word. When word length information for the next word
in the text is denied (e.g., masked), readers' eyes tend to move a shorter distance
(McConkie & Rayner, 1975; Pollatsek & Rayner, 1982; Rayner, Fischer, &
Pollatsek, 1998). Early studies on landing positions within words found that the
location of fixations within words was not random, rather there is a preferred
viewing location - readers' eyes tend to land somewhere between the middle and the
beginning of words (Rayner, 1979; O'Regan, 1980). Readers also tend not to fixate
on uninformative parts of text such as the extra spaces between sentences (Abrams
& Zuber, 1972; Rayner, 1975b). In addition, Vitu (1991) found that although
readers' initial fixation locations were on or near the preferred viewing location, for
longer words (e.g., 10+ letters), readers initially fixated near the beginning of the
word and then made a refixation near the end of the word (Hyonii, Niemi, &
Underwood, 1989; O'Regan, Levy-Schoen, Pynte, & Brugaillere, 1984; Rayner &
Morris, 1992; Underwood, Bloomfield, & Clews, 1988; Underwood, Clews, &
Everatt, 1990; Underwood, Hyonii, & Niemi, 1987).
As mentioned previously, all eye movements are not directed forward in text-
some words are refixated, some are skipped, and sometimes the eyes regress
backwards in text. Although low-level visual factors may primarily determine where
to move the eyes for forward eye movements, there is a growing body of evidence
which indicates that cognitive factors are responsible for such refixations, skips, and
regressions.
Approximately 15% of the words in text are refixated. Some researchers (e.g.,
O'Regan & Levy-Schoen, 1987) have argued that refixations within a word are
caused by oculomotor error, wherein the eyes originally land in a non-optimal
location. In such cases, a refixation is necessary to move the eyes to a location
within a word which is more conducive to the extraction of lexical information.
However, such a contention would suggest that most refixations should move the
eyes to the optimal viewing location (the middle of the word) and that little
information should be obtained from the initial "erroneous" fixation. In contrast to
this prediction, the most prevalent pattern of refixations is an initial fixation near the
beginning of a word, followed by a refixation near the end of the word (Rayner &
Pollatsek, 1987; Rayner et aI., 1996). There is also evidence that word frequency
affects the initial fixation on words which are subsequently refixated (Sereno, 1992),
and that refixations are less likely if the currently fixated word is predictable from
prior sentence context (Balota et aI., 1985).
Word skipping is also governed by a number of factors. Although the most
influential variable involved in word skipping is word length (short words are less
likely to be fixated than are long words, Brysbaert & Vitu, 1998; Rayner &
McConkie, 1976), there is also some evidence that high frequency words are
skipped more often than low frequency words (Radach & Kempe, 1993; Rayner et
aI., 1996). Contextual constraint has also been found to be a powerful predictor of
skipping, in that words which are highly constrained by prior sentence context are
more likely to be skipped than words which are not highly constrained (Balota et aI.,
1985; Binder et aI., 1999).
Despite the fact that 10-15% of fixations are regressions, little is known about
what causes them (Rayner, 1998). However, it does seem that regressions are
profoundly influenced by cognitive processing- specifically, by text difficulty and
comprehension failures (Frazier & Rayner, 1982; Just & Carpenter, 1980). When
readers encounter a word which indicates that their original interpretation of a
sentence is incorrect, they will often immediately regress their eyes to the prior
portion of text which will amend their comprehension of the context (Frazier &
Rayner, 1982).
3.3 Integration of Information across Eye Movements
Thus far we have discussed issues relating to how information is processed on

individual fixations. However, there are several studies which show that word
information may be partially processed on one fixation and then completed on the
next (Rayner & Pollatsek, 1989). Given that both the physiological make-up of the
eye and the width of the perceptual span allow the reader to perceive information in
the parafovea, it is not surprising that linguistic information may also be obtained
from the word to the right of the currently fixated word (i.e., the parafoveal word).
In general, Rayner et aI. 's (1982) finding that letter information to the right of
fixation is more important than word integrity suggests that at least partial letter
information may be extracted from the parafoveal word. Moreover, information
gleaned from the parafoveal word will typically decrease the time spent viewing it
when it is subsequently fixated. If lexical information to the right of fixation is
denied, reading rate decreases rapidly, falling to almost one third of its baseline level
(Rayner et al., 1982). The advantage gained by the availability of lexical information
in the parafovea over the absence of lexical information in the parafovea is called
the parafoveal preview benefit (Rayner & Pollatsek, 1989).
What other kinds of information are integrated across saccades? Given our
earlier discussion on the issue of where to move the eyes, it is perhaps not surprising
that word length information is reliably obtained parafoveally and is subsequently
used in computing the location of the next fixation (O'Regan, 1980; Morris, Rayner,
& Pollatsek, 1990; Pollatsek & Rayner, 1982; Rayner, 1979; Rayner & Morris,
1992; Rayner et aI., 1996). As with most other variables involved in eye movements
during reading, the processing and extraction of information from the parafovea is
affected by moment-to-moment cognitive processing, as well.
A variety of factors have been implicated in the integration of information across
saccades. One of these is the parafoveal word's phonological (sound) code.
Pollatsek, Lesch, Morris, and Rayner (1992) found that when a homophone of the
currently fixated word was presented in the parafovea, processing of the fixated
word was facilitated as compared to when the parafoveal preview word was
orthographically similar to the fixated word. Although it seems that morphological
units do not facilitate the integration of information across saccades (Inhoff, 1989;
Lima, 1987), some evidence suggests that sentential constraint may mediate
parafoveal processing. Balota et aI. (1985) found that the parafoveal preview benefit
was greater when the currently fixated word was predictable from previous context,
indicating that the processing of parafoveal words is more efficient when facilitated
by sentence context.
The linkage between foveal and parafoveal words has also been evinced by a
number of other findings. Rayner et al. (1989), for example, discovered that
difficulty in processing a word can be passed downstream (spillover). As mentioned
earlier, they found that as the frequency of the foveal word decreases, fixation times
increase on parafoveal words when they are subsequently fixated. Readers have also
been shown to obtain more information from a high frequency word in the parafovea
than from a low frequency word (Kennison & Clifton, 1995; cf. Inhoff & Rayner,
1986). Moreover, Henderson and Ferreira (1990) found an inverse relationship
between the difficulty of the foveal word and the size of the subsequent preview
benefit, such that when the difficulty of the foveal (i.e., fixated) word was high, the
preview benefit was small.
4. RECENT FINDINGS
In this section we will briefly describe three recent findings which further
demonstrate a tight linkage between eye movements and cognitive processes in
reading. Specifically, we will describe studies which have investigated three
potential cognitive influences on the "when" decision to move the eyes during
reading: the role of words' sounds and spellings (orthographic and phonological
neighborhood sizes), the effects of multiple word meaning (lexical ambiguity), and
the effects of morphological complexity. In this final section, we will briefly discuss
some of the ways in which these variables influence eye movements.
4.1 Phonological and Orthographic Neighborhood Size
Through the use of a variety of reading measures (e.g., eye movements, lexical
decision times, naming latencies), a great deal has been learned about the processes
involved in word recognition. However, one question that has yet to be satisfactorily
addressed deals with the effects of a word's spelling and sound characteristics on the
ease or difficulty of lexical processing. For example, some models of word
recognition posit that during the initial processing of words, abstract orthographic
(spelling) codes are quickly utilized to activate phonological (sound) codes (e.g.,
Pollatsek et aI., 1992; Van Orden, Johnston, & Hale, 1988). If this is the case, it
would be expected that both orthographic and phonological neighborhood candidate
sets will be activated during reading. A word's orthographic neighborhood size is
indexed by counting the number of legal words which can be generated by changing
one letter of the target word while keeping the remaining letters unchanged
(Coltheart, Davelaar, Jonasson, & Besner, 1977). For example, the orthographic
neighbors of the word shark would be spark, stark, shirk, shard, etc. Similarly,
phonological neighborhood size is computed by counting the number of legal words
that can be generated by changing one phoneme of the target word while keeping the
remaining phonemes unchanged. For instance, the phonological neighbors of the
word shark would include hark, park, shard, lark, mark, etc.
In an experiment in our lab, we (Miller, Rayner, & Pollatsek, 1999) manipulated
the sizes of target words' orthographic and phonological neighborhoods. We found
that larger orthographic neighborhood sizes led to shorter fixations on target words,
but only when the phonological neighborhood was also large: gaze durations were
20 ms longer when there was a small orthographic neighborhood than when there
was a large orthographic neighborhood. For phonological neighborhoods, larger
phonological neighborhood sizes led to shorter fixations on the target word, but only
when the orthographic neighborhood size was small: gaze durations were 26 longer
when there was a small phonological neighborhood than when there was a large
phonological neighborhood. One possibility for this pattern of effects is that for
words which have a relatively large number of orthographic neighbors, increases in
the number of words which are phonologically similar to the target may provide an
extra boost in activation (i.e., word processing is facilitated by both orthographic
and phonological codes). For words which have a relatively large number of
phonological neighbors, as the number of words which are also orthographically
similar to the target increases, a corresponding increase in activation may result in a
facilitatory effect. Hence, it appears that both an orthographic and phonological
candidate set is activated during visual word recognition and that models need to
posit an early role for both orthography and phonology. Although more work needs
to be done in this area, it appears that both orthographic and phonological
neighborhood sizes influence eye fixation durations during reading.
4.2 Lexical Ambiguity
Earlier in this chapter, we noted that gaze durations on ambiguous words vary as a
function of context (Duffy et aI., 1988; Rayner & Duffy, 1986). Specifically, gaze
durations are longer when an ambiguous word has two equally likely meanings
(balanced ambiguous words) and the preceding context is neutral with respect to
which meaning to instantiate than are fixations on unambiguous words (matched on
length and frequency). Presumably, this is because both meanings of the ambiguous
word are accessed more or less simultaneously, resulting in some form of
competition. When the preceding context is consistent with one of the meanings,
readers look no longer at the ambiguous word than at the control word. On the other
hand, when one meaning is more dominant than the other meaning (unbalanced
ambiguous words) and the context is neutral, gaze durations on the ambiguous
words are equivalent to those on the unambiguous words. In this case, the dominant
meaning of the ambiguous word becomes available first, thus avoiding any
competition. However, if the preceding context instantiates the subordinate meaning,
then readers' gaze durations are longer on the ambiguous word than on the control
word (again matched on length and frequency). This latter finding, referred to as the
subordinate bias effect (Rayner, Pacht, & Duffy, 1994; Binder & Rayner, 1998) has
been replicated a number of times.
Table 1. Gaze Duration (ms) on the Target Word
Global Subordinate Global Dominant

Local Before Local After Local Before Local After
Ambiguous 303 304 304 298
Control 286 284 291 297
In recent research in our lab (see Table 1), we (Kambe & Rayner, 1999) examined
the influence of both local and global context on the subordinate bias effect. Local
context was manipulated within the sentence containing the ambiguous target word,
and all local contexts were biased towards the subordinate meaning of the
ambiguous word. The contextually biasing information, however, either preceded or
followed the ambiguous target word. Consistent with prior research, local context
influenced gaze durations: when the preceding context instantiated the subordinate
meaning, gaze durations on the target words were longer than were gaze durations
on frequency/length matched control words.
To manipulate global context, we embedded unbalanced ambiguous words into

paragraphs wherein the overall discourse context was biased towards either the
dominant or subordinate meaning of the word. The contextual information in the
discourse context was instantiated in the topic sentence of the paragraph. We found
that when the global discourse context was biased towards the subordinate meaning,
gaze durations on ambiguous target words were longer than on control words.
Hence, the distant discourse context had an immediate influence on processing times
for ambiguous target words. However, this effect was not modulated by the amount
of biasing information available prior to encountering the ambiguous target word.
For example, when both global context and local contexts were biased towards the
subordinate meaning of the ambiguous word, gaze durations were no different than
when either the global context or the prior local context contained biasing
information. These results indicate that biasing contextual information prior to the
ambiguous target word may be able to influence fixation times on the target word
and the decision when to move the eyes.
4.3 Morphologically Complex Words
Compound words differ from other types of morphologically complex words (e.g.,
lamplight vs. govern~government or run~running) in that they are formed by
joining two free lexemes (i.e., the words which make up the compound), and in that
the linguistic information conveyed by a compounded expression may differ from
the information conveyed by the individual lexemes. For instance, a black bird,
consisting of two free lexemes, is a bird that must be black, but the concept referred
to by the compounded expression, blackbird denotes a particular type of bird. The
sequence of two free lexemes, black and bird provides little additional information
about the bird's properties other than color, but the compounded version, blackbird,
conveys specific notions of form and function.
To date, many researchers have found that constituent lexemes within compound
words are accessed at some point during compound processing (e.g., Andrews,
1986; Hyonii & Pollatsek, 1998; Sandra, 1990; Taft & Forster, 1976). The nature of
how readers determine that a compound word consists of two individual lexemes
(morphological decomposition) has remained elusive, however. Specifically,
researchers have yet to determine how compound words are represented in the
mental lexicon. For example, compounds could be represented and accessed via
their full forms, or they could be represented and accessed via their lex erne
constituents. Hence, there are two major questions. First, if morphological
constituents are critical for compound processing, are they used to activate the
overall compound word (i.e., are lexemes activated before the compound is
recognized) or, in contrast, are constituent lexemes only activated after the overall
compound has been recognized? Second, if lexemes maintain a functional role in the
processing of compounds, then what are the respective roles of beginning and
ending lexemes within compounded words (i.e., is either lex erne more important
than the other)?
Recent work done in two separate studies in our lab (one, referred to as
Experiment 1 below, by the first author in collaboration with Barbara Juhasz, Lars
Placke, and Albrecht Inhoff and the second, referred to as Experiment 2 below, by
the third and fourth authors in collaboration with Sally Andrews) set out to address
these questions by investigating both the nature of morphological decomposition
effects (Le., whether beginning, ending, or beginning and ending lexemes determine
the ease of compound recognition) and also the time-course of these effects. To
examine the decomposition issue, we controlled overall compound word frequencies
and manipulated individual beginning and ending lexeme frequencies
(morphological decomposition would be denoted by the presence of individual
lexeme frequency effects). To examine the time-course of compound processing we
recorded readers' eye movements as they read sentences containing compound
words. As we have pointed out throughout this chapter, one great advantage of the
eye-tracking paradigm is that it allows us to infer moment-to-moment cognitive
processing during reading. Specifically, for our current experiments, the eye-
tracking paradigm gave us the ability to investigate the time-course of compound
processing. Early processing was gauged by first fixation duration and later
processing was gauged by gaze duration (which includes refixations on the target
word prior to moving to another word).
Both of our studies revealed that compounds were decomposed into their
individual lexemes. In Experiments 1 and 2, first fixation durations revealed small
differences due to the frequency of the initial lexeme, with a nonsignificant 11 ms
difference between high and low frequency beginning lexemes for Experiment 1 and
a marginally significant difference in Experiment 2 (8 ms). For ending lexemes,
there was only a trend for high frequency ending lexemes to yield shorter durations
on first fixation with an 8 ms difference between high and low frequency ending
lexemes (4 ms in Experiment 2). For gaze duration though, Experiment 1 revealed
that although the beginning lexeme effect was still nonsignificant (8 ms), ending
lexemes exerted an influence on compound processing, with an ending lexeme effect
of 27 ms (see Table 2). Consistent with Experiment 1, Experiment 2 revealed an
ending lexeme effect in gaze duration (15 ms), but in contrast there was also a robust
beginning lexeme effect (22 ms).
Table 2. First Fixation and Gaze Durations as a Function of Lexeme Frequency
First fixation Gaze

Exp.1 Exp.2 Exp. 1 Exp.2
HH-compound (farmhouse) 278 270 358 383
HL-compound (landslide) 292 273 382 390
LH-compound (whirlwind) 296 277 363 397
LL-compound (graveyard) 297 282 396 419
Hence, we found that readers process the lexemes in compound words separately.
However, given the contrasting results of Experiments 1 and 2, it is not clear
whether beginning or ending lexemes play a more dominant role in compound
processing. We are currently undertaking further experiments to resolve the
discrepancy (see also Placke, Starr, & Inhoff, 1999). Specifically, we are examining
two types of compound words: 1.) "headed" compounds, where the meaning of the
first lexeme is more closely related to the overall compound meaning (e.g.,
humankind), and 2.) "tailed" compounds, where the meaning of the second lexeme
is more closely related to the overall compound meaning (e.g., handbook).
5. SUMMARY
In the past 30 years, considerable advances have been made in understanding eye
movements during reading. Although mechanical visuomotor factors have been
found to influence eye movement patterns, it is also becoming increasing clear that
eye movements during reading are profoundly affected by moment-to-moment
cognitive processing, as well. In this chapter, we have delineated various factors
which have been found to influence both the amount of time the eyes fixate on
words as well as where the eyes move from fixation to fixation. One great virtue of
being able to use eye movements to infer cognitive processing in this manner is that
it allows researchers to use eye movement data to study the activity of the mind, and
it greatly enhances our capacity to construct a coherent explanation of how words
are processed during reading. Many of the discoveries that have been documented in
this chapter have made it possible to develop sophisticated computer simulations
(see Reichle, Pollatsek, Fisher, & Rayner, 1998) that can effectively predict where
readers fixate and how long they look at different words in text. We expect that in
the next few years considerable progress will be made due to both simulation work
and empirical work relating to eye movements and cognitive processes in reading.
6. AFFILIATIONS
Dr. Keith Rayner
Psychology Department
Tobin Hall
University of Massachusetts
Amherst
MA 01003
U.S.A.
e-mail: rayner@psych.umass.edu
7. NOTES
I Preparation of this chapter was made possible by Grants HD17246 and HD26765 from the National
Institute of Health. The first author was supported by a Post-doctoral Fellowship from Grant MH16745
from the National Institute of Mental Health, the second author was supported by a Pre-doctoral
Fellowship from Grant HD07327 from the National Institute of Health, and the fourth author was
supported by a Research Scientist Award (MH01255) from the National Institute of Mental Health.
8. REFERENCES
Abrams, S. G., & B. L. Zuber. (1972). Some characteristics of information processing during reading.
Reading Research Quarterly, 8, 1, 40-51.
Andrews, S. (1986). Morphological influences on lexical access: Lexical or nonlexical effects? Journal of
Memory and Language, 25, 726-740.
Andriessen, 1., & deVoogd A. (1973). Analysis of eye movement patterns in silent reading. [PO Annual
Progress Report, 8, 29-35.
Balota, D. A., Pollatsek A., & Rayner, K. (1985). The interaction of contextual constraints and parafoveal
visual information in reading. Cognitive Psychology, 17, 3, 364-390.
Binder, K S., Pollatsek A., & Rayner, K (1999). Extraction of information to the left of the fixated word
in reading. Journal of Experimental Psychology: Human Perception and Performance, 25, 4, 1162-
1172.
Binder, K S., & Rayner, K (1998). Contextual strength does not modulate the subordinate bias effect:
Evidence from eye fixations and self-paced reading. Psychonomic Bulletin & Review, 5, 2, 271-276.
Blanchard, H. E., McConkie, G. W., Zola, D., & Wolverton, G. S. (1984). Time course of visual
information utilization during fixations in reading. Journal of Experimental Psychology: Human
Perception and Performance, 10, 1, 75-89.
Brysbaert, M., & Vitu, F. (1998). Word skipping: Implications for theories of eye movement control in
reading. In G. Underwood (Ed.), Eye guidance in reading and scene perception (pp. 125-147).
Oxford, England UK: Elsevier Science Ltd.
Chomsky, N. (1957). Syntactic structures. Gravenhage, Holland: Mouton.
Coltheart, M., Davelaar, E., Jonasson, J. F., & Besner, D. (1977). Access to the intemallexicon. In S.
Domic (Ed.), Attention and performance VI (pp. 535-555). Hillsdale, NJ: Erlbaum.
Dodge, R. (1900). Visual perceptions during eye movement. Psychological Review, 7, 454-465.
Duffy, S. A. (1992). Eye movements and complex comprehension processes. In K Rayner (Ed.), Eye
movements and visual cognition: Scene perception and reading (pp. 462-471). New York: Springer-
Verlag.
Duffy, S. A., Morris, R. K, & Rayner, K. (1988). Lexical ambiguity and fixation times in reading.
Journal of Memory and Language, 27, 4, 429-446.
Frazier, L., & Rayner, K. (1982). Making and correcting errors during sentence comprehension: Eye
movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14, 2, 178-
210.
Henderson, J. M., & Ferreira, F. (1990). Effects of foveal processing difficulty on the perceptual span in
reading: Implications for attention and eye movement control. Journal of Experimental Psychology:
Learning, Memory, and CognitiOn, 16,3,417-429.
Huey, E. B. (1968). The psychology and pedagogy of reading. (1908); Reprint, Cambridge, MA: MIT
Press.
Hyona, 1., & P. Niemi. (1990). Eye movements during repeated reading of a text. Acta Psychologica, 73,
3,259-280.
Hyona, J., Niemi, P., & Underwood, G. (1989). Reading long words embedded in sentences:
Informativeness of word halves affects eye movements. Journal of Experimental Psychology: Human
Perception and Performance, 15, 1,142-152.
Hyona, J., & Pollatsek, A. (1998). Reading Finnish compound words: Eye fixations are affected by
component morphemes. Journal of Experimental Psychology: Human Perception and Performance,
24,6, 1612-1627.
InhOff' A. W. (1984). Two stages of word processing during eye fixations in the reading of prose. Journal
of Verbal Learning and Verbal Behavior, 23, 5, 612-624.
Inhoff, A. W. (1989). Lexical access during eye fixations in reading: Are word access codes used to
integrate lexical information across interword fixations? Journal of Memory and Language, 28, 4,
444-461.
Inhoff, A. W., & Rayner, K (1986). Parafoveal word processing during eye fixations in reading: Effects
of word frequency. Perception & Psychophysics, 40, 6, 431-439.
Ishida, T., & Ikeda, M. (1989). Temporal properties of information extraction in reading studied by a text-
mask replacement technique. Journal of the Optical Society ofAmerica, 6, 10, 1624-1632.
Just, M. A., & Carpenter, P. A. (1980). A theory of reading: From eye fixations to comprehension.
Psychological Review, 87, 4, 329-354.
Kambe, G., & Rayner, K (1999). Does global and local context have an effect on lexical ambiguity
resolution? Paper presented at the Eastern Psychological Association Meeting, Providence, Rhode
Island.
Kennedy, A., & Murray, W. S. (1987). The components of reading time: Eye movement patterns of good
and poor readers. In J. K O'Regan, & A. Levy-Schoen (Eds.), Eye movements: From physiology to
cognition (pp. 509-520). Amsterdam, Netherlands: Elsevier Science Publishing Co, Inc.
Kennison, S. M., & Clifton C. (1995). Determinants of parafoveal preview benefit in high and low
working memory capacity readers: Implications for eye movement control. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 21, 1, 68-81.
K1iegl, R. M., Olson, R. K, & Davidson, B. J. (1982). Regression analyses as a tool for studying reading
processes: Comment on Just and Carpenter's eye fixation theory. Memory & Cognition, 10, 3, 287-
296.
COONmVE PROCESSES AND EYE MOVEMENTS 135
Lima, S. D. (1987). Morphological analysis in sentence reading. Journal of Memory and Language, 26, 1,
84-99.
Llewellyn-Thomas, E. (1968). Movements of the eye. Scientific American, 219, 88-95.
McConkie, G. W., Kerr, P. W., Reddix, M. D., & Zola, D. (1988). Eye movement control during reading:
I. The location of initial eye fixations on words. Vision Research, 28, 10, 1107-1118.
McConkie, G. W., & Rayner, K. (1975). The span of the effective stimulus during a fixation in reading.
Perception & Psychophysics, 17, 6, 578-586.
Miller, B., Rayner, K., & Pollatsek, A.. (1999). Investigations into orthographic and phonological
similarity: Are orthographic neighborhood effects simply due to orthography? Paper presented at the
Eastern Psychological Association Meeting, Providence, Rhode Island
Morris, R. K., Rayner, K., & Pollatsek, A. (1990). Eye movement guidance in reading: The role of
parafovealletter and space information. Journal of Experimental Psychology: Human Perception and
Performance, 16, 2, 268-281.
Morrison, R. E., & Rayner, K. (1981). Saccade size in reading depends upon character spaces and not
visual angle. Perception & Psychophysics, 30, 395-396.
O'Brien, E. J., Shank, D. M., Myers, J. L., & Rayner, K. (1988). Elaborative inferences during reading:
Do they occur on-line? Journal of Experimental Psychology: Learning, Memory, and Cognition, 14,
3,410-420.
O'Regan, J. K. (1980). The control of saccade size and fixation duration in reading: The limits of
linguistic control. Perception & Psychophysics, 28, 2, 112-117.
O'Regan, J. K. (1990). Eye movements and reading. In E. Kowler (Ed.), Eye movements and their role in
visual and cognitive processing: Reviews of oculomotor research (pp. 395-453). Amsterdam,
Netherlands: Elsevier.
O'Regan, J. K. (1992). Optimal viewing position in words and the strategy-tactics theory of eye
movements in reading. In K. Rayner (Ed.), Eye movements and visual cognition: Scene perception
and reading (pp. 333-354). New York: Springer-Verlag.
O'Regan, J. K., & Levy-Schoen, A. (1987). Eye movements: From physiology to cognition. Amsterdam,
Netherlands: Elsevier Science Publishing Co., Inc..
O'Regan, J. K., Levy-Schoen, A., Pynte, J., & Brugaillere, B. (1984). Convenient fixation location within
isolated words of different length and structure. Journal of Experimental Psychology: Human
Perception and Performance, 10, 2, 250-257.
Placke, L., Starr, M. S., & Inhoff, A. W. (1999). Accessing compound words: The role of meaning and
decomposition. Poster presented at the European Conference on Eye-Movements, Utrecht,
Netherlands.
Pollatsek, A., Bolotsky, S., Well, A., & Rayner, K. (1981). Asymmetries in the perceptual span for Israeli
readers. Brain and Language, 14, 174-180.
Pollatsek, A., Lesch, M., Morris, R. K., & Rayner, K. (1992). Phonological codes are used in integrating
information across saccades in word identification and reading. Journal of Experimental Psychology:
Human Perception and Performance, 18,1,148-162.
Pollatsek, A., & Rayner, K. (1982). Eye movement control in reading: The role of word boundaries.
Journal of Experimental Psychology: Human Perception and Performance, 8,6, 817-833.
Radach, R., & Kempe, V. (1993). An individual analysis of initial fixation positions in reading. In G.
d'Ydewalle & J. Van Rensbergen (Eds.), Perception and cognition: Advances in eye movement
research. Studies in visual information processing, Vol. 4 (pp. 213-225). Amsterdam, Netherlands:
North-HollandlElsevier Science Publishers.
Rayner, K. (1975a). The perceptual span and peripheral cues in reading. Cognitive Psychology, 7,1,65-
81.
Rayner, K. (1975b). Parafoveal identification during a fixation in reading. Acta Psychologica, 39, 4, 271-
281.
Rayner, K. (1977). Visual attention in reading: Eye movements reflect cognitive processes. Memory &
Cognition, 5, 4, 443-448.
Rayner, K. (1978). Eye movements in reading and information processing. Psychological Bulletin, 85,
618-660.
Rayner, K. (1979). Eye guidance in reading: Fixation locations within words. Perception, 8, 1, 21-30.
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research.
Psychological Bulletin, 124, 3,372-422.
Rayner, K., & Duffy, S. A.. (1986). Lexical complexity and fixation times in reading: Effects of word
frequency, verb complexity, and lexical ambiguity. Memory & Cognition, 14, 3, 191-201.
Rayner, K., Fischer, M. H., & Pollatsek, A. (1998). Unspaced text interferes with both word identification
and eye movement control. Vision Research, 38, 1129-1144.
Rayner, K., Inhoff, A. W., Morrison, R. E., Siowiaczek, M., & Bertera, J. (1981). Masking of foveal and
parafoveal vision during eye fixations in reading. Journal of Experimental Psychology: Human
Perception and Performance, 7, 1,167-179.
Rayner K., & McConkie, G. W. (1976). What guides a reader's eye movements? Vision Research, 16,
829-837.
Rayner, K., & Morris, R. K. (1992). Eye movement control in reading: Evidence against semantic
preprocessing. Journal of Experimental Psychology: Human Perception and Performance, 18, 1,
163-172.
Rayner, K., Pacht, J. M., & Duffy, S. A. (1994). Effects of prior encounter and global discourse bias on
the processing of lexically ambiguous words: Evidence from eye fixations. Journal of Memory and
Language, 33, 4, 527-544.
Rayner, K., & POllatsek, A. (1981). Eye movement control during reading: Evidence for direct control.
Quarterly Journal of Experimental Psychology, 33, A-4, 351-373.
Rayner, K., & Pollatsek, A.. (1987) Eye movements in reading: A tutorial review. In M. Colt heart (Ed.),
Attention and Performance 12: The psychology of reading (pp. 327-362). Hove, England UK:
Lawrence Erlbaum Associates, Inc.
Rayner, K., & Pollatsek, A. (1989). The psychology of reading. Hillsdale, NJ: Erlbaum.
Rayner, K, Raney, G. E., & Pollatsek, A. (1995) Eye movements and discourse processing. In R. F.
Lorch & E. 1. O'Brien (Eds.), Sources of coherence in reading (pp. 9-35). Hillsdale, NJ: Erlbaum.
Rayner, K., Sereno, S. c.; Morris, R. K, Schmauder, A. R., & Clifton, C. (1989). Eye movements and
on-line language comprehension processes. Language & Cognitive Processes, 4, 3-4, 21-49.
Rayner, K, Sereno, S. c., & Raney, G. E. (1996). Eye movement control in reading: A comparison of
two types of models. Journal of Experimental Psychology: Human Perception and Performance, 22,
5, 1188-1200.
Rayner, K, & Well, A. D. (1996). Effects of contextual constraint on eye movements in reading: A
further examination. Psychonomic Bulletin & Review ,3, 4, 504-509.
Rayner, K., Well, A., & Pollatsek, A. (1980). Asymmetry of the effective visual field in reading.
Perception & Psychophysics, 27, 6, 537-544.
Rayner, K, Well, A. D., Pollatsek, A., & Bertera, 1. H. (1982). The availability of aseful information to
the right of fixation in reading. Perception & Psychophysics, 31, 6, 537-550.
Reichle, E. D., Pollatsek, A., Fisher, D. L., & Rayner, K (1998). Toward a model of eye movement
control in reading. Psychological Review, 105, 1, 125-157.
Sandra, D. (1990). On the representation and processing of compound words: automatic access to
constituent morphemes does not occur. Quarterly Journal of Experimental Psychology, 42, A3, 529-
567.
Schustack, M. W., Ehrlich, S. F., & Rayner, K (1987). Local and global sources of contextual facilitation
in reading. Journal of Memory and Language, 26, 3, 322-340.
Sereno, S. C. (1992). Early lexical effects when fixating a word in reading. In K Rayner (Ed.), Eye
movements and visual cognition: Scene perception and reading (pp. 304-316). New York: Springer-
Verlag.
Siowiaczek, M. L., & Rayner, K (1987). Sequential masking during eye fixations in reading. Bulletin of
the Psychonomic Society, 25, 3, 175-178.
Taft, M., & Forster, K. I.. (1976). Lexical storage and retrieval of polymorphemic and polysyllabic words.
Underwood, G., Bloomfield, R., & Clews, S. (1988). Information influences the pattern of eye fixations
during sentence comprehension. Perception, 17, 2, 267-278.
Underwood, G., Clews, S., & Everatt, J. (1990). How do readers know where to look next? Local
information distributions influence eye fixations. Quarterly Journal of Experimental Psychology, 42,
I-A, 39-65.
Underwood, G., Hyonii, 1., & Niemi, P. (1987). Scanning patterns on individual words during the
comprehension of sentences. In 1. K O'Regan & A. Levy-Schoen (Eds.), Eye movements: From
physiology to cognition (pp. 467-477). Amsterdam, Netherlands: North Holland.
Van Orden, G. c., Johnston, J. C., & Hale, B. L. (1988). Word identification in reading proceeds from
spelling to sound to meaning. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 14, 3, 371-386.
Vitu, F. (1991). The existence of a center of gravity effect during reading. Vision Research, 31, 7-8, 1289-
1313.
Wolverton, G. S., & Zola, D. (1983). The temporal characteristics of visual information extraction during
reading. In K Rayner (Ed.), Eye movements in reading: Perceptual and language processes (pp. 41-
51). New York: Academic Press.
R. RADACH, A. INHOFF, & D. HELLER
THE ROLE OF ATTENTION AND SPATIAL

SELECTION IN FLUENT READING
Abstract. Eye movements are an essential part of reading behavior. They are also interesting from a
perceptual and information processing point of view, as they provide a way to study a very complex and
yet ecologically valid mental process in a simple and well structured visual environment. In the current
chapter we discuss a core issue of current research, the role of visual selective attention in reading. After
introducing sequential attention shift models as the currently dominant theory of eye movement control
we explore some limitations of this family of models. This includes a critique of the central claim that
attention is allocated sequentially in a word-by-word fashion and a number of issues regarding the time
line of information processing and oculomotor control. We conclude with a brief look at two alternative
theoretical conceptions.
1. INTRODUCTION
In current discussions on eye movements in reading, the concept of attention is
widely used, but in most cases without an explicit definition of its content. This is
perhaps not very surprising, since the literature on relevant basic research is itself far
from conclusive as to the nature of attentional phenomena (Allport, 1992). One
influential view on the scope of what the term "attention" refers to is apparent in the
functions that Posner and Petersen (1990) allocate to their integrated attention
system: (a) orienting to sensory events, (b) detecting signals for focal (conscious)
processing and (c) maintaining a vigilant or alert state. Within the more limited
scope of visual selection attention is often seen as a limited resource, something that
can be "directed" to or "focused" on a certain location or object. Posner (1980, p.
172) introduced the metaphor of "a spotlight that enhances the efficiency of the
detection of events within its beam". Alternatively, attention can be seen as a
controller that directs or allocates the limited resource (Yantis, 1998). This view is
also quite popular and it is not unlikely that one is confronted with multiple
references to the concept of visual attention within the same publication: First, as a
mechanism that does selection and then as a entity that is moved as a result of this
selection, or, similarly, as a beam of enhanced processing due to resource allocation.
The aim of this chapter is to explore the role of "attention" in current research on
fluent reading. We will start with a review of theories and models of information
processing and eye movement control in reading that rely on the concept of attention
and discuss some relevant evidence. 1 The major conclusion of this discussion will be
that current models are doing a good job in predicting eye behaviour during reading
but that there are also a number of empirical findings that are not in harmony with
these models. As a consequence of several lines of critique, two alternative
theoretical conceptions will be considered.

138 R. RADACH, A. INHOFF, & D. HELLER
Given the situation sketched above, we will not introduce our specific definition
of attention. Instead we will follow a minimalistic definition, using the concept as a
synonym for visual spatial selection and preferred processing of visual input,
deliberately leaving details unspecified until specifications are required and become
useful. At some places it will become difficult to draw a clear line between visual
processing terminology (e.g. letter identification, lexical processing) and attention
terminology (attention allocation, shift of attention). This is a consequence of the
(clearly tautological) custom of using the concept of attention simply to describe
varieties of perceptual and cognitive processing. Also, the necessary distinction
between attention as a description of what has been observed vs. an agent that
causes what can be observed will sometimes be hard to establish, a difficulty that
our chapter shares with much of the literature in the field (Johnston & Dark, 1986;
Neumann, 1987).
2. ATIENTION-BASED THEORIES OF EYE MOVEMENT CONTROL IN

READING
2.1 Early Models: McConkie (1979) and Morrison (1984)
Perhaps the first systematic approach to linking attention and eye movements in
reading was proposed by McConkie (1979) who suggested that visual attention
provides the trigger for moving the eye across lines of text. He defined the region
within which visual detail is being considered for the purpose of linguistic
processing as the "attended" region. He proposed that when the attended region is
shifted along the line of text, there will be a point when visual information is sought
from a retinal region too far from central vision to supply the necessary visual detail.
This will cause the saccadic system to initiate an eye movement. Morrison (1984)
took this idea as a starting point for his influential model of eye movement control.
He criticized that in McConkie's proposal attention to a parafoveal area elicits an
eye movement only when visual processing failed to generate adequate information.
This means that an additional evaluation mechanism is needed.
Morrison combined the idea that shifts of attention trigger eye movements with
the notion of parallel saccade programming (Becker & Jiirgens, 1979, see section
3.2).2 He emphasized that multiple shifts of attention can take place during one
fixation. To illustrate his model, he proposed the following "scenarios", describing a
double attention shift pattern towards two successive words from left to right.
Common to all scenarios is the idea that attention shifts from the currently fixated
word N to the next word N+ 1 as the reader attempts to encode the word. Sometime
later, the allocation of attention on word N+ 1 will "exceed threshold", irrevocably
initiating a movement to the right. 3
1. If word N+ 1 is successfully encoded before the attention shift reaches its
temporal threshold, attention may shift further to word N+2. As this
happened before the temporal threshold to elicit a saccade was reached, a
saccade to the new locus of attention, word N+2, will be programmed,
completely skipping word N+ 1.
2. If successful encoding of word N+ 1 and the related second attention shift
occur when amplitude computation for the first one is already underway,
ATIENTION AND SPATIAL SELECTION IN FLUENT READING 139
the resulting saccade will be directed partly to the location of the first word
and partly to the second. Saccade amplitude will depend on the relative
timing between the first and second attention shift (see section 3.2).
3. If the parafoveal encoding of word N+ 1 succeeds shortly after the first
saccade is being initiated, e.g. during the efferent transmission time or even
during the saccade itself, attention will shift to N+2 and elicit a second
saccade. As this saccade's amplitude will be computed during the very
beginning of the fixation on N+ 1, that fixation will be of short duration
(100 ms or less). Attention during such a brief fixation will be directed
already to word N+2.
The original model by Morrison (1984) has been revised and reformulated in an
influential textbook by Rayner and Pollatsek (1989). It should be noted that these
models can not only elegantly account for word skipping and the occurrence of very
short fixations in the way described above, but can also deliver a parsimonious
explanation for the benefit of parafoveally preprocessing word N+ 1 in terms of
saving fixation time when this word is being fixated (see Rayner, 1998, for a review
of the substantial literature on this basic issue). The assumption is that, when
attention shifts to word N+ 1 some time before the actual eye movement is executed,
lexical pre-processing can take place, leading to a shortening of the fixation on N+ 1,
relative to reading without a parafoveal preview.
As it turned out shortly after the model was published, this mechanism also
constitutes one of the major limitations of the original sequential attention shift
mechanism. Since the attention shift occurs when word N is (almost) fully
processed, it follows that the amount of parafoveal pre-processing on word N+ 1
(done during the oculomotor programming time) should be independent of any
properties of the origin word N. However, as first shown by Henderson and Ferreira
(1990) there are significant effects of the difficulty (e.g. word frequency) of word N
on the preview benefit for word N+l. If word N is of low-frequency, the preview
benefit virtually disappears (see also Kennison & Clifton, 1995, for corroborative
evidence).
A second serious limitation of the original sequential attention models is that
they do not include mechanisms to deal with refixations on the current word.
McConkie, Kerr, Reddix, Zola, and Jacobs (1989) have first shown that refixations
in reading are both a function of fixation position and word frequency. To account
for word frequency effects on refixations, Henderson and Ferreira (1990)
supplemented the SAS model with the notion of an oculomotor deadline, in which a
refixation is initiated if attention has not shifted to a new word after a certain delay.
However, this idea was rejected by O'Regan, Vitu, Radach, and Kerr (1994) and
Vitu and O'Regan (1995) who noted that, in contradiction to Henderson and
Ferreira's proposal, the first of two fixations on a word is shorter in comparison to
one fixation (Kliegl, Olson, & Davidson, 1983). A very interesting consequence of
these results is that refixations must be initiated relatively early during the initial
fixation on a word, when only limited results of lexical processing are available.
Pollatsek and Rayner (1990) discussed this possibility in relation to parallel
interactive models of word recognition (e.g. Paap, Newsome, McDonald &
Schvaneveldt, 1982) and suggested the total "level of excitation" in the lexicon in
relation to some threshold to be a candidate for this kind of early cognitive evidence.
This idea may have been elaborated later to the notion of a lexical "familiarity
check" (see below).
2.2 Contemporary Sequential Attention Shift Models
To overcome the limitations discussed above, a new key feature was introduced in
the latest version of sequential attention models: a relative decoupling of attention
and saccade programming4 allowing for considerable flexibility in the temporal
ordering of attention shift and saccade execution (Reichle, Pollatsek, Fisher, &
Rayner, 1998; Reichle, Rayner, & Pollatsek, 1999; Rayner, Reichle, & Pollatsek,
2000). Saccade programming in the new model is divided into an early, labile stage,
where saccade cancellation and reprogramming is possible and a later, non labile
stage, that does not allow for cancellation. The labile saccade program will initiate
refixations on the word as long as the initial stage in the word recognition module is
not completed, which then provides the signal to go to the next word. This
mechanism is quite similar to a proposal originally made by O'Regan (1990).
Word recognition is claimed to involve an initial stage, during which global
familiarity is established and lexical access is imminent, and a second stage during
which a specific lexical form is accessed so that meaning can be retrieved.
Completion of a fixated word's familiarity assessment triggers programming of an
interword saccade to the next word and successful access of lexical word form and
word meaning initiates a corresponding shift of attention. In this model, therefore,
the programming of an interword saccade precedes a corresponding shift of
attention; however, since programming of the saccade consumes time, the actual
execution of the saccade will often lag behind the shifting of attention, which
accounts for parafoveal preview benefits. The processing times required to complete
both stages of word processing are a function of lexical properties of the word. Their
relation, as implemented in a computational simulation, is such that the time
necessary to complete lexical access is a constant multiple of the time required for
the familiarity check. This leads to an increasing disparity between both time
components as word frequency decreases, which can account for the reduced
parafoveal preview benefit when a more difficult word is processed foveally.
3. EVIDENCE IN FAVOR AND AGAINST SEQUENTIAL ATTENTION

MODELS
The development of the E-Z reader family of eye movement control models is
certainly a major achievement. It can account for a wide range of empirical
phenomena and it is the first complex model of eye movement control in reading to
be implemented into a computational simulation. From a theoretical point of view it
should be emphasized that, while E-Z reader sees lexical processing as the "engine"
that drives the eyes across the text, it also incorporates low-level aspects like an
automatic generator for refixations (O'Regan, 1990) and in its most current version
also the metrical characteristics of landing positions (McConkie, Kerr, Reddix, &
Zola, 1988).
Like every model, E-Z reader has its limitations. One point worth noting is that
not all phenomena that the model can account for conceptually are included into the
AITENTION AND SPATIAL SELECIlON IN FLUENT READING 141
simulations. In Reichle et al. (1998) simulation results are provided for effects of
word frequency on eye movements, including skipping rates, refixations rates, single
fixation durations, first fixation durations, and gaze durations. However, the relation
between foveal processing difficulty and parafoveal preview benefit, that is at the
core of the theoretical concept of E-Z reader, is not part of the simulation, at least it
is not explicitly tested there. Of course such a test would require a sophisticated data
base including an actual variation of foveal and parafoveal processing load. Until
such a simulation study is being done, the account for fovea-on-parafovea effects is
conceptual and not computational in nature.
To summarize, sequential attention models claim that at each point in time
attention is focused at a single word, that attention is shifted to the next word only
after an attended word has been recognized and that saccade programming is solely
determined by the completion of an attended word's familiarity check. In the
following sections we will discuss to what extent these assumptions and/or
predictions that follow from them are supported by empirical data.
3.1 The Allocation ofAttention During a Fixation in Reading
3.1.1 Parafovea-on-Fovea Effects

Using a word-comparison task, Kennedy (1998, 2000) presented two words to the
right of a fixated prompt word. Two tasks were used, one which asked the reader to
decide whether the two parafoveal words were visually identical and one that asked
whether they were synonymous. The critical experimental manipulation involved
only properties of the first of the two words, which was shown to have effects on
viewing durations on the prompt word. For example, gaze durations were
significantly shorter when type trigram frequency of the parafoveal word was high,
that is, when the word shared the same initial letters with many other words.
According to Kennedy, these results suggest that a local process monitoring
mechanism operates over a region larger than the current word. The triggering of
saccades is assumed to be sensitive to the extraction rate for parafoveal sublexical
information. In the case of "difficult" parafoveal items an early saccade can be
initiated when information acquisition proceeds insufficiently rapid. The
information that is extracted parafoveally is subsequently traded off when the
neighboring word is directly fixated.
The task used by Kennedy has the advantage that oculomotor complexities
inherent to continuous reading can be reduced while still requiring words to be
lexically processed. Yet, similar types of parafovea-to-fovea cross-talk can also be
found in normal reading. In Underwood, Binns, & Walker's (2000) sentence reading
task, properties of the beginning trigram of a parafoveally visible word influenced
fixation durations. Even more striking are effects of parafoveal word meaning on
fixation durations. Using a search-type task, in which a one word difference in
meaning had to be detected between two short sentences, Murray (1998) obtained
shorter fixation durations on noun targets when the meaning of the following
(parafoveally visible) verb yielded a plausible sentence continuation than when it
yielded an implausible continuation, although the effect size was robust only when
the fixation was relatively close to the parafoveally visible word.
Our recent work (Inhoff, Starr, & Shindler, 2000) corroborates and extends these
results. Similar to Underwood et al. (2000), we used a sentence reading task and
examined effects of a parafoveally visible post-target word on target viewing
durations. A wider range of posttarget preview types was used, however, one
consisting of the base word, which was generally closely related to the fixated
target, e.g., that target traffic was followed by the posttarget base light, one
consisting of the base word in upper case,. e.g., LIGHT, one consisting of a string
of quasirandom letters, potpq, and one consisting of a semantically unrelated word,
e.g., smoke. Eye movement contingent display changes were used to show the lower
case post-target base when the eyes moved onto its location so that each sentence
constituted a meaningful unit. Examination of target viewing durations as a function
of the parafoveally visible preview type revealed longer viewing durations in the
upper case condition and, similar to Underwood et al. (2000), longer viewing
durations in the random letter preview condition than the base condition.
Furthermore, similar to Murray (1998), target viewing conditions were longer in the
base condition than in the unrelated word preview condition when the target fixation
location was relatively near to the post-target preview. Effects of post-target
meaning on target viewing were also evident in a another study (Inhoff, Radach,
Starr, & Greenberg, 2000) where readers spent less time viewing a target word, e.g.,
mother's, when the next (para foveal) word was associated with the fixated target,
e.g., father, than when it was un associated, e.g., garden.
3.1.2 Information Acquisition from the Left of the Current Fixation

According to the EZ reader family of models, attention is usually either focused at
the fixated word or at the next word in the text. However, attention may also be
focused at the word to the left of fixation. On occasion, this occurs when the eyes
moved to the right of an attended word before the previously fixated word was
recognized. When this occurs, a regressive saccade may be programmed to re-align
the loci of attention and fixation.
In agreement with this view, Binder, Pollatsek, and Rayner (1999) recently
showed that readers did obtain useful information from a word to the left of fixation
when the eyes regressed back to it. They used eye movement contingent display
changes to replace a target word with another word when the eyes moved to the
right of a target's location (regressions back to the target word resulted in another
display change that revealed the originally viewed target - so that the original target
was always visible upon its fixation). No left-of-fixation display changes occurred in
a control condition. The results showed effects of the left-of-fixation replacement on
trials in which the target was skipped. Readers regressed more often to the target
location in the experimental condition than in the control condition. Effects of the
left-of-fixation replacement were also evident when the target had been fixated
during first pass reading, regressions being again more common in the experimental
condition. Moreover, readers spent less time viewing the target during second pass
reading when the prior left-of-fixation replacement was related to sentence meaning
than when it was unrelated. In agreement with the EZ reader model, readers thus
appeared to extract meaning from the left-of-fixation replacement when the eyes
regressed back to the target, presumably because the eyes had left the target before it
had been identified.
ATTENTION AND SPATIAL SELECTION IN FLUENT READING 143
Other results indicate, however, that readers also obtain useful information from
a word to the left of fixation when the eyes do not regress back to it. Similar to
Binder et al. (1999), we (Inhoff, Radach, et aI., 2000) changed the identity of a
target word after the eyes moved to its right. E.g., the word cat in the sentence "he
saw a large cat in the .... " was replaced with the word rat, after the eyes moved to
the right of it. Sentence reading was followed by a forced choice task in which the
reader had to determine which one of three visible words had appeared in the
previously read sentence. Three choices were offered, one consisting of the target,
cat in the example, one of its left-of-fixation replacement word, rat, and one
consisting of a length- and frequency-matched word that never appeared in the
previously read sentence, e.g., hat. As expected, readers generally selected the target
word. The results also revealed a substantially higher selection rate for the left-of-
fixation replacement word than for the new word, which is consistent with Binder et
al. 's (1999) results. However, in contrast to Binder et al., supplementary analyses
showed that the higher selection rate for the left-of-fixation replacement word
occurred irrespective of whether readers regressed back to the target. Furthermore, it
appeared that acquisition of useful information to the left of fixation did not prevent
the reader from obtaining useful information to the right of fixation, suggesting that
attention is allocated to the right and left of a fixated word.
Direct evidence for the acquisition of useful information to the right and left of a
fixation word was obtained in a recently completed dissertation study in which the
visibility of text to the right AND left of a fixated word was manipulated (Starr,
under review). Four viewing conditions were created: one in which the visibility of
text to the left of fixated target was changed upon the fixation of a target word, so
that a previously visible pretarget word was changed to a quasi-random string of
letters when the eyes moved onto the target (left-of-fixation change), one in which
visibility of text to the right was manipulated, so that a previously visible posttarget
word was changed to a quasi-random string of letters when the eyes moved onto the
target (right-of-fixation change), one in which visibility to the right and left was
manipulated when the eyes moved onto the target (dual change condition). No
display change occurred in the control condition. Consistent with prior results
(Inhoff, Radach, et al., 2000; Inhoff, Starr, et al., 2000), target viewing durations
were longer in the right-of-fixation condition than in the control condition.
Similarly, the replacement of the word to the left of fixation with a string of quasi-
random letters increased target fixations, even when the eyes did not regress to the
pre target location. As expected, fixations were also longer in the dual change
condition. However, target fixation durations in this condition were equivalent to
fixation durations in the two single-side display change conditions.
The latter finding is difficult to reconcile with the EZ reader model. According to
the model, left-of-fixation display changes influenced target fixation durations when
attention was to the left of fixation but not to its right; analogously, the right of
fixation display change influenced target fixations when attention was to the right of
fixation but not to its left. That is, effects of these two types of display change
should have occurred on a subset of trials - during which the attentional spotlight
happened to be allocated at the display change location. According to the model, the
probability with which attention is allocated to the right or left of fixation must be
higher than the probability with which attention is allocated to just one side of the
target, hence effect sizes should have been higher in the dual change conditions than
the right- or left-side change conditions.
3.2 Basic Oculomotor Research and the Time Line of Events during a Fixation
As discussed in section 2.1, one core ingredient of the functional architecture of

SAS models is the concept of temporal overlap in the programming of saccades as
introduced into basic research by Becker & Jurgens (1979). This research is cited
quite frequently in the reading literature, but it appears that with the exception of
Morrison (1984) so far there has been no attempt to directly relate evidence from
basic oculomotor paradigms to eye movements in reading (see Deubel, O'Regan, &
Radach, 2000, for a detailed discussion).
In an oculomotor double-step experiment, a trial starts with the horizontal
displacement of a saccade target, usually a small dot or rectangle to an eccentric
location. Before a saccade can be executed, the target is displaced to a second
location, which is either in the same or in the opposite direction. Both amplitudes
and latencies of saccades are affected when the interval between the first and the
second step is systematically varied. A so-called bistable response will be observed
when the presaccadic shift is very large or requires a change in saccade direction. In
this case the saccade is directed either to the first or the second target location,
depending on the specific timing. There are an increased number of delayed
responses including a single saccade to the second target location, which is taken to
indicate the extra time needed for a cancellation of one saccade program and its
replacement with a program for a new saccade. It is obvious this pattern has
similarities with Morrison's (1984) word skipping scenario. Interesting in the
current context is the fact that this alteration of saccade timing can occur no later
than 120 ms before saccade onset (Deubel et aI., 2000). If an analogy to reading is
being entertained, this is an important constraint. If a word skipping mechanism in
reading would work in analogy to the bistable response mode, this should be
associated with inflated fixation durations prior to word skipping. Seen from this
perspective, finding that fixation durations before word skipping are not prolonged,
does not exclude the possibility that word skipping is based on saccade cancellation
and reprogramming, but it would suggest that an analogy to a bistable response is
not justified. 5
The second response mode in a double step experiment, the averaging response
can be observed when the second target step is over a relatively short distance and
goes into the direction of the first step. Obviously, this situation is more similar to a
sequence of progressions in normal reading. If the second step takes place during an
interval from approximately 180 to 70 msec before saccade onset, saccades tend to
land at positions between first and second target location, suggesting that a spatial
average of target eccentricity is calculated within a particular temporal window.
With increasing time available for this process, the saccade will systematically land
closer to the second, final target position. Such averaging saccades in double step
paradigms are often followed by short-latency secondary saccades directed to the
final target position. 6 Both the proportion of two-saccade responses and the
amplitude of the primary saccade are a function of the time interval between the
occurrence of the second target and the onset of the primary saccade. This means,
ATIENTION AND SPATIAL SELECTION IN FLUENT READING 145
saccade modification takes place when enough time is available during the latency
period, no matter what the total latency is (Becker, 1989; Becker & Jiirgens, 1979;
Deubel, Wolf, & Hauske, 1984).
We have carried out a double-step experiment using step eccentricities of two
degrees that are quite similar to average amplitudes of reading saccades (see Findlay
& Harris, 1984 for a similar approach). Figure 1 presents some data that are relevant
for our discussion. In the left panel, a typical amplitude transition function for
uncrossed steps (2 two degree steps either to the left or to the right) is shown. More
importantly, the right panel depicts the proportion of one-saccade responses as a
function of the available reprocessing time. If an analogy to reading is assumed,
such skip-over responses can be seen as a model for "word skipping" (Morrison,
1984). As our data indicate, skip-over responses remain very unlikely for
reprocessing times below 90 ms, and most skipping cases are associated with much
higher reprocessing intervals. Thus, the minimal value of 70 ms, suggested by
Deubel et al. (2000) on the basis of experiments with much larger step sizes needs to
be corrected for more reading-like conditions. Assuming a minimal reprocessing
time of 90 ms is also in line with McConkie, Underwood, Zola, and Wolverton's
(1985) estimate of 80-100 ms before saccade onset as the deadline for visual
stimulus influence.
2 Saccade Amplitude Differem."C 1 Proportion of One·Saccade Responses
1,5 0.8
0.61- ......................................................... ............ ,,'j.\
0.5 0.41-................................. .....................,,/ I
0.2 r············································ ·,/······7""'············1
-OSb'-o.6=-9---=7:::'-0.=89--'9=0"7.10::::9--:-11:-:0.7.12=-9-1:-::-::30.149 20~.6~9===:::7~0~.89;:::::::90;,.1~09=---1c:-1O="-.1=-=-29=---cIJ
30 • 149
Reprocessing Time Reprocessing Time
. 2 left steps*2 right steps +21eft steps,*2 right steps
Figure 1. Results of uncrossed conditions in a double-step experiment with either two

consecutive steps to the left or two steps to the right. The lSI between the two steps varied in
50 ms intervals between 50 and 150 ms. Reprocessing time is the time interval between the
occurrence of the second saccade target and the onset of the primary saccade. Left panel:
Saccade amplitude as a function of reprocessing time, the zero value denotes a 2 deg. saccade
to the location of the first step. Right panel: Proportion of one-saccade (skip-over) responses
as a function of reprocessing time. A veraged data of 8 participants.
The basic research discussed above has important consequences for possible time
lines of processing events during fixations in reading. Several such time lines have
been put forward in the literature (e.g. Blanchard, McConkie, Zola, & Wolverton,
1984; Russo, 1978; McConkie, 1983; Posner & Abdullaev, 1999; Posner, Abdullaev,
McCandliss, & Sereno, 1999; Sereno, Rayner, & Posner, 1998). We will restrict our
discussion to some points made by Sereno et al. (1998). As a starting point of their
time line they postulate a normative fixation duration of 300 ms. This estimation is
the result of assuming an average fixation duration of 250 ms and excluding possible
fixation duration savings due to a prior parafoveal processing of the fixated word.
Consistent with other published time line estimates it is assumed that, as a first step,
it takes about 60 ms for the visual information on the retina to get to higher brain
centers. Sereno et al. (1998) report ERP results in a word recognition paradigm
indicating that, starting at 132 ms after stimulus onset, word frequency differences
can be observed. They further estimate a shift of attention and the initiation of an
eye movement to take place at 150 ms. They also suggest that "oculomotor latency
is estimated to be around 150 ms (Rayner, Slowiaczek, Clifton, & Bertera, 1983)",
and that, once a signal is given to move the eyes, "about 20 ms elapses before the
eye muscles are activated and the saccade begins." (p. 2197).7
The proposed time line appears to work fine for inter-word saccades from word
N to word N+ 1. However, things are getting complicated, once the case of a word
skipping saccade from word N to word N+2 is being considered. If an attention shift
to word N+ 1 takes place at 150 ms, this word could be lexically processed starting
after another 130 ms and the eye movement to N+2 be triggered after another 150
ms. Hence there would be effectively no time left for computing the actual
movement. To make things even worse, " ... at the right end of the time line,
oculomotor latency (the time needed to program an eye movement) limits the
interval during which a sufficient degree of lexical processing must be achieved."
(Sereno et al., 1998, p. 2197). As we have seen above, a mechanism for skipping
word N+1 in analogy to the averaging response mode in double step paradigms puts
the latest possible time at which new information can modify a saccade amplitude at
about 90 ms (or, according to Deubel et al., 2000, at 70 ms) before saccade onset. s
Even assuming that word skipping occurs in a way similar to the bistable response
mode would not help, as the 50 ms extra fixation time on word N that might be
reserved in this case (350 ms fixation duration) would be eaten up by the fact that
the latest point for saccade modification (cancellation and reprogramming) for the
bistable mode is 120 ms until saccade onset.
One additional point to be noted is that in the above considerations it is tacitly
assumed that the "movement of attention", necessary to start lexical processing on
word N+1, does not require any time for "programming" and "execution". However,
as noted by Morrison (1984), shifting attention does require time. According to
Eriksen (1990; cited in Kinchla, 1992, p. 726.),50 ms would be a possible estimate
of a latency between the arrival of the trigger signal from lexical processing and the
actual execution of a shifting of an attentional focus. Taken together, a close look at
the temporal constraints for various processing events during reading fixations poses
serious problems for the time line proposed by sequential attention models,
especially when more than one shift of attention needs to be accommodated.
A possible solution to these problems might be to claim that word skipping via
attention shift, saccade cancellation and reprogramming only takes place when word
N is very easy to process and, hence, requires much less than 150 ms of processing
time until saccade triggering. However, as reported by Sereno et ai. (1998), the
earliest point at which lexical processing (word frequency effects) shows up in their
EEG-analysis was around 130 ms, which leaves not much room for shifting
processing events towards the beginning of the time line. Even if both attention
shifts (or, for the E-Z reader model, the shift from N to N+1 and the familiarity
check on N+ 1) would be finished each within 130 ms, it would still be very hard to
squeeze in all necessary operations within the time until the saccade programming
deadline (see Brysbaert & Vitu, 1998, for a similar argument). To restrict the
application of the sequential attention logic to cases with extremely easy N+ 1 word
would also mean to give up the claim to account for every case of word skipping, as
the phenomenon can be regularly observed with difficult and rare words. A more
principal solution could be offered by allowing some processing of word N+ 1 to
take place in parallel of processing word N. This possibility will be elaborated in
some detail in the remaining two sections of this chapter.
4. THEORETICAL ALTERNATIVES
It may be possible that modifications of SAS models will overcome the empirical
challenges noted in section 3. Alternatively, these empirical challenges may require
revisions of the models' core assumptions. We will consider the promise of two
alternative conceptions in the following section, one that is concerned with the
spatial allocation of attention and one that is concerned with the programming of
sequences of saccades.
4.1 The Attentional Gradient Hypothesis
LaBerge and Brown (1989; see also LaBerge, Brown, Carter, & Bash, 1991)
proposed an alternative to the spotlight conception of visual attention allocation.
According to their view, attentional resources are generally distributed over several
spatially adjacent units. Viewers are assumed to estimate the importance of the
elements of a visual array within the perceptual span and use it for the allocation of
attention. Importantly, resources are allocated in a graded manner, so that the bulk of
the resources is allocated to the element with the highest importance value, and
progressively fewer resources are allocated to adjacent units.
If this conception of attention allocation was extended to reading, it would imply
that readers can allocate resources to more than one word within the perceptual span.
More than one word can be attended to concurrently and a word may be attended to
more than once, i.e., before, while, and after it is fixated. This conception of
attention allocation can account for the parafovea-on-fovea effects described in
section 3.1. In the experimental conditions described there, some attention may have
been allocated to locations to the left and right of a fixated word so that useful
orthographic information could be obtained.
LaBerge and Brown's (1989; LaBerge et aI., 1991) attentional gradient model
considers neither dynamic shifts of spatial attention that may occur during a fixation
nor the coordination of attention shifts with saccade programming. The time line of
events during a fixation, described in section 3.2, appears to favor, however, the
gradient model, as it does not adhere to the notion of strictly serial word recognition
process.
4.2 Visual Selection Based on a Spatial Saliency Map
Although the attentional gradient idea criticizes. and modifies key claims of
sequential attention shift models, it still refers to the concept of attention to account
for eye movement control in reading. In this section, a more radical departure from
current attention-based theories will be sketched that may provide an alternative
account for many of the phenomena described above.
One important origin of the framework to be described is the fact that in normal
reading word targeting decisions are co-determined by low-level visual factors and
linguistic processing. In a computational simulation, very simple word targeting
strategies like "fixate the largest word within a 20 letter window from the current
fixation" can do a reasonable job in producing eye movements quite similar to those
observed in human reading (Reilly & O'Regan, 1998). However, there is also no
doubt that lexical (and to some degree also semantic and syntactic) processing has a
significant immediate influence on eye movement control (e.g. Rayner, Sereno, &
Raney, 1996). This suggests a modeling approach where spatial and temporal
control decisions, rather than depending exclusively on lexical processing, are
governed by a combination of low level and cognitive factors (Reilly & Radach, in
press).
The specific idea is derived from recent developments in basic oculomotor and
neurophysiological research. Findlay & Walker (1999) have proposed a new
theoretical framework for saccade generation in which the selection of saccade
targets is based on parallel processing and competitive inhibition within a two-
dimensional salience map. At each point in time, multiple potential saccade targets
are assumed to be represented and to compete within a two-dimensional spatial
salience map. Mter a period of accumulation of information and competitive
inhibition between targets one emerges as a winner, and a saccade to the
corresponding location is initiated.
In the case of reading, spatial selection takes the form of deciding which of the
words within the current perceptual span, including the one currently fixated, should
be the goal for the next saccade. The most important low-level sources of influence
on this decision are the length and the eccentricity of words located around the
current point of fixation (Kerr, 1992; McConkie, Kerr, & Dyre, 1994). It can be
assumed that the potential target words are represented in a two-dimensional
salience map and, depending on the particular visual configuration, their salience
values form a preference list of potential targets. For example, when a long word N
is first fixated near its end, the initial saliency value (assumed to be a function of
distance to the word center) for this word will be high, making a refixation on N
likely. If the next word to the right, word N+1, is short, it may receive a lower
saliency value than the next, longer word N+2 and the initial decision preference for
the next inter-word saccade will be to go to word N+2. Note that there is no default
saccade program to N+ 1 and, hence, no saccade cancellation, reprogramming or
"word skipping". There is also no principal difference between "foveal" and
"parafoveal" words. Instead, the potential word targets N, N+1 and N+2 compete
within a unified spatial representation.
The initial saliency values provided by low-level visual analysis at the beginning
of each fixation will subsequently be modified on the basis of linguistic processing.
The saliency map can be thought of as an integrator where visual and cognitive
information can be combined into a single preference value for each target. As
discussed by Radach (1999), cognition may influence salience values, and hence the
selection of saccade targets, in different ways. There is ample evidence in support of
direct cognitive influences, with many experiments demonstrating immediate
feedback of cognitive processing to the saccade generation system during the current
fixation. As an example, Rayner et al. (1996) showed that word frequency (as a
measure of lexical processing difficulty) has a significant influence on all fixation
duration measures and on the decision to execute a saccade to word N+1 vs. N+2.
Similar studies including manipulations of word frequency and contextual
predictability have recently been discussed in an interesting meta-analysis by
Brysbaert & Vitu (1998). Although the bulk of the variance for saccade targeting
decisions can be explained in terms of low level visuomotor factors, there remains a
significant contribution from immediate cognitive processing.
Although still somewhat speculative, there is also some evidence for possible
indirect cognitive influences on saccade targeting decisions. In analogy to a
suggestion by Findlay & Walker that long term implicit learning and memory playa
significant role, it is possible that readers produce an estimation of how likely the
successful parafoveal recognition of a word will be, given the length and
eccentricity of this next word. The decision of whether to fixate or skip a subsequent
word could then be based on this type of "educated guessing" (Brysbaert & Vitu,
1998). A psychophysical function that specifies the perceptibility of eccentric words
could be derived empirically in tachistoscopic recognition experiments (Brysbaert,
Vitu, & Schroyens, 1996) or by using techniques that measure the spatial extent of
word recognition more directly in a reading situation (McConkie & Hogaboam,
1985; McConkie & Zola, 1987).
There are still open questions about the saliency map framework. For example,
the relation and level of dependency between spatial aspects (saccade target and
amplitUde) and temporal decisions (fixation duration) needs to be specified. Also,
given that linguistic processing of a word will generally result in a reduction of its
saliency, it needs to be investigated which conditions lead to an increase of saliency
for parafoveal words. This is likely to be the case for words with orthographically
irregular beginnings (Radach, Inhoff, & Heller, 2000), but could also include other
conditions mentioned in our discussion of parafoveal-to-foveal effects.
In sum, the saliency map conception provides an attractive modeling framework,
as it appears to have the potential for a unified account for many oculomotor
phenomena such as refixations, word skipping, foval-on-parafoveal and parafoveal-
on-foveal effects (Reilly & Radach, in press). The proposition advocated here has
similarities with the suggestion of Henderson & Hollingworth (1998), in the domain
of picture perception, that ongoing cognitive processing might modify saliency of
potential saccade targets. However, the question is whether much is gained by their
assumption that "attention" is disengaged, moved and re-allocated as a function of
saliency, and that these processes, in tum, trigger saccade programming (Findlay &
Walker, 1999). We prefer to see attention-based vs. salience map models as two
alternative explanatory principles, rather than combining them into an even more
complex framework. Future empirical research and modeling will show whether eye
movement control in reading can be best explained by a modified attention-based
theory or whether parallel processing and competitive inhibition within a two-
dimensional salience map will provide a better account.
5. ACKNOWLEDGMENTS
The authors gratefully acknowledge the help of Stefanie Lemmer and Catharina Hau
in preparing the manuscript.
6. AFFILIATIONS
Dr. Ralph Radach and Dr. Dieter Heller

Technical University ofAachen
Dr. Albrecht Inhoff
State University of New York at Binghamton
Correspondence should be addressed to
Dr. Ralph Radach,
Institute of Psychology, Technical University of Aachen,
liigerstr. 17,52064 Aachen,
e-mail: ralph.radach@psych.rwth-aachen.de.
7. NOTES
1 A more general perspective on eye movements and visual information processing in reading is provided
in the chapter by Starr, Kambe, Miller & Rayner, this volume.
2 The term "parallel programming" should not be taken to suggest that identical programming operations
on two saccades can be performed at the same time. In this context, it refers to the fact that there can be
temporal overlap in the successive programming of saccades.
J It is interesting to note that, according to Morrison, "extrafoveal attentional allocation may have to exist
for a minimal time in order to make an eye movement irrevocable and to begin the amplitude
computation stage." (p. 679). In other words, shifting attention takes time. This point is not part of later
versions of attention shift models.
4 We will use the term "saccade programming" when referring to the parameterization of saccades in
general. By "saccade initiation" we will refer to a stage where an imminent saccade has passed a point
of no return and can no longer by delayed by linguistic information (McConkie, Underwood, Zola &
Wolverton, 1985).
S The evidence on whether fixation durations before word skipping are prolonged is equivocal. While
Pollatsek, Rayner, and Balota (1986) and Reichle, Pollatsek, Fischer, and Rayner (1998) have found
such an effect, there are also studies that could not replicate this result (McConkie, Kerr & Dyre, 1994;
Radach & Heller, 2(00).
6 It follows that word skipping in analogy to an avarage response should be associated with an inflation of
very short fixation durations. Radach, Heller, and Inhoff (1999) investigated the occurrence of short
reading fixations in the range between 80 and 120 ms. In a sample of 24 participants they found
proportions of such fixations from 0.5 to 12 percent. This extremely uneven distribution among
individual readers raises doubts about averaging responses as the default mechanism of word skipping.
7 The operations denoted as "shift attention" and "initiate eye movements" are suggested to take place at
the same time, although the E-Z reader model, claiming their temporal decoupling, is being referred to
ATIENTION AND SPATIAL SELECfION IN FLUENT READING 151
(see section 3.2). Since this difference is not critically important for our discussion, we will also refer to
the simpler mechanism described in Sereno et al. (1998).
8 The possible argument that some of this time is used for afferent transmission of visual information
from the retina applies to the results of lexical processing as well, as these also need to be transported
and transferred into a motor command.
7. REFERENCES
Allport, A. (1992). Attention and control: have we been asking the wrong questions? A critical review of
twenty-five years. In D. E. Meyer, & S. M. Kornblum (Eds.), Attention and performance XIV:
Synergies in experimental psychology, artificial intelligence, and cognitive neuroscience (pp. 183-
218). Cambridge, MA: MIT Press.
Becker, W. (1989). Metrics. In R. H. Wurtz, & M. E. Goldberg (Eds.), The neurobiology of saccadic eye
movements (pp. 13-61). Amsterdam: Elsevier.
Becker, W., & Jiirgens, R. (1979). An analysis of the saccadic system by means of double step stimuli.
Vision Research, 19, %7-983.
Binder, K. S., Pollatsek, A., & Rayner, K. (1999). "Extraction of a fixated word to the left of a fixation in
reading." Journal of Experimental Psychology: Human Perception and Performance, 25, 1162-1172.
Blanchard, H. E., McConkie, G. W., Zola, D., & Wolverton, G. S. (1984). The time course of visual
information utilization during fixations in reading. Journal of Experimental Psychology. Human
Perception and Performance, 10, 75-89.
Brysbaert, M., & Vitu, F. (1998). Word skipping: implications for theories of eye movement control in
reading. In G. Underwood (Ed.), Eye Guidance in reading and scene perception (pp. 125-148).
Amsterdam: Elsevier.
Brysbaert, M., Vitu, F., & Schroyens, W. (1996). The right visual field advantage and the optimal
viewing position effect: on the relation between foveal and parafoveal word recognition.
Neuropsychology, 10, 3, 385-393.
Deubel, H., O'Regan, K., & Radach, R. (2000). Attention, information processing and eye movement
control. In A. Kennedy, R. Radach, D. Heller, & 1. Pynte (Eds.), Reading as a perceptual process
(pp. 355-374). Oxford: Elsevier Inc.
Deubel, H., Wolf, W., & Hauske, G. (1984). The evaluation of the oculomotor error signal. In A.G. Gale,
& F. Johnson (Eds.), Theoretical and applied aspects of eye movement research (pp. 55-62).
Amsterdam: Elsevier.
Findlay, J. M., & Harris, L. R. (1984). Small saccades to double-stepped target moving in two
dimensions. In A. G. Gale, & F. Johnson (Eds), Theoretical and applied aspects of aye movement
research (pp. 71-78). Amsterdam: Elsevier.
Findlay, J. M., & Walker, R. (1999). A model of saccade generation based on parallel processing and
competitive inhibition. Behavioral and Brain Sciences, 22, 661-721.
Henderson, 1. M., & Ferreira, F. (1990). Effects of foveal processing difficulty on the perceptual span in
reading: Implications for attention and eye movement control. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 16, 417-429.
Henderson, J. M., & Hollingworth, A. (1998). Eye movements during scene viewing: an overview. In G.
Underwood (Ed.), Eye guidance in reading and scene perception (pp. 269-293). Oxford: Elsevier
Science.
Inhoff, A. W., Radach, R., Starr, M., & Greenberg, S. (2000). Allocation of visuo-spatial attention and
saccade programming in reading. In A. Kennedy, R. Radach, D. Heller, & J. Pynte (Eds.), Reading as
a perceptual process (pp. 221- 246). Oxford: Elsevier Inc.
Inhoff, A. W., Starr, M., & Shindler, K. (2000). Is the processing of words during eye fixations in reading
strictly serial? Perception and Psychophysics, 62, 1474-1484.
Johnston, W. A., & Dark, V. 1. (1986). Selective attention. Annual Review of Psychology, 37, 43-75.
Kennedy, A. (1998). The influence of parafoveal words on foveal inspection time: Evidence for a
processing tradeoff. In G. Underwood (Ed.), Eye guidance in reading and scene perception (pp. 149-
180). Oxford, England: Elsevier.
Kennedy, A. (2000). Attention allocation in reading: sequential or parallel? In A. Kennedy, R. Radach, D.
Heller, & J. Pynte (Eds.), Reading as a perceptual process (pp. 193-220). Oxford: Elsevier.
Kennison, S. M., & Clifton, C. (1995). Determinates of parafoveal preview benefit in high and low
working memory capacity readers: Implications for eye movement control. Journal of Experimental
Psychology: Learning, Memory and Cognition, 21, 68-81.
Kerr, P. W. (1992). Unpublished Doctoral Dissertation. University of Illinois. Urbana-Champaign.
Kinchla, R. A. (1992). Attention. Annual Review of Psychology, 43, 711-742.
Kliegl, R., Olson, R. K., & Davidson, B. 1. (1983). On problems of unconfounding perceptual and
language processes. In K. Rayner (Ed.), Eye movements in reading and perceptual and language
processes (pp. 333-343). New York: Academic Press.
LaBerge, D., & Brown, V. (1989). Theory of attentional operations in shape identification. Psychological
Review, 96, 101-124
laBerge, D., Brown, V., Carter, M., & Bash, D. (1991). Reducing the effects of adjacent distractors by
narrowing attention. Journal of Experimental Psychology: Human Perception and Performance, 17,
65-76.
McConkie, G. W. (1979). On the role and control of eye movements in reading. In P. A. Kolers, M. E.
Wrolstad, & H. Bouma (Eds.), Processing of visible language. Vol. 1 (pp. 37-48). New York: Plenum
Press.
McConkie, G. W. (1983). Eye movements and perception during reading. In K. Rayner (Ed.), Eye
movements in reading: Perceptual and language processes (pp.65-96). New York: Academic Press.
McConkie, G. w., & Hogaboam T. W. (1985). Eye position and word identification during reading. In R.
Groner, G. W. McConkie, & C. Menz (Eds.), Eye movements and human information processing (pp.
159-172). North-Holland: Elsevier Science Publishers.
McConkie, G. W., Kerr, P. W., & Dyre, B. P. (1994). What are "normal" eye movements during reading:
Toward a mathematical description. In J. Ygge, & G. Lennerstrand (Eds.), Eye movements in reading
(pp. 315-327). Oxford: Elsevier.
McConkie, G. W., Kerr, P. W., Reddix, M. D., & Zola, D. (1988). Eye movement control during reading:
I. The location of initial eye fixation on words. Vision Research, 28, 1107-1118.
McConkie, G. W., Kerr, P. W., Reddix, M. D., Zola, D., & Jacobs, A. M. (1989). Eye movement control
during reading: II. Frequency of refixating a word. Perception and Psychophysics, 46, 245-253.
McConkie, G. W., Underwood, N. R., Zola, D., & Wolverton, G. S. (1985). Some temporal characteristics of
processing during reading. Journal of Experimental Psychology: Human Perception and Performance,
11, 168-186.
McConkie, G. W., & Zola D. (1987). Visual attention during eye fixations while reading. In M. Colt heart
(Ed.), Attention and Performance Xll, The psychology of reading (pp. 385-401). East Sussex:
Erlbaum.
Morrison, R. E. (1984). Manipulation of stimulus onset delay in reading: Evidence for parallel
programming of saccades. Journal of Experimental Psychology: Human Perception and
Performance, 10, 667-682.
Murray, W. S. (1998). Parafoveal pragmatics. In G. Underwood (Ed.), Eye guidance in reading and scene
perception (pp. 181-200). Oxford, England: Elsevier.
Neuman, O. (1987). Beyond capacity: A functional view of attention. In H. Heuer, & A. F. Sanders
(Eds.), Perspectives on selection and action (pp. 361-394). Hillsdale, NJ: Erlbaum.
O'Regan, 1. K. (1990). Eye movements and reading. In E. Kowler (Ed.), Reviews of oculomotor research:
Vol.4. Eye Movements and Their Role in Visual and Cognitive Processes (pp. 395-453). Amsterdam:
Elsevier.
O'Regan, 1. K., Vitu, F., Radach, R., Kerr, P. (1994). Effects of local processing and oculomotor factors
in eye movement guidance in reading. In J. Ygge, & G. Lennerstrand (Eds.), Eye movements in
reading (pp. 329-348). New York: Pergamon Press.
Paap, K. R., Newsome, S. L., McDonald, 1. E., & Schvaneveldt, R. W. (1982). An activation-verification
model for letter and word recognition: The word-superiority-effect. Psychological Review, 89, 573-
594.
Pollatsek, A., & Rayner, K. (1990). Eye movements and lexical access in reading. In D. A. Balota, G. B.
Flores d' Arcais, & K. Rayner (Eds.), Comprehension processes in reading (pp. 143-164). Hillsdale,
NJ: Erlbaum.
Pollatsek, A., Rayner, K., & Balota, D. A. (1986). Inferences about eye movement control from the
perceptual span in reading. Perception and Psychophysics, 40, 123-130.
Posner, M. I. (1980). Orienting of attention. Quarterly Journal of Experimental Psychology, 32, 3-25.
Posner, M. I., & Abdullaev, Y. G. (1999). Neuroanatomy, circuitry and plasticity of word reading. Neuro
Report, 10, RI2-R23.
AITENTION AND SPATIAL SELECTION IN FLUENT READING 153
Posner, M. I., Abdullaev, Y. G., McCandliss, B. D., & Sereno, S. C. (1999). Anatomy, circuitry and
plasticity of word reading. In J. Everatt, (Ed.), Reading and dyslexia (pp. 137-162). London:
Routhledge.
Posner, M. I., & Petersen, S. E. (1990). The attention system of the human brain. Annual Review of
Neuroscience, 13, 25-42.
Radach, R. (1999). Top-down influences on saccade generation in cognitive tasks. Behavioral and brain
sciences,22,697-698.
Radach, R., & Heller, D. (2000). Relations between spatial and temporal aspects of eye movement
control. In A. Kennedy, R. Radach, D. Heller, & J. Pynte (Eds.), Reading as a perceptual process
(pp. 165-192). Oxford: Elsevier.
Radach, R., Heller, D., & Inhoff, A. W. (1999). Occurrence and function of very short fixation durations
in reading. In W. Becker, H. Deubel, & T. Mergner (Eds.), Current oculomotor research:
Physiological and psychological aspects (pp. 321-331). New York: Plenum Publishers.
Rayner, K (1998). Eye movements in reading and information processing: 20 years of research.
Psychological Bulletin, 124, 372-422.
Rayner, K, & Pollatsek, A. (1989). The psychology of reading. Boston: Prentice-Hall.
Rayner, K:, Reichle, E. D., & Pollatsek, A. (2000). Eye movement control in reading: updating the E-Z
Reader model to account for initial fixation locations and refixations. In A. Kennedy, R. Radach, &
D. Heller (Eds.), Reading as a perceptual process (pp. 701-719). Oxford: Elsevier.
Rayner, K, Sereno, S. C., & Raney, G. E. (1996). Eye movement control in reading: A comparison of
two types of models. Journal of Experimental Psychology: Human Perception and Performance, 22,
1188-1200.
Rayner K, Slowiaczek, M. L., Oifton, C., & Bertera, J. H. (1983). Latency of sequential eye movements:
Implications for reading. Journal of Experimental Psychology: Human Perception and Performance, 9,
912-922.
Reichle, E. D., Pollatsek, A., Fisher, D. L., & Rayner, K (1998). Toward a model of eye movement
control in reading. Psychological Review, 105, 125-157.
Reichle, E. D., Rayner, K, & Pollatsek, A. (1999). Eye movement control in reading: accounting for
initial fixation locations and refixations within the E-Z Reader model. Vision Research, 39, 4403-
·4411.
Reilly, R. G., & O'Regan, 1. K (1998). Eye movement control during reading: A simulation of some
word-targeting strategies. Vision Research, 38, 303-317.
Reilly, R. G., & Radach, R. (in press). Glenmore: An interactive activation model of eye movement
control in reading. In J. Hyona, R. Radach, & H. Deubel (Eds.), The mind's eye: Cognitive and
applied aspects of oculomotor research. Oxford: Elsevier Science.
Russo, J. E. (1978). Adaptation of cognitive processes to the eye movement system. In J. W. Senders, D.
F. Fisher, & R. A. Monty (Eds.), Eye movements and the higher psychological functions (pp. 89-
112). Hillsdale, N. J.: Erlbaum.
Sereno, S. c., Rayner, K, & Posner, M. I. (1998). Establishing a timeline of processing during reading:
Evidence from eye movements and event-related potentials. NeuroReport, 9, 2195-2200.
Underwood, G., Binns, A., & Walker S. (2000). Attention demands on the processing of neighbouring
words. In A. Kennedy, R. Radach, D. Heller, & J. Pynte (Eds.), Reading as a perceptual process.
Oxford: Elsevier.
Vitu, F., & O'Regan, 1. K (1995). A challenge to current theories of eye movements during reading. In
J.M. Findlay, R.W. Kentridge, & R. Walker (Eds.), Eye movement research: Mechanisms, processes
and applications (pp. 381-392). Amsterdam: Elsevier.
Yantis, S. (1998). Control of visual attention. In H. Pashler (Ed.), Attention (pp. 223-256). East Sussex:
Psychology Press.
M. OSAKA & N. OSAKA
THE EFFECT OF FOCUSING ON A SENTENCE IN

JAPANESE READING SPAN TEST
Abstract. The effect of focusing on a target word during the performance of a Japanese version of the
reading span test (J- RST) was investigated. A focus word in RST was defined as the most important
word with a core meaning necessary to integrate the sentence. Two kinds of RST were performed, one
was focused RST in which the target word to be maintained was a focus word of the sentence. The other
was a non-focused RST in which the target word was not a focus word of the sentence, although the
sentence did contain a focus word. Script difference between kanji and kana was also examined; half of
the target words were kanji nouns and the other half were kana nouns. RST score and total correct
response higher for the focused RST than for non- focused RST. This result showed that focusing had an
advantage in maintaining the target words and also accelerated the reader's ability to integrate the
sentence. A script difference was found only in non-focused RST and not in focused-RST, suggesting a
strategic differences in the maintenance method between high span readers and low span readers. The
word length effect is also discussed in relation to the focus effect.
1. INTRODUCTION
1.1 Working Memory and Text Comprehension
Working memory refers to the immediate brain processes involved in the
simultaneous storage and processing of information and plays an important role in
complex cognition such as language comprehension, learning and reasoning
(Baddeley, 1986; Just & Carpenter, 1992).
Working memory plays a particularly critical role in comprehension processes
during text reading. For example, incoming information is decoded perceptually,
reorganized, and integrated with contextual interpretation while the products of
these processes are stored for a short period (Daneman & Carpenter, 1980; Kintsch
& Van Dijk, 1978). At that time, working memory is important in storing the
intermediate and final products of successive data, and allows the reader to integrate
the contents and develop a context from the text words.
There are individual differences in working memory capacities, and differences
in working memory account for many aspects of language comprehension (Just &
Carpenter, 1992). To investigate the involvement of the working memory system in
reading comprehension, Daneman and Carpenter (1980) developed a reading span
test (RST). In the RST, subjects have to read a few sentences aloud while
maintaining the last word of each sentence. Thus, RST measures both processing
and storage functions during reading. Daneman and Carpenter (1980) found a
significant correlation between reading comprehension and RST span score.
Subjects with large working memory capacities (high span subjects in RST) were
successful in interpreting the meaning of an ambiguous word when it was distant
from the portion of the text necessary to clarify its meaning. However, subjects with

@2002 Kluwer Academic Publishers.
156 M. OSAKA & N. OSAKA
small working memory capacities (low span subjects) have difficulty in maintaining
the target words due to insufficient working memory capacity during reading. A
significant correlation between RST and reading comprehension was found
(Daneman & Carpenter, 1980), but it showed a lower (and not significant)
correlation between reading comprehension and short-term memory task.
Some hypotheses have been proposed to explain which of the processes required
in RST influence the comprehension difference between high and low span readers
of working memory. Daneman & Carpenter (1983) suggest that a semantic
processing is attributable to differences between capacities; high span readers devote
fewer resources to the semantic processing of sentences, therefore, they still have
sufficient resources to maintain the words. It was also suggested that high-span
subjects were not only quicker to process linguistic material, but they also made
greater use of various strategies (Carpenter & Just, 1989).
Other researchers suggest an inhibitory mechanism hypotheses; low span readers
have a deficit in inhibiting irrelevant information (Conway & Engle, 1994; Engle,
Conway, Tuholski, & Shisler, 1995; De Beni, Palladino, Pazzaglia, & Cornoldi,
1998).
In reading sentences, however, it is required for readers not only to inhibit
irrelevant information but also to decide which information is important. When
subjects determine the meaning of each word in a sentence, they decide if the word
is relevant or irrelevant to integrate the sentence. When readers read a sentence, they
always search for a focus word that provides an advantage to integrate the whole
sentence automatically. Therefore, readers pay dynamic attention to finding focus
words and exhaust working memory capacity. High span readers are able to give
attention to the search for a focus word, and still have sufficient resources to
maintain the target word. Low span subjects can find the focus word, but the amount
of attention resources required to do this does not leave sufficient resources to
maintain the target word. Therefore, when the number of sentence increases, low
span readers can not remember words, causing their RST score to deteriorate. Even
during text reading, low span readers sometimes forget focus words, therefore they
are not good at comprehending the context of the text.
In the RST, if the target word that subjects were required to remember is a focus
word of the sentence, the sentence comprehension processes accelerated maintaining
the word. However, if the target word is not a focus word of the sentence, the
comprehension processes do not facilitate maintaining the word. Therefore, subjects
pay more attention to the word, resulting in a decrease in attention resources in the
central executive.
In this experiment, we investigated the focusing effect on the RST scores using
Japanese sentences. Moreover, to measure both quantitative and qualitative
differences between the focus and non-focus RST, we compared the performance
between two different Japanese scripts; that is the phonological scripts of kana and
logographical scripts of kanji. To this aim, half of the target words of the sentence
used in each RST were kanji nouns and the other half were kana nouns.
THE EFFECI' OF FOCUSING ON A SENTENCE IN JAPANESE READING SPAN TEST 157
2. METHOD
2.1 Japanese Reading Span Test
In Japanese orthography, there are two kinds of written symbols, kanji and kana.
There are 48 mora-based kana and thousands of kanji. Kana scripts are alphabet-like
phonetic symbols for mora while kanji scripts are essentially logographic symbols
representing lexical morphemes with dense meaning.
The Japanese reading span test (J-RST) (Osaka & Osaka, 1992; Osaka & Osaka,
1994) was used to measure working memory capacity during text reading. In the
English version, the target word was the last word of each sentence (Daneman &
Carpenter, 1980). In the Japanese version, however, the target word was the
underlined word in each sentence. This is because the last word of the sentence in
Japanese is usually a verb, therefore the underlined word of each sentence was
designated as the target (Osaka & Osaka, 1992). Subjects were asked to read a few
sentences aloud and remember the underlined words in each sentence. They were
requested to report the words immediately after reading the set of sentences.
High correlation was found between J-RST and the English version (Osaka &
Osaka, 1992). Moreover, significant correlation was found between the span score
and comprehension test scores (Osaka & Osaka, 1994).
2.2 Focus RST and Non-Focus RST
Two hundred forty sentences in Japanese were selected, each of which contained
about 26-30 characters. All of the sentences contained more than two nouns, one of
which was a kanji noun, while the other was a kana noun. Each kind of noun
selected is written usually in the chosen script, that is, kanji nouns were selected
from those usually written in kanji while kana nouns were selected from those
usually written in kana. Therefore, the familiarity of both kanji and kana nouns were
high. The ability to visualize the noun was also high for both kanji and kana. The
character lengths for both kanji and hiragana nouns were 2 to 4 characters.
One hundred undergraduate students asked to choose a focus word from each of
the 240 sentences. When they read a sentence, they were requested to judge which
word was the most important and had a core meaning necessary to understanding the
sentence. They wrote the word after deciding on the focus word.
Focus words were identified when more than 70 % of subjects selected that word
as a focus word. As a result, 140 sentences were selected with a focus word in each
sentence.
Using these sentences, two kinds of sentences were constructed; one was the
focus sentence and the other was non-focus sentence. In a focus sentence, the target
word the subject was asked to remember in RST was a focus word of the sentence.
In a non-focus sentence, however, the target word was not a focus word but another
noun.
Table 1 shows a sample sentence as a focus sentence and non -focus sentence. In
the sentence of "A TV program reported many kinds of nutrients are contained in
seaweed", the word "seaweed" was selected as a focus word by more than 70 % of
students. Therefore, in the focus sentence, "seaweed" was selected as the target
word. On the contrary, in the non-focus sentence, another noun "nutrients" was
selected as a target word. In the latter case, the target word (nutrients) and focus
(seaweed) word were written in kanji and kana, respectively. In total, 70 focus
sentences and 70 non-focus sentences were selected.
Table 1. Sample Sentence of F-RST and NF-RST.
F-RST: A TV program reported many kinds of nutrients are

contained in seaweed.
TARGET WORD: SEAWEED Focus WORD: SEAWEED
NF-RST A TV program reported many kinds of nutrients are
contained in seaweed.
TARGET WORD: NUTRIENTS Focus WORD: SEAWEED
Two kinds of RST were constructed; one was a focus sentence RST (F-RST) and
the other was non-focus sentence RST (NF-RST). F-RST consisted of 70 focus
sentences. NF-RST consisted of 70 non- focus sentences. Half of the target word of
both F-RST and NF-RST were kanji nouns (35 kanji words) and the other half were
kana nouns (35 hiragana words). In the RST condition, when the last sentence of
each series was finished, the subject was requested to recall the target word of each
sentence. The number of sentences was increased from 2 sentences to 5 sentences.
Each sentence condition had 5 trials.
2.3 Subjects
Sixty undergraduate students of the same university performed the task; half of the
subjects performed F-RST, and the other half performed NF-RST. All of the
subjects performed all the trials in each RST. There were no significant differences
in the means of their digit span and word span between the two groups.
3. RESULTS
Table 2. Mean Span and Total Correct Recall for Each 30 Subjects
F-RST NF-RST
N 30 30
Mean Span 3.67 > 3.00
Mean Correct 82.43 > 71.95
Recall (%)
Table 2 shows the mean span and total correct recall for each of the 30 subjects. The
mean from the F-RST was 3.67 and that from NF-RST was 3.0. The mean span was
significantly higher for F-RST than NF-RST (t-test, p<O.OOI). The numbers of total
correct recall were also significantly higher for F-RST(57.70, SD=7.04) than NF-
RST (50.37, SD=9.99) (t-test, p<O.OOI).
THE EFFECT OF FOCUSING ON A SENTENCE IN JAPANESE READING SPAN TEST 159
The correct response percentages were measured separately for kanji and kana.
The figure shows the correct response percentages for kanji and kana (average of 30
subjects) in F-RST and NF-RST.
100 o kanji
.-..
~
~
80
II kana
a
"co. 60
:
..
~
u"
40
20
0
F·RST NF·RST
Figure 1. The correct response percentages for kanji and kana in F·RST and NF·RST.
The correct response percentages were higher for F-RST than NF-RST. Two-way
ANOVA (2 (focus; F and NF) x 2 (script; kanji x kana» revealed that there were
significant main effects of focus (F(1,116)=21.81, p<O.OOOl). There was no
significant difference in script difference. However, the interaction between the
focus and script was significant (F(1,116)=10.26,p<0.001). With F-RST, there were
no clear differences in correct responses between kanji and kana. However, with
NF-RST, the correct response was significantly lower for kanji than for kana
(p<O.Ol, t-test ).
F-RST NF-RST
100 100
-
tfl. .
~
1II:
il..
so I~ so
.. ...
u"
u"
0 0
".Span-Ss L-Span-Ss ".Span-Ss L-Span.Ss
Figure 2. The correct answer percentage in kanji and kana for the high span and low span
subject groups on F·RST and NF -RST.
Two groups were selected in both F-RST and NF-RST; those with high span
subjects (HSS, with J-RST score of 4.0 - 4.5) and low span subjects (LSS, with J-
RST score of 2.0 - 2.5). Figure 2 shows the correct answer percentage in kanji and
kana for each subject group. In the F-RST, the correct answer percentage was higher
for HSS (n=7) than for LSS ( n=14). Two way ANOVA (2 (subject groups, high and
low) x 2(script difference, kanji and kana» showed only a significant main effect of
the subject group (F(1,36)=44.67, p<O.OOOl). There were no significant differences
between kanji and kana in both the high and low span groups.
In the NF-RST, the correct answer percentage was higher for HSS than for LSS.
Two-way ANOVA (2(subject groups, high and low) x 2(script difference, kanji and
kana» showed a significant main effect of the group (F(1,38)=35.18, p<O.OOOl). The
correct answer percentage was higher for HSS than for LSS. Moreover, a significant
main effect of script difference (F(1,38)=13.06, p<O.OOl); the correct answer
percentage was higher for kana than for kanji. The script difference was also
measured separately in HSS and LSS, and a significant script difference was found
only in the LSS group (t-test, p<O.OOl), and not in the HSS group.
4. DISCUSSION
In this experiment, the span of F-RST was larger than NF-RST, and the focus effect
was observed. In the sentence, if the target word was the focus of the sentence, the
comprehension processes facilitated maintaining the target word. However, if the
target word was not the focus word of the sentence, readers must process the words
for comprehension as well as maintain the target word and the two processes
compete for attention resources.
In NF-RST, comprehension processing did not accelerate maintaining the target
word, the readers consumed a large capacity and did not have sufficient capacity to
hold the word. To avoid decay, they have to rehearse the word frequently, because it
has been argued that the most stable and retrievable short-term memory code is a
sound-based code (Baddeley, 1986). Therefore, they depended on the easy method
of holding the word, that is phonological loop.
Moreover, both kanji and kana interact with focus effect. A kanji and kana
difference was found only in NF-RST.
Table 3 shows the mora length of the target word of kanji and kana in F-RST and
NF-RST. Because Kanji mora length are generally longer than the character length;
most kanji scripts have more than one mora length. The word length in mora was
longer in kanji than in kana for both F-RST and NF-RST. Therefore, it is supposed
that the percent recall of kanji was decreased by the effect or word length effect
(Conrad & Hull, 1964). It is also suggested that the reason for the difference
between kanji and kana in NF-RST was that subjects depended on phonological
rehearsal in the NF-RST.
The word length effect was mostly dominant in the LSS, because they severely
exhausted their working memory capacity in comprehending the sentence and they
had less capacity to maintain the target word. Low span readers are forced to use
most of the working memory capacity just to reactivate products using the
phonological loop.
In F-RST, on the contrary, the script difference disappeared. In addition, the
script difference did not interact with the focus effect.
THE EFFECT OF FOCUSING ON A SENTENCE IN JAPANESE READING SPAN TEST 161
Table 3. Word Length (in Mora ) of the Target Word of Kanji and Kana in F-RSTandNF-
RST.
Kana Kanji
Min 2 3
Max 4 7
Mean 3.03 4.38 F-RST
SD 1.26 1.28
Min 2 2
Max 4 7
Mean 2.91 4.03 NF-RST
SD 0.77 1.03
The word length effect disappeared in the F-RST. The word length was reported
even in RST (La Pointe & Engle, 1990), and it was stressed that the maintaining
words is most important in RST (Engle, Kane, & Tuholski, 1999). However, in this
experiment, the results suggest that sentence comprehension is more important than
maintaining the target word; focus effect overcame the word length effect. The
result also showed that focusing had an advantage for integrating the sentence and
also accelerated the process of maintaining the target words even though the target
word was phonological or logo graphical one.
As for the inhibitory mechanism hypotheses, low span readers had deficit only in
the NF-RST condition, and they showed good performance in the F-RST condition.
Therefore, the important thing influencing their performance on RST was not a
inability to inhibit irrelevant information, as Engle et al (1995) suggested, but how
they focus their attention and integrate the sentence.
The differences between the high span reader and low span reader were how
quickly they find the focus word, and how effectively they use the focus word for
temporal integration of incoming textual information of the sentence. In this
integration processing, the activity of the central executive is more involved than is
the activity of the phonological loop or VSSP (Baddeley, 1986). The absence of a
word length effect in the focus condition suggests that there is less rehearsal,
indicating that comprehension processing facilitated the storage system and also
suggesting that focusing effects play an important role in efficient processing of
working memory.
5. ACKNOWLEDGMENTS
The work was supported in part by grants (#09044007 and #12301005) from the
Ministry of Education of Japan to the second author.
6. AFFILIATIONS
Dr. Mariko Osaka
Department of Psychology, Osaka University of Foreign Studies,
Address correspondence to:

Dr. Naoyuki Osaka
Department of Psychology, Graduate School of Letters, Kyoto University,
Kyoto 606-8501, Japan
E-mail: osaka@psy.bun.kyoto-u.ac.jp
7. REFERENCES
Baddeley, AD. (1986). Working memory. New York: Oxford University Press.
Carpenter, P. A, & Just, M. A (1989). The role of working memory in language comprehension. In D.
Klar, & K. Kotovsky (Eds.), Complex information processing: The impact of Herbert A. Simon (pp.
31-68). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Conrad, R, & Hull, A J. (1964). Information, accoustic confusion and memory span. British Journal of
Conway, A R. A, & Engle, R W. (1994). Working memory and retrieval: A resource-dependent
inhibition model. Journal of Experimental Psychology:General, 123, 354-373.
of Verbal Learning & Verbal Behavior, 19,450-466.
Daneman, M., & Carpenter, P. A(1983). Individual differences in integrating information between and
within sentences. Journal of Experimental Psychology: Learning, Memory and Cognition, 9, 561-
583.
De Beni, R, Palladino, P., Pazzaglia, F., & Cornoldi, C. (1998). Increases in intrusion errors and working
memory deficit of poor comprehenders. The Quarterly Journal of Experimental Psychology, 51A,
305-320.
Engle, R. W., Conway, A R A, Tuholski, S. W., & Shisler, R. J. (1995). A resource account of
inhibition. Psychological Science, 6, 122-125.
Engle, R W., Kane, M. J., & Tuholski, S. W. (1999). Individual differences in working memory capacity
and whit they tell us controlled attention, general fluid intelligence, and functions of the prefrontal
cortex. In A Miyaka, & P. Shah (Eds.), Models of working memory: Mechanisms of active
maintenance and executive control (pp. 102-134). New York: Cambridge University Press.
Just, M. A, & Carpenter, P. A (1992). A capacity theory of comprehension; Individual differences in
working memory. Psychological Review, 99, 122-149.
Kintsch, W., & VanDijk, T. A (1978). Toward a model of context comprehension and production.
Psychological Review, 85, 363-394.
La Pointe, L. B., & Engle, R. W. (1990). Simple and complex word spans as measures of working
memory capacity. Journal of Experimental Psychology: Learning, Memory and Cognition, 16, 1118-
1133
Osaka, M., & Osaka, N. (1992). Language-independent working memory as measured by Japanese and
English reading span tests. Bulletin of the Psychonomic Society, 30,287-289.
Osaka, M., & Osaka, N. (1994). Working memory capacity related to reading: Measurement with the
Japanese version of reading span test. Japanese Journal of Psychology, 65, 339-345 (In Japanese
with English summary).
SECTION IV
BASIC FUNCTIONS OF READING DISABILITY

T.LACHMANN
READING DISABILITY AS A DEFICIT IN

FUNCTIONAL COORDINATION
Abstract. Reading disability (dyslexia) is introduced as a failure in learning to optimize the coordination
of the subfunctions involved in reading with the consequence of errors or delays in integrating reading
related information represented in working memory (Functional Coordination Deficit model). Within this
multicausal model, so-called reversal errors, such as those typically found in beginning readers and in a
subgroup of reading disabled children, are explained as a result of a failure in optimally integrating visual
and auditory information during reading. "Symmetry generalization" is introduced to describe a
mechanism in visual object recognition. It is argued that symmetrically related objects (all axis) are
represented in the brain by similar patterns of neural activity (cell assembly). Symmetry generalization, as
an result of evolutionary and individual development, is assumed to be learned as an infant to warrant
behavioral advantages (object constancy). However, this mechanism may be a hindrance in reading,
because graphemes are visual symbols, and as such they have to have a non-ambiguous relation to the
respective verbal information Oabel) they represent. It is argued that learning to read is comprised of
learning to treat graphemes as symbols instead of objects, which is assumed to be learned very early
during the first stage of reading acquisition. A failure in complete suppression of visually symmetrical
information in the representation of visual symbols during reading produces ambiguous relations between
visual and phonological information and disturbs the functional coordination, and thus may cause
problems in learning to read. It is emphasized that reversal errors do not reflect a visual deficit, an
incomplete hemispheric dominance nor the one and only cardinal symptom of dyslexia.
1. THE BEGINNING OF A SCIENTIFIC INTEREST IN READING

DISABILITY
1.1 The Educational Concept of Reading Disability
The beginning of the scientific interest in reading disability started about 100 years
ago and can be seen as a consequence of the advances in two originally distinctive
fields, educational and neurological science.
At the beginning of the 20 th century, psychoeducational testing became more
important, not only for measuring general intelligence (Binet, 1905; Stern, 1912) but
also for testing separate techniques like reading and writing (Gates, 1921, 1927;
Thorndike, 1919). The main interest in doing so was to predict a students learning
performance and to identify students whose performance in school differed greatly
to what the test results suggested. The main intention of the reading tests was to
optimize educational guidance for acquiring these techniques. Gates, Bond and
Russell (1939) later pointed out that reading ability is something that children
acquire in varying degrees; "it must be taught and is not a series of attributes for the
development of which a teacher can do nothing but wait". The reading test battery
compiled by Gates (1927) includes most of the aspects that are considered today as
important for reading: visual as well as auditory perception, word-recognition,

166 T.LACHMANN
ocular motor function, vocabulary and comprehension, general intelligence as well

as educational background, motivation, and emotional stability. The purpose of such
reading tests was not only to identify students with reading problems, but also to
provide an opportunity to teach them in a special way and to give clues for
educational guidance. Therefore, it was an aim to differentiate between "types of
difficulty" (Gates, 1927) reflecting causes of reading disability (also including low
intelligence, constitutional and educational immaturity, special bodily or mental
defects, defects of sensory apparatus, or personality for "extreme cases" of reading
disability; Gates, 1937).
1.2 From Aphasia to Congenital Word-Blindness

Another source of growing interest in reading disability can be seen in the
fundamental improvements in a sub-field of neurology now known as
neuropsychology. The finding by Broca (1865, see also the work of the precursors
Gall, Bouillaud, Dax, Auburtin, all cited in Kolb & Whishaw, 1996) that damage to
a specific frontal part of the left hemisphere of the cortex (still called Broca's area
today) induces a partial loss of spoken language (aphemia), as well as the work of
Trousseau (1865), Wernicke (1874), Broadbent (1872, all cited in McCready, 1926),
and others, put forward - albeit from a different perspective - Gall's (1758 - 1828)
idea of a localization of functions in the brain (especially in the cortex) and
awakened the hope to find particular areas responsible for reading. Hughlings-
Jackson (1868; cited in Taylor, 1931 or Zangwill & Wyke, 1990), one of the
pioneers of the idea of lateral dominance and known for his theory of the
hierarchical organization of the brain, argued that the areas "specially related" to
reading are not only those responsible for visual object recognition, but also those
responsible for spoken language. However, in contrast to most of his colleagues, he
assumed that basically all areas are more or less involved in language processing
and therefore in reading.
Kussmaul (1878) introduced the concept of "word-blindness" (and "word-
deafness"), which did not signify a sensory defect. The patients were not assumed to
be blind (or deaf) for the written (or spoken) words but rather for their meaning,
whereas other intellectual abilities, including speech, remained intact (see also
Wernicke, 1874; for critical remarks see Ebbinghaus, 1913, pp. 728-732). He
assumed that damage to certain parts of the left cortex, such as those investigated
previously by Broadbent (1872, cited in McCready, 1926), located in the left angular
and supramarginal gyri, was responsible for the symptoms he observed.
The case studies by Morgan (1896) and Kerr (1896, cited in Orton, 1928a) are
recognized as the first (see also Berlin, cited in Venezky 1993; Broadbent, cited in
McCready, 1926; Hinshelwood, 1895) which shift attention from patients with
word-blindness acquired by brain damage to children without any injury. These
children showed similar symptoms to patients with brain injury. However, these
symptoms result from an inability to learn reading, not from a loss of reading
capability. Morgan (1896) assumed that the student he described suffered from a
congenital caused defective development of the same areas of the left hemisphere
which are damaged in the patients described by Kussmaul (1878). During the
following decades a striking number of publications on congenital word-blindness,
FUNCfIONAL COORDINATION DEFICIT IN READING DISABILITY 167
as the counterpart to acquired word-blindness, were published (Hinshelwood, 1907;

Rutherford, 1909; Warburg, 1911). These studies mostly concentrated on the genetic
etiology (family and gender studies) and the verification of locations in the brain,
but on the other hand increasingly considered educational and social aspects (Ford,
1928; Hinshelwood, 1917; McCready, 1926) and suggested educational (or
treatment) methods (Hinshelwood, 1917). In this respect, the concept of congenital
word-blindness can be seen as a brick of the bridge between the neurological
concept of word-blindness and the educational approach towards learning disability.
The concept introduced a developmental aspect and yet still held strongly onto the
notion of the localization of dysfunctions in the cortex and their genetic
conditionality.
Nevertheless, neuropsychologists (in the broadest sense) and educational
scientists remained skeptical of each other. The latter especially doubted the
hypothesized analogy of acquired and congenital word-blindness (Gates, 1921,
1927; Marie, 1922). They argued that it is the wrong way and a vain endeavor to try
to define and locate a separate area in the cortex, such as the "visual word center"
(Hinshelwood, 1917), whose faulty development is supposed to be responsible for
the whole range of reading disabilities. Furthermore, they criticized the analogue
assumption (Morgan, 1896) which claims that this area is identical to the area
damaged in acquired word-blindness patients (Hinshelwood, 1917). This also
implicates a controversy about educational methods. As a consequence of the
analogue assumption, most of the supporters of the congenital word-blindness
concept suggested to choose the same methods of treatment as used in acquired
cases. In contrast, the educators were interested in diagnosing different types of
reading disability (Gates, 1927) to optimize the special educational guidance. The
educators critique on the analogue assumption was also based on the fact that
patients suffering from word-blindness often cannot read at all while reading
disabled students usually can, although not as well as they should and with a great
variety of restraints. This critique was averted by the argument that a stroke, a
tumor, physical injuries, or other damages can seriously affect or even totally
destroy the reading area, while in the congenital case the adequate development is
more or less impaired. In this respect, Hinshelwood (1917) for instance, argued by
introducing different cases of word-blindness. Those which he categorized as "mild"
and called "congenital dyslexia," are characteristic for what occurs often in everyday
school life and might even be overlooked.
1.3 Neurological Basis of Education

The term congenital word-blindness never described a homogeneous concept. There
were always controversies concerning the number and locations of the brain areas
involved, as well as the diagnostic criterions (Bachmann, 1927; Fildes, 1921;
Gordon, 1923; Hinshelwood, 1907; Wallin, 1920; see also the critical review in
Ebbinghaus, 1913). Most influentially, Hinshelwood (1907, 1917) proposed that
word-blind children have an impaired visual memory for words, while their auditory
memory is unimpaired. However, neither he nor Kussmaul (1896) assumed that the
visual perception for words is in any way affected, but rather the association
between words as visual configurations and their appropriate names and meanings.
168 T.LACHMANN
This association is assumed to be realized by the "visual word center" in the cortex.
A faulty development of this particular area in the early stages of embryonic growth
- while other areas remain intact - induces (more or less) a disability to understand
the meaning of written words.
From the neurologists field, Orton (1925, 1927, 1928a, 1928b, 1929a, 1929b,
1937) most influentially criticized the concept of congenital word-blindness by
favoring a different understanding of the developmental aspect. Orton's (1925,
1928a) critique on a simple analogy hypothesis (see above), as postulated by the
supporters of the concept of congenital word-blindness (Hinshelwood, 1917; Kerr,
cited in Orton, 1928a; Morgan, 1896), was based on findings such as that during the
plastic period of development, a lesion does not necessarily prevent the acquisition
of the respective function (Apert; Pick, both cited in Orton, 1925; Marie, 1922). He
argued that reading disability is not a pathologic irremediable condition where a
center will never be developed adequately. Instead, reading disability, appearing in a
graded series of severity, is "explainable as a variant in the development of the
physiologic lead in the hemispheres" (1928a). Using the right methods, the difficulty
might be overcome. In other words, he still believed in a physiological origin - albeit
in a different way - but also emphasized educational and environmental "forces",
such as social aspects, as important causal factors, while earlier studies, if any,
considered these aspects only in terms of consequences of the congenital word-
blindness (Ford, 1928), or as factors influencing the rehabilitation success. For a
developmental disorder (or "delay") in this sense, the symptoms are assumed to be
more or less normal in a certain stage of physiological maturation of brain functions
important for learning to read. However, reading disabled children need special
training for overcoming these symptoms (Orton, 1925 based on findings by Javal,
Sereni, Pick and others cited there; see also Fildes, 1923). As a consequence, even
though he did not doubt that there is a congenital proportion, the term congenital
word-blindness was replaced by the term developmental word-blindness. Thus,
Orton's theory can be counted as more than just a brick of the bridge between
neurologists and educators. Rather, it can be characterized, as he himself did, as a
theory of the "neurological basis of education" of reading disabled children (Orton,
1925, 1929a). For this reason, his theory will be introduced in more detail in the
following section.
1.4 Strephosymbolia
In 1925, Orton was the director of a mental hygiene clinic, an institution to support
practical and longitudinal orientated multidisciplinary research in natural settings in
cooperation with the social environment, such as parents, teachers, etc. (Richardson,
1989). As a researcher with neurological background, he then was confronted with
the entire range of learning disabilities arising in schools. Among the pupils who
were sent by their teachers as "failing in school work", Orton and his collaborators
identified a high proportion of students with a special difficulty in learning to read.
Using a psychometric test (Binet, 1905), they realized that these students were
unfairly underestimated in their intellectual capacity (Orton, 1925, 1928a). These
students reached IQ measures within a range of "normal intelligence" and could by
no means be characterized as defective. At this point, parallels to educational
research can clearly be recognized (Gates 1921, 1927). Orton's theory related to
these findings, however, is based on his neurological background and knowledge,
such as the specificity of the hemispheres (Fildes, 1921, 1923; Gorden, 1920;
Hughlings-Jackson in Taylor, 1931; Javal, Kapper, 1906, both cited in Orton, 1925;
Parson, 1924; Sherington, 1906) or separate levels (or types) of visual cortical brain
functions (Brodmann, 1909; Campbell, 1905). In this respect, he interpreted the high
frequency of left-handness (-eyeness) within the group of reading disabled children
(already mentioned by others, Fildes, 1923; Gorden, 1923; Parson, 1924) and the
typical reading and writing mistakes made by many of that group as result of a
developmental disorder of the interhemispheric relation. Without any "visual deficit
in the ordinary sense", as tested by recognition and naming of visual objects (Healy
Pictorial Test, see Orton, 1925, 1928a), the characteristic mistakes and phenomena
he found (1925) were:
(1) On the letter level (he later called them static reversals):
difficulty in differentiating letters which are horizontally or vertically
symmetrical to each other or rotated (p and q; band d; p and d);
(2) On the word level (he later called them kinetic reversals):
tendency to confuse palindromic words (was and saw; not and ton) and to
read partially from right to left resulting in a reverse of paired letters or
even syllables within a word;
(3) a remarkable capability for mirror reading and writing, sometimes better
than in normal orientation.
These orientation mistakes (reversals) are assumed to be the result of a hemispheric

imbalance and understood as a cardinal symptom reflecting the whole graded series
of reading disability (contrary to Hinshelwood's, 1917, distinction between mild and
severe forms). The great variety of all other mistakes, which he also observed (see
Orton, 1925), were seen as a "secondary defect of the learning process resulting
from a lack of practice in reading because of the obstacle in recognition interposed
by the reversals" (1928a). Therefore, he introduced the term strephosymbolia as a
"descriptive name for the whole group of children who show unusual difficulty in
learning to read" (1925). It is important to note at this point that Orton believed to
have found an objective measure for reading disability (or strephosymbolia) in the
number of reversals and in the ratio of the time from mirror reading to normal
reading (1928a). Using these objective measures, he found a striking difference in
the percentage (two to four percent, cf. Orton, 1929c) of strephosymbolic children in
different school populations, which he interpreted as evidence for his hypothesis of
the causal role of teaching methods (sight reading plan, multi-sensory training, etc.)
as an environmental factor.
Orton is often incorrectly cited in literature as the first to realize that reading
disabled children make typical orientation mistakes (reversals) and have a special
capability in mirror writing. These symptoms and their reference to interhemispheric
communication were previously hypothesized (Fildes, 1923; Gorden, 1923; Javel;
Sereni, both cited in Orton, 1925) and extensively reviewed by Orton (1925). This
also holds true for the idea that mirror writing and hemispheric confusion is normal
in a certain stage of evolutional (phylogenesis) and individual (ontogenesis)
development (Fildes, 1923; Fildes & Myers; Javel; Sereni; Ullmann, all cited in
170 T.LACHMANN
Orton, 1925; Parson, 1924). His accomplishments rather consist of a

supplementation and systematic synopsis of all findings and their reinterpretation
within a neurological framework (for reading disability, with educational relevance),
using his enormous experience and considering the latest neurological theories of
that time. This neurological framework will be explained in the following paragraph.
In the first quarter of the 20 th century, experimental psychologists (Wundt
founded the first psychological institute in Leipzig in 1879) began to study cognitive
levels (stages) using introspection method (Wundt, 1912) and methods of measuring
behavioral data (reaction time paradigm based on Donders, cited in Lachman,
Lachman, & Butterfield, 1979; Wundt, 1912). Neurologists, such as Orton, however,
were motivated to define localizations in the brain as a neurological basis for the
functional levels. Orton (1925, 1929b) postulated three such levels within the visual
cortex, based on both histological findings about distinct types of neurons in the
visual field (Brodmann, 1909; Campbell, 1905) and neuropsychological evidence
from focal brain damages (Brodmann, 1909; Hinshelwood, 1917; Munk, cited in
Orton, 1925, 1929b; Sharington, 1906).
' - - - - - VtllU . . . CI.IoTlVI _ _ _ _J

' - - - - - - V"IU llU..-rIlI-_ _ _ _J
' - - - - - - - .... u PlHI"'" - _ _ _ _ _.J
Figure 1. Distribution of three types of cortex in the hemispheres of the human brain. Taken
from Orton (1925) with permission of the American Neurological Association.
As an early arrival platform (after reflex centers), he defined the area striata (see
Campbell, 1905) as the neurological basis for the first level, the visual perceptive
level. From a bilateral destruction, but not from unilateral destruction of this area(s)
a cortical blindness (Orton, 1929b) results in which there are no conscious visual
processes, whereas the lower reflex phenomena remain unaffected. The occipital
cortex which surrounds the arrival platform in both hemispheres (area occipitalis,
Brodmann, 1909; Campbell, 1905), is assumed to be related to the second level, the
visual recognitive level. An extensive bilateral destruction results in a mind-
blindness (Orton, 1929b). There is retention of mere awareness of visual stimuli
(first level). The patients can move about without collisions, but there is a loss of
their recognition (objective memory). For the reading process the third level, the
visual associative level or symbolic level (Orton, 1928b), is most interesting.
According to Orton, this functional level is not exactly locatable, but includes
temporal and parietal areas laying nearest to the visual recognitive field (prae-
occipital area, occipito-temporal area, area angularis and parts of area parietalis
superior, see Brodmann, 1909). With lesions in these areas, the awareness and the
recognition of the meaning of visual objects or symbols remain intact, but there is a
loss of the abstract or associative meaning of printed words (associative memory)
and thus a difficulty in reading (acquired word-blindness). However, in this case,

contrary to the first two levels, an unilateral damage in the dominant hemisphere
(either left or right) is sufficient to produce the symptoms, whereas damage to these
areas in the non-dominant hemisphere does not result in comparable clinical
symptoms (Orton, 1928b). This is similar to other disabilities resulting from a loss
of a higher function after unilateral damages and points out the role of cerebral
dominance in phylogenetic and ontogenetic development of higher processes (such
as language).
The lesion studies suggest that within the first two levels, the hemispheres "work
in union to produce a single conscious impression" (Orton, 1929a), whereas the
association with sounds and abstract meanings (associative level) required for
reading is realized by the dominant hemisphere only. Orton's interest, however, not
only refers to what happens in the associative fields of the dominant hemisphere, but
also to what happens in the cells of the symmetrical counterpart (often called "silent
areas," 1928b). These cells are assumed to be stimulated similarly by the incoming
stimuli, and are also supposed to form a mnemonic record or engram (Kapper, cited
in Orton, 1925; see also H.-Jackson, 1864 in Taylor, 1931). Orton assumed this
engram to be a mirrored (antitropic) version of the one in the dominant hemisphere,
because of the oppositely orientated nerve cells (see Sharington, 1906). Thus, for
reading at the associative level, the orientation of letters and words becomes
relevant. A mirrored word or letter might be related to a different concept or sound
association (as for p/q). Since the linkage between the levels must be simultaneous
or quickly successive and strongly concordant, it is crucial to completely suppress
the mirrored record to make a fast and correct association with sounds or abstract
meanings. In this respect, learning to read means learning to suppress memory
patterns in the associative fields of the non-dominant hemisphere. Otherwise
confusion leads to mistakes and/or a delay in reading, as well as in writing because
of failing recognition of correct associations (see above). According to this point of
view, teaching to read must use a multi-sensory, more phonetic approach instead of
a "look and say" method (see Gillingham & Stillman, 1969).
Figure 2. Selection of correctly oriented memory images in focus of attention. Taken from
Orton (1925) with permission of the American Neurological Association.
172 T.LACHMANN
1.5 After Orton

Even though we highlighted Orton's theory here as a bridge between the educational
concept of reading disability and the neurological concept of congenital word-
blindness, it must be mentioned that this bridge was not really used enthusiastically
by many representatives of both fields (e.g. Mills, 1929; see also Critchley, 1966).
Instead, they still ignored or kept their critical distance to each other as well as to
Orton (Gates & Bernett, 1933; Gates, 1936; Hall, 1945; Schonell, 1940, 1941; see
also Venezky, 1993 for an overview). Many authors doubted the assumption of
reversal errors as a cardinal-symptom of reading disability, which is still based on a
single cause hypothesis (e.g. Gates & Bernett, 1933; Monroe, 1932; Witty & Kopel,
1936; Wolf, 1935; see also Dearborn, 1933; Robinson, 1946 for review). They
argued that not all reading disabled children show differences in mirror writing and
make more reversal errors, or show other symptoms in reading as well as in other
performances that could be explained by a deficient hemispheric lead (Schon ell,
1940). Some authors even failed to find any significant difference between groups of
normal reading and reading disabled children for the rate of reversal errors (Gates &
Bernett, 1933, see Miles, 1993 for review and statistics). Conversely, some authors
failed to find more reading problems or reversal errors within the group of left-
handed children (e.g. Gates & Bernett, 1933) or more reversals in left-eye dominant
reading disabled children (e.g. Monroe, 1932; but see also Monroe, 1928).
Furthermore, the unclear relationship between static (letters) and kinetic (words)
reversals, the role of non-vertical symmetry between letters (e.g. p/d), as well as the
findings that not all reversible letters are going to be reversed and that static
reversals only occur in the context of reading and writing, but not when letters were
presented separately, were topics of critical papers (see results and review in
Liberman, Shankweiler, Orlando, Harris, & Bell Berti, 1971; but see also Fisher,
Liberman, & Shankweiler, 1978). In addition or in competition to reversal errors and
mirror-writing, the main interest in the following decades shifted to what Orton
(1925, 1928a, 1928b) called "secondary" symptoms, such as repetitions in reading
and writing, and errors over vowels or consonants involving the omission of sounds
(Gates & Bernett, 1933; Monroe, 1932; Schonell, 1940, 1941, 1942; see also
Hermann, 1959; see Miles & Miles, 1999, for an overview). Later on, it was found
that these secondary errors are even more common than reversal errors (e.g.
Liberman et al. 1971; see KibeI & Miles, 1994 for more current results) and that the
reversals occur with large intersubject and innersubject variability (Liberman et aI.,
1971). This led to the implementation of multiple cause models for reading
disability (Schon ell, 1941, 1942), which either includes possible hemispheric factors
(for instance for a subgroup: Liberman et aI., 1971) or even negates the relation
between hemispheric dominance and reading problems (Robinson, 1946; see also
Zangwill, 1990). At this point, however, we should refer to a critique regarding the
definition of reading disability which will be mentioned later in more detail. .
In general, reading disability increasingly became a topic of psychological and
psycho linguistic research. To investigate the phenomenon, paradigms and methods
of experimental psychology (e.g. reaction time paradigm) were used (see Miles &
Miles, 1999, ch. 2, or Springer & Deutsch, 1998 for an overview).
1.6 The Phonological Turnaround
The main reason for explaining word-blindness and Orton's strephosymbolia here in
more detail is not only to give an introduction to the history of reading disability
research, but also to emphasize that these approaches are not to be counted as
theories of "visual perceptual deficits" in reading disabled children. It is often
argued in the literature that Orton and others overestimated visual components of
reading and overlooked language-basedllinguistic aspects (Fisher et aI., 1978;
Vellutino, 1979, 1987; Vellutino & Scanlon, 1998; see also Willows & Terepocki,
1993; or Corballis & Beale, 1993; or Willows, 1998 for a critical review). However,
this is a simplification of an approach which emerged from theories of acquired
disabilities of spoken language anyway (see above). Nevertheless, the terms "visual
word center" (Hinshelwood, 1907, 1917) and "visual associative level" (Orton,
1925, 1928) not only include the word "visual", but also describe the association
between a written word and a sound or a meaning (language aspects) as a higher
(linguistic) level of visual processing, and thus directly dependent upon the visual
input. As a matter of fact, following Orison's single cause assumption, many
mistakes (e.g. consonant confusion by voicing: c/g; f/v; confusion to more "open"
vowels: i>e, see Kibei & Miles, 1994) cannot be explained by any relation to the
visual stimulus, but rather by phonological aspects fully independent from the
configuration of the letter(s). Such mistakes are indeed difficult to explain as
"secondary symptoms". Moreover, some typically static (bid) and kinetic reversals
(brain/brian) can also be explained by phonological aspects (e.g. the place of
articulation in consonants, Kibei & Miles, 1994; see also Frith, 1985). Such
findings, often substantiated by teachers, as well as new theoretical approaches
(conceptual learning, Chomsky, 1959; see Friederici & Menzel, 1999 for a short
review) and numerous results within the field of psycho linguistic research, led to a
change of interest in reading researchers from unicausal theories of incomplete
cerebral lateralization to more linguistic based theories regarding the role of
phonological coding/decoding and segmentation skills for reading (Bradley &
Bryant, 1978; Liberman, et aI., 1971; Shankweiler, Liberman, Mark, Fowler, &
Fisher, 1979; often cited in this respect is Vellutino, 1977, 1979). Theories
originating from this "phonological turnaround" can roughly be subdivided into
multi- and unicausal models. Multicausal models include phonological, visual, and
other aspects in different proportions (Doehring, Trites, Patel, & Fiedorowicz, 1981;
Feagans & Merriwether, 1990; Korhonen, 1991; Lovett, Steinbach, & Frijters, 2000;
Lyon, Stewart, & Freedman, 1982; Mattis, French, & Rapin, 1975; Mitterer, 1982;
Satz & Morris, 1981; Seymour & McGregor, 1984; Watson, Goldar, & Ryschon,
1983; see Watson & Willows, 1993 for a review). Unicausal models assume primary
phonological deficits as the cardinal factor in reading disability (Bradley & Bryant,
1978, 1981; Vellutino, 1979, 1987; see also Miles, 1991; Wimmer, Mayringer, &
Landerl, 1999; Snowling, 1998 and Siegel, 1998 for an overview). Of course, there
are also unicausal visual models or modified lateralization models, which, do not
necessarily support Orton (Annett & Kilshaw, 1984; Bishop, 1997; Corballis,
Macadie, & Beale, 1985; Corballis, Macadie, Crotty, & Beale, 1985; Geschwind &
Galaburda, 1987; Galaburda, 1995; Jenner, Rosen, & Galarburda, 1999; see the
chapters of Galaburda & Stein in this volume; see Corballis & Beale, 1993 for
review, with restriction also Hier, LeMay, Rosenberger, & Perlo, 1978; please note
174 T.LACHMANN
that for these models "unicausal" does not mean that varying lateralization must
always lead to dyslexia; please see also the discussion about the definition of "visual
deficits in dyslexics" later in this chapter). In fact, this differentiation is not really
tenable, the complexity of models which exist in contemporary reading research
literature is not describable in simple terms (see Miles, 1991, for discussion).
However, there is at least one question which seems to arise from this distinction,
namely: Are there visual deficits in reading disabled at all? This question is hard to
answer, not only because the results published in the last decades are so
contradictory (see Rayner & Pollatsek, 1989 for discussion), but also because of the
inconsistency in defining both what is a "visual deficit", and also what is a "reading
disability". For this reason, the researchers who are trying to answer this question
have quite different understandings of these concepts and are using very different
methods and samples for their experiments. The following section tries to roughly
put in order the various studies and models of reading and reading disability within
the context of an information integration deficit view.
2. INFORMATION INTEGRATION AND THE COORDINATION OF

FUNCTIONS
2.1 The Reading Process

Undoubtedly, reading is one of the most complex cognitive processes for humans
overall. Even the "old neurologists" (especially Hughlings-Jackson, 1864 in Taylor,
1931), including those who believed in a stricter localization of brain functions
(Broca, 1865; Wernicke, 1874) and the representatives of the concept of congenital
word-blindness (Hinshelwood, 1917), would not have objected that many brain
areas and more than one special cortical location are involved in the process of
reading (see also Ebbinghaus, 1913, pp. 729). In this respect, it appears to be
important not only to study cases of both acquired word-blindness and
developmental reading disability (Seymour & McGregor, 1984), but also to try to
understand the reading process by itself (see critique on reading disability research
by Rayner, 1993). For instance, Everatt et al. (1999) described the reading process
simply as a comprisal of two basic processes: visual decoding of a written form and
language comprehension, including multitudes of sub processes. It would be a big
misunderstanding to presume that this does not agree with the understanding of
Hinshelwood (1917) and Orton (1925). However, they assumed a serial order of
these processes (which followed the general belief at this time). As already
mentioned, within this point of view, phonological and semantic processes cannot be
explained independently of the visual input since they are triggered by a fully
created visual representation. Of course, with present knowledge from
psychophysiology (Friederici, Wang, Herrmann, Maess, & Oertel, 2000),
neuropsychology (see Kolb & Whishaw, 1996 for an overview), experimental
psychology and reading research (see Rayner & Pollatsek, 1989 or Everatt, 1999 for
an overview; see also the chapter by Starr, Kambe, Miller, & Rayner in this volume)
and their interdisciplinary overlap in terms of cognitive neuroscience (Springer &
Deutsch, 1998 for an overview), it appears unlikely that visual and "associative"
processes occur serially. For instance, we know from eye-movement research (Javal,
FUNCTIONAL COORDINATION DEFICIT IN READING DISABILITY 175
cited in Heller, 1982 as first study in eye-movement; Katz & Wicklund, 1971;
McConkie & Rayner, 1976; Just & Carpenter, 1980; Rayner & Pollatsek, 1989; see
also Heller, 1982; Rayner, 1998; Radach, Heller, & Inhoff, 1999) that it may take
only a couple of hundred ms (e.g. Starr et a1. in this volume) to read a whole word
(within one fixation). On the other hand, participants in experiments need about half
a second or longer to recognize or to recall a single letter (Brendler & Lachmann,
2001). Nevertheless, they perform such experimental tasks with letters much faster
than they do with non-linguistic pattern material of comparable structure (Brendler
& Lachmann, 2001; Lachmann, in preparation; see also Koiers, 1970). We also
know that it takes much less time to recognize a word than a nonword (word
superiority effect; Reicher, 1969; Wheeler, 1970; see also Johnston & McClelland,
1973; Krueger, 1975, Greenberg & Krueger, 1980) and that context has a significant
influence on word recognition (Everatt et aI., 1999; Massaro, 1998; Massaro &
Cohen, 1991; Rayner & Pollatsek, 1989). Thus, reading seems to be more than just a
simple addition of visual recognition of letter forms or letter combination shapes and
language comprehension. This is at least true for skilled readers in most of the
reading situations. In fact, beginning readers use a more serial strategy on the very
beginning of the pre-alphabetic (Ehri & McCormick, 1998; Ehri, 1995, 1999) or
logographic phase (Frith, 1986), where they perceive each letter separately, linking
them with a sound and merging the sounds to a word which generates a meaning. In
contrast, skilled readers fixate mostly on whole words in a text for about 200 ms
(fixation time; Rayner & Pollatsek, 1989; see also Just & Carpenter, 1980;
McConkie & Zola, 1987; Starr et a1. and Radach, Inhoff, & Heller in this volume,
see also Radach et aI., 1999). After a fixation they move their eyes in terms of a
saccade to the next fixation point. The perceptual span, i.e. the area fixated upon
(over which visual information can be retrieved), can be described as a function of
text difficulty (not of the complexity of the configuration) and reading experience
(and thus the ability of using context information and recognizing words by sight,
see e.g. Ehri, 1999). Thus, the time of a fixation cannot be seen as a first serial stage
that represents the time taken for a visual analysis of the letters focused upon. Even
without recognizing all letters of (for instance) a word (Rayner, Well, Pollatsek,
Bertera, 1982; Rayner, Slowiaczek, Clifton, & Bertera, 1983), visual information
can be analyzed within about 50 ms during a fixation (Rayner, Inhoff, Morrison,
Slowiaczek, & Bertera, 1981). Furthermore, information from neighboring words is
often acquired parafoveally (Rayner, 1998). Short words ("and"; "or", "in", "of')
may even be skipped from fixation analysis (see Radach et a1. in this volume). A
famous example for the strong semantic and context influence on skilled reading is:
Paris in the the Spring.
Most readers do not realize the double "the" (e.g. Miles, 1991). For more details
about the reading process, see the chapters of Starr et. aI., Radach et aI., and Osaka
& Osaka this volume.
176 T.LACHMANN
2.2 Functional Fragmentation: The Dilemma of Experimental Reading Disability

Research
Generally, it can be summarized that reading is not a single capacity, but rather a
coordination of many functions and processes (which was already postulated by
Gates in 1921; see also modular organization, Marr, 1976). Such functions are not
only visual, phonological, and seman tical decoding, but also orthographic, syntactic,
and contextual analysis, emotional evaluation, motor and attentional control, long-
and short-term memory, etc. (see also the chapter Friederici & Lachmann this
volume; Friederici, 2001). Reading not only implies that all of these functions work
with a high speed and accuracy, but also that they are perfectly coordinated and
influence each other in an optimal way. This view excludes a single cause
assumption and explains why reading disabled mostly do not show comparable
failures in other performances and why experimenters often fail to find visual (or
auditive) deficits in all reading disabled children or adults. In fact, experimental
reading disability research is in somewhat of a dilemma. By definition, reading
disabled differ from skilled readers in their reading performance. Consequently, this
reading performance is already a result sufficient to differentiate these groups (apart
from the fact that their must be a chosen criterion for their differentiation, see
below). Thus, the patterns of errors and the latency of reading should be analyzed
and interpreted, as it was done e.g. by Orton (1925). However, to investigate the
origin of the lack of reading performance experimentally, the reading process has to
be fractionalized to test the ability in single or several functions which are assumed
to play a major role for reading (e.g. phonological decoding); or to test another
performance which is assumed to be based on one or several of the functions
involved in reading (e.g. Ho, Law, & Ng, 2000; Ho & Lai, 2000: onset detection or
rhyming for phonological awareness, word repetition vs. non-word repetition for
phonological memory). If reading disabled in comparison to skilled readers show a
failure in this tested function(s), then this function(s) is (are) assumed to playa
causal role for the failure in the reading performance. However, with the point of
view that reading depicts a coordination process, a reading disabled person can do
very well within all functions that are hypothesized to be important for reading when
they are tested separately. His/her problem may rather be the coordination of the
processes and the integration of the information available from different
representational formats in the speed and accuracy required for reading. That is why
the question about visual deficits generally cannot be answered more satisfactory
than the question about phonological or other deficits in reading disabled.
Nevertheless, many experiments suggest in fact a deficit in a single function. There
is no doubt that such a failure might lead to problems in reading. However, it cannot
be concluded from these results that reading disability is caused solely by a lack of
the function which has been tested. Moreover, the null finding of an experiment
aimed at testing e.g. visual or phonological processing, does not necessarily imply
the exclusion of visual or phonological processing deficits as one cause for (at least)
the failure in coordination of the functions involved in reading. An experimenter
cannot be sure to test the function in the same way it is claimed in reading (the
problem of dissociation/double dissociation and association in neuroscience), nor
can she/he comprehend the interaction with other functions in the way which is
necessary for reading.
2.3 The Specificity of the Coordination Process in Reading
Due to other factors besides complexity, the process of coordination of cognitive

functions and the binding of information in reading is different than in most other
human cognitions. To explain this, let us simply consider the visual, phonological,
and semantic processing in terms of a vertical order of the main functions involved
in reading (amongst others, like the analyses of orthographic and syntactic
information, see above), and attention/coordination processes as horizontal aspects.
No doubt, at first the visual input serves as a key to trigger a bottom-up process.
However, this does not only run the formation of a visual representation, but rather
generates hypotheses in terms of interacting bottom-up - top-down information
integration processes within and between the representational dimensions (see e.g.
Massaro & Cohen, 1991; Massaro, 1998; see also the contributions in McClelland &
Rumelhart, 1986). Visual information can trigger hypotheses on the semantic level
(bottom-up) and activated semantic information can strongly influence the speed and
quality of the creation of a visual representation (top-down). A semantic decoding
can even be finished before a visual representation is fully completed (e.g. Stroop,
1935). Thus, it only makes sense to speak about a vertical order of the involved
functions when considering the fact that visual information starts the whole process
and that semantic decoding (meaning) is its aim.
Of course, the well known fact of interacting bottom-up - top-down mechanisms
is not reading specific (e.g. Vecera & Farah, 1997; see also Stroop, 1935). For
instance, from visual recognition experiments we know that the degree of which dot
patterns can be associated with meaningful objects influences the recognition time
significantly (e.g. Lachmann, 1996, Exp. 3). However, the fact that the meaning is
coded abstractly is reading specific. Seeing a drawn table is quite different in
comparison to decoding the meaning of the written word "table" (Zaidel, 1998). The
visual configuration of letters (graphemes) have no meaning. They only represent
phonemes which alone have no meaning either (except one letter words). In order to
comprehend the meaning of e.g. a word, depending on reading experience and the
familiarity of the word (see above), we have to fixate and analyze a certain number
of visual configurations and merge the phonemes they represent to a word, which
represents a meaning in long-term memory. This meaning can even depend on the
semantic context. This also requires short-term memory processes on all
representational dimensions (on letter as well as on word level) and their quick and
accurate coordination. Thus, reading not only requires the recognition and decoding
of visual shapes, but also the storage and retrieval of visual, phonological, and
semantic information from long- as well as from short-term memory (see Cowan,
2000 for models of memory organization), and their interaction. In a way, reading
reflects a continuing working memory (Baddeley, 1986) process as defined by
Cowan (1997, 1998, 2000) in the sense that it reflects a collection of mental
processes, which permit information to be held temporarily in an accessible state, in
the service to perform some mental task.
In this respect, it is also specific that in reading the bottom-up information from
visual analyses plays a deciding role during the whole decoding process. Even the
orientation of the grapheme may imply a different phonological (phoneme) code (d :
p : b) and, within a word context, a different meaning (e.g. dip : did; doll : poll :
boll). This is not the case in object recognition, which will be discussed in more
178 T.LACHMANN
detail in the following paragraph.

Lachmann and colleagues (Lachmann & Geissler 1999; Lachmann, 2000;
Lachmann & van Leeuwen, in press; see also Berti, Geissler, Lachmann &
Mecklinger, 2000) found evidence that shapes of different orientation
(reflection/rotation in 90° increments), which were shown in a visual recognition
experiment, were (at least within the used task demand) represented by the same
visual code. Furthermore, the process of identification of such a shape was modeled
as a serial search within that group code representation; that is, any shape was coded
and processed in relation to all transformational alternatives (Lachmann & Geissler,
in press; Lachmann & van Leeuwen, in press; Lachmann, 2001). Moreover, it is
known that animals as well as humans can hardly discriminate between objects and
their mirrored replicas (Sutherland, 1957; see also Hubel & Wiesel, 1967; Corballis
& Beale, 1970, 1993; Geissler & Lachmann, 1996). It was shown experimentally
that the time to recognize or name a familiar shape or to identify a face in a visual
search task does not depend on its orientation (Corballis & Beale, 1970, 1993;
Corballis 1986, 1988; Jolicoeur, Corballis & Lawson, 1998; Klatzky & Stoy, 1974;
Kuehn & Jolicoeur, 1994; see also Rock & Leaman, 1997). Furthermore, it was
shown that a mental rotation or reflection of visual representations (Shepard &
Metzler, 1971; see also Ruthruff, Miller & Lachmann, 1995) or a search within
orientational alternatives (Lachmann & Geissler, in press) only occurs under certain
task demands (Jolicoeur, Corballis & Lawson, 1998; Kuehn & Jolicoeur, 1994;
Lachmann, 2000). This confirms everyday life experience about the constancy of
objects (Gestalt) in different orientations. In order to detect eyeglasses on a table,
their orientation is not a factor as long as the same amount of information is
available to the visual system.
Although this sounds trivial, it is significant for reading. Consider a task that
requires the fast naming of objects and letters (abstract code) which are shown in
randomly varying orientation (reflection and rotation). It is easy to recognize a table
or a house as well as a letter "T" or "x" independent of their orientation. When the
letter "b" is shown, however, we cannot solve the task at all, because we do not
know the orientation of the visual configuration. In other words, while visual
representations seem, to a certain extent, to be invariant against orientations and thus
ambiguous, the phonological code is non-ambiguous. Detailed visual information,
induding the orientation of letters, influences the whole process of phonological and
semantic decoding. This depicts a special exertion for the coordination process by
the need to take detailed configurationally bottom-up information into consideration
during the whole decoding process and - most importantly - to keep this information
in (working) memory.
When we are looking in a mirror we might feel we get a "normal" impression of
the environment; we even forget that it is simply a mirror-image that we perceive. It
is no problem to put makeup on the right eye and there is no risk of combing the hair
from left to right instead of the other way around. This effect immediately breaks
down when we look at a newspaper in the mirror. Everything looks as strange as a
script of foreign language. It may even be difficult to read the headlines. This
underlines the specificity of reading in terms of abstract signs (symbols, see below).
Furthermore, suppose you get a horizontally or diagonally mirrored image of the
environment. In contrast to cases of vertical symmetry, this may initially produce
estrangement and strong problems of adequate control of behavior. There is a lot of
evidence that vertical symmetry has a special importance in human perception

(Barlow & Reeves, 1979; Lachmann, 2000, pp. 189-191; Mach, cited in Herbert &
Humphrey, 1996; Masame, 1984; Palmer & Hemenway, 1978; Pashler, 1990; but
see also Sekuler & Swimmer, 2000). This can be seen as a consequence of
behavioral relevance. The theoretical importance of these samples will be explained
in the following section.
2.4 Welcome Back to Orton: Revival of an Unpopular Dinosaur.
Corballis & Beale (1970, 1993) and others (see Herbert & Humphrey, 1996) argued
that the effects of invariance of object orientation in vision, especially for vertical
symmetry, are caused by a mirror-image generalization (mirror-image duplication,
Beale, Williams, Webster, & Corballis, 1972) which in itself is the result of an
evolutionarily indicated symmetrical organization of the nervous system in higher
animals and humans (Tschirgi or Gardner, both cited in Corballis & Beale, 1970).
Even though there is an asymmetrical tendency in terms of a hemispheric dominance
in humans, brain functions are to a certain extent still organized in both hemispheres
(Zaidel, 1998). There is also evidence that many (not all) callosal and commissural
fibers connect homotopic mirror-image points of both hemispheres (Mach's
"callosal hypothesis", see Achim & Corballis, 1977; Beale et aI., 1972; Cummings,
1970; Herbert & Humphrey, 1996; Noonan & Axelrod, 1992; Saarinen & Levi,
2000; Sperry, 1962). Thus, just as Orton (1925) suggested, it seems reasonable to
look at both hemispheres and their interaction with each other. In this respect, in a
modification of Orton's model, Beale et ai. (1972) and Corballis & Beale (1993)
argued that a visual representation of a shown object, such as a letter or a word, is
recorded adequately in each hemisphere. A mirror-image generalization, as a result
of interhemispheric communication (Achim & Corballis, 1977; Beale et aI., 1972;
Corballis & Beale, 1970), does not occur in the initial laying down of traces as
supposed by Orton (1925), but rather with the transferee of traces from one
hemisphere to the other. Thus, mirror-image generalization describes an effect of
memory formation and not of visual and phonological decoding of a letter per se. In
this respect, within the context of the Functional Coordination Deficit model,
reversal errors in reading are explainable beyond Orton's (1925) theory as occurring
during the coordination process in reading (see above). This process includes the
storage and retrieval of visual bottom-up information which is assumed to be
represented symmetrically in the cortex, while the phonological representations are
organized asymmetrically. This induces confusion because different phonological
codes ("bee" vs. "dee") might be appropriated to the same visual representation (b =
d) in memory. Thus, in correspondence with Orton (1925), reversals are assumed to
be an effect of labeling (see also Bigsby, 1985). However, in contrast to him, they
are seen as a memory effect. This especially explains reversals in writing. According
to this approach, an important factor in learning to read is the suppression of mirror-
image information (of graphemes), or better, their connection to the phonological
label. This can be ensured by a hemispheric dominance and a shift of magnitude of
the representation between the hemispheres, such as a stronger visual record in one
hemisphere, which leads to a stronger reversion in the other. Still, both orientations
represent the stimulus shown, but there will be a stronger connection between the
180 T.LACHMANN
label and that part of the representation which reflects the configuration in its
original (physical) orientation. We can assume that hemispheric asymmetry is not
only a phylogenetic (evolutionary) but also an ontogenetic (developmental) stage
(e.g. Biyasheva & Shvetsova, 1993). An impairment in the development of
hemispheric dominance or of the pattern of interhemispheric communication in
general, results in a deterioration of the coordination process. Since bottom-up
information cannot be used adequately, there will be either a need for more time
and/or a higher probability of reversals in reading and writing.
2.5 Symmetry-Generalization in Memory
There is a question that still arises regarding the model of mirror-image

generalization: How can non-vertically symmetrical reversals be explained? First of
all, it should be said that it is not the purpose of this chapter to defend Orton or any
modification of his model. Rather, it is intended to emphasize that reading describes
a complex cognitive technique which requires the coordination of multiple functions
and the very quick and accurate integration of information within and between
different perceptual and representational levels. From this point of view, a differing
interhemispheric communication, and thus a difference in binding symmetrical and
unsymmetrical represented information, becomes important. More generally, the
deterioration of the coordination of visual and phonological information represented
in working memory, can be seen as one reason for a failure in reading. In principle,
this also holds for non-vertical symmetry. Experiments on symmetry detection can
be helpful to understand how symmetry of different orientation is represented in the
brain. It was shown that the relative advantage of vertical symmetry in comparison
to symmetries at other orientations disappears when the pattern is presented away
from the fixation point (e.g. Bornstein & Krinsky, 1985; Herbert & Humphrey,
1996; Wenderoth, 1995). This finding substantiates the original callosal hypothesis,
but it also suggests that there are additional mechanisms (not specifically explained
by the authors) for detecting non-vertical symmetry, as well as for detecting vertical
symmetry away from the point of fixation. For instance, Palmer & Hemenway
(1978; see also Bornstein, Ferdinandsen, & Gross, 1981; Wenderoth, 1995 and
contributions in Lockhead & Pomerantz, 1994) have shown that there is not only an
advantage for vertical symmetry, but also, albeit a smaller one, for horizontal
symmetry, relative to diagonal symmetries. Results by Lachmann (2000, pp. 188-
193; Lachmann, in preparation) confirm such effects for a task that involves
working memory. The participants of his study had to judge two successively
presented dot-patterns (as used by Garner & Clement, 1963) as same not only when
they were identical (repetition) but also when they were symmetrical to another
(transformation ally related by rotation in multiples of 90° or reflection on any axis,
producing vertical, horizontal, and oblique symmetrical pattern pairs), and as
different otherwise. The reaction times (RT) found for same responses followed a
linear function; performance was fastest for identical patterns, followed by
vertically, horizontally, and lastly diagonally symmetrical patterns. In some cases
(patterns with imaginary straight lines), there was no RT difference at all between
identical patterns and patterns that were vertically symmetrical to each other (see
also Klatzky & Stoy, 1974 for picture matching), while the RT for other symmetries
still followed the linear trend.

Basically, there is theoretical and experimental support for the argument that the
visual representation of a stimulus configuration in memory is to a certain extent
invariant against any reflection. Therefore, we should use the term symmetry
generalization with the understanding of mirror-image generalization as a special
mechanism which prevails in the evolutionary stage of a vertically symmetrical
organization of the nervous system. Admittedly, we do not have a physiological
model, such as the callosal transfer for vertical symmetry, to explain an overall
symmetry generalization. At least in the context of Hebb' s theory (1949) of neuronal
representations, it could be argued that a memory unit (engram), represented by a
neuronal cell-assembly which is activated by a perceived object, consists of neurons,
the majority of which are identical to those of the assembly, which would have been
activated by a symmetrical version of that particular object. Such an overlap of
assemblies (which can also be assumed for phonological representations,
Pulvermiiller, 1996), may be a result of their frequently synchronized activation (in
the sense of Hebb's theory), and thus depend on the behavioral relevance of the
orientation of the symmetry.
Bornstein and his colleagues (1981, 1985) showed that four month old infants
have no preference for symmetry at all, but process vertically symmetrical stimuli
faster than others, whereas 12 month old infants prefer vertical symmetry to
horizontal symmetry and asymmetry. This pattern of results, as well as results
mentioned earlier, suggest that the special role of vertical symmetry might indeed be
a consequence of the symmetrical organization of the cortex (or vice versa) and
interhemispheric communication (callosal transfer), which leads to a faster
performance but not to a preference of vertical symmetry in newborns. However, the
advantages of other symmetries and the preference of vertical symmetry in older
infants could be explained as a learning effect, the graded character of which results
as a consequence of behavioral relevance. Regarding the hypothesis that the
evolution of the nervous system can be described as a progression from spherical to
radial to bilateral symmetry and finally to asymmetry (e.g. Tschirgi's 1958, cited in
Corballis & Beale, 1970) and considering that this progression reflects in some
respect also the individual (ontogenetic) development, we could also assume a
genetic based predisposition for learning symmetry generalization and its graded
orientation preference!.
2.6 Reversals as an Effect of Symmetry-Generalization in Reading
Symmetry generalization, especially mirror-image generalization, warrants a

significant advantage for visual processing of objects (and scenes) and consequently
for behavior, not only of humans, but also of higher animals. Thus, it is strongly
related to the architecture of the nervous system and its evolution. However,
language (and thus reading) seems to be unique for humans (Bierwisch, 1999;
Chomsky, 1959; see also Byrne, 1995 for discussion) and related to a specific
tendency towards the development of asymmetrical functions in the human cortex 2•
As mentioned earlier in this chapter, the visual processing in reading (as well as the
auditory processing in spoken language) is different than the processing of visual
object information (or auditive environmental stimuli). Letters, words, sentences,
182 T.LACHMANN
numbers, punctuation marks, etc. have no direct behavioral relevance; their meaning
is coded abstractly. What we perceive within a text are symbols. Consequently, both
the evolutionary and individual development of language, and thus of reading, imply
symbolic representation (in addition to associative learning, Chomsky, 1959; see
also Byrne, 1995; Klix, 1985). The processing of letters (or words) as visual icons
requires the internal representation of rules which guarantee its desymbolization,
and thus making the icon into a symbol (cf. Deacon, 2000). A hebrew letter means
nothing to a German, and thus is not a symbol to her/him, as long as she/he has not
learned its correspondence to a certain sound or meaning. In this respect, the
technique of reading can be characterized as a highly automated desymbolization
process. While symmetry generalization is beneficial to vision directly related to
behavior, it is detrimental for vision as part of a symbolic processing such as
reading, especially if symmetrical counterparts of visual symbols are related to
different sounds or meanings. However, this does not implicate that there has to be a
separate visual system for reading. Instead, the evolutionary as well as the individual
development is mostly characterized by a differentiation of already existing
functions. We suggest that a part of the visual differentiation required for reading is
the suppression of symmetry-generalized representations of visual icons in working
memory as a precondition of an automated de symbolization process. The
development of this visual differentiation has to take place in the early non-
alphabetic phases of reading acquisition (before letters function as symbols: Ehri's,
1999, pre-alphabetic & partial alphabetic phase) and is accompanied by typical
mistakes, including reversals, made by children learning to read. Therefore, we
postulate that not a dysfunction of the visual system, rather an incomplete functional
specification or its developmental delay hinders the fast desymbolization process
(consolidated alphabetic phase, cf. Ehri, 1999). Consequently, reading as a process
of coordination of parallel operating functions and fast integration of non-ambiguous
information will be impaired. This explains not only reversals but also other errors
reflecting a general deterioration of the coordination process.
3. ABOUT SOME MISUNDERSTANDINGS: INTERIM SUMMARY AND

CONCLUSIONS
To summarize at this point, we do not assume, just as Orton (1925), Corballis &
Beale (1993), and others (see Willows & Terepocki, 1993) did not assume that
reversals are a result of a visual deficit in the sense that a reading disabled person
cannot see the configuration of the letter "b" as what it is. However, even (or maybe
especially) in contemporary literature, reversal research is often characterized as
visual deficit research, and reversal errors are described as "perceiving letters and
words as reversed forms" (Vellutino & Scanlon, 1998). This may reflect some
ambiguities about the terms "visual" and "perceptual" and the resulting
misunderstandings. As argued earlier, we assume that reversals occur in memory
and not in perception, due to the deterioration of the fast integration of phonological
and visual information in working memory. Some authors (e.g. Vellutino, 1987),
however, argue that reversals are "usually blamed on defects in visual perception".
This is simply not true, not even for the original theory of Orton (1925), who
emphasized that strephosymbolic children have no "visual deficit in the ordinary
sense" and can see just as well as others (1925). The diagnosis of strephosymbolia
even implies by definition that the performance on the basis of the assumed first two
levels of visual processing, such as visual recognition, is not impaired. The names
"visual word center" (Hinshelwood, 1917) or "visual associative level" (Orton,
1925) might contribute to such misunderstandings. However, as we argued earlier in
this chapter, the term "visual" was used in the context of an assumed seriality of
visual perception, visual recognition, and sound and meaning association, realized
within a type of word-reading system which is a part of the visual cortex. In this
respect, Vellutino (1977) would be in agreement with Orton (1925) when he argues
that reversals in poor readers are not caused by visual perceptual deficits, but rather
by the "difficulty in establishing visual-verbal relationships" (pp. 337). This
becomes especially clear for reversals in writing (e.g. examples in Orton, 1925, or in
Willows & Terepocki, 1996). Reversals are neither assumed to be a poorly visual
phenomenon, nor do they reflect a perceptual deficit. In fact, there are numerous
theories of visual perceptual deficits in disabled readers (Badcock & Lovegrove,
1981; Becker, Lachmann & Elliott, 2001; Lehmkuhle, 1993; Lovegrove, Martin, &
Slaghuis, 1986; Stanley & Hall, 1973; see also the chapter of Friederici &
Lachmann this volume), but they cannot be seen as following in the direct tradition
of Hinshelwood, Morgan or Orton at all. At this point it becomes clear that the
comparison and interpretation of the large amount of data that has been collected,
and the evaluation of the theories that have been developed over the last century,
imply an agreement of the concepts that are discussed and the consideration of
several questions to be introduced in the next section.
4. THE INCONSISTENCY OF RESULTS IN READING DISABILITY

RESEARCH
4.1 The Conditions for 1nterpreting Experimental Results in Reading Disability

Research
The sample size (N) of the experimental groups of normal readers and readers
defined as dyslexic may dramatically influence the significance of the experimental
results. A study that rejects the importance of reversals due to the null-finding in a
disabled reader sample of N=lO may not be acceptable for someone who is favoring
a multicausal approach. Moreover, in order to compare and interpret the variety of
experimental results published concerning the discussion about the significance of
reversals for reading disability, more generally the causal role of visual versus
phonological deficits, it is necessary to consider at least the following questions:
• What definition for reading disability is used for sample formation?

• What function(s), assumed to be important for reading, does the
experimental task require?
• What methods are used to investigate the function(s)?
We introduced the term functional fragmentation (see above) to describe a kind of

dilemma of experimental reading research. In this respect, we argued that on one
hand reading is already a performance and its pattern can be used to distinguish
184 T.LACHMANN
between reading disabled and normal readers. On the other hand, it was argued that
in order to investigate the origin of the difference in performance between reading
disabled and normal readers, the experimenter has to test separate functions which
he assumes to be substantial for reading. However, there is a controversial
discussion in the literature about both aspects which are strongly related to the
definition of dyslexia, and the question of whether or not dyslexia is special to a
general reading disability 3. This dissension is one of the reasons for contradicting
results and interpretations in experimental reading disability research, because it
leads to different criterions for distinguishing subpopulations of normal readers and
dyslexics, or of normal, poor (backward), and dyslexic readers for the experimental
design. Furthermore, it constitutes the selection of functions to be tested. The
definition of developmental dyslexia or general reading disability, and thus for
creating an experimental subgroup (we do not account here for acquired cases) rests
either upon symptoms, i.e. observable and measurable characteristics of reading (and
writing) performance (e.g. reversal errors, Orton, 1925; reading tests), etiological
assumptions (e.g. temporal integration of auditory information, Tallal, 1999; see also
Merzenich et aI., 1996; Farmer & Klein, 1995; e.g. phonological deficit, Snowling,
1998; e.g. proximal versus distal causes, Coltheart & Jackson, 1998), including uni-
versus multicausality (see Snowling, 1998; see also Miles, 1995; Siegel & Ryan,
1989), or education and treatment results (e.g. persistence despite education in
school, cf. Tonnessen, 1995; for general overview see Miles, Haslum, & Wheeler,
1999, Miles, 1995; contributions in van den Bos, Siegel, Bakker, & Share, 1994).
4.2 The Discrepancy Criterion and the Definition of Dyslexia
One of the main points of critique about the definition of dyslexia (as well as about
the definition of other learning disabilities) is the so-called discrepancy criterion
(e.g. Miles, et aI., 1999; Siegel, 1989, 1992; Snowling, 1998; Tonnessen, 1995; Toth
& Siegel, 1994; see also the chapter of Jimenez in this volume). This criterion relies
on an assumed disparity between expected and actual achievement in reading which
is not caused by diagnosed organic deficiencies in vision or hearing. In this respect,
reading disabled are identified in the majority of the literature by a tested reading
performance which is two years (or more) below what is expected from their
potential. But how should this potential be determined? It could be done with a
comparison to either an age-matched or grade-matched reference population. In
principle, both the degree of general development and the degree of instruction to /
experience in reading must be regarded in this comparison in order to estimate the
reading potential of an individual or a subgroup. However, this is not always
possible. In Germany, for instance, students diagnosed with a difficulty in learning
to read have the chance to attend a special class after second grade where they repeat
the curriculum of the second grade. Therefore, age-matched samples of normal
reading and reading disabled students may differ in their level of reading instruction
(and vice versa, Lachmann & Brendler, in preparation). Moreover, estimating the
reading potential by comparison to a reference population introduces further
problems in diagnosing: adolescents, adults (for grade- and age-matching,
controlling the amount of reading experience, etc.), students with different cultural
or regional backgrounds, students with different curricula and teaching methods,
FUNCTIONAL COORDINATION DEFICIT IN READING DISABIUTY 185
students with general developmental delays or learning disabilities, students with

attention, motivation, or emotional problems, and so forth. The common definition
of dyslexia (e.g. World Health Organization) tries to account for these factors by
implying that the discrepancy occurs in spite of conventional instruction,
comparable socio-cultural opportunity, and - most importantly - adequate cognitive
abilities reflected by general intelligence. In other words, the reading potential is
estimated by a comparison of reading performance to both that of a reference
population (of same age, grade, educational and socio-cultural background, etc.) and
an individual's general cognitive ability. Therefore, the standard classification
systems (e.g. International Classification of Diseases of the World Health
Organization; World Federation of Neurology) require normal scores on
standardized intelligence tests for the diagnosis of dyslexia. Most of the criticism
about the discrepancy definition is based on the latter assumption because it implies
that the diagnosis of dyslexia directly depends on measures of intelligence (e.g.
Stanovich, 1991; see also Toth & Siegel, 1994; Tonnessen, 1995; Fletcher et aI.,
1998; Jimenez Gonzales & Rodrigo LOpez, 1994; Vellutino, Scanlon, & Lyon,
2000; see also the chapter of Jimenez this volume). Besides the fact that intelligence
is in some respect a "fuzzy concept" (cf. Gustafson & Samuelsson, 1999) and also
can only be estimated by a comparison to a reference population, this leads to a
distinction between dyslexics/specific reading disabled (fitting the definition) and
poor (backward) readers (Rutter & Yule, 1975). Even though the latter group may
read at the same level as dyslexics of the same age, they are not defined as such,
because they are not expected to read better. For the justification of such a
distinction, there is mixed empirical evidence (Coltheart, & Jackson, 1998; Siegel,
1992; Toth & Siegel, 1994; Vellutino et aI., 2000; see contributions in van den Bos
et aI, 1994 for review) and antithetic theoretical argumentation (independence of
intelligence and reading, e.g. Siegel, 1989). There is a considerable amount of
literature suggesting that reading disability should be defined by a discrepancy in
underlying skills (naming-speed, phonological memory, e.g. Ho & Lai, 2000, Ho et
aI., 2000) and etiological factors instead (see above; e.g. Siegel, 1992; see also
Snowling, 1998; Miles et aI., 1999; Uhry, 1993, and this volume; Coltheart &
Jackson, 1998). However, this is seen as almost impossible by others, regarding the
present state of knowledge (Tonnessen, 1995). Some authors even argue that
dyslexia, as a highly specific condition, cannot be defined in any agreed way (see
e.g. the dispute between Presland, 1991 and Miles, 1991; see also the chapter of
Friederici and Lachmann, this volume).
The discussion about the definition of dyslexia was outlined here in great detail not
in order to take sides, but rather to point out that this dispute reflects a reason for
contradicting results in experimental reading disability research, including the
controversy about the significance of reversals in reading disability. The variety of
definitions leads to different ways of sample formation (e.g. based on symptom
principle), different selection of functions to be tested (e.g. based on etiology
principle), varying evaluations of the teaching effects, varying experimental
methods, and even a different way of data interpretation. The term "poor readers",
for instance, is not always used in the literature to distinguish this group from
"dyslexic" children. For example, in the paper of Willows and Terepocki (1993), a
review of reversal research, the terms "dyslexic readers", "disabled readers", and
"poor readers" are used synonymously and contrasted with "normal readers", or
186 T.LACHMANN
"average readers", matched according to age and cognitive ability (although they
mention that neuropsychological models use a distinction between dyslexic and poor
readers, which may be irrelevant beyond clinical studies, pp. 42). That is why not all
of the reviewed papers are really comparable (which the authors themselves
conceded). Extreme cases representative of the problematic conditions will now be
explored: the work of Orton (1925), who assumes reversals as cardinal symptom in
dyslexia, the work of Liberman et al. (1971), which is often cited in the literature as
evidence against the significance of reversal errors and as a turning point from
neurological to linguistic based models in reading disability research, and the work
of Fischer et al. (1978), a follow-up study to Liberman et al. (1971; for further
review of reversal studies see Willows & Terpocki, 1993).
4.3 Revisiting Classical Studies
The study of Liberman et al. (1971) was aimed directly at testing Orton's (1925,
1929a) theory of anomalous cerebral dominance as the universal cause for
developmental dyslexia (see above) experimentally. In this connection, the
investigation was concentrated on the analyses of the proportion of reversal errors to
other errors (Orton's secondary symptoms), especially of the relation between errors
based on optical (reversals) and on linguistic properties (vowel and consonant
errors), and of the relationship between reversals of sequence (kinetic reversals) and
reversals of orientation (static reversals). The experimental subgroup of poor
readers, the lower third (18 students) of the second grade of an elementary school,
was defined by their performance (measured error rate) in reading a list consisting of
60 monosyllabic words, including primer-level sight words, non-sight words, and
words where reversals (both types) were possible and not possible (e.g. get, was,
bed, bet, not, pin). The remaining 36 students were counted as normal readers (age
and IQ matched samples). Two students were excluded from the sample because of
speech impairment. This should be mentioned because Orton later emphasized that
reading/writing and speech impairments are strongly related to each other and may
be ascribable to the same cause, an anomaly in cerebral dominance (Orton, 1937). In
addition to the word list, all students had to perform the Gray Oral Reading Test and
a matching task, where a given letter was to be matched to one out of a group of
five, four of which were reversible. The results were analyzed for the whole
population as well as for poor and normal readers separately. Liberman and her
colleagues found a significant correlation between the total performance in the
reading test and the more artificial task of reading monosyllabic words in isolation.
This may reflect a stage in reading development which is based on scanning
syllables rather than larger chunks of text. The main results the authors found were:
1. nearly all of both kinds of reversals were done by the poor reader group;
2. vertical b-d confusion (which they called horizontal transformation)
occurred with as great a frequency (10.2 in average for both directions) as
b-p confusion (11.4 in average for both directions), while both confusions
obviously occurred more often than d-p confusion (1.1) as well as
confusions of b, d, and p with g (0.9 in average for both directions);
3. for static reversals the presence of equivalent shapes within the alphabet is
important, and the reversibility of a letter is not by itself a sufficient

condition for confusion;
4. normal readers as well as poor readers made less reversal errors than other
errors (error percentages according to the opportunities of occurrence in
poor readers: 6.3 kinetic reversals, 12.7 static reversals, 16.3 consonant
confusion, 26.8 vowel confusion; error percentages according to the total
number of errors in the same sequence: 10, 15, 32, 43);
5. within the group of poor readers, individual differences were larger and test
- retest reliability was lower for reversals in comparison to vowel and
consonant errors, indicating that reversals do not reflect a constant
proportion of all errors as is expected from assuming them to be cardinal
symptoms;
6. static and kinetic reversals in poor readers were found to be uncorrelated to
each other and both types were correlated with other measures, such as the
reading test, in a different way.
These findings were counted as evidence for the importance of linguistic

determinants in reading problems and against the theory of strephosymbolia as a
consequence of hemispheric imbalance reflected by reversals. However, these
results are representative for the lower third of a regular second grade class only,
and not for dyslexics, not for strephosymbolics, and not even for poor readers. Orton
was confronted with students sent by their teachers or parents because they were
failing to learn to read. He (Orton, 1929c) identified two to four percent of this
students as strephosymbolic children by measuring the degree of reversal errors.
That is why the results of Liberman and her colleagues are not sufficient to disprove
Orton's concept of developmental word-blindness. The merit of their work consists
much more of a stronger emphasis on linguistic aspects in understanding reading
disability. The pUblication can in fact be counted as a milestone in reading disability
research, but not as evidence against the significance of reversals in general. In a
follow-up study, published by Fischer et al. (1978), the performance of the poor
reading students described by Liberman and her colleagues was compared to that of
dyslexic children, as defined by the discrepancy criterion (retardation of at least 18
months in Gray Reading Test performance), of the same age and intelligence. The
authors could show that while both groups performed almost equally in the reading
test, the dyslexic group (smaller data set) made significantly more errors in the Word
List Test. However, these errors were relatively made in the same proportion as by
the poor readers group (error percentages according to the opportunities of
occurrence: 8.3 kinetic reversals, 11.4 static reversals, 32.9 consonant confusion,
40.3 vowel confusion), underlining the importance of linguistic aspects for the error
rate. In contrast to the poor readers, the dyslexic children performed more
consistently in their reversal errors and both kinds were significantly correlated (in
contrast to findings 5 and 6 of poor readers, see above).
4.4 Old Results with a New Perspective
The results of Liberman et al. (1971) and Fischer et al. (1978) that most of the errors
done by poor readers and dyslexic children are consonant and vowel errors rather
188 T.LACHMANN
than reversals, allowed some doubts about the secondary character of linguistic
errors to emerge, even though Orton never denied their frequency and importance
(maybe today, he would term reversals as linguistic errors, understanding the
distinction only as an etiological one, see above). Besides the fact that there would
be a considerable change in the pattern of results of both studies when disregarding
the debatable reversibility of the letter "g" and regarding the letter "q" instead, there
is still the question of whether or not dyslexic students, as defined by Fischer et aI.,
are comparable to strephosymbolic children (defined by reversals plus discrepancy
criterion). On the other hand, both studies showed that nearly all of the reversals
were done by the poor readers and the dyslexic children, suggesting that reversals
are in fact significant for reading problems. Beyond the unicausal model of Orton,
the complete pattern of results is well explainable within the context of the
introduced multicausal model which explains reading disability as a failure in
functional coordination and information integration. For a certain subset of the
dyslexic and the poor reading children, the reading problems maybe caused by a
symmetry generalization which is inappropriate in reading and hinders a non-
ambiguous association between the visual and the phonological representation of a
letter, and thus a quick and accurate desymbolization. To conclude, according to this
model it is in fact expected that:
(1) the majority of reversal mistakes are made by dyslexics;

(2) there occurs not only mirror-image confusion but also a confusion for all forms
of symmetry;
(3) within a context, such confusion depends on the existence of equivalent letter
shapes;
(4) only a subgroup of the reading disabled show an accumulation of reversals, and
vowel and consonant confusion occurs more frequently in reading disabled as
well;
(5) the dyslexic group makes reversal errors more consistently than the poor reader
group;
(6) depending on the formation of the reading disabled group, there mayor may not
be a correlation between static and kinetic reversals.
Consequently, while Liberman et ai. and Fischer et ai. are often cited as evidence
against the significance of reversals for explaining a failure in reading, the same
results may be interpreted in support of their significance when employing a
different etiological model for their occurrence. Moreover, it becomes clear that the
theoretical background also determines the understanding of a reversible letter (e.g.
letter "g" may diminish the reversal error rate) and the selection of functions to be
tested (e.g. phonological confusion). If the authors of both studies had tested their
subjects only for reversals, then the same pattern of results could have been
interpreted as evidence for the cardinal importance of reversals in explaining reading
disability (see e.g. Patton, Yarbrough, & Thursby, 2000). Furthermore, the
dependency of results on the definition of reading disability (poor readers versus
dyslexics versus strephosymbolics) becomes clear. Most important in this respect as
well as for the interpretation of the results, is the unicausal versus multicausal
orientation of the underlying model. This also regards the problem of the functional
fragmentation in experimental reading disability research (see above). In this
FuNCTIONAL COORDINATION DEFICIT IN READING DISABILITY 189
respect, it appears more reasonable to test the significance of reversals for reading
disability by analyzing the reading performance first. If there is a subgroup of
reading disabled children which can be identified as performing in reading and
writing with a significantly higher proportion of reversal errors (strephosymbolic
children), then this subgroup should be analyzed further to investigate the nature of
these mistakes experimentally (cf. Lachmann and Brendler, in preparation; Brendler
& Lachmann, 2001). However, there are further aspects that have to be considered
for such an investigation. As representatives, some of these aspects will be discussed
in the following paragraph.
4.5 Further Conclusions

Once again, the theoretical approach determines the experimental design and the
interpretation of the results. If reversals are assumed to be a visual effect, e.g. they
are supposed to reflect a general deficit in visual space perception and orientation
(Goins, 1958; see Willows & Terepocki ,1993, for a general review and Terepocki et
aI, cited there), then the researcher may compare the performance of poor readers (or
dyslexic children) and normal readers, for instance, using an orientation task
involving non-alphabetic stimuli. However, for testing a model that assumes
reversals as occurring within the association between the visual and the
phonological representation (for instance, the concept of strephosymbolia, mirror-
image or symmetry generalization), the null-finding in such studies is non-decisive.
Moreover, a significant difference found between the reader groups in such cases
would even question an explanation of reversals as occurring in symbol-sound
coding, unless the framework is multicausal, as that of the Functional Coordination
Deficit model. Numerous experiments were done to study the nature of reversal
errors by involving verbal material (such as the Liberman et al. study, see above).
However, they differ strongly in their experimental design. A verbal labeling was
either possible or required to perform the task (cf. Bigsby, 1985). Isolated letters
were exposed for different durations to be named or reproduced, or to be compared
with other letters in varying orientations or name codes, presented either
simultaneously or with a certain delay (e.g. Bigsby, 1985; Brendler & Lachmann,
2001; Corballis, Macadie, & Beale, 1985). However, for the justification of the
assumption that reversals are caused by a failure in the integration of (symmetry-
generalized) visual and phonological information represented in memory, it is
crucial to consider if the task requires perception, recall or recognition and on what
degree an information integration is required. This also holds true for studies using
nonwords, which represent a lower degree of functional fragmentation. For instance,
in order to test the nature of orientation errors, and the visual deficit hypothesis in
general, Vellutino, Steger, & Kandel (1972; Vellutino, 1979) compared disabled and
normal readers in their ability to reproduce visual presented nonwords (amongst
other items) graphically and orally (naming). The fact that the performance of poor
and normal readers was comparable in the graphic reproduction but significantly
different in naming, was interpreted as evidence for the linguistic rather than the
perceptual nature of reading disability in general. Again, such results may be used as
an argument against an unicausal specific visual recognition deficit in reading
disabled. However, they can neither be availed to argue generally against the
190 T.LACHMANN
significance of reversal errors for reading disability (as often done), nor be used as
evidence against visual components leading to a failure in information integration in
reading disabled students. One of the main messages of this chapter is that the
question about visual deficits in reading disabled children is not identical to the
question about the significance of reversal errors in reading disability.
5. SHORT SUMMERY
A big lack in the discussion about the nature of reading disability and its remediation
exists due to an absent conceptual agreement, which finally leads to an
incompatibility of experimental results and their interpretation. Reading disability
was explained as a deficit in functional coordination resulting in a failure of the
integration of information in a degree of speed and accuracy required for reading.
The problem of functional fragmentation was introduced, according to which results
of testing a single function may not necessarily be meaningful in order to conclude
on a process that coordinates multiple functions.
6. AFFILIATIONS
Dr. Thomas Lachmann

University of Leipzig
Department of Psychology
Seeburgstr. 14/20
D-04103 Leipzig
Germany
E-mail: lachmann@uni-leipzig.de
7. NOTES
1 To speculate, as we know from mental rotation research (Cooper, 1976; Shepard & Metzler, 1971;
Ruthruff, Miller, & Lachmann, 1995), the decision whether or not objects of varying orientations are
mirrored requires no mental transformation if the objects are rotated less than about 30° from upright
position. Regarding this, if in both hemispheres there is a vertically mirror-image generalized
representation, then an additional symmetry generalization in memory (horizontal and diagonal
reflection) would include the representation of any rotational orientation (max. distance 22.5° to the
next representation).
2 Arguments for a uniqueness of the human brain and the relation to the development of consciousness,
thought, and a theory of mind will not be discussed here in detail; see e.g. the contributions in
Corballis & Lea, 1999 or Byrne, 1995.
J We did not use the term dyslexia above, because of this controversy, see also Stanovich (1994).
8. REFERENCES
Achim, A., & Corballis, M. C. (1977). Mirror image equivalence and the anterior commissure.
Neuropsychologia, 15, 475-478.
Annett, M., & Kilshaw, D. (1984). Lateral preference and skill in dyslexics: Implications of the right shift
theory. J oumal of Child Psychology & Psychiatry & A /lied Disciplines, 25, 357-377.
Bachmann, F. (1927). Uber die kongenitale Wortblindheit. Berlin: Karger.
FUNCTIONAL COORDlNAT[ON DEF[C[T [N READ[NG D[SAB[LlTY 191
Badcock, D., & Lovegrove, W. (1981). The effects of contrast, stimulus duration, and spatial frequency
on visible persistence in normal and specifically disabled readers. Journal of Experimental
Psychology: Human Perception and Performance, 7, 495-505.
Baddeley, A. D. (1986). Working memory. Oxford: Clarendon Press.
Barlow, H. B., & Reeves, B. (1979). The versatility and the absolute efficiency of detecting mirror
symmetry in random dot displays. Vision Research, 19, 783-793.
Beale, I. L., Williams, R. J., Webster, D. M., & Corballis, M. C (1972). Confusion of mirror images by
pigeons and interhemispheric commissures. Nature, 238, 348-349.
Becker, C, Lachmann, T., & Elliott, M. (2001). Evidence for impaired integration-segmentation
processes and effects of oscillatory synchrony on stimulus coding in dyslexics. In E. Sommerfeld, R.
Kompass, & T. Lachmann (Eds.), Fechner Day 2001. Proceedings of the International Society for
Psychophysics (pp. 273-278). Lengerich, Berlin: Pabst.
Berti, S., Geissler, H.-G., Lachmann, T., & Mecklinger, A (2000). Event-related brain-potentials
dissociate visual working memory processes under categorical and identical comparison conditions.
Cognitive Brain Research, 9, 147-155.
Bierwisch, M. (1999). Words as programs of mental computation. [n A D. Friederici, & R. Menzel
(Eds.), Learning: Rule extraction and representation (pp. 3-35). Berlin: de Gruyter.
Bigsby, P. (1985). The nature of reversible letter confusions in dyslexic and normal readers:
Misperception or mislabeling? British Journal of Educational Psychology, 55, 264-272.
Binet, A (1905). Analyse de C E. Spearman, the proof and measurement of association between two
things and general intelligence objectively determined and measured. L 'Annie Psychologique, 11,
623-624.
Bishop, D. V. M. (1997). Cognitive neuropsychology and deve[opmental disorders: Uncomfortable
bedfellows. Quarterly Journal of Experimental Psychology. A, Human Experimental Psychology, 50,
899-923.
Biyasheva, Z. G., & Shvetsova, E. V. (1993). Information approach to analysis of the EEG age dynamics
in 7-18-year-old boys during mental arithmetic. Human Physiology, 19, 327-331.
Bornstein, M. H., Ferdinandsen, K., & Gross, C G. (1981). Perception of symmetry in infancy.
Bornstein, M. H., & Krinsky, S. J. (1985). Perception of symmetry in infancy: The salience of vertical
symmetry and the perception of pattern wholes. Journal of Experimental Child Psychology, 39, 1-19.
Bradley, L. & Bryant, P. E. (1978). Difficulties in auditory organization as a possible cause of reading
backwardness. Nature, 271, 746-747.
Bradley, L., & Bryant, P. E. (1981). Visual memory and phonological skills in reading and spelling
backwardness. Psychological Research, 43, 193-199.
Brend[er, K., & Lachmann, T. (2001). Letter reversals in the context of the Functional Coordination
Deficit Model. [n E. Sommerfeld, R. Kompass, & T. Lachmann (Eds.), Fechner Day 2001.
Proceedings of the International Society for Psychophysics (pp. 308-313). Lengerich, Berlin: Pabst.
Broca, P. (1865). Sur Ie siege de la faculte du [angage articule. Bulletin of the Society of Anthropology, 6,
377-396.
Brodmann, W. (1909). Vergleichende Lokalisationslehre. Leipzig: Barth.
Byrne, R. W. (1995). Primate cognition: Comparing problems and skills. American Journal of
Primatology, 37, 127-141.
Campbell, A W. (1905). The localization of cerebral functions. Cambridge: University Press.
Chomsky, M. (1959). A review of B. F. Skinner's "Verbal Behavior". Language, 35, 26 - 58.
Coltheart, M., & Jackson, N. E. (1998). Defining dyslexia. Child Psychology & Psychiatry review, 3, 12-
16.
Cooper, L. A (1976). Demonstration of a mental analog of an external rotation. Perception &
Psychophysics, 19, 296-302.
Corballis; M. C (1986). [s mental rotation controlled or automatic? Memory & Cognition, 14, 2, 124-128.
Corballis, M. C. (1988). Recognition of Disoriented Shapes. Psychological Review, 95, 1, 115-123.
Corballis, M. C, & Beale, I. L. (1970). Bilateral symmetry and behavior. Psychological Review, 77, 451-
464.
Corballis, M. C, & Beale, I. L. (1993). Orton revisited: Dyslexia, laterality, and left-right confusion. [n
D. Willows, R. S. Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp.
57-73). Hillsdale, NJ.: Erlbaum.
Corballis, M. C, Macadie, L., & Beale, I. L. (1985). Mental rotation and visual laterality in normal and
reading disabled children. Cortex, 21, 225-236.
Corballis, M. C., Macadie, L., Crotty, A, & Beale, I. L. (1985). The naming of disoriented letters by
normal and reading-disabled children. Journal of Child Psychology & Psychiatry & Allied
192 T. LACHMANN
Disciplines, 26, 929-938.

Corballis, M. c., & Lea, S. E. G. (Eds.). (1999). The descent of mind: Psychological perspectives on
hominid evolution. New York: Oxford University Press.
Cowan, N. (1997). Attention and memory. An integrated framework. New York: Oxford University Press.
Cowan, N. (1998). Visual and auditory working memory capacity. Trends in Cognitive Sciences, 2, 77-
80.
Cowan, N. (2000). Childhood development of some basic parameters of working memory. In E.
Schroeger, A Meckiinger, & A Friederici (Eds.), Working on working memory. Leipzig series on
cognitive sciences 1 (pp. 139-160). Leipzig: Leipziger Universitaetsverlag.
Critchley, M. (1966). Is developmental dyslexia the expression of minor cerebral damage? Slow Learning
Child,13, 1, 9-19.
Cummings, W. J. K (1970). An anatomical review of the corpus callosum. Cortex, 6, 1-18.
Deacon, T. W. (2000). Evolutionary perspectives on language and brain plasticity. Journal of
Communication Disorders, 33, 273-291.
Dearborn, W. F. (1933). Structural factors which condition special disability in reading. Proceedings of
the American Association on Mental Deficiency, 38, 268-283.
Doehring, D. G., Trites, R. L., Patel, P. G., & Fiedorowicz, C. A M. (1981). Reading disabilities: The
interaction of reading, language and neuropsycological deficits. New York: Academic Press.
Ebbinghaus, H. (unter Mitwirkung von Duerr, E.). (1913). Grundzuege der Psychologie. Zweiter Band.
Leipzig: Veit.
Ehri, L. C. (1995). Phases of development in learning to read words by sight. Journal of Research in
Reading, 18, 116-125.
Ehri, L. C. (1999). Phases of development in learning to read words. In J. Oakhill, & R. Beard (Eds.),
Reading development and the teaching of reading: A psychological perspective (pp. 79-108). Oxford:
Blackwell.
Ehri, L. c., & McCormick, S. (1998). Phases of word learning: Implications for instruction with delayed
and disabled readers. Reading & Writing Quarterly: Overcoming Learning Difficulties, 14, 135-163.
Everatt, J. (Ed.). (1999). Reading and Dyslexia. London: Routledge.
Everall, 1., McCorquidale, 8., Smith, 1., Culverwell, F., Wilks, A, Evans, D., Kay, M., & Baker, D.
(1999). Association between reading ability and visual processes. In 1. Everatt (Ed.), Reading and
Dyslexia (pp. 1-39). London: Routledge.
A review. Psychonomic Bulletin & Review, 2, 460-493.
Feagans, L. V., & Merriwether, A (1990). Visual discrimination of letter-like forms and its relationship
to achievement over time in children with learning disabilities. Journal of Learning Disabilities, 23,
417-425.
Fletcher, J. M., Francis, D. J., Shaywitz, S. E., Lyon, G. R., Foorman, B. R., Stuebing, K K, & Shaywitz,
8. A (1998). Intelligent testing and the discrepancy model for children with learning disabilities.
Learning Disabilities Research & Practice, 13, 186-203.
Fildes, L. G. (1921). A psychological inquiry into the nature of the condition known as congenital word-
blindness. Brain, 44, 286-307.
Fildes, L. G. (1923). No title. British Journal of Psychology, 14, 24.
Fisher, F. W., Liberman, I. Y., & Shankweiler, D. (1978). Reading reversals and developmental dyslexia:
A further study. Cortex, 14, 496-510.
Friederici, A D. (2001). Syntactic, prosodic, and semantic processes in the brain: Evidence from event-
related neuroimaging. Journal of Psycholinguistic Research, 30, 237-250.
Friederici, A D., & Menzel, R. (1999). Introduction. In AD. Friederici, & Menzel, R. (Eds.), Learning:
Rule extraction and representation (pp. 1-6). Berlin: de Gruyter.
Friederici, A D., Wang, Y., Herrmann, C. S., Maess, B., & Oertel, U. (2000). Localization of early
syntactic processes in frontal and temporal cortical areas: A magnetoencephalographic study. Human
Brain Mapping, 11,1-11.
Frith, U. (1985). Beneath the surface of dyslexia. In K. E. Petterson, J. C. Marshall, & M. Coltheart
(Eds.), Surface dyslexia: Neuropsychological and cognitive studies of phonological reading.
Hillsdale: Erlbaum.
Frith, U. (1986). A developmental framework for developmental dyslexia. Annals of Dyslexia, 36, 69-81.
Ford, C. A (1928). A case of congenital word-blindness showing its social implications. Psychological
Clinic, 17, 73-84.
Garner, W. R., & Clement, D. E. (1963). Goodness of pattern and pattern uncertainty. Journal of Verbal
Learning and Verbal Behavior, 2, 446-452.
Galaburda, AM. (1995). Developmental dyslexia and animal studies: At the interface between cognition
and neurology. In 1. Mehler, & S. Franck (Eds.), Cognition on cognition. Cognition special series
(pp. 31-47). Cambridge: MIT Press.
Gates, A. I. (1921). An experimental and statistical study of reading and writing tests. Journal of
Educational Psychology, 12, 303-314,378-391,445-464.
Gates, A I. (1927). The improvement of reading: A program of diagnostic and remedial methods. New
York: Macmillan.
Gates, A. I. (1936). The improvement of reading: A program of diagnostic and remedial methods. New
York, NY: Macmillan.
Gates, A. I. (1937). Diagnosis and treatment of extreme cases of reading disability. Yearbook of the
National Society for the Study of Education, pp. 391-416.
Gates, A. I., & Bernett, C. C. (1933). Reversal tendencies in reading. Causes, diagnosis, prevention and
correction. New York, NY: Teachers Colleg, Columbia University, Bureau of Publications.
Gates, A. I., Bond, G. L., & Russell, D. H. (1939). Methods of determining reading readiness. New York,
NY: Teachers Colleg, Columbia University, Bureau of Publications.
Geissler, H.-G., & Lachmann, T. (1996). Memory-guided inference in matching tasks: Symmetries and
the case of inferred sets. In S. C. Masin (Ed.), Proceedings of the 12th Annual Meeting of the
International Society for Psychophysics (pp. 119-124). Padua: The International Society for
Psychophysics.
Geschwind, N., & Galaburda, A. M. (1987). Cerebral lateralization: Biological mechanisms,
associations, and pathology. Cambridge, MA: MIT Press.
Gillingham, A, & Stillman, B. A. (1%9). Remedial training for children with specific disability in
reading, spelling and penmanship. Cambridge: Educators Publishing Service.
Goins, 1. T. (1958). Visual perceptual abilities and early reading process. Chicago: University of
Chicago Press.
Gordon, H. (1923). Hand and Ear Tests. BritischJournal of Psychology, 13, 283-300.
Greenberg, S. N., & Krueger, L. E. (1980). Limitations of the word superiority effect with a fixed target
set. Bulletin of the Psychonomic Society, 15, 25-28.
Gustafson, S., & Samuelsson, S. (1999). Intelligence and dyslexia: Implications for diagnosis and
intervention. Scandinavian Journal of Psychology, 40, 127-134.
Hall, R. (1945). Word-blindness: Its cause and cure. British Journal of Ophthalmology, 29, 467-472.
Hebb, D. O. (1949). The organization of behavior: A neuropsychological theory. New York: Wiley.
Heller, D. (1982). Eye movements in reading. In R. Groner, & P. Fraisse (Eds.), Cognition and eye-
movements (pp. 164-170). (East-) Berlin: DVW.
Herbert, A. M., & Humphrey, G. K. (1996). Bilateral symmetry detection: Testing a "callosal"
hypothesis. Perception, 25, 463-480.
Hermann, K. (1959). Reading disability: A medical study of word-blindness and related handicaps.
Kopenhagen: Munksgaard.
Hier, D. 8., LeMay, M., Rosenberger, P. B., & Perlo, V. P. (1978). Developmental Dyslexia: Evidence
for a subgroup with a reversal of cerebral asymmetry. Arch. Neurol., 35, 90-92.
Hinshelwood, J. (1895). Word-blindness and visual memory. Lancet, 2, 1564-1570.
Hinshelwood, J. (1907). Four cases of congenital word blindness occurring in the same family. British
Medical Journal, 2, 1229.
Hinshelwood,1. (1917). Congenital word blindness. London: Lewis
Ho, C. S.-H., & Lai, D. N.-C. (2000). Naming-speed deficits and phonological memory deficits in
Chinese developmental dyslexia. Learning & Individual Differences, 11, 173-186.
Ho, C. S.-H., & Law, T. P.-S., & Ng, P. M. (2000). The phonological deficit hypothesis in Chinese
developmental dyslexia. Reading & Writing, 13, 57-79.
Hubel, D. H., & Wiesel, T. N. (1967). Cortical and callosal connections concerned with the vertical
meridian of visual fields in the cat. Journal of Neurophysiology, 30, 1561-1573.
Jenner, A. R., Rosen, G. D., & Galaburda, A. M. (1999). Neuronal asymmetries in primary visual cortex
of dyslexic and nondyslexic brains. Annals of Neurology, 46, 189-196.
Jimenez Gonzales, J. E., & Rodrigo LOpez, M. (1994). Is it true that the difference in reading
performance between students with and without LD cannot be explained by IO? Journal of Learning
Disabilities, 27, 3, 155-163.
Jolicoeur, P., Corballis, M. C., & Lawson, R. (1998). The influence of perceived rotary motion on the
recognition of rotated objects. Psychonomic Bulletin & Review, 5, 140-146.
Johnston, J. c., & McClelland, 1. L. (1973). Visual factors in word perception. Perception and
Psychophysics, 14, 365-370.
194 T.LACHMANN
Just, M. A, & Carpenter, P. A (1980). A theory of reading: From eye fixations to comprehension.
Psychological Review, 87, 329-354.
Katz, L., & Wicklund, D. A (1971). Word scanning rate for good and poor readers. Journal of
Educational Psychology, 62, 138-140.
Kibei, M. & Miles, T. R. (1994). Phonological errors in the spelling of taught dyslexic children. In C.
Hulme, & M. Snowling (Ed.), Reading development and dyslexia (pp. 105-127). London, England:
Whurr Publishers, Ltd.
KIatzky, R. L., & Stoy, A M. (1974). Using visual codes for comparisons of pictures. Memory &
Cognition, 2, 727-736.
KIix, F. (1985).Ueber Sprache und Denken und ueber die Entstehungsgeschichte von Bewustseinsformen
[On language and thinking and on the history of the origin of types of consciousness]. Zeitschrift {iir
Psychologie, 193, 397-414.
Kolb, B., & Whishaw, I. Q. (1996). Fundamentals of human neuropsychology. New York: Freeman.
Kolers, P. A (1970). Three stages of reading. In H. Levin, & J. Williams. (Eds.), Basic studies on
reading. New York: Harper & Row.
Korhonen, T. (1991). Neuropsychological stability and prognosis of subgroups of children with learning
disabilities. Journal of Learning Disabilities, 24, 48-57.
Krueger, L. E. (1975). The word-superiority effect: Is its locus visual-spatial or verbal? Bulletin of the
Psychonomic Society, 6, 465-468.
Kuehn, S. M., & Jolicoeur, P. (1994). Impact of quality of the image, orientation, and similarity of the
stimuli on visual search for faces. Perception, 23, 95-122.
Kussmaul, A (1878). Word-deafness and word-blindness. In H. v. Ziemssen (Ed.), Cyclopedia of the
practice of medicine. London: Sampson Row, Maston, Searle & Rivingston.
Lachman, R., Lachman, J. L., & Butterfield, C. (1979). Cognitive psychology and information
processing: An introduction. Hilldale: Erlbaum.
Lachmann, T. (1996). Informationsverarbeitung beim visuellen Vergleich. Beitraege zur wis-
senschaftlichen Diskussion, 15, 5-113.
Lachmann, T. (2000). Representation-guided memory search in visual matching tasks. Psychologia
Universalis, 15. Lengerich, Berlin: Pabst Science Publishers.
Lachmann, T. (2001). Strategies of coding and processing in a physical same-different task. In E.
Sommerfeld, R. Kompass, & T. Lachmann (Eds.), Fechner Day 2001. Proceedings of the
International Society for Psychophysics (pp. 308-313). Lengerich, Berlin: Pabst.
Lachmann, T., & Geissler, H.-G. (1999). Frequency effects in a categorical same-different task point to
collective pattern codes. In P. R. Killeen, & W. R. Uttal (Eds.), 15th Annual Meeting of the
International Society for Psychophysics (pp. 138-143). Phoenix, AZ: The International Society for
Psychophysics.
Lachmann, T., & Geissler, H. G. (in press). Memory search instead of template matching? Memory-
guided inference in same-different performance. Acta Psychologica.
Lachmann, T., & van Leeuwen, C. C. (in press). Memory-guided-inference in "same-different"-
performance. In C. Kaernbach, E. Schroeger, & H. Mueller (Eds.), Psychophysics beyond sensation:
Laws and invariants of human cognition. Hillsdale, NJ.: Erlbaum.
Lehmkuhle, Stephen. (1993). Neurological basis of visual processes in reading. In D. M. Willows, R. S.
Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp. 77-94). Hillsdale,
NJ: Lawrence Erlbaum Associates.
Liberman, I. Y., Shankweiler, D., Orlando, c., Harris, K. S., & Bell Berti, F. (1971). Letter Confusions
and Reversals of Sequence in the Beginning Reader: Implications for Orton's Theory of
Developmental Dyslexia. Cortex, 7, 127-142.
Lockhead, G. R., & Pomerantz, J. R. (Eds.). (1994). The perception of structure. Washington DC:
American Psychological Association.
Lovegrove, W., Martin, F., & Slaghuis, W. (1986). A theoretical and experimental case for a visual
deficit in specific reading disability. Cognitive Neuropsychology, 2,225-267.
Lovett, M. W., Steinbach, K. A, & Frijters, J. C. (2000). Remediating the core deficits of developmental
reading disability: A double-deficit perspective. Journal of Learning Disabilities, 33, 334-358.
Lyon, R., Stewart, N., & Freedman, D. (1982). Neuropsychological characteristics of empirically derived
subgroups of learning disabled readers. Journal of Clinical Neuropsychology, 4, 343-365.
Marie, P. (1922). No title. Presse med. 30,177.
Marr, D. (1976). Early processing of visual information. Philosophical Transactions of the Royal Society
of London, 275, 483-524.
Masame, K. (1984). Detection of symmetry in relatively simple patterns. Tohoku Psychologica Folia, 43,
42-48.
Massaro, D. W. (1998). Models for reading letters and words. In D. Scarborough, & S. Sternberg (Eds.),
Methods, models, and conceptual issues. An invitation to Cognitive Psychology (pp. 301-364).
Cambridge: The MIT Press.
Massaro, D. W., & Cohen, M. M. (1991). Integration versus interactive activation: The joint influence of
stimulus and context in perception. Cognitive Psychology, 23, 558-614.
Mattis, S., French, 1. H., & Rapin, I. (1975). Dyslexia in children and young adults: Three independent
neuropsychological syndromes. Developmental Medicine and Child Development, 17, 150-163.
McClelland, J. L., & Rumelhart, D. E. (Eds.). (1986). Parallel distributed processing: Exploration in the
microstructure of cognition. Cambridge, MA: MIT Press.
McConkie, G. W., & Rayner, K. (1976). An on-line computer technique for studying reading:
Identification of the perceptual span. In H. Singer, & R. B. Ruddell (Eds.), Theoretical models and
processes of reading. Newark: International Reading Association.
McConkie, G. W., & Zola D. (1987). Visual attention during eye fixations while reading. In M. Coltheart
(Ed.), Attention and Performance Xli. The psychology of reading (pp. 385-401). East Sussex:
Erlbaum.
McCready, E. B. (1926). Defects in the zone of language and their influence in education and behavior.
American Journal of Psychiatry, 6, 267-277.
Merzenich, M. M., Jenkins, W. M., Johnston, P., Schreiner, c., Miller, S. L., & Tallal, P. (1996).
Temporal processing deficits of language-learning impaired children ameliorated by training.
Science, 271, 77-81.
Miles, E. (1991). Visual dyslexial auditory dyslexia: Is this a valuable distinction to make? In M.
Snowling, & M. Thompson (Eds.). Dyslexia: Integrating theory and practice (pp. 195-203). London:
Whurr.
Miles, E. (1995). Can there be a single definition of dyslexia? Dyslexia, 1, 37-45.
Miles, T. R. (1991). Explaining away dyslexia: A reply to Presland. Educational Psychology in Practice,
7, 163-167.
Miles, T. R. (1993). Dyslexia: The pattern of difficulties. London: Whurr.
Miles, T. R., Haslum, M. N., & Wheeler, T. 1. (1999). Gender ratio in dyslexia. Annals of Dyslexia, 48,
27-55.
Miles, T. R., & Miles, E. (1999). Dyslexia: A hundred years on. Buckingham: Open University Press.
Mills, C. K. (1929). Cerebral localization and failures in written languages. Archives of Neurology and
Psychiatry, 22, 1127-1134.
Mitterer, 1. O. (1982). There are at least two kinds of poor readers: Whole-word poor readers and
recoding poor readers. Canadian Journal of Psychology, 36, 445-461.
Monroe, M. (1928). Methods for diagnosis and treatment of cases of reading disability. Genetic
Psychological Monographes, 4, 335-456.
Monroe, M. (1932). Children who cannot read. Chicago: Chicago University Press
Morgan, W. P. (1896). A case of congenital word blindness. The British Medical Journal, 2, 1378.
Noonan, M., & Axelrod, S. (1992). Partial callosotomy and left-right response differentiation in the rat:
Separate anterior and posterior facilitatory effects. Behavioral Neuroscience, 106, 433-436.
Orton, S. T. (1925). "Word-blindness" in school children. Archives of Neurology and Psychiatry, 14, 5,
581-615.
Orton, S. T. (1927). Studies in stuttering. Introduction. Archives ofNeurology & Psychiatry, 18, 671-672.
Orton, S. T. (1928a). Specific reading disability - strephosymbolia. Journal of the American Medical
Association, 90, 1095-1099.
Orton, S. T. (1928b). A physiological theory of reading disability and stuttering in children. New England
Journal of Medicine, 29, 1046-1052.
Orton, S. T. (1929a). The neurological basis of education. Archives of Neurology & Psychiatry, 21, 641-
647.
Orton, S. T. (1929b). The three levels of cortical elaboration to certain psychiatric symptoms. American
Journal of Psychiatry, 8, 647-659.
Orton, S. T. (1929c). The "sight reading" method of teaching reading as a source of reading disability.
Journal of Educational Psychology, 2, 135-142.
Orton, S. T. (1937). Reading, writing, and speech problems in children and selected papers. Austin: Pro-
Ed.
Palmer, S. E., & Hemenway, K. (1978). Orientation and symmetry: Effects of multiple, rotational, and
near symmetries. Journal of Experimental Psychology: Human Perception & Performance, 4, 691-
702.
Parson, B. S. (1924). Lefthandedness. New York: Macmillan.
196 T.LACHMANN
Pashler, H. (1990). Coordinate frame for symmetry detection an object recognition. Journal of
Experimental Psychology: Human Perception and Performance, 16, 1, 150-163.
Patton, 1. E., Yarbrough, D. Y., & Thursby, D. (2000). Another look at children's symbol reversals.
Per,ceptual and Motor Skills, 90, 577-578.
Pulvermiiller, F. (19%). Neurobiology of language. Psychologia Universalis 1. Lengerich, Berlin: Pabst.
Presland, J. (1991). Explaining away dyslexia. Educational Psychology in Practice, 6, 215-221.
Radach, R., Heller, D., & Inhoff, A. W. (1999). Occurrence and function of very short fixation durations
in reading. In W. Becker, H. Deubel, & T. Mergner (Eds.), Current Oculomotor Research:
Physiological and Psychological Aspects (pp. 321-331). New York: Plenum Publishers.
Rayner, K. (1993). Visual processing in reading: Directions for research and theory. In D. Willows, R. S.
Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp. 475-480).
Hillsdale, NJ.: Erlbaum.
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research.
Psychological Bulletin, 124, 372-422.
Rayner, K., Inhoff, A. W., Morrison, R. E., Slowiaczek, M, & Bertera, 1. (1981). Masking of foveal and
parafoveal vision during eye fixations in reading. Journal of Experimental Psychology: Human
Perception and Performance, 7, 167-179.
Rayner, K., & Pollatsek, A. (1989). The psychology of reading. Englewood Cliffs, NJ.: Prentice-Hall.
Rayner, K., Slowiaczek, M. L., Clifton, C., & Bertera, J. H. (1983). Latency of sequential eye
movements: Implications for reading. Journal of Experimental Psychology: Human Perception &
Rayner, K., Well, A. D., Pollatsek, A., & Bertera, 1. H. (1982). The availability of useful information to
the right of fixation in reading. Perception & Psychophysics, 31, 537-550.
Reicher, G. M. (1969). Perceptual recognition as a function of meaningfulness of stimulus material.
Journal of Experimental Psychology, 81, 275-281.
Richardson, T. (1989). The century of the child: The Mental Hygiene Movement and social policy in the
United States and Canada. Albany, N. Y.: Suny
Robinson, H. M. (1946). Why pupils fail in reading. Chicago: University Press.
Rock, I., & Leaman, R. (1997). Symmetry. In I. Rock (Ed.), Indirect perception. MIT series in cognitive
psychology (pp. 151-162). Cambridge, MA: MIT Press.
Rutherford, W. 1. (1909). The etiology of congenital word-blindness, with an example. British Journal of
Children's diseases, 6, 484-488.
Ruthruff, E., Miller, 1. 0., & Lachmann, T. (1995). Does mental rotation require central mechanisms.
Journal of Experimental Psychology: Human Perception and Performance, 21, 552-570.
Rutter, M., & Yule, W. (1975). The concepts of specific reading retardation. Journal of Child Psychology
and Psychiatry, 16, 181-197.
Saarinen, J., & Levi, D. M. (2000). Perception of mirror symmetry reveals long-range interactions
between orientation-selective cortical filters. Neuroreport, 11, 2133-2138.
Satz, P, & Morris, R. (1981). Learning disability SUbtypes: A review. In F. 1. Pirozzolo, & M. C. Wittrock
(Eds.), Neuropsychological and cognitive processes in reading (pp. 109-141). New York: Academic
Press.
Schonell, F. J. (1940). The relation of reading disability to handedness and certain ocular factors. British
Schonell, F. 1. (1941). The relation of reading disability to handedness and certain ocular factors. British
Schonell, F. 1. (1942). Backwardness in the basic subjects. Edinburgh: Oliver & Boyd.
Sekuler, A. B., & Swimmer, M. B. (2000). Interactions between symmetry and elongation in determining
reference frames for object perception. Canadian Journal of Experimental Psychology, 54, 42-56.
Seymour, P. H. K., & McGregor, C. J. (1984). Developmental dyslexia: A cognitive experimental
analysis of phonological, morphemic and visual impairments. Cognitive Neuropsychology, 1, 43-82.
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.
Sharington, C. S. (1906). The interactive action of the nervous system. New Haven: Yale University
Press.
Shepard, R. N., & Metzler, 1. (1971). Mental rotation of three-dimensional objects. Science, 171, 701-
703.
Siegel, L. S. (1989). 10 is irrelevant to the definition of learning disabilities. Journal of Learning
Disabilities, 22, 469-478.
FuNCfIONAL COORDINATION DEFICIT IN READING DISABILITY 197
Siegel, L. S. (1992). An evaluation of the discrepancy definition of dyslexia. Journal of Learning

Siegel, L. S. (1998). Phonological processing deficits and reading disabilities. In J. L. Metsala, & L. C.
Ehri (Eds.), Word recognition in beginning literacy (pp. 141-160). Mahwah, NJ: Erlbaum.
Siegel, L. S., & Ryan, E. B. (1989). Subtypes of developmental dyslexia: The influence of definitional
variables. Reading and Writing, 2, 257-287.
Snowling, M. (1998). Dyslexia as a phonological deficit: Evidence and implications. Child Psychology &
Psychiatry Review, 3, 4-11.
Sperry, R. W. (1962). Some general aspects of interhemispheric integration. In V. B. Mountcastle (Ed.),
Interhemispheric relations and cerebral dominance. Baltimore: Johns Hopkins Press.
Springer, S. P., & Deutsch, G. (1998). Left brain, right brain. New York: Freeman.
Stanley, G., & Hall, R. (1973). Short-term information processing in dyslexics. Child Development, 44,
841-844.
Stanovich, K. E. (1991). Discrepancy definitions of reading disability: Has intelligence led us astray?
Reading Research Quarterly, 26, 7-29.
Stanovich, K. E. (1994). Annotation: Does dyslexia exist? Journal of Child Psychology & Psychiatry, 35,
579-595.
Stern, W. (1912). Die psychologischen Methoden der IntelligenzpriiJung. Leipzig: Barth.
Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental
Sutherland, N. S. (1957). Visual discrimination of orientations and shape by Octopus. Nature, 179, 11-13.
Tallal, P. (1999). Children with language impairment can be accurately identified using temporal
processing measures: A response to Zhang and Tomblin, Brain and Language, 65, 395-403 (1998).
Brain & Language, 69, 222-229.
Thorndike, E. L. (1919). Handwriting. Teachers College Record, 11,83-175.
Taylor, J. (Ed.). (1931). Selected writings ofJohn Hughlings-Jackson. London: Hodder.
Tonnessen, F. E. (1995). On defining "dyslexia." Scandinavian Journal of Educational Research, 39,
139-156.
Toth, G., & Siegel, L. S. (1994). A critical evaluation of the IO-achievement discrepancy based definition
of dyslexia. In K. P. van den Bos, & L. S. Siegel (Eds.), Current directions in dyslexia research. (pp.
45-70). Lisse, NL: Swets & Zeitlinger.
Uhry, J. K. (1993). Predicting low reading from phonological awareness and classroom print: An early
reading screening. Educational Assessment, 1, 349-368.
van den 80s, K. P., Siegel, L. S., Bakker, D. J., & Share, D. L. (Eds.). (1994). Current directions in
dyslexia research. Lisse, Netherlands: Swets & Zeitlinger.
Vellutino, F. R. (1977). Alternative conceptualizations of dyslexia: Evidence in support of a verbal-deficit
hypotheses. Harvard Educational Review, 47, 334-354.
Vellutino, F. R. (1979). Dyslexia: Theory and research. Cambridge, MA: MIT Press.
Vellutino, F. R. (1987). Dyslexia. Scientific American, 256, 20-27.
Vellutino, F. R. (1991). Introduction to three studies on reading acquisition: Convergent findings on
theoretical foundations of code-oriented versus whole-language approaches to reading instruction.
Vellutino, F. R., & Scanlon, D. M. (1998). Research in the study of reading disability: Whot have we
learned in the past four decades? Paper presented at the annual conference of the American
Educational Research Association, San Diego.
Vellutino, F. R., Scanlon, D. M., & Lyon, G. R. (2000). Differentiating between difficult-to-remediate
and readily remediated poor readers: More evidence against the IO-achievement discrepancy
definition of reading disability. Journal of Learning Disabilities, 33, 223-238.
Vellutino, F. R., Steger, J. A., & Kandel, G. (1972). Reading disability: An investigation of the perceptual
deficit hypothesis. Cortex, 8, 106-118.
Venezky, R. L. (1993). History of interest in the visual component of reading. In D. Willows, R. S. Kruk,
& E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp. 3-30). Hillsdale, N1.:
Erlbaum.
Vecera, P., & Farah, M. 1. (1997). Is visual image segmentation a bottom-up or an interactive process?
Perception & Psychophysics, 59, 1280-1296.
Wallin, J. E. W. (1920). Congenital word-blindness: Some analyses cases. Training School Bulletin, 17,
76-84.
Wenderoth, P. (1995). The role of pattern outline in bilateral symmetry detection with briefly flashed dot
patterns. Spatial Vision, 9, 57-77.
198 T.LACHMANN
Watson, B. U., Goldgar, D. E., & Ryschon, K. L (1983). Subtypes of reading disability. Journal of
Clinical Neuropsychology, 5, 377-399.
Watson, c., & Willows, D. M. (1993). Evidence for a visual-processing-deficit sUbtype among disabled
readers. In D. Willows, R. S. Kruk, & E. Corcos (Eds.), Visual processes in reading and reading
disabilities (pp. 287-309). Hillsdale, NJ.: Erlbaum.
Wernicke, C. (1874). Der aphasische Symptomkomplex. Breslau: Cohn & Weigert.
Wheeler, D. D. (1970). Processes in word recognition. Cognitive Psychology, 1, 59-85.
Willows, D. M. (1998). Visual processing in learning disabilities. In B. Wong (Ed.), Learning about
learning disabilities (pp. 203-236). San Diego: Academic Press.
Willows, D. M., & Terepocki, M. (1993). The relation of reversal errors to reading disabilities. In D.
Willows, R. S. Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp.
31-56). Hillsdale, NJ.: Erlbaum.
Witty, P., & Kopel, D. (1936). Factors associated with the etiology of reading disability. Journal of
educational research, 29, 449-459.
Wolf, I. J. (1935). Strephosymbolia vs. congenital word-blindness. Proceedings of the American
Association on Mental Deficiency, 40, 446-462.
Warburg, F. (1911). Uber die angeborene Wortblindheit. Z. f. Kinderforschung, 4, 21-53.
Wimmer, H., Mayringer, H., & Landerl, K. (1999). Poor reading: A deficit in skill-automatization or a
phonological deficit? Scientific Studies of Reading, 2, 321-340.
Wundt, W. (1912). Grundriss der Psychologie. Leipzig: Alfred Korner Verlag.
Zaidel, E. (1998). Stereognosis in the chronic split brain: Hemispheric differences, ipsilateral control and
sensory integration across the midline. Neuropsychologia, 36, 1033-1047.
Zangwill, O. L., & Wyke, M. A. (1990). Hughlings-Jackson on the recognition of places, persons, and
objects. In C. B. Trevarthen, & B. Colwyn (Eds.), Brain circuits and functions of the mind: Essays in
honor of Roger W. Sperry (pp. 281-292). New York: Cambridge University Press.
1. STEIN
THE NEUROBIOLOGY OF READING DIFFICULTIES
Abstract. There is very little agreement about why so many otherwise intelligent children experience
such difficulty learning to read. Up to 10% of children fail to learn to read at the level expected of their
age and other cognitive abilities. Some people still believe that their problems are purely linguistic, due to
inheriting or acquiring a defective linguistic processor. Others take the view that difficulty with learning
to read derives from more basic deficiencies in processing the sensory inputs that are required for reading.
According to this view literacy problems are caused by fundamental differences in sensory processing
between the brains of good and bad readers, so that in addition to their literacy problems there are many
other sensory, motor and cognitive differences to be found in poor readers because the development of the
whole brain is disordered. In our laboratory we take the latter position. We study the neurophysiological
processes which underlie reading; hence we view reading problems as having a neurodevelopmental
basis. In this chapter therefore, I will discuss how normal reading depends on the quality of its sensory
input; it absolutely requires both a highly sensitive visual magnocellular system to acquire good
orthographic skills and a sensitive auditory transient system to develop the ability to parse the
phonological structure of words. Then I will speculate about the genetic and immunological mechanisms
that may be responsible for the wide variety of abnormalities that are seen in developmental dyslexics.
My overall conclusion wiJI be that reading difficulties are neither specific to reading nor exclusively
linguistically based, but a consequence of mildly impaired development of a particular kind of neurons in
the brain, magnocellular neurons, so that dyslexia has widespread manifestations, which are not at all
confined to reading (Stein & Walsh, 1997; Stein & Talcott, 1999). They are best thought of as individual
differences between people rather than a consequence of neurological "disease".
1. NORMAL READING
Recent functional imaging studies have shown that reading involves both
hemispheres of the brain; but the importance of the left becomes clearer when the
phonological demands of the task are made harder (Demonet, Wise, & Frackowiak,
1993). Silent reading engages mainly the posterior part of the left hemisphere
focusing on the left angular and supramarginal gyri. However reading out loud shifts
activity forwards towards Broca's area in the frontal cortex, so that the whole of the
temporo-frontal articulatory loop is engaged for this task. Homologous areas in the
right hemisphere are also activated to a lesser extent; but their role is enhanced only
when the intonation or emotional content of speech is emphasized.
Activation of posterior more than anterior areas during silent reading
emphasizes the importance of the visual input to reading. Vision is obviously
important, and Morgan (1896) first described developmental dyslexia as "word
blindness". However its role in learning to read is very neglected nowadays.
Following the linguistic tradition most people believe that poor readers' main
problem is lack of phonological skill rather than anything to do with vision
(Liberman, Shankweiler, Fischer, & Carter, 1974; Snowling, 1981). But in fact
familiar words are recognized entirely visually without any requirement for
phonological mediation. Likewise irregular words cannot be successfully sounded
out; hence they must be read by the visual route.

and Reading Disability, 199-21l.
200 1. STEIN
It is true on the other hand, that unfamiliar words (and all words are unfamiliar to
beginning readers) have to be sounded out using the letter sound correspondences
that have to be learnt. This engages more anterior parts of the articulatory loop. Even
if the letters are sounded out entirely mentally, by means of "inner speech", the
whole articulatory loop is engaged. Thus there are two reading mechanisms that are
at least conceptually separable: the whole word semantic route which draws heavily
on the visual system, and the phonological route which draws heavily on the
auditory articulatory system (Castles & Coltheart, 1993; Ellis, 1992). These two
routes therefore depend upon the processing powers of the visual and auditory
systems; and it is these that I shall now deal with. Further details of many of the
experiments that I will describe can be found in the Chapter by Talcott and Witton
in this book.
2. THE VISUAL MAGNOCELLULAR SYSTEM
The ganglion cells whose axons carry visual signals from the retina to the rest of the
brain can be divided into two main types: 90% are known as parvo cells from their
small cell bodies and restricted dendritic spread (Enroth Kugel & Robson, 1966;
Shapley & Perry, 1986). These are responsible for color and fine detail. The
remaining 10% are large magno cells. Their dendrites cover a retinal area up to 500
times that of parvo cells. They are not only larger, but they are more heavily
myelinated which means that their conduction velocities and membrane dynamics
are much faster. Hence although their spatial resolution is coarser and they do not
support color vision, they respond more quickly and their signals arrive at the visual
cortex 10-20 msecs earlier than those of parvo cells. Thus they are important for
timing events in the visual world and for detecting changes with time, such as those
caused by visual motion; but they do not signal color or fine detail (Merrigan &
Maunsell, 1993).
On arrival at the main visual relay nucleus in the thalamus, the lateral geniculate
nucleus (LGN), magno axons are completely separate from parvo and pass into the
magnocellular layers of the LGN. The magnocellular LGN cells then project to layer
IV C alpha of the primary visual cortex, whereas the parvocellular layers project to
layer IV C beta. From there parvo and magno streams intermingle, but the output of
the visual system splits into two main streams, the "what" ventrolateral and the
"where" dorsomedial pathways (Ungerleider & Mishkin, 1982). The ventrolateral
stream projects towards the inferotemporal cortex and receives roughly equal inputs
from magno and parvo pathways. It is called the "what" pathway because it is
specialized for identifying the shape, pattern and color of objects, and therefore for
identifying what they are.
In contrast the dorsomedial pathway passes towards the visual motion area
(V5/MT) in the Superior Temporal sulcus, and thence to the posterior parietal
cortex. It is termed the "where" pathway because it is specialized for detecting the
current position and motion of targets, and for directing attention and movements
towards them (Milner & Goodale, 1995). It is therefore dominated by input from the
visual magnocellular system. As befits its function in helping the visual guidance of
movements this system projects onwards to all the areas involved in the guidance of
eye movements: the frontal eye fields, the superior colliculus and the cerebellum.
3. DYSLEXICS' VISUAL MAGNO CELLULAR DEFICITS

One advantage of the distinction between the visual magno- and parvocellular
systems is that the sensitivity of each can be separately assessed psychophysically in
normal subjects using stimuli that selectively activate one or the other. Spatial
contrast and temporal flicker sensitivity are mainly limited by the performance of
the peripheral visual system up to, but not including, the level of the visual cortex.
Lovegrove, Martin, Blackwood, and Badcock (1980) therefore used sinusoidal
gratings to show that the contrast sensitivity of dyslexics was impaired compared
with controls, particularly at low spatial and high temporal frequencies. Thus it was
he who first suggested that dyslexics may have a selective impairment of the
magnocellular system. Using stationary gratings he found that dyslexics' contrast
sensitivity at the high spatial frequencies that are mediated by the parvocellular
system was actually higher than in controls and we were able to confirm his results
(Mason, Cornelissen, Fowler, & Stein, 1993 ). Likewise Martin and Lovegrove
(1987) showed that dyslexics' flicker sensitivity tends to be lower than controls, and
we have confirmed this also (Mason et aI., 1993; Talcott et aI., 1998). Again this
result suggests that dyslexics may have a specific magnocellular impairment.
However these results have been hotly disputed (Skottun, 2000; see also Stein,
Richardson, & Fowler, 2000; Stein, Talcott, & Walsh, 2000). The impairment is
slight and is only found in some 2/3rds of dyslexics. Most studies have involved
small numbers so that there have indeed been many failures to replicate Lovegrove's
results. But larger numbers should be studied and the dyslexics should be selected
for those who have visual symptoms and therefore are most likely to have a
significant magnocellular deficit.
Testing sensitivity to visual motion has proved much more reliable however,
because motion engages not only the peripheral visual system but also central
processing stages in visual cortical areas up to at least area V5/MT. In monkeys it
has been found that detecting coherent motion in a display of dots moving about
randomly (random dot kinematograms - RDK) is a sensitive test for probing the
magnocellular system (Newsome, Britten, & Movshon, 1989). We have therefore
developed a RDK test of motion sensitivity for use with adults and children. We
present 2 panels of randomly moving dots side by side. In one of the panels, selected
at random, a proportion of the dots is moved together, "coherently" so that they look
like a cloud of snowflakes blown in the wind. The subject is asked in which panel a
cloud of dots appears to be moving all together. The proportion of dots that is moved
together is then reduced until the subject can no longer tell on which side the dots
are moving together. His threshold is then defined as the proportion of dots that have
to move together for him to see the coherent motion correctly on 75% of occasions.
Using this test we have found that in children and adults with dyslexia, defined as
those whose reading is more than 2 Standard Deviations behind that expected from
their age and IQ, about 75% have significantly worse motion sensitivity than
controls matched for age and IQ (Cornelissen, Richardson, Mason, Fowler, & Stein,
1994; Talcott et aI., 1998; Talcott, Hansen, Elikem, & Stein, 2000; Talcott, Witton et
aI., 2(00). Our conclusion from psychophysical studies that many dyslexics have
poor motion sensitivity has also now been confirmed by a succession of functional
imaging studies (Demb, Boynton, Best, & Heeger, 1998; Demb, Boynton, &
Heeger, 1997; Eden, VanMeter, Rumsey, Maisog, & Zeffiro, 1996).
202 J. STEIN
4. LGN ABNORMALITY IN DYSLEXIA

Lest anyone is still in doubt however, that many dyslexics have impaired
development of the visual magnocellular system, the best evidence would come
from directly examining the magnocellular layers of the relay nucleus of the
peripheral visual system, the lateral geniculate nucleus (LGN) in the thalamus.
Samuel Orton, a pioneer in the study of developmental dyslexia in the 1930s,
persuaded some of his patients to donate their brains to medical science. These
donations have now begun to materialize and Albert Galaburda and colleagues have
now studied some of these histologically. They have shown unequivocally that the
magnocellular layers of the LGN are indeed disordered and the neurons some 30%
smaller in area than in control brains (Livingstone, Rosen, Drislane, & Galaburda,
1991; Galaburda & Livingstone, 1993). Such differences are known to arise during
the early development of the brain, during the phase of rapid neuronal growth and
migration during the fourth or fifth month of foetal development. One could not
adduce stronger evidence than this that the visual magnocellular system does not
develop quite normally in dyslexics.
5. MAGNOCELLUlAR FUNCTION PREDICTS ORTHOGRAPHIC SKILL

Since our hypothesis is that impaired visual magnocellular function impedes reading
it is incumbent on us to show this. Therefore we have been correlating
magnocellular sensitivity with visual reading ability. In both children and adults,
whether dyslexic, good readers or unselected primary school children we have
demonstrated that visual motion sensitivity correlates strongly with their reading
ability (Witton et aI., 1998; Talcott, Hansen, et aI., 2000). In particular motion
sensitivity predicts visual orthographic skill. We measure this using the Olson
pseudo-homophone test (Olson, Wise, Connors, Rack, & Fulker, 1989). In this test
we present on a computer screen 2 words side by side that sound the same but have
different spellings, i.e. "rain" beside "rane". The subject is asked which is the correct
spelling. Since the words sound the same, this task cannot be solved phonologically
by sounding out the letters. The visual form or orthography of the word must be
recalled correctly. Hence Joel Talcott found that the correlation between visual
motion sensitivity and performance in this pseudo-homophone test is very strong
(Talcott, Hansen, et aI., 2000). This was true not only in dyslexics but across the
whole range of reading abilities. Good spellers in this test had high motion
sensitivity, whereas poor performers had low motion sensitivity. Likewise we found
that the correlation between subjects' motion sensitivity and their performance in
Castles and Coltheart's (1993) test of the spelling of irregular words was very high,
because success in this test also depends on a good visual memory of their spelling.
In contrast the correlation between subjects' visual motion sensitivity and tests of
phonological skill, such as the ability to read nonsense words or to form
Spoonerisms is much lower. In fact when we controlled statistically for the
correlation that exists between subjects' phonological and orthographic abilities we
found that motion sensitivity continued to account for a high proportion of the
variance in orthography, but now of course independently of phonology. In other
words motion sensitivity accounts for children's orthographic skill independently of
its relationship with their phonological skill, as you would expect if this basic visual
function helps to determine how well the visual component of reading develops.
6. DYSLEXICS' BINOCULAR CONTROL
However one problem that constantly bedevils the hypothesis that dyslexics have
impaired magnocellular function, is that people find it very difficult to understand
how a system that is devoted to detecting visual motion could possibly be relevant to
reading (Hulme, 1988). After all we don't usually have to track moving targets when
reading; the page is kept stationary. In fact however, the retinal images of print are
not stationary, and many dyslexic children complain that letters seem to move
around when they are trying to read, i.e. their visual world is highly unstable
(Fowler & Stein, 1979). This is because visual images are actually very far from
stable on the retina during reading and dyslexics fail to compensate for this. The
eyes remain fixated on individual words for only about 300 msecs before saccading
to the next. Even during the fixations however, they are not totally stable but move
around by up to 1 degree of visual angle - equivalent to 4 or 5 letters' worth. In
normal readers the visual magnocellular system detects such unintended motion of
the letters over the retina, "retinal slip", and this signal is used to help stabilize the
eyes. For movements of less than 1 degree the magnocellular system further
sharpens the image by enabling us to discount the motion. But it seems that many
dyslexics fail to be able to stabilize their vision when reading, so that they tend to
transpose letters when attempting to read (Cornelissen, Hansen, Hutton,
Evangelinou, & Stein, 1997). We believe that their unstable visual perceptions are
the result of the insensitivity of their visual magnocellular systems.
7. DYSLEXICS' BINOCULAR INSTABILITY
To test this hypothesis we measured how steadily dyslexics and controls could fixate
small targets that were the size of letters, at the normal reading distance of about 30
ems. We found that the amplitude of the unintended eye movements made during
attempts to fixate for 3 seconds was indeed much greater in the dyslexics than in the
controls (Eden, Stein, Wood, & Wood, 1994). Furthermore these movements were
different in the two eyes so that the degree of binocular convergence varied
randomly from moment to moment. This would explain why so many dyslexic
children complain not only that the letters move from side to side but also slide over
each other and in and out of the plane of the page (Cornelissen et aI., 1994;
Cornelissen, Bradley, Fowler, & Stein, 1991). We found that the unstable binocular
control of these dyslexics was also reflected in slower and less smooth vergence
pursuit tracking of a target moving in depth, with a much greater tendency for
vergence to break down to conjugate gaze, hence diplopia (Riddell, Fowler, & Stein,
1988, 1990; Stein, Riddell, & Fowler, 1988). The high incidence of binocular
instability in dyslexics has also been confirmed by Evans, Drasdo, and Richards
(1994).
Thus it seems that an important way in which the impaired magnocellular
function found in dyslexics may interfere with reading may be because it leads to
binocular instability. Since these eye movements are unintended and uncontrolled
they may be misinterpreted as movements of the letters. The instability often causes
204 1. STEIN
the two eyes' lines of sight to cross over each other, so the letters can appear to
move around, slide over each other, and change places. We reasoned that if this is
so, then simply blanking the vision of one eye should alleviate the visual confusion
and help these children to see the letters properly. This is exactly what we have
found in four different trials (Cornelissen, Bradley, Fowler, & Stein, 1992; Stein &
Fowler, 1981, 1985; Stein, Richardson, et al., 2000; Stein, Talcott, et al., 2000). In
children with binocular instability, occluding the left eye for reading and close work
relieved their binocular perceptual confusion and helped them to learn to read. The
results were often dramatic and in our most recent double blind controlled trial of
monocular occlusion in dyslexic children with binocular instability we were able to
help those who received the occlusion to catch up with the reading age of their
peers. In contrast those who did not receive occlusion and who did not gain
binocular stability remained lagging 2 years behind their chronological age. This
progress is far greater than any remediation technique for use with dyslexics that we
have been able to find.
After three months occlusion not only had the children's reading improved to
this great extent, but also they could now fixate stab ally with their two eyes, so that
they no longer needed to wear the patch. We believe that this gain of binocular
stability results from the magnocellular signals from the seeing eye now successfully
routing themselves to control the muscles of that eye. This is called utrocular control
(Ogle, 1962) and it is crucial for the final stages of precise vergence control because
it enables each eye to home in accurately on a target so that both fixate properly on
it.
So now we think we can explain how magnocellular function impacts on
reading, and in particular how it helps to develop orthographic skill. Poor readers
have slightly impaired development of their magnocellular neurons. As a
consequence the dense magnocellular input that visuomotor centers in the PPC,
superior colliculus and cerebellum receive is both delayed and smeared in time. In
consequence utrocular control over the muscles controlling the eye that supplied the
magnocellular input is less sharply focussed in time and therefore less able to
stabilize the eyes during fixation or during vergence pursuit. Therefore the eyes'
lines of sight may cross over each other, hence the letters can appear to do so also.
This is why these dyslexics tend to reverse the order of neighboring letters and to
reverse the order of letter features, thus confusing ds with bs and ps with qs. These
are precisely the kinds of error that we find in dyslexics with unstable binocular
control.
8. AUDITORY TRANSIENT SYSTEM

Only about 2/3 of dyslexics have significant visual problems however. The
remaining 1/3 together with about half of those with visual symptoms in addition,
have another major cause of reading difficulties, namely inability to retrieve letter
sound correspondences accurately and fast. Hence their ability to grasp the
phonological structure of words is unreliable and slow. As described earlier
unfamiliar regular words can be read by producing the sounds that each letter stands
for, "sounding out" the letters, and then blending them together to identify the word.
Acquisition of this phonological skill is perhaps the most important and most
difficult component of learning to read (Bradley & Bryant, 1983; Snowling, 1987;
Liberman et aI., 1974).
9. SOUND FREQUENCY MODULATION
The cues that distinguish the different letter sounds are changes in the frequency and
amplitude of speech sounds. Thus the difference between /d/ and fb/ is that in /d /the
2nd and 3rd formants rise in frequency in the first 40 ms, but in b they go down.
Everything else is identical in the two sounds. Thus phonological analysis draws
very heavily on the ability of the auditory system to track frequency and amplitude
changes, acoustic transients, accurately. Such processing can be assessed as in the
visual system, using much simpler transients than those found in speech, for
example by measuring subjects' sensitivity to sinusoidal frequency or amplitude
modulations of a single sinusoidal tone. Therefore we have been measuring people's
sensitivity to frequency (FM) and amplitude modulations (AM) of simple tones to
see whether this relates to their phonological abilities.
Ken McAnally and Stein (1996; see also Stein & McAnally, 1996) and more
recently Caroline Witton et a1. (1998) measured subjects' sensitivity to frequency
modulations by asking subjects to listen to frequency modulated (warbling) tones
and then adjusting the amount of warble, the modulation depth, until they could no
longer distinguish this from a pure tone. Using 2 , 20 and 40 Hz frequency
modulations of a 500 Hz or 1000 Hz carrier stimulus we were able to show that both
adult and child dyslexics were significantly worse than matched controls at hearing
these changes in frequency, i.e. they required significantly greater modulation
depths. Again however there were some dyslexics who were just as good as controls
at this task.
10. PHONOLOGICAL SKILL IS PREDICTED BY FM SENSITIVITY

Since we expected this stimulus to mimic to some extent the phonological cues that
enable people to distinguish letter sounds, we needed to correlate their FM
sensitivity with their phonological ability. This can best be tested by getting subjects
to read nonsense words, nonwords, such as "tegwop". This word means nothing; yet
it can be read by applying the phonological rules properly. However you are not
assisted by the word having any meaning, so that nonsense word reading has very
little "top down" input. Performing this task depends mainly on rapid and accurate
application of letter/sound conversions. Hence it has been shown to be a highly
sensitive index of dyslexia.
We therefore measured the number of errors and the time our subjects' took to
read a list of nonwords and correlated this with their FM sensitivity. This showed
that the relationship was very strong in both children and adults, so that in a class of
30 normal 10 year olds 2 Hz FM sensitivity accounted for over 50% of their
variance in nonword reading (Talcott et aI., 1999, Talcott, Hansen et aI., 2000)
When we controlled for the variance shared with orthographic ability to allow for
the fact that the subjects had to visualize the non words in order to read them, FM
sensitivity still accounted for a substantial proportion of nonword reading variance
independently of orthography. Thus high auditory FM sensitivity seems to enable
206 1. STEIN
children to develop strong phonological skills, whereas poor FM sensitivity prevents

this, in the same way as high visual motion sensitivity promoted orthographic skill
and low sensitivity prevented it.
11. AMPLITUDE MODULATION (AM)
Peter Menell (Men ell, McAnally, & Stein, 1999) and Caroline Witton (submitted)
have also demonstrated that sensitivity to amplitude modulations is also important
for developing phonological skill. Caroline measured sensitivity to 20 Hz amplitude
modulations and found not only that dyslexics are significantly less sensitive to
these than controls, but that in both dyslexics and normal readers covering the whole
range of reading ability AM sensitivity also helps to account for variance in their
nonword performance.
What is particularly interesting was that the subjects performed differently in
response to the FM and AM stimuli; their scores in one did not correlate with the
other. The auditory system processes 2 Hz FM and 20 Hz AM stimuli rather
differently. Unlike 2 Hz FM, surprisingly 2 Hz AM thresholds did not correlate with
the subjects' reading ability at all. Thus slow amplitude modulations may not
provide cues to word boundaries, intonation and syllablification in the way that we
expected. But both 2 Hz FM and 20 Hz AM sensitivity did correlate with nonword
reading ability, and the latter accounted for a significant amount of variance that 2
Hz FM could not. In other words our FM and AM tests probably relate to the
development of different aspects of phonological skill. 2Hz FM is quite slow and
probably corresponds to the frequency of syllable production and intonation rather
than to individual phonemes, whereas 20 Hz is clearly in the range of the transients
that enable us to distinguish between different phonemes.
12. AUDITORY MAGNOCELLS?

Whereas the magnocells in the visual system form a clearly separate anatomical
system responsible for processing visual transients, there is no such anatomically
defined magnocellular division in the auditory system. We cannot so simply
attribute the FM and AM deficits that we have found in poor readers to impaired
development of an auditory magnocellular system. Nevertheless there are large
celled magnocellular divisions of all the auditory relay nuclei, and there is evidence
that these are specially responsible for following changes in frequency or amplitude
of acoustic signals with time (Trussel, 1998). Thus there probably is an auditory
magnocellular system whose function is to process auditory transients such as those
important for phonological analysis, but one must bear in mind that it is not an
anatomically separate system as is the case for the peripheral part of the visual
magnocellular system.
This idea that there is an auditory analogy with the visual magnocellular system
received further support from further neuropathological studies by Galaburda,
Menard, and Rosen (1994) in dyslexic brains post mortem. They found that, like in
the LGN, the magnocellular division of the medial geniculate nucleus (MGN) which
is the auditory relay nucleus of the thalamus, was disordered, particularly on the left,
i.e. the side that projects to the language hemisphere.
13. AETIOLOGY OF DYSLEXICS' LOW SENSITIVITY TO SENSORY

TRANSIENTS
In summary dyslexics seem to have impaired processing of both visual and auditory
transients, because they tend to have impaired development of magnocellular
neurones in both the visual and auditory systems. These impairments help to explain
their failure to develop adequate orthographic and phonological skills, respectively.
Of course the question arises why dyslexics fail to develop good sensitivity to
sensory transients. Those who argue that dyslexia is a specifically linguistic problem
suggest that the low level auditory and visual deficits are purely epiphenomena, not
on the main causal chain leading to reading problems (Studdert Kennedy & Mody,
1995). But this argument seems somewhat implausible in the light of the high
correlations we have found between these sensory deficits and the cognitive skills
required for reading. Also our finding that setting right the binocular instability
associated with the visual magnocellular deficit very greatly improves dyslexics'
reading strongly suggests a causal connection (Stein, Richardson, et aI, 2000; Stein,
Talcott, et aI., 2000), as does Merzenich et ai. 's (1996) finding that auditory training
using artificially slowed frequency transients can also improve children's ability to
hear them, and that this improves their ability to make good phonological
discriminations.
More likely is the possibility that there is a difference in the development of
magnocellular systems in dyslexics' brains that blunts their transient sensitivity,
hence compromises their ability to develop orthographic and phonological skills.
There is now a great deal of evidence for this view and a plausible mechanism for
the development of these differences can even be sketched out. The evidence is
genetic, immunological and neurological.
14. GENETIC LINKAGE

It is well known that dyslexia tends to run in families, and twin studies have shown
that it has a heritability of c. 60%, ie 60% of the variance in people's reading
abilities can be attributed to the particular alleles they inherit (Pennington, 1991).
This is a high value that is similar to the heritability of height and other cognitive
abilities such as verbal ability or general intelligence, but much higher than the
heritability of conditions that tend to be seen as much more familial, such as
diabetes or heart attacks.
We have taken advantage of the large number of dyslexics we have seen over the
years to collect, so far, over 200 families in which two or more children have severe
reading problems. We retest the children in standardized orthographic and
phonological tests and take blood for DNA analysis from the children and their
parents. In collaboration with Prof. Tony Monaco and Simon Fisher we then look
for linkage of the reading scores treated as quantitative traits (OTLs) to
chromosomal markers. We are currently in the process of carrying out a screen of
the whole human genome to find which sites link to reading ability; but we have
already confirmed (Fisher et aI., 1999) the findings of Cardon et al. (1994) and
Grigorenko et a1. (1997) from smaller studies carried out in Colorado and
Connecticut respectively, that both phonological and orthographic impairments link
to a site on the short arm of chromosome 6 near the Tumour Necrosis Factor and
208 J. STEIN
Major Histocompatibility Complex (MHC) immunological sites. This linkage may

turn out to be very significant as there is accumulating evidence that an
immunological mechanism contributes to the control of the development of
magnocells.
15. MAGNO SURFACE MARKERS
Visual magnocells express specific surface antigens, such as one recognized by the
Cat 301 antibody, and these same antigens are found on large cells all over the
nervous system (Hockfield & Sur, 1990). For example they are also expressed in the
magnocellular nuclei of the auditory system, on the large cells in the dorsal column
somaesthetic nuclei, the hippocampus and on other magno cells throughout the
brain. One site of special interest is the cerebellum. This structure is the brain's main
timing device, the brain's autopilot, and it receives dense input from all the magno
systems throughout the brain (Stein & Glickstein, 1992). Thus magnocells probably
represent a distinct cerebral system with a separate developmental lineage, common
surface antigens and the heavy myelination and rapid membrane dynamics that
confer upon them their enhanced sensitivity to temporal transients. But their
common surface antigens may make them all vulnerable to antibody attack in
autoimmune individuals.
16. GENERALISED MAGNO CELLULAR SYSTEM
Therefore we were not surprised to find that there is a strong tendency, especially in
normal readers, for auditory and visual transient sensitivity to be similar in
individuals (Witton et aI., 1998). This again suggests that the development of
auditory and visual magnocells is indeed under some sort of common control.
Recently Corriveau and Shatz (1998) have found that Class 1 MHC molecules play
an important part in the development of visual magnocellular neurons in the cat
LGN and also in the hippocampus. Since Hockfield and Sur's work (1990) suggests
that all magnocells throughout the brain derive from a common developmental
lineage, it is reasonable to speculate that the development of all of them is regulated
by the MHC system. Hence our finding that reading disability is genetically linked
to a site close to the MHC complex on chromosome 6 takes on added significance,
as does the evidence that dyslexics and their families very often demonstrate mildly
abnormal immunological responses. They have a higher incidence of asthma,
eczema, hayfever and other immune conditions than controls, and families with
serious autimmune conditions such as systemic lupus erythmatosus (SLE) often
have dyslexic relatives as well (Hugdahl, Synnevag, & Satz, 1990).
Thus as Miles (1983, 1993) first emphasized, dyslexics' literacy problems should
really be considered part of a much more generalized neurodevelopmental syndrome
caused by impaired development of magnocells throughout the brain. We speculate
that this results from genetically mediated, disordered immunological regulation of
their development in utero. This would explain not only the auditory and visual
transient deficits that directly undermine the development of phonological and
orthographic skills, but also the plethora of other problems that beset dyslexics.
Their legendary clumsiness, impaired coordination and balance can be attributed to
abnormal cerebellar function, for which there is now a great deal of evidence
(Fawcett, Nicolson, & Dean, 1996; Rae et aI., 1998; Nicolson et aI., 1999). Their
reduced sequencing and timing ability may be attributed partly to impaired
cerebellar function and partly perhaps to reduced magnocellular input to the left
hemisphere, which normally receives more than the right (Klingberg et aI., 2000).
This in turn would lead to impaired hemispheric specialization that would explain
dyslexics' failure to establish fixed hemispheric dominance, their tendency to have
problems with telling left from right, their mixed handedness, urifixed eye
dominance and many other cognitive symptoms. In short immunologically mediated
mild magnocellular deficiency could explain all the wide variety of problems that
dyslexics face.
17. AFFILIATIONS
Dr. John Stein FRCP

University Laboratory of Physiology,
Parks Rd. Oxford, OX1 3PT, UK
e-mail: john.stein@physiol.ox.ac.uk
18. REFERENCES
Bradley, L., & Bryant, P. (1983). Categorizing sounds and learning to read - a causal connection. Nature,
301,419-421.
Cardon, L. R., Smith, S. D., Fulker, D. W., Kimberling, W. 1., Pennington, 8. F., & Defries, 1. C. (1994).
Quantitative trait locus for reading disability on Chromosome 6. Science, 266, 276-279.
Castles, A., & Coltheart, M. (1993). Varieties of developmental dyslexia. Cognition, 47, 149-180.
Cornelissen, P., Bradley, L., Fowler, M. S., & Stein, 1. F. (1991). What children see affects how they
read. Develop. Med. & Child Neurology, 33, 755-762.
Cornelissen, P., Bradley, L., Fowler, M. S., & Stein, 1. F. (1992). Covering one eye affects how some
children read. Develop.Med. & Child Neurology, 34, 296-304.
Cornelissen, P. L., Hansen, P.c., Hutton, J. L., Evangelinou, V., & Stein, J. F. (1997). Magnocellular
visual function and children's single word reading. Vis Res., 38, 471-482.
Cornelissen, P. L., Richardson, A R., Mason, A, Fowler, M. S., & Stein, 1. F. (1994). Contrast sensitivity
and coherent motion detection measured at photopic luminance levels in dyslexics & controls. Vision
Res.,35,1O, 1483-1494.
Corriveau, R., & Shatz, C. (1998). Regulation of Class 1 MHC gene expression in the developing and
mature CNS by neural activity. Neuron, 21, 505-520.
Demb, 1. 8., Boynton, G. M., Best, M., & Heeger, D. J. (1998). Psychophysical evidence for a
magnocellular deficit in dyslexics. Vision Research, 38, 1555-1559.
Demb,1. 8., Boynton, G. M., & Heeger, D. J. (1997). Brain activation in visual cortex predicts individual
differences in reading performance. Proc. NY Ac. Sc., 94, 13363-6.
Demonet, 1. F. R., Wise, R., & Frackowiack, R. S. 1. (1993). Language function explored by positron
emission tomography. Hwnan Brain mapping, 1, 39-45.
Eden, G. F., Stein, 1. F., Wood, H. M., & Wood, F. B. (1994). Differences in eye movements and reading
problems in dyslexic and normal children. Vision Res., 34,10, 1345-1358.
Eden, G. F., VanMeter, 1. W., Rumsey, J. W., Maisog, J., & Zeffiro, T. A (1996). Functional MRI
reveals differences in visual motion processing in individuals with dyslexia. Nature, 382, 66-69.
Ellis, A W. (1992). Reading, writing and dyslexia. Erlbaum, London ..
Enroth Kugel, C., & Robson, J. G. (1966). The contrast sensitivity of retinal ganglion cells in the cat. J.
Physiol., 187, 517-552.
Evans, 8. J. W., Drasdo, N., & Richards, I. L. (1994). Investigation of accommodative and binocular
function in dyslexia. Ophthalmic & Physiological Optics, 145, 5-20.
Fawcett, A, Nicolson, R. I., & Dean, P. (1996). Impaired performance of children with dyslexia on a
range of cerebellar tasks. Ann. Dyslexia, 46, 259-283.
210 1. STEIN
Fisher, S., Marlowe, A., Lamb, J., Maestrinin, E., Williams, D., Richardson, A., Weeks, D., Stein, J. F., &
Monaco, A. (1999). A quantitative trait locus on chromosome 6p influences different aspects of
developmental dyslexia. Am. 1. Human Genetics, 64, 146-156.
Fowler, S., & Stein, J. F. (1979). New evidence for ambilaterality in visual dyslexia. Neurosci.Letts.,
Suppl.3, 214.
Galaburda, A. M., & Livingstone, M. (1993). Evidence for a magnocellular defect in developmental
dyslexia. In P. Tallal, A. M. Galaburda, R. R. Llinas, & C. von Euler (Eds.), Temporal Information
Processing in the Nervous System (pp. 70-82). Ann.NY Acad.Sci.
Galaburda, A. M., Menard, M. T., & Rosen, G. D. (1994). Evidence for aberrant auditory anatomy in
developmental dyslexia. Proc. Natl. Acad. Sci. USA., 91, 8010-8013.
Grigorenko et al (1997). Susceptibility loci for distinct components of developmental dyslexia on
chromosomes 6 and 15. American 1. Human Genetics, 60, 27-39.
Hockfield, S., & Sur, M. (1990). Monoclonal CAT 301 identifies Y cells in cat LGN.1. C. Neurol. 300,
320-30.
Hugdahl, K., Synnevag, 8., & Satz, P.(1990). Immune and autoimmune disorders in dyslexic children.
Hulme, C. (1988). The implausibility of low level visual deficits as a cause of reading disability.
Cognitive Neuropsychology, 5, 369-374.
Klingberg, T., Hedehus, M., Temple, E., Salz, T., Gabrieli, J. D. E., Mosely, M. E., & Poldrack, R. A.
(2000). Microstructure of temporoparietal white matter as a basis for reading ability; evidence from
diffusion tensor magnetic resonance imaging. Neuron, 25, 493-500.
Liberman, I. Y., Shankweiler, D., Fischer, F. W., & Carter, B. (1974). Explicit syllable and phoneme
segmentation in the young child. 1. Exp. Child Psychol., 18, 201-212.
Livingstone, M. S., Rosen, G. D., Drislane, F. W., & Galaburda, A. M. (1991). Physiological and
anatomical evidence for a magnocellular defect in developmental dyslexia. Proc. Natl. Acad. Sci.
USA, 88, 7943-7947.
Lovegrove, W. J., Martin, F., Blackwood, M., & Badcock, D. (1980). Specific reading difficulty:
Differences in contrast sensitivity as a function of spatial frequency. Science, 210, 439-440.
Martin, F., & Lovegrove, W. (1987). Flicker contrast sensitivity in normal and specifically disabled
readers. Perception, 16, 215-221.
Mason, A., Cornelissen, P., Fowler, M. S., & Stein, 1. F. (1993). Contrast sensitivity, ocular dominance
and reading disability. Clinical Visual Science, 8, 4, 345-353.
McAnally, K. I., & Stein, 1. F. (1996). Abnormal auditory transient function in dyslexia. Proc. Roy. Soc.
B., 263, 961-965.
Menell, P., McAnally, K. I., & Stein, 1. F. (1999). Psychophysical and physiological responses to
amplitude modulations in dyslexia. 1. Speech and Hearing Res, 42, 797-803.
Merrigan, W. H., & Maunsell, 1. R. (1993). How parallel are the primate visual pathways? Ann. Rev.
Merzenich, M. M., Jenkins, W. M., Johnston, P., Schreiner, c., Miller, S., & Tallal, P. (1996). Temporal
processing deficits of language-learning impaired children ameliorated by stretching speech. Science,
271,77-81.
Miles, T. (1983). Dyslexia, the pattern of difficulties. (Granada; 2nd ed. 1993) Whurr Wyke: London.
Milner, A. D., & Goodale, M. A. (1995). The Visual Brain inAction. Oxford: UP.
Morgan, W. P. (1896). Word Blindness. Br Med1., 2, 1378.
Newsome, W. T., Britten, K. H., & Movshon, 1. A. (1989). Neuronal correlates of a perceptual decision.
Nature, 341, 52-54.
Nicolson, R. I., Fawcett, A. 1., Berry, E. L., Jenkins, I. H., Dean, P., & Brooks, D. J. (1999). Motor
learning difficulties and abnormal cerebellar activation in dyslexic adults. Lancet, 353, 43-47.
Ogle, K. (1962). The Optical Space Sense. In H. Davson (Ed), The Eye, VoI.IV. New York, London:
Academic Press.
Olson, R. K., Wise, 8., Connors, F., Rack, J., & Fulker. D. (1989). Specific deficits in component reading
and language skills. 1. Learn. Disabil., 22, 339-348.
Pennington, B. F. (Ed.). (1991). Reading disabilities: Genetic and neurological influences. Dordrecht,
The Netherlands: K1uwer.
Rae, c., Martin, A. L., Dixon, R. M., Blamire, A. M., Thompson, C. H., Styles, P., Talcott, J.,
Richardson, A. 1., & Stein, 1. F. (1998). Metabolic abnormalities in developmental dyslexia detected
by magnetic resonance spectroscopy. Lancet, 351, 1849-52.
Riddell, P., Fowler, M. S., & Stein, J. F. (1988). Vergence eye movements and dyslexia. Dyslexia
Contact, 7, 2, 5-6.
Riddell, P., Fowler, M. S., & Stein, J. F. (1990). Spatial discrimination in children with poor vergence
control. Perceptual & Motor Skills, 70, 707-718.

Shapley, R., & Perry, V. H. (1986). Cat and monkey retinal ganglion cells and their functional roles. TINS
9,229-235.
Skottun, B. C. (2000). The magnocellular deficit theory of dyslexis; the evidence from contrast
sensitivity. Vis. Res., 40, 111-27.
Snowling, M. (1981). Phonemic deficits in developmental dyslexia. Psychological Research, 43, 219-34.
Snowling, M. (1987). Dyslexia: A cognitive developmental perspective. Oxford: Blackwells.
Stein,1. F., & Fowler, M. S. (1981). Visual dyslexia. Trends in Neuroscience, 4,77-80.
Stein, J. F., & Fowler, M. S. (1985). Effect of monocular occlusion on visuomotor perception and reading
in dyslexic children. The Lancet, 13, 69-73.
Stein, J. F., & Glickstein, M. (1992). The role of the cerebellum in the visual guidance of movement.
Physiological Reviews, 72, 967-1018.
Stein, 1. F., & McAnally, K. I. (1996). Impaired auditory temporal processing in Dyslexics. Irish J.
Psych., 16, 220-228.
Stein, J. F., Riddell, P., & Fowler, M. S. (1988). Disordered vergence eye movement control in dyslexic
children. Brit. J. Ophthalmol., 72, 162-166.
Stein, 1. F., Richardson, A. 1., & Fowler, M. S. (2000). Monocular occlusion can improve binocular
control and reading in dyslexic children. Brain, 123, 164-170.
Stein, J. F., & Talcott, J. B. (1999). The Magnocellular Theory of Dyslexia. Dyslexia, 5, 59-78.
Stein, J. F., Talcott, 1. B., & Walsh, V. (2000). Controversy about the visual magnocellular deficit in
developmental dyslexics. Trends in Cog. Sc., 4, 209-211.
Stein, J. F., & Walsh, V. (1997). To see but not to read; the magnocellular theory of dyslexia. TINS, 20,
147-152.
Studdert Kennedy, M., & Mody, M. (1995). Auditory temporal perception deficits in the reading-
impaired: A critical review of the evidence. Psychonomic Bulletin & Review, 2, 508-514.
Talcott, 1. B., Hansen, P. c., Elikem, L. A., & Stein, J. F. (2000). Visual motion sensitivity in dyslexia:
evidence for temporal and motion energy integration deficits. Neuropsychologia, 38, 935-43.
Talcott, J. B., Hansen, P. c., Willis-Owen, c., McKinnell, I. W., Richardson, A. J., & Stein, J. F. (1998).
Visual Magnocellular Impairment in Adult Developmental Dyslexics. Neuroophthalmology, 20, 187-
201.
Talcott, J. B., Witton, c., McClean, M., Hansen, P. c., Rees, A., Green, G. G. R., & Stein, 1. F. (1999).
Can sensitivity to auditory frequency modulation predict children's phonological and reading skills?
Neuroreport, 10, 2045-50.
Talcott, 1. B., Witton, c., McClean, M., Hansen, P.c., Rees, A., Green, G. G. R., & Stein, J. F. (2000).
Visual and auditory transient sensitivity determines word decoding skills. Proc. US Nat. Ac. Sc., 97,
2952.
Trussell, L. O. (1998). Cellular mechanisms for preservation of timing in central auditory pathways.
Current Opinion in Neurobiology, 7, 487-492.
Ungerleider, L. G., & Mishkin, M. (1982). Two cortical visual systems. In D. J. Ingle, M. A. Goodale, &
R. J. W. Mansfield (Eds.), The Analysis of Visual Behavior (pp.549-586). Cambridge, MA: MIT
Press.
Witton, c., Talcott, J. B., Hansen, P. c., Richardson, A. J., Griffiths, T. D., Rees, A., Stein, 1. F., & Green
G. G. R. (1998). Sensitivity to dynamic auditory and visual stimuli predicts nonword reading ability
in both dyslexic and normal readers. Current Biology, 8, 791-797.
1. B. TALCOTT & c. WITTON
A SENSORY-LINGUISTIC APPROACH TO NORMAL

AND IMPAIRED READING DEVELOPMENT
Abstract. In this chapter we outline a sensory-linguistic approach to the study of reading skill
development. We call this a sensory-linguistic approach because the focus of interest is on the
relationship between basic sensory processing skills and the ability to extract efficiently the orthographic
and phonological information available in text during reading. Our review discusses how basic sensory
processing deficits are associated with developmental dyslexia, and how these impairments may degrade
word-decoding skills. We then review studies that demonstrate a more direct relationship between
sensitivity to particular types of auditory and visual stimuli and the normal development of literacy skills.
Specifically, we suggest that the phonological and orthographic skills engaged while reading are
constrained by the ability to detect and discriminate dynamic stimuli in the auditory and visual systems
respectively.
1. SENSORY-LINGUISTICS: A "BOTIOM-UP" APPROACH TO STUDYING

LITERACY SKILL
Central to the study of the basic mechanisms of language and language disorders are
the following questions:
Why do such a large percentage of people (upwards of 6%, Shaywitz, Shaywitz,
Fletcher, & Escobar, 1990) fail to develop proficient reading skills despite having
normal intelligence? What is the neurobiological basis for specific reading disability
(developmental dyslexia)? What is the role (if any) of basic sensory mechanisms to
the development of normal reading (and other language) skills?
In this chapter we describe our "sensory-linguistic" approach to the study of
how basic sensory mechanisms might be involved in the development of proficient
reading skills. Skilled reading also relies upon higher-level language information
(e.g., semantic processing). We therefore restrict this review to the earliest stages of
text processing, when letter strings must be recognized accurately and then
subsequently linked with the speech sounds they represent to decode single words.
Single word reading is arguably the primary limitation on reading performance
(Olson, Forsberg, Wise, & Rack, 1994), constraining subsequent reading processes
(e.g., fluency and comprehension). Sensory processing mechanisms would be
predicted to be most influential at this early stage; visual and auditory skills might
constrain the extraction of orthographic and phonological information.
In the experimental studies reported here we are concerned primarily with
measuring dynamic stimulus detection, particularly for visual and auditory stimuli.
Therefore, we have used stimuli that require the perception of a dimension that
changes in time. Such measures of temporal processing might provide a means to
evaluate the importance of sensory skills to the reading difficulties characteristic of
developmental dyslexia (see Farmer & Klein, 1995, for review). Perception of

214 1. TALCOlT & C. WIlTON
dynamic auditory and visual events might also constrain the development of
component reading skills by limiting the ability to use the phonological and
orthographic information available in words.
1.1 Phonological and Orthographic Sensitivity in Normal Reading
Learning to read in most languages depends on understanding both its spoken

properties (phonology), and its written form (orthography). When children first learn
to read they often identify words by their visual attributes rather than as made up of
letters that correspond to sounds (Frith, 1985; Ehri & Wilce, 1985). Most written
languages, including English, employ alphabetic scripts; reading development then
depends on learning the relationship between letter-units and speech sounds. In
languages such as Spanish or German the relationship between letters and sounds is
relatively invariant, but readers of English are presented also with the problem that
most letter combinations can be mapped to more than one phoneme. As a
consequence, grapheme-phoneme correspondence rules do not always provide the
correct pronunciation of a letter string (e.g., "yacht").
There are a number of models that describe how phonological and orthographic
information might (or might not) interact during reading development (e.g., Ehri,
1997; Frith, 1985; Goswami, 1988; Seidenberg & McClelland, 1989), in skilled
reading (e.g., Jacobs & Grainger 1994; Morton, 1969; Van Orden, Pennington &
Stone, 1990) and in reading disabilities (e.g., Coltheart, 1978; Manis, Seidenberg,
Doi, McBride-Chang, & Petersen, 1996; Van Orden & Goldinger, 1996)1. However,
most current models acknowledge that orthographic and phonological processes,
however distinct, rarely operate entirely autonomously; rather, they bootstrap on
each other (e.g., Seidenberg & McClelland, 1989; Van Orden et aI., 1990).
Proficient extraction of both types of information is important for the development
of competent reading skills, with the most successful methods for teaching reading
employing training on both printed letter and word sounds; i.e., both orthographic
and phonological coding (e.g., Hatcher, Hulme, & Ellis, 1994; Tangel & Blachman,
1995). Letter knowledge and phonemic segmentation skill, measured in pre-readers,
have been shown to be the two best predictors of first grade reading achievement
(Share, Jorm, MacLean, & Matthews, 1984).
The importance of phonological skill to the development of reading skills is well
established (see Wagner & Torgeson, 1987, for review). However, the finding that
orthographic skills are at least partially independent from phonological ability has
been confirmed only recently (Castles, Datta, Gayan, & Olson, 1999; cf. Olson,
Wise, Conners, Rack, & Fulker, 1989; Olson, Forsberg, Wise, et aI., 1994). One
reason for this delay is that measures of orthographic skill are often confounded with
print exposure (and also with reading ability (Stanovich, West, & Cunningham,
1991». Both phonological and orthographic ability are significantly heritable, and
each accounts for independent variance in word recognition skill (Olson, Forsberg,
Wise et aI., 1994). Similarly, Stanovich et al. (1991) reported that orthographic
factors contribute independently to reading skill, even after accounting for
differences in print exposure and phonological ability. Talcott, Witton, et al. (2000)
reached a similar conclusion, but also showed that an orthographic
word/pseudohomophone choice test (e.g., "rain" vs. "rane"; after Olson, Forsberg,
SENSORy-LINGUISTICS 215
Wise, et aI., 1994) was the best predictor of reading ability in a sample of normal
lO-year olds.
Proficient phonological skills are essential to the development of good reading
skills (e.g., Wagner & Torgeson, 1987), although there is some debate about the
direction of causality between these two factors (Bowey & Francis, 1991; Morais,
Cary, Alegria & Bertelson, 1979; Wagner, Torgesson, & Rashotte, 1994).
Nevertheless, pre-readers who are sensitive to rhyme (e.g., those most able to
discriminate the oddball among "cat", "mat", "fan") are less likely to develop
reading problems early in life (Bradley & Bryant, 1983). Proficiency on such
phonological awareness tasks, segmentation tasks such as Pig Latin and
Spoonerisms, and phonological decoding tasks such as pseudoword naming, can
reliably discriminate good from poor readers across the lifespan (Bruck, 1990;
Gallagher & Fredrickson, 1995; Olson et aI., 1989; Stanovich, 1988; van Ijzendoorn
& Bus, 1994).
There is some evidence that adult dyslexics are less impaired on phonological
awareness tasks than tasks of phoneme manipulation and coding (Pennington, Van
Orden, Smith, Green, & Haith, 1990). Although these skills are closely related and
both are necessary for learning to read, there is some difference between phoneme
awareness, an ability to recognize and manipulate word sound units, and
phonological awareness, a more general skill for perceiving the acoustic properties
rather than the semantic meaning of speech. Phonological awareness skills may
develop automatically and are related to the ability to perform tasks such as
perceiving speech rhythm and intonation (Wood & Terrell, 1998). In contrast,
Morais and colleagues have argued persuasively that phonemic awareness is not
spontaneously acquired; it emerges primarily through the teaching of the alphabetic
code that links letter identity with speech sounds (cf. Bradley & Bryant, 1983;
Morais et aI., 1979). Consistent with this hypothesis is evidence that functional
illiterates seem to be fairly proficient, though less so than controls, at phonological
awareness tasks such as rhyming2 and pseudoword repetition (Castro-Caldas,
Petersson, Reis, Stone-Elander, & Ingvar, 1998). However, illiterates seem to be
unable to successfully solve most phoneme awareness tasks (such as phoneme
deletion and substitution) which require an ability to utilize segmental skills.
Segmented and unsegmented phonology may show different patterns of neural
activation, and these may be dissociated in dyslexia (Paulesu et aI., 1996). It is
therefore plausible that phonological awareness tasks could depend on
suprasegmental language listening skills that are more dependent upon generalized
speech perception mechanisms than are segmental phoneme awareness tasks.
The phonological representations in memory that are necessary for grapheme-
phoneme mapping have been argued to result from either a mechanism specific to
linguistic stimuli, such as the categorical perception of speech segments (Mody,
Studdert-Kennedy, & Brady, 1997; Shankweiler, Liberman, Mark, Fowler, & Fisher,
1979), or from more basic acoustic processing skills such as detecting and
sequencing rapid auditory events (Tallal, 1980; Tallal, Merzenich, Miller, &
Jenkins, 1998). In either case it is clear that the ability to detect accurately the fine-
grained distinctions characteristic of phonemes in spoken language is important for
the development of proficient reading skills. In this chapter, however, we discuss
how phonological skills are probably related to more generalized acoustic
216 J. TALCOlT& C. WIlTON
processing. Such supra-segmental skills could provide the template for making more
fine-grained acoustic discriminations such as phonemes.
2. ACOUSTIC STIMULUS PROCESSING, SPEECH PERCEPTION AND

PHONOLOGICAL AWARENESS
This section addresses the question of whether dyslexics' phonological processing

deficits could be related to their basic sensitivity to the acoustic changes present in
speech. If phonological awareness in the broadest sense can be defined as an ability
to attend to the acoustic structure of language (see above and Borstr0m & Elbro,
1997), accurate perception of the acoustic properties of speech could be considered a
phonological awareness skill. This general skill could limit more specific phoneme
segmentation skills that depend upon accurate speech perception. Thus, as described
below, reduced sensitivity to the basic acoustic components of spoken language
could plausibly constrain the phonological abilities of all readers.
Speech is a complex auditory signal, containing dynamic changes in frequency
and amplitude that need to be detected accurately, often in background noise.
Although most of us learn to perceive speech automatically, it could be the most
complicated acoustic stimulus that humans are required to process. The acoustic
properties of speech on which we have focused our studies are modulations in
frequency (FM) and amplitude (AM). Like phonological information, the acoustic
information in speech can occur at a number of different grain sizes (see Kay, 1982,
for review). Slow modulations (about 3-4 Hz) in frequency and amplitude carry the
prosodic or intonation information in spoken language; faster rates (up to about 20 -
50 Hz) can define speech segments such as the formant transitions in phonemes (see
for e.g., Liberman, Delattre, Gerstman, & Cooper, 1956; Rosen, 1992). Some
syllables are acoustically very similar; the stop consonant-vowel transitions \ba\ and
\da\ for example, only differ from each other in the direction of a frequency change
at the beginning of the first three formants.
Remez, Rubin, Pisoni and Carrell (1981) demonstrated the importance of
temporal changes to the perception of speech. They synthesized acoustic speech by
modulating three sinusoids, mirroring the changes in the center frequencies of the
first three formants of a sentence. Although these stimuli lacked the complex
harmonic structure and voicing of natural speech, the signals could be perceived as
speech by a majority of subjects, who were able to identify most of the words within
the sentences. Thus, even when speech is highly degraded and contains only
frequency (and amplitude) modulations of the single frequencies of the first three
formants, enough information is present for comprehension. This illustrates the
potential importance of FM and AM in speech comprehension. In Landau-Kleffner
syndrome (Stefanatos, 1993) and receptive dysphasia (Stefanatos, Green, &
Ratcliffe, 1989), severely impaired processing of frequency modulation can be
related to poor receptive language skills. The slow (4 - 16 Hz) amplitude
modulations in speech are particularly important for intelligibility (Drullman,
Festen, & Plomp, 1994) and there is some evidence that amplitude-envelope
fluctuations provide information about the identity of consonants in speech (e.g.,
Van Tassel, Soli, Kirby, & Widin, 1987). These cues may be especially useful to
hearing-impaired people, in combination with cues from lip-reading. It is possible
then that even relatively mild impairments in the detection of frequency and
amplitude modulation could have especially detrimental effects initially on learning
to perceive speech, and later on the development of phonological skills.
2.1 Speech Perception and Basic Acoustic Processing in Developmental Dyslexia
It is probable that the mechanism by which auditory processing affects phonological

skills is via speech perception (McBride-Chang, 1996). Early studies of dyslexics'
auditory processing (e.g., Tallal, 1980) have led to suggestions that any problems
would be particularly prominent for stop consonants, which are characterized by the
most rapid transitions. Thus, most of the studies of dyslexics' speech perception
have centered upon the discrimination of stop consonant-vowel transitions. Manis et
al. (1997) showed that a subset of dyslexic children with poor phoneme
segmentation skills were significantly less able than age- and reading age-matched
controls to discriminate consistently between the stop consonants fbi and /p/.
However, dyslexic children who had good phoneme awareness were not similarly
impaired 3• Adlard and Hazan (1998) showed that a subset of reading disabled
children were poorer than age- and reading ability-matched controls at
discriminating consonant contrasts differing in a single feature that was not
acoustically salient, specifically nasals and fricatives as well as stop consonants.
Conversely, other studies have provided evidence that some types of speech-
processing deficits in dyslexia might be a result of a language-specific deficit rather
than acoustic insensitivity. Mody et al. (1997) selected poor readers who were
impaired, relative to controls, in temporal order judgements (TOJt for the phonemes
fb/ and /d/. They found that the poor readers were no less sensitive than the controls
to non-speech sine-wave analogues of the first three formants of the stop
consonants, nor were they less sensitive to brief transitional cues that varied along a
synthetic speech continuum. This finding supports the hypothesis that the poor
readers' speech processing difficulties are specific to the domain of phonological
processing (e.g., specific to Tal's for syllables), rather than the processing of the
acoustic properties of speech.
Although most studies have centered on processing of sub-syllabic units,
dyslexics' difficulties might not be restricted to the briefest transitions in speech.
Wood and Terrell (1998) showed that primary school children who were poor
readers had a developmental delay in perceiving speech rhythm, even though there
was no significant difference between the poor readers and controls when
vocabulary was taken into account. Because speech rhythm provides important
segmental information about where syllable, word, and phrase boundaries occur,
sensitivity to such properties of language may be an important, but little recognized,
understood, or tested aspect of phonological awareness. They go on to suggest that
skills such as rhythmic awareness develop early in infancy to facilitate speech
perception and later impact upon phonological skill. Other research has suggested
that the effective segment size of phonological information becomes progressively
smaller with development (see Treiman & Zukowski, 1991, for review). Wood and
Terrell's (1998) study suggested that non-segmental phonological skills such as
rhythmic awareness are also important to the development of literacy skill.
218 l. TALCOTT & c. WITTON
It seems clear, therefore, that although the differences are often small and
difficult to measure, many dyslexics have difficulties in particular aspects of speech
perception. These speech perception deficits could be a causal or contributing factor
to dyslexics' poor phonological skills: attention to the acoustics of speech requires
accurate perception of the acoustics of speech. Therefore the question of whether or
not dyslexics' phonological processing deficits are related to deficient auditory
processing is a valid one, since the accurate speech perception required for
developing phonological skill must be mediated by the auditory system.
Tallal (1980) was the first to suggest that basic auditory processing of non-
linguistic stimuli is related to reading skill, based on a study of dyslexic children and
controls5• The study reported discrimination thresholds (using a sequencing TOl and
a same-different paradigm) for brief stimuli where the inter-stimulus interval (lSI)
was short, and demonstrated that the performance of the dyslexic children
deteriorated compared to the controls when the lSI was reduced below a critical
duration. In addition, the number of errors made on this task by the dyslexics was
significantly correlated with reading skill, suggesting that basic auditory processing
of non-linguistic stimuli could be related to the development of phonological skills
necessary for learning to read.
Other investigators have built upon this hypothesis with different measures of
auditory sensitivity, applied to samples of dyslexics and controls. Dyslexics have
been shown to require a larger difference between the frequency of two tones to be
able to discriminate between them. This has been measured psychophysically (e.g.
deWeirdt, 1988; France et aI., (in press); McAnally & Stein, 1996) and using
mismatch-negativity evoked potentials (Baldeweg, Richardson, Watkins, Foale, &
Gruzilier, 1999; Kujala, in this volume). Nagarajan et aI., (1999) showed that neural
responses to the second of two tones in a Tallal-type sequencing task were reduced
in dyslexic listeners compared to controls, especially at an lSI of 200 ms. This
suggested an interaction between pitch discrimination and the interval between the
two tone stimuli. Some other basic auditory processing deficits identified in dyslexia
include performance on measures of ability to extract dichotic pitch information
from noise (Dougherty, Cynader, Bjornson, Edgell, & Giaschi, 1998), and poor
auditory stream segregation (Helen ius, Uutela, & Hari, 1999).
Modulation processing has been investigated in a number of studies. McAnally
and Stein (1996) showed that a sample of dyslexic adults were less able than
controls to discriminate between the rate or depth of frequency modulation of a tonal
stimulus. Similarly, dyslexic adults were also found to have lower psychophysical
thresholds, and reduced electro-physiological responses to amplitude modulation of
a broadband stimulus compared to the controls (Mennel, McAnally & Stein, 1999).
As outlined above, both FM and AM are important components of spoken language,
so in our experiments we have investigated further the relationship between
sensitivity to certain modulations and phonological skill.
2.2 Frequency Modulation Sensitivity in Dyslexic and Normal Readers
The sensitivity of normal human listeners to auditory FM has been well

characterized (e.g., Demany & Semal, 1989; Moore & Sek, 1995; Schorer, 1986;
Zwicker, 1952). Sensitivity to FM of a pure tone is determined by the degree of
frequency change required for the subject to detect the presence of the modulation
(see Figure 1a). This can be measured simply in children and adults using standard
psychophysical techniques.
We have shown that adult dyslexics are less sensitive to 2 Hz and 40 Hz FM of a
500 Hz tone compared to controls (Witton et ai., 1998). Both of these rates of FM
are processed within one critical bandwidth6, and the auditory system detects the FM
by tracking the frequency change in time; in other words, these low rates of FM are
processed temporally. At a modulation rate of 2 Hz, subjects hear a "warble" in
pitch; 40 Hz FM is perceived as a "roughness". Thus the percepts associated with
these particular modulations reflect the dynamic tracking of frequency change in the
stimuli.
(a) FM (b) AM
(i)
(ii)
Time Time
Figurel. Sketches of sinusoidally (a) frequency- and (b) amplitude-modulated tone stimuli.
Top panels (i) depict the waveform of the tone and bottom panels (ii) show the modulation
waveform. For FM, the modulation waveform tracks the frequency of the tone as it deviates
above and below the mean carrier frequency, denoted by the horizontal line through the
modulation waveform. The arrows denote the modulation depth, the maximum deviation of
the frequency from that of the carrier. Similarly, the modulation waveform for AM follows the
deviations of the amplitude of a sound from its mean. In both figures arrows denote the extent
of the maximum deviation from the amplitude of the carrier, the modulation depth. In all the
FM and AM detection experiments described in the text, thresholds are described in terms of
the minimum depth of modulation required for the subject to detect its presence, in
comparison with an unmodulated pure tone.
Importantly, dyslexics were not less sensitive, compared to controls, to 240 Hz

FM (Witton et ai., 1998). At 240 Hz, the FM exceeds the critical modulation
frequency and the frequency' change cannot be tracked. Instead, the modulation is
processed using a spectral mechanism (Hartman & Hnath, 1982), and subjects
perceive a 240 Hz tone (the "residue"; Schouten, 1940).
220 J. TALCOTI & C. WITION
Because the differences between dyslexics and controls are found only at rates of
FM that are detected by tracking the dynamic changes in time, it appears that
dyslexics are less able to temporally process (i.e., track in real time) frequency
changes in sound (see also below). This dissociation in psychophysical thresholds is
important for two reasons. First, it suggests that dyslexics have a specific deficit in
detecting FM when it is processed temporally and tracked, but not when the spectral
mechanism dominates. Second, the lack of group difference for 240 Hz FM shows
that the dyslexics' difficulties with detecting the slower modulations (that are
measured with the same paradigm) cannot simply be a consequence of reduced
ability to perform the psychophysical experiments compared to the controls.
Therefore, the group differences found likely resulted from a sensitivity difference
to the temporal aspects of the stimuli rather than from other factors (e.g., difficulty
performing psychophysical tasks in general).
Two studies have reported negative results concerning the detection of FM in
sounds. Adlard and Hazan (1998) found that poor readers did not differ from
controls in the detection of FM when the carrier was a synthetic formant of 1 kHz.
They used rates between 60 and 300 Hz, higher than most of those used in our
studies. (NB: Witton et aI., 1998 did not find differences at 240 Hz FM). Hill,
Bailey, Griffiths, and Snowling (1999) measured FM detection thresholds using a
similar stimulus to ours. Although they did find that dyslexics were less sensitive
than controls on average, this difference was not reliable. The lack of consistent
differences between groups might be attributed in part to the comparatively small
difference between the dyslexic and control group's phonological skills or the
relatively small number of participants in the latter study.
In addition to the overall group differences in sensitivity, Witton et aI. (1998)
showed that acoustic FM sensitivity could be related to the phonological skills of
both dyslexic and control subjects. Importantly, only thresholds for 2 and 40 Hz FM
were statistically predictive (2 Hz, r = 0.53; 40 Hz, r = 0.45; 240 Hz r = 0.03) of
performance on a pseudoword naming measure of phonological decoding skill
(Castles & Coltheart, 1993). This dissociation in the pattern of correlations supports
the hypothesis that basic temporal sensitivity to acoustic modulations helps to
determine phonological skill. Because the correlation between pseudoword reading
and FM detection at low rates was also present in the controls, this suggested to us
that FM detection might play a more general role in constraining the phonological
abilities of all readers rather than simply in the special population of dyslexics.
Following up this preliminary study, we measured sensitivity to 2 and 240 Hz
FM in a single classroom of lO-year old children who were not pre-selected in any
way (Talcott, Witton, et aI., 2000). Consistent with our previous results (Witton et
aI., 1998) we found a strong relationship between pseudoword reading skill and
detection thresholds for 2 Hz FM. In addition, we found that 2 Hz FM was a strong
predictor of Spoonerisms (Gallagher & Frederickson, 1995) performance and overall
single-word reading ability in the British Abilities scales (BAS: Elliot, Murray, &
Pearson, 1983). Figure 2 shows the relationships between the various measures of
literacy skill and FM sensitivity at 2 and 240 Hz. In general 2 Hz FM sensitivity was
a stronger predictor of literacy and literacy-component skills than was 240 Hz FM
sensitivity. This was especially true when individual differences in reading
experience and cognitive skills were removed statistically (Talcott, Witton, et aI.,
2000). This was done for three reasons: first, the reciprocal nature of the causal
relationship between reading and phonological awareness (Morais et aI., 1979;

Wagner et aI., 1994) may obscure important covariance between our sensory and
phonological processing measures. Second, intelligence measures account for a
significant proportion (upwards of 10%) of the variance in children's word
recognition skills (e.g., Hansen & Bowey, 1994). Third, variance in psychophysical
task performance is associated with individual differences in cognitive skill (Hirsh
& Watson, 1996). Consistent with this evidence we found that a single measure of
cognitive skill could account for on average about 15% of the variance in the
literacy skills and psychophysical thresholds of our un selected sample (Talcott,
Witton, et aI., 2000). This is consistent with population-based estimates from the
BAS (Elliot et aI., 1983). Accounting for overall differences in intelligence and
reading skill serves a second purpose as it removes shared variance between the
orthographic and phonological performance measures. When the effects of overall
reading skill and intelligence were removed, the highest intercorrelation between
any orthographic and phonological measure dropped to 0.10, whereas the
correlations between the two orthographic measures (r = 0.48) and between the two
phonological measures (r = 0.58) remained strong.
I
Ic:::l
c::::J 240 H~ FM I
2 HzFM
I
I -- p:: O.OS
Spelling
I
• - -- p:: 0.01
p:: O.OO I
Nonword. I
, ~A'
I
Exception words I
I
I
Spoonerism I
iI
PltudohOlIIophona
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Spearman rank-order Correlation Coefficient
Figure 2. The relationship between FM detection performance and the literacy skills of
normal readers (n = 32). Filled bars designate bivariate correlation coefficients between a
measure of linguistic performance and 2 Hz FM detection. Unfilled bars designate
correlation coefficients between a measure of linguistic performance and 240 Hz FM
detection. Vertical lines depict the correlation coefficient necessary to exceed alpha at 0.05
(solid line), 0.01 (dashed line) and 0.001 (dotted line) respectively. Single word reading and
spelling measures are subscales of the British Abilities Scales (BAS). Nonwords and
exception word reading were adapted from Castles and Coltheart (1993), spoonerisms from
the Phonological Assessment Battery and word-pseudohomophone detection from Olson,
Wise, Forsberg, et al. (1994). See text for details and further references.
222 J. TALCOIT & C. WIITON
The high correlations between basic FM sensitivity and phonological and literacy
skills suggest that basic auditory sensitivity to the modulations important in speech
could constrain phonological skill in the general population of readers as well as in
special populations such as dyslexics. However, the correlations were not as high for
measures of orthographic skill (e.g., exception word naming and
pseudo homophones), especially when the intercorrelations among literacy skill
measures were removed by accounting for overall reading and cognitive skill. We
found that visual motion detection was a strong predictor of orthographic skill in
these same subjects (see below and Talcott, Witton, et aI., 2000).
1.8
Poor
1.6
!l'
e
II
1.4 00 0
2
1.2 0 0 •
g
-
0
gp 1.0
l
"E
0.8
0
• •
~ 0.6 0
• •
.
b
1 0.4
••• ••
0.2
Good Poor
0.0
-0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
2 Hz FM Detection Threshold (log 10 modulation index)
Figure 3. Log transformed detection thresholds for 2 Hz FM are plotted against pseudo-word
reading accuracy similarly transformed. The best fitting regression line and 95% confidence
intervals bounding the correlations are shown. Unfilled circles denote subjects who have 20
Hz AM detection thresholds falling above the group median; filled circles denote subjects
whose 20 Hz AM thresholds fall below the group median. High psychophysical thresholds are
equivalent to low sensitivity and vice versa. For clarity, each axis is labeled with "good" and
"poor" to clarify subject's relative performance on the dimension of interest.
2.3 Amplitude Modulation Sensitivity in Dyslexic and Normal Readers

As is the case for FM, human sensitivity to sinusoidal AM of tones is well
characterized (e.g. Moore & Sek, 1995; Schorer, 1986; Zwicker, 1952). AM is also
important for speech comprehension, as outlined above. Adult dyslexics have
previously been shown to have reduced sensitivity to AM, measured both
psychophysically and using evoked potentials (Men ell, McAnally, & Stein, 1999).
To examine whether a similar relationship to that demonstrated for FM sensitivity
and phonological skills is also found for AM sensitivity, we have measured
thresholds for detecting FM and AM of a 1 kHz tone in a group of adult subjects,
both dyslexics and controls (see also Witton, Stein, Stoodley, Rosner, & Talcott, in
press). Early results show that sensitivity to 20 Hz AM predicts significant variance
in pseudo-word reading skill; this contribution is statistically independent from that
predicted by 2 Hz FM sensitivity. However, neither 2 Hz AM nor 240 Hz FM

detection thresholds were strong predictors of nonword naming skill. This lack of an
effect for sensitivity to 240 Hz FM replicates our previously published findings
(Talcott, Witton, et aI., 2000; Witton et aI., 1998).
Together, thresholds for 2 Hz FM and 20 Hz AM can account for over one-third
of the variance in nonword reading (Witton et aI., in press). These rates of FM and
AM are processed by separate auditory temporal mechanisms. At modulation rates
of 2 Hz and 20 Hz, AM is processed using the same mechanisms based on the
excitation-pattern of neural responses (see Moore, 1997, for review). 2 Hz FM
detection uses a supplementary mechanism (based on phase locking to the
waveform, Moore & Sek, 1995). That AM and FM thresholds at these rates explain
independent variance in nonword reading performance suggests that different
aspects of auditory modulation sensitivity might impact upon separate components
of phonological skill.
2.4 Acoustic Processes in Reading Skill Development: Overview and Caveats
Although dyslexia is accompanied by acoustic (and visual, see below) processing

deficits, it is not clear to what extent these problems can be considered causal factors
rather than benign symptoms that are correlated with their reading problems but not
directly related to them. Rosen (1999) has suggested that two complementary
approaches could be used to clarify these issues regarding causation. The first is to
assess the incidence of auditory processing difficulties in persons who read
normally; and second, to identify and characterize dyslexics who have no such
processing problems. Our approach has been to focus primarily on the first of these
issues; we are interested most in how reading skill might develop in concert with
other sensory skills. We are also interested in examining whether or not the sensory
processes we believe are important for developing proficient reading skills in
English are similarly important for learning to read other scripts. We have replicated
our finding of 2 Hz FM sensitivity deficits in a reading disabled sample of
Norwegian children. These children were similarly insensitive to visual coherent
motion (see below). Such results suggest that the sensory deficits often found in
dyslexia reflect differences in underlying neural mechanisms, rather than arising as a
consequence of learning an inconsistent orthography such as English (Norwegian
orthography is much more regular than English).
One persistent issue that remains to be clarified is how the relatively small
deficits in auditory processing relate to dyslexics' comparatively greater deficits
found on measures of phonological skill. The effect-size in studies of auditory
processing is often small (see for e.g., McAnally & Stein, 1996; Menell et aI., 1999;
Witton et aI., 1998) and is therefore unlikely to offer sufficient reduction in
sensitivity to directly cause difficulties in distinguishing phonemes in speech.
Nevertheless, it is plausible that small effects measured in simple psychophysical
threshold paradigms could be translated into large effects in the natural speech
environment. For example, psychophysically measured thresholds for frequency
discrimination of 40 ms tones embedded in a known pattern of other tones are
between 10 and 20 Hz (Watson & Foyle, 1985). However, when tone-pattern in
which the target is embedded is subject to random change, resulting in a high level
224 J. TALCOIT & C. WIITON
of "stimulus uncertainty", thresholds are increased to hundreds of Hz, even in highly

trained subjects (Watson & Kelly, 1981). In speech a high level of stimulus
uncertainty is ubiquitous in the sense that the modulations are embedded in complex
and varying series of dynamic acoustic events. Thus the effects of a small deficit in
detecting modulations demonstrated in our psychophysical experiments could be
multiplied when experienced within natural speech.
3. ORTHOGRAPHIC PROCESSES IN VISUAL WORD RECOGNITION
Most reading begins with visual analysis of text. Low level visual information, in
the form of visual features, defined in terms of their luminance contrast, orientation
and size constitute the letters that are used to spell words. Knowledge of spelling
patterns provides a number of sources of orthographic information. Orthographic
information is defined inconsistently (see Corcos & Willows, 1993, for review), and
can come from a number of sources, ranging from the vi suo-perceptual (e.g., the
order or spatial position redundancy of letters in words, Mason, 1975) to a more
cognitive level (Massaro, Venezky, & Taylor, 1979).
Our hypothesis is that basic visual processes constrain the use of orthographic
information by limiting the ability to accurately encode information about the letter
position within words (see also Cornelissen, Hansen, Gilchrist, et aI., 1998;
Cornelissen, Hansen, Hutton, Evangelinou, & Stein, 1998). Therefore, we are most
interested in the perceptual processes involved in word recognition, e.g., the visual
orthographic information that ranges in grain size from the single letter to the whole
word (see Jacobs & Grainger, 1994, for review). However, the ability to utilize this
information may be strongly linked to print exposure (Stanovich & West, 1989) and
overall reading skill (Allington, 1978).
Orthographic skills may be particularly important for reading exception words
like "quay", or for determining which of two spellings of a word/pseudohomophone
pair (e.g., "rain" vs. "rane") is correct. In both cases, grapheme-phoneme
correspondence rules do not provide sufficient information to correctly solve the
task. A test of literacy skill can thus be considered orthographic if it taps the ability
to use and identify familiar letter sequences with minimal aid from phonological
information (Olson, Forsberg, Wise, et aI., 1994). Although phonological processing
may be engaged during these orthographic tasks, the result of such processing is not
sufficient to determine the identity of a lexical string.
The evidence for a causal relationship between impaired phonological awareness
and reading difficulties has been so widely accepted that it is often assumed that
such a deficit is a prerequisite for poor reading (Le., a necessary rather than a
sufficient condition). Conversely, the relationship between measures of peripheral
and central visual function and reading proficiency is much more controversial (see
for e.g., Vellutino, 1979) because visual (and auditory) factors may appear to have
poor face validity in comparison with the prominence of what seem to be deficits
specific to language processing. However, if text processing is considered in light of
what the retina "sees", it becomes apparent that visual problems, however slight, can
dramatically degrade the visual image that must be accurately processed while
reading. At a constant viewing distance of 35 cm, the font on this page (10 point
Times New Roman) has about 7 characters per cm. On the retina this is equivalent to
about 4 letters per degree of visual angle; only the central 2 degrees support high
visual acuity so the information available is already substantially attenuated. If we
consider the spatial frequency information in this text7, the average letter is 3.5
cycles per degree (cpd), the average 7-character word is about 0.5 cpd. Intermediate
grain sizes such as the syllable occur at about 1 cpd. Thus, on the fovea, which
subtends about the central 2 degrees of visual angle on the retina, there is enough
room for only about a single word. Couple this sparse information coding with the
noise generated by eye movements while reading and associated saccadic
suppression (see below), and then the reading process, as it pertains to the low level
visual system, is a marvel of neuroengineering. Yet these factors are nearly always
discounted in the etiology of dyslexics' information processing difficulties.
Degrading text, which simulates sensitivity differences in low level visual filters,
can have a profound effect on both the legibility of the text and on the speed that it
can be read. Simply reducing the luminance contrast between the print and its
background can halve the reading speed of normally reading adults (Legge, Rubin,
& Luebker, 1987). Similarly, Williams, May, Solman, and Zhou (1995) have shown
that lowering the contrast between letters and their background impaired children's
rate of visual search. The magnocellular (or transient)8 pathway, suggested to be
selectively impaired in many dyslexics (see Vigyasagar & Pammer, 1999), likely
limits such visual search processes.
Cornelissen and colleagues (Cornelissen, Hansen, Gilchrist, et aI., 1998;
Cornelissen, Hansen, Hutton, et aI., 1998) have provided evidence for a more direct
link between visual processes and the integrity to which orthographic information is
encoded. Normal subjects, that were defined as "poor" motion detectors, as
determined by a median split on a coherent motion detection task (see below), were
more likely than "good" motion detectors to falsely report anagrams as words in a
tachistoscopic word/anagram (e.g., "ocean" vs. "oaecn") discrimination task. This
suggested that dynamic visual processing skills, similar to those described for
auditory processes (see above) might be directly related to the reading process,
presumably by limiting the accuracy by which letter position could be encoded in
the early stages of print analysis.
3.1 Visual Deficits Associated with Dyslexia: An Overview

Much of the contemporary psychophysical work examining how dyslexics process
basic visual information in the peripheral visual filters can be credited to Lovegrove
and colleagues. In a series of studies they provided strong evidence to link dyslexia
with poor sensitivity of the visual transient system9 (e.g., Lovegrove, Bowling,
Badcock, & Blackwood, 1980; Lovegrove, Martin, & Slaghuis, 1986; Martin &
Lovegrove, 1984, 1987; Slaghuis & Lovegrove, 1985). This was suggested on the
basis of three main findings. First, dyslexic readers were usually found to be less
sensitive than controls to spatial frequency gratings at low but not high spatial
frequencies (Lovegrove et aI., 1980; Martin & Lovegrove, 1984; cf. Skottun, 2000).
Second, dyslexics were found to be less sensitive to flickering stimuli, both full field
flicker (Brannan & Williams, 1988a, b) and flickering gratings, especially at higher
temporal frequencies (Martin & Lovegrove, 1987). They also have lower critical
flicker fusion (CFF) frequencies (Talcott et aI., 1998). Third, dyslexics have a
226 J. TALCOlT& C. WIlTON
smaller increase in visible persistence duration than do controls. Visible persistence

duration usually increases with increasing spatial frequency, but dyslexics often
show a shallower gradient in this increase with spatial frequency than do controls
(Slaghuis & Lovegrove, 1985).
What had been lacking in the early studies was a description of the mechanism
by which a dysfunctional transient system would directly affect reading skill.
However, many of the group differences between dyslexics and controls in
Lovegrove and colleagues' studies appeared to take place over stimulus durations in
which the time course of the rapid onset transient response and the slower sustained
response were most overlapping (see Lovegrove, 1991, for review). It was therefore
argued that normal transient responses at the end of fixation sequences during
reading serve to erase the iconic buffer that serves as the temporary store of visual
information by inhibiting the sustained system (Breitmeyer, 1993; Lovegrove,
1991). In dyslexia, a slowed or temporally inaccurate transient signal would lead to
ineffective suppression of the visual information from the prior fixation during
saccadic eye movements, the consequence of which would cause visual confusions
and ultimately inaccurate and slowed reading.
However, a number of researchers have shown that it might be the transient
system rather than the sustained system that is inhibited during a saccadic eye
movement (Burr, Morrone, & Ross, 1994). Therefore it is almost certainly not the
case that poor transient-on-sustained inhibition is the mechanism by which poor
transient function disrupts reading in dyslexia. Some researchers have suggested that
these findings provide direct evidence against the transient theory of dyslexia
(Hogben, 1997; Skottun & Parke, 1999). However, there are several alternative
hypotheses that can address the mechanism between reading disabilities and poor
visual transient function; we focus our studies on central (rather than peripheral)
visual processes which are closer (in anatomical terms) to the reading processes
which they might affect (see below).
It should be noted that the effect sizes of the deficits between groups found for
stimuli that assess the peripheral visual system (e.g., flicker and spatial frequency
gratings) are usually smaller than those found for more central visual stimuli such as
coherent motion (see below). For example, Talcott, et a1. (1998) showed that while
dyslexics were less sensitive than controls overall, both to high frequency flicker
and coherent motion, the effect size of the former was substantially smaller. The
results are also less consistent when peripheral visual stimuli are used. Several
studies have failed to replicate the pattern of transient selective visual deficits found
by Lovegrove and colleagues (e.g., Gross-Glenn et aI., 1995; Walther-Muller, 1995).
However, the lack of a selective visual deficit in the majority of cases was found in
the presence of a sensitivity difference in both the magnocellular (M) and
parvocellular (P) streams (for review see Skottun, 2000; Stein, Talcott, & Walsh,
2000).
The strongest evidence for a selective visual deficit in dyslexia comes, not from
studies of their peripheral visual systems, but from the study of their thresholds for
stimuli such as coherent motion which tap the sensitivity of cortical (mostly
extrastriate) visual areas. Such stimuli require spatial (and temporal) integration
from a number of directionally selective units (see for e.g., Heeger, Boynton, Demb,
Seideman, & Newsome, 1999) and therefore cannot be detected by the lower level
visual filters. Coherent motion stimuli are also ideal for testing the magnocellular
deficit hypothesis of developmental dyslexia (Stein & Walsh, 1997; Stein, in this
volume), since the detection of visual motion is mediated primarily by the
extrastriate visual areas which receive a predominant input from magnocells (see
below).
A large literature exists to show that dyslexics are less sensitive than both age
and reading-age matched controls on temporal processing measures (e.g., those that
require the detection or discrimination of short duration or rapidly presented stimuli)
for both auditory and visual stimuli (see Farmer & Klein, 1995, for review). It
should be noted that Studdert-Kennedy and Mody (1995) have suggested that the
common operational definition of temporal processing uses stimulus duration as the
relevant temporal dimension. Detection of time-limited stimuli is probably different
from the detection of stimuli in which the crucial parameter is dynamic change. It is
therefore important to appreciate the difference between rate of processing for
discrete and unchanging stimuli (e.g., gap detection paradigms) and processing of
rate where the stimulus itself is changing over time (e.g., visual motion).
In many of our recent studies, we have examined dyslexics' thresholds for
dynamic stimulus detection. The duration of these stimuli is not restricted; the use of
long durations ensures that any group differences in stimulus detection result from
sensitivity differences to the dynamic aspects of the stimuli, rather than general
difficulties in detecting short duration stimuli.
3.2 Coherent Motion and Dorsal Stream Sensitivity
Coherent motion stimuli provide an effective stimulus for measuring the sensitivity
of cortical visual areas such as V5/MT and the posterior parietal cortex, both of
which receive a high proportion of afferent fibers from the M-stream (see Merigan
& Maunsell, 1993; Milner & Goodale, 1995, for review). Such stimuli can assess the
sensitivity of more central (or cortical) transient system function. This is
advantageous because central visual areas, such as V5/MT, provide the neural
substrate for encoding the spatial location of objects and guiding movements toward
visual targets, including saccadic eye movements while reading (Milner & Goodale,
1995). This dorsal stream for determining object position has been likened to a
"where" pathway; the more ventral stream for object recognition has been called the
"what" pathway (Milner & Goodale, 1995).
Sensitivity to coherent motion in random dot kinematograms (RDK), such as
those depicted in Figure 4, can be measured with multi-frame stimuli (Newsome &
Pare, 1988; Wattam-Bell, 1994). In the typical RDK, a proportion of the total pixel
elements, or "dots", move coherently with their direction of motion perfectly
correlated over successive screen refreshes. The remaining noise dots move
independently at the same speed such that their directions of motion do not correlate
over time. Lifetimes of single dots are limited to prevent subjects from detecting
motion by tracking the trajectory of a single dot. Thus, on every screen refresh a
certain percentage of dots disappear; they then reappear at random positions within
the stimulus patch. Subjects' ability to see global coherent RDK motion therefore
depends on the successful detection and subsequent integration of local motion
signals over both space and time (Braddick, 1974, 1993).
228 J. TALCOIT& C. WIITON
Firing rates of cells within MT can predict primates' responses to direction of

motion in psychophysical tasks (Newsome, Britten, & Movshon, 1989) and lesions
to the same areas selectively reduce psychophysical sensitivity to motion stimuli
(Newsome & Pare, 1988). Lesions to the homologue of primate V5/MT and
surrounding areas have been shown to cause profound deficits in the extraction of
motion signals from dynamic noise in humans (Baker, Hess, & Zihl, 1991; Zihl, von
Cramon, & Mai, 1983). Visual tasks that tap the sensitivity of M-cells up to and
including primary visual cortex are reported to be only slightly impaired in the same
subjects (Hess, Baker, & Zihl, 1989) .
" , ~ ".
.
.. ... ...1
.....
Ii .. "II
~ ".' ..
~
II" 11+
... ~
.. Ii ,. ...
a b c
Figure 4. Schematic diagrams of the coherent motion (panels a, b) and coherent form (panels
c, d) paradigms. For the coherent motion stimuli, the arrows represent the motion vector of
each dot in the panel. Panel a depicts 50% coherence; half the dots are moving together,
whereas the rest of the dots are moving in random directions. Panel b depicts the random
motion in the non-target RDK patch. Each dot moves in a random direction so the average
coherence value is close to 0%. Panels c and d depict the coherent form task. In this measure
subjects detect a circle which is comprised of coherently oriented line segments. Panel d
shows the patch that contains the coherent form. See text for details.
3.3 Selective Deficits for Coherent Motion Detection in Developmental Dyslexia
A number of psychophysical studies have shown that dyslexic adults are less
sensitive than controls to coherent motion stimuli (Cornelissen, Richardson, Mason,
Fowler & Stein, 1995; Talcott et aI., 1998; Talcott, Hansen, Assoku, & Stein, 2000;
Witton et aI., 1998). These deficits were shown to be robust to changes in the
stimulus duration and dot density of the RDK displays (Talcott, et aI., 2000),
demonstrating that the dyslexics had poor detection of the dynamic properties of the
stimulus, rather than generalized poor detection of short duration stimuli.
More recently we conducted a study to determine whether or not these visual
deficits are limited to dynamic stimuli. We measured both the sensitivity of both
dyslexics and controls to both coherent motion and a non-dynamic control task of
coherent form sensitivity (see Figure 4). The form coherence task was designed to
be nearly identical to the motion task, except that subjects must detect a coherent
pattern (defined as the proportion of oriented line segments that constituted a circle)
rather than dynamic motion. Atkinson et ai. (1997) had previously used coherent
form and coherent motion to assess a sample of children with Williams syndrome, a
developmental disorder associated with deficits of visual dorsal stream function. The
children with Williams syndrome had increased thresholds for coherent motion
detection. However, they were not as impaired relative to controls in their ability to
detect coherent form.
Our study of dyslexics suggested a similar pattern of results (Hansen, Stein,
Orde, Winter, & Talcott, 2001). Dyslexics showed the expected motion detection
deficits relative to controls but they were not impaired on the coherent form task.
Furthermore, performance across the two measures was poorly correlated,
suggesting that they measured different attributes of visual function.
Psychophysical results from other laboratories have verified this basic difference
between dyslexics' and controls' visual motion processing (Everatt, Bradshaw, &
Hibbard, 1999; Raymond & Sorenson, 1998). Furthermore, neuroimaging studies
have confirmed that extrastriate areas important for extracting visual motion such as
V5/MT do not show the same robust activations to visual motion in dyslexics
compared to controls (Eden et aI., 1996). Demb, Boynton and Heeger (1997) related
this finding directly to the reading skills of their dyslexic subjects; the magnitude of
the hemodynamic response in extrastriate area MT and adjacent motion processing
areas co-varied strongly with their overall reading skills.
3.4 Motion Detection and Orthographic Sensitivity in Unselected Readers
Both the studies by Demb et al. (1997) and Cornelissen and colleagues (Cornelissen,
Hansen, Gilchrist, et aI., 1998; Cornelissen, Hansen, Hutton, et aI., 1998) showed
that coherent motion sensitivity was strongly correlated with particular aspects of
reading skill. Like Cornelissen and colleagues, our hypothesis is that visual skills
should impact most strongly on measures of orthographic sensitivity. Therefore we
measured the relationship between children's literacy skills and coherent motion
thresholds in the same sample of lO-year old children as described in the section on
auditory PM (see section 2.6 above and Talcott, Witton et aI., 2000). These results
are summarized in Figure 5. The main finding was that measures of orthographic
skill (e.g., irregular word naming and word/pseudohomophone choice, e.g., rane vs.
rain) and spelling co-varied most with sensitivity to coherent visual motion
(measures of phonological skill and BAS reading ability less so). This pattern of
correlation contrasts with the results depicted in Figure 2. Orthographic skills seem
to co-vary with visual coherent motion sensitivity, whereas phonological skills co-
vary most strongly with measures of acoustic sensitivity. As explained above (see
section 2.6) we had strong motivations for controlling for individual differences in
intelligence and overall reading skill, partly to account for the large statistical
overlap between the orthographic and phonological measures lO, which are both
strongly related to reading skill. Regression analyses then showed that coherent
motion detection could account for upwards of 20% of the residual variance in
overall orthographic skill as estimated by irregular word reading and
word/pseudohomophone discrimination.
We have since been running a study of over 350 school children between the ages of
7 and 11. Our early analyses suggest that, although the correlations between
measures are smaller overall, we have replicated our previous demonstration of a
relationship between children's sensitivity to dynamic visual and auditory stimuli
230 1. T ALCOIT & C. WIITON
and their performance on BAS reading and spelling and on component literacy skill
measures (Talcott, Witton, et aI., 2000). We have also replicated the distinction
between the patterns of correlation between the phonological and orthographic tests
and our measures of sensory sensitivity.
Rudin,
IC=:J Coherent Motion I
f -......._ --o.._-f-...J --p~O , 05
- -- p~O , OI
Spelll_c
. p~O , OO I
Nonwords
E:u :epdon Words
Spoonemau
PHudohomophones
0.0 0, 1 0,2 0,) 0,4 0.5 0,6 0,7 0.8 0,9 1,0
Spur.,a. ruk-order Correlation Co.rnd."1
Figure 5. The relationship between coherent motion detection performance and the literacy
skills of normal readers (n = 32). Vertical lines depict the rank-order correlation coefficient
necessary to exceed alpha at 0.05 (solid line), 0.01 (dashed line) and 0.001 (dotted line)
respectively. Literacy skill measures are the same as those in Figure 2. See text for study
details.
As an alternative hypothesis, it has been suggested that transient visual

processing, as measured by the Ternus test (a measure of apparent movement) is
related more to nonword reading skill than to exception word reading skill. Cestnick
and Coltheart (1999), for example, suggested that visual processing difficulties
would affect non word reading to a greater degree than exception words because the
former might need to be read letter by letter, as there is no top down information
about where the letters should be localized. Although we have found that motion
sensitivity does correlate with nonword reading (Talcott, et aI., 1998; Talcott,
Witton, et aI., 2000; Witton, et aI., 1998), we argue that the visual contribution to
exception word reading should be the stronger. Exception words cannot be read by
using phonological processes alone; hence, extra visual information is needed to
circumvent their orthographic irregularity. The relatively stronger correlations
between coherent visual motion and measures of orthographic sensitivity seem to
support this conclusion.
3.5 Potential Visual Mechanisms o/Orthographic Coding
So what is the mechanism by which dynamic visual processing limits sensitivity to

orthographic structure? We can still do little more than speculate, although the
specificity of the relationship between dynamic visual processes and aspects of
single word decoding point more to some visual functions than to others. As
mentioned above, it is unlikely that the mechanism is mediated by poor transient-on-

sustained inhibition during saccadic eye movements; our studies of single word
decoding accuracy are also unlikely to be affected significantly by such a
mechanism. However, it is probable that the deficits shown by many dyslexics on
measures of peripheral visual system function elicit textual degradation sufficient to
cause reading problems; however, this is unlikely to be the case for most poor
readers, whose visual problems, when present, are slight. Our correlations between
coherent motion sensitivity and orthographic skill, which occur across the range of
reading ability, suggest a more subtle influence of dynamic visual processing.
Sensitivity to visual motion can be related to the ability to encode accurately
information about letter position (Cornelissen, Hansen, Gilchrist, et aI., 1998;
Cornelissen, Hansen, Hutton, et aI., 1998). This sensitivity likely depends on input
from the magnocellular pathway, and information carried in this pathway is
generally considered to be important for encoding the spatial position of objects
(Milner & Goodale, 1995). Hence, coherent motion may serve as a direct
psychophysical measure of the sensitivity of neural structures important for
providing information about object position, of which orthographic sensitivity is a
specific instance. Further support for the direct relationship between orthographic
sensitivity and dorsal stream function comes from Samar, Parasnis, and Berent
(1999). They have recently reported that deaf dyslexics, who presumably are entirely
dependent upon orthographic rather than phonological information when reading,
had visual deficits in evoked potential studies. These were particularly striking for
low contrast checkerboard patterns designed to selectively stimulate M-cells.
Because our correlational studies are mute to issues of causation, we can not rule
out the possibility that reading experience (or print exposure) improves motion
detection performance rather than vice versa. In fact, it would be surprising to find
that there was no relationship between such measures of visual experience and a
visual psychophysical task. By this account one might expect the visual (and
auditory) processing skills of good readers to be more acute than those of dyslexics
by virtue of their more highly trained orthographic and phonological systems, rather
than a visual deficit arising from a fundamental neurological deficit in dyslexics.
However, several studies, primarily using peripheral visual stimuli (e.g., Brannan &
Williams, 1988a,b), have shown using both reading ability- and age-matched
groups, that older dyslexics are often less sensitive to such stimuli than younger
controls who are reading at the same level. Such evidence for a persistent visual
processing deficit in dyslexia that is not resolved with age or reading experience is
supported by evidence of anomalous neuroanatomy in both the periphery (LGN,
Livingstone, Rosen, Drislane, & Galaburda, 1991), primary visual cortex (Jenner,
Rosen, & Galaburda, 1999) and central visual systems of dyslexics (Rae et aI.,
1998)11. Furthermore, there is some suggestion that these visual neuroanatomical
differences are specific to cellular projections of the magnocellular pathway
(Livingstone, et aI., 1991). Dorsal stream structures with a strong input from
magnocells have also been implicated in the poor eye movement control (see Stein
this volume and Stein, Richardson, & Fowler, 2000), poor visual attention
(Steinman, Steinman & Garzia, 1998; Steinman, Steinman, & Lehmkuhle, 1997)
and poor visual search performance (Brannan & Williams, 1987; Vidyasagar &
Pammer, 1999; Williams, Brannan, & Lartigue, 1987) of many dyslexics. However,
232 1. TALCOTT & c. WITTON
it still has not been shown how these reported deficits might be directly related to
poor text processing.
Another possibility is that motion sensitivity might be related to the processing
of visible speech, although we would predict that this would affect phonological
skill more than orthographic skill. Lip reading has long been known to aid auditory
speech perception, especially when the acoustic information is degraded (for review
see Massaro, 1987). Campbell, Zihl, Massaro, Munhall, and Cohen (1997) have also
shown that patient LM, who has profound motion detection deficits resulting from
bilateral lesions to cortical area V5/MT and adjacent motion processing areas, is
insensitive to the dynamic information provided by the lips during utterances.
Although LM's ability to recognize speech patterns from static photographs is
unimpaired, she is unable to read lips in natural speech conditions. Deficits in
identifying visual tokens during lip reading tasks has also been reported in dyslexics
(Campbell, Whittingham, Frith, Massaro, & Cohen, 1997), although we know of no
studies that have explicitly linked coherent motion processing, lip reading and
measures of component literacy skills.
4. SENSORY-LINGUISTICS: SUMMARY AND CONCLUSIONS
Our sensory-linguistic approach to normal and dysfunctional reading development

examines the reading process from a bottom-up (e.g., a perceptual) perspective.
From this approach, basic sensory functions that occur early in the processing
hierarchy while reading can constrain the effectiveness with which phonological and
orthographic information is utilized to decode words. Our working hypothesis is that
dynamic auditory and visual sensitivity, respectively, can constrain phonological
and orthographic sensitivity. In addition to demonstrating that developmental
dyslexics are less sensitive to temporal changes in auditory and visual stimuli, we
have shown that sensitivity to 2 Hz FM and coherent motion can be used to predict
normal children's performance on tests of phonological and orthographic skill. The
route by which low-modulation rate FM sensitivity affects reading skill is probably
via an effect on phonological skill. In fact, if phonological awareness is includes the
general ability to attend to the sound properties of language (Wood & Terrell, 1998),
then 2 Hz FM sensitivity itself might be considered a phonological skill.
Whereas auditory sensitivity co-varies most strongly with phonological skills,
dynamic visual sensitivity, particularly for coherent motion, can be more strongly
related to sensitivity to orthographic structure. Coherent motion thresholds provide a
robust measure of the sensitivity of dorsal stream visual structures, such as V5/MT,
which receive a large input from magnocells. These structures are important for
encoding information about the spatial position of objects. Because sensitivity to
orthographic structure requires accurate letter position encoding, poor magnocellular
function could result in uncertainty about letter position while reading. This in tum
could result in inaccurate and slowed reading and poor spelling. Visual uncertainties
could also cause a knock-on effect to disrupt the learning of the alphabetic principle
by which letter units are consistently mapped to sounds.
Although our data do not prove a causal relationship, they do show that dynamic
sensory processes are likely to contribute to the development of children's literacy
skills. This hypothesis stands in contrast to theoretical perspectives that view literacy
SENSOR Y- LINGUISTICS 233
skill development as exclusive to the domain of modular language skill. Moreover,

whereas previous versions of a sensory processing hypothesis have been derived
mainly from experiments comparing dyslexic and control groups, our studies have
also examined un selected children who are learning to read. Our data suggest that
sensory processing is important for the development of literacy skills in all readers.
5. ACKNOWLEDGEMENTS
The research conducted by the authors was supported by grants from the Wellcome
Trust and the Rodin Remediation Foundation. We gratefully acknowledge John
Stein, in whose lab the authors' research was based. We also thank the families, staff
and, in particular, all of the students who participated in our primary school studies.
6. AFFILIATIONS
Dr. Joel B. Talcott

Neurosciences Research Institute
Behavioural and Cognitive Sciences Research Group
Aston University
Birmingham
B4 7ET, UK
e-mail: j.b.talcott@aston.ac.uk
7. NOTES
I It is not our intention to debate the existence of distinct subtypes of developmental dyslexia or whether
specific types of sensory processing impairment can be linked to putative subtypes (e.g., Borsting et aI.,
1996; Ridder, Borsting, Cooper, McNeel, & Huang, 1997). Such subtype distinctions might be based
upon arbitrary cutoff points along a continuum of skill describing population performance on reading
(Shaywitz, Escobar, Shaywitz, Fletcher, & Makuch, 1992) and on component subskills (Olson, Wise,
Conners, Rack, & Fulker, 1989).
2 Rhyming has been argued to require phonological segmentation skills, on the basis that rhyme
judgements must be made by comparing vowel sounds between words. Therefore, any preceding
consonants (e.g., onsets) must first be stripped off the vowel before a rhyme judgement can be made.
However, judging whether whole words rhyme need not involve such complex phoneme
manipulations; comparison of vowel sounds could be made on the basis of their acoustic properties
without applying phoneme segmentation.
J In our lab, sine wave speech experiments have suggested that, while dyslexics' perception of stop
consonants is impaired relative to controls, the impairment is equally severe for fricatives and nasals
(Rosner et aI., in press).
4 The use of temporal order judgements for detecting temporal processing deficits is somewhat
controversial (see for e.g., Studdert-Kennedy & Mody, 1995) in part because it involves not only
stimulus individuation but also labelling, which probably requires additional memory processes. France
et al. (in press) have shown that dyslexics' impairments in frequency discrimination tasks likely result
from poor sensory processing and poor short term memory.
; Following similar work by Efron (1963), Tallal and colleagues (e.g., Tallal & Piercy, 1973, 1974) found
evidence for rapid auditory processing deficits, measured by TOJs, in children with Specific Language
Impairment (SLJ). However, for the purpose of this review only research pertaining to developmental
dyslexia and reading will be considered here.
234 1. TALCOIT & C. WIITON
6 Critical bandwidth is a measure of the effective width of the band-pass auditory filters that characterise
the peripheral auditory system. Acoustic information must pass through these band-pass filters during
auditory processing. The critical modulation frequency is the maximum modulation rate at which the
carrier and first pair of spectral sidebands of a periodic signal can fall entirely within one critical
bandwidth; below this, sinusoidal FM and AM are processed temporally. For further review of the
properties of the auditory system the reader is referred to Moore (1997).
7 For a review of spatial frequency analysis in vision the reader is referred to Cornsweet (1970). A similar
treatment of the visual processes involved in sampling text can be found in Lehmkuhle (1993). For
further review of the role of the visual periphery in developmental dyslexia see Lovegrove (1991).
8 Neurophysiological work has shown the strong overlap between the structural and functional
neuroanatomy of primates and humans (see Shapley, 1990, and Merigan & Maunsell, 1993, for
review). Many investigators therefore have reconceptualized the older transient/sustained visual system
distinction (based on cellular response properties) as being equivalent to current anatomical
terminology (i.e., magno-/parvo-cellular). For the purposes of this review they will be treated
synonymously.
9 For a full review of the anatomy and function of the primate visual pathways, the reader is referred to
Merigan and Maunsell (1993) and Shapley (1990).

10 Other authors have found large correlations (r = 0.69) between measures of orthographic and
phonological skill (Coltheart & Leahy, 1996). Using similar items we have found similar robust
correlations (r = 0.76) in a study of over 350 children aged from 7 to 12 years.
11 Similar neuroanatomical anomalies have been reported in the primary auditory pathways of dyslexics
(Galaburda, Menard, & Rosen, 1994). Although there is no such clearly identifiable subsystem of
auditory magnocells as that seen in the visual system, there are magno-like divisions of each of the
auditory relay nuclei (Kaas, Hackett, & Tramo, 1999). These large neurons may be important for
processing the frequency and amplitude changes which can signal phonetic contrasts (Rauschecker,
1998).
8. REFERENCES
Adlard A, & Hazan V. (1998). Speech perception abilities in children with specific reading disabilities
(dyslexia). Quarterly Journal of Experimental Psychology, 51A, 153-177.
Allington, R. L. (1978). Sensitivity to orthographic structure as a function of grade and reading ability.
Journal of Reading Behavior, 10, 437-439.
Atkinson, J., King, J., Braddick, 0., Nokes, L, Anker, S., & Braddick, F. (1997). A specific deficit of
dorsal stream function in Williams syndrome. Neuroreport, 8, 1919-1922.
Baldeweg, T., Richardson, A, Watkins, S., Foale, c., & Gruzilier, J. (1999). Impaired auditory frequency
discrimination in dyslexia detected with mismatch evoked potentials. Annals of Neurology, 45, 495-
503.
Baker, C., Hess, R., & Zihl, J. (1991). Residual motion perception in a "motion-blind" patient, assessed
with limited lifetime random dot stimuli. Journal of Neuroscience, 11, 454-461.
Borsting, E., Ridder III, W., Dudeck, K., Kelly, c., Matsui, L., & Motoyama, J. (1996) The presence of a
magnocellular defect depends on the type of dyslexia. Vision Research, 36, 1047-1053.
Borstrl'lm, I., & Elbro, C. (1997). Prevention of dyslexia in kindergarten: Effects of phoneme awareness
training with children of dyslexic parents. In C. Hulme, & M. Snowling (Eds.), Dyslexia: Biology,
cognition and intervention (pp. 235-253). London: Whurr Publishers.
Bowey, J. A, & Francis, J. (1991). Phonological analysis as a function of age and exposure to reading
instruction. Applied Psycholinguistics, 12, 91-121.
Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning to read - a casual connection.
Nature,301,419-421.
Braddick, O. (1974). A short range process in apparent motion. Vision Research, 14, 519-527.
Braddick, O. (1993). Segmentation vs. integration in visual motion processing. Trends in Neurosciences,
16,263-268.
Brannan, J. R., & Williams, M. C. (1987). Allocation of visual attention in good and poor readers.
Perception & Psychophysics, 41, 23-28.
Brannan, J. R., & Williams, M. C. (1988a). Developmental versus sensory deficit effects on perceptual
processing in the reading disabled. Perception & Psychophysics, 44, 437-444.
Brannan, J. R., & Williams, M. C. (1988b). The effects of age and reading ability on flicker threshold.
Clinical Vision Science, 3, 137-142.
Breitmeyer, B. G. (1993). Sustained (P) and transient (M) channels in vision: a review and implication for
reading. In D. M. Willows, R. S Kruk, & E. Corcos (Eds.), Visual processes in reading and reading
disabilities (pp. 95-110). Hillsdale, NJ: Lawrence Erlbaum.
Bruck, M. (1990). Word-recognition skills of adults with childhood diagnoses of dyslexia. Developmental
Burr, D. c., Morrone, M. c., & Ross, 1. (1994). Selective suppression of the magnocellular visual
pathway during saccadic eye movements. Nature, 371,511-513.
Campbell, R., Whittingham, A., Frith, U., Massaro, D. W., & Cohen, M. (1997). Audiovisual speech
perception in dyslexics: Impaired unimodal perception but no audiovisual integration deficit. In
Proceedings of the (ECSA) workshop on audiovisual speech processing (pp. 85-89). Rhodes, Greece.
Campbell, R., Zihl, J., Massaro, D., Munhall, K., & Cohen, M. M. (1997). Speechreading in the
akinetopsic patient, L.M. Brain, 120, 1793-1803.
Castles, A., Datta, H., Gayan, J., & Olson, R. (1999). Varieties of developmental reading disorder:
Genetic and environmental influences. Journal of Experimental Child Psychology, 72, 73-94.
Castles A., & Colt heart M. (1993). Varieties of developmental dyslexia. Cognition, 47, 149-180.
Castro-Caldas, A., Petersson, K. M., Reis, A., Stone-Elander, S., & Ingvar, M. (1998). The illiterate
brain: Learning to read and write during childhood influences functional organization of the adult
brain. Brain, 121, 1053-1063.
Cestnick, L., & Coltheart, M. (1999). The relationship between language-processing and visual-
processing deficits in developmental dyslexia. Cognition, 71,231-255.
Coltheart, M. (1978). Lexical access in simple reading tasks. In G. Underwood (Ed.), Strategies of
information processing (pp. 151-216). London: Academic Press.
Coitheart, M., & Leahy, J. (1996). Assessment of lexical and nonlexical reading abilities in children:
some normative data. Australian Journal of Psychology, 48, 136-140.
Corcos, E., & Willows, D. M. (1993). The processing of orthographic information. In D. M. Willows, R.
S. Kruk, & E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp. 163-190).
Hillsdale, NJ: Lawrence Erlbaum.
Cornelissen, P., Richardson, A., Mason, A., Fowler, S., & Stein, J. (1995). Contrast sensitivity and
coherent motion detection measures at photopic luminance level in dyslexic readers and controls.
Vision Research, 35, 1483-1494.
Cornelissen, P. L., Hansen, P. C, Gilchrist, I., Cormack, F., Essex, 1., & Frankish, C. (1998). Coherent
motion detection and letter position encoding. Vision Research, 38, 2181-2191.
Cornelissen P. L., Hansen, P. c., Hutton J. L., Evangelinou V., & Stein J. F. (1998). Magnocellular
function and children's single word reading. Vision Research, 38,471-482.
Cornsweet, T. N. (1970). Visual perception. New York: Academic Press.
Demb, J. B., Boynton, G. M., & Heeger, D. J. (1997). Brain activation in visual cortex predicts individual
differences in reading performance, Proceedings of the National Academy of Sciences (USA), 94,
13363-13366.
Demany, L., & Semal, C. (1989). Detection thresholds for sinusoidal frequency modulation. Journal of
the Acoustical Society ofAmerica, 85, 1295-1301.
deWeirdt, W. (1988). Speech perception and frequency discrimination in good and poor readers. Applied
Dougherty, R. F., Cynader, M. S., Bjornson, B. H., Edgell, D., & Giaschi, D. E. (1998). Dichotic pitch: a
new stimulus distinguishes normal and dyslexic auditory function. Neuroreport, 9, 3001-3005.
Drullman, R., Festen, 1. M., & Plomp, R. (1994). Effect of reducing slow temporal modulations on speech
reception .. Journal of the Acoustical Society of America, 95, 1053-1064.
Eden, G. F., VanMeter, 1. W., Rumsey, 1. M., Maisog, J. M., Woods, R. P., & Zeffiro, T. A. (1996).
Abnormal processing of visual motion in dyslexia revealed by functional brain imaging. Nature, 382,
66-69.
Efron, R. (1963). Temporal perception, aphasia and deja vu. Brain, 86, 403-424.
Ehri, L. C. (1997). Sight word learning in normal readers and dyslexics. In BA Blachman (Ed.),
Foundations of reading acquisition and dyslexia: Implications for early intervention (pp. 163-189).
London: Lawrence Erlbaum.
Ehri, L., & Wilce, L. (1985). Movement into reading: Is the first stage of printed word learning visual or
phonetic? Reading Research Quarterly, 20, 163-179.
Elliot, c., Murray, D., & Pearson, L. (1983). British Abilities Scales. Windsor, UK: National Foundation
for Educational Research-Nelson.
Everatt, J., Bradshaw, M. F., & Hibbard, P. B. (1999). Visual processing and dyslexia. Perception, 28,
243-254.
236 1. TALCOTI & C. WITION
A review. Psychonomic Bulletin & Review, 2, 460·493.
France, S. J., Rosner, 8. S., Hansen, P. C., Calvin, C, Talcott, J. B., Richardson, A. J., & Stein, J. F. (in
press). Auditory frequency discrimination in adult developmental dyslexics. Perception &
Psychophysics.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K E. Patterson, 1. C Marshall, & M.
Coltheart (Eds.), Surface Dyslexia (pp. 310·330). London: Routledge and Kegan Paul.
Galaburda, A M., Menard, M. T., & Rosen, G. D. (1994). Evidence for aberrant auditory anatomy in
developmental dyslexia. Proceedings of the National Academy of Sciences (USA), 91, 8010-8013.
Gallagher, A., & Frederickson, N. (1995). The phonological assessment battery (PhAB): an initial
assessment of its theoretical and practical utility. Educational and Child Psychology, 12, 53-67.
Goswami, U. (1988). Orthographic analogies and reading development. Quarterly Journal of
Experimental Psychology, 40A, 239-268.
Gross-Glenn, K, Skottun, B. C, Glenn, W., Kushch, A, Lingua, R., Dunbar, M., Jallad, B., Lubs, H. A,
Levin, 8., Rabin, M.,Parke, L. A., & Duara, R. (1995). Contrast sensitivity in dyslexia. Visual
Hansen, 1., & Bowey, J. (1994). Phonological analysis skills, verbal working memory, and reading ability
in second-grade children. Child Development, 65, 938-950.
Hansen, P. C., Stein, J. F., Orde, S. R., Winter, J. L., & Talcott, J. 8. (2001). Are dyslexics' visual deficits
limited to measures of dorsal stream function? NeuroReport, 12, 1527-1530.
Hartmann, W. M., & Hnath, G. M. (1982). Detection of mixed modulation. Acustica, 50, 297-312.
Hatcher, P. J, Hulme, C, & Ellis, A W. (1994). Ameliorating early reading failure by integrating the
teaching of reading and phonological skills: The phonological linkage hypothesis. Child
Development, 65, 41-57.
Heeger, D., Boynton, G., Demb, J., Seideman, E., & Newsome, W. (1999). Motion opponency in visual
cortex. The Journal of Neuroscience, 19, 7162-7174.
Helenius, P., Uutela, K, & Hari, R. (1999). Auditory stream segregation in dyslexic adults. Brain, 122,
907-913.
Hess, R., Baker, C, & Zihl, J. (1989). The "motion-blind" patient: LoW-level spatial and temporal filters.
JournalofNeuroscience,9,1628-164O.
Hill, N. I., Bailey, P. J., Griffiths, Y. M., & Snowling, M. J. (1999). Frequency acuity and binaural
masking release in dyslexic listeners. Journal of the Acoustical Society of America, 106, 53-58.
Hirsh, I. J., & Watson, C S. (1996). Auditory psychophysics and perception. Annual Review of
Hogben, 1. (1997). How does a visual transient deficit affect reading? In C Hulme, & M. Snowling
(Eds.), Dyslexia: Biology, cognition, and intervention (pp. 59-71). London: Whurr Publishers.
Jacobs, AM., & Grainger, 1. (1994). Models of visual word recognition: Sampling the state of the art.
Journal of Experimental Psychology: Human Perception and Performance, 20, 1311-1334.
Jenner, A. R., Rosen, G. D., & Galaburda, A M. (1999). Neuronal asymmetries in the primary visual
cortex of dyslexic and non-dyslexic brains. Annals of Neurology, 46, 189-196.
Kaas, J. H., Hackett, T. A, & Tramo, M. J. (1999). Auditory processing in primate cerebral cortex.
Current Opinion in Neurobiology, 9, 164-170.
Kay, R. H. (1982). Hearing of modulation in sounds. Physiological Reviews, 62, 894-975.
Legge, G. E., Rubin, G. S., & Luebker, A (1987). Psychophysics of reading-V: The role of contrast in
normal vision. Vision Research, 27, 1165-1175.
Lehmkuhle, S. (1993). Neurological basis of visual processes in reading. In D. M. Willows, R. S Kruk, &
E. Corcos (Eds.), Visual processes in reading and reading disabilities (pp. 77-94). Hillsdale, NJ:
Erlbaum.
Liberman, A. M., Delattre, P. C, Gerstman, L. J., & Cooper, F. S. (1956). Tempo offrequency change as
a cue for distinguishing classes of speech sounds. Journal of Experimental Psychology, 52, 127-137.
Livingstone, M., Rosen, G. D., Drislane, F. W., & Galaburda, AM. (1991). Physiological and anatomical
evidence for a magnocellular defect in developmental dyslexia. Proceedings of the National Academy
of Sciences (USA), 88, 7943-7947.
Lovegrove, W. (1991). Spatial frequency processing in dyslexic and normal readers. In J. F. Stein (Vol.
Ed.), Vision and visual dysfunction: Vol. 13. Vision and visual dyslexia (pp. 148-154). Boca Raton,
F1: CRC Press.
Lovegrove, W., Bowling, A., Badcock, D., & Blackwood, M. (1980). Specific reading disability:
differences in contrast sensitivity as a function of spatial frequency. Science, 210, 439-440.
Lovegrove, W., Martin, F., & Siaghuis, W. (1986). A theoretical and experimental case for a visual deficit
in reading disability. Cognitive Neuropsychology, 3, 225-267.
Manis, F. R., Seidenberg, M. S., Doi, L. M., McBride-Chang, c., & Petersen, A. (1996). On the bases of
two sUbtypes of developmental dyslexia. Cognition, 58, 157-195.
Manis, F., McBride-Chang, C., Seidenberg, M., Keating, P., Doi, L., Munson, B., & Petersen, A. (1997).
Are speech perception deficits associated with developmental dyslexia? Journal of Experimental
Child Psychology, 66, 211-235.
Martin, F., & Lovegrove, W. (1984). The effects of field size and luminance of contrast sensitivity
differences between specifically reading disabled and normal readers. Neuropsychologia, 22, 72-77.
readers. Perception, 16, 215-221.
Mason, M. (1975). Reading ability and letter search time: Effects of orthographic structure defined by
single-letter positional frequency. Journal of Experimental Psychology: General, 104, 146-166.
Massaro, D. W. (1987). Speech perception by ear and eye: a paradigm for psychological inquiry.
Hillsdale, NJ: Erlbaum.
Massaro, D. W., Venezky, R. L., & Taylor, G. A. (1979). Orthographic regularity, positional frequency
and visual processing of letter strings. Journal of Experimental Psychology: General, 108, 107-124.
McAnally, K, & Stein, 1. F. (1996). Auditory temporal coding in dyslexia. Proceedings of the Royal
Society of London (B), 263,961-965.
McBride-Chang, C. (1996). Models of speech perception and phonological processing in reading. Child
Development, 67, 1836-1856.
Menell, P., McAnally, K E., & Stein, J. F. (1999). Psychophysical sensitivity and physiological response
to amplitude modulation in adult dyslexics. Journal of Speech Language and Hearing Research, 42,
797-803.
Merigan, w., & Maunsell, J. (1993). How parallel are the primate visual pathways? Annual Review of
Neuroscience,16,369-402.
Milner, D., & Goodale, M. (1995). The visual brain in action. Oxford, UK: Oxford University Press.
Mody, M., Studdert-Kennedy, M., & Brady, S. (1997). Speech perception deficits in poor readers:
Auditory processing or phonological coding? Journal of Experimental Child Psychology, 64, 199-
231.
Moore, B. C. 1. (1997). Introduction to the psychology of hearing (4 th Ed.). London: Academic Press.
Moore, B. C. 1., & Sek, A. (1995). Effects of carrier frequency, modulation rate, and modulation
waveform on the detection of modulation and the discrimination of modulation type (amplitude
modulation versus frequency modulation). Journal of the Acoustical Society of America, 97, 2468-
2478.
Morais, J., Cary, 1., Alegria, 1., & Bertelson, P. (1979). Does awareness of speech as a sequence of
phones arise spontaneously? Cognition, 7, 323-331.
Morton, 1. (1969). The interaction of information in word recognition. Psychological Review, 76, 165-
178.
Nagarajan, S., Mahncke, H., Saltz, T., Tallal, P., Roberts, T., & Merzenich, M. M. (1999). Cortical
auditory signal processing in poor readers. Proceedings of the National Academy of Sciences (USA),
96, 6483-6488.
Newsome, W. T., Britten, K H., & Movshon, 1. A. (1989). Neuronal correlates of a perceptual decision.
Nature, 341, 52-54.
Newsome, W. T., & Pare, E. B. (1988). A selective impairment of motion processing following lesions of
the middle temporal visual area (MT). Journal ofNeuroscience, 8, 2201-2211.
Olson, R. K, Forsberg, H., & Wise, B. (1994). Genes, environment, and the development of orthographic
skills. In V. Berninger (Ed.), The varieties of orthographic knowledge 1: Theoretical and
development issues (pp. 27-71). Dordrecht, The Netherlands: Kluwer Academic
Olson, R., Forsberg, H., Wise, B., & Rack, 1. (1994). Measurement of word recognition, orthographic,
and phonological skills. In G. R. Lyon (Ed.). Frames of reference for the assessment of learning
disabilities: New views on measurement issues (pp. 243-277). Baltimore: Paul H. Brookes.
Olson, R. K, Wise, B., Conners, F., Rack, 1., & Fulker, D. (1989). Specific deficits in component reading
and language skills: genetic and environmental influences. Journal of Learning Disabilities, 22, 339-
348.
Paulesu, W., Frith, U., Snowling, M., Gallagher, A., Morton, J., Frackowiak, R. S. J., & Frith, C. D.
(1996). Is developmental dyslexia a disconnection syndrome? Evidence from PET scanning. Brain,
119, 143-157.
Pennington, B. F., Van Orden, G. c., Smith, S. D., Green, P. A., & Haith, M. M. (1990). Phonological
processing skills and deficits in adult dyslexics. Child Development, 61, 1753-1778.
238 1. T ALCOTI & C. WITION
Rae, c., Lee, M. A., Dixon, R. M., Blamire, A. M., Thompson, C. H., Styles, P., Talcott, 1., Richardson,
A. J., & Stein, J. F. (1998). Metabolic abnormalities in developmental dyslexia detected by 'H
magnetic resonance spectroscopy. The Lancet, 351, 1840-1852.
Raymond, J. E., & Sorenson, R. (1998). Visual motion perception in children with dyslexia: Normal
detection but abnormal integration. Visual Cognition, 5, 389-404.
Rauschecker,1. P. (1998). Cortical processing of complex sounds. Current Opinion in Neurobiology, 8,
516-521.
Remez, R. E., Rubin, P. E., Pisoni, D. B., & Carrell, T. D. (1981). Speech perception without traditional
speech cues. Science, 212, 947-950.
Ridder III, W. H., Borsting, E., Cooper, M., McNeel, B., & Huang, E. (1997). Not all dyslexics are
created equal. Optometry and Vision Science, 74, 99-104.
Rosen, S. (1992). Temporal information in speech: acoustic, auditory and linguistic aspects.
Philosophical Transactions of the Royal Society of London Series B, 336, 367-373.
Rosen, S. (1999). Language disorders: A problem with auditory processing? Current Biology, 9, R698-
R7oo.
Rosner, B. S., Talcott, J. B., Witton, C., Hogg, 1. D., Richardson, A. 1., Hansen, P. c., & Stein, J. F. (in
press). The perception of "sinewave speech" by adult developmental dyslexics.
Samar, V. J., Parasnis, I., & Berent, G. P. (1999). Deaf poor readers' pattern reversal VEPs reveal
magnocellular system deficits. Poster session presented at the annual meeting of the American
Psychological Society, Denver, CO.
Schorer, E. (1986). Critical modulation frequency based on the detection of AM versus FM tones. Journal
of the Acoustical Society of America, 79, 1054-1057.
Schouten, J. F. (1940). The residue and the mechanism of hearing. Proceedings of the Koninklijke
Nederlandse Akademie Van Wetenschappen, 43, 991-999.
Seidenberg, M., & McClelland, J. (1989). A distributed developmental model of word recognition and
naming. Psychological. Review, 96, 523-568.
Shankweiler, D., Liberman, I., Mark, L., Fowler, c., & Fischer, F. (1979). The speech code and learning
to read. Journal of Experimental Psychology: Human Learning and Memory, 5, 531-545.
Shapley, R. (1990). Visual sensitivity and parallel retinocortical channels. Annual Review of Psychology,
41, 635-658.
Share, D., Jorm, A., Maclean, R., & Matthews, R. (1984). Sources of individual differences in reading
acquisition. Journal of Educational Psychology, 76, 1309-1324.
Shaywitz, S. E., Escobar, M. D., Shaywitz, B. A., Fletcher, J. M., & Makuch, R. (1992). Evidence that
dyslexia may represent the lower tail of a normal distribution of reading ability. The New England
Journal of Medicine, 326, 145-150.
Shaywitz, S. E., Shaywitz, B. A., Fletcher, 1. M., & Escobar, M. D. (1990). Prevalence of reading
disability in boys and girls: Results of the Connecticut longitudinal study. Journal of the American
Medical Association, 264, 998-1002.
Skottun, B. (2000). The magnocellular deficit theory of dyslexia: the evidence from contrast sensitivity.
Vision Research, 40, 111-127.
Skottun, B. c., & Parke, L. A. (1999). The possible relationship between visual deficits and dyslexia:
Examination of a critical assumption. Journal of Learning Disabilities, 32, 2-5.
Slaghuis, W., & Lovegrove, W. 1. (1985). Spatial-frequency mediated visible persistence and specific
reading disability. Brain and Cognition, 4, 219-240.
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., & West, R. F. (1989). Exposure to print and orthographic processing. Reading Research
Quarterly, 24, 402-433.
Stanovich, K. E., West, R. F., & Cunningham, A. E. (1991). Beyond phonological processes: Print
exposure and orthographic processing. In S. A. Brady, & D. P. Shankweiler (Eds.), Phonological
processes in literacy: A tribute to Isabelle Y. Liberman (pp.219-235). Hillsdale, NJ: Lawrence
Erlbaum Press
Stefanatos, G. (1993). Frequency modulation analysis in children with Landau-Kleffner syndrome. In P.
Tallal, A. Galaburda, R. Llinas, & c. Von Euler (Eds.), Temporal Information processing in the
Nervous system: Special reference to dyslexia and dysphasia (pp. 412-414). New York: New York
Academy of Sciences.
Stefanatos, G., Green, G., & Ratcliffe, G. (1989). Neurophysiological evidence of auditory channel
anomalies in developmental dysphasia. Archives of Neurology, 46, 871-875.
Stein, J. F., Richardson, A J., & Fowler, M. S. (2000). Monocular occlusion can improve binocular
control and reading in dyslexics. Brain, 123, 164-170.
Stein, J., Talcott, J., & Walsh, V. (2000). Controversy about the evidence for a visual magnocellular
deficit in developmental dyslexics. Trends in Cognitive Sciences, 4, 209-211.
Stein, J., & Walsh, V. (1997). To see but not to read; the magnocellular theory of dyslexia. Trends in
Neurosciences,20,147-152.
Steinman, S. B., Steinman, B. A, & Garzia, R. P. (1998). Vision and attention. II: Is visual attention a
mechanism through which a deficient magnocellular pathway might cause reading disability?
Optometry and Vision Science, 75, 674-681.
Steinman, B.A., Steinman, S.B., & Lehmkuhle, S. (1997). Transient visual attention is dominated by the
magnocellular stream. Vision Research, 37, 17-23.
Studdert-Kennedy, M., & Mody, M. (1995). Auditory temporal perception deficits in the reading-
impaired: a critical review of the evidence. Psychonomic Bulletin & Review, 2, 508-514.
Talcott, J.B., Gram, A, Van Ingelghem, M., Witton, C., Stein, J.F. & Tonnessen, F. (in press). Impaired
sensitivity to dynamic stimuli in poor readers of a transparent orthography.
Talcott, J. B., Hansen, P. c., Willis-Owen, C, McKinnell, I. W., Richardson, A J., & Stein, J. F. (1998).
Visual magnocellular impairment in adult developmental dyslexics. Neuro-Ophthalmology, 20, 187-
201.
Talcott, J. B., Witton, c., Mclean, M. F., Hansen, P. c., Rees, A, Green, G. G. R., & Stein, J. F. (2000).
Dynamic sensory sensitivity and children's word decoding skills. Proceedings of the National
Academy of Sciences (USA), 97, 2952-2957.
Talcott, J. B., Hansen, P. C, Assoku, E. L., & Stein, J. F. (2000). Visual motion sensitivity in dyslexia:
Evidence for temporal and energy integration deficits. Neuropsychologia, 38, 935-943.
Tallal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in children. Brain and
Language, 9, 182-198.
Tallal, P., & Piercy, M. (1973). Defects of non-verbal auditory perception in children with developmental
aphasia. Nature, 241,468-469.
Tallal, P., & Piercy, M. (1974). Developmental aphasia: Rate of auditory processing and selective
impairment of consonant perception. Neuropsychologia, 12, 83-94.
Tallal, P., Merzenich, M. M, Miller, S., & Jenkins, W. (1998). Language learning impairments:
Integrating basic science, technology, and remediation. Experimental Brain Research, 123, 210-219
Tangel, D. M., & Blachman, B. (1995). Effect of phoneme awareness instruction on the invented spelling
of first-grade children: a one-year follow-up. Journal of Reading Behavior, 27, 153-185.
Treiman, R., & Zukowski, A (1991). Levels of phonological awareness. In SA Brady, & D. P.
Shankweiler (Eds.), Phonological processes in literacy: A tribute to Isabelle Y. Liberman (pp.67-83).
Hillsdale, N1: Erlbaum.
van Ijzendoorn, M. H., & Bus, A G. (1994). Meta-analytic confirmation of the nonword reading deficit in
developmental dyslexia. Reading Research Quarterly, 29, 267-275.
Van Orden, G. C., & Goldinger, S. D. (1996). Phonologic mediation in skilled and dyslexic reading. In C.
H. Chase, G. D. Rosen, & G. F. Sherman (Eds.), Developmental dyslexia: neural, cognitive and
genetic mechanisms (pp. 185-223). Baltimore, MD: York Press.
Van Orden, G. c., Pennington, B. F., & Stone, G. O. (1990). Word identification in reading and the
promise of subsymbolic psycholinguistics. Psychological Review, 97, 488-522.
Van Tassel, D. 1., Soli, S. D., Kirby, V. M., & Widin, G. P. (1987). Speech waveform envelope cues for
conson;mt recognition. Journal of the Acoustical Society ofAmerica, 82, 1152-1161.
Vellutino, F. (1979). Dyslexia: Research and Theory. Cambridge, MA: MIT Press.
Vidyasagar, T. R., & Pammer, K. (1999). Impaired visual search in dyslexia relates to the role of the
magnocellular pathway in attention. Neuroreport, 10, 1283-1287.
Wagner, R. K., & Torgeson, J. K. (1987). The nature of phonological processing and its causal role in the
acquisition of reading skills. Psychological Bulletin, 101, 192-212.
Wagner, R. K., Torgeson, J. K., & Rashotte, C. A (1994). Development of reading-related phonological
processing abilities: New evidence of bidirectional causality from a latent variable longitudinal study.
Walther-Miiller, P. U. (1995). Is there a deficit of early vision in dyslexia? Perception, 24, 919-936.
Watson, C. S., & Foyle, D. C. (1985). Central factors in the discrimination and identification of complex
sounds. Journal of the Acoustical Society of America, 78 (1), 375-380.
Watson, C. S., & Kelly, W. J. (1981). The role of stimulus uncertainty in the discrimination of auditory
patterns. In D. J. Getty, & 1. H. Howard 1r. (Eds.), Auditory and visual pattern recognition (pp. 37-
59). Hillsdale, N.J.: Erlbaum.
240 1. TALCOTT & C. WITTON
Wattam·Bell, 1. (1994). Coherence thresholds for discrimination of motion direction in infants. Vision
Research, 34, 877-883.
Williams, M. c., Brannan, 1. R., & Lartigue, E. K. (1987). Visual search in good and poor readers.
Clinical Vision Sciences, 1, 367-371.
Williams, M. C., May, 1. G., Solman, R., & Zhou, H. (1995). The effects of spatial filtering and contrast
reduction on visual search times in good and poor readers. Vision Research, 35, 285-291.
Witton, C., Stein, 1. F., Stoodley, C. 1., Rosner, B. S. & Talcott, J. B. (in press).Separate influences of
acoustic AM and FM sensitivity on the phonological decoding skills of impaired and normal readers.
Journal of Cognitive Neuroscience.
Witton, c., Talcott, 1. B., Hansen, P. c., Richardson, A J., Griffiths, T. D., Rees, A, Stein, 1. F., &
Green, G. G. R. (1998). Sensitivity to dynamic auditory and visual stimuli predicts nonword reading
ability in both dyslexic and normal readers. Current Biology, 8, 791-797.
Wood, C. & Terrell, C. (1998). Poor readers' ability to detect speech rhythm and perceive rapid speech.
British Journal of Developmental Psychology, 16, 397-413.
Zihl, 1., von Cramon, D., & Mai, A (1983). Selective disturbance of movement vision after bilateral brain
damage. Brain, 106, 313-340.
Zwicher, E. (1952). Die Grenzen der Hiirbarkeit der Amplitudenmodulation und der Frequenzmodulation
eines Tones. Acustica, 2, 125-133.
A.M.GALABURDA
ANATOMY OF THE TEMPORAL PROCESSING

DEFICIT IN DEVELOPMENTAL DYSLEXIA
Abstract. The acquisition of reading competence requires important environmental input as well as the
brain machinery to process a variety of types of information at multiple levels. The development of the
architecture of this machinery is under the control of multiple genetic and epigenetic factors. It is not
surprising, therefore, that developmental reading disorders could arise from dysfunction in many areas
and at many levels. However, it appears that many children who fail to acquire reading competence easily
and fully suffer from a specific type of developmental brain anomaly that can affect linguistic and low
level auditory and/or visual processing. The present chapter will examine the effects of this anomaly on
brain anatomy and behavior and will look at the perceptual and cognitive consequences of minor cortical
malformations in the visual and auditory systems, as they relate to reading acquisition. Furthermore, the
developmental mechanisms associated with these malformations will be described in the context of other
developmental disorders. Finally, etiologiC factors leading to the formation of these malformations will be
reviewed, including genetic and epigenetic influences acting during the period of neuronal migration to
the cerebral cortex. The chapter will take examples from human autopsy studies as well as from studies
involving experimental animal models.
1. INTRODUCfION
There is a great deal of support now for the claim that dyslexics have trouble
processing rapidly changing sensory experience (see review by Farmer & Klein,
1995; Stein & Walsh, 1997). This is true for the auditory system, as it is for the
visual system. Rapidly changing tones, for instance, are not processed normally in
dyslexics and language impaired children (reviewed by Fitch, Miller, & Tallal,
1997). Rapidly switching checkerboard patterns at low contrasts elicit abnormal
visual evoked potentials in dyslexics (Livingstone, Rosen, Drislane, & Galaburda,
1991). There are other examples of psychophysical deficits in dyslexics affecting the
visual system (Martin, & Lovegrove, 1987; Slaghuis, Twell, & Kingston, 1996), and
dysfunction can be demonstrated using functional magnetic resonance imaging
(Eden, et al. 1996; Demb, Boynton, & Heeger, 1998). There may be involvement of
manual dexterity too (Rousselle & Wolff, 1991) and slowness of automatized
naming (Denckla & Rudel, 1976; Wolf & Obregon, 1992) implicating the motor
system. Rapid tapping, for instance, is not smooth in some dyslexics. How specific
these deficits are to dyslexics is not known, but it is likely that they are not, as it has
long been known that brain injury causing aphasia in adults can slow down sensory
processing (Efron, 1963). It is possible that non-specific slowness of processing is
one of the outcomes of brain injury irrespective of age and cause.
It is even less clear what the relationship between the sensory-motor processing
deficits and the reading disorder in dyslexics is, as there is experimental evidence
indicating that only speech perception problems relate to the reading disorder and
E. Witruk, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions ofLanguage, Reading,

@2oo2 Kluwer Academic Publishers.
242 A. M. GALABURDA
not temporal processing of non-linguistic sounds (Studdert-Kennedy & Mody, 1995;

Mody, see also Degelder & Vroomen, 1996; Studdert-Kennedy, & Brady, 1997).
Our work has consisted in modeling developmental cortical anomalies in
laboratory rodents for the purpose of shedding light on the problem of
developmental dyslexia. Dyslexia is a disorder of reading that commonly affects
language (Mody, et aI., 1997; Shankweiler, et aI., 1995; Vellutino, 1987; Vellutino,
Scanlon, & Spearing, 1995), and, although there is great deal of interest in language
precursors in non-human animals (Gannon, Holloway, Broadfield, & Braun, 1998;
Hauser, 1997; Ramus, Hauser, Miller, Morris, & Mehler, in press), this is not the
reason we are studying mice and rats. Instead, because it is found to be abnormal in
dyslexics, we are focusing on the animal's inability to process rapidly changing
sounds, whether or not this includes processing of language sounds. Another related
and important question that remains to be solved is how separate and unique is the
representation of language and other types of sounds in the early stages of auditory
processing, including the thalamus and primary auditory cortex. There is some
magnetoencephalographic evidence to show that as early as the primary auditory
cortex there is specialized representation of phonemes (Poeppel, et aI., 1997). If
there is a distinct separation early on, it is not unreasonable to ask how early in
phylogenesis can this separation be demonstrated. We do not suppose it would be
found in rodents (but, cf., Miller & Kuhl, 1975; Ohlemiller, Jones, Heidbreder,
Clark, & Miller, 1999).
In the laboratory rat and mouse we have modeled a neuronal migration anomaly
we have called "ectopia" (Galaburda & Kemper, 1979; Galaburda, Sherman, Rosen,
Aboitiz, & Geschwind, 1985; Sherman, Galaburda, & Geschwind, 1985), also
known as molecular layer heterotopia, brain wart, glioneuronal heterotopia. In the
mouse we currently mostly study the genetics of this disorder, which is not this
chapter's concern, and will not be considered further. In the rat we look at the
anatomical organization of the ectopia and its physiological and behavioral
consequences.
Our interest in developing these models sprung from the finding that ectopias are
found in large numbers in the brains of persons with developmental dyslexia
(Galaburda, et aI., 1985). In eight male dyslexic brains we found these ectopias in
perisylvian cortex, affecting frontal opercular areas, superior and middle temporal
gyri, and parietal operculum. In most of the specimens, there were more ectopias in
the left than the right hemisphere. In two female brains we did not find ectopias.
Instead we found small prenatal gliotic scars of roughly the same size as the ectopias
and in roughly the same locations (Galaburda & Humphreys, 1989). This led us to
postulate that they were indeed the product of damage in both cases, but affecting a
more mature brain in the female case.
We assumed that ectopias had something to do with the language behavior
displayed by dyslexics, because, other than the excessive tendency for the planum
temporale to be symmetric in the dyslexic brains, we found no other anomalies. The
planum temporale is not found in rodents, and it is in fact barely discernible in
chimpanzees (Gannon, et aI., 1998), making this an impossible structure to model
further in our experimental model. Thus, we were limited to studying ectopias in
rodents, because, except to say that the ectopias probably scrambled cortical areas
involved in language, we could not get any closer to the neural mechanism
associated with the ectopias. The rat, and in a few cases the mouse, offered the
ANATOMY OF THE TEMPORAL PROCESSING DEFICIT 243
opportunity to learn more about the anatomical, physiological, and behavioral

effects ectopias might have on neural networks.
2. ANATOMICAL AND FUNCTIONAL FEATURES OF

MALFORMATIONS
Ectopias consist of 50-100 neurons and glia nested quite compactly in the molecular
layer of the neocortex. Normally, the molecular layer, the most superficial cortical
layer, layer I, does not contain agglomerations of neurons, but rather an occasional
isolated neuron belonging to an early developmental type - the Cajal-Retzius cell.
Ectopias belong to a class of developmental anomalies known as neuronal migration
anomalies, usually meaning that neurons have migrated to inappropriate sites within
the cortex or subcortical white matter. Neurons are born after neuroblast
proliferation in the germinal zones adjacent to the fetal ventricles, after which they
migrate either to the cortex or subcortical gray masses to find their final locations,
where they differentiate and establish connectivity with other neurons nearby or far
away. Most neurons migrate along radially disposed glial fibers strung between the
germinal zones and the molecular layer (Rakic, 1972), but some migrate tangentially
and end up at fair distances from their birth sites (Walsh & Cepko, 1988). Many of
the neuronal migration anomalies known, including the molecular layer ectopias,
reflect abnormal migration along the radial dimension.
Ectopias appear in the cortex soon after young neurons born in the germinal
zones begin to migrate to the cortex to find their laminar positions. In the mouse, as
early as embryonic day 13 ectopias can be found (Rosen, Sherman, & Galaburda,
1996). In humans, this would correspond roughly to about 16 weeks of gestation
(Sidman & Rakic, 1973). Analysis of the neuronal types in the ectopias using Golgi
stains and radioactive dating methods demonstrates neurons of different birth dates,
beginning with the earliest born neurons destined to populate layer VI and ending
with neurons destined for layer II (Rosen, Press, Sherman, & Galaburda, 1992).
These neurons escape into the molecular layer through a breach in the external glial
limiting membrane (Sherman, Rosen, Stone, Press, & Galaburda, 1992), which
ordinarily serves as a barrier to keep neurons from entering this layer. We assume
that the breach is there early because of the presence of multi-aged neurons and
because ectopias are seen early in corticogenesis. The cause of the breach is not
known either for the dyslexic human or for the mutant mouse with ectopias, but it
can be produced experimentally by ischemic injury to the surface of the cortex
(Humphreys, Rosen, Press, Sherman, & Galaburda, 1991) or by poking a small hole
in the membrane with a needle before the end of neuronal migration (Rosen,
Sherman, Richman, Stone, & Galaburda, 1992).
Whereas ectopias are produced by mild injury to the developing cortex, focal
microgyria occurs when the injury is more severe (Rosen, & Galaburda, in press).
Microgyria disturbs the organization of all the layers including and underlying the
molecular layer. Ectopias and microgyria can occur together when adjacent areas of
the cortex are exposed to different degrees of injury. Focal migrogyria has also been
seen in the brains of dyslexics (Galaburda & Kemper, 1979; Galaburda, et al., 1985).
Microgyria is easier to induce in the rats, so we have studied them in more detail,
although many of the findings in microgyria apply to ectopias too.
244 A. M. GALABURDA
Neurons in the microgyria are mainly excitatory (LoTurco, J. J., personal

communication). Initially there is a complete paucity of GABAergic neurons
(Rosen, Jacobs, & Prince, 1998), and excessive excitability is demonstrated
physiologically in and near the microgyria (Jacobs, Hwang, & Prince, 1999).
Placement of axonal tracers in microgyria or ectopias shows that the abnormal
region is connected both locally and widely (Jenner, Galaburda, & Sherman, in
press; Rosen, Burstein, & Galaburda, 2000). Barrel field microgyria, for instance,
send projections to the ventrobasal complex of the thalamus and diffusely to the
contralateral hemisphere and receive reciprocal connections. There may be
inappropriate connections, too, because some developmentally transient connections
are not eliminated or as a result of abnormal sprouting. The diffuse nature of the
contralateral projections, for instance, may be an example of lack of withdrawal of
transient connections. This pattern of organization suggests that ectopias are in a
position to affect the development of areas to which they connect, and the types of
connections suggest that the influence is possibly through the mechanism of
excitotoxicity and with functional consequences.
We found that induction of microgyria in the frontal lobe, some synapses away
from the thalamus, causes histological changes in the medial geniculate nucleus
(MGN) (Herman, Galaburda, Fitch, Carter, & Rosen, 1997). Animals with
microgyria, as compared with animals with sham injury, showed an excess of small
neurons in the MGN. Dyslexic brains, too, have an excess of small neurons in the
MGN, in this case only the left MGN (Galaburda, Menard, & Rosen, 1994). As, in
general, smaller neurons are slower (Lawson & Waddell, 1991; Sakai, & Woody,
1988), we sought to demonstrate whether animals with microgyria exhibited slowing
of auditory processing.
Microgyria and molecular layer ectopias slow auditory processing speed in the
rat, and this can be demonstrated in two different ways. Rats were taught to
discriminate between two tones presented sequentially, and the inter-stimulus
separation was varied. Animals with microgyria could not discriminate the tones at
shorter inter-stimulus separations (Fitch, Tallal, Brown, Galaburda, & Rosen, 1994).
In another experiment, mice with ectopias were implanted with electrodes and tested
on an oddball paradigm using tones. Short inter-stimulus distances failed to produce
mismatch negativity in animals with ectopias (Frenkel, Sherman, Bashan,
Galaburda, & LoTurco, 2000). These findings indicate that ectopias interfere with
rapid auditory processing for tones. They do not, however, implicate a specific stage
of processing, whether in the thalamus or the cortex.
We observed sexual dimorphism in the phenomena described above (Rosen,
Herman, & Galaburda, 1999). Female rats with microgyria did not show any deficits
in rapid auditory processing once examined separately from the males. We
investigated the microgyria in the females and found them to be identical to those of
the males, both in size and location. On the other hand, examination of the thalamus
did not show changes in cell size in the females as it did in the males. This showed
that the microgyria themselves were not directly responsible for the behavioral
deficit, but rather the changes in the MGN, which it stimulated in the males but not
in the females. Moreover, the difference could help account for the male-biased sex
ratio in dyslexia. Furthermore, this finding helped to causally link the slowing of
sound processing to the earliest of the forebrain processors for sound - the MGN of
the thalamus. Additional experiments indicated that the sex difference was caused
by sex hormones (Rosen, et aI., 1999). Thus, exposure of fetal female rats to male
sex steroids changed their pattern of response to microgyria to the male pattern.
3. DISCUSSION
The animal model might explain why dyslexics have problems with processing
rapidly changing sounds. What the research does not address is whether cortical
malformations like those in the dyslexic brain and in the laboratory animals are
capable of also changing the processing of speech sounds. Current research is
looking at the animals' ability to make phonetic distinctions based on voice onset
time or place of articulation. Although these results will be important to gather, they
may not address the fundamental question, which is whether humans have special
systems for speech sounds as compared to other sounds. If rats with malformations
fail at these phonological tasks, it may simply mean that they use one system for all.
However, if they succeed where they fail with non-speech sounds, assuming
equivalent physical demands, the possibility exists that even in rodents speech
sounds are processed separately, an unlikely case (but, cf., Miller & Kuhl, 1975;
Ohlemiller, et aI., 1999).
There are other reasons for which dyslexics may fail at speech sounds separately
from non-speech sounds, and this has to do with other anatomical changes
associated with the ectopias and microgyria. As mentioned in the introduction,
dyslexic brains show symmetry of the planum temporale (Beaton, 1997; Galaburda,
et aI., 1985; Larsen, Hoien, Lundberg, & Odegaard, 1990; Shapleske, Rossell,
Woodruff, & David, 1999), even though on imaging studies there is no uniform
agreement as to what comprises the planum temporale (Rumsey, et aI., 1997). Such
symmetry is found also in 2/3 of unselected autopsy brains (Geschwind & Levitsky,
1968), but all dyslexic brains analyzed thus far were symmetric in this region. We
looked at the relationship between symmetry and the presence of ectopias. We did
not find that symmetry is more common in the presence of ectopias, but we did find
that the nature of symmetry changes from the normal pattern (Rosen, Sherman,
Mehler, Emsbo, & Galaburda, 1989). Symmetric cortical areas are larger in total
than asymmetric cortical areas. In the asymmetric cases one of the sides is smaller
than a symmetrical side rather than larger, as if the asymmetrical case is a
unilaterally curtailed form of a symmetrical case (Galburda, Corsiglia, Rosen, &
Sherman, 1987). There is an inverse linear relationship between total amount of
cortex (left plus right) and the coefficient of asymmetry of a cortical area. This is
seen in human and animal brains where degree of asymmetry varies from individual
to individual (Galaburda, Aboitiz, Rosen, & Sherman, 1986; Galaburda, et aI.,
1987). When ectopias are present, however, this relationship breaks down and no
relationship between asymmetry and symmetry vis a vis total area emerges (Rosen,
et aI., 1989). Symmetry and asymmetry differ in the number of neurons (Rosen,
Sherman, & Galaburda, 1991) and in the pattern of callosal connections (Rosen,
Sherman, & Galaburda, 1989). There are more neurons in total in the symmetrical
areas, and more callosal connections. It is therefore possible that areas in the planum
temporale having to do with phonological representation and processing are
rendered anomalous by alterations associated with ectopias, although the exact
mechanism for this is unknown.
246 A. M. GALABURDA
Another possibility for involvement of phonological systems directly in

dyslexics may have to do with the local effects of ectopias and microgyria,
separately from their effects on the thalamus and asymmetry. We find alterations in
the numbers of GABAergic neurons (Rosen, et aI., 1998) and neuronal excitability
in cortex adjacent to microgyria (Jacobs, et aI., 1999). Frontal opercular involvement
by malformations is perhaps the most common finding in the dyslexic brains,
leading to the supposition that infero-Iateral frontal functional capacities are
changed. The infero-Iateral frontal region has been implicated in phonological
processing (Fiez, et aI., 1995; Fujimaki, et aI., 1999; Poldrack, et aI., 1999). One
aspect of phonological processing is critically abnormal in dyslexia - phonological
awareness (Shankweiler, & Crain, 1986) (cf., Cossu, Rossini, & Marshall, 1993;
Wimmer, Landerl, Linortner, & Hummer, 1991). Phonological awareness is required
for learning to read and for reading unknown words and pseudowords. It is
reasonable to propose that it is the frontal lobe that participates in phonological
awareness (Breier, Simos, Zouridakis, & Papanicolaou, 1999) It is further
reasonable to suppose that malformation in this region in dyslexics may be behind
the difficulty with phonological awareness.
A problem arises when examining this type of brain evidence. Injury to the
cortex early in development is accompanied by reaction close by and far afield.
Microgyria in the frontal cortex produces changes in neuronal sizes in the thalamus,
and probably in all intervening processing stations along the way. Changes may not
stop at the thalamus and may continue instead to stations in the brainstem. We
suspect that changes occur downstream, too, affecting cortical areas further along in
the processing stream. In fact, a whole pathway may be affected by the ectopia. We
have suggested that some of the changes - namely those in the thalamus - could
account for sound processing abnormalities, while others - those in the frontal lobe
- could explain problems with phonological awareness. However, it is not clear why
would not all aspects of speech processing be affected. So, for instance, this
corruption in the total network does not explain normal speech production and
speech comprehension, which is by and large the case in dyslexia. The female case
may be enlightening in this regard. It shows that malformation in the cortex is not
always associated with plastic changes in connectionally related areas. At least, if
there is plasticity, it is not of the type that can be measured in cell sizes and does not
affect temporal processing. It is therefore possible that phonological representations
and the areas that accommodate them may remain intact, or may react adaptively to
the initial event and remain functional.
One of the characteristics of processing modules is that they degrade gracefully.
Developmental injury may lead to an exception to this rule. Connectivity is
demonstrably anomalous after cortical malformation. Units that normally do not
connect in adult brains are poised to connect in the event of abnormal influences
coming to bear upon them. Thus, somatosensory cortices are functionally connected
to the visual system in congenitally blind subjects who read Braille (Sadato,
Pascual-Leone, Grafman, Deiber, & Ibanez, 1998). Cortical ectopias and microgyria
lead to alterations in the patterns of connections within and between the
hemispheres. Could this result in the elimination of normal functional boundaries
between modular systems? Alternatively, systems having separate genetic
backgrounds, which may apply to modules, may never interconnect, irrespective of
cause or timing of brain damage. This question cannot be answered at present. If
intermodule connectivity is indeed possible, then it is also possible that in some

abnormal developmental states, systems that remain isolated in the normal state,
become intertwined with resultant degradation of multiple functions.
Neuroscience may help resolve some of these questions in ways not possible
only a decade ago. It is possible, therefore, to map function onto selected areas of
the cortex using instruments such as MEG and fMRI. The functionality of these
areas of cortex identified by these tools can be checked with transcranial magnetic
stimulation. Some, albeit not enough, of these approaches are now available for the
study of children and infants. We need to know for certain whether phonemes and
other sounds are mapped separately in the auditory cortex and whether the maps can
be changed one independently from the other. We could perhaps learn about whether
dyslexics follow the rules vis a
vis mapping sounds and speech early in the
processing pathway. It may even be possible to find out whether behavioral
improvement in dyslexics through the use of a given therapeutic approach exerts any
change, even normalization, of the maps. The question in dyslexia is still primarily
about sounds and speech, and its resolution will help further define theories
regarding the uniqueness of language in the brain.
4. AFFILIATIONS
Dr. Albert M. Galaburda

Emily Fisher Landau Professor of Neurology and Neuroscience
Harvard Medical School
Beth Israel Deaconess Medical Center
330 Brookline A venue
Boston, MA 02215
E-mail: agalabur@caregroup.harvard.edu
5. REFERENCES
Beaton, A. A. (1997). The relationship of the planum temporale asymmetry and morphology of the corpus
callosum to handedness, gender, and dyslexia: a review of the evidence. Brain and Language, 60,
255-322.
Breier, J., Simos, P. G., Zouridakis, G., & Papanicolaou, A. C. (1999). Temporal course of regional brain
activation associated with phonological decoding. Journal of Clinical and Experimental
Cossu, G., Rossini, F., & Marshall, J. C. (1993). When reading is acquired but phonemic awareness is
not: a study of literacy in Down's syndrome. Cognition, 46, 129-138.
Degelder, B., & Vroomen, J. (19%). Auditory illusions as evidence for a role of the syllable in adult
developmental dyslexics. Brain Language, 52, 373-385.
Demb, J. B., Boynton, G. M., & Heeger, D. J. (1998). Functional magnetic resonance imaging of early
visual pathways in dyslexia. Journal ofNeuroscience, 18, 6939-51.
Denckla, M. B., & Rudel, R. G. (1976). Rapid "automatiied" naming (RAN.): Dyslexia differentiated
from other learning disabilities. Neuropsychologia, 14, 471-479.
Eden, G. F., Vanmeter, J. W., Rumsey, J. M., Maisog, J. M., Woods, R. P., & Zeffiro, T. A. (19%).
Abnormal processing of visual motion in dyslexia revealed by functional brain imaging. Nature, 382,
66-69.
Efron, R. (1%3). Temporal perception, aphasia, and deja vu. Brain, 36,403-424.
A review. Psychonomic Bulletin and Review, 2, 4, 460-493.
248 A. M. GALABURDA
Fiez, J. A., Raichle, M. E., Miezin, F. M., Petersen, S. E., Tallal, P., & Katz, W. F. (1995). PET studies of
auditory and phonological processing: effects of stimulus characteristics and task demands. Journal
of Cognitive Neuroscience, 7,357-375.
Fitch, R. H., Miller, S., & Tallal, P. (1997). Neurobiology of speech perception. Annual Review
Fitch, R. H., Tallal, P., Brown, C., Galaburda, A. M., & Rosen, G. D. (1994). Induced microgyria and
auditory temporal processing in rats: A model for language impairment? Cerebral Cortex, 4, 260-
270.
Frenkel, M., Sherman, G. F., Bashan, K. A., Galaburda, A. M., & LoTurco, 1. 1. (2000). Neocortical
ectopias are associated with attenuated neurophysiological responses to rapidly changing auditory
stimuli. Neuroreport, 11, 575-579.
Fujimaki, N., Miyauchi, S., Putz, B., Y., S., Takino, R., Sakai, K., & Tamada, T. (1999). Functional
magnetic resonance imaging of neural activity related to orthographic, phonological, and lexico-
semantic judgments of visually presented characters and words. Human Brain Mapping, 8, 44-59.
Galaburda, A. M., Aboitiz, F., Rosen, G. D., & Sherman, G. F. (1986). Histological asymmetry in the
primary visual cortex of the rat: Implications for mechanisms of cerebral asymmetry. Cortex, 22,
151-160.
Galaburda, A. M., Corsiglia, J., Rosen, G. D., & Sherman, G. F. (1987). Planum temporale asymmetry:
Reappraisal since Geschwind and Levitsky. Neuropsychologia, 25, 853-868.
Galaburda, A. M., & Humphreys, P. (1989). Developmental dyslexia in women: Neuropathological
findings in two cases. Neurology, 39, Suppl. 1,317.
Galaburda, A. M., & Kemper, T. L. (1979). Cytoarchitectonic abnormalities in developmental dyslexia; a
case study. Annals of Neurology, 6, 94-100.
developmental dyslexia. Proc. NatL. Acad. Sci. USA, 91, 8010-8013.
Galaburda, A. M., Sherman, G. F., Rosen, G. D., Aboitiz, F., & Geschwind, N. (1985). Developmental
dyslexia: Four consecutive cases with cortical anomalies. Annals of Neurology, 18, 222-233.
Gannon, P. 1., Holloway, R. L., Broadfield, D. c., & Braun, A. R. (1998). Asymmetry of chimpanzee
planum temporale: humanlike pattern of Wernicke's brain language area homolog. Science, 279, 220-
222.
Geschwind, N., & Levitsky, W. (1968). Human brain: Left-right asymmetries in temporal speech region.
Science, 161, 186-187.
Hauser, M. D. (1997). Artifactual kinds and functional design features: what a primate understands
without language. Cognition, 64, 285-308.
Herman, A. E., Galaburda, A. M., Fitch, H. R., Carter, A. R., & Rosen, G. D. (1997). Cerebral
microgyria, thalamic cell size and auditory temporal processing in male and female rats. Cerebral
Cortex, 7, 453464.
Humphreys, P., Rosen, G. D., Press, D. M., Sherman, G. F., & Galaburda, A. M. (1991). Freezing lesions
of the newborn rat brain: A model for cerebrocortical microgyria. Journal of Neuropathology and
Experimental Neurology, 50, 145-160.
Jacobs, K. M., Hwang, B. 1., & Prince, D. A. (1999). Focal epileptogenesis in a rat model of
polymicrogyria. Journal of Neurophysiology, 81, 159-73.
Jenner, A. R., Galaburda, A. M., & Sherman, G. F. (In press). Connectivity of ectopic neurons in the
molecular layer of the somatosensory cortex in autoimmune mice. Cerebral Cortex
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-
30l.
Lawson, S. N., & WaddelI, P. 1. (1991). Soma neurofilament immunoreactivity is related to cell size and
fibre conduction velocity in rat primary sensory neurons. Journal of Physiology (London), 435, 41-
63.
Livingstone, M., Rosen, G., Drislane, F., & Galaburda, A. (1991). Physiological and anatomical evidence
for a magnocellular defect in developmental dyslexia. Proc. Natl. Acad. Sci. USA,88, 7943-7947.
readers. Perception, 16, 215-22l.
Miller, 1. D., & Kuhl, P. K. (1975). Speech perception by the chinchilla: voiced-voiceless distinction in
alveolar plosive consonants. Science, 190, 69-72.
Auditory processing or phonological coding? Journal of Exp. Child Psychology, 64, 199-23l.
Ohlemiller, K K, Jones, L. B., Heidbreder, A. F., Clark, W. W., & Miller, J. D. (1999). Voicing
judgements by chinchillas trained with a reward paradigm. Behavioral Brain Research, 100, 185-
195.
Poeppel, D., Phillips, C., Yellin, E., Rowley, H. A., Roberts, T. P., & Marantz, A. (1997). Processing of
vowels in supratemporal auditory cortex. Neuroscience Letters, 221, 145-148.
Poldrack, R. A., Wagner, A. D., Prull, M. W., Desmond, J. E., Glover, G. H., & Gabrieli, J. D. (1999).
Functional specialization for semantic and phonological processing in the left inferior prefrontal
cortex. Neuroimage, 10, 15-35.
Rakic, P. (1972). Mode of cell migration to the superficial layers of fetal monkey neocortex. Journal of
Comparative Neurology, 145, 61-84.
Ramus, F., Hauser, M. D., Miller, C., Morris, D., & Mehler, J. (In press). Language discrimination by
human newborns and by cotton-top tamarin monkeys. Science
Rosen, G. D., Burstein, D., & Galaburda, A. M. (2000). Changes in efferent and afferent connectivity in
rats with cerebrocortical microgyria. Journal of Comparative Neurology, 418, 423-440.
Rosen, G. D., & Galaburda, A. M. (In press). Single cause, polymorphic neuronal migration disorder: An
animal model. Developmental Medicine and Child Neurology
Rosen, G. D., Herman, A. E., & Galaburda, A. M. (1999). Sex differences in the effects of early
neocortical injury on neuronal size distribution of the medial geniculate nucleus in the rat are
mediated by perinatal gonadal steroid. Cerebral Cortex, 9, 27-34.
Rosen, G. D., Jacobs, K M., & Prince, D. A. (1998). Effects of neonatal freeze lesions on expression of
parvalbumin in rat neocortex. Cerebral Cortex, 8, 753-61.
Rosen, G. D., Press, D. M., Sherman, G. F., & Galaburda, A. M. (1992a). The development of induced
cerebrocortical microgyria in the rat. Journal of Neuropathalogy and Experimental Neurology, 51,
601-iill.
Rosen, G. D., Sherman, G. F., & Galaburda, A. M. (1989a). Interhemispheric connections differ between
symmetrical and asymmetrical brain regions. Neuroscience, 33, 525-533.
Rosen, G. D., Sherman, G. F., & Galaburda, A. M. (1991). Ontogenesis of neocortical asymmetry: A
[3H]thymidine study. Neuroscience, 41, 779-790.
Rosen, G. D., Sherman, G. F., & Galaburda, A. M. (1996). Birthdates of neurons in induced microgyria.
BrainResearch, 727, 71-78.
Rosen, G. D., Sherman, G. F., Mehler, C., Emsbo, K, & Galaburda, A. M. (1989b). The effect of
developmental neuropathology on neocortical asymmetry in New Zealand Black mice. International
Journal of Neuroscience, 45, 247-254.
Rosen, G. D., Sherman, G. F., Richman, 1. M., Stone, L. V., & Galaburda, A. M. (1992b). Induction of
molecular layer ectopias by puncture wounds in newborn rats and mice. Dev. Brain Res., 67, 285-
291.
Rousselle, c., & Wolff, P. H. (1991). The dynamics of bimanual coordination in developmental dyslexia.
Rumsey, J. M., Donohue, B. c., Brady, D. R., Nace, K, Giedd, J. N., & Andreason, P. (1997). A
magnetic resonance imaging study of planum temporale asymmetry in men with developmental
dyslexia. Archives ofNeurology, 54,1481-1489.
Sadato, N., Pascual-Leone, A., Grafman, 1., Deiber, M. P., & Ibanez, V. (1998). Neural networks for
Braille reading by the blind. Brain, 121, 1213-1229.
Sakai, H., & Woody, C. D. (1988). Relationships between axonal diameter, soma size, and axonal
conduction velocity of HRP-filled, pyramidal tract cells of awake cats. Brain Research, 460, 1-7.
Shankweiler, D., & Oain, S. (1986). Language mechahisms and reading disorder: A modular approach.
Cognition, 24, 139-168.
Shankweiler, D., Crain, S., Katz, L., Fowler, A. E., Liberman, A. M., Brady, S. A., Thornton, R.,
Lundquist, E., Dreyer, L., Fletcher, J. M., Stuebing, K K, Shaywitz, S. E., & Shaywitz, B. A.
(1995). Cognitive profiles of reading-disabled children: Comparison of language skills in phonology,
morphology, and syntax. Psychological Science, 6, 149-156.
Shapleske, 1., Rossell, S. L., Woodruff, P. W., & David, A. S. (1999). The planum temporale: a
systematic, quantitative review of its structural, functional and clinical significance. Brain Research;
Brain Research Reviews, 29, 26-49.
Sherman, G. F., Galaburda, A. M., & Geschwind, N. (1985). Cortical anomalies in brains of New Zealand
mice: A neuropathologic model of dyslexia? Proc. Nat. AcarI. Sci. (USA), 82, 8072-8074.
Sherman, G. F., Rosen, G. D., Stone, L. V., Press, D. M., & Galaburda, A. M. (1992). The organization of
radial glial fibers in spontaneous neocortical ectopias of newborn New-Zealand black mice. Dev.
Brain Res., 67, 279-283.
250 A. M. GALABURDA
Sidman, R. L., & Rakic, P. (1973). Neuronal migration, with special reference to developing human
brain: a review. Brain Research, 62, 1-35.
Siaghuis, W. L., Twell, A. 1., & Kingston, K R. (1996). Visual and language processing disorders are
concurrent in dyslexia and continue into adulthood. Cortex, 32,413-38.
Stein 1., & Walsh V. (1997). To see but not to read; the magnocellular theory of dyslexia. Trends in
Neuroscience, 20, 4, 147-152.
impaired: a critical review of the evidence. Psychonomic Bulletin and Review, 2, 508-514.
Vellutino, F. R. (1987). Dyslexia. Sci. Amer., 256, 34-41.
Vellutino, F. R., Scanlon, D. M., & Spearing, D. (1995). Semantic and phonological coding in poor and
normal readers. Journal of Experimental Child Psychology, 59, 76-123.
Walsh, c., & Cepko, C. (1988). Clonally related cortical cells show several migration patterns. Science,
241, 165-192.
Wimmer, H., LanderJ, K, Linortner, R., & Hummer, P. (1991). The relationship of phonemic awareness
to reading acquisition: more consequence than precondition but still important. Cognition, 40, 219-
249.
Wolf, M., & Obregon, M. (1992). Early naming deficits, developmental dyslexia, and a specific deficit
hypothesis. Brain Language, 42, 219-247.
J. E. JIMENEZ GONZALEZ
READING DISABILITIES IN A LANGUAGE WITH

TRANSPARENT ORTHOGRAPHY
Abstract. The aim of the present chapter is to provide some empirical evidence about underlying
mechanisms that are involved in the explanation of reading disabilities in a language with transparent
orthography (i.e., Spanish). Previously, we analyze some phonological and orthographical specificities of
English and Spanish in order to describe their implications on psycholinguistic processing. Specifically,
our analysis is centered on word recognition processes, therefore we describe some psycholinguistic
parameters which have been used in Spanish studies in order to determine to what extent some reading
models (e.g., dual models) are functional in a transparent orthography. If reading mechanisms are the
same for different alphabetical writing systems, then the pattern of results found in English should be
expected in Spanish as well . Likewise, taking into account that dual-route theory has been used to
explain the mechanisms underlying developmental dyslexics' word reading, we review some research
carried out in Spanish which has been designed to test whether the phenotypic pattern in a transparent
orthography matches that found in English. But before we discuss these results we also review some
problems related to the definition of reading disorders. The definition for developmental reading disorder
relies on a discrepancy between expected and actual achievement. One of the most important tests for the
credibility of and justification for the discrepancy definition of dyslexia is its ability to identify a unique
group of individuals with reading problems that is different from ordinary poor readers who have no
discrepancy. We review some bibliography literature on this topic in English, and we provide additional
empirical evidence from Spanish studies which demonstrate that the IQ discrepancy based definition for
dyslexia should be abandoned in favor of a definition which incorporates the phonological deficit
hypothesis. On the other hand, we also review some studies about the relative influence that different
forms of orthographic units (e.g., syllables, onset-rime, etc.) on the explanation of reading disabilities as a
function of orthographic systems. And finally, we provide empirical data relevant to the issue whether
specific reading disabilities fit a developmental lag or deficit model in a transparent orthography.
1. ALPHABETICAL WRITTEN SYSTEMS AND PSYCHO LINGUISTIC

PROCESSING
There is a great variability in alphabetical systems, that is, the relationship between
written forms and the phonology varies from language to language. If we consider a
continuum in the predictability of grapheme-phoneme relationships, there are at one
extreme the languages with a much higher degree of orthographic transparency
where the mapping between graphemes and phonemes is largely consistent (e.g.,
Serbo-Croatian, Spanish, Finnish, ... ), whereas at the other extreme of continuum are
languages with an opaque orthography, where there are many ways of sounding out
graphemes and many of the correspondences cannot be predicted from context-
dependent grapho-phonological rules (e.g., English). Many authors have suggested
that differences in the depth of alphabetical codes imply different ways of
processing written languages (Baluch & Besner, 1991; Frost, Katz, & Bentin, 1987;
Seidenberg, 1985). In the case of languages with a deep orthography, skilled readers
are normally assumed to recognize words through the orthographic-graphemic code,
whereas in transparent orthographies readers are assumed to rely on the phonemic
E. Witrule, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading,

252 J. E. JIMENEZ GONzALEZ
prelexical code. Nevertheless, even though English is opaque in its orthography,

there is empirical evidence for the use of orthographic and phonemic cues in
printed-word recognition (e.g., Perfetti, Bell, & Delaney, 1988).
In the Spanish, as a language with a more transparent orthography, there is a
high probability that the lexical access is influenced to a larger extent by
phonological recoding. Spanish has a number of phonological and orthographic
differences from English. For instance, Spanish has a very regular orthography, that
is, the pronunciation of a string of letters can always be derived from print
(Sebastian-Galles, 1991). Each grapheme only has one possible pronunciation and in
this language, irregular words do not exist. There are some exceptions where the
pronunciation depends on the syllabic context (e.g., "c" is pronounced Ik/ when it is
followed by the vowels "a", "0", "u" and 101 when it is followed by the vowels "e",
"i"; "g" is pronounced as Ig! when it is followed by the vowels "a", "0" and "u" and
Ixl when it is followed by the vowels "e", "i"; and "r" is pronounced as Irrl whether
it appears at the beginning of the word or when it is preceded by the letters I, n, s,
and iri: rest). The number of different syllabic structures is limited, the irregularities
can be resolved by taking into account the overall syllable where they appear. As a
consequence, the difficulties with reading would be explained basically by
difficulties using a phonological route. Morais (1995) has suggested that languages
with a highly transparent orthography might exhibit a more systematic use of
phonological decoding than languages with opaque orthography, like English.
Morais also noted that access to phonological transcoding might thus be easier for
languages that have a low number of vowels and a few complex syllabic structures.
For instance, whereas in Spanish, Cuetos (1989) found a very rapid development of
the alphabetical route in children aged 5-6, there is evidence that this route develops
more slowly in normal English readers and does not reach the level of mature
readers until approximately 9 years of age (Backman, Bruck, Herbert, & Seidenberg,
1984). As has been suggested by several authors (Bryson & Werker, 1989),
acquiring this knowledge is a very complex process in English, because of the
irregular nature of the sound-to-symbol correspondences in this language. In this
way, the reader is obliged to resort to the lexical route to be able to identify words
never seen before, comparing them, through analogy, to words that are known or
learning them globally as if they were a logograph. In summary, to an English
language user, both the phonological and lexical routes may be useful.
2. PSYCHOLINGUISTIC PARAMETERS AND WORD RECOGNITION
Traditionally, word identification has been conceptualized as a context-driven

process that relies heavily on predictive or guessing strategies. For instance,
Goodman (1976b) conceptualized the reading as a "psycho linguistic guessing
game". That means that readers, using their normal and natural language processes,
guess the message by taking advantage of context. However, this assumption is
contrary to fact because research findings has demonstrated that word identification
is a highly automatized modular process that need not import any contextual
information for its execution (Nicholson, 1991). Furthermore, Goodman's (1976a)
traditional contributions regarding the influence of the context in word recognition
and in the type of errors the reader may commit are questionable on the basis of the
READING DISABILITIES AND TRANSPARENT ORTHOGRAPHY 253
empirical data obtained through eye movement techniques. Their fundamental

assumption about skilled reading is contrary to fact (Rayner & Pollatseck, 1987).
Readers fixate their eyes about 80% of the content words and about 30% of the
function words, and the quantity and duration of those fixations depend on the type
of word and the level of the reader's ability. This means that word recognition plays
an important role in reading and that the context may not be as prominent an effect
as Goodman suggests.
Most reading models assume that two different mechanisms or routes allow word
access: a direct orthographic route and an indirect phonological route (Ehri &
Saltmarsh, 1995). The first route involves direct connections between a written word
and its occurrence in the orthographic lexicon of the subject. This route is generally
used to read high frequency words, whether they are regular or irregular. The
indirect route involves "spelling-to-sound" correspondences. Earlier versions of the
dual-route theory (Coltheart, 1978, 1980) sustained that the two routes were
completely independent. However, a modified dual-route theory assumes that the
two routes are somewhat dependent on one another in terms of both knowledge
structure and processes (Humphreys & Evett, 1985). The general view most
commonly favored in the bibliography literature is that the presentation of a letter
string activates both lexical and phonological routines (e.g., see Patterson &
Coltheart, 1987). These routines, in parallel, more or less automatically compute a
phonological code, and the decision mechanism selects the output from the routine
that first makes an available response. The faster of the two routes determines access
in any given case.
It is unlikely that a competent reader will apply systematically the grapheme-
phoneme conversion rules to all the words, especially to those that are more
familiar. When words are familiar to us, we can recognize them globally. In
languages such as Spanish, the dual route model would not be absolutely necessary
because the reader uses basically the phonological route. However, the use of the
lexical or visual route does appear to be functional and operative, at least in some
words.
Some psycho linguistic parameters have been used in Spanish studies in order to
assess the routes involved in gaining access to the meaning of print. Neither
regularity, nor homophony effects can be studied with Spanish subjects as
correspondences between graphemes and phonemes are highly regular in this
language. However, in order to assess whether children use different routes,
researchers generally take into account either the accuracy of nonword reading, or
the difference in RTs and error performance as a function of lexicality, length, or
word frequency. Word length has been traditionally associated with phonological
processing because reading time is increased by word length (Just & Carpenter,
1980). This has been interpreted as evidence that readers carry out letter-to-Ietter
phonological processing (de Vega, Carreiras, Gutierrez, Alonso-Quecuty, 1990). If
the grapheme-phoneme conversion rules (GPC) are insufficiently learned, longer
RTs are more likely to occur in longer, rather than in shorter stimuli, because there
will be more GPC rules in longer words than in shorter ones (Valle-Arroyo, 1996).
Word frequency could have an effect on reading time or error performance and
would imply the use of direct access, as the phonological route is widely taken to be
prelexical. It is well known that high-frequency words are recognized faster than
low-frequency words (Frost, et aI, 1987). Another interesting psycholinguistic
parameter is the positional frequency of syllables (PFS) (Le., the number of times a
syllable appears in a particular position in a word-first, second, final, etc.). This
parameter would be associated to phonological processing because it is expected to
influence reading time, and it could reflect the use of the phonological route in the
Spanish language. If words and nonwords with low frequency syllables are read more
slowly, that is, as the less frequent a syllable is, the more likely it is that the conversions
implied have not been used in the past and the more poorly consolidated they will be.
Finally, when lexicality has an effect on RTs or error performance it is due to only
words can known by the reader, not nonwords.
Some research was conducted in Spanish to determine to what extent the dual
models are functional (de Vega & Carreiras, 1989; de Vega, et aI, 1990; Defior,
Justicia, & Martos, 1996; Garcia-Albea, Sanchez & del Viso-Pabon, 1982; Valle-
Arroyo, 1989, 1996). For instance, Garcia-Albea et al. (1982) detected significant
effects of word frequency in Spanish. Valle-Arroyo (1989) conducted a study to test
predictions from the dual-route theory in Spanish children ranging from eight to
thirteen years old and the results showed no frequency effect and a significantlength
effect. He concluded that this evidence indicates the use of the non-lexical
phonological route in readers of Spanish. However, taking the errors as the
dependent variable this same author reached the same conclusions as Garcia-Albea
et al. (1982) in his research on the validity of the dual route model of reading and
writing in the Spanish language. He found that the subjects committed four times
more errors in nonwords than in words (consequence of lexicality). This can only be
attributed to the visual route. This implies that even though Spanish is a transparent
language, the Spanish readers identify familiar words in a direct or lexical way,
without phonological mediation. De Vega and Carreiras (1989) and de Vega et al.
(1990) have also found an interaction between length and frequency. The effect of
the length dimension on high frequency and low frequency words indicates that
length does not affect known words as they are recognized in a holistic way and
processed by the lexical route. More recently, Valle-Arroyo (1996) has presented
data from normal subjects with different reading levels and from patients with
acquired dyslexia, and he found evidence for the role of both the lexical and the non-
lexical routes by Spanish children and adults. Defior, et al. (1996) studied the effect
of several lexical and sublexical variables (lexical category, lexical frequency,
syllabic structure, and word length) in the acquisition of reading in normal and poor
Spanish readers. They demonstrated that the four variables studied produced a
significant effect on the number of errors made by the children.
Overall, the pattern of results in these studies suggests no difference between the
processes involved in the acquisition of reading Spanish and those implicated in
deep orthographies such as English.
3. READING DISABILITIES: IQ OR PHONOLOGICAL SKILLS ?
In Spain, a legal category of learning disabilities (LD) as defined in the United

States by the National Joint Committee on Learning Disabilities (NJCLD, 1994),
does not exist in the special education field. Therefore, Spain has not had a tradition
of using an IQ-achievement discrepancy criterion, and Spanish professionals have
never used it for the identification ofLD (Jimenez & Hernandez-Valle, 1999). In the
United States, "Learning disorders are diagnosed when the individual's achievement
on individually administered, standardized tests in reading, mathematics, or written
expression is substantially below that expected for age, schooling, and level of
intelligence" (American Psychiatric Association, 1994, p. 46). The traditional
approach to identifying individuals with LD is to select children who are one or two
years behind in reading, in spite of having average or above-average IQ (i.e.,
dyslexics). Children with LD should be distinguished from those who demonstrate
general learning backwardness (i.e., garden-variety poor readers). The poor
academic performance of these children of below average intelligence is believed to
be non-discrepant, or in accordance with their lower cognitive capabilities.
However, using the concept of "discrepancy" to classify reading disabilities
leads to various problems. In 1989 the Journal of Learning Disabilities included
several articles which discussed the usefulness of discrepancy criteria in the
identification of LD in different academic domains. This special series was initiated
by Siegel (1989) and articles appeared with arguments in favor and against this
criteria. The discrepancy criteria implies several assumptions (Siegel, 1989; Toth &
Siegel, 1994): 1) intelligence tests are able to measure potential intellectual capacity,
2) dyslexia is caused by some form of highly circumscribed cognitive deficit which
does not affect IQ; 3) dyslexics with a reading-IQ discrepancy are qualitatively
different from poor readers who have low IQ scores (i.e., no discrepancy). In fact,
the term dyslexia has been reserved for children with discrepancies between
intelligence and reading ability, it has been assumed that there are important
etiological, neurological, and cognitive differences between high-IQ and low-IQ
poor reading performance.
However, many studies have demonstrated that there was no evidence that
dyslexics and poor readers were different in reading or spelling skills or other basic
cognitive processes. Siegel (1992) published data to show that the areas where there
were differences between subjects with dyslexia and poor readers were less related
to the fundamental processes involved in reading than the areas in which there were
no differences. More recently, Toth and Siegel (1994) after reviewing 21 studies that
explicitly compared poor readers and dyslexics, found that there were more
similarities than differences between these groups in reading tasks such as word
recognition, decoding, comprehension, orthographic and phonological awareness.
Most of the differences found between the dyslexics and poor readers were limited
to IQ correlated tasks such as mathematics, vocabulary, and syntax. It is suggested
that the principal cause of LD in reading is overreliance on the phonological route
(Stanovich, 1988).
To see if these results could be extrapolated to Spanish, Jimenez and Rodrigo
(1994) designed an experiment using a lexical decision task and they manipulated
different variables in the words (i.e., familiarity, length, and PFS) to analyze the
effects in children with LD and NLD children with different levels of IQ.
Specifically, the main purpose of this study was to test whether lexical and
sublexical parameters had a greater influence on subjects with LD than on NLD
subjects, independently of IQ. The analyses of interactions found between reading
level and psycholinguistic parameters provide indications about the cognitive
processes used by both groups. With the low-level cognitive processes (sublexical
and lexical) the LD group showed a greater deficit independently of IQ scores. The
10 influenced lexical decision tasks, but did not have relevance in the explanation of
differences in lexical access between the LD and NLD groups (see Figure 1).
2,5
2
1,5
I-
II:
0,5
0
<80 81-90 91-109 110-140
IQ
I-+-LD __ NLD I
Figure 1. Reaction times in lexical decision task as a function of IQ and reading level.
Using a naming task, Rodrigo and Jimenez (2000) manipulated the same
variables mentioned in the previous study (i.e., length, PFS, and word frequency,
and the nonword parameters length and PFS). The experiment was designed to test
whether poor readers would have more difficulties· than normal readers,
independently of their lOs, in naming words under conditions that require extensive
phonological computation. They found that the poor readers, independently of their
10, was slower in the naming task. This pattern of data has also been demonstrated
in other studies of languages with opaque orthography (see for a review, Rodrigo &
Jimenez, 2000). Moreover, 10 did not explain the differences between groups in this
task. That means that the group of poor readers, independently of their 10, was
slower in the naming task probably due to difficulty with phonological processing.
In fact, one particularly relevant result of this study is seen in the differences found
between the groups in reading nonwords. Many authors have suggested that
probably the most significant measure of phonological processing is the reading of
non words because it requires a knowledge of the sound-spelling relations. Some
researchers point out that while in a transparent orthographical system like Spanish,
nonwords can be read by analogy (Sebastilin-Galles, 1991); in fact, it is impossible
to read them correctly without doing some kind of segmentation and without
knowing the rules of letter-sound correspondence. For this reason, many researchers
consider this task to be the most discriminatory in detecting reading ability (Perfetti,
1986; Siegel, 1986). There is also evidence that poor readers have greater difficulty
reading nonwords (Ehri, 1975; Ehri & Wilce, 1983; Snowling, 1980). What seems to
be clear is the problem of poor readers in phonological recoding (Perfetti, 1985).
Stanovich (1988) suggested that a deficit in phonological recoding is at the root of
the problem of reading difficulty and in the same way, Gough and Tunmer (1986)
highlighted the importance of phonological recoding difficulties as the basis of
reading problems.
In another study (Rodrigo & Jimenez, 1999), error performance was also
analyzed for the same groups in the naming task. The analysis of errors have
constituted a topic of research in the field of reading. This method seems to be
promising due to its high ecological value and because it evaluates one of the few
observable manifestations of reading processes. Reading and spelling errors have
also been a focus of attention and research in neuropsychology, resulting in
classifications of dyslexia as the ones by Boder (Boder & Jarrico, 1982). In a similar
manner, many researches are designing intervention strategies based on word
recognition in order to reduce the errors committed by reading disabled children
(Lenz & Hughes, 1990). Nevertheless, this methodology is still evolving and it
presents some problems. The most important are the lack of agreement in definitions
of error categories and the lack of control in text difficulty in relation to the type of
errors. This fact may be due to the need of having, as a reference, a theoretical
model about reading as has been suggested by Castles and Coltheart (1993). Taking
into account the special cognitive requirements of reading, we might expect, that the
quality of errors produced in a word naming task should reveal the nature of the
underlying processes. In a study by Rodrigo and Jimenez. the subject had to read
aloud isolated words, taken out of context, and nonwords. In this way, one of the
methodological problems noted above was eliminated. As we expected, the poor
readers committed more errors than the normal readers in the word and nonword
naming task. Also, the poor readers were more affected than the normal readers by
the length of nonwords. In addition, the poor readers committed more errors when
the words had a low frequency. Finally, the poor readers were more affected by
lexicality than the normal readers, since they made more non-responses in
nonwords. Consequently, the authors concluded that error analysis was a more
reliable measure than RT for analyzing phenotypic performance pattern in
individuals with reading disabilities.
These and other similar results have caused researchers to turn their attention to
other factors that are more immediately related to the reading process. From the
theory of information processing we know that reading is a complex cognitive
activity. There are perceptual processes for analyzing graphical information, lexical
processes for finding the concept associated with graphical symbols, and syntax
processes for finding out the interrelation between words and semantic processing,
which extracts the meaning and integrates it into the reader's knowledge scheme (de
Vega, et aI, 1990). Possibly, then, a better way to identify a child with learning
disability would be to evaluate the different cognitive processes (e.g. word
recognition) that are involved in reading. In summary, taking into account the data
obtained in the studies reported here, we could conclude that IQ does not explain the
differences between poor readers and normal readers in word recognition and that
phonological processing is a more significant component of reading ability.
4. DO THE ORTHOGRAPHIC UNITS HAVE AN INFLUENCE ON

READING DISABILITIES?
Another topic of major interest for reading and reading disabilities is related to the
relative influence that different forms of orthographic units (e.g., syllables, onset-
rime, etc.) might have in explaining reading disabilities (Jimenez, Alvarez, Estevez
258 1. E. JIMENEZ GONzALEZ
& Hernandez-Valle, 2000). In Spanish, there are some linguistic units that seem to
be more important for word recognition than in other languages. One of these units
is the syllable, several Spanish studies cited above have used the PFS to analyze
their role in the processing of printed words, as Spanish syllables are well-defined
and the syllable boundaries are always clear. The effect of syllable frequency varies
depending on the task. When the task does not require access to lexical entries, the
frequency of the syllable speeds processing. In contrast, when the task requires
access to lexical entries, the syllable frequency slows processing because its effect
occurs at the lexical level during the activation of lexical candidates (Carreiras,
Alvarez & de Vega, 1993; Dominguez, Cuetos & de Vega, 1993). However, the
effect of PFS is the same in lexical decision tasks and naming tasks in children; that
is, words and nonwords with low PFS slow processing (Jimenez, Guzman, &
Artiles, 1997).
On the other hand, we do not know whether onset- and rime-like units in Spanish
may be involved in the translation of printed letter strings into phonological
representations as well as in visual word recognition. Several studies have found that
sensitivity to rhyme in preschool children is a good predictor of future reading
ability (Bradley & Bryant, 1985). Nevertheless, the relative incidence that each of
the phonological awareness levels has on reading would depend on the
characteristics of each language, so that when the language has deep orthography, it
may be more influenced by intrasyllabic awareness. For instance, Jimenez and Ortiz
(2000) using structural equation modeling in a longitudinal study with Spanish
children found a relation between preliterate syllabic awareness and word- and
pseudoword-reading. This is in accordance with the results of Spanish studies that
showed that syllabic awareness was a good predictor of reading ability (e.g.,
Carrillo, 1994). In fact, once the children know the Spanish alphabetical code and
possess phonemic awareness, it is not necessary to categorize words by their intra-
syllabic components in order to be able to read.
Theoretical models for opaque orthography have been proposed in which intra-
syllabic awareness contributes directly to reading, and in which intrasyllabic
awareness is independent of the connection between reading and phonemic
awareness (Bryant, MacLean, Bradley, & Crossland, 1990). The assumption
underlying these findings is that children who are able to categorize words based on
rhyme or onset, when learning to read, would realize that words with similar
orthographical patterns are pronounced similarly. Consequently, they could read
new words by making analogies with known words belonging to the same category
(e.g., right, light, might, sight, etc.). Bradley and Bryant (1978), Bowey, Cain and
Ryan (1992) and Bruck (1992) studied onset-rime awareness in dyslexic children
and they found differences between reading disabled readers and younger normal
readers in their performance when given intrasyllabic tasks. In this context, it is
relevant to know what kind of sub lexical units are involved in the different
alphabetical systems and if readers use onset and rime units in addition to grapheme-
phoneme and phoneme units when decoding words and nonwords in a transparent
orthography. As Treiman suggested (1992, p.98) "the basic notion that print
represents speech might be more easily understood, especially by children with poor
phonological skills, if it were introduced by reference to a larger, more accessible
unit of sound".
Relatively little empirical research in a transparent orthography has compared

normal readers and reading disabled children's use of different multi-letter units.
Nevertheless, some empirical evidence also does exist in a transparent orthography
that reading disabled children have difficulty in analyzing complex onsets and rimes
into their component phonemes (Jimenez, 1997) though their performance was
equivalent to the younger reading level-matched group (RL). Therefore, they may be
unable to learn the grapheme-phoneme mappings, and this may suggest that Spanish
reading disabled children would also use correspondences that are based on higher-
level units such as onsets and rimes in a transparent orthography. Jimenez et al.
(2000) designed an experiment in Spanish to test whether onset- and rime-like units
may be involved in visual word recognition in reading disabled children in
comparison to normal readers. They suggested that reading disabled children would
be unable to learn the grapheme-phoneme mappings, and this would suggest that
Spanish reading disabled children would also use correspondences that are based on
higher-level units such as onsets and rimes. For instance, lexical decisions would be
faster for stimuli like S_OL or CL_UB in reading disabled children than for stimuli
like SO_L or C_LUB. However, in normal readers there would not be differences,
because normal readers can fully analyze spoken words into phonemes and,
consequently, they may have less need for larger units. The authors found that
orthographic onset and rime units were not involved in the translation of printed
letter strings into phonological forms. They had hypothesized that for individuals
with reading disabilities, in comparison to normal readers, lexical decisions would
be faster for printed stimuli that had a space between the onset and rime than printed
stimuli where the separation did not correspond to the onset/rime boundary, because
normal readers can fully analyze spoken words into phonemes and, consequently,
they would have less need for larger units. However, they found that there were no
differences between Spanish reading disabled and normal readers in using
correspondences that are based on higher-level units such as onsets and rimes in a
transparent orthography. Consequently, onset-rime units would not be relevant in the
Spanish language because there is a direct correspondence between graphemes and
phonemes. However, in English (Le., an opaque system) readers can rely more on
spelling patterns as has been demonstrated. Because of this, the relative influence
that different forms of orthographic units (e.g., syllables, onset-rime, etc.) would
have in explaining reading disabilities could depend on orthographic systems.
5. READING DISABILITIES: DEVELOPMENTAL LAG VS. DEFICIT

HYPOTHESIS
Finally, we will turn to the issue the functional nature of reading disabilities. For
instance, a major concern is whether specific reading disabilities fit a developmental
lag or deficit model. This question has been addressed by a number of RL design
studies which were used as tests for this hypothesis involving only chronological-
age matched samples. For instance, in some studies reviewed in Spanish
(Dominguez & Cuetos, 1992; Jimenez & Rodrigo, 1994), the experience with print
was not controlled because an RL-match design was not used. Consequently,
interpretation of these studies was compromised because the design did not allow us
to know whether the results supported a developmental lag model or a deficit model
in a transparent orthography. As has been suggested by Bryant and Goswami (1986)

the studies which analyze correlates of reading disability should involve a
combination of RL and chronological age matched groups. In the three-group
design, there are two control groups in addition to the target group, one for RL and
one for chronological age. Therefore, the design allows not only the comparison of
different chronological aged children with the same reading level as in the two-
group approach, but also comparison within chronological age across the reading
levels. The addition of the third group-chronological age controls-allows
examination of the differing performance levels across two chronological age levels
in normal children, as well as relative performance within chronological age and RL
matched groups. As several authors have pointed out (Backman et aI, 1984; Bryant
& Goswami, 1986), positive results (i.e., a difference between reading disabled
children and normal controls) in experiments that use a RL match allows us to
conclude that the measure under consideration is probably causally related to the
reading disabilities.
Many studies carried out in opaque orthography using the RL-match design have
found empirical evidence in favor of the deficit model in phonological processing,
because individuals with dyslexia have more difficulty in reading nonwords than
non-disabled readers matched on age or RL (see for a review Jimenez & Hernandez-
Valle, 2000). Recently, these same authors used an RL-match design, and two
experiments were conducted to investigate the effects of lexicality, word frequency,
word length and PFS on lexical-decision and word-naming performance. These
linguistic variables were manipulated in order to test the routes involved in gaining
access to the meaning of print in the different groups. The main purpose of the study
was to determine whether the difficulties that Spanish children with reading
disabilities experience in phonological processing are due to a developmental lag or
a specific deficit. A hypothesis based on a developmental lag would predict that
there would not be differences in the lexical access between poor readers and
younger, non-disabled readers. That is, that the differences in RT as a function of
those variables that allow us to test the routes, such as lexicality, word frequency,
PFS, and word length, should be similar between the groups. On the other hand, a
hypothesis based on a specific deficit model would predict that the differences in RT
in the same variables would be greater in individuals with reading disabilities, which
means that this group should be more affected by unfamiliar and longer words, low
PFS, and nonwords in comparison to younger, non-disabled readers. In this study,
the differences in RTs as a function of lexicality, word frequency, and word length
were not significant between the group with reading disabilities and younger, non-
disabled readers, in both lexical-decision or naming tasks. However, the error
performance showed that the group with RD was substantially worse at nonword
reading compared to the younger, non-disabled readers. Reading-level-match
deficits in nonword reading by individuals with dyslexia have been reported in a
number of other studies (e.g., Olson, 1994). Moreover, a differential pattern
emerged when the children with RD were compared to younger, non-disabled
readers in words. The main result was that no differences were found in the analysis
of the total number of errors but the group with RD produced more visual and
morphological errors, of the sort that might be considered consistent with the
hypothesis that they had a bias for lexical over phonological routes to naming.
Another question is whether deficits in phonological processes are specific to

reading, or whether there are more general phonological deficits in language
processes that contribute to deficits in phonological decoding. A number of studies
using a reading-level-match design have shown that children with RD have lower
scores on phonological awareness tasks than younger, non-disabled readers (see for
a review, Jimenez & Hernandez-Valle, 2000). For another study, Jimenez (1997)
analyzed phoneme awareness in the same groups reported above. The main purpose
in the research was to test whether different levels of phonological awareness could
explain the reading disabilities in a transparent orthography. Phonological awareness
would also have an influence on reading disabilities, but it would depend on the
levels of phonological awareness studied. The study demonstrated that the reading
disabled group was equivalent to the younger RL control group in the odd-word-out
task, however, there were differences between those groups in phoneme
segmentation and reversal tasks (see Figure 2).
6
4
2
O+-------+--------j
Odd-word-out Segmentation Reversal
Tasks
-+-Reading disabled _ Younger readers . 0ir~ Normal readers
Figure 2. Performance on phonological awareness tasks as a function of reading level.
This suggests that a precursor to phonological coding difficulties is a deficit in

segmental language skills, that is, the deficit observed in phonemic awareness in the
reading disabled group could be causally related to reading disabilities. Therefore,
we conclude that the relative incidence that different forms of phonological
awareness (e.g., intrasyllabic or phonemic awareness) would have in explaining
reading disabilities could depend on orthographical systems.
6. ACKNOWLEDGEMENTS
This work was supported by a grant from DG/CIT (DirecciOn General de InvestigaciOn Cientifica y
Tecnica, number PB91-0759, Ministerio de Educacion y Ciencia).
7. AFFILIATIONS
Dr. Juan E. Jimenez
Facultad de Psicologia
Departamento de Psicologia Evolutiva y de la Educacion
Universidad de La Laguna
Campus de Guajara, 38205 La Laguna, Tenerife Islas Canarias, ESPANA
e-mail: ejimenez@ull.es
8. REFERENCES
American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders.
Washington, DC: Author.
Backman, 1., Bruck, M., Herbert, M., & Seidenberg, M. (1984). Acquisition and use of spelling-sound
correspondences in reading. Journal of Experimental Child Psychology, 38, 114-133.
Baluch, B., & Besner, D. (1991). Visual word recognition: Evidence for strategic control of lexical and
nonlexical routines in oral reading. Journal of Experimental Psychology: Learning, Memory and
Cognition, 17, 644-652.
Boder, E., & larrico, S. (1982). The Boder test of reading and spelling patterns: A diagnostic screening
test for subtypes of reading disabilities. Orlando, FL: Grune & Stratton.
Bowey, 1., Cain, M., & Ryan, S. (1992). A reading-level design study of phonological skills underlying
fourth-grade children's word reading difficulties. Child Development, 63, 999-1011.
Bradley, L., & Bryant, P. E. (1978). Difficulties in auditory organization as a possible cause of reading
Bradley, L., & Bryant, P. E. (1985). Rhyme and reason in reading and spelling. I.AR.L.D. Monographs
1. Ann Arbor. University of Michigan Press.
Bruck, M. (1992). Persistence of dyslexics' phonological awareness deficits. Developmental Psychology,
28, 874-886.
Bryant, P. E., & Goswami, U. (1986). Strengths and weaknesses of the reading level design: A comment
on Backman, Mamen, and Ferguson. Psychological Bulletin, 100, 101-103.
Bryant, P. E., Maclean, M., Bradley, L., & Crossland, M. (1990). Rhyme, alliteration, phoneme detection
and learning to read. Development Psychology, 26, 1-10.
Bryson, S. E., & Werker, l. F. (1989). Toward understanding the problems in severely disabled readers:
Part I. Vowel errors. Applied Psycholinguistics, 10, 1-12.
Carreiras, M., Alvarez, e., & de Vega, M. (1993). Syllable frequency and visual word recognition in
Spanish, Journal of Memory and Language, 32, 766-780.
Carrillo, M. S. (1994). Development of phonological awareness and reading acquisition. A study in
Spanish language. Reading and Writing: An Interdisciplinary Journal, 6, 279-298.
Castles, A, & Coltheart, M. (1993). Varieties of developmental dyslexia. Cognition, 47, 149-180.
Information Processing (pp. 151-216). London: Academic Press.
Coltheart, M. (1980). Reading, phonological recoding and deep dyslexia. In M. Coltheart, K.E., Patterson,
& l.e. Marshall (Eds.), Deep Dyslexia (pp. 197-226). London-New York: Routledge and Kegan
Paul.
Cuetos, F. (1989). Lectura y escritura de palabras a traves de la ruta fonologica [Reading and spelling
using a phonological route]. Infancia y Aprendizaje, 45, 71-84.
Defior, S., Juslicia, F., & Martos, F. (1996). The influence of lexical and sublexical variables in normal
and poor Spanish readers. Reading and Writing: An Interdisciplinary Journal, 8, 487-497.
de Vega, M., & Carreiras, M. (1989). The role of graphemic frequency in visual word processing.
Communication presented at the 3rd European Conference for Learning and Instruction. Madrid.
de Vega, M., Carreiras, M., Gutierrez, M., & Alonso-Queculy, M. L. (1990). Lectura y comprension. Una
perspectiva cognitiva[Reading and comprehension. A cognitive approach]. Madrid: Alianza
Editorial.
Dominguez, A & Cuelos, F. (1992). Desarrollo de las habilidades de reconocimiento de palabras en
ninos con dislintas competencias lectoras [Word recognition and reading disabilities]. Cognitiva, 4,
193-208.
Dominguez, A., Cuetos, F., & de Vega, M. (1993). Efectos diferenciales de la frecuencia sihibica
posicional: dependencia del tipo de prueba y caracteristicas de los estimulos [Effects of positional
frequency of syllables as a function of task and stimuli]. Estudios de Psicologia, 50, 5-31.
Ehri, L. C. (1975). Word consciousness in readers and prereaders. Journal of Educational Psychology,
67, 204-212.
Ehri, L. c., & Saltmarsh, J. (1995). Beginning readers outperform older disabled readers in learning to
read words by sight. Reading and Writing: An Interdisciplinary Journal, 7, 295-326.
Ehri, L. C., & Wilce, L.S. (1983). Development of word identification speed in skilled and less-skilled
beginning readers. Journal of Educational Psychology, 75, 3-18.
Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recognition and orthographical depth:
A multidimensional comparison. Journal of Experimental Psychology: Human Perception and
Garcia-Albea, J. E., Sanchez, c. R., & del Viso-Pabon, S. (1982). Efectos de la frecuencia de uso en el
reconocimiento de palabras [Effects of word frequency in word recognition]. Investigaci6n
Psicol6gica, 1, 24-63.
Goodman, K. S. (1976a). What we know about reading. In O. D. Allen, & D. 1. Watson (Eds.), Findings
of research in miscue analysis: Classroom implications. Urbana Illinois US. National Council of
Teachers of English (NCTE).
Goodman, K. S. (1976b). Behind the eyes: What happens in reading. In H. Singere, & R. B. Rudell
(Eds.), Theoretical models and processes of reading (pp. 497-508). Newark, N.J.: International
Reading Association.
Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading and reading disability. Remedial and Special
Education, 7, 6-10.
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 Behovioral and Brain Sciences, 8,
689-740.
Jimenez, 1. E. (1997). A reading-level match study of phonemic processes underlying reading disabilities
in a transparent orthography. Reading and Writing: An Interdisciplinary Journal, 9, 23-24.
Jimenez, J. E., & Hernandez-Valle, I. (1999). A Spanish perspective on LD. Journal of Learning
Jimenez, J. E., & Hernandez-Valle, I. (2000). Word identification and reading disorders in the Spanish
language. Journal of Learning Disabilities, 33, 44-60.
Jimenez, J. E., & Ortiz, M. R. (2000). Metalinguistic awareness and reading acquisition in the Spanish
language. The Spanish Journal of Psychology, 3, 37-46.
Jimenez, J. E., & Rodrigo, M. (1994). Is it true that the differences in reading performance between
students with and without LD cannot be explained by IQ ? Journal of Learning Disabilities, 27, 155-
163.
Jimenez, 1. E.; Guzman, R., & Artiles, C. (1997). Efectos de la frecuencia sihibica posicional en el
reconocimiento visual de palabras y aprendizaje de la Iectura [Effects of positional syllable frequency in
visual word recognition and learning to read]. Cognitiva, 1, 3-27.
Jimenez, J. E., Alvarez, C. J., Estevez, A., & Hernandez-Valle, I. (2000). Onset-rime units in visual word
recognition in Spanish normal readers and children with reading disabilities. Learning Disability
Research & Practice, 15, 135-141.
Just, M. A., & Carpenter, P. A. (1980). A theory of reading: from eye fixation to comprehension.
Psychology Review, 87, 329-354.
Lenz, B. K., & Hughes, CH. A. (1990). A word identification strategy for adolescents with learning
disabilities. Journal of Learning Disabilities, 23, 149-163.
Morais, 1. (1995). Do orthographic and phonological peculiarities of alphabetically written languages
influence the course of literacy acquisition? Reading and Writing: An Interdisciplinary Journal, 7, 1-
7.
National Joint Committee on Learning Disabilities (1994). Collective perspectives on issues affecting
learning disabilities. Austin, TX: PRO-ED.
Nicholson, T. (1991). Do children read words better in context or in lists? A classic study revisited.
Olson, R. K. (1994). Language deficits in "specific" reading disability. In M. A. Gernsbacher (Ed.),
Handbook ofPsycholinguistics (pp. 895-916). New York: Academic Press.
Patterson, K., & Coltheart, V. (1987). Phonological processes in reading: A tutorial review. In M.
Coltheart (Ed.), Attention and performance XII: The psychology of reading (pp. 422-447). HiIlsdale,
NJ: Erlbaum.
Perfetti, C. A. (1985). Reading ability. New York. Oxford University Press.
Perfetti, C. A. (1986). Continuities in reading acquisition, reading skill and reading disability. Remedial
and Special Education, 7, 11-21.
Perfetti, C. A., Bell, L. c., & Delaney, S. M. (1988). Automatic (prelexical) phonetic activation in silent
word reading: Evidence from backward masking. Journal of Memory and Language, 27, 59-70.
Rayner, K., & Pollatsek, A. (1987). Eye movements in reading: A tutorial review. In M. Coltheart (Ed.),
Attention & performance XIII: The psychology of reading (pp. 327-362). Hillsdale, NJ: Lawrence
Erlbaum Associates.
Rodrigo, M., & Jimenez, J. E. (1999). An analysis of the word naming errors of normal readers and
reading disabled children in Spanish. Journal of Research in Reading, 22, 180-197.
Rodrigo, M., & Jimenez, J. E. (2000). IQ vs phonological recoding skill in explaining differences between
poor readers and normal readers in word recognition: Evidence from a naming task. Reading and Writing:
An Interdisciplinary Journal, 12, 129-142
Sebastian-Galles, N. (1991). Reading by Analogy in a Shallow Orthography. Journal of Experimental
Psychology: Human Perception and Performance, 17, 471-477.
Seidenberg, M. S. (1985). The time course of phonological code activation in two writing systems.
Cognition, 19, 1-30.
Siegel, L. S. (1986). Phonological deficits in children with a reading disability. Canadian Journal of
Special Education, 2, 45-54
Siegel, L. S. (1989). LQ. is irrelevant to the definition of learning disabilities. Journal of Learning
Siegel, L. S. (1992). An evaluation of the discrepancy definition of dyslexia. Journal of Learning
Snowling, M. J. (1980). The development of grapheme-phoneme correspondence in normal and dyslexic
readers. Journal of Experimental Child Psychology, 29, 294-305.
Stanovich, K. E. (1988). Explaining the differences between the dyslexic and garden variety poor reader:
The phonological-core variance-difference model. Journal of Learning Disabilities, 21, 590-604.
Toth, G., & Siegel, L. S. (1994). A critical evaluation of the IQ-achievement discrepancy based definition
of dyslexia. In K. P. van den Bos., L. S. Siegel., D. 1. Bakker., & D. L. Share. (Eds.), Currents
directions in dyslexia research (pp. 45-70). Linsse: Swets & Zeitlinger.
Treiman, R. (1992). The role of intrasyllabic units in learning to read and spell. In P. B. Gough, L. Ehri, & R.
Treiman (Eds.), Reading acquisition (pp.65-106). Hillsdale, NJ: LEA.
Valle-Arroyo, F. (1989). Errores de lectura y escritura. Un modelo dual [Reading and spelling miscues: A
dual route model]. Cognitiva, 2, 35-63
Valle-Arroyo, F. (1996). Dual-route models in Spanish: Developmental and Neuropsychological data. In
M. Carreiras, & N. Sebastian (Eds.), Language Processing in Spanish (pp. 89-118). Hillsdale, NJ:
LEA.
A. J. FAWCETT
DYSLEXIA, THE CEREBELLUM AND

PHONOLOGICAL SKILL
Abstract. The identification of a range of causal hypotheses for reading difficulties in dyslexia has been a
major achievement of dyslexia research over the last decade. Causal hypotheses should be capable of
explaining the difficulties, via the "cognitive" level of explanation in terms of primitive sub-skills, to
dysfunction of some neurological structure. Any complete theory of dyslexia should account for problems
in phonological skill, motor skill, automatisation and speed. It is now claimed that the cerebellum is
involved not only in the automatisation of motor skill, but also the acquisition of skills related to
language, making cerebellar dysfunction a prime candidate for an underlying cause of dyslexia. In earlier
research, using standard clinical tests (Fawcett, Nicolson, & Dean, 1996) we established clear but indirect
evidence in support of this hypothesis. Direct evidence derived from a PET study of motor skill learning
in dyslexic adults, which showed significantly lowered activation in the cerebellum in comparison with
controls (Nicolson & Fawcett, 1999a, b). In this chapter, I outline the most likely route by which
cerebellar problems might lead to reading and spelling difficulties, via the role of the cerebellum in the
development of articulation (a skill requiring fluent, precisely timed co-articulation between numerous
effectors). I then present a study on articulation deficits in dyslexia, which demonstrates that children with
dyslexia have significant problems in articulation, not only in motor planning, but also in the speeded
production of single articulatory gestures. This is consistent with the cerebellar impairment hypothesis.
Investigation of this hypothetical route of skill acquisition from infancy to school would provide a fruitful
research agenda.
1. INTRODUCfION: DEVELOPMENTAL DYSLEXIA

Developmental dyslexia is the most common of the developmental disorders, with
an estimated prevalence of at least 3-4% in Western school populations (Badian,
1984; Jorm, Share, McLean, & Matthews, 1986; Miles, 1990). There will be at least
one child with dyslexia in every class, with around four times as many boys as girls
diagnosed', normally on the basis of unexpected problems in learning to read-"a
disorder in children who, despite conventional classroom experience, fail to attain
the language skills of reading, writing and spelling commensurate with their
intellectual abilities" (World Federation of Neurology, 1968, p. 26).
In a ten-year research program, aimed at a fuller understanding of the underlying
causes of dyslexia, Nicolson and Fawcett adopted a skill learning framework in
order to outline the pattern, severity and incidence of difficulties in dyslexia across a
wide range of skills. We found severe and persistent deficits in phonological skill,
processing speed, balance and motor skill in our panels of children with dyslexia.
We therefore set out to find possible cause(s) of this range of deficits, focusing on
cerebellar impairment. The cerebellar impairment hypothesis can address the range
of problems in dyslexia, and led to specific predictions, which were confirmed by
empirical studies, with our original panel and further dyslexic subjects, thus
providing strong but indirect evidence that cerebellar impairment may be a cause of
dyslexia (see Nicolson & Fawcett, 1999a, 1999b for a review).
E. Witrul, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading,

266 A. J. FAWCETI
In this chapter, I present direct evidence on cerebellar impairment in dyslexia.

Then I outline the rationale for the cerebellar and alternative hypotheses, in order to
explain that the cerebellar impairment is not only compatible with existing
approaches, but also provides an explanatory framework with the potential to unite
these seemingly diverse approaches. The conference in Leipzig addressed basic
functions of language and language disorders, and therefore it seemed appropriate
here to present a study of articulation skills in dyslexia, the first stage in a further
analysis of this causal chain.
1.1 Dyslexia: the Theories
Deficits in dyslexia are diverse - motor skill, visual processing, rapid processing,
and memory - with many coupled with intriguing links to neuroanatomical
abnormalities. In a valuable recent review, Frith (1997) outlines three major current
"causal" explanations of dyslexia, namely phonological deficit, rapid temporal
processing deficit and cerebellar impairment. These three explanations are discussed
below.
1.2 Phonological Deficit
The demonstration that many reading-related deficits are attributable to

phonological processing deficits was the major achievement of dyslexia research
over the last decade (Bradley & Bryant, 1983; Snowling, Goulandris, Bowlby, &
Howell, 1986; Stanovich, 1988; Vellutino, 1979) - see Shankweiler et aI., (1995) for
a review. Evidence for this influential hypothesis derived from Bradley and Bryant
(1978), who found that children who subsequently developed reading difficulties
were significantly worse than other children at a rhyme judgement task. Their
intervention study showed that pre-school children trained in rhyming and
alliteration skills ("sound categorisation") made significantly more progress in
reading than children trained in semantic categorisation, but equal progress in maths.
This evidence for a causal link between early phonological skills and reading
acquisition, was confirmed and extended, both for normal readers (e.g., Hatcher,
Hulme, & Ellis, 1994) and for children with dyslexia (e.g., Ackerman, Dykman, &
Gardner, 1990; Lundberg, Frost & Petersen, 1988; Olson, Wise, & Rack, 1989;
Rack, 1985). Thus converging evidence led to a consensus on the phonological
deficit hypothesis, that children with dyslexia suffer from early impairments in
phonological awareness, which prevents them from acquiring the word decoding
and blending skills necessary for normal acquisition of reading.
1.3 Rapid Temporal Processing Deficit
Research into dyslexia has also identified more general problems in rapid
performance. Denckla and Rudel (1976) discovered problems in "rapid automatised
naming"; Tallal (1985) found impairments in ordering rapid temporal events; and
Wolf (1991) established a direct relationship between early speed of naming deficits
and incidence and severity of later reading deficits. Reduced speed of temporal
processing, even for non-phonological tasks, was found for speed of lexical access,
DYSLEXIA, THE CEREBELLUM 267
and even selective choice reaction to one of two pure tones (Nicolson & Fawcett
1994a) with simple reaction to that tone normal, suggesting that the need for a
decision caused the reduction in speed. Abnormalities have also been found in the
auditory magnocellular system (Galaburda, Menard, & Rosen, 1994; for a review of
the neuroanatomy of dyslexia see the chapter by Galaburda, this volume).
It has been argued that a temporal processing speed deficit across modalities
might be the cause of difficulties in dyslexia (Llinas, 1993; Stein, 1994; Stein &
Walsh,1997; Tallal, Miller, & Fitch, 1993). A review of the magnocellular deficit
hypothesis is provided in the chapter by Stein in this volume. This hypothesis
provides a link between known neuroanatomical differences via cognitive skills to
reading in the visual and auditory modalities. In the auditory domain, the hypothesis
accounts for phonological problems in terms of deficits in discrimination of similar
consonants. However, Mody, Studdert-Kennedy and Brady (1997) show that
temporal processing deficits are found only for speech rather than for all auditory
processing.
1.4 Cerebellar Impairment
The third major causal explanation is our own cerebellar deficit theory, which we
will consider in the majority of this chapter.
It may be seen that research has identified a wide range of apparently disparate
problems in dyslexia. It became one of the major challenges in the field to find an
explanatory framework not only broad enough to encapsulate these deficits, but also
specific enough to generate testable predictions, to support better diagnostic
procedures, and to inform remediation methods. This challenge formed the
motivation for our research program.
2. THE SHEFFIELD PROGRAM - DYSLEXIA AND SKILL

Interestingly enough, despite the range of theoretical perspectives adopted by the
late 80's, no attempt had been made to systematically assess components of learning
ability in children with dyslexia within a learning framework. Consequently, we
adopted a learning, or skill acquisition, perspective for our initial research.
Although learning is clearly a valuable framework for dyslexia, it has been
largely overlooked by theorists because it fails to explain the apparent specificity of
the deficit (Stanovich, 1988). One might predict that a general problem in learning
would lead to problems in all skills, both cognitive and motor, which is clearly too
powerful! Nevertheless, we had noted that when skills were examined carefully,
deficits had been found across the board for dyslexic children. Furthermore, dyslexic
children show lapses in concentration and tire more easily when performing a skill.
As part of my doctoral research, (Fawcett, 1990) our early work focused on
automatisation, a key concept in skill acquisition - the process by which skilled
performance becomes smoother, and requires less and less effort, following
extensive practice (Anderson, 1982; Fitts & Posner, 1967; Shiffrin & Schneider,
1977). In a stringent test of the Dyslexic Automatisation Deficit hypothesis, that
children with dyslexia have unusual difficulty in automatising any skill, whether
motor or cognitive, we tested dyslexic children and controls on balance under single
268 A. J. FAWCETf
and dual task conditions (Nicolson & Fawcett, 1990). As predicted, there were no
differences between the groups when just balancing, but when undertaking a further
secondary task, the children with dyslexia showed a highly significant impairment in
balance, whereas the control children showed no deficit. The automatisation deficit
hypothesis was the only one of the major theories to predict this pattern of results.
It seemed that there was clear support for all three theories of dyslexia. However,
the phonological deficit and temporal processing deficit were a little too specific to
explain the range of problems in dyslexia, whereas the automatisation deficit was
too general to account for the precise pattern of difficulties. All three theories predict
that there will be phonological difficulties, leading to reading problems.
Furthermore, there may be different subtypes of dyslexia (Boder, 1973; Seymour,
1987), [though see Thomson, 1984] and the different approaches may sample
different populations of children with dyslexia. This led us to the next phase of our
research.
2.1 Primitive Skills and Dyslexia

In order to check the generality of these deficits, a wide range of tests must be
administered to the same groups of dyslexic children, assessing the incidence of
deficit on each task individually for each child as well as performing a between-
groups analysis. This allows us to assess the severity and the incidence
independently. Consequently we set out to characterize performance on the entire
range of skill (even skills thought to be spared in dyslexia). We hoped to be able to
build up an overall picture of the pattern of difficulties associated with dyslexia, and
progress towards the major theoretical goal of dyslexia research, identification of the
underlying cause or causes.
In all our studies, we have adopted the standard discrepancy/exclusionary criteria
for membership of the dyslexia group - normal or above normal 10 [10 of 90 or
more on the WISC-R/WISC III (Wechsler, 1976, 1992)], without known primary
emotional, behavioral or socioeconomic problems, with reading age at least 18
months behind chronological age) compared with normally-achieving children, with
groups matched overall for age and 10.
For these studies three groups of children with dyslexia, mean ages 16, 12 and 8
years2, were recruited, together with three groups of normally achieving children
matched for age and 10.
2.1.1 Skill Tests

A variety of tests were designed which tap performance on "primitive" cognitive
and motor skills, which form the building blocks for more complex skills. In
addition to psychometric tests, four types of test were used, namely tests of
phonological skill, working memory, information processing speed, and motor skill.
Details are given in Fawcett and Nicolson (1995a, b, c), and overviews are presented
in Nicolson and Fawcett (1994b, 1995).
2.1.2 Results
Significant differences were found between children with dyslexia and their
chronological age controls on all but two of the 23 measures presented. Consistently
normal performance for children with dyslexia at all three age levels was found only
for simple reaction time. Comparison with reading age controls is the critical
theoretical test, since impairment here cannot be attributable to reading difficulties
alone, and therefore highlights some more basic problem (Bryant & Goswami,
1986). Tests on which significant impairments compared with reading age controls
were found were as follows: spelling; segmentation; word flash; picture naming
speed; bead threading; blindfold and dual task balance (see Nicolson & Fawcett,
1999a, b for details).
Clearly, these analyses showed significant deficits, even in comparison with
reading age controls, in several areas of skill. However, further analyses were
needed to address two major issues: the relative severity of the deficits on the
various tasks, and the relative individual incidence of deficit for the tasks. We first
normalized the data for each test for each group relative to that of the corresponding
control group.3 This procedure led to an age-appropriate "effect size" in standard
deviation units (analogous to a z score) for each test for each child (e.g., Cohen,
1969), a technique used throughout our studies for comparisons. Normally
distributed data has 15% of the popUlation at least one standard deviation below the
mean (impaired), and 2% at least 2 standard deviations below (significantly
impaired). The results showed that most of the dyslexic children were impaired on a
wide range of tasks. Naturally they were all impaired on reading since this was the
criterion for membership of the dyslexia group; and an even greater impairment for
spelling was as expected (Thomson, 1984). The other tasks for which the mean
effect size was -2 or worse are balance (dual task, one foot blindfold, and one foot
non-blindfold), segmentation, letter naming speed and articulation rate, with an
incidence rate of around 80% on most of these skills.
The individual data showed that balance difficulties, especially when
blindfolded, appeared therefore to be associated with all our sample of children with
dyslexia. Furthermore, if we consider links between balance, phonological deficits
and speed, 90% of the dyslexic children had positive scores on at least two, by
contrast with only 6% of the controls (Nicolson & Fawcett, 1995).
In short, dyslexic children in this sample showed no evidence of subtypes, but
rather showed deficits across the range of primitive skills. Performance overall was
consistent with automatisation deficit. Nevertheless, we feel that the automatisation
deficit hypothesis is merely another descriptive theory - an economical
characterization of the symptoms - rather than an explanation of the underlying
cause. A causal explanation should account for the precise pattern of results
obtained, and identify the mechanism(s) underlying these symptoms. The search for
an underlying cause was the focus of the next phase of our research program.
2.2 Dyslexia and the Cerebellum
The primitive skills analysis indicated that any complete theory of dyslexia should
account for problems in phonological skill, motor skill, automatisation, and
information processing speed. Interestingly enough, both phonological skill and
blindfold balance resist the effects of maturation, with the oldest children with
dyslexia barely reaching the level of the youngest controls. Clearly, if there were a
270 A. J. FAWCETf
mechanism underlying balance, automatisation, motor skill and phonological skill,

that mechanism would be a prime candidate for the underlying cause of dyslexia.
Deficits in motor skill and automatisation implicate the cerebellum, traditionally
considered as a motor area (Eccles, Ito, & Szentagothai, 1967;Holmes, 1917, 1939;
Stein & Glickstein, 1992), and also claimed to be involved in the automatisation of
motor skill and in adaptive learning (Ito, 1984, 1990). A clear demonstration of the
role of the cerebellum was provided by a recent PET study (Jenkins, Brooks, Nixon,
Frackowiak, & Passingham, 1994) which showed cerebellar activation associated
with automatic movements, but most extensively in new learning.
Nonetheless, despite early work by Levinson 4 (e.g. Frank & Levinson, 1973;
Levinson, 1990) the cerebellum has largely been discounted in the dyslexia literature
owing to its supposed lack of involvement in linguistic and cognitive skill. However,
Leiner, Leiner, and Dow (1989) note links between the cerebellum and the frontal
cortex, including Broca's language area, which make the cerebellum central for the
acquisition of "language dexterity". The role of the cerebellum in language and
cognitive skill is now established (Allen, Buxton, Wong, & Courchesne, 1997);
Thach, 1996), including a recent demonstration of specific cerebellar involvement in
reading (Fulbright et aI., 1999). This means that, in effect, the cerebellum is
involved in the automatisation of any skill, whether motor or cognitive. For recent
reviews of the involvement of the cerebellum in cognitive activities see
Schmahmann (1998).
The most severe deficits for children with dyslexia in the primitive skills were in
terms of phonological skill, naming speed, motor skill, and balance. Just such a
pattern would be predicted from current knowledge of the role of the cerebellum in
skill acquisition and execution. Clearly, therefore, mild cerebellar impairment was a
prime candidate for the underlying cause of dyslexia. Investigation of cerebellar
functioning in dyslexia formed the focus of the next phase of our program. In a
stringent preliminary test of the hypothesis, we first administered a test of time
estimation, which is thought to highlight specific impairments in cerebellar patients,
by contrast with an analogous task, loudness estimation, on which performance is
unimpaired (Ivry & Keele, 1989). Exactly as predicted by the cerebellar impairment
hypothesis, our three groups of dyslexic children showed a dissociation between
impaired time estimation and preserved loudness estimation, which had previously
been associated with cerebellar patients (Nicolson, Fawcett, & Dean, 1995).
2.2.1 Clinical Cerebellar Symptoms in Children with Dyslexia

However, dyslexic children should also show traditional signs of cerebellar
dysfunction if there is indeed a cerebellar impairment (Holmes, 1917, 1939; see also
Dow & Moruzzi, 1958). Traditional symptoms are dystonia (problems with muscle
tone) and ataxia (disturbance in posture, gait, or limb movements). There was no
evidence of problems of this type, apart from our own work and Levinson's (1990)
controversial findings. Consequently, in a further stringent test of the cerebellar
impairment hypothesis, we replicated the tests described in Dow and Moruzzi
(1958), using groups of children with dyslexia and matched controls aged 18, 14 and
10 years (see Fawcett, Nicolson, et aI., 1996). Tasks fell into three types; posture
and muscle tone; hypotonia of the upper limbs; and complex voluntary movement, a
total of 14 tasks in all.
Results. The performance of the children with dyslexia was significantly worse than
that of the chronological age controls on all 14 tasks, and significantly worse than
reading age controls on 11 out of the 14 tests.
Effect size analyses were again used, with each group normalized to
chronological age controls. All but one task (finger to finger) produced an overall
effect size for the groups with dyslexia of -1 or worse (that is, performance at least 1
sd worse than the controls). Deficits more severe than reading age (which showed an
effect size of -2.26, and included 100% of the dyslexic children) were found for
finger and thumb opposition; tremor; arm displacement; toe tap; limb shake; muscle
tone; diadochokinesis - speed of alternating tapping the knees with palm and back
of hand; and postural stability - movement when pushed gently in the back (see
Nicolson & Fawcett, 1999a, b for details). The performance of the lO-year-old
dyslexic children was markedly poorer than the older dyslexic children on several
tests of muscle tone, with large effect sizes of -4 and worse.
Discussion. The predictions of the cerebellar impairment hypothesis were

confirmed, with impairments on all the cerebellar tests, not only in comparison with
chronological age controls, but also for the majority of tests in comparison with
reading age controls. Moreover, the 18 year old dyslexics were consistently worse
than non-dyslexic children 8 years their junior on many tasks. Effect size analyses
showed that impairments were greater than the reading deficits for 8 measures of
dystonia, balance and speed. Furthermore, individual analyses showed that for 8 of
the 14 tasks over 80% of the children with dyslexia were "at risk". This is a
powerful finding given that earlier dyslexia research did not identify these problems.
Nevertheless, these findings provide only indirect support for the cerebellar
impairment hypothesis. In the next phase of the research we therefore undertook a
direct test of the cerebellar impairment hypothesis, using the task devised by Jenkins
et al. (1994) which tests cerebellar acquisition and performance of a simple motor
learning task known to activate the cerebellum. In a stringent test of the hypothesis,
we compared the positron emission tomography (PET) scan data of six adult
dyslexics drawn from our panel with matched controls. The task demanded that
participants learn a sequence of 8 keypresses, cued by a tone in a trial and error
process, with computer feedback on whether or not they had selected the right key.
The scan was undertaken in a resting condition, while executing a previously
learned sequence, and while acquiring a new sequence. In both the pre-learned and
new learning condition, highly significant differences were found in levels of
activation of the cerebellum for the dyslexics in comparison with the controls. In the
prelearned sequence, activation was significantly lower (p<.OI) in the right
cerebellar cortex and left cingulate gyrus, and while learning a new task, in the right
cerebellar cortex. Focusing on the rCBF in voxels in the area of maximum
activation, the dyslexic adults showed only 10-20% the level of increased blood
flow of the controls. These results would not be predicted by any other theory of
dyslexia and are amongst the strongest findings using the PET technique at the
Hammersmith hospital, UK.
Based on this direct evidence, it seems plausible to argue, therefore, that
cerebellar impairment provides a natural explanation of the three major behavioral
deficits (balance, phonological skill and time estimation) obtained in this series of
272 A. J. FAWCEIT
studies. The cerebellum is particularly active in initial skill acquisition (Ito, 1990;
Jenkins et al., 1994; Krupa, Thompson, & Thompson, 1993) and therefore the
hypothesis also accounts naturally for our finding (Nicolson & Fawcett, 1994c,
2000) that children with dyslexia have initial problems in blending two skills, but
subsequent skill acquisition is not abnormal apart from difficulties in eliminating
errors.
3. FROM CEREBELLAR IMPAIRMENT TO LITERACY: A HYPOTHETICAL

CAUSAL CHAIN
To summarize the results of our research program; significant deficits were found
for dyslexics compared with reading age controls in primitive skills, from
phonological skill to picture naming speed to motor skill and balance. This pattern
of deficits was strongly suggestive of cerebellar impairment, and therefore cerebellar
function was assessed. Dyslexic children showed precisely the pattern of difficulties
predicted from cerebellar impairments, with clear clinical signs of cerebellar
dysfunction. Finally, the most direct evidence of cerebellar impairment, PET scans
of the cerebellum while performing a motor learning task revealed significantly
lowered activation in our panel of dyslexic adults. The issue for this last section is to
what extent can cerebellar impairment be considered a truly causal explanation?
We should emphasize here that, even though it now appears that cerebellar-type
problems are important symptoms of dyslexia (and therefore valuable for diagnosis),
there is no reason to see cerebellar symptoms as manifestations of the "true"
underlying cause. Cerebellar symptoms, literacy difficulties, and phonological
difficulties may all be associated with a yet unidentified underlying cause. It is
therefore necessary to provide an account of how cerebellar impairment could lead
to the manifold symptoms of dyslexia, in the following plausible but speculative
account.
Consider the possible role of the cerebellum, in particular the neocerebellum, in
the proceduralisation of cognitive skills (see Leiner et al., 1989). Whether or not the
final procedural code is stored in the cerebellum or not (see Glickstein, 1993) the
cerebellum can "scaffold" the development of a cognitive skill by using in built
timing and error analysis mechanisms (Ito, 1990).
The critical issue for our purposes is the involvement of the cerebellum in
language skills. The cerebellum is a key structure in the development of articulatory
skill -a motor skill, demanding exquisite timing and fluency. Infants spend around
12 months learning to articulate their first word, (a plausible basis for the striking
development of the neocerebellum in humans, Passingham, 1975). In our primitive
skills analysis we found a highly significant deficit in articulation time for the
children with dyslexia, with the youngest group showing their most severe mean
impairment (effect size -3.4). Articulation plays a key role in language development,
particularly the development of phonemic awareness (Locke, 1983). Consequently,
the cerebellar impairment hypothesis can subsume the well-supported phonological
deficit hypothesis.
Interestingly, however, the cerebellar impairment hypothesis also accounts
naturally for the difficulties in handwriting, which are difficult for the phonological
hypothesis to explain, but follow from reduced motor skill. Poor spelling can be
attributed not only to impaired ability to acquire orthographic regularities, but also
to a reduced ability to automatise knowledge of spelling patterns. This twin deficit
may account for the greater severity and pervasiveness of spelling deficits compared
with reading deficits (Thomson, 1984). The account is developed further below.
The cerebellar impairment hypothesis provides a complete mechanism for the
causal chain outlined in Figure 1, starting from known functions of the brain (the
cerebellum) through cognitive processes (phonemic awareness, and also
automatisation deficits in letter knowledge), explaining the reading, writing and
spelling deficits, as well as other symptoms not directly related to reading.
Figure 1 outlines the hypothetical ontogenetic causal chain linking cerebellar
impairment, phonological difficulties and later reading problems.
Balance impairment
______ ~RITING)
Motor skill impairment
Phonological
awareness
Prob ems automallSmg /

skill and knowledge
1-------....
Figure 1. Proposed causal chain for the cerebellum and reading.
If an infant has a cerebellar impairment, they may be slower to sit up and to

walk, and may have greater problems with fine muscular control. However, our
most complex motor skill, and that needing the finest control is articulation and co-
articulation. Consequently the infant might be slower to start babbling, and, later,
talking. Indeed, there is emerging evidence that the early articulatory and manual
skills develop in step (Ramsay, 1984). Locke (1993, p.189) speculates that this is
based on initial development of left hemisphere control of the muscular movements
needed for reaching and speech. The left hemisphere takes control of speech like
activity, with babbling representing the convergence of motor control and sensory
feedback systems. Evidence for this derives from Fowler (1991) who found very
young children first perceive words as a loose bundle of articulated gestures, until in
time, the coarticulated gestures become grouped into the representations of
phonemes. Even after speech and walking emerge, one might expect that the skills
would be less fluent, less "dextrous", in infants with cerebellar impairment. If
articulation is less fluent than normal, then it takes up more conscious resources,
274 A. J. FAWCEIT
leaving fewer resources to process the ensuing sensory feedback. In particular, the
processing of the auditory, phonemic structure of the words spoken may be less
complete. There may, therefore, not be a natural sensitivity to onset, rime, and the
phonemic structure of language - in short, one would expect early deficits in
phonological awareness (see Snowling & Hulme, 1994, for a related account).
Cerebellar impairment would therefore be predicted to cause the "phonological core
deficit" (Stanovich, 1988) a fruitful explanatory framework for many aspects of
dyslexia. Furthermore, the cerebellar impairment hypothesis provides a natural
causal explanation of the poor quality handwriting shown by children with dyslexia
(which most existing theories of dyslexia have trouble explaining). Handwriting is a
motor skill, which, like articulation, requires precise timing and co-ordination of
diverse muscle groups. One reason that spelling, the third criterial skill, is so
resistant to remediation (Thomson, 1984) is that it requires the simultaneous use of
phonological skill and of motor output.
However, this causal chain was initially purely speculative. In order to move
forward in our understanding of the cerebellar impairment hypothesis, it is necessary
to examine components of the causal chain, in order to identify deficits in
components of skill.
3.1 The Hypothetical Causal Chain: Articulation
In a preliminary investigation of the potential role of the cerebellum in articulation

in dyslexia, we tested members of our existing panel on components of the
articulation process, in order to determine whether the known deficits are
predominantly phonological in nature, based on motor skill deficits or on some
combination of the two.
There is evidence that in addition to their known problems in phonological
processing, children with dyslexia show reduced speed of articulation. Note for
example, that in our primitive skills research we had found deficits in articulation
speed in our panel of dyslexic subjects, in repetition of the single words bus,
monkey and butterfly (Fawcett & Nicolson, 1995a, b). Children with dyslexia
experience well-documented problems in phonological access to the lexicon and
thence the motor program for articulation (Snowling et aI., 1986). Nevertheless,
theoretically it was not clear whether or not problems arise specifically in speech
production or in motor planning in dyslexia, whether such deficits are found only
with more complex stimuli, or whether it is only the need for speeded repetition
which leads to errors, and results in deficient performance.
The traditional explanation of deficits in articulation tasks for children with
dyslexia lies in output phonology (Snowling, 1987; Snowling & Hulme, 1994) based
on problems in constructing a novel motor program for words not previously
encountered. However, the components of fluent articulation include both speed and
accuracy, and it remains unclear from the literature whether deficits are based
predominantly on motor skill deficits in the rate or the accuracy of articulation, or
both. This issue has become one of considerable theoretical significance, in that
different theories make different predictions. In terms of articulation, the Cerebellar
impairment hypothesis predicts that there will be deficits, not only in speed of single
word repetition, but also in speech gestures. These would derive from mild in-co-
ordination of the phonatory equipment (such as the tongue and lips) which produce
the speech gestures (see the work of the Haskins lab, e.g. Saltzman, 1995).
In short, therefore, cerebellar impairment would almost certainly give rise to
articulatory difficulties, and thence to phonological problems - see Heilman,
Voeller, and Alexander (1996), Snowling and Hulme (1994) for advocacy of the
latter link. Furthermore, cerebellar deficit would lead to slowed central processing
speed and deficits in motor skill, but not necessarily to sensory processing speed
deficits. Both the "pure" phonological deficit and speed of processing deficit would
presumably predict deficits only in planning and loading the speech gesture, whereas
the cerebellar impairment would predict deficits in both gesture planning and
articulation.
The study reported here was designed to establish the relative contributions to
this delay of speed of motor planning versus speed of articulatory gesture
production. Two groups of children with dyslexia, mean ages 13 and 16 years,
participated together with two groups of normally achieving children matched for
age and IQ, with 33 participants in total. Participants were asked to articulate
repeatedly, as fast as they could, an articulatory gesture or sequence (the rapid
repetition task). This allowed us to analyze the data without the error data, which has
clouded interpretation of earlier repetition tasks in dyslexia (Wolff, Cohen, & Drake,
1984, 1990; Snyder & Downey, 1995). The waveform generated was analyzed in
two ways; the time per gesture excluding inter-articulatory pauses (articulatory
duration); and the mean time including the pauses (gesture duration). Tests included
the single gestures Ipl, Itl, and Ik/ together with the sequence "putuku". This allowed
us to dissociate the time for a single repetition from the time to restart the motor
program.
The analysis showed that dyslexic groups were significantly slower on all tests,
but no age effects were found. Effect sizes for the dyslexic group were comparable
to those obtained for reading. Deficits were greater in relative magnitude for the
gestures with planning. Moreover, for the children with dyslexia, the gesture
duration was 30 - 50% longer than the basic articulation time. These results are
comparable with Wolff's studies, which also included errors, but in this case for
error-free performance. The results suggest that children with dyslexia have
significant problems in articulation, not only in motor planning, but also in the
speeded production of single articulatory gestures. This provides further support for
the cerebellar impairment hypothesis, with a pattern of results, which would not be
directly predicted by other theories of dyslexia (see Fawcett & Nicolson, 2000 for
full details). It should be stressed however, that these results are not incompatible
with any of the other theories, it is more that these theories should now be
augmented to incorporate motor output difficulties.
3.2 Conclusions
To summarize, research on primitive skills (Nicolson & Fawcett, 1994b, 1995) has
shown deficits in phonological skill, naming speed, motor skill and balance in
children with dyslexia. These deficits are characterized as problems in skill
automatisation, normally masked by the process of conscious compensation
(Nicolson & Fawcett, 1990). However, these symptoms of "general automatisation
276 A. 1. FAWCETT
deficit" explain neither the cause nor the specific pattern of difficulties. Its
involvement in the above skills and in skill automatisation suggested the cerebellum
as a natural focus for further dyslexia research. The cerebellar impairment
hypothesis provides a principled account of the qualitative aspects of the data; a
reasonable account of the precise quantitative nature of the effects; and has predicted
unsuspected deficits in time estimation, in clinical tests of co-ordination and muscle
tone, and in a PET study of motor learning. It provides a natural and principled
explanation (in terms of impaired articulatory fluency) for phonological deficits; it
explains difficulties in handwriting in terms of motor skill; and it explains
difficulties in reading and spelling in terms of reduced phonological awareness and
difficulties in automatising sub-skills, together with difficulties in error elimination.
In a preliminary test of the causal chain, precisely as predicted by the cerebellar
impairment hypothesis, deficits were found not only in the production of single
speech gestures, but also in motor planning. This provides support for the
hypothetical causal chain from the brain to behaviour in dyslexia.
We consider that investigation of this hypothetical route of skill acquisition from
infancy to school might well provide a fruitful research agenda for dyslexia. We are
currently undertaking a series of pilot studies, examining aspects of language and
motor skill acquisition in infants and nursery children, comparing the performance
of children from non-dyslexic families, with children with a family history of
dyslexia, who have around a 50% risk of developing dyslexia, in order to investigate
further the relative contribution of the three main theories of dyslexia to our
understanding of development in the pre-school child.
4. ACKNOWLEDGEMENTS
This research was conducted at the Department of Psychology, University of
Sheffield, with funding from the Leverhulme Trust and the Medical Research
Council. The research program was directed by Professor Rod Nicolson, ~hose role
in this research is acknowledged with gratitude.
5. AFFILIATIONS
Dr. Angela Fawcett,

Department of Psychology, University of Sheffield, Western Bank,
Sheffield S10 2TP, UK
e-mail: a.fawcett@sheffield.ac.uk
6. NOTES
It has been suggested (e.g., Shaywitz, Shaywitz, Fletcher, & Escobar, 1990) that there are equal
numbers of males and females with dyslexia, and that the gender imbalance is a referral bias. Miles,
Haslum and Wheeler (1998) argue that the equation of poor reading with dyslexia should be avoided,
recommending a more clinical definition including spelling and items from the Bangor Dyslexia Test
(Miles, 1982, 1997).
Dyslexic children were located via the local Dyslexia Institute or the local branch of the British
Dyslexia Association. No screening or selection was undertaken, apart from checking that they met
our criteria for dyslexia, and were willing to undertake testing on a long-term basis. Subjects
participated with fully informed consent.
For example, for the D16 group the data for blindfold balance for each subject were normalized by
obtaining the difference of that subject's blindfold balance score from the mean blindfold balance
score for group C16, and then dividing this difference by the standard deviation of the C16 group for
blindfold balance. Groups D16 and C16 were normalized relative to C16, groups D12 and C12 were
normalized relative to C12, and groups C8 and D8 were normalized relative to CS.
Note that Levinson's findings are largely based on individual case studies, and that his findings have
been questioned (e.g., Silver, 1987).
7. REFERENCES
Ackerman, P. T., Dykman, R. A, & Gardner, M. Y. (1990). ADD students with and without dyslexia
differ in sensitivity to rhyme and alliteration. Journal of Learning Disabilities, 23, 279-283.
Allen, G., Buxton, R. B., Wong, E. c., & Courchesne, E. (1997). Attentional activation of the cerebeUum
independent of motor involvement. Science, 255, 1940-43.
Anderson, 1. R. (1982). Acquisition of cognitive skiU. Psychological Review, 89, 369-406.
Badian, N. A. (1984). Reading disability in an epidemiological context: Incidence and environmental
correlates. Journal of Learning Disabilities, 17, 129-136.
Boder, E. (1973). Developmental dyslexia: a diagnostic approach based on three atypical speUing-reading
patterns. Developmental Medicine and Child Neurology, 15, 663-687.
Bradley, L., & Bryant, P. E. (1978). Difficulties in auditory organisation as a possible cause of reading
Bradley, L., & Bryant, P. E (1983). Categorising sounds and learning to read: A causal connection.
Nature, 301, 419-421.
Bryant, P., & Goswami, U. (1986). Strengths and weaknesses of the reading level design. Psychological
Bulletin, 100, 101-103.
Cohen, J. (1969). Statistical power analysis for the behavioral sciences. New York: Academic Press.
Denckla, M. B., & Rudel, R. G. (1976). Rapid "Automatized" naming (R.A.N.). Dyslexia differentiated
from other learning disabilities. Neuropsychologia, 14, 471-479.
Dow, R. S., & Moruzzi, G. (1958). The physiology and pathology of the cerebellum. Minneapolis:
University of Minnesota Press.
Eccles, 1. C., ito, M., & Szentagothai, J. (1967). The cerebellum as a neuronal machine. New York:
Springer-Verlag.
Fawcett, A. 1. (1990). A Cognitive Architecture of Dyslexia. PhD Thesis, University of Sheffield.
Fawcett, A. J., & Nicolson, R. I. (1995a). Persistence of phonological awareness deficits in older children
with dyslexia. Reading and Writing: An Interdisciplinary Journal, 7, 361-376.
Fawcett, A. J., & Nicolson, R I. (1995b). Naming speed and dyslexia. Journal of Learning Disabilities,
27,641-647.
Fawcett, A 1., & Nicolson, R. I. (1995c). Persistent deficits in motor skiU for children with dyslexia.
Journal of Motor Behavior, 27, 235-241.
Fawcett, A J., & Nicolson, R. I. (2000). Children with dyslexia show deficits in speech motor planning
and in single gesture articulation speed. Submitted to Developmental Neuropsychology.
Fawcett, A J., Nicolson, R. I., & Dean, P. (1996.) Impaired performance of children with dyslexia on a
range of cerebeUar tasks. Annals of Dyslexia, 46, 259-283. .
Fitts P. M., & Posner, M. I. (1967). Human performance. Belmont, CA: Brooks Cole.
Fowler, A. (1991). How early phonological development might set the stage for phoneme awareness. In
S. Brady, & D. Shankweiler (Eds.), Phonological Processes in Literacy (pp. 97-117). Hi11sdale, NJ:
Lawrence Erlbaum Associates.
Frank, 1., & Levinson, H. N. (1973). Dysmetric dyslexia and dyspraxia: hypothesis and study. Journal of
American Academy of Child Psychiatry, 12, 690-701.
Frith, U. (1997). Brain, mind and behaviour in dyslexia. In C. Hulme & M. Snowling (Eds.), Dyslexia:
Biology, cognition and intervention (pp.1-20). London: Whurr.
Fulbright, R. K., Jenner, A R., Mencl, W. E., Pugh, K. R., Shaywitz, B. A, Shaywitz, S. E., Frost, S. 1.,
Skudlarski, P., Constable, R. T., Lacadie, C. M., Marchione, K. E., & Gore, 1. C. (1999). The
cerebeUum's role in reading: A functional MR imaging study. American Journal of Neuroradiology,
20,1925-1930.
Galaburda, AM., Menard, AM., & Rosen, G.D. (1994). Evidence for aberrant auditory anatomy in
developmental dyslexia. Proceedings of the National Academy of Sciences of the USA, 91, 8010-
8013.
Glickstein, M. (1993). Motor skiUs but not cognitive tasks. Trends in Neuroscience, 16, 450-451.
278 A. J. FAWCETT
Hatcher, P. J., Hulme, C., & Ellis, A. W. (1994). Ameliorating early reading failure by integrating the
teaching of reading and phonological skills: the phonological linkage hypothesis. Child Development,
65,41-57.
Heilman, K. M., Voeller, K., & Alexander, A. W. (1996). Developmental dyslexia: A motor-articulatory
feedback hypothesis. Annals of Neurology, 39, 407-412.
Holmes, G. (1917). The symptoms of acute cerebellar injuries due to gunshot injuries. Brain, 40, 461-
535.
Holmes, G. (1939). The cerebellum of man. Brain, 62, 1-30.
Ivry, R. B., & Keele, S. W. (1989). Timing functions of the cerebellum. Journal of Cognitive
Ito, M. (1984). The cerebellum and neural control. New York: Raven Press.
Ito, M. (1990). A new physiological concept on cerebellum. Revue Neurologique (Paris), 146, 564-569.
Jenkins, I. H., Brooks, D. 1., Nixon, P. D., Frackowiak, R. S. J., & Passingham, R. E. (1994). Journal of
Jorm, A. F., Share, D. L., Mclean, R., & Matthews, D. (1986). Cognitive factors at school entry
predictive of specific reading retardation and general reading backwardness: A research note. Journal
of Child Psychology and Psychiatry and Allied Disciplines, 27, 45-54.
Krupa, D.1., Thompson, 1.K., & Thompson, R.F. (1993). Localization of a memory trace in the
mammalian brain. Science, 260, 989-991.
Leiner, H.C., Leiner, A.L., & Dow, R.S. (1989). Reappraising the cerebellum: what does the hindbrain
contribute to the forebrain? Behavioural Neuroscience, 103, 998-1008.
Levinson, H. N. (1990). The diagnostic value of cerebellar-vestibular in detecting learning disabilities,
dyslexia and attention deficit disorder. Perceptual and Motor Skills, 71, 67-82.
L1inas, R. (1993). Is dyslexia a dyschronia? Annals of the New York Academy of Science, 682, 48-56.
Locke, J. L. (1983). Phonological acquisition and change. New York: Academic Press.
Locke, J. L. (1993). The child's path to spoken language. Cambridge, MA: Harvard University Press.
Lundberg, I., Frost, J., & Petersen, O. P. (1988). Long term effects of a pre-school training program in
phonological awareness. Reading Research Quarterly, 28, 263-284.
Miles, T. R. (1982, 1997). The Bangor Dyslexia Test. Wisbech, Cambridge: Learning Development Aids.
Miles, T. R. (1990). On determining the prevalence of dyslexia. In M. Snowling, & M. Thomson (Eds.),
Dyslexia: Integrating Theory and Practice (pp.144-153). London: Whurr.
Miles, T. R., Haslum, M. N., & Wheeler, T. J. (1998). Gender Ratio in Dyslexia. Annals of Dyslexia, 36,
27-55.
Auditory processing or phonological coding? Journal of Experimental Child Psychology, 64, 199-
231.
Nicolson, R. I., & Fawcett, A. 1. (1990). Automaticity: A new framework for dyslexia research.
Cognition, 35, 159-182.
Nicolson, R. I., & Fawcett, A. 1. (1994a). Reaction Times and Dyslexia. Quarterly Journal of
Nicolson, R. I., & Fawcett, A. J. (1994b). Comparison of deficits in cognitive and motor skills among
children with dyslexia. Annals of Dyslexia, 44, 147-164.
Nicolson, R. I. & Fawcett, A. 1. (1994c). Long-term skill learning in dyslexic children In A. 1. Fawcett, &
R.1. Nicolson (Eds.), Dyslexia in children: MultidiSciplinary perspectives (pp. 157-191). Hemel
Hempstead, UK: Harvester Press.
Nicolson, R. I., & Fawcett, A. 1. (1995). Dyslexia is more than a phonological disability. Dyslexia: An
international journal of research and practice, 1, 19-37.
Nicolson, R. I., & Fawcett, A. J. (1999a). Developmental Dyslexia: the role of the cerebellum. Dyslexia:
An International Journal of Research and Practice, 5, 155-177.
Nicolson, R. I., & Fawcett, A. 1. (1999b). Developmental Dyslexia: the role of the cerebellum. In I.
Lundberg, F. E. Tonnessen, & I. Austad (Eds), Dyslexia: Advances in theory and practice (pp. 173-
197). Netherlands: K1uwer.
Nicolson, R. I., & Fawcett, A. 1. (2000). Long-term learning in dyslexic children European Journal of
Cognitive Psychology, 12, 357-393.
Nicolson, R. I., Fawcett, A. J., & Dean, P. (1995). Time estimation deficits in developmental dyslexia:
Evidence for cerebellar involvement. Proceedings of the Royal Society, 259, 43-47.
Olson, R. K., Wise, B. W., & Rack, J. P. (1989). Dyslexia: Deficits, genetic etiology and computer based
remediation. Irish Journal of Psychology, 10, 594-508.
Passingham, R. E. (1975). Changes in the size and organization of the brain in man and his ancestors.
Brain Behavior and Evolution, 11, 73-90.
Rack, J. (1985). Orthographic and phonetic coding in normal and dyslexic readers. British Journal of
Ramsay, D. S. (1984). Onset of duplicated syllable babbling and unimanual handedness in infants:
Evidence for developmental change in hemispheric specialization? Developmental Psychology, 20,
64-71.
Saltzman, E. L. (1995). Dynamic and co-ordinate systems in skilled sensorimotor activity. In R. F. Port,
& T. van Gelder (Eds.), Mind as motion: explorations in the dynamics of cognition (pp. 149-173).
Cambridge, MA: MIT Press.
Schmahmann, 1. D. (1998). Dysmetria of thought: clinical consequences of cerebellar dysfunction on
cognition and affect. Trends in Cognitive Sciences, 9, 362-371.
Seymour, P. H. K. (1987). Individual cognitive analysis of competent and impaired reading. British
Journal of Psychology, 78, 483-506.
Shankweiler, D., Crain, S., Katz, L., Fowler, A E., Liberman, A M., Brady, S. A, Thornton, R.,
Lundquist, E., Dreyer, L., Fletcher, 1. M., Stuebing, K. K., Shaywitz, S. E., & Shaywitz, B. A
(1995). Cognitive profiles of reading-disabled children - comparison of language-skills in phonology,
morphology, and syntax. Psychological Science, 6, 149-156.
Shaywitz, S. E., Shaywitz, B A, Fletcher, 1. M., & Escobar, M. D. (1990). Prevalence of reading-
disability in boys and girls - results of the Connecticut Longitudinal Study. Journal of the American
Medical Association, 264, 998-1002.
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human information processing: II.
Perceptual learning, automatic attending and general theory. Psychological Review, 84, 127-190.
Silver, L. B. (1987). The "magic cure": A review of the current controversial approaches for treating
learning disabilities. Journal of Learning Disabilities, 20, 498-505.
Snowling, M. (1987). Dyslexia: A cognitive developmental perspective. Oxford: Blackwell.
Snowling, M. J., Goulandris, N., Bowlby, M. & Howell, P. (1986). Segmentation and speech perception
in relation to reading skill: a developmental analysis. Journal of Experimental Child Psychology, 41,
487-507.
Snowling, M., & Hulme, C. (1994). The development of phonological skills. Philosophical Transactions
of the Royal Society of London B, 346, 21-27.
Snyder, L., & Downey, D. (1995). Serial rapid naming skills in children with reading disabilities. Annals
of Dyslexia, 45, 31-50.
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.
Stein, J. F. (1994). Developmental dyslexia, neural timing and hemispheric lateralisation. International
Journal of Psychophysiology, 18, 241-249.
Stein, J. F., & Glickstein, M. (1992). Role of the cerebellum in visual guidance of movement.
Physiological Reviews, 72, 972-1017.
Stein,1. F, & Walsh, V. (1997). To see but not to read: the magnocellular theory of dyslexia. Trends in
Tallal P. (1985). Auditory temporal perception, phonics and reading disabilities in children. Brain and
Language, 9, 182-198.
Tallal, P., Miller, S., & Fitch, R. H. (1993). Neurobiological basis of speech - a case for the pre-eminence
of temporal processing. Annals of the New York Academy of Sciences, 682, 27-47.
Thach, W. T. (1996). On the specific role of the cerebellum in motor learning and cognition: Cues from
PET activation and lesion studies in man. Behavioral and Brain Sciences, 19, 411-431.
Thomson, M. E. (1984). Developmental dyslexia: Its nature, assessment and remediation. London:
Edward Arnold.
Vellutino, F. R. (1979). Dyslexia: Theory and research. Cambridge, MA: MIT Press.
Wechsler, D. (1976). Wechsler Intelligence Scale for Children Revised (WISC-R). Slough, UK: NFER.
Wechsler, D. (1992). Wechsler Intelligence Scale for Children Third edition, UK, Sidcup, Kent: The
Psychological Corporation.
Wolf, M. (1991). Naming speed and reading: The contribution of the cognitive neurosciences. Reading
Research Quarterly, 26, 123-141.
Wolff, P. H., Cohen, c., & Drake, C. (1984). Impaired motor timing control in specific reading
retardation. Neuropsychologia, 22, 587-600.
World Federation of Neurology. (1968). Report of research group on dyslexia and world illiteracy.
Dallas: WFN.
E. WITRUK, C. S.-H. HO, & U. SCHUSTER
WORKING MEMORY IN DYSLEXIC CHILDREN -

HOW GENERAL IS THE DEFICIT?
Abstract. The important role of working memory functions for reading and writing will be discussed
together with the proven impairments of working memory performance in dyslexic individuals for visual
and auditory presentation of stimuli with different paradigms and types of material. In the following
paper, three studies will be introduced. The conceptual differentiation of these studies is based on the
Baddeley and Hitch model (1974), with its specific modalities for incoming and maintaining information
as well as on Cowan's model (1995), with regard to automatic versus controlled executive functions
during the generation of response and action. Thus, the central issue of the following studies is to
investigate the importance and extent of assumed working memory impairments in dyslexic children with
the focusing on the generality versus the specificity of these impairments. All three studies are conducted
with samples of third grade dyslexic and non-dyslexic children. With the obtained results, we notice a
dependency of working memory performance in dyslexic children on the training and language system
(experiment 1), the specific type of modality (experiment 2) and on a specific kind of material
(experiment 3). Therefore, the results of the three experiments cannot support the assumption of a general
modality-, material-, and language-independent deficit in dyslexic individuals.
1. FUNCfIONAL PRINCIPLES OF WORKING MEMORY AND THEIR

RELEVANCE FOR IMPAIRMENTS OF READING AND WRITING
1.1 Cognitive Approaches to Working Memory

The complex mental abilities of reading and writing require firstly, perceptive and
cognitive steps of processing which are based on each other, and secondly,
maintaining and integrating functions of working memory with access to long term
memory. The analogy of working memory to a "workshop" is obvious: Incoming
perceptual information must be maintained for a short time and must be ready for
active processing. At the same time, processes of searching and comparing start in
long term memory, whose results are also activated in working memory and
integrated into the incoming perceptual information.
We concur with Klix (1984), that procedures of working memory take effect in
two ways:
1. processing of perceptual contents and
2. processing of retrieved contents from long term memory.
Working memory starts information and operation searches in long term memory
and processes copies of retrieved information. Retrieved automatic programs and
controlled activated procedures and strategies are applied to copies in working
memory. The function of working memory comprises not only the storage of
incoming information and interim results but also their further processing in a sense

282 E. WITRUK, C. S.-H. Ho, & U. SCHUSTER
of working out an algorithm. Therefore, working memory maintains a fundamental

meaning for reading and writing processes.
Mechanisms which enable the maintenance of incoming and retrieved
information in working memory are described by Baddeley and Hitch (1974) as
well as by Gathercole and Baddeley (1993). They proceed from three components:
A modality-unspecific and capacity-limited Central Executive which
realises retrieval from and transfer to long term memory as well as
processing and storage of incoming information,
Two modality-specific and also capacity-limited subsystems, divided and
under the command of the Central Executive, which are specialized in the
processing of auditory and visual information (phonological loop and
visual-spatial sketchpad). Modality specific information can be maintained
available to further processing in the Central Executive by automatic
rehearsal processes.
Goldman-Rakic (1998) and Petrides (1994) differentiate the visual-spatial sketchpad
between the storage of spatial information (where-pathway) and the storage of
object information (what-pathway). These pathways are realized physiologically as
different neural networks.
Functions of the Central Executive can be differentiated according to Cowan
(1995) by the proportion of automatic and controlled executive functions.
Automatic functions can, be realized to a great extent unconsciously with low mental
effort, as for instance, with automatic word recognition. On the other hand,
controlled executive functions demand a conscious plan for action steps, as for
instance when reading a word by grapheme-phoneme-correspondences and the
following synthesis.
1.2 Neuro-Psychological Approaches to Working Memory
Neuro-anatomical foundations of working memory functions regarding the Central

Executive are shown by Goldman-Rakic (1998) in the prefrontal cortical areas in
apes, which form an interactive network with posterior cortical areas. Control and
coordinate functions of storage and processing are realized by this network.
Modality-specific performances of storage are assumed in other cortical areas.
Goldman-Rakic states that spatial information (where-pathway) is stored in
posterior dorsal cortical areas, while object information (what-pathway) is stored in
posterior ventral areas. Similar network locations for spatial and object information
Were found by Mecklinger and Mueller (1996) in humans. Roesler, Bajric, and Hei!,
(1997) gave evidence with neuro-psychological data for the domain specificity of
performances of working memory for processing visual information. In a first study,
they found activity in the parietal cortex for retrieval of spatial information from the
long term memory while retrieval of verbal information was combined with activity
in the left-frontal cortex. A further study showed that topographies of arithmetic and
semantic working memory performance (N400 effect) widely correspond and show
activity in centro-parietal regions. During tasks which demand mental rotation of
letters, a negativation in the parietal cortex could be shown. The amplitude of the
negative curve enlarges with increasing mental rotation effort. Syntax processing
WORKING MEMORY IN DYSLEXIC CHILDREN 283
working memory performances could be proven with a further experiment

conducted by Roesler et al. (1997) showing phasical negativation over the left
anterior cortex.
In addition, storage and processing performances of auditory information in the
phonological loop seem to be domain specifically realized according to the results
of Schroeger (1995). The topographic differentiation between activated brain
regions regarding the direction, frequency, volume and periodic of the auditory
stimulus are provable with electro-physiological data.
1.3 Working Memory in Dyslexia
Reading and writing are realized based on storage and processing performances for
different types of modality as well as comparison and retrieval processes in long
term memory and modality shifts. Impairments of working memory performance in
dyslexic children and adults were proven for visual and auditory presentation of
stimuli with different paradigms and types of material.
Regarding deficits of working memory in the visual-spatial loop, differentiated
results are available. So and Siegel (1997) found deficits for Chinese dyslexics
during visual working memory tasks (free visual reproduction of character lists with
and without phonological, visual and semantic similarities). Ellis (1981) reported
four visual matching experiments based on the Posner Paradigm with different
material, in which he was not able to find deficits for dyslexics if the two stimuli
were not nameable. Significant deficits for dyslexics were shown if the visual
stimuli were phonologically similar letters. He interpreted these results as naming
deficits. Vellutino's findings (1987) also speak against a general deficit of the visual
working memory. His dyslexic children were able to reproduce unknown Hebrew
words and letters just well as normal reading children in the control group. If the
word list was in English, the dyslexic children performed significantly poorer than
the control group. Vellutinos interpretation refers to a deficit of dyslexics during
storage and recall of linguistic information. Likewise Bamea, Lamm, Epstein, and
Pratt (1994) mainly found deficits for Hebrew speaking dyslexic children with
series of lexical, visual stimuli.
During visual matching tasks Willows, Kruk, and Corcos (1993) found deficits
of dyslexic children with letters from the - to them unknown- Hebrew alphabet.
These deficits in accuracy and speed were stronger in 6-year-old children than in 8-
year-olds. Compensation effects for deficits of visual working memory could be
shown in the study by Witruk and Rosendahl (1999) for visual matching tasks and
for visual serial recall tasks. For these visual working memory tasks we found
significant adaptations to the control group in a longitudinal and cross-section
comparison between 7- and 9-year old dyslexic children corresponding with the
increasing age of the children. These adaptations could not be shown for
phonological serial recall performances.
Regarding the deficits of the phonological loop of the working memory, the
results are more clear and less differentiated. The most often used paradigm is the so
called memory span for numbers, words, and pseudo-words. Deficits in
phonological abilities and of phonological working memory in dyslexic Canadian
children are described by Siegel and Linder (1984). Ho, Law and Ng (2000) and Ho
and Lai (2000) were able to validate these phonological deficits for Chinese
dyslexic children. According to Everatt et. al. (2001) phonological working memory
deficits on sequential information (such as in digit span tasks) could be the root-
cause of some other deficits and they are evident across child and adult populations.
The findings indicate only a small, insignificant difference between dyslexic and
normal groups when visual or spatial sequential processing is required. Gathercole
and Baddeley (1993) found delays of development regarding articulation speed,
rehearsal of non-words, and memory span for words in 8-, 11- and 15-year-old
dyslexic children. Phonological deficits which could be shown in 8- and 11-year-old
dyslexic children are not provable in 15-year-old dyslexic children. Thus, with
regard to phonological working memory, one can call it a later on-set in the
tendency of compensation. From this point of view, results that refer to deficits of
phonological working memory before the age of 15 can be interpreted as delays in
the development.
Proof of deficits in dyslexics in relation to Central Executive functions may only
be found in a few investigations. Schneider (2001) could show a stronger activation
of the frontal lobe in dyslexic children during mental rotation and sound connecting
tasks. She interpreted these results as a stronger involvement of the Central
Executive in dyslexic children on the basis of an inefficient automatization. The
tasks used by Siegel and Ryan (1989 a, b) involved executive functions during word
recognition after sentence completions and counting. They found generalized
working memory deficits in dyslexic children· (age 7-13), while children with
arithmetic deficits have only a deficit in processing numerical information. The
children with attention deficit disorders did not have working memory deficits.
Results regarding the deficits of the phonological loop seem to be present with
relatively high consistency. Deficits of visual-spatial loop appear to depend strongly
on the types of material used. Studies in which visual but nameable stimuli were
used could be related to phonological decoding and to the phonological loop.
1.4 Integration a/Working Memory Deficits into a Dyslexia Model
The assumed and in some studies proven working memory deficits in dyslexics can
be integrated into the multilevel model of dyslexia (modified by Witruk & Schuster,
1997 on the basis of Valtin, 1984) on the level of secondary causes, which are based
on primary causes in the sense of biological deficits. These biological causes were
discussed on the basis of a strong genetic determination (Grigorenko, 2001) and
some neuro-anatomic peculiarities in dyslexics of the magnocellular pathway
(Galaburda & Livingstone, 1993; Stein & Walsh, 1997) and of the Cerebellum
(Fawcett & Nicolson, 2001). Based on these results, biochemical (Robinson, 2001)
and neuro-physiological (Kujala, this volume; Schneider, 2001) anomalies in
dyslexics will be interpreted.
The level of the secondary causes stands out and implies psychic representations
For the visual system, lower temporal and spatial contrast sensitivity, as well as
longer duration of visible persistence as early, basic visual deficit in dyslexics
(Talcott, this volume; Witruk, Lachmann, & Graeve, in preparation), which are
linked to abnormalities in the magnocellular pathway and its interaction with the
parvocellular pathway, could be shown. Working memory performances involve
early, basic, (low level) and later, complex sensory (high level) representations and
their connections to the semantic networks of the long-term-memory. The
interdependence between early elementary auditory and visual perception
performances (e.g. contrast sensitivity and persistence) and later highly complex
processes of working memory is still unsettled for experimental dyslexia research.
Because of this, the question is if and how strongly deficits of working memory are
affected by deficits of early perception processes. We should consider that some
studies cannot verify the magnocellular/transient deficit theory, for example
Walther-Mueller (1995) and Spinelli et al. (1997), or can show that the presence of
a magnocellular deficit depends on the type of dyslexia for example Borsting et al.
(1996).
Primary and secondary causes are the basis of the development of the primary
symptoms in the sense of the individual pattern of failures in reading and writing. A
lot of studies prove the typical reading and writing errors of dyslexics (for example
the reversal errors (Lachmann, this volume) and the typical eye movement patterns
(Fischer & Hartnegg, 2000; Witruk, 1993; Witruk & Staats, 1983). In time, these
latent failures and the responses from the children's environments lead to a vicious
cycle of secondary symptoms like anxiety, blocking, avoidance, compensation and
lowering of motivation as explained by Betz and Breuninger (1993).
2. AlMS AND QUESTIONS OF THREE EXPERIMENTS
Reading and writing are realized by highly complex and adaptive processes of
perception, and attention as well as various short and long term memory functions.
Thus, the central issue of the following studies is to investigate the importance and
extent of assumed working memory impairments in dyslexic children. The
conceptual differentiation is based on the Baddeley and Hitch model (1974) with its
specific of modalities for incoming and maintaining information and further based
on Cowan's model (1995) with regard to automatic versus controlled executive
functions during generation of response and action.
Three studies will be introduced whose questions aim at dependency on
language system, dependency on types of modality, and the type of material as well
as the influence of controlled versus automatic executive functions on working
memory performances. All three studies are conducted with samples of third grade
dyslexic and non-dyslexic children suitably matched for intelligence, age, and
grade. The severity degree of dyslexia of the children was measured by the dyslexia
test system of Weigt (1980, 1994) and verified as middle and severe degrees. Well-
tried paradigms were transferred into experimental designs: adaptive working
memory tasks series (experiment 1), Sternberg tasks (experiment 2) and same
different tasks series (experiment 3).
286 E. WITRUK, C. S.-H. HO, & U. SCHUSTER
3. LANGUAGE DEPENDENCY OF WORKING MEMORY PERFORMANCES

IN DYSLEXIC AND NON-DYSLEXIC CHILDREN
We investigated the influence of the language system on working memory

performances in dyslexic and non-dyslexic children in Hong Kong and in Leipzig
with adaptive working memory experiments (Witruk & Ho, 1999). The aim is to
analyze the impact of language on the development of working memory
performances. The modern Chinese language contains multimorphemic words. That
means that 60 % of the words have two characters, where many component
characters also represent words on their own (Hoosain, 1991). The mono-syllabic
characters were spoken with a variety of nine tones for the Cantonese dialect,
spoken in Hong Kong. About 60% of Chinese characters have homophones and
almost a quarter of them have six or more homophones. There are many
homophones at the character level, but fewer at the word level. The differentiation
of the homophones, the recognition of their meaning arises from one or various
syllables simultaneous to the realization of the syntactic structure of a sentence. This
places special challenges to the functions of the working memory, in order to reach
disambiguation in the oral language. The written Chinese language as a logographic
script contains characters written without word boundaries. Therefore, the Chinese
reader has to monitor the semantic relations of the character sequence more than in
alphabetic languages. Only 5 % of the characters are pictographs or ideographs, 5 %
are compound ideographs allowing the direct semantic access. 90 % of the
characters are ideophonetic compounds containing a radical and a phonetic.
Therefore, the decoding of Chinese script requires phonological skills and a high
demand on the phonological loop of the working memory. Phonological deficits are
distinguished also for Chinese dyslexic children as Ho and Bryant (1997) could
show.
For the German orthography, we have to consider a middle degree of regularity
of grapheme-phoneme-relation. About 40 phonemes are represented by 85
graphemes each consisting of one, two or three letters (Valtin, 1989). The basic
principle of the German orthography is phonological, but it will be added by
semantical principles (same writing of morphemes, different writing of
homophones) and the syntactical principles (capitalization,punctuation, conjunction
and declination). For German dyslexics the relevance of phonological working
memory deficits could also be shown (Witruk & Rosendahl, 1999).
The basic aim of the experiment is to compare the visual and auditory working
memory performances in dyslexic and non-dyslexic children from different
language and culture regions.
The developed adaptive working memory tasks permit a continuing adaptation
of the degree of task difficulty on the individual current achievement. Visual and
auditory matching and serial recall performances were analyzed. Each of the four
task sets with increasing levels of difficulty is presented with a duration of five
minutes. After a false response, the participant goes three levels downwards and
thus, the individual limit of performance can be reached without overloading of the
child. Tasks based upon the Same-Different-Paradigm and upon the Serial Recall
Paradigm are presented both visually and auditory.
During part A patterns of dots are used, visually presented at the computer
screen, and will be enlarged by an increasing number of dots. The decision about
same or different requires a matching process to a second pattern of dots. During
part B, series of tones are auditory presented with an increasing number of tones
with high and low pitches. The decision about same or different has to be based on
matching a second series of tones. Part C and part D use serial recall as response
mode. During part C, patterns of lines with increasing number of lines are visually
presented and are to be reproduced by pressing keys. During part D, auditory
presented series of tones with increasing number of tones with high and low pitches
are to be reproduced at the same way. As dependent variables, accuracy parameters
of the working memory like the mean value of highest individual reached level of
difficulty (local maxima), the highest level reached during the experiment and time
parameters of the working memory like reaction and reproduction time for correct
and false responses and the number of processed items were measured.
We compared these working memory performances between dyslexic and non-
dyslexic children in Leipzig and in Hong Kong. The samples were paralleled by
age, grade (third grade) and intelligence. 20 dyslexic children from Hong Kong and
42 dyslexic children from Leipzig as well as 20 non-dyslexic children from Hong
Kong and 79 non-dyslexic children from Leipzig participated in the experiment. An
intelligence test (CFT1) and a reading test (Zuerich Reading Test in Germany) were
carried out.
The results show a significant superiority of the Chinese children in all time
parameters during all four task series (A-D). Their advantages in accuracy are only
significant in tasks with auditory stimulus presentation and the demands for
phonological loop activity (see Figures 1a and 1b).
A statistical discrimination between dyslexic and non-dyslexic children is only
significant for the German samples. Here, the two groups differ strongest during
auditory stimulus presentation and activation of highly controlled processes like
active reproduction of series of stimuli (see Figures 1a and 1b).
The results give evidence for the strong influence of the language system on
working memory performances. The Chinese language system, with its low
differentiation of syllables, makes it necessary for oral communication to modulate
the pitches of syllables and to use context in a large way. The use of context means
that learning the Chinese language is accompanied with a "natural" and highly
efficient training of working memory and therefore no deficits of dyslexia were
discernible in the investigated fields. Our results can be also interpreted by the
generalization of the Chinese superiority effect: Results of Liu, Zhu, and Wu (1992)
show, that the lexical access for Chinese logographs is more directly and quicker
than for alphabetic words, because the logo graphs are more unique in shape or more
discrimable than alphabetic words. The German language system with its multi-
syllabic word constructions and combinations (except homophones) realizes an
unambiguous meaning on the level of words. Understanding spoken German
requires less use of context and thus less maintenance of short units like mono-and
bi-syllabic words. The significant inferiority in performance of German children in
the auditory experiments regarding accuracy and speed of response can be
interpreted as a result of weaker "natural" training through oral communication.
With that weaker training significant deficits of performance for dyslexic children
are extremely visible mostly related to auditory presentation of stimuli and
combined with demand of high controlled reproduction processing.
Reaction Time for Correct Responses
visual presentation auditory presentation
1,5
II'
I
'U
il 1,0
~ I
0,5
non dyslexics dyslexics non dyslexics dyslexics

sample sample
Accuracy Performance (Means of Local Maxima)
visual presentation auditory presentation

10 10
B+. ... _... _-- ---c__ ··_·

sample sampl.
Figure la. Reaction time and accuracy performance on visual and auditory working memory
tasks in dyslexic and non-dyslexic children from Hong Kong (white) and Leipzig (grey).
Reaction Time for Correct Responses
passive matching active reproduction
1,5
'\:J
~ 1,0 +-""",.= .---_. ·1- - .. _. ("",:"",.",:"",,'1-
~
0,5
sample sample
Accuracy Performance (Means of Local Maxima)
passive matching active reproduction
10 10
1
8 r-- ...-
i 6
II 8
I
j
4
2
i illl-

sampl. sampl.
Figure 1b. Reaction time and accuracy performance on matching tasks and active
reproduction tasks in dyslexic and non-dyslexic children from Hong Kong (white) and
Leipzig (grey).
In addition to these language differences, we have to consider cultural
differences in a more general sense. The better memory performance of Chinese
children over that of German children is probably due to the drilling practice in
Hong Kong. Educational practices in Hong Kong place more emphasis on rote
memory and speed drilling. Therefore, the cultural differences (in terms of
educational practices) together with the language differences may contribute to the
performance differences in the two regions.
4. DEPENDENCY OF WORKING MEMORY PERFORMANCE ON TYPES OF

MODALITY AND MATERIAL IN DYSLEXIC AND NON-DYSLEXIC
CHILDREN
The special concern of auditory or phonological perception and maintenance

performance in German dyslexics could be confirmed with our results, obtained by
working memory experiments based on the Sternberg Paradigm (Graentzel & Jaehn,
1997, Witruk & Schuster, 1997). The Sternberg requirements examine searching
processes over the working memory. The time is measured until a decision, whether
or not the target item was in a previously presented list of items. This measured time
span is composed of time for encoding, time for serial matching, time for decision-
making, and time for motor reaction. This highly complex composed time for
response increases according to Sternberg (1975) by 38 ms for one more item in the
presented list. With a negative set (target item not in the list), one can find an
exhaustive search over all items of the list while with a positive set (target item in
the list) one can see a self-terminating search. We examined processes of
maintenance, search and executive functions of working memory during visual and
auditory presentation of words (one syllable with four letters, e. g. TREE), and
pseudo-words (one syllable with four letters, e. g. TARK). The length of the lists
was 2 -5 items. The experiment is based on a four-factorial design in which group
membership, type of material, type of modality, and the length of the lists was
varied. The reaction time and the sum of errors were declared as dependent
variables. The samples of the dyslexic children (n=2l) and of the normal reading
children (n=2l) were paralleled by age, grade (third grade) and intelligence
(measured with CFTl).
The results showed no significant group effects for the reaction time and the sum
of errors during visual presented memory and search tasks, but did show a
significantly higher reaction time and sum of errors for pseudo-words in both
samples.
250 r------------------------,
200
e
,-.
150 127
~
a.
E
a.
100 114 r===ool
~
~
50
o ~------------------------~
auditory words visual
Figure 2a. Dependency of errors in the Sternberg tasks on types of modality (auditory vs.
visual presentation) for words in dyslexic (EC) and non-dyslexic children (CC).
250 ,
200 193
..-
1~187
--
c
'-' 150
..
* ..
'"0
40l
100
96 r==rol
50 ~
0
auditory pseudo-words visual
Figure 2b. Dependency of errors in the Sternberg tasks on types of modality (auditory vs.
visual presentation) for pseudowords in dyslexic (EG) and non-dyslexic children (eG).
Significant deficits of performance for dyslexic children occurred only for words
and pseudo-words with regard to the sum of errors during auditory presentation (see
Figure 2). One can notice just a tendency towards a deficit in the higher reaction
time of the dyslexic children.
Three sub-types of dyslexics can be differentiated in our Sternberg-experiment
on the basis of working memory performance and its specific in modality. The
individual deviation of each dyslexic child from the standard deviation interval of
the reaction times and of the error sums in the control group during visual and
auditory presentation of items, form the basis for this differentiation. We found
dominating visual deficits in 28.6% of the dyslexic children, compared to 38.1 %
with auditory or phonological deficits which mainly affected the phonological loop
of working memory. In 33.3 % of our dyslexics both modalities were equally
impaired, showing a hint of deficits in both the phonological loop and the visual-
spatial sketchpad.
The results supply the evidence for the dominance of auditory or phonological
deficits of working memory, which are based onJhe phonological loop.
5. VISUAL WORKING MEMORY PERFORMANCE IN DYSLEXIC

CHILDREN - DEPENDENCY ON TYPE OF MATERIAL
The deficits in the area of visual working memory seem to be more subtle and more
dependent on the type and the complexity of the material. They could only affect
one sub-type of dyslexics, which would make some controversial results more
understandable. The visual matching experiment is based upon the same-different
paradigm aimed at the analysis of the influence of kind and complexity of the
stimuli and the influence of naming access in long term memory on temporal and
accuracy parameters for memory matching performance in dyslexic and non-
dyslexic children. The central question of the study is whether dyslexia is associated
with a material nonspecific deficiency in encoding, maintenance, and matching of
visual stimuli, or whether deficiencies in visual memory matching of dyslexics only
occur in naming access or when the stimulus materials exceed an intensity criterion
in the level of structure and complexity.
With the third experiment (Witruk, Lachmann & Graeve, in preparation), we
examined visual matching performance of dyslexic (n=24) and normal reading
(n=18) children for two successively presented visual stimuli, which are different in
type of material (letters, five-dot-patterns developed by Garner and Clement (1963),
patterns of lines) and in complexity. It was based on same-different-tasks with a
presentation time of 500 ms for the first stimulus, and also a 500 ms interstimulus
interval until the second stimulus. Reaction time and error rate were measured as
dependent variables. The results showed a diverging picture regarding the
parameters for accuracy and speed (see Figure 3a, b).
A highly differentiated and non-general deficit performance profile of dyslexic
children was found. For the accuracy parameter, significantly higher error rates
were seen with dyslexic children for letters and dot patterns but not for line patterns.
For the time parameter, the group effect was not significant, but there was a
tendency towards shorter reaction times for dyslexic children. The model fit of the
time parameter in dependence from the degree of complexity (equivalence size) for
both groups, succeeded perfectly in adapting the reaction times to the complexity of
the matching tasks for dot patterns and line patterns. The quality of the model fit
was the worst for dyslexic children concerning letters. This could be explained by
the reduced automation of recognizing letters. The naming access of dyslexic
children seems to be more difficult.
1000
I 900
~c
.2
..'"
ts
a:
c 800 Material
~'" o Letter.
~Dotpatterna
700 .l.i1epattern.
dyslexic
Sample
Figure 3a. Mean reaction times in dyslexic and non-dyslexic children for different types of
material.
This is also observable in the speed-accuracy-trade off. The material and

complexity specificity of the performance profiles of dyslexics suggests an
interaction of relatively specific qualities for visual matching performance.
Material
DLe\Ie,·
IZ:I Dot patterns
• SIring patterns
Sample
Figure 3b. Mean error rates in dyslexic and non-dyslexic children for different types of
material.
Regarding accuracy, we find significant deficits for dyslexic children during the
processing of letters and patterns of dots, but no significant differences in reaction
time. The dyslexic children even react slightly faster with all three types of material
and therefore show partial performance advantages which can be interpreted as
higher impulsiveness and/or effects of compensation or training respectively. This
gives a hint to differences in strategies for using the lexical code and the combined
advantages in name-ability for normal reading children. These findings confirm
results of Ellis (1981) and Remschmidt and Warnke (1995).
6. DISCUSSION
In the discussed experiments, working memory performance was differentiated

based on the Baddeley-Hitch model (1974), with their specific of modalities for
incoming and received information, and further based on Cowan's model (1995),
with regard to the degree of the automaticity versus control of the executive
functions during generation of response and action. We are aiming to find out to
which extent dyslexic children differ from normal reading children regarding
various functions of working memory, and furthermore, whether deficits of dyslexic
children are general or specific in type of modality, material and domain.
The results of the three experiments do not confirm the assumption about a
general deficit of working memory, but do point towards dependencies on types of
modality and material, influences of language systems together associated with
culture differences, and the importance of the degree of automaticity versus control
of the executive function requirements for working memory performance in
dyslexic children.
Hence, we can confirm the yet consistent findings for deficits of phonological
loop for dyslexics which seem to affect most of the dyslexic children (71.4% of the
dyslexic sample according to sub-type analysis in experiment 2) and appear to be

more strongly developed compared to deficits of visual-spatial sketchpad. Thus, we
can explain, with the results of experiment 1, why German dyslexic children were
highly discriminable by phonological working memory working tasks with a high
degree of controlled executive functions (sequential reproduction) from normal
reading children with same age. During experiment 2, significant deficits in the
performance of dyslexics were only provable for auditory presentation of words and
pseudo-words regarding the error parameter.
According to the results of experiment 3, deficits of the visual-spatial sketchpad
seem to depend strongly on the type of material. The significantly higher error rate
found for dyslexic children for letters and patterns of dots resulted on one hand, due
to more complex visual patterns of stimuli, confirming the results by Casco and
Prunetti (1996). On the other hand, it was due to the impaired lexical access while
matching of letters, which is well explainable with Ellis' results (1981) and his
name-ability-hypothesis.
For the first time, an effect of superiority of Chinese during childhood could be
confirmed by a proven dependency of working memory processes on the language
system associated with cultural differences (experiment 1). For Chinese adults,
higher working memory performance of the phonological loop was proven by Lau
and Hoosain (1999) and Chin cotta and Underwood (1997), as well as of the visual-
spatial sketchpad by Lass (1997). In the first experiment, a general superiority of
Chinese children could be shown for the time parameter regarding visual and
auditory tasks. In addition, a superiority during auditory tasks in accuracy
parameters of Chinese children could be confirmed. The results suggest a high
superiority of Chinese children regarding performance of the phonological loop, and
at the same time, a lower superiority regarding the visual-spatial sketchpad. The
results can be interpreted as effects of a "natural" training of working memory,
initiated by the Chinese language system in its requirements for the oral and written
communication associated with the specificity of the education practice in Hong
Kong. This superiority effect could be associated with the required cognitive
processes of language comprehension and production. The different intonations and
maintenance of single syllables, during simultaneous recognition of word
boundaries, places great challenges on the working memory of Chinese individuals
during oral communication. Furthermore, the differentiation of characters and word
recognition without graphical markers of the word boundaries, demands a special
performance of the working memory of the Chinese reader. The confirmed strong
language- and culture-related dependency of working memory performance on
training is well comparable with the age-related compensation effects for
phonological loop confirmed by Gathercole and Baddeley (1993), as well as with
findings by Witruk and Rosendahl (1999) concerning the compensation effects of
the visual-spatial sketchpad.
Therefore, the question posed in the title about the generality of deficits can be
differentiated in three ways. With the obtained results, we can notice a specific in
the type of modality and material, and a dependency of working memory
performance on training and language system in dyslexic individuals.
Further research should clarify the age-compensation-hypothesis of the working

memory functions for several modalities with the help of longitudinal designs, and
the problem of relating early, elementary visual and auditory perception of dyslexics
to later, complex visual and auditory information processing, like working memory
functions. The interdependence between early elementary auditory and visual
perception performances (e.g. contrast sensitivity and persistence) and later highly
complex processes of working memory is still unsettled for experimental dyslexia
research. Therefore, the question is if and how strongly deficits of working memory
are affected by deficits of early perception processes.
7. AFFILIATIONS
Dr. Evelin Witruk
Universitiit Leipzig,
Fakultiit fUr Biowissenschaften, Pharmazie und Psychologie,
Institut fUr Angewandte Psychologie
Seeburgstrasse 14/20
04103 Leipzig
e-mail: witruk@rz.uni-leipzig.de
8. REFERENCES
Baddeley, AD. (1995). Working memory. Oxford: Clarendon Press.
Baddeley, A D., & Hitch, G. (1974). Working memory. In G. Bower (Ed.), Recent advances in learning
and motivation (pp. 8). New York: Academic Press.
Barnea, A, Lamm, 0., Epstein, R., & Pratt, H. (1994). Brain potentials from dyslexic children recorded
during short-term memory tasks. International Journal of Neuroscience, 74, 227-237.
Betz, D., & Breuninger, H. (1993). Teufelskreis Lernstorungen. Weinheim: Beltz, Psychologie
Verlagsunion.
Borsting, E., Ridder III, W. H., Dudeck, K., Kelley, C., Matsui, L., & Motoyama, J. (1996). The Presence
of a Magnocellular Defect Depends on the Type of Dyslexia. Vision Research, 36, 7, 1047-1053.
Casco, C., & Prunetti, E. (1996). Visual search of good and poor readers: Effects with targets having
single and combined features. Perceptual and Motor Skills, 82, 1155-1167.
Chincotta, D., & Underwood, G. (1997). Digit span and articulatory suppression: A cross-linguistic
comparison. European Journal of Cognitive Psychology, 9, 1, 89-96.
Cowan, N. (1995). Attention and memory: An integrated framework. Oxford: Oxford University Press.
Daneman, M., & Carpenter, P. A (1980). Individual Difference in Working Memory and Reading.
Ellis, N. (1981). Visual and name coding in dyslexic children. Psychological Research, 43, 201-218.
Everatt, J. Groeger, 1., Smythe, I., Baalam, S., Richardson, 1., & McNamara, S. (2001). Dyslexia and
deficits in short-term memory: phonological versus sequential explanations. Abstract on 5 th BDA
International Conference, York.
Fawcett, A 1., & Nicolson, R. I. (2001). Dyslexia: the role of the cerebellum. In A J. Fawcett (Ed.),
Dyslexia- Theory and Good Practice (pp. 89-105). London: Whurr Publishers.
Fischer, B., & Hartnegg, K. (2000). Effects of visual training on saccade control in dyslexia. Perception,
29,531-542.
Galaburda, A, & Livingstone, M. (1993). Evidence for a magnocellulary defect in developmental
dyslexia. In P. Tallal, A M. Galaburda, R. R. Llindas, & c. von Euler (Eds.), Temporal information
processing in the nervous system. Special reference to dyslexia and dysphasia. Annals of the New
York Academy of Sciences (pp. 70-82). New York: The New York Academy of Sciences.
Garner, W. R., & Clement, D. E. (1963). Goodness of pattern and pattern uncertainty. Journal of Verbal
Learning and Verbal Behavior, 2, 446-452.
Gathercole, S. E., & Baddeley, A. D. (1993). Working memory and language. East Sussex: Hove,
Lawrence Erlbaum Associates Ltd.
Goldman-Rakic, P. S. (1998). The prefrontal landscape: Implications of functional architecture for
understanding human mentation and the central executive. In A. C. Roberts, W. T. Robbins et al.
(Eds.), The prefrontal cortex: Executive and cognitive functions (pp. 87-102). New York: Oxford
University Press.
Graentzel, B., & Jaehn, A. (1997). Untersuchung zur Arbeitsgediichtnisleistung von legasthenischen
Kindern bei akustischer und visueller Darbietung von Worten und Nichtworten. Unpublished thesis,
Universitiit Leipzig, Institut fiir Angewandte Psychologie, Germany.
Grigorenko, E. (2001). Genetics: chromosome 6 and 15 linkage findings in eight large extended families.
Paper presented at the 5th BDA International conference in York, UK.
Ho, C. S.-H., & Bryant, P. (1997). Phonological skills are important in learning to red Chinese.
Developmental Psychology, 33, 6, 946-951.
Ho, C. S.-H, & Lai, D. N.-C. (2000). Naming-speed deficits and phonological memory deficits in
Chinese developmental dyslexia. Learning and Individual Differences, 11, 173-186.
Ho, C. S.-H., Law, T. P.-S., & Ng, P. M. (2000). The phonological deficit hypothesis in Chinese
developmental dyslexia. Reading and Writing, 13, 57-79.
Hoosain, R. (1991). Psycholinguistic implications for linguistic relativity: A case study of Chinese.
Hillsdale: Lawrence Erlbaum Associates.
K1ix, F. (1984). Uber Wissensrepriisentation im menschlichen Gediichtnis. In F. K1ix (Ed.), Gediichtnis,
Wissen, Wissensnutzung (pp. 9-73). Berlin: VEB Deutscher Verlag der Wissenschaften.
Lass, U. (1997). EinfluB von Sprache auf Verarbeitungsprozesse im Kurzzeitgediichtnis - eine
experimentelle Untersuchung mit deutschen und chinesischen Probanden. In G. Liier, & U. Lass
(Eds.), Erinnern und Behalten (pp. 244-268). Gottingen: Vandenhoeck & Ruprecht.
Lau, C. W., & Hoosain, R. (1999). Working memory and language difference in sound duration: A
comparison of mental arithmetic in Chinese, Japanese, and English. Psychologia: An International
Journal of Psychology in the Orient, 42, 3, 139-144.
Liu, I., Zhu, Y., & Wu, J. (1992). The long-term modality effect: In search of differences in processing
logographs and alphabetic words. Cognition, 43, 31-66.
Mecklinger, A., & Mueller, N. (1996). Dissociations in the processing of "what" and "where"
information in working memory: An event-related potential analysis. Journal-of-Cognitive-
Neuroscience, 8 (5), 453-473.
Pedrides, M. (1994). Functional specialization within the dorsolateral frontal cortex. Revue de
Neuropsychologie,4,305-325.
Remschmidt, H., & Warnke, A. (1995). Zur visuellen und sprachlichen Informationsverarbeitung bei
lese-rechtschreibschwachen Kindem im Liingsschnitt. Kinder- undJugendpsychiatrie, 23, 1, 27.
Robinson,G. (2001). Investigation of biochemical anomalies in people with dyslexia: Implications for
immune system dysfunction and dietary intervention. Paper presented at the 5th BDA International
conference in York, UK.
Roesler, F., Bajric, J., & Heil, M. (1997). Gediichtnisspuren im EEG (Memory traces in the EEG).
Zeitschrift fUr Experimentelle Psychologie, 44, 4-37.
Schneider, A. (2001). EEG-coherence-analysis for examining a general automatizational deficit in
dyslexic children. Paper presented at the 5th BDA International Conference, York.
Schroeger, E. (1995). Processing of auditory deviants with changes in one vs. two stimulus dimensions.
Psychophysiology, 32, 55-65.
Share, D. L. (1994). Deficient phonological processing in disabled readers implicates processing deficits
beyond the phonological module. In K. P. van den Bos, L. S. Siegel, D. J. Bakker, & D. L. Share
(Eds.), Current directions in dyslexia research (pp. 149-167). Lisse: Swets and Zeitlinger.
Siegel, L. S., & Linder, B. A. (1984). Short-term memory processes in children with reading and
arithmetic learning disabilities. Developmental Psychology, 20, 2, 200-207.
Siegel, L. S., & Ryan, E. B. (1989a). The development of working memory in normally achieving and
SUbtypes of learning disabled children. Child Development, 60, 973-980.
Siegel, L. S., & Ryan, E. B. (1989b). Subtypes of developmental dyslexia: The influence of definition
variables. Reading and Writing: An Interdisciplinary Journal, 2, 257-287.
So, D., & Siegel, L. S. (1997). Learning to read Chinese: Semantic, syntactic, phonological and working
memory skills in normally achieving and poor Chinese readers. Reading and Writing: An
InterdisciplinaryJournal,9, 1-21.
Spinelli, D., Angelelli, P., De Luca, M., Di Pace, E., Judica, A., & Zoccolotti, P. (1997). Developmental
surface dyslexia is not associated with deficits in the transient visual system. NeuroReport, 8, 1807-
1812.
Stein, 1., & Walsh, V. (1997). To see but not to read; the magnocellular theory of dyslexia. TINS, 20,
147-152.
Sternberg, S. (1975). Memory scanning. New findings and current controversies. Quarterly Journal of
Experimental Psychology, 27, 1-32.
Tallal, P., Galaburda, A. M., Llinas, R. R., & von Euler, C. (Eds.). (1993). Temporal information
processing in the nervous system. Special reference to dyslexia and dysphasia. Annals of the New
York Academy of Sciences. New York: The New York Academy of Sciences.
Vallin, R. (1984). Awareness of features and functions in language. In: J. Downing, & R. Valtin (Eds.),
Language awarness and learning to read (pp. 227-260). Springer: Berlin.
Valtin, R. (1989). Dyslexia in the German language. In: P. G. Aaron, & R. M. Joshi (Eds.), Reading and
writing disorders in different orthographic systems (pp.119-135). The Netherlands: K1uwer
Academic Publishers.
Vellutino, F. R. (1987). Dyslexia. Scientific American, 256,3,1-42.
Walther-Milller, P. U. (1995). Is there a deficit of early vision in dyslexia? Perception, 24, 919-936.
Weigt, R. (1980). Zur Auswahl von SchUlern fUr die LRS-Klassen. Berlin: Verlag Volk und Wissen.
Weigt, R. (1994). Lesen- und Schreibenlernen kann jeder!? : Methodische Hilfen bei Lese-Rechtschreib-
Schwiiche. Neuwied, Kriftel, Berlin: Luchterhand.
WiIlows, D. M., Kruk, R. S., & Corcos, E. (Eds.). (1993). Visual processes in reading and reading
disabilities. Hillsdale: Lawrence Erlbaum Associates.
Witruk, E. (1993). Memory deficits of dyslexic children. In P. Tallal, A. M. Galaburda, R. R. Uindas, &
C. von Euler (Eds.), Temporal information processing in the nervous system. Special reference to
dyslexia and dysphasia. Annals of the New York Academy of Sciences (pp. 430-435). New York: The
New York Academy of Sciences.
Witruk, E. (1999). Working memory performance in dyslexic children. In E. Witruk, & T. Lachmann
(Eds.), Basic Mechanisms of Language and Language Disorders (pp. 20-21). Leipzig: Leipziger
Universitiitsverlag.
Witruk, E., & Ho, C. S.-H. (1999). Dyslexia and working memory in different language systems. Paper
presented at the 5'h Annual Meeting of the Society for Scientific Study of Reading, Montreal,
Canada.
Witruk, E., Lachmann, T., & Graeve, 1. (in preparation). Visual matching with different types of material
in dyslexic and non-dyslexic children. The Official Journal of the Society for Scientific Study of
Reading.
Witruk, E., & Rosendahl, W. (1999). Modalitiits-und Anforderungsspezifik von
Arbeitsgediichtnisleistungen bei Legasthenikern. In Kongrepbericht der XXIII. Arbeits- und
Fortbildungstagung der Deutschen Gesellschaft fUr Sprachhei/piidagogik e. v., Wilrzburg.
Witruk, E., & Schuster, U. (1997). Working memory in dyslexic children. Paper presented at the 4th
World Congress on Dyslexia, Halkidiki, Greece.
Witruk, E., & Staats, M. (1983). Differentielle Leselernanalyse. Zeitschrift fUr Psychologie, 3, 261-278
J. K. UHRY
KINDERGARTEN PHONOLOGICAL AWARENESS

AND RAPID SERIAL NAMING AS PREDICTORS OF
GRADE 2 READING AND SPELLING
Abstract. While phonemic awareness and rapid serial naming are both considered forms of phonemic
processing, each has been found to account for independent variance in reading, but the underlying
processes utilized by rapid serial naming are not clear. These processes have been conceptualized as a
form of phonological processing, as speed of processing orthographic units, and as sensitivity to arbitrary
associations between phonology and orthography. This study looks at these models longitudinally with 86
children across grades K-2 using spellings of the oed suffix in past tense verbs as a measure of
orthographic conventions learned while scanning text during reading. Regression analysis results
indicated that both phonological awareness and rapid serial in kindergarten made unique contributions to
Grade 2 reading and spelling measures. Controlling for early word reading and vocabulary, these
contributions were strongest for pseudo word accuracy in the case of phonological awareness, and to
spelling of arbitrary orthographic conventions in the case of rapid serial naming, but both kindergarten
measures made unique contributions to late word-reading rate.
1. INTRODUCTION
It is the purpose of this study to build on a model of the transition from early
phonological recoding to the more efficient recognition of sight words. Ehri has
provided a model for the early development of word reading in which children use
only partial letter-sound cues. Phonemic awareness facilitates informed use of more
complete representations of phoneme-grapheme relationships as children learn to
use the full alphabetic strategy to identify unfamiliar printed words (Ehri &
McCormick, 1998). With practice, these small phoneme-grapheme units (e.g., /t/ = t,
Ib/ = b) become consolidated, and larger units (-at, -igh) are also recognized
automatically and rapidly. Ehri has described the emergence of this consolidated-
alphabetic phase as typical of children in Grade 2; This phase is a precursor to
mature reading, or the automatic-alphabetic phase in which words are recognized
automatically by sight. Frith (1985) described these final phases of word reading
acquisition as the orthographic phase. In her view the orthographic strategy rests on
the rapid recognition of letter units, a skill made possible by earlier phonological
development. This is consistent with Ehri's (1992; Ehri & McCormick, 1998; Uhry
& Ehri, 1999) consolidation theory of sight-word learning in which larger and larger
units are recognized, facilitated by early letter-sound learning. In Frith's view,
spelling and reading are mutually facilitative. Spelling facilitates word reading at the
earlier alphabetic stage, because spelling builds ability in phoneme segmentation,
but at the orthographic stage, exposure to print during reading facilitates spelling.
Understanding the processes underlying both alphabetic and orthographic strategies
is critical to understanding the transition to mature word reading.
E. Witruk, A.D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading,
300 J. K. UHRY
Both phonological awareness and rapid serial naming are associated with success
in learning to read words. Phonological awareness has been well documented in the
research literature (Adams, 1990; Bradley & Bryant, 1983; Uhry, 1993, 1999) as an
associate of beginning word reading in young children. Training studies (e.g., Ball &
Blachman, 1991; Bradley & Bryant, 1983; Foorman, Francis, Fletcher,
Schatschneider, & Mehta, 1998; Lundberg, Frost, & Petersen, 1988; Torgesen,
Wagner, Rashotte, Alexander, & Conway, 1997; Uhry & Shepherd, 1993, 1997)
have reported a significant advantage in learning to read for children trained in
phonological awareness in comparison with children without this training. Clear
models exist for the ways in which early reading is supported by this skill.
Phonological awareness tasks such as segmenting and blending phonemes utilize
ability to understand the sound structure of spoken language. Together with phonics
knowledge (i.e., the association between letters and the sounds of phonemes)
phonological awareness facilitates the alphabetic principle, or the way in which
spoken language maps onto written language (Adams, 1990). This is important in
the early phase of word reading.
While Denckla's Rapid Automatized Naming Test (RAN; Denckla & Rudel,
1974, 1976) is more than 25 years old, less is known about rapid serial naming in
comparison with phonological awareness in regard to how it facilitates reading.
There is no clear model here. What underlying abilities does the RAN measure?
How do these abilities facilitate word reading and spelling? And what is the
relationship between rapid serial naming and phonological awareness? It is the goal
of this study to explore these questions. Research initiatives have provided several
competing theories of the relationship between rapid serial naming and reading.
The RAN's color-naming task was taken originally from protocols for
neurological evaluations of head-injured patients. For their work with children with
reading disorder, Denckla and Rudel added picture naming as a downward extension
for pre-readers, and number and letter naming as upward extensions (Denckla &
Cutting, 1999). Over the years, the two symbol-naming sub tests, numbers and
letters, have proved to be more strongly correlated with reading than the other
subtests, and serial naming has proved to be a better predictor of word reading than
discrete trial naming (see Wolf, 1991, 1999 for reviews). Denckla attributes the
success of the RAN to two similarities between this task and reading words in text,
one in the language domain, and the other in the executive domain.
2. RAPID SERIAL NAMING AS A FORM OF PHONOLOGICAL PROCESSING
Wagner and Torgesen (1987) have focussed on the language domain and have
referred to rapid serial naming as phonological recoding in lexical access. They have
described it as one of three phonological processes associated with skilled reading,
the other two being phonological awareness and phonetic recoding to maintain
information in working memory (i.e., verbal short-term memory). They emphasized
the code-switching aspect of rapid naming and viewed difficulty with rapid naming
as difficulty accessing letter (or color or digit or object) names. In their view,
retrieving words during the recoding process is facilitated by phonology, a view
shared by Bradley and Bryant (1983, 1985) who conceptualized the sorting tasks in
their phonological awareness training as a method of building up stores of
PHONOLOGICAL AWARENESS AND RAPID NAMING 301
phonologically categorized items for retrieval during reading. This view is not
inconsistent with research reports in which phonological awareness and rapid
naming have each made unique contributions to variance in reading (e.g., Felton &
Brown, 1990). That is, isolating and storing phonemes, and rapid retrieval from
phonologically stored material, are related operations that are facilitated by distinct
underlying processes.
In a large-scale, longitudinal study, Torgesen, Wagner, Rashotte, Burgess, and
Hecht (1997) carried out regression analyses on over 200 children to examine the
predictive contributions of phonological awareness and rapid serial naming from
Grades 2 to 4, and from Grades 3 to 5. In both analyses, each of these variables
contributed unique variance to later reading. However, with word identification
entered at the first step, only phonological awareness continued to contribute to later
reading. Torgesen and his colleagues have interpreted this to indicate that the
influence of rapid serial naming is subsumed by the effect of the original reading
measure on later reading.
3. RAPID SERIAL NAMING AS RAPID PROCESSING OF ORTHOGRAPHY
Slow naming has been associated historically with reading disorder (see Denckla &
Cutting, 1999; Wolf, 1991). Bowers and Wolf (1993; Bowers, 1995; Bowers &
Swanson, 1991; Wolf, 1991, 1999) have argued against Wagner and Torgesen's
(1987) model of rapid serial naming as a form of phonological processing. By
contrast, they have presented a model in which the rapid naming task taps a separate,
non-phonological group of functions that are related to Denckla's second domain,
speed of processing. Bowers and Wolf (1993) argued that slow naming rate was
distinct from phonological disorders and hypothesized a "double deficit" whereby
the weakest readers suffered from disorders in phonological awareness as well as
slow naming. This argument has been supported by regression-based studies in
which both phonological awareness and rapid naming made unique contributions to
reading (see Wolf, 1999).
Bowers (1995; Bowers, Sunseth, & Golden, 1999) has argued, moreover, that the
RAN's relationship with reading taps functions that are more specific than general
processing speed. She has developed a model by which the transition from letter-by-
letter reading is facilitated by ability to process units of letters or complete words
both accurately and rapidly, and hypothesized that naming speed is dependent on
"the extent to which sequences of letters are fully processed" (1999, p. 34). She
cited as evidence Ehri and Saltmarsh's (1995) finding that poor readers took longer
to learn new sight words than younger children matched by reading level. In her
view the RAN taps the complex processes whereby units of letters are rapidly
recognized and utilized in reading. It is speed-related but specific to complex
orthographic tasks.
4. RAPID SERIAL NAMING AS SENSITIVITY TO ARBITRARY

ORTHOGRAPHY
Mannis, Seidenberg, and Doi (1999) have provided a third model of the way in
which processes underlying the RAN task impact on skilled reading. Using
302 J. K. UHRY
Seidenberg and McClelland's (1989) connectionist model, they hypothesized that

difficulty with phonological representations affects the quality of conversions from
orthography to phonology during reading. and that, while not tested using the
original Seidenberg and McClelland computational model, these conversions are
also affected by a specific form of orthographic skill. They measured 67 children on
reading-related skills in Grade 1 and again in Grade 2. Orthographic measures were
based on Olson, Wise, Conners, Rack, and Fulker's (1989) Orthographic Choice
task, Stanovich and Siegel's (1994) Word Likeness task, and an exception word
task. They found that Grade 1 phonemic deletion (Rosner & Simon, 1971) was a
strong predictor of Grade 2 reading measures utilizing predictable phonology-to-
orthography conversion, such as the Word Attack task from the Woodcock Reading
Mastery Test (1987), while the RAN was a strong predictor of Grade 2 orthographic
skill. In their view, the RAN tapped arbitrary associations between verbal names and
orthography rather than speed of orthographic processing.
5. THE PRESENT STUDY
The goal of this study is to examine the relationship between rapid orthographic
processing of print and the development of young children's reading and spelling.
The studies described above have involved children after the onset of reading
instruction. The present study will look at kindergarten children and at the
relationship between phonological awareness and rapid serial as predictors of
reading in Grade 2. In doing this, these processes can be examined prior to the
influence of formal reading instruction. That is, rapid serial naming could be
influenced by practice in reading text, and phonological awareness is facilitated by
learning how speech is mapped onto written language during the acquisition of a
reading vocabulary (see Ehri, 1992; Ehri & McCormick, 1998; Uhry & Ehri, 1999).
Another initiative is the use of spellings as a way of measuring orthographic
sensitivity. Bryant, Nunes, and Bindman (1997) have documented the transition
from spelling phonetically in younger children to spelling using orthographic
conventions. Past tense English verbs ending in -ed, but sounding like /d/ or /t/,
provide a way of looking at the choice young children make between phonics
regularities and learned orthography involving arbitrary units of letters. Uhry (1999)
found that children in Grade 2 are making the transition from phonetically-based
invented spellings to the more sophisticated -ed endings. In Frith's (1985) model,
this transition is facilitated by exposure to print during reading. If the RAN taps
processes used during rapid scanning of text in reading, then the transition to
orthographic conventions should be facilitated by processes measured by the RAN.
This study's design is modeled on studies by Torgesen et al. (1997) and Mannis
et al. (1999) in which early rapid serial naming is expected to predict later measures
of rate-related orthographic processing, and early phonological awareness is
expected to predict later measures of word-reading accuracy. The finding that RAN
predicts orthographically conventional spellings of arbitrary associations (Le., -ed
endings) would support Frith's model in which reading leads to spelling, and would
have implications for instruction.
5.1 METHOD
5.1.1 Participants
One hundred and nine children in five kindergarten classes in an urban school were
assessed at mid kindergarten for an earlier study of the relationship between
invented spelling and early reading skills (Uhry, 1999). Eighty-six of these children
were available for additional testing at the end of kindergarten and again at the end
of their second-grade year. At that point they were distributed across six Grade 2
classrooms. The mean age of these 86 children when Grade 2 testing began in
March was 92.36 months with a standard deviation of 3.45. In this school district,
children tum 6 before January 1 of the Grade 1 school year. The ethnic make-up of
the group was roughly 60 percent Caucasian, 15 percent Asian, 14 percent Mrican
American, and 11 percent Latino/a. About 25 percent of the children were eligible
for the United States government's Title I free lunch program, used here as an
indicator of low socio-economic status.
5.1.2 Kindergarten Measures
Receptive Vocabulary. The Peabody Picture Vocabulary Test (Dunn & Dunn, 1981)
was administered as a measure of ability to understand spoken words. The examiner
says a word aloud and the child points to its best meaning-based match with one of a
choice of four pictures. This test uses a basal-ceiling format and provides standard
scores as well as the raw scores that were used here for statistical analysis.
Phonological Awareness. The Test of Auditory Analysis Skills (TAAS; Rosner,

1975) was administered as a measure of phonological awareness. The format is a
deletion task that progresses from syllable deletion to initial phoneme deletion to
final phoneme deletion to deletion of a consonant within a consonant cluster, with a
range of 0-13.
Rapid Serial Naming. The Rapid Automatized Naming Test (RAN; Denckla &
Rudel, 1974) consists of four subtests. Children were asked to name 5 colors (and
then numbers, pictured objects, and letters) repeated 10 times each and randomly
arranged in a 10 column x 5 row array. Children were provided with initial
instruction in order to be certain they could name each stimulus item. Then they
were told to say the names of the colors in order as quickly as they could. As a child
read the stimulus items, the examiner made certain that the child did not skip a line.
In order to avoid skewed data, the total number of seconds for each subtest was
converted to a score of items-per-second.
Word Reading. The Word Identification (WID) subtest of the Woodcock Reading
Mastery Test (WRMT; Woodcock, 1987) uses a basal-ceiling format and consists of
progressively more difficult words. Children were given five seconds in which to
read each word aloud. The examiner read the word aloud after five seconds if the
child had not done so, and encouraged the child to move on to the next word on the
304 J. K. UHRY
list. Standard scores are available for the WID but raw scores were used for
statistical analysis here.
5.1.3 Grade 2 Measures
Comprehension. Reading comprehension was measured in two ways. First, a

standardized measure of silent reading comprehension (SilComp) was administered
using Form L of the Grade 2 level of the Gates-MacGinitie Reading Tests. On the
Vocabulary subtest of this group-administered test, children are asked to look at a
picture and match it with one of four words. On the Comprehension sub test, they are
asked to match a sentence or paragraph with one of three pictures. The two sub tests
are totaled for a raw score. Standardized scores are provided but raw scores were
used here for statistical analysis.
The second comprehension measure (OraIComp) was individually administered,
and involved oral questioning by the examiner following the oral reading of
progressively more difficult passages from trade books. In a procedure similar to
that used in informal reading inventories, four passages were provided, at the
primer, Grade 1, Grade 2, and Grade 3 levels. As with the WID, above, a word was
provided by the examiner after any 5-second delay. All children began with the
primer-level text and read through the passages until they missed 50% of the words
in the first paragraph of a passage. Eight questions were provided after each of the
four reading passages with possible scores ranging from 0-32. Validity data for this
measure are provided in the results section below.
Word Reading Accuracy. Three word-reading accuracy measures were collected.

The Word Identification (WID) subtest of the Woodcock Reading Mastery Test
(WRMT; Woodcock, 1987) was administered again in Grade 2 as a standardized
measure of decontextualized word-list reading.
Word reading accuracy was also measured using lists of phonetically regular
pseudowords from the Word Attack (WA) subtest of the Woodcock Reading
Mastery Test. Again, a 5-second limit was placed on response. As with the WID,
standard scores are available but raw scores were used here for statistical analysis.
On a third measure, children were asked to read aloud from a series of
progressively more difficult passages taken from children's trade books described
above in the section on comprehension. A count of correctly-read words was used as
a measure of word-reading accuracy in text (WordsTxt). Again, the examiner
provided a word after any 5-second delay. All children began with the primer-level
text and read through the passages until they missed 50% of the words in the first
paragraph of a passage.
Word Reading Rate. The primer and Grade 1 trade book passages were timed with a
stop watch as the children read aloud and a words-per-second score was created for
each level and then averaged to use as a measure of words-per-second (WPS)
reading rate. These two passages were chosen because every child was able to read
to the end of both passages without missing the requisite number of words for
discontinuing, whereas not every child was able to read to the end of the Grade 2
passage.
Spelling. Spelling was used as a measure of orthography to test the hypothesis that
RAN utilizes the ability to process and learn arbitrary patterns of letters. Two lists of
spelling words were dictated, a standardized spelling test with a range of word types,
and the second limited to words with arbitrary associations. These lists were meant
to differentiate between general orthographic skill and skill with arbitrary
associations.
The first list was the Spelling subtest (WIATSp) from the Wechsler Individual
Achievement Test (Wechsler, 1992). This test involves progressively more difficult
letter names and words that are dictated to children to spell. Standard scores are
available and were used here descriptively, but raw scores were used for statistical
analysis.
The second dictated spelling list (SPELL-ed) involved 12 past-tense verbs
designed for a study (Uhry, 1999) of children's ability to spell using orthographic
conventions rather than letters that sound like the phonemes they represent. This
measure was based on one designed by Bryant, Nunes, and Bindman (1997). On the
list of 12 words used here, 6 of the -ed endings sounded like Idl (i.e., sailed,
grabbed, played, burned, killed, phoned), and 6 like It! (i.e., kissed, tricked, bumped,
parked, asked, hoped). The 12 words were used as a measure of arbitrary
associations between letter units and sounds. Again, context sentences were
provided as part of the dictation. All 12 words were dictated to all children. The
score for this measure was the correct spelling of -ed suffixes with a range of 0-12.
5.2 Procedures
Testing was carried out between March and May by graduate students trained in
assessment. Data collection represented part of a larger battery designed to follow
these children from kindergarten (Uhry, 1999) through grade 4 in regard to literacy
skills. The children were tested individually in a quiet space outside of their
classrooms over two or three sessions. Sessions lasted 25-45 minutes.
5.3 Data Analysis

The relationships among kindergarten and grade 2 literacy measures was analyzed
through a correlation matrix and then through a series of hierarchical multiple
regressions.
6. RESULTS
6.1 Descriptive Data
Table 1 provides mean raw scores and standard deviations for kindergarten and
Grade 2 measures. Standard scores derived from the raw scores for word
identification (WID) and oral receptive vocabulary (PPVT) in kindergarten indicate
performance in the average range; the mean standard score for word identification
306 J. K. UHRY
was 113, and for oral vocabulary was 101. The mean raw score of 6.9 on the TAAS
syllable and phoneme deletion task indicates ability to delete syllables and initial
phonemes but not final phonemes or phonemes within consonant clusters. Test
norms indicate early Grade 1 performance for this end-of-kindergarten group.
Grade 2 standard scores were also in the average range. At the end of Grade 2
the mean standard score for word identification was 109, for word attack was 100,
for spelling was 99, and for silent reading comprehension was 98.
Table 2 presents a matrix of correlations among kindergarten and Grade 2
literacy measures. Validity for non-standardized measures was established through
correlations with standardized measures. The correlation between the standardized
WID word-reading lists and the measure of words read in text was .86. The
correlation between the standardized Gates silent reading test and the oral questions
following trade book passages was .80.
Table 1: Means and Standard Deviations for Kindergarten and Grade 2 Measures
Kindergarten Grade 2
Possible M (SD) M (SD)
Range
Age in Months on May 1 70.4 (3.45)
RosnerTAAS (0-13) 6.9 (2.98)
RAN Colors per second (CPS) .9 (0.23)
RAN Numbers per second (NPS) 1.0 (0.30)
RAN Objects per second (OPS) 0.6 (0.16)
RAN Letters per second (LPS) 0.9 (0.31)
Word Identification (WID) 13.9 (16.67) 56.3 (13.96)
Peabody Picture Vocabulary Test 68.9 (16.74)
Comprehension in Silent Passages (0-90) 72.0 (15.73)
Comprehension in Oral Passages (0-32) 21.3 (6.11)
Word Attack (WRMT nonword list) 22.9 (9.85)
Words in Text Passages (WdsText) (0-492) 402.3 (117.19)
Passage Reading Rate (WPS) 1.5 (0.56)
Spelling (WIATSp - dictated words) 22.0 (5.68)
Spelling (Spell-ed: Idl & It! as -ed) (0-12) 8.7 (4.09)
Concurrent and predictive correlations among other variables were significant

with several exceptions: the relationship between oral vocabulary and phoneme
deletion in kindergarten, the relationship between kindergarten oral vocabulary and
Grade 2 spelling measures, and the relationship between color naming rate in
kindergarten and Grade 2 measures of word attack and spelling were nonsignificant.
The strongest concurrent correlations included relationships between measures of
Grade 2 word reading and comprehension and ranged from .83 to .90.
Strong predictive relationships included the following: kindergarten TAAS as a
predictor of Grade 2 word identification (r= .63) and Grade 2 word attack (r = .62);
letter-naming rate as a predictor of word-reading rate (r = .60), word identification
as a predictor of spelling (r = .71); and reading words learned in
Shared Reading as a predictor of silent comprehension (r = .74), word

identification (r = .79), word attack (r = .74), and spelling (r = .72).
6.2 RAN as a Measure of Rate
One interpretation of the relationship between rapid serial naming and reading is that
speed of retrieval is the critical factor. If speed of processing were the only factor
relating the RAN to reading, then kindergarten color naming and letter naming each
would be correlated with each other, and with Grade 2 measures in which rate is a
factor, and each could be expected to contribute unique variance to rate-related
reading tasks, but neither could be expected to contribute unique variance beyond
the contribution of the other.
However, this was not the case. See Table 2 for correlations between
Kindergarten and Grade 2 measures. Considering, first, the correlations among
kindergarten variables, color and letter naming each were significantly correlated
with the other three RAN sub tests. The strongest correlation for color naming was
with object naming, and the strongest correlation for letter naming was with number
naming. Next consider correlations between kindergarten naming and Grade 2
measures. Kindergarten color-naming rate was significantly correlated with Grade 2
measures for which rate could be considered an important factor (accuracy of word-
list reading in a timed format, rate of reading words in text, both oral and silent
comprehension). However, in each of a series of hierarchical multiple regression
analyses (see Table 3), with color-naming rate (RAN:CPS) entered at Step 1, letter-
naming rate (RAN:LPS) made substantial unique contributions to Grade 2 measures
at Step 2, with contributions ranging from 12% to 36%. When these RAN subtests
were entered in the opposite order, in no case was there significant unique variance
for color naming at Step 2 after letter naming was entered at Step 1.
6.3 RAN as a Measure of Phonological Processing Distinct from Phonological

Awareness
Another interpretation of the relationship between rapid serial naming and reading is
that rapid naming is a form of phonological processing that contributes to reading
beyond the contribution of phonological awareness. The strong correlations between
Kindergarten phonological awareness and letter-naming rate, and the strong
correlations between Kindergarten letter-naming rate and Grade 2 measures believed
to depend on phonological awareness (i.e., word reading accuracy on lists and in
text, word attack skills) support this interpretation.
A series of hierarchical multiple regression analyses (see Table 3) was used to
determine if rate of letter naming (LPS) in Kindergarten accounted for unique
variance in Grade 2 measures beyond the variance attributable to phonological
awareness. With the TAAS measure of syllable and phoneme deletion entered at
Step 1, the measure of serial letter-reading rate (RAN:LPS), entered at Step 2,
accounted for additional significant unique variance in Grade 2 reading-related
measures ranging from 3% to 15%. The highest levels of unique variance were
present in analyses of word-list reading (WID; 10%), spelling -ed endings in past
tense words (Spell-ed; 10%), and word-reading rate in text (WPS; 15%).
w
0
00
Table 2. Correlations among Kindergarten and Grade 2 Measures
Note: All correlations are significant except those in which r < .IB4
Kindergarten Grade 2
2. 3. 4. 5. 6. 7. B. 9. 10. 11. 12. 13. 14. 15. 17.
I. TAAS-K .312 .525 .320 .524 .578 .679 .175 .552 .449 .514 .629 .619 .540 .514 .334
2. RAN-CPS-K .660 .762 .529 .219 .287 .288 .282 .274 .249 .229 .183 .342 .106 .232
3. RAN-NPS-K .688 .780 .456 .553 .279 .448 .362 .380 .463 .420 .588 .329 .245
4. RAN-OPS-K .622 .357 .351 .238 .353 .367 .300 .350 .253 .447 .243 .257
5. RAN-LPS-K .670 .625 .318 .530 .505 .421 .565 .503 .602 .496 .455
6. WID-K .796 .280 .583 .545 .488 .693 .635 .551 .708 .430
7. RdWds-K .312 .736 .637 .695 .785 .741 .684 .724 .543
8. PPVT-K .427 .548 .351 .306 .319 .397 .189 .158 ~
9. SilComp-2 .795 .858 .844 .764 .810 .690 .637 ~
to. OralComp-2 .805 .769 .683 .787 .615 .519 c:::
II. WdsText-2 .858 .755 .834 .664 .581 :I:
::tl
12. WID-2 .895 .859 .843 .587 -<
13. WA-2 .767 .833 .544
14. WPS-2 .697 .503
15. WIATSp-2 .599
16. Orth-ed-2
Table 3: Hierarchical Multiple Regressions Predicting Grade 2 Reading from Kindergarten Variables Reported in Unique Contribution (If Increase) ,
Pointo/Entry with Significance Levels Reported/or Final Step. Note: *p < .05, **p < .01, ***p < .001
~
Kindergarten Grade 2 ~
Step Variable SilComp OralComp WdsTut WID WA WPS WIATSP WIATOR SPEU-ed
§
Unique Contribution of Kindergarten Rapid Serial Naming Measures to Grade 2 Reading
1. RAN:CPS .08ns .07ns .06ns .05ns .03ns .12ns .01ns .00ns .05ns ~
2. RAN:LPS .20"*** .18**** .12 ..... .28*"** .23 .... ** .24*",,* .27·**· .28 .... *· .16"*"~
1. RAN:LPS .28* .... * .25"*"* .18*** .32**"* .24**** .36**** .25·*** .24**** .21 **.~ i
2. RAN:CPS .00ns .00ns .00ns .00ns .02ns .oons .03ns .04ns .00ns ~
~
.Unique Contribution of Phonological Awareness and Rapid Letter Naming to Grade 2 Reading >
Z
1. T AAS .46* *"* .31 *"*.. .32"* ** .39* "* * .39* "* * .32"* * .32**** .28"** .14ns I:'
2. RAN LPS .06"* .08** .03* .10*** .06** .15**** .07"* .08** .10**
~
I:'
1. RANLPS .28** .25*" .18· .32*** .25** .36**** .25"* .24** .21 **
2. TAAS .24**** .14**** .17" .... " .17**** .20**** .11 "** .14**** .12*** .03ns ~
~
z
Unique Contribution otPhoDologacal Awarenessand Rapid Letter Naming Controlling for Word Reading and Vocabulary o
1. WID .34* .30· .24ns .48*** .40** .30ns .50**** .49**** .18ns
2. PPVT .08** .13**** .05* .01ns .02ns .07* .00ns .oons .01ns
3. TAAS .06** .02ns .08** .08*** .10*** .07** .02ns .02ns .01ns
4. RAN: LPS .01ns .01ns .oons .01ns .00ns .05** .00ns .00ns .04*
w
o
\0
310 J. K. UHRY
Conversely, with LPS entered at Step 1, the TAAS, entered at Step 2, accounted
for additional significant unique variance ranging from 3% to 24%. Here the highest
levels of variance were present in analyses of Grade 2 silent reading comprehension
on the Gates-MacGinitie Reading Tests (SilComp; 24%), and a measure of word
attack skills (WA; 20%).
6.4 Contributions of Phonological Awareness and Rapid Naming Beyond

Contributions of Word Reading and Vocabulary
In a third set of hierarchical regressions controlling for Kindergarten word-reading
level (WID) at Step 1 and kindergarten oral receptive vocabulary (PPVT) at Step 2,
the TAAS was entered at Step 3 and the RAN: LPS at Step 4. Little unique
significant variance was contributed beyond the contribution of Kindergarten word
reading, with Kindergarten WID contributions ranging from 30% of the variance in
Grade 2 spellings of -ed endings to 62% of the variance in Grade 2 WID scores.
Kindergarten vocabulary contributed unique significant variance, beyond the
contribution of the WID in three cases. It contributed to silent reading
comprehension on the Gates-MacGinitie (SilComp; 4%), to comprehension of oral
text reading (OraIComp; 13%), and to reading rate in oral text (WPS; 3%). Beyond
the contributions of Kindergarten word reading and oral vocabulary, the TAAS and
the RAN made unique contributions in highly specific areas. The TAAS, at Step 3,
contributed its highest level of unique significant variance to word attack (10%), but
also contributed to silent comprehension (6%), reading words accurately in text
(8%), reading words in lists (WID; 8%), and word-reading rate in text (WPS; 7%).
The RAN's contributions were specific to word-reading rate (5%) and spelling
phonetically ambiguous -ed endings with arbitrary orthography (4%).
7. DISCUSSION
The major goal of this study was to explore the relationship between rapid serial
naming in kindergarten and children's reading and spelling in Grade 2, a time when
a shift from alphabetic to orthographic strategies could be expected. The correct
spelling of -ed suffixes with the Idl and It! sounds was measured as an indicator of
orthographic knowledge acquired from print processing.
One ongoing question in the rapid serial naming literature is its overlap with
phonological awareness. Consistent with longitudinal data from Manis et a1. (1999),
and Torgesen et a1. (1997), kindergarten phonological awareness and rapid naming
each contributed unique variance to Grade 2 literacy measures when entered without
controlling for early word-reading or oral vocabulary levels.
Two findings indicate that processing speed alone is not the factor shared by
rapid serial naming and later reading, and these findings are consistent, in general
with studies reported in the research literature. Consistent with Bowers' work
(Bowers, 1995; Bowers et aI., 1999) the finding that letter naming contributed
unique variance beyond the contribution of color naming, supports rapid processing
of orthography rather than general processing speed as the critical factor here.
Consistent with a study reported by Torgesen, Wagner, Rashotte, Burgess, and
Hecht (1997) the correlation (r = .78) between letter naming and number naming,
the two symbol-based subtests, was stronger than these subtests' correlations with
the other sub tests.
Torgesen, Wagner, Rashotte, Burgess, and Hecht (1997) reported that rapid
naming in Grades 2 and 3 failed to contribute unique variance to later reading
measures once early reading ability had been entered into hierarchical multiple
regressions, whereas phonological awareness did make additional contributions.
Mannis et al. (1999) reported results inconsistent with this, finding that rapid naming
contributed unique variance, beyond the contribution of phoneme deletion, on tasks
requiring skill in arbitrary associations.
Several pieces of data here need careful interpretation. First the unique
contribution of Kindergarten RAN scores to spelling -ed endings beyond the
contributions of early word reading, vocabulary, and phonological awareness,
without this pattern emerging for the more general WIAT spelling list, suggests that
arbitrary associations are an important factor. Keep in mind that rapid naming also
made a unique contribution to rate of word reading in Grade 2. These findings
support the notion of rapid serial naming as a highly complex task involving both
sensitivity to arbitrary orthographic patterns and orthographic processing rate.
7.1 A Model of Acquisition of Orthographic Conventions
A possible interpretation of these data is a model of the shift from alphabetic to

orthographic reading consistent with Frith's focus on reading as facilitator of
spelling patterns. Although spelling, itself, is not a rapid process in Grade 2 children,
the knowledge of orthographic conventions that contributes to skilled spelling is
gained during rapid scanning of letters, units, and words during reading. This
interpretation reconciles Mannis et al. 's (1999) model of rapid serial naming as
sensitivity to arbitrary associations and Bowers et al. 's (1999) model in which the
RAN measures sensitivity to orthographic patterns under speeded conditions.
7.2 Implications for Instruction

Torgesen, Wagner, Rashotte, Alexander et al. (1997) have provided cautions about
the use of correlational data without considering the effects of changes due to
instruction over time. They cited Bowers' study (1995) in which children
categorized as having deficits early in the study no longer exhibited these deficits by
Grade 4. Until recently, little has been known about effective instruction to help
children compensate for deficits in rapid serial processing. Wolf (1999) has
described this literature, including her own program in retrieval, automaticity,
vocabulary, elaboration, and orthography (RAVE-O). Tan and Nicholson (1997)
have used flash cards to increase automaticity in sight-word reading, with resultant
growth in comprehension in weak readers. Ebaugh (2000) has reported positive
effects for fluency training in first graders who fit Bowers and Wolf's (1993)
description of children with a double deficit. While this is encouraging, additional
training studies need to be carried out to refine what is known about successful
instruction for weak readers such as the children in these treatment studies, and for
all children making the transition from alphabetic to orthographic reading.
312 J. K. UHRY
8. AFFILIATIONS
Dr. Joanna K. Uhry
Fordham University
Graduate School of Education, Room 1102
113 West 60th Street
New York, NY 10023 USA
e-mail: joannauhry@aol.com
9. REFERENCES
Adams, M. 1. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: The MIT
Press.
Ball, E. W., & Blachman, B. A. (1991). Does phoneme segmentation training in kindergarten make a
difference in early word recognition and developmental spelling? Reading Research Quarterly, 26,
49-66.
Bowers, P. G. (1995). Tracing symbol naming speed's unique contributions to reading disabilities over
time. Reading and Writing: An Interdisciplinary Journal, 7, 189-216.
Bowers, P. G., Sunseth, K., & Golden J. (1999). The route between rapid naming and reading progress.
Scientific Studies of Reading, 3, 1, 31-53.
Bowers, P. G., & Swanson L. B. (1991). Naming speed deficits in reading disability: Multiple measures
of a single process. Journal of Experimental Child Psychology, 51, 195-219.
Bowers, P. G., & Wolf M. (1993). Theoretical links among naming speed, precise timing mechanisms
and orthographic skill in dyslexia. Reading and Writing: An Interdisciplinary Journal, 5, 69-85.
Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning to read - a causal connection.
Nature, 301, 419-421.
Bradley, L., & Bryant, P. E. (1985). Rhyme and reason in reading and spelling. Ann Arbor, MI:
University of Michigan Press.
Bryant, P. E., Nunes, T., & Bindman, M. (1997). Children's understanding of the connection between
grammar and spelling. In B. Blachman (Ed.), Foundations of Reading Acquisition and Dyslexia (pp.
219-240). Mahwah, NJ: Erlbaum.
Denckla, M. B., & Cutting, L. E. (1999). History and significance of rapid automatized naming. Annals of
Dyslexia, 49, 29-42.
Denckla, M. B., & Rudel, R. G. (1974). "Rapid Automatized Naming" of pictured objects, colors, letters,
and numbers by normal children. Cortex, 10, 186-202.
Denckla, M. B., & Rudel, R. G. (1976). Rapid automatized naming (RAN): Dyslexia differentiated from
other learning disabilities. Neuropsychologia, 14, 471-479.
Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabulary Test - Revised. Circle Pines, MN:
American Guidance Service.
Ebaugh, 1. (2000). The effects of fluency instruction on the literacy development of at risk first graders.
Unpublished Dissertation. New York, NY: Fordham University.
Ehri, L. C. (1992). Reconceptualizing the development of sight word reading and its relationship to
recoding. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading Acquisition (pp. 107-143).
Hillsdale, NJ: Erlbaum.
Ehri, L. c., & McCormick, S. (1998). Phases of word learning: Implications for instruction with delayed
and disabled readers. Reading and Writing Quarterly: Overcoming Learning Difficulties, 14, 135-
163.
Ehri, L. c., & Saltmarsh, 1. (1995). Beginning readers outperform older disabled readers in learning to
read words by sight. Reading and Writing: An Interdisciplinary Journal, 7, 295-326.
Felton, R. H., & Brown, I. S. (1990). Phonological processes as predictors of specific reading skills in
children at risk for reading failure. Reading and Writing: An Interdisciplinary Journal, 2, 39-59.
Foorman, B. R., Francis, D. 1., Fletcher, 1. M., Schatschneider, C., & Mehta, P. (1998). The role of
instruction in learning to read: Preventing reading failure in at-risk children. Journal of Educational
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. E. Patterson, 1. C. Marshall, & M.
Coltheart (Eds.), Surface dyslexia: Neuropsychological and cognitive studies of phonological reading
(pp. 301-330). London: Lawrence Erlbaum Associates, Ltd.
Lundberg, I., Frost, J., & Petersen O. P. (1988). Effects of an extensive program for stimulating
phonological awareness in preschool children. Reading Research Quarterly, 23, 263-284.
Mannis, F. R., Seidenberg, M. S., & Doi, L. M. (1999). See Dick RAN: Rapid naming and the
longitudinal prediction of reading subskills in first and second graders. Scientific Study of Reading, 3,
2, 129-157.
Olson, R. K., Wise, B. Conners, F. Rack, 1. P., & Fulker, D. (1989). Specific deficits in component
reading and language skills: Genetic and environmental influences. Journal ofLearning Disabilities,
22, 339-348.
Rosner, J. (1975). Test of auditory analysis skills. In 1. Rosner (Ed.), Helping children overcome learning
difficulties (pp. 46-49). New York: Walker and Co.
Rosner, J., & Simon, D. P. (1971). The auditory analysis test: An initial report. Journal of Learning
Seidenberg, M. S., & McClelland, 1. L. (1989). A distributed, developmental model of word recognition
and naming. Psychological Review, 96, 523-568.
Stanovich, K. E., & Siegel, L. S. (1994). Phenotypic performance profiles of children with reading
disabilities: A regression-based test of the phonological-core variable differences model. Journal of
Educational Psychology, 86, 24-53.
Tan, A., & Nicholson, T. (1997). Flashcards revisited: Training poor readers to read words faster
improves their comprehension of text. Journal of Educational Psychology, 89, 276-288.
Torgesen, J. K., Wagner, R. K., Rashotte, C. A., Alexander, A. W., & Conway, T. (1997). Preventive and
remedial interventions for children with severe reading disabilities. Learning Disabilities: A
Multidisciplinary Journal, 8, 51-61.
Torgesen, J. K., Wagner, R. K., Rashotte, C. A., Burgess, S., & Hecht, S. (1997). Contributions of
phonological awareness and rapid automatic naming ability to the growth of word-reading skills in
second- to fifth-grade children. Scientific Study of Reading, 2, 161-185.
Uhry, J. K. (1993). Predicting low reading from phonological awareness and classroom print: An early
reading screening. Educational Assessment, 1, 349-368.
Uhry, J. K. (1999). Phonological and orthographic development in first and second graders' spellings.
Paper presented at the annual meeting of the Society for the Scientific Study of Reading in Montreal.
Uhry, J. K., & Ehri, L. C. (1999). Children'S reading acquisition. In D. A. Wagner, R. L. Venezky, & B.
V. Street (Eds.), Literacy: An International Handbook (pp. 43-48). New York: Westview Press.
Uhry, J. K., & Shepherd, M. J. (1993). Segmentation/spelling instruction as part of a first grade reading
program: Effects on several measures of reading. Reading Research Quarterly, 28, 218-233.
Uhry, J. K., & Shepherd, M. J. (1997). Teaching phonological recoding to young children with dyslexia:
The effect on sight vocabulary acquisition. Learning Disabilities Quarterly, 20, 104-125.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the
Wechsler, D. (1992). Wechsler Individual Achievement Test. San Antonio, TX: Psychological
Corporation.
Wolf, M. (1991). Naming speed and reading: The contribution of the cognitive neurosciences. Reading
Research Quarterly, 26, 123-141.
Wolf, M. (1999). What time may tell: Towards a new conceptualization of developmental dyslexia.
Annals of Dyslexia, 49, 3-28.
Woodcock, R. W. (1987). Woodcock Reading Mastery Tests-Revised. Circle Pines, MN: American
Guidance Service.
A.DAHLGRENSANDBERG
PHONOLOGICAL RECODING PROBLEMS IN

CHILDREN WITH SEVERE CONGENITAL SPEECH
IMPAIRMENTS - THE IMPORTANCE OF
PRODUCTIVE SPEECH
Abstract. Children with severe expressive speech impairments usually show significant difficulties in
acquiring reading and spelling skills. Reading and spelling abilities in nonvocal pre-school children
(chronological age 5 - 7 years) with cerebral palsy were subject to study. In the study, major reading and
spelling difficulties were found. The results were compared with the results of IQ-matched children.
Precursors to acquisition of literacy skills, previously identified to be of importance in the vocal population,
were assessed in all groups. These were verbal and visual memory and phonological awareness. The groups
performed equally well on the different indicators of phonological awareness that were used. The groups
differed in verbal short-term memory but performed on a comparable level on a visual memory task. Inability
to articulate might be an important factor in the reading and spelling problems found in the nonvocal
population, since lack of speech could make it more difficult to create and maintain an auditory
representation in working memory, stable enough to allow elaboration during decoding and encoding.
1. THEORETICAL BACKGROUND
The reading and spelling problems of children with cerebral palsy and congenital
speech production impairments have been subject of interest in the last decades (e.g.
Berninger & Gans, 1986; Blischak, 1994; Dahlgren Sandberg & Hjelmquist, 1996a, b,
1997; Foley, 1993; McNaughton, 1998). A common finding has been an inconsistency
between their reading and spelling levels and their general intellectual and verbal
abilities. Later research has also shown that the difficulties are persistent into adulthood
(Foley & Pollatsek, 1999). Children with such severe disabilities are surrounded by
barriers to their intellectual and linguistic development. They have gross motor as well
as fine motor problems, which make it difficult for them to handle books, paper and
tools for writing experiences. Their articulatory difficulties slow down and even hinder
communicative activities. They have limited expressive, and sometimes also receptive,
vocabulary, possible memory problems and limited phonological ability (Blischak,
1994; Dahlgren Sandberg & Hjelmquist, 1996a, b, 1997; McNaughton, 1998). These
factors, one by one or in combination, can contribute to their problems to acquire
reading and spelling skills. Nevertheless, literacy skills are of great importance to
persons that rely on augmentative and alternative means for communication (AAC) in
the absence of natural speech. Low technology AAC methods often imply a restrictive
vocabulary, normally communication is slow and a third person is often needed for
communication. Access to literacy skills would make it possible for persons with
speech production problems to use the written code for communication. Functional

316 A. DAHLGREN SANDBERG
literacy skills are also needed to be able to take fully part in society and to take
advantage of the rapidly developing sophisticated technology that could enhance social,
educational and working possibilities.
1.1 Phonological Awareness and Phonological Recoding
In order to identify the reasons for the reading and spelling difficulties of children with
severe congenital speech impairments, research has focused on studying how their
reading and spelling acquisition is influenced by factors that has been identified as
precursors to literacy acquisition in typically speaking children. Many studies have
highlighted phonological awareness, the awareness of the sound structure in spoken
language, and phonological recoding, the detection and use of the correspondence
between a phonological and a graphemic representational system as the most important
and unequivocal predictors of early reading and spelling ability (e.g., Elbro, 1996; Ellis
& Large, 1988; Gleitman & Rozin, 1977; Goswami & Bryant, 1990; H0ien, Lundberg,
Stanovich, & Bjaalid, 1995; Lundberg, Frost, & Petersen, 1988; Lundberg & H0ien,
1991). Different aspects of phonological awareness have been proposed, for example
awareness of phonological strings (a global, nonanalytic level of awareness), of
syllables, of onset and rime (rhyme) and of phonemes (Adams, 1990; Goswami &
Bryant, 1990; H!2Iien et al., 1995; Morais, 1991). These factors, varying in complexity
and difficulty (Goswami & Bryant, 1990; Lundberg, Frost, & Petersen, 1988; Yopp,
1988), emerge at different stages in the child's linguistic development (Goswami &
Bryant, 1990), with phonemic awareness being the most important one for literacy
acquisition.
1.2 Short-Term Memory
Memory capacity is also most certainly influential on the acquisition of reading and
spelling skills (Ellis & Large, 1988; H0ien & Lundberg, 1992). Of interest are the
memory processes at work during the fast processes of decoding and encoding. These
processes involve the use of phonological coding in working memory (Baddeley, 1999;
Wagner & Torgesen, 1987). Stable phonological representations need to be constructed
and stored in verbal short-term memory, assumed to serve as a working memory system
when spelling and reading (Gathercole & Baddeley, 1993; H0ien & Lundberg, 1992).
Baddeley and Hitch (1974) introduced the term "working memory" to describe the
short-term memory system which is thought to be involved in the storage and
temporary processing of information and argued that working memory plays a central
role in linguistic activities. The model contains three main components: the central
executive; and two "slave-systems", the phonological loop and the visuo-spatial
sketchpad. The central executive serves as an attentional control system, while the
phonological loop is specialized for storage of verbal material or recoding of nonverbal
information. The material is assumed to be represented in a phonological code that
decays with time, unless rehearsed in an articulatory process. Thus, the phonological
loop is described as consisting of two components, a subvocal rehearsal process and a
phonological store. The visuo-spatial sketchpad is engaged in processing and storage of
visual and spatial information (Gathercole & Baddeley, 1993). According to Gathercole
PHONOLOGICAL RECODING PROBLEMS 317
and Baddeley (1993), the visuo-spatial sketchpad has little to do with language
activities.
Concerning the relationship between linguistic and literacy abilities and short-term
memory (STM), most studies have shown shorter STM spans for verbal material in
poor readers (e.g. Bowey, Cain, & Ryan, 1992; Hansen & Bowey, 1994; Newman,
Fields, & Wright, 1993; Siegel & Ryan, 1988; Snowling, 1991). This has not been the
case for nonverbal material that is difficult to code linguistically, (e.g., a test of block
span) (Gould & Glencross, 1990), and therefore, according the model of Baddeley and
Hitch (1974) would be treated in the visuo-spatial sketchpad. Hicks (1980), using the
visual sequential memory task from the Illinois Test of Psycho linguistic Abilities
(ITPA) (Kirk, McCarthy, &Kirk, 1968), found that competent readers tended to use a
labeling strategy during recall. Further, she showed that poor retention of visual stimuli
could be improved by adoption of a labeling strategy. In a case study of a graduate
student with an exceptionally restricted verbal STM, Baddeley (1993) demonstrated
normal STM for visual material and low results on a test for recognition and recall of
names and also impairments in new language learning. The reading performance of the
student was good while he had problems to spell correctly. He reported that he had
been subject to intensive training and that he still used mnemonics to be able to
maintain the spelling level. Taken together, these findings indicate a connection
between linguistic and literacy ability on the one hand, and verbal coding in subvocal
rehearsal and STM on the other.
1.3 Articulatory Ability and Subvocal Rehearsal
A specific issue for the study of early reading and spelling acquisition in children with
severe speech impairments is the role of articulatory ability and subvocal rehearsal in
the development of phonological awareness and phonological decoding. Studies of
individuals with severe congenital speech impairments have demonstrated good
performance on different tasks measuring phonological awareness (Dahlgren Sandberg
& Hjelmquist, 1996a, b, 1997; Foley, 1993; Foley & Pollatsek, 1999). They
demonstrate difficulties, though, when it comes to use of this ability in decoding and
encoding activities. Here, the differences between processes of perception and of
elaboration might be important. It is acknowledged that understanding of a spoken
word among persons with a spoken language is possible without having ever produced
the word. In our research we have also found that persons without any possibility to
produce speech sounds still perceive the spoken language, and are attentive to fine
distinctions (Hjelmquist, Dahlgren Sandberg, & Hedelin, 1994). A possible explanation
to the incongruence between the reading and spelling difficulties and the indices of
good phonological awareness in these persons, would be that, on the one hand, tasks
measuring phonological awareness that do not demand extensive processing will be
subject to a perception process. On the other hand, decoding and encoding demand
elaboration of the sound structure and as such put a heavy load on working memory,
and subvocal rehearsal would be important.
When Baddeley and Hitch (1974) formulated their theory about working memory,
they used the term "articulatory" loop rather than "phonological", indicating an idea
that motor speech acts are needed for rehearsal to take place. In a later study of adult
patients with acquired anarthria, Baddeley and Wilson (1985) showed that articulatory
or phonological coding can take place in verbal STM without such speech acts. They
argued that subvocal speech develops in children from overt speech and that overt
articulation looses its importance in adult subvocal rehearsal. Also Bishop and Robson
(1989) argued from the results in their study of persons both with congenital and
acquired anarthria that overt articulation is not a necessary condition for subvocal
rehearsal. However, it has been demonstrated that children with severe speech
disorders, demonstrating slow articulation, also perform worse on memory tasks than
children with a natural speech (Raine, Hulme, Chadderton, & Bailey, 1991). Children
with congenital anarthria have never been able to produce any speech sounds. This
might effect negatively the development of subvocal rehearsal in these children, with
problems in working memory and reading and spelling problems as a consequence.
Another line of research that has bearings on the reading and spelling difficulties in
children with severe congenital speech impairments is the work of Case, Kurland and
Goldberg (1982). In their theoretical model, a system of limited capacity was
postulated. This system was thought to account for both temporary storage and
processing. Since the capacity is limited, children who meet novel stimuli, or who for
other reasons have problems with the processing, have little capacity left over for
storage. This would explain the increasing verbal STM span in children who at young
ages are not acquainted with the material, e.g. the digits in the digit span task, and need
to concentrate on the processing. Nonvocal children have demonstrated significantly
lower levels of verbal STM span than children with a spoken language (Dahlgren
Sandberg & Hjelmquist, 1996a, b; 1997). In accordance with the theoretical
explanations of Case et al. this might be accounted for by the children's difficulty with
subvocal rehearsal, the processing, and consequently less capacity for temporary
storage in the phonological loop system.
2. EMPIRICAL STUDIES
Examples from my studies will be used to clarify the theoretical viewpoints above.
Children with severe congenital speech impairments, seven pre-school children, five to
seven years of age, participated in a study of phonological awareness, working memory
and the relationships between these variables on the one hand and the children's
reading and spelling acquisition on the other. The aim was to test what part, if any,
articulatory speech plays in the development of phonological awareness and literacy
abilities. This could also shed some light over the question of subvocal rehearsal in
working memory.
2.1 Participants' Characteristics
The children all used Bliss as their primary communication mode or as a complement
and parallel to other communication modes, such as a few manual signs, gestures, body
movements, facial expressions or vocalizations. They were all anarthric. Except for an
ability to express "yes" and "no" in spoken language in some cases, none of the
children had any speech that was intelligible to unfamiliar listeners. Their performance
was compared with that of typically developing children with natural speech. All the
children in the control group had normal hearing and had had a normal development of
communication, language and speech. There were no differences between the groups in
chronological age or mental age as measured by Raven's progressive matrices, colored
version (Raven, 1965) (Table 1).
Pre-school children were chosen as participants since problems with subvocal
rehearsal in working memory seems to be most detrimental in early reading and
spelling acquisition. At later stages the effects might be masked by general knowledge
and access to other strategies. Pre-school age was also preferred to avoid that different
amounts and quality of formal teaching of literacy skills entered as a confounding
variable.
Table 1. IQ, Chronological and Mental Age. Speech Impairment and Control Groups.
Participants Female Male Chronological age Mental age
n n Mean Range Mean Range

Speech impairment
group 6 1 6.4 5.1-7.3 5.6 5.0-8.0
Control group 6 1 6.2 4.8-7.2 6.5 5.0-9.5

P-value p=.638 p=.091
2.2 Phonological Awareness
All the children, irrespective of spoken language abilities, demonstrated evidence of

phonological awareness, with the measures used in the study (Figure 1). There were no
statistically significant differences between the groups. The four different measures
used were chosen to tap different levels of awareness, demanding different types and
amounts of processing and consequently putting different loads on working memory
(Yopp, 1988). According to this, the prediction would be that rhyme identification,
working as it is at subword level, (onset and rhyme (Goswami & Bryant, 1990», and
therefore being the least demanding, would be the easiest one. Next would follow
phoneme synthesis, sound identification and judgment of word length, in that order.
These three imply phonological awareness on phoneme level. On the word length task,
the children with severe speech impairments presented somewhat lower results than did
the children with natural speech. On the other hand, they performed better than the
vocal children on the sound identification task did. The order of difficulty followed the
predicted one in the case of the children with speech impairments. Most important,
these results indicate that development of phonological awareness is possible in the
absence of productive speech. The order of difficulty also indicates use of a
phonological strategy on behalf of the nonvocal children. Following the line in the
discussion above, their success is quite reasonable postulating that the basic forms of
phonological awareness measured in the current study engage immediate perception of
spoken entities, rather than demand elaboration. Consequently, they put little load on
working memory.
100
t)
~
00
70
ro
1:=-1
~
OJ
50
~ 40
30
Xl
10
0
Rhyrre &lJnI
iWUificatirn iWUificatirn
Figure 1. Phonological awareness.
To examine this further, three to four years later the same children were tested on the
same variables, with two more indicators of phonological awareness added; deletion
and segmentation of phonemes. These were considered to demand more processing and
thus to be of greater difficulty to the children with severe speech impairments.
The resulting picture was the same as in the pre-school study, though: good
performance in both the disability and the control group, with the deletion and the word
length tasks being most difficult to all children. At the second occasion, children with a
spoken language performed significantly better on the word length task. The word
length task is one of the measures of phonological awareness that probably benefits
most from being able to read, and as will be seen later, the vocal children outperformed
the children with anarthria on the reading and spelling variables. Thus, the often cited
Matthew-effect (Stanovich, 1986), "the more you read, the better you get", could easily
explain this result.
2.3 VerbaL and VisuaL STM

The two tests, pertinent to the question of phonological or visual coding in working
memory was a digit span task for verbal memory, and a visuo-spatial test of memory,
the Corsi blocks task (see Rapala & Brady, 1990). In the first task the participants
repeated the numbers spoken by the experimenter. In the second, four to nine blocks
were placed on a sheet of paper. The examiner slowly pointed at the blocks in random
order. The task of the participants was to repeat the pointing. The hypothesis was that
the children with severe speech impairments would perform worse than the children
with a spoken language on then Digit span task due to their supposedly greater
difficulty with subvocal rehearsal in working memory. Since no verbal coding was
supposed to occur with the Corsi blocks task, all children were thought to perform
equally well.
II Speech impairment
II Comparison
Digit span Corsi blocks
Figure 2. Mean number recalled for the two memory tasks.
The results were in line with this hypothesis (Figure 2) and the picture was consistent at
a later occasion, three to four years later. The children with natural speech performed
significantly better than the children with severe speech impairments on the Digit span
task at both test occasions and there were no differences on the test for visual memory.
2.4 Phonological Recoding in Reading and Spelling

Having seen the results on the tests for phonological awareness and for verbal and
visual memory, the expectations for the reading and spelling results could be the
following two. On the one hand, the seemingly good performance on the phonological
awareness tasks demonstrated by the children with no functional speech would predict
possibility of phonological recoding and no differences between the two groups on
reading and spelling measures. On the other hand, the significant differences between
the groups on the Digit span task would predict worse results for the nonvocal children,
unless they would be able to use their visual memory capacities for some other strategy,
e.g., whole word reading. The subvocal rehearsal needed for active manipulation ofthe
sound structure in the dictated words would be inefficient, if not totally absent.
Consequently, the children would have great difficulties to find the correspondences
between the phonemes and the graphemes. Spelling in the case of no overt sound image
and silent reading, both processes where the children have to create the phonological
representations by themselves, would be even more difficult. It would be highly
improbable that the children could use an orthographic route, since they were at a stage
of emergent literacy.
The results pointed at clear difficulties in the nonvocal group. At the age of five to
seven all children with a spoken language demonstrated reading and spelling abilities,
whereas only some of the children with severe speech impairments tried to spell a few
words but with a very meager result (Tables 2a, b and c). They showed practically no
reading ability (Table 3).
2.4.1 Phonological Encoding o/Words and Nonwords

Three different spelling tasks were used. In order to assess phonological encoding
ability, the children were asked to spell nonwords besides spelling of dictated words of
high frequency. The children in the disability group tried some spelling of the dictated
words (Table 2a), but when presented with the nonwords (imaginary names of
monsters) they did not even try to spell (Table 2b). Typically, they demonstrated
interest in the pictures of the monsters but showed clearly that they had no idea about
how to spell their names. The children in the control group were instead eager to try the
spelling both of the dictated words and the nonwords. They were quite successful at
giving correct versions of the words and they demonstrated their ability to encode
phonologically achieving as high a percentage correct for the nonwords as for the
words. If they did not succeed the spelling completely, they sometimes gave
phonetically correct versions of both words and nonwords.
Table 2a. Spelling of Words, Oral Presentation. Mean Percent Correct.
Response measure Words Words Phonetically Letters Sounds

Group attempted correct correct correct correct
Speech impairment 4.8 0.0 0.0 1.0 1.9
Control 54.8 30.4 38.7 50.0 52.7

P-value .020 .023 .020 .012 .014
Table 2b. Spelling of Nonwords, Oral Presentation. Mean Percent Correct.

Speech impairment 0.0 0.0 0.0 0.0 0.0
Control 69.6 33.9 37.5 51.0 52.7

P-value .002 .025 .019 .008 .008
2.4.2 Phonological Encoding o/Visually Presented Material

In the third spelling task, the children were asked to give the spelling of objects
presented on photographs. This task was introduced to see how the children managed
spelling when they were forced to produce the phonological representations by
themselves. This task was thought to add to the difficulty of encoding for all children,
since no spoken representation was given. In case articulatory speech is helpful in early
spelling it would be even more difficult for the children with anarthria. The objects on
the photographs were well known to all the children. There were e.g., pictures of a cat,
a house, and a lamp. This task turned out to be difficult for all children and more so for
the children in the "speech impairment" group (Table 2c). The children in the control
group typically pronounced the word and thereafter they started to spell. The children
in the disability group told that they were not able to spell the words. They showed by
means of their AAC-methods that they recognized the object and as soon as the
experimenter pronounced the word (outside the experimental session), they tried to
spell. Lack of articulatory speech made it impossible for them to use overt rehearsal
and, as it seems, it made it difficult for them to engage in subvocal rehearsal as well.
Table 2c. Spelling, Visual Presentation. Mean Percent Correct.

Speech 7.1 0.3 0.3 0.3 0.3

impairment
Control 69.6 12.5 16.1 42.7 46.8

P-value .007 .045 .060 .009 .0lD
Mter three to four years of formal literacy training, the differences between the two
groups persisted on all spelling tasks. The same testlists were used also at this second
measurement. The children with severe speech impairments had made significant
improvements, with less gain for the visually presented material. This task was still
hardest also in the control group.
2.4.3 Analysis of Spelling Errors

To further examine if and how the children with anarthria used their demonstrated
phonological awareness, an analysis of the spelling errors was made. The analysis
revealed that the nonvocal children frequently made omissions (50%) or introduced
letters (30%) that were not related to the word that they were supposed to spell. Only
approximately 13% of the errors were phonetically acceptable versions of the words,
indicating that they did not use their phonological abilities during spelling. Phonetically
acceptable versions occurred almost only in the first task: dictation of words (33%),
where the children were presented with a "sound image" of the well-known words.
Such spellings were rare in the two other tasks, approx. 2%. This strengthens the
picture of their difficulty to encode phonetically. In the control group the majority of
the few errors committed were of the phonetically acceptable type, 55%.
2.4.4 Decoding Abilities

At pre-school time, decoding and reading for meaning, was more difficult than spelling
in the control group. Decoding abilities were almost absent among the children with
speech impairments (Table 3), confirming the line of reasoning above, that
phonological recoding is hindered in the absence of a spoken representation of the word
in this group of children.
When the same reading and spelling tasks were presented to the children after three
to four years at school the gap between the groups had increased on all measures.
Further, the children with anarthria still found reading more difficult than spelling, with
an exception for spelling of visually presented material that was the most difficult task
of all tasks to these children.
Table 3. Reading Tests, Mean Percent Correct
Type of test Lexical Proofreading Comprehension Decoding

Group decision (Umesol) (OS400)
Speech impairment 5.0 0.0 0.0 0.0
Control 15.0 10.8 13.3 26.6

P-value .363 .053 .337 .149
3. CONCLUDING REMARKS
On the preceding pages, examples have been presented concerning the difficulties of
phonological recoding in children with severe congenital speech impairments. Despite
of the evidence of good phonological awareness, reading and spelling performance is
far behind that of children with natural speech. One possible explanation of the
difficulty to use a phonological strategy is that the difficulties result from problems
with subvocal rehearsal in working memory in the absence of functional speech.
There is a difference between the tasks for phonological awareness on the one hand
and of phonological recoding abilities on the other. Whereas the phonological
awareness tasks mostly could be handled without much processing and thus not
occupying much of the capacity in working memory, decoding and encoding both
demand active manipulation at phoneme level and rely on subvocal rehearsal in
working memory. Productive speech does not seem to be a necessary condition for
perception of the sound structure of spoken language, but in the absence of subvocal
rehearsal it could be difficult to create representations of the sound structure in spoken
language in working memory, representations that are stable enough to allow
elaboration during reading and spelling.
Although the children with expressive speech impairments demonstrated good
phonological capacity according to the tests used in the cited experiments (Dahlgren
Sandberg & Hjelmquist, 1996a), there might still exist difficulties on levels of
phonological awareness that were not detected with the measures used. The children
also showed that they experienced differences in difficulty among the variables used.
Since different levels of phonological awareness demand different processing of
information and put different demands on working memory capacity, this could explain
part of the reading and spelling difficulties found among the children with no functional
speech.
Synthetic speech might be used to overcome the assumed difficulties to rehearse

subvocally and to help in creating and maintaining a stable sound representation of the
words. There is some evidence that congenitally anarthric persons have benefited from
extensive training to read and spell with support of synthetic speech output (Dahlgren
Sandberg & Hjelmquist 1997; Foley & Pollatsek, 1999). The results from studies by
Olson and Wise (1992) and Olofsson (1992) also showed positive effects of use of
synthetic- speech feedback in disabled readers.
However, as has been demonstrated, children and adults with severe congenital
anarthria also reach some reading and spelling skills. As yet it is not clear what strategy
or strategies they use to succeed. The progress made by the children in the above
mentioned studies and the phonologically acceptable versions of the words that they
produced, even if a small percentage, might still indicate the use of a phonological
strategy in spelling. This strategy did not seem to work fully, though. The high number
of unrelated errors, approx. 25%, indicates that the children also tried to solve the
spelling problems that they were exposed to in quite unexpected and irrational ways.
Some words were, no doubt, recalled as whole words.
We know from other studies that people find compensatory strategies if their
possibility to subvocalize is hindered (e.g., Baddeley, 1999, p. 52). A possibility for
persons without functional speech would be to rely on visual short-term memory. In a
study of deaf children's spelling strategies, Aaron, Keetay, Boyd, Palmatier and Wacks
(1998) found that neither a phonological nor a visually based word-specific memory
strategy seemed to be used. They proposed that the children in their study used visual
memory for patterns of letter sequences, i.e., stochastic-dependent probabilities of letter
sequences. This could help spelling to a certain extent but a phonological strategy
would be needed for words that are morphophonemically complex. Another strategy
could simply be guessing. Guessing, when a phonological strategy failed, might explain
the many omissions and unrelated errors made by the children in the above studies.
So, results from different studies indicate that a spoken productive language is not
necessary for development of at least basic levels of phonological awareness. However,
in this chapter the issue of the role of productive speech in working memory and
acquisition of reading and spelling skills in children with severe congenital speech
impairments has been highlighted. Further research is needed to pursue the issue of
working memory in children with severe congenital speech impairments and of the
importance of visual versus auditory memory in achieving good reading and spelling
abilities. Possible strategies, other than phonological, to improve spelling and reading
skills in the absence of productive speech need to be examined further. The role of
speech synthesis as a substitute for natural speech in literacy acquisition also needs to
be highlighted.
4. AFFILIATIONS
Dr. Annika Dahlgren Sandberg

Dept. of Psychology, Goteborg University, Box 500
SE-405 30 Goreborg, Sweden
e-mail: psyandas@psy.gu.se
5. REFERENCES
Aaron, P. G., Keetay, V., Boyd, M., Palmatier, S., & Wacks, 1. (1998). Spelling without phonology: A study
of deaf and hearing children. Reading and Writing: An Interdisciplinary Journal, 10, 1-22.
Adams, M.J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: The MIT Press.
Baddeley, A. (1993). Short-term phonological memory and long-term learning: A single case study.
European Journal of Cognitive Psychology, 5, 129-148.
Baddeley, A. (1999). Essentials of human memory. Hove, East Sussex: Psychology Press.
Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. Bower (Ed.), The psychology of learning
and motivation Vol. 8 (pp. 47-90). New York: Academic Press.
Baddeley, A. D., & Wilson, B. (1985). Phonological coding and short-term memory in patients without
speech. Journal of Memory and Language, 24, 490-502.
Berninger, V., & Gans, B. M. (1986). Assessing word processing capability of the nonvocal, nonwriting.
Augmentative and Alternative Communication, 2, 56-63.
Bishop, D. V. M., & Robson, 1. (1989). Unimpaired short-term memory and rhyme judgement in congenitally
speechless individuals: Implications for the notion of "Articulatory Coding." Quarterly Journal of
Blischak, D. M. (1994). Phonologic awareness: Implications for individuals with little or no functional
speech. Augmentative and alternative communication, 10, 245-254.
Bowey, J. A., Cain, M. T., & Ryan, S. M. (1992). A reading-level design study of phonological skills
underlying fourth- grade children's word reading difficulties. Child Development, 63, 999-1011.
Case, R. D., Kurland, D. M., & Goldberg, J. (1982).Operational efficiency and the growth of short-term
memory span. Journal of Experimental Child Psychology, 15, 386-404.
Dahlgren Sandberg, A., & Hjelmquist, E. (1996a). Phonological awareness and literacy in non-speaking
preschool children with cerebral palsy. Augmentative and Alternative Communication, 12, 138-153.
Dahlgren Sandberg, A., & Hjelmquist, E. (1996b). A comparative, descriptive study of reading and writing
skills among nonspeaking children - a preliminary study. European Journal of Disorders of
Communication, 31, 289-308.
Dahlgren Sandberg, A., & Hjelmquist, E. (1997). Language and literacy in nonvocal children with cerebral
palsy. Reading and Writing: An Interdisciplinary Journal, 9, 107-133.
Elbro, K. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Reading and
Writing: An Interdisciplinary Journal, 8, 453-485.
Ellis, N., & Large, 8. (1988). The development of reading: As you seek so shall you find. British Journal of
Foley, 8. E. (1993). The development of literacy in individuals with severe congenital speech and motor
impairments. Topics in Language Disorders, 13, 16-32.
Foley, B. E., & Pollatsek, A. (1999). Phonological processing and reading abilities in adolescents and adults
with severe congenital speech impairments. Augmentative andAlternative Communication, 15, 156-173.
Gathercole, S. E., & Baddeley, A. D. (1993). Working memory and language. Hove, UK: Lawrence Erlbaum
Associates Ltd ..
Gleitman, L. R., & Rozin, P. (1977). The structure and acquisition of reading I: Relations between
orthographies and the structure of language. In A. S. Reber, & D. L. Scarborough (Eds.), Towards a
psychology of reading (pp. 1-53). Hillsdale, New Jersey: Lawrence Erlbaum Associates.
Gould, 1. H., & Glencross, D. 1. (1990). Do children with a specific reading disability have a general serial-
ordering deficit? Neuropsychologia, 28, 271-278.
Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. London: Lawrence Erlbaum.
Hansen, 1., & Bowey, J. A. (1994). Phonological analysis skills, verbal working memory, and reading ability
in second-grade children. Child Development, 65, 938-950.
Hicks, C. (1980). The ITP A visual sequential memory task: An alternative interpretation and the implications
for good and poor readers. British Journal of Educational Psychology, 50, 16-25.
Hjelmquist, E., Dahlgren Sandberg, A., & Hedelin, L. (1994). Linguistics, MC and metalinguistics in
communicatively handicapped children. Augmentative and Alternative Communication, 10, 169-183.
H(Ilien, T., & Lundberg, I. (1992). Dyslexi. (Dyslexia). Stockholm: Natur och Kultur.
H(Ilien, T., Lundberg, I., Stanovich, K. E., & Bjaalid, I. (1995). Components of phonological awareness.
Reading and Writing: An Interdisciplinary Journal, 7, 171-188.
Kirk, S. A., McCarthy, J. J., & Kirk, W. D. (1968). Examiner's manual. Illinois test of psycholinguistic
abilities. Urbana, IL: University of Illinois Press.
Lundberg, I., Frost, J., & Petersen, O.-P. (1988). Effects of an extensive program for stimulating
phonological awareness in preschool children. Reading Research Quarterly, 23, 263-284.
Lundberg, I., & Hf/lien, T. (1991). Initial enabling knowledge and skills in reading acquisition: Print
awareness and phonological segmentation. In D.E. Sawyer, & B. Fox (Eds.), Phonological awareness.
The evolution of the concept (pp. 73-95). N.Y.: Springer-Verlag.
Morais, J. (1991). Constraints on the development of phonemic awareness. In S.A. Brady, & D.P.
Shankweiler (Eds.), Phonological processes in literacy (pp. 5-27). Hillsdale, New Jersey: Lawrence
Erlbaum Associates.
McNaughton, S. (1998). Reading acquisition of adults with severe congenital speech and physical
impairments: Theoretical infrastructure, empirical investigation, educational application. Unpublished
doctoral dissertation, University of Toronto, Ontario, Canada.
Newman, S. P., Fields, H., & Wright, S. (1993). A developmental study of specific spelling disability. British
Olofsson, A.. (1992). Synthetic speech and computer aided reading for reading disabled children. Reading and
Writing: An Interdisciplinary Journal, 4, 165-178.
Olson, R. K., & Wise, B. W. (1992). Reading on the computer with orthographic and speech feedback.
Reading and Writing: An Interdisciplinary Journal, 4, 107-144.
Raine, A., Hulme, c., Chadderton, H., & Bailey, P. (1991). Verbal short-term memory span in speech-
disordered children. Child Development, 62, 415-423.
Rapala, M. M., & Brady, S. (1990). Reading ability and short-term memory: The role of phonological
processing. Reading and Writing: An Interdisciplinary Journal, 2, 1-25.
Raven, 1. C. (1965). Guide to using the coloured progressive matrices. London: H. K. Lewis.
Siegel, L. S., & Ryan, E. B. (1988). Development of grammatical sensitivity, phonological and short-term
memory skills in normally achieving and learning disabled children. Developmental Psychology, 24, 28-
37.
Snowling, M. (1991). Developmental reading disorders. Journal of Child Psychology and Psychiatry, 32, 49-
77.
Stanovich, K E. (1986). Matthew effects in reading: Some consequences of individual differences in the
acquisition of literacy. Reading Research Quarterly, 21, 360-407.
Yopp, H. K. (1988). The validity and reliability of phonemic awareness tests. Reading Research Quarterly,
23, 159-177.
Wagner, R. K, & Torgesen, J. K (1987). The nature of phonological processing and its causal role in the
N. GREGG, C. COLEMAN, R. STENNETI, M. DAVIS,
K. NIELSEN, D. KNIGHT, & c. HOY
SUBLEXICAL AND LEXICAL PROCESSING OF

YOUNG ADULTS WITH LEARNING DISABILITIES
AND ATTENTION DEFICIT/ HYPERACTIVITY
DISORDER
Abstract. The purpose of this research was to investigate the performance of three groups of English-
speaking university students with disabilities, which included: (a) 93 with learning disabilities (LD); (b)
64 with attention deficit/hyperactivity disorder (AD/HD); and (c) 54 with LD + AD/HD. The profiles of
the students with disabilities were compared to those of 100 of their normally-achieving peers across
several phonological and orthographic processing measures, as well as standardized measures of reading-
decoding (of real and nonsense words), spelling, listening comprehension, and reading comprehension. In
addition, analyses were conducted to determine which measures were predictive of performance on
selected measures of reading and spelling abilities. Implications for assessment, intervention, and
accommodations are provided.
1. INTRODUCTION
The persistence of developmental dyslexia and other types of LD (e.g., specific

written expression, reading comprehension) into adulthood has been documented
throughout the literature (Bruck, 1985, 1990, 1993a; Finnucci, Gottfredson, &
Childs, 1986; Johnson & Blalock, 1987; Rawson, 1968). However, research
investigating the breakdown related to decoding (word recognition) and encoding
(spelling) among adults has received less attention than that for children. Problems
with definition, eligibility criteria, and methodology for the adult population have
often led to conflicting outcomes based on the scant empirical data available to
professionals. Yet the adult population provides a means by which to explore
fundamental questions related to whether dyslexia should be attributed to
developmental delay, deviance, or arrest.
Recent interest in the adult population demonstrating AD/HD has led researchers
to also investigate the relationship of attention to the attainment of the sub lexical
and lexical processes required for word recognition and spelling. Careful
examination of the adult profiles of individuals demonstrating LD, AD/HD, and LD
+ AD/HD will contribute to a better understanding of the cognitive and linguistic
processes impacting reading (decoding and comprehension) and writing (spelling
and written expression). Such information will lead to better identification,
intervention, and accommodation of these adult populations in the educational and
work environment.

330 GREGG, COLEMAN, TENNEIT, DAVIS, NIELSEN, KNIGHT, & HOY
2. KNOWLEDGE OR PROCESSING DEFICIT?

One of the foremost issues related to the adult population of poor readers is whether
poor word recognition and spelling performance should be attributed to a
developmental delay, a knowledge deficit, or a processing deficit. (In his chapter in
this compendium, Jimenez Gonzalez provides an excellent review of the issues
related to delay and deficit hypotheses as explanations of inadequate reading
development.) Bruck (1992) concluded from research findings that the term arrest,
rather than delay, is more accurate in characterizing the word-recognition skills of
adults with dyslexia since the population continues to demonstrate significant
phonological and orthographic processing deficits across the lifespan. The results of
other empirically-based studies provide evidence that adults with dyslexia
demonstrate poorer skill on many phonological and orthographic tasks than other
adults matched for age or reading ability (Booth, Perfetti, MacWhinney, & Hunt,
2000). Bruck (1992) also investigated whether the poor performance of adults on
reading-decoding and spelling measures was indicative of knowledge or processing
deficits. A knowledge deficit, according to Bruck, refers to the inability to make use
of sounds and spellings due to a lack of knowledge of the associations between
them, whereas a processing deficit refers to an artificial independence demonstrated
between orthographic and phonological codes. She found that even when adults with
dyslexia have grapheme-phoneme knowledge, they do not "automatically activate
available orthographic codes for phonological tasks" (p. 884). Further specificity in
the debate of knowledge versus processing deficit requires closer investigation of
the adult population with reading disorders.
3. MODELS OF DECODING AND ENCODING DEVELOPMENT

Prior to reviewing the relationship of sub lexical deficits to word-recognition and
spelling deficits, it is important to consider normal decoding and encoding
development. Several theories and models have been offered in order to explain the
development of spelling to sound correspondences. Underlying most of the models
explaining the sublexical variance involved in decoding and encoding is an
agreement that spelling and sound playa significant role; however, the importance
and manner of development of these abilities are debated. For instance, Coltheart
(1978) argued that decoding is based upon knowledge of a set of rules which relates
to spelling and sounds in English, whereas Glushko (1979) proposed that readers
synthesize pronunciations on the basis of analogies between words. In his chapter in
this compendium, Galaburda raises the important question of whether humans
possess, in addition to a general sound-processing system, a special system for
processing speech sounds.
Frith (1985) proposed a three-stage theory of word-recognition development that
has served as a foundation for other theorists investigating both normal and
disordered decoding and encoding abilities across the lifespan, as well as different
orthographies. She proposed the following three stages as contributing to the
development of word-recognition: (1) the logo graphic stage (whole word based on
visual representation); (2) the alphabetic stage (sound/symbol associations); and (3)
the orthographic stage (pairing of orthographic symbols and underlying
representations). Frith (1985) then applied this model of word recognition
SUBLEXICAL & LEXICAL PROCESSING OF YOUNG ADULTS WITH LD/AD/HD 331
development to the population with dyslexia, explaining the breakdown

demonstrated by individuals with dyslexia in terms of an inability to move from
stage to stage.
Ehri (1998) incorporated the basic developmental stages proposed by Frith
(1985), stressing the grapheme-phoneme relationship as fundamental to mature word
recognition. Her stages of word recognition development include: (1) pre-alphabetic
(visual form); (2) partial alphabetic (sound/symbol association); (3) full-alphabetic
(syllable and rule focus); and (4) consolidated (fluency). Ehri's full-alphabetic and
consolidated stages have had a significant impact on professionals' understanding of
the decoding and encoding reading abilities of the adult population. The relationship
of specific cognitive and linguistic sub lexical processes to the ability to decode or
encode words at the various stages proposed by Frith and Ehri is currently being
debated by researchers. In addition, with populations demonstrating deficits in word
recognition and spelling the question of etiology versus symptomology becomes
important. For instance, a lack of fluency related to the acquisition of sublexical
information could be a symptom of a weak phonological and/or orthographic
system, or it might be the cause for the bi-directional relationship between decoding
and encoding competence.
Cross-linguistic research has also contributed significantly to a more
sophisticated understanding of the sublexical processes involved in tasks such as
decoding and encoding words. The transparency (regularity) of the orthography of
the reader's language contributes to the nature of the relationship between the
sub lexical processes in question. Beginning readers of more transparent
orthographies (e.g., German; Spanish) acquire essential phonological decoding skills
earlier and more completely than readers of nontransparent (irregular) orthographies
(Cossu, Shankweiler, Liberman, Katz, & Tola, 1988; Klicpera & Gasteiger-
Klicpera, 1993; Wimmer & Goswami, 1994; Wolf, Pfeil, Lot, & Biddle, 1994). On
the other hand, readers of nontransparent languages (e.g., English) appear to draw
significantly from the orthographic system (Wimmer & Goswami, 1994). Therefore,
reading disabilities in nontransparent orthographies like English are often identified
on the basis of impaired nonword reading (Rack, Hulme, Snowling, & Wightman,
1994) while in transparent orthographies, like German or Spanish, nonword reading
does not account for a significant amount of variance on word recognition or
spelling tasks (Wimmer, 1993; Wolf et aI., 1994). The sequence and the rate of word
recognition and spelling acquisition vary significantly across individual languages.
While the orthographic regularity of a language influences preferred instructional
methods, no reciprocal effect of teaching on word recognition or spelling acquisition
has been found (Klicpera & Gasteiger-Klicpera, 1993; Wimmer & Goswami, 1994).
Therefore, the literature supports the influence of orthographic regularity as
contributing more to the sequence and rate of acquisition of reading and spelling
abilities than instructional experiences. For a more thorough review of the contrasts
between transparent and opaque orthographies, see Jimenez Gonzalez's chapter.
Recent work by scholars endorsing a connectionist perspective has also added to
a more sophisticated understanding of word-recognition and encoding.
Connectionist theory is based upon computer networking models that involve units
thought to simulate neurons required in the processing of specific types of learning
(Seidenberg, 1997). The connectionist models proposed to date-specifically,
Seidenberg and McClelland's (1989) parallel distributed processing model of word
332 GREGG, COLEMAN, TENNETT, DAVIS, NIELSEN, KNIGHT, & HOY
recognition, and Van Orden, Pennington, and Stone's (1990) developmental bypass
hypothesis reading model--have helped researchers to better understand pertinent
cognitive and linguistic processes and their relationships to the reading process.
These connectionist models have also been helpful in making distinctions among the
different patterns exhibited by the population with dyslexia (e.g., phonological or
orthographic difficulties). The models stress that there is not a clear way to
discriminate among subgroups of dyslexia (Manis, 1999).
4. RELATIONSHIP BETWEEN PHONOLOGICAL AND ORTHOGRAPHIC

PROCESSING
The relationship between phonological and orthographic processing has been a

recent focus of researchers investigating the decoding and encoding abilities of
children and adults with dyslexia. Findings regarding the directionality of the
relationship (uni- or bi-) between phonological and orthographic processing have
been critical to the advancement of scholarship related to the development of word
recognition. Landerl, Frith, and Wimmer (1996) suggested that the connection
between these processes is automatically forged in normal readers, but not in the
population with dyslexia. In other words, the normal development of word-
recognition skills requires phonological abilities to serve as a foundation for
orthographic knowledge. In the population demonstrating dyslexia, there appears to
be a weak link between phonological and orthographic representations. "Thus,
seeing a written word does not automatically evoke the word's inner sound.
Moreover, the sound of a word does not automatically evoke the inner orthographic
image" (Landerl et aI., 1996, p.14).
The relationship of phonological and orthographic processing to the ability to
decode and encode words appears to involve a variety of sub lexical processes, as
well as the literacy experiences of an individual. The adult populations with LD and
AD/HD provide researchers with a means to examine the validity and suitability of
various models developed to explain the acquisition of words for word-recognition
and spelling skills. Connectionist models have examined the multidimensionality of
the processes involved in reading and spelling. Such models propose that words are
processed by a single system that contains orthographic, phonological, and semantic
codes that are activated when an individual is confronted with specific words.
Connectionist models have stressed the importance of orthographic as well as
phonological coding in the decoding and encoding of words. Just how similar the
processes activated by decoding and encoding tasks are in the adult population with
learning disorders in comparison to normally-achieving adult readers requires
further exploration by researchers.
5. PHONOLOGICAL PROCESSING AMONG ADULTS WITH DYSLEXIA

Numerous research studies have demonstrated the predictive ability of phonological
knowledge (i.e., phoneme segmentation, phoneme deletion, and phoneme blending)
to reading achievement among children (Gough & Tunmer, 1986; Wagner,
Torgesen, & Rashotte, 1994). In his chapter, Galaburda (this volume) suggests that
malformations in the frontal lobe may be related to the phonological awareness
difficulties demonstrated by individuals with dyslexia. Another hypothesis

supported in the literature is that phonological decoding may be a "self-teaching"
device that facilitates the establishment of orthographic representations (Bruck,
1993b; Share & Stanovich, 1995). Gradually, normally-achieving readers depend
less and less on their knowledge of spelling-sound information, recognizing words
more on the basis of direct visual orthographic information (Andrews, 1982;
Seidenberg, Waters, Barnes, & Tanenhaus, 1984; Waters, Seidenberg, & Bruck,
1984). A bi-directional relationship between phonological awareness and reading
acquisition develops for such readers (Bruck, 1993a; Perfetti, Beck, Bell, & Hughes,
1987). However, children with weak sound-spelling correspondence skills have
difficulty constructing reliable orthographic representations of words via a direct
visual route. A lack of fluent access to an age-appropriate store of orthographic
representations results in over-reliance on an already weak sound-spelling system.
The relationship between phonological processing and reading achievement has
received less attention with regard to adult readers with and without learning
disorders. In addition, the populations of adult readers that have been studied tend to
be extremely heterogeneous in terms of ability and literacy experiences (Byrne &
Ledez, 1983; Liberman, Rubin, Duques, & Carlisle, 1985; Pratt & Brady, 1988).
Yet, the outcome of most research related to adult readers with reading disabilities
points toward continuing difficulties with phonological and orthographic processing.
Recently Booth et a1. (2000) provided evidence that deficits in rapid auditory
naming ability in children are primarily associated with problems in phonological
processing. By contrast, rapid visual naming ability in children has been shown to be
more closely related to problems with orthographic processing (Farmer & Klein,
1995). However, adults with reading disorders demonstrated a strong relationship
between rapid auditory naming ability and both orthographic and phonological
processing in the Booth et a1. research.
Bruck has conducted several studies investigating the reading and spelling skills
of adults with known histories of childhood dyslexia (Bruck, 1990, 1992, 1993a,
1993b; Bruck & Waters, 1990). The populations participating in these studies were
drawn from college students who had received a clinical diagnosis of dyslexia
during childhood. Many of these individuals, as adults, demonstrated high levels of
reading comprehension and average word recognition skills on traditional
standardized measures. However, they continued to show inaccurate and particularly
slow word-recognition skills. In addition, they did not use age- or reading-
appropriate word-recognition processes. As a group, they over-relied on spelling-
sound information, syllabic information, and context for word recognition. In
particular, these adults with dyslexia did not acquire appropriate levels of phoneme
awareness (measured by phoneme counting [non digraph items] and phoneme
deletion) despite extensive educational and reading experiences. Bruck (1992)
concluded that the adult popUlation with dyslexia did not show evidence of a bi-
directional relationship between reading achievement and phonological awareness.
6. ORTHOGRAPHIC PROCESSING AMONG ADULTS WITH DYSLEXIA

The role of orthographic processing in decoding and encoding has certainly not
received the attention that phonological processing has in the literature (Berninger,
334 GREGG, COLEMAN, TENNEIT, DAVIS, NIELSEN, KNIGHT, & HOY
1994; Foorman, 1994; Roberts & Mather, 1997). Research findings have supported
the claim that part of the variance in reading and spelling performance among adults
is accounted for by orthographic processing (Cunningham & Stanovich, 1990;
Stanovich & West, 1989). Researchers investigating the heritability of dyslexia also
report equal representation of sub lexical processes (i.e., phonological and
orthographic). Recent imaging studies have further supported the idea that
orthographic processing exists separately from phonological processing (Carr &
Posner, 1994; Rumsey, Donohue, Nace, Maisog, & Andreason, 1997).
Bruck (1992) found that adults with dyslexia, when asked to make phonological
judgments, did not use orthographic information to the same extent as normal
readers of the same age or younger readers with equivalent or lower levels of word
recognition skill. "Even Grade 3 children used orthographic information to a greater
extent than high-functioning adult dyslexics" (Bruck, 1992, p. 883). According to
Bruck, the dyslexic adults' poor use of orthographic information led to a weak
understanding of the phonemic structure of language.
From a connectionist perspective, the findings from Bruck (1992) might best be
explained by an acknowledgement that a single system containing orthographic,
phonological, and semantic codes is used in the decoding and encoding of words.
Orthographic processing does play a significant role in such a model, although
deficits in orthographic processing might look different depending on the task, as
well as the experience, age, and ability level of the individual in question.
7. RELATIONSHIP OF PHONOLOGICAL AND ORTHOGRAPHIC

PROCESSING TO READING AND LISTENING COMPREHENSION
The relationship of phonological and orthographic processing abilities to listening

and reading comprehension among the adult population has received very little
attention from researchers. Research with children has provided strong evidence that
reliance on syntactical context to assist word recognition declines as a function of
age and reading abilities (Bruck, 1990; Perfetti, Goldman, & Hogaboam, 1979;
Perfetti & Roth, 1981; Stanovich, 1988). Bruck (1990), investigating the use of
context to facilitate word recognition performance, found that adults with dyslexia
read words more accurately in meaningful contexts than in neutral contexts. While
their accuracy was superior in meaningful contexts, a continued delay in reaction
time was evident. Cunningham, Stanovich, and Wilson (1990) found " ... the less
skilled readers displayed larger facilitation effects and overall context effects than
the skilled readers" (p. 141). Reading comprehension showed a moderate
relationship to context in their research. Knight (2000), using a structural equation
model, found a minimal relationship between decoding and reading comprehension
among adult readers with and without disabilities. Her work identified language
comprehension as the strongest predictor of reading comprehension.
8. PURPOSE OF STUDY
The primary purpose of this research was to investigate the contribution specific
sub lexical processes make for English-speaking university students in the ability to
decode words (real and nonsense), to recall spellings of single words, to comprehend
information presented orally (listening comprehension), and to comprehend text read

silently (reading comprehension). In addition, the performance of three groups of
university students with disabilities (LD, AD/HD, and LD + AD/HD) was compared
to that of university students without disabilities across sublexical processes
(phonological and orthographic), word fluency, word recognition (real and
nonsense), spelling recall, listening comprehension, and reading comprehension
measures. The comparisons were made in order to explore the phonological and
orthographic processing abilities that contribute to performance on reading-
decoding, spelling, listening comprehension, and reading comprehension measures
among these groups of university students. In addition, across-group comparisons
were made to determine whether similar patterns of phonological and orthographic
abilities could be identified across populations of university students with and
without disabilities.
9. METHODS
9.1 Participant Selection
9.1.1 Populations with Documented Disabilities

Three groups were identified within this category of university students. Individuals
were identified as demonstrating LD, AD/HD or LD + AD/HD. All participants in
these three categories were evaluated at The University of Georgia Regents' Center
for Learning Disorders (GA-RCLD). The evaluation of all participants included
measures of overall ability, cognitive processing, oral language, achievement, and
social-emotional functioning. Assessment instruments were selected on the basis of
their psychometric properties and usefulness with the adult population. An inter-
disciplinary team of experienced master's- and doctoral-level diagnosticians and
psychologists individually administered tests used in the evaluation process. Clinical
judgment was used to interpret test results as well as analyze error responses, writing
samples, and other data obtained from informal assessment measures. Quantitative
data included results from standardized tests and informal measures. Qualitative data
included information gathered from case histories, clinical interviews, and previous
records that confirmed the chronicity of learning problems. Both quantitative and
qualitative data were considered in a careful study of the performance of each
individual participant. No diagnoses were made on the basis of a single test score or
discrepancy measure; rather, they were based on patterns of problems and errors.
Each of the participants was classified into one of the three categories based on
the results of the evaluation process and in accordance with the Georgia Regents'
Criteria (Gregg, Heggoy, Stapleton, Jackson & Morris, 1994). The definition for LD
applied throughout this study was written by the Interagency Committee on
Learning Disabilities (1987). Participants in Group 1 (LD) were clinically diagnosed
as having a learning disability but no other developmental disability. Of the 93
students (58 females; 35 males) in Group 1, 63% had received a previous diagnosis
of learning disabilities prior to coming to the GA-RCLD. Group 2 (AD/HD but no
other developmental disability) consisted of 64 participants (31 females; 33 males),
with 52% having received a prior diagnosis of AD/HD. Group 3 (LD + AD/HD)
included individuals with a dual diagnosis of LD + AD/HD. This group consisted of
336 GREGG, COLEMAN, TENNETI, DAVIS, NIELSEN, KNIGHT, & HOY
54 participants (20 females; 34 males), 60% of whom had received a prior diagnosis
of LD and/or AD/HD.
Table 1. Clinical Diagnostic Areas of Deficit for Populations with Disability
Area of Deficit Group 1 Group 2 Group 3

(Percentage of Group) (LD) (AD/HD) (LD&
AD/HD)
Previously-identified learning problems 63% 52% 60%
No history of testing/resource class 37% 48% 40%
Phon%rtho-related weaknesses 55% 2% 35%
Auditory discrimination weaknesses 24% 2% 10%
Receptive +/- expressive syntax weaknesses 38% 10% 28%
Reading-decoding weaknesses 32% 2% 28%
Reading rate weaknesses 32% 2% 16%
Spelling weaknesses 39% 5% 28%
Language-specific cognitive deficit 64% 0% 65%
Underachievement in reading +/- spelling 83% 0% 81%
Table 2. Descriptive Psychometric Information for Population with Disabilities
Group 1 (LD) Group 2 (AD/HD) Group 3 (LD & AD/HD)

Variables N Mean S.D. N Mean S.D. N Mean S.D.
KAIT
50 101.76 11.68 28 104.64 10.76 37 102.32 11.62
(Composite) 1
KAIT
(Crystallized) 50 100.08 9.80 28 104.25 9.50 37 101.37 12.07
KAIT (Fluid) 50 102.90 13.77 28 103.89 11.58 37 102.46 13.05

WAIS-II1
47 105.32 9.83 32 115.53 12.12 22 110.23 8.02
(Full Scale) 2
WAIS-II1
47 106.62 12.33 32 116.69 14.15 22 111.09 10.77
(Verbal)
WAIS-II1
47 103.13 11.40 32 111.09 8.43 22 107.82 7.42
(Performance)
PPVT-IIC 86 104.26 10.64 64 111.68 11.05 44 106.99 11.24
WJ-R
(Analysis- 80 103.24 11.60 57 108.70 12.51 51 107.26 12.46
Synthesis)
WJ-R
(Memory for 89 102.44 16.81 64 106.92 15.18 54 100.20 12.53
Sentences)
Two types of descriptive information have been provided to best represent the types
of learning problems evident across the clinical populations. Table 1 provides
selected clinical information for the students in each of the groups with disabilities.
The decision as to whether an individual was deficient in the areas listed in Table 1
was the result of diagnoses established at the time of staffing; therefore, percentages
reflect clinical decisions and not single discrepant scores. Table 2 provides the
reader with standardized data from select cognitive, language, and achievement data
gathered at the time of the participants' evaluation at the GA-RCLD.
9.1.2 Population without Documented Disabilities

Participants in Group 4 were recruited during the 1999-2000 academic year from
departments across UGA as part of a larger research project related to cognitive and
language profiles of college students with and without LD. Each participant was
administered a 12-hour battery of cognitive, linguistic, and achievement measures.
The UGA Phonological/Orthographic Battery was administered as part of this
testing. Group 4 consisted of 100 English-speaking university students (73 females;
27 males). The chronological mean age of the group was 22 (standard deviation
4.20). All students spoke English as a first language, had no known neurological
impairment, had received no special education services (with the exception of gifted
education programs), and were enrolled in undergraduate or graduate college
courses at UGA during the 1999/2000 academic year. Their Scholastic Achievement
Test (SAT) mean score was 1183 (Verbal =600; Quantitative =583).
9.2 Materials
University of Georgia Phonological/Orthographic Battery: This battery consists of

several tasks used by established researchers to assess phonological and
orthographic processing. Permission was obtained from each of the researchers who
developed the measures. The battery consists of six tasks assessing phonological
processing and five tasks assessing orthographic processing. A brief description of
each task will follow, accompanied by the number of items on that task and means
and standard deviations for the group of college students without disabilities (n =
100).
Segmenting by Syllables (Johnson & Blalock, 1987) requires the examinee to break
familiar words into their constituent syllables (e.g., an acceptable response for
"interview" would be in-ter-view). Stimulus items are provided orally. The subtest
begins with two training items and includes 12 items; among the adults without
disabilities, the mean was 12 and the standard deviation was .77.
Number of Syllables (Johnson & Blalock, 1987) presents the examinee with
(spoken) familiar words and asks himlher to determine the number of syllables in
each (e.g., "interview" contains three syllables). Again, the test begins with two
training items and includes 12 items. The mean on this subtest was 12 and the
standard deviation was .30.
Segmenting by Sounds (Johnson & Blalock, 1987) presents the examinee with
(spoken) familiar words and asks himlher to produce the sequence of individual
sounds that makes up each word (e.g., an acceptable response for "chat" would be
[ch]-[a]-[t]). Three training items precede the ten scored sub test items (mean = 8;
standard deviation = 1.81).
General Rhyming (Johnson & Blalock, 1987) measures the concept of rhyming (as
well as word-retrieval) by asking the examinee to produce "three other words" that
rhyme with 4 stimulus items. The examiner says each stimulus word, then reads a
sentence with the word in it, then says the word again (e.g., "Moose. 'On our visit to
Maine we saw a moose.' Tell me three other words that rhyme with 'moose'."). All
stimulus words consist of one syllable and have many potential rhymes. Training
items illustrate the concepts of rhyming and non-rhyming. The mean for the task
was 12 (of a possible 12) with a standard deviation of .38.
Phonemic Localization (Vellutino & Scanlon, 1988) presents the examinee with 10
pairs of spoken one-syllable words/pseudowords. Each pair differs by one phoneme
either initially, medially, or finally. The examinee's task is to state where the two
words in each pair sound different (e.g., [flig] and [slig] sound different at the
beginning). Two training items orient the examinee to the goal of the task and
discourage the use of an orthographic strategy. The mean for the task was 10 with a
standard deviation of .56.
Phonological Segmentation (Berninger & Abbott, 1994) is a taped subtest requiring
the examinee to (1) repeat pseudowords (e.g., "Say [gant]"), and (2) say them again
after "taking out" one or more sounds (e.g., "Now say it again, but don't say [t]").
Pseudowords vary form one to four syllables; segments to be deleted are varied to
control for location (e.g., word-initial sounds/syllables) and type (e.g., segments
breaking syllable boundaries). One training item is provided prior to administration
of the 24 items. The complex task taps into skills including (but not limited to)
auditory discrimination, phonemic awareness, and working memory. The mean for
the subtest was 19 with a standard deviation of 3.34.
Orthographic Expressive Coding (Berninger & Abbott, 1994) presents the examinee
with a series of pseudowords (computer-printed on index cards). Each of the 18
items is shown for one second, after which the examinee is asked to write either (a)
part of the word he/she saw (e.g., the third and fourth letters), or (b) the word in its
entirety. The task taps into automaticity of decoding, orthographic awareness, and
working memory. The mean for the task was 17 with a standard deviation of .83.
Orthographic Choice (Stanovich, West & Cunningham, 1991) is a psychomotor
scanning task consisting of 25 items. The examinee is presented with a sheet of
paper featuring a series of stimulus questions (e.g., "Which can be cooked?") and a
two homophonic answer options (e.g., meat/meet). The examinee's task is to circle
the correct answer to each item as quickly as possible (time required to complete
was recorded during the current study). The subtest begins with five practice items.
The mean among non disabled students was 25 with a standard deviation of .27.
Homophone/Pseudohomophone Choice (Olson, Forsberg, & Wise, 1994), another
psychomotor speed measure, presents the examinee with a sheet featuring a series of
spelling choice tasks. Each of the 78 items consists of two orthographically plausible
spellings for a common word, only one of which is conventionally accepted (e.g.,
fit/phit). The examinee's task is to circle the "correct" answer to each item as
quickly as possible. The sub test, which has a 3-minute time limit, features eight
practice items. The mean for the subtest was 77 with a standard deviation of .78.
The Colorado Perceptual Speed Test (DeFries & Baker, 1983) presents the
examinee with pages featuring rows of clusters of letters and (sometimes) numbers
(e.g., zxc6: zcx6 zxc9 zxc6 z6cx). In each row, one of the four items to the right is
identical to the stimulus on the left. The examinee's task is to circle the matching
cluster for each row as quickly as possible. Three one-minute, 3~-row trials are
administered. In trials I and II the clusters bear no resemblance to pronounceable
words (e.g., zxc6); in trial III, however, most clusters can be decoded as single-
syllable pseudowords (e.g., falp). The sub test begins with several examples and
sample items. On trial I (30 items), the mean was 23 and the standard deviation was
3.46; on trial II (30 items) the mean was 19 and the standard deviation was 3.35; and
on trial III (30 items), the mean was 29 and the standard deviation was 1.76.
Orthographic Fluency (Coleman & Nielsen, 2000) presents the examinee with
printed groups of consonants (e.g., spr) and asks him/her to create real words by
adding vowels to those consonants (e.g., "spar" or "super"). Six 40-second trials are
administered. The subtest begins with an example item. The task taps into
vocabulary, orthographic awareness, knowledge of spelling patterns, and fluency of
orthographic-based word- retrieval. The mean for the subtest was 21 with a standard
deviation of 4.58.
Verbal Fluency (Johnson & Blalock, 1987) provides the examinee with a category
(e.g .. , "things to eat") and asks himlher to name as many examples or
representatives of that category (e.g., spaghetti; fruit) as he/she can. Four 20-second
trials are administered; the categories are, in order of presentation, Things to Eat
(mean = 12.65; standard deviation = 2.87), Animals (mean = 12.20; standard
deviation = 2.86), Things to Wear (mean = 12.65; standard deviation = 2 .. 57), and
Things to Ride (mean =7.90; standard deviation =2.43).
Standardized Decoding and Encoding Measures: Two measures of decoding (real
and nonsense words) and a dictated spelling task were administered to all
participants to measure decoding and encoding competency. The Reading and
Spelling subtests of the Wide Range Achievement Test - 3 (WRAT-III; Wilkinson,
1983) were chosen as the single real word decoding and encoding measures. The
reliability of the WRAT-III as measured by the coefficient alpha is .90 for the
reading subtest and .89 for the spelling subtest. Content validity and construct
validity of the WRAT-III, according to the Rasch statistic of item separation, are both
1.00. The Word Attack subtest of the Woodcock-Johnson Tests of Achievement-
Revised (WJ-R; Woodcock & Johnson, 1989) was chosen as the measure of
nonword reading. The mean reliability coefficient of the subtest is reported as .91.
Reading Comprehension: The Comprehension subtest of the Nelson-Denny Reading
Test (NDRT; Brown, Fishco, & Hanna, 1993) was chosen to measure the reading
comprehension skills across all groups. The sub test consists of seven passages and
38 multiple-choice questions. It has a time limit of 20 minutes, during which the
examinee works independently. Reliability for the Comprehension sub test is .81.
Listening Comprehension: The Listening Comprehension scale of the Oral and
Written Language Scales (OWLS; Carrow-Woolfolk, 1996) was used to measure
listening comprehension of connected text. The scale, which takes approximately 15
to 20 minutes to administer, includes two examples and 111 multiple-choice items
(the examinee must select, given 4 choices, the picture that best represents the
spoken sentence(s) of the examiner). Reliability coefficients for internal consistency
are reported at .84 and test-retest reliability coefficients range from .73 to .80.
10. RESULTS
A series of ANOVAs was used to evaluate differential performance among the four
groups on the measures of interest. Because of the large number of statistical tests
conducted, a Bonferroni-adjusted significance level of p :s: .001 was used for the
ANOVAs. Several regression analyses were then examined to determine which
tasks were predictive of performance on selected measures of reading and spelling
abilities. The goal of these analyses was to determine the most appropriate set of
predictors of phonological, orthographic, and word fluency abilities across the
different groups of individuals to use in subsequent analyses. Therefore, predictors
were initially selected based on the strength of their correlations with reading and
spelling scores and were retained only if regression analysis showed that they were
significantly related to one of the five independent variables included in the study:
WRAT-III (Spelling); WRAT-III (Reading); WJ-R (Word Attack); OWLS (Listening
Comprehension); and Nelson-Denny (Comprehension).
10.1 Differences Between Groups
The results of this research support past research indicating an arrest in the
phonological and orthographic processing abilities of English-speaking university
students with LD as compared to their nondisabled peers, as well as to university
students demonstrating AD/HD. Raw score means, standard deviations, and standard
errors for all measures (as well as the number of participants from each group
administered each measure) are provided in Table 3. Table 4 provides a breakdown
of between-group differences by task (in cases of missing data points, within-group
mean substitutions were made).
Table 3. Normative Data for Selected Measures
Diagnosis N Mean Std. Deviation Std. Error of the

Mean
WRAT-III (Reading) Group 1 89 45.157 5.121 0.543
Group 2 64 48.719 2.786 0.348
Group 3 54 45.833 4.454 0.606
Group 4 100 52.540 2.687 0.269
WRAT-III (Spelling) Group 1 90 39.778 5.888 0.621
Group 2 64 43.266 3.596 0.449
Group 3 54 38.778 4.645 0.632
Group 4 100 46.600 2.892 0.289
WJ-R (Word Attack) Group 1 93 21.849 4.950 0.513
Group 2 64 25.734 2.577 0.322
Group 3 50 21.160 4.692 0.664
Group 4 96 27.625 2.038 0.208
Segmenting by Group 1 93 10.183 1.989 0.206
Syllables Group 2 64 10.906 1.866 0.233
Group 3 54 10.037 2.339 0.318
Group 4 100 11.690 0.775 0.077
Number of Syllables Group 1 52 11.173 1.309 0.182
in Words Group 2 40 11.650 1.167 0.184
Group 3 37 10.865 1.888 0.310
Group 4 100 11.920 0.307 0.031
Segmenting by Group 1 84 7.226 2.214 0.242
Sounds Group 2 62 7.629 1.918 0.244
Group 3 52 7.250 2.611 0.362
Group 4 100 7.940 1.819 0.182
General Rhyming Group 1 91 10.901 1.915 0.201
Group 2 63 11.571 0.911 0.115
Group 3 52 11.096 1.287 0.179
Table 3. Normative Data for Selected Measures (continued)

Mean
Phonemic Group 1 92 8.826 1.642 0.171

Localization Group 2 64 9.531 0.796 0.100
Group 3 53 9.019 1.201 0.165
Group 4 100 9.780 0.561 0.056
Phonological Group 1 89 15.427 3.879 0.411
Segmentation Group 2 63 18.238 2.933 0.370
Group 3 54 15.463 3.859 0.525
Group 4 100 19.370 3.341 0.334
Orthographic Group 1 89 15.933 2.378 0.252
Expressive Coding Group 2 64 16.922 1.301 0.163
Group 3 50 15.980 1.767 0.250
Group 4 100 17.100 0.835 0.083
Orthographic Choice Group 1 90 24.478 1.073 0.113
Group 2 62 24.742 0.541 0.069
Group 3 54 24.426 1.057 0.144
Group 4 100 24.920 0.273 0.027
Orthographic Fluency Group 1 47 15.681 5.288 0.771
Group 2 32 19.125 4.640 0.820
Group 3 29 15.448 4.864 0.866
Group 4 99 20.869 4.575 0.460
Homophone/Pseudo- Group 1 86 73.186 7.065 0.762
homophone Choice Group 2 61 76.164 3.034 0.388
Group 3 52 73.692 5.090 0.706
Group 4 100 77.560 0.756 0.076
Colorado part 1 Group 1 89 17.775 4.142 0.439
Group 2 64 19.766 3.398 0.425
Group 3 53 16.887 4.543 0.624
Group 4 100 23.260 3.460 0.346
Group 2 64 16.328 3.227 0.403
Group 3 53 14.151 4.054 0.557
Group 4 100 19.300 3.347 0.335
Group 2 64 26.453 3.572 0.447
Group 3 53 23.717 5.496 0.755
Group 4 100 28.940 1.763 0.176
Verbal Fluency: Group 1 91 11.264 3.231 0.339
Things to Eat Group 2 64 12.344 2.923 0.365
Group 3 53 11.528 3.528 0.485
Group 4 20 12.650 2.870 0.642
Animals Group 2 64 11.000 2.971 0.371
Group 3 53 10.811 2.624 0.360
Group 4 20 12.200 2.858 0.639
Things to Wear Group 2 64 12.750 2.755 0.344
Group 3 53 11.925 2.615 0.359
Group 4 20 12.750 2.573 0.575
Table 3. Normative Data for Selected Measures (continued)

Mean
Things to Ride Group 2 64 8.813 2.383 0.298

Group 3 53 8.396 2.641 0.363
Group 4 20 7.900 2.426 0.542
OWLS Listening Group 1 89 96.652 13.091 1.388
Comprehension Group 2 52 101.065 13.404 1.702
Group 3 63 97.755 14.223 1.954
Group 4 100 107.200 11.537 1.154
Nelson Denny Group 1 89 44.562 14.591 1.547
Comprehension Group 2 64 53.375 12.502 1.563
Group 3 54 43.481 16.452 2.239
Group 4 99 66.141 7.121 0.716
Table 4. Patterns of Significant Differences4 * Significance level p .001
Group 1 Group 1 Group 1 Group 2 Group 2 Group 3

vs. vs. vs. vs. vs. vs.
Group 2 Group 3 Group 4 Group 3 Group 4 Group 4
WRAT-3 Reading * * * * *
WRAT-3 Spelling * * * * *
WJ-R Word Attack * * * * *
Segmenting by Sounds
General Rhyming * * *
PhonemicLocalization * * *
Phonological * * *
Segmentation
Orthographic * * * *
Expressive Coding
Orthographic Choice * *
Orthographic Fluency * *
Homophone/Pseudo- * * * * *
homophone Choice
Things to Eat
Animals
Things to Wear *
Things to Ride
Colorado Part 1 * * * * *
Colorado Part 2 * * * *
Colorado Part 3 * * * * *
Nelson Denny Compo * * * * *
Orthographic Fluency * *
OWLS Listening * *
Comprehension
No significant differences were identified between Group 1 (LD) and Group 3 (LD
+ AD/HD). Both Group 1 and Group 3 performed significantly below their
normally-achieving peers across all measures of phonological and orthographic
processing, as well as on measures of reading-decoding, spelling, listening
comprehension, and reading comprehension. Groups 1 and 3 performed significantly

differently than Group 2 (AD/HD) across all measures of phonological and
orthographic processing, with the exception of two measures from the UGA
Phonological and Orthographic battery (HomophonelPseudohomophone Choice and
Colorado Part 2), both measures of psychomotor speed and attention. Group 2 was
significantly different from Group 4 only on HomophonelPseudohomophone Choice
and all three of the Colorado subscales. Significant differences were noted between
Group 2 and Group 4 on the reading-decoding, spelling, and reading comprehension
tasks, but not on the listening comprehension measure. Therefore, Groups 1 and 3
appear to be performing poorly compared to their nondisabled peers due to different
cognitive and linguistic processing deficits than those exhibited by Group 2. In
addition, Groups 1 and 3 demonstrated significantly lower academic and language
abilities than their peers with AD/HD.
10.2 Regression Analyses
All regressions were conducted on the total sample. Stepwise regression was used to
investigate different models and to determine whether the inclusion of additional
variables beyond those with the strongest correlations to the dependent variable
would account for substantially more variance.
Table 5. Regression Results for WJ-R (Word Attack) Models 1-3
Dependent Variable: WJ-R Word Attack

R R2 Adj. R2 F(dt) AR~ P
Modell: Phonology/ .753 .567 .557 56.434

Orthographic Battery (8,345)
Model 2: .765 .585 .575 53.974 .019 .000
Phonology/Orthographic (9,344)
Battery and Nelson Denny
(Comprehension)
Model 3: Phono./Orthographic .791 .626 .609 37.666 .040 .000
Battery, Nelson Denny (15,338)
(Comp.), & Interactions
WJ-R Word Attack (Nonwords): Three models were investigated. The first
regression model, Model 1, included measures selected from the UGA Phonological
and Orthographic battery as predictors of WJ-R (Word Attack). As can be seen in
Table 5, the overall regression and all the predictors were significant and accounted
for approximately 57% of the variance in WJ-R (Word Attack). General Rhyming
and Phonological Segmentation accounted for the greatest amount of variance.
Regression coefficients are reported in Table 6. Model 2 added the Nelson-Denny
(Comprehension) to those variables that were used in Model 1. No suppressor
variables were identified. The Nelson-Denny (Comprehension) predicts an
additional 1.8% of the variance in WJ-R (Word Attack) , which was a statistically
significant increase in the variance accounted for. Model 3 included the Nelson-
Denny (Comprehension), the UGA Phonological and Orthographic variables, and
344 GREGG, COLEMAN, TENNETf, DAVIS, NIELSEN, KNIGHT, & HOY
several significant two-way interactions. The change in R2 for this model was
relatively small (L\R2 = .040), indicating that the interactions accounted for about 4%
of the variance in WJ-R (Word Attack); however, this additional explained variance
was statistically and theoretically significant. Particular attention should be paid to
the interpretation and meaning of these interactions because of the theoretical
importance of interaction effects, as well as the practical implications they suggest.
The meaning of such interactions will be discussed in greater detail in a subsequent
section. Other possible two-way interactions and higher-order interactions were
investigated, but were not statistically significant. Therefore, Model 3 is the most
appropriate model examined for predicting the WJ-R (Word Attack) and, as shown
in Table 5, accounted for approximately 63% of the variance.
Table 6. Regression Weights for WJ-R (Word Attack) Model 3
b Std. 13 p
Error
Intercept -12.807 2.905 -4.409 .000
Segmenting by Syllables .234 .091 .099 2.577 .010
Number of Syllables in Words .529 .170 .115 3.101 .002
Segmenting by Sounds .203 .081 .096 2.509 .013
General Rhyming .785 .136 .225 5.789 .000
Phonological Segmentation .302 .047 .271 6.392 .000
Homophone/Pseudohomophone Choice .136 .036 .143 3.777 .000
Things to Wear -.212 .064 -.120 -3.316 .001
Orthographic Fluency .129 .040 .123 3.191 .002
Nelson-Denny (Comprehension) .05216 .012 .181 4.423 .000
Segmenting by Syllables x -0275 .006 -.173 -4.737 .000
Homophone/Pseudo-homophone Choice
Segmenting by Syllables x Nelson- .01048 .005 .076 2.169 .031
Denny (Comp.)
Number of Syllables in Words x .01077 .004 .086 2.416 .016
Homophone/Pseudo-homophone Choice
Phonological Segmentation x Things to .01316 .007 .069 1.974 .049
Wear
Homophone/Pseudo-homophone Choice .009298 .004 .092 2.562 .011
x Orthographic Fluency
Orthographic Fluency x Nelson-Denny -.00335 .001 -.100 -2.831 .005
(Comprehension)
WRAT-III Reading (Real Words): Three models were investigated, again using a
stepwise regression approach, to explore changes in explained variance due to the
inclusion of additional predictor variables. The first regression model, Modell,
included the three parts of the Colorado measure as predictors of WRAT-III
(Reading). As can be seen in Table 7, the overall regression for Model 1 was
significant and accounted for approximately 21 % of the variance in WRAT-III
(Reading). Parts 1 and 3 of the Colorado measure resulted in statistically significant
regression coefficients (p = .009 and p = .021, respectively). Part 2 of the Colorado
measure was not significantly related to WRAT-III (Reading). One possible reason
for this result is that Part 2 (which features clusters such as x6pd) lends itself least to
a general word-decoding strategy.
Model 2 included all the UGA Phonological and Orthographic variables. When
the Colorado measures were dropped· from the model, Orthographic Expressive
Coding and Verbal Fluency-Things to Eat became insignificant predictors (p = .154
and p = .084, respectively) of WRAT-III (Reading). This result indicates that the
Colorado was acting to suppress some of the error variance associated with
Orthographic Expressive Coding and Things to Eat and should therefore be retained
in the model. As can be seen in Table 7, the change in R2 that resulted from the
addition of the phonological and orthographic variables was substantial (Model 2 R2
= .502, L\R2 = .295) and resulted in a large improvement in the prediction of WRAT-
III (Reading ).
Table 7. Regression Results for WRAT-lll (Reading) Models 1-3
Dependent Variable: WRAT-III Reading

R R2 Adj. F(df) ~R2 P
R
Modell: Colorado .455 .207 .200 30.446 .000

(3,350)
Model 2: Colorado & UGA .708 .502 .486 25.301 .295 .000
Phonology/Orthography (8,342)
Model 3: Colorado, UGA .748 .559 .537 7.260 .057 .000
Phonology/Orthography & (6,336)
Interactions
Model 3 included all the UGA Phonological and Orthographic battery variable
and six significant two-way interactions. Regression coefficients for this model are
reported in Table 8. The change in R2 from Model 2 was relatively small (L\R2 =
.057), indicating that the interactions account for less than 6% of the variance in
WRAT-III (Reading), but the results are statistically and theoretically significant.
Higher-order interactions were investigated, but were not statistically significant.
Model 3 accounted for approximately 56% of the variance in WRAT-III (Reading)
(R2 = .559) and is the model that will be used in subsequent analyses. It is important
to remember that the Colorado tests should be retained in the regression equation
because they enhanced the predictive capabilities of Orthographic Expressive
Coding and Things to Eat.
An alternate set of predictors of WRAT-III (Reading) was examined in a separate
set of regression analyses. These alternate predictors cannot be included in the
models previously discussed because of multicollinearity problems. These analyses
resulted in Models 4 and 5, which included Phonological Segmentation, Nelson-
Denny (Comprehension), WJ-R (Word Attack), and WRAT-III (Spelling) as
predictors of WRAT-III (Reading). The only difference between Model 4 and Model
5 was that ModelS included the interaction between WJ-R (Word Attack) and
WRAT-III (Spelling). Other possible interactions were not significant. Model 4
accounted for approximately 68% of the variance in WRAT-III (Reading) (R2 = .676,
P < .001). As can be seen in Table 8, inclusion of the interaction between WJ-R
(Word Attack) and WRAT-//I (Spelling) (ModelS) accounted for an additional 1%
of the variance in WRAT-//I (Reading).
Table 8. Regression Weights for WRAT-Il/ (Reading) Model 3
b Std. ~ P
Error
Intercept -3.269 6.144 -.532 .595
Colorado Part I .06943 .078 .066 .888 .375
Colorado Part II .02874 .087 .025 .329 .743
Colorado Part III .126 .075 .124 1.887 .083
Number of Syllables .570 .200 .115 2.849 .005
Segmenting by Sounds .207 .092 .091 2.251 .025
General Rhyming .661 .162 .176 4.076 .000
Orthographic Expressive Coding -.322 .135 -.115 -2.383 .018
Orthographic Choice .560 .262 .090 2.142 .033
Homophone/Pseudohomophone Choice .195 .054 .190 3.601 .000
Colorado Part I x Colorado Part II .02255 .008 .123 2.692 .007
Colorado Part I x Segmenting by Sounds .0512 .015 -.138 -3.338 .001
Colorado Part I x Orthographic Expressive .03795 .014 .118 2.763 .006
Coding
Segmenting by Sounds x Phonological .04295 .017 .106 2.559 .011
Segmentation
Segmenting by Sounds x Orthographic .0569 .029 -.087 -1.982 .048
Expressive Coding
General Rhyming x Homophone/Pseudo- .0275 .014 -.080 -1.970 .050
homophone Choice
Table 9 presents the regression weights for ModelS. As can be seen from the
tables, Models 4 and 5 accounted for substantially more of the variance in WRAT-
III (Reading) than did Model 3 and were, therefore, better predictive models. Both
Model 3 and ModelS deserve additional research attention because the differences
in these models may lead to an enhanced theoretical understanding of the
relationships among the variables investigated in these analyses. Regression
coefficients for ModelS are presented in Table 10.
Table 9. Regression Results for WRAT-IlI (Reading) Models 4 and 5
Dependent Variable: WRAT-III Reading

R R2 Adj. F(d!)
R2
Model 4: Alternate Predictors .822 .676 .672 182.133
(4,349)
ModelS: Alternate Predictors .828 .686 .682 11.209 .010 .001
and Interaction (1,348)
WRAT-III Spelling (Real Words): As before, three models were investigated using
stepwise regression procedures to explore the changes in the variance accounted for
by including additional predictors of Spelling scores. The first regression model,
Model 1, included Verbal Fluency-Things to Wear and Number of Syllables in
Words as predictors of WRAT-III (Spelling). As shown in Table 11, the overall
regression and both predictors were significant and accounted for almost 29% of the
variance in WRAT-III (Spelling) (R2 =.286).
Table 10. Regression Weights for WRAT-III (Reading) ModelS
b Std. ~ P
Error
Intercept 20.524 1.208 16.987 .000
ND (Comprehension) .06402 .011 .207 5.665 .000
WJ-R (Word Attack) .318 .051 .296 6.279 .000
WRAT-III (Spelling) .336 .043 .364 7.725 .000
WJ-R (Word Attack) x WRAT- -.0132 .004 -.101 -3.348 .001
III (Spelling)
Table 11. Regression Resultsfor WRAT-I/l (Spelling) Models 1-3
Dependent Variable: WRAT-III Spelling

R R2 Adj. F(dt) ~R2 P
R
Modell: Things to Wear & .534 .286 .269 17.192

Number of Syllables in Words (2,86)
Model 2: Additional Selected .870 .758 .727 18.999 .472 .000
Predictors (8,78)
Model 3: Additional Selected .880 .775 .742 5.746 .017 .019
Predictors and Interaction (1,77)
Model 2 included the Nelson-Denny (Comprehension) and the UGA

Phonological and Orthographic variables (identified in Table 12) as additional
predictors. When either Things to Wear or Number of Syllables in Words was
dropped from the model, several of the other variables became insignificant
predictors. This indicated that these two variables were suppressing some of the
error variance associated with other predictor variables and should be retained in the
model because such suppression has the effect of enhancing the predictive power of
other variables included in the regression. As indicated in Table 11, Model 2
accounted for substantially more of the variance in WRAT-III (Spelling) than did
Modell (Model 2 R2 = .758, ~R2 = .472).
Model 3 included the suppressor variables, the UGA Phonological and
Orthographic battery variables, and a significant two-way interaction between
Things to Wear and Segmenting by Syllables. This interaction accounted for less
than 2% of the variance in WRAT-III (Spelling), as shown in Table 10, but the
change in R2 from Model 2 to Model 3 is statistically significant (~R2 = .017, P =
.019). Overall, Model 3 accounted for 77.5% of the variance in WRAT-III (Spelling).
Regression coefficients for Model 3 are reported in Table 12. As before, other
possible interactions were investigated, but were not significant.
Table 12. Regression Weights for WRAT-lI/ (Spelling) Model 3
b Std. Ii p
Error
Intercept 23.069 11.158 2.067 .042
Things to Wear .01865 .127 .010 .147 .884
Number of Syllables in Words .317 .263 .083 1.208 .231
Segmenting by Syllables .325 .158 .142 2.061 .043
General Rhyming 1.017 .232 .327 4.391 .000
Orthographic Choice -2.045 .590 -.281 -3.466 .001
HomophonelPseudohomophone .559 .094 .447 5.954 .000
Choice
Animals -.296 .110 -.176 -2.681 .009
Nelson-Denny (Comprehension) .07250 .022 .218 3.300 .001
Things to Wear x Segmenting by .113 .047 .147 2.397 .019
Syllables interaction
Listening Comprehension: A single regression model was applied to the OWLS

(Listening Comprehension). As can be seen in Tables 13 and 14, the overall
regression and all predictors were significant and accounted for 29% of the variance.
In subsequent analyses, other potential predictors, interaction effects, and possible
suppressors were found to be nonsignificant.
Table 13. Regression Results for OWLS (Listening Comprehension)
Dependent Variable: OWLS (Listening Comprehension)

R R2 Adj. R2 F(df) P
.539 .290 .280 28.463 (5,348) .000
Table 14. Regression Weights for OWLS (Listening Comprehension)
b Std. Error Ii t P
Intercept 40.531 6.733 6.020 .000
Segmenting by Syllables 1.077 .364 .148 2.955 .003
Things to Wear .846 .255 .156 3.320 .001
WRAT-3 (Reading) .365 .163 .128 2.240 .026
Reading Comprehension: A full regression model that included the entire UGA
Phonological and Orthographic battery, the OWLS (Listening Comprehension),
WRAT-III (Reading), and WRAT-III (Spelling) was examined. This model was
statistically significant (F =4.140, df =20, 66, P < .001) and accounted for a large
proportion of the variance in reading comprehension scores (R2 = .556). However,
due to the multicollinearity among the predictors, the OWLS (Listening
Comprehension) and WRAT-III (Reading) were the only statistically significant
SUBLEXICAL & LEXICAL PROCESSING OF YOUNG ADULTS WITH LDIAD/HD 349
predictors (t = 2.392, P = .020 and t = 2.938, P = 005, respectively) of Nelson-Denny

(Comprehension). Inspection of collinearity diagnostics, product-moment
correlations, and theoretical considerations suggested three other, more constrained
models that should also be examined.
Model 1 included selected UGA Phonological and Orthographic variables and
accounted for about 27% of the variance in the Nelson-Denny (Comprehension), as
shown in Table 15. Other variables from the UGA Phonological and Orthographic
Battery did not reach statistical significance and were therefore omitted from the
analysis. Model 2 included the predictors from Model 1 as well as the OWLS
(Listening Comprehension). This model accounted for an additional 9.5% of the
variance in reading comprehension. Model 3 added WRAT-III (Reading) to the
predictors included in Model 2. Model 3 accounted for an additional 8.8% of the
variance in the Nelson-Denny (Comprehension) above that explained by Model 2.
As can be seen in Table 15, Model 3 accounted for a total of 45.1 % of the variance
in the Nelson-Denny (Comprehension).
Table 15. Regression Results for Nelson-Denny (Comprehension) Models 1-3
Dependent Variable: NO Reading Comprehension

R R' Adj- F(dt) ~R2 P
R
Modell: Selected Phonol Orthographic .518 .269 .244 11.094 .000

Measures (5,151)
Model 2: Selected Phonol Orthographic .603 .364 .338 14.302 .095 .000
Measures & OWLS (Listening Comp.) (6,150)
Model 3: Selected Phonol Orthographic .672 .451 .426 17.516 .088 .000
Measures, OWLS (Listening Comp.) & (7,149)
WRAT-III (Rdg)
Table 16. Regression Weights for Nelson-Denny (Comprehension) Model 3
b Std. Error ~ t P
Intercept -93.662 28.143 -3.328 .001
Colorado Part 1 .709 .227 .214 3.117 .002
Phonological Segmentation -.694 .309 -.172 -2.250 .026
Things to Wear .737 .354 .138 2.084 .039
Number of Syllables in Words -1.521 .772 -.131 -1.969 .051
Orthographic Choice 2.541 1.151 .141 2.209 .029
OWLS (Listening Comp.) .325 .075 .296 4.331 .000
WRAT -III (Reading) 1.192 .245 .381 4.876 .000
Possible interactions were examined, but no significant interaction effects were

found among this set of predictors. Regression coefficients for Model 3 are
presented in Table 16. All regression coefficients were significant.
The inclusion of WRAT-III (Spelling) in an expanded model did not produce a
significant improvement in variance accounted for. WRAT-III (Spelling) was not a
significant predictor of the Nelson-Denny (Comprehension) when WRAT-III
(Reading) was included in the model. This was because of the strong collinearity
between these two tests (r = .756).
11. DISCUSSION
The results of this study support past research (Bruck, 1990, 1992; Pennington, Van
Orden, Smith, Green, & Haith, 1990) demonstrating that English-speaking
university students. with LD do not acquire adequate levels of phonological or
orthographic knowledge regardless of their age or reading level. This study
demonstrated that adults with AD/HD (and no LD), on the other hand, do appear to
acquire adequate phonological and orthographic processing skills. Their weaker
performance on academic tasks (decoding, encoding, and comprehension), as
compared to the nondisabled university population, is likely a consequence of other
processing deficits (e.g., attention, executive functioning). Further research is
needed to explore the underlying cognitive processing deficits impacting the
academic performance of adults with AD/HD.
The university students with LD and the group with LD + AD/HD (co-morbid)
performed substantially below expected levels across all the phonological and
orthographic tasks, as well as the decoding (nonsense and real), encoding (spelling
recall), listening comprehension, and reading comprehension measures (compared to
their non disabled peers). As with Bruck's (1990, 1992) findings, the results of this
research show not only that the adult popUlation with LD continues to demonstrate
deficits in phoneme awareness when compared to adults with AD/HD and
nondisabled adults, but also that these adults appear not to be using orthographic
information when making phonological judgments.
The importance of intelligence (IQ) to the performance of each of the groups
(with and without disabilities) in this research was not addressed. Therefore, the
results of this study could be confounded by between-group IQ differences. The
decision not to control for IQ was based on the results of past research. Stanovich
(1988) demonstrated that the IQ scores of children with dyslexia may be the
consequence of reading deficits but not the cause of underlying phonological
awareness deficits. In addition, he showed that IQ scores of the population with
dyslexia do not correlate with phonological measures. Therefore, even if the IQ
scores across the populations studied for the purpose of research differ, IQ
differences would be unrelated to performance on phonological and orthographic
measures. For further discussion of the relationship between IQ test scores and
reading competence, see the chapter by Jimenez Gonzalez (this volume).
11.1 Contribution of Phonological/ Orthographic Abilities to Decoding & Encoding
The Berninger & Abbott (1994) Phonological Segmentation task and the General
Rhyming task (Johnson & Blalock, 1987) used in this research contributed the most
variance to the reading of nonsense (Wl-R Word Attack) and real words (WRAT-III
Reading). Further exploration of the collinearity of these two tasks is needed.
Additionally, since the WRAT-III Reading subtest includes orthographically regular
and irregular words, it will be important in the future to investigate the variance of
phonological and orthographic abilities (including segmenting and rhyming skills)
across both categories of spellings. Many studies have demonstrated that

phonological awareness is highly correlated with reading development (e.g.,
Cunningham, 1990; Juel, 1988; Perfetti et aI., 1987; Vellutino & Scanlon, 1988;
Stahl & Murray, 1998). In particular, some researchers have stressed the strong
connection between rhyme awareness and early reading (e.g., Bradley & Bryant,
1978, 1983; Ellis & Large, 1987; Holligan & Johnston, 1988). According to
Goswami (1998) the strong correlation of rhyme to early reading acquisition may be
a function of the depth of the English spelling system.
Adams (1990) organized the various types of phonological tasks into five levels
of difficulty. His proposed levels of phonological awareness, from most primitive to
most difficult, include: (a) remembering familiar rhymes; (b) recognizing and
sorting patterns of rhyme and alliteration in words; (c) blending and syllable
splitting tasks; (d) segmenting phonemes; and (e) adding and deleting phonemes
(phonological segmentation tasks). It was observed in the current study that the
ability to rhyme, a very primitive phonological ability, continues to play a
significant role in the English-speaking adult population's ability to decode words.
Further investigation of the importance and role of rhyme is certainly needed.
According to Goswami (1998), it may be that " ... rhyme rather than rime is a more
important variable in explaining reading development in English" (p.52).
The General Rhyming and HomophonelPseudohomophone Choice tasks
contributed the greatest variance in WRAT-III (Spelling) performance in this study.
As with WRAT-III (Reading), WRAT-lII (Spelling) consists of both regular and
irregular words. Further investigation of the influence of phonological and
orthographic abilities on a more direct measure of spelling-sound correspondence is
needed. It does appear, however, that both phonological awareness (i.e., General
Rhyming) and orthographic awareness (i.e., HomophonelPseudohomophone
Choice) continue to playa significant role in recalling spellings of words among the
adult population. In the case of adults with LD, both the phonological and
orthographic systems appear to be significantly impaired, resulting in
underachievement on decoding and encoding tasks.
11.2 Contribution of Phonological/Orthographic Abilities to Listening and Reading

Comprehension
The relationship of listening comprehension to reading comprehension appears to

strengthen as readers develop reading competencies. Very little research, however,
has investigated the similarities of the cognitive and linguistic strategies adults
utilize across these two modalities. Danks and End (1987) identified five processes
used in both listening and reading comprehension: speech perception/print decoding,
lexical access, clause/sentence integration, discourse understanding and
comprehension monitoring. One purpose of the present research was simply to
investigate whether any variance from the phonological and orthographic tasks
would load onto a listening comprehension measure for an English-speaking
university population. The OWLS (Listening Comprehension) measures
understanding of sentences (rather than text comprehension), thereby tapping into
sub lexical and lexical processes more than text structure awareness. As one would
hypothesize, phonological awareness (i.e., Phonological Segmentation and
Segmenting by Syllables), as well as word fluency (Le., Things to Wear), explained

the greatest variance on this measure. Further research is needed to explore the
direct and indirect relationship of phonological awareness and word fluency to
listening comprehension at both the sentence and text level among the adult
population.
Studies comparing the reading and listening skills of college students have found
that listening comprehension accounts for 32% to 92% of the variance in reading
comprehension (Bell & Perfetti, 1994; Cunningham & Stanovich, 1990; Jackson &
McClelland, 1979; Palmer, MacLeod, Hunt, & Davidson, 1985). Higher correlations
are present in studies that do not include word recognition as a variable (e.g., Palmer
et aI., 1985). Therefore, it was of interest in this study to look at the contributions of
phonological, orthographic, decoding, encoding, and listening comprehension tasks
to a measure of reading comprehension. While the decoding and listening
comprehension measures explained the most variance in reading comprehension, the
relationship was not quite linear. Upon review of the interaction effects, it appears
that the phonological and orthographic processes contribute a moderate amount of
variance to the process of reading comprehension. Further exploration of these
variables is needed by researchers.
11.3 Instructional Implications
Several interaction effects were noted in the regression analyses that appear to have
direct implications for the intervention and accommodation of decoding and
encoding among adult readers with learning disabilities. These interactions indicate
the presence of moderating effects, where the strength or level of one variable
influences the magnitude or the influence of another variable on the dependent
variable.
In the case of WJ-R (Word Attack), a nonsense-word decoding task, six different
moderators were identified and reported in Table 6. Three of the six significant
interactions involved the Homophone/Pseudohomophone Choice task. In all three
interactions, higher performance on the Homophone/Pseudohomophone Choice task
was associated with weaker effects of the other interacting variables (Segmenting by
Syllables, Number of Syllables in Words, and Orthographic Fluency), and lower
performance on Homophone/Pseudohomophone Choice was associated with
stronger effects of these other variables. This pattern suggests that the abilities
measured by these interacting variables can, to a limited extent, compensate for
weaknesses in the skills and abilities measured by Homophone/Pseudohomophone
Choice, while higher performance on the Homophone/Pseudohomophone Choice
task may reflect skills that one can use to compensate for weaknesses measured by
these other tasks. It could be that syllable boundaries, whether determined
orthographically or phonologically, facilitate performance on the
Homophone/Pseudohomophone Choice task, since most of the words on the task
were polysyllabic. While lexical identity of a homophone is largely determined
orthographically, it appears that syllabication is a related and contributing skill for
word identification. (See the chapter by Jimenez Gonzalez in this volume for
discussion of the importance of syllable boundaries in transparent and opaque
languages.) Therefore, the teaching of word patterns (e.g., syllable juncture and
derivational constancy) would directly address the skills underlying the

Homophone/Pseudohomophone task while at the same time indirectly addressing
the issue of syllable boundaries.
The Segmenting by Syllables task was also involved in an interaction with the
Nelson-Denny (Comprehension). In this case, the effect on WJ-R (Word Attack)
appeared to reach an asymptote fairly quickly for students who displayed low levels
of ability on the Nelson-Denny (Comprehension), but failed to exert a strong effect
on WJ-R (Word Attack) for those with relatively high levels of ability on the Nelson-
Denny (Comprehension). This finding, when considered in light of the interaction
between Segmenting by Syllables and Homophone/Pseudohomophone Choice,
suggests that the skills measured by the Segmenting by Syllables task should be
another high priority for professionals designing interventions for students with
demonstrated weaknesses in this area.
The Phonological Segmentation task interacted with the skills and abilities
measured by the Verbal Fluency-Things to Wear task in a surprising way. Average
scores on Phonological Segmentation, more than either high or low Phonological
Segmentation scores, were associated with higher performance on WJ-R (Word
Attack). This inverted-U pattern of task scores was most strongly pronounced in
adults with average scores on Things to Wear and much less evident in those adults
with either low or high scores on Things to Wear. This pattern of results is
extremely interesting and suggests a possible fruitful avenue for future empirical
research and theoretical development.
Six two-way interactions were identified for WRAT-III (Reading) and reported in
Table 8. Three of these interactions involved the Colorado Part I measure from the
UGA Phonological/Orthographic Battery. In all three interactions, higher scores on
the Colorado Part I task were associated with stronger effects of the interacting
variables (Colorado Part II, Segmenting by Sounds, and Orthographic Expressive
Coding) on the dependent variable, WRAT III (Reading). This suggests that
interventions designed to enhance the skills measured by Colorado Part I may have a
stronger influence on decoding proficiency than other interventions because these
particular skills enhance the effectiveness of the skills measured by the Colorado
Part II test, the Segmenting by Sounds task, and the Orthographic Expressive
Coding task. Certainly Ehri's (1998) full-alphabetic and consolidated (fluency)
stages of word recognition provide a framework in which to interpret some of the
findings from this study, as well as provide direction for intervention. It appears that
intervention focused on syllabication and affix rule knowledge, as well as strategies
to enhance decoding and encoding fluency, would prove beneficial to many adults
with LD demonstrating decoding and encoding underachievement.
The General Rhyming task from the UGA Phonology and Orthography battery
interacted with the Homophone/Pseudohomophone Choice task in predicting WRAT-
III (Reading) in a way suggesting that a strength in the abilities underlying one task
can be used to compensate for weakness in the other. Understanding the impact of
rhyme on the decoding and encoding abilities of adult readers needs further
exploration by researchers.
As reported in Table 10, WJ-R (Word Attack) interacted with WRAT-III
(Spelling) in predicting WRAT-III (Reading). Spelling abilities were more predictive
of WRAT-III (Reading) when WJ-R (Word Attack) scores were low than when WJ-R
(Word Attack) scores were high. This suggests that the abilities used to perform
WRAT-III (Spelling) can be used to compensate somewhat for deficits in the skills
measured by the WJ-R (Word Attack) test. Further empirical and theoretical research
is needed to fully understand the nature of this relationship.
The identification of interaction effects involved in regression analyses for WJ-R
(Word Attack), WRAT-III (Reading), and the WRAT-III (Spelling) suggests that
moderator variables may be a common occurrence. Researchers need to be aware of
this trend and design future studies to include possible moderators and to test for
interaction effects among the variables included in those studies.
11.4 Design Limitation
The ANOVA or regression designs used in this study limited a true exploration of
the relationship of the cognitive and linguistic variables contributing to the different
reading and spelling constructs under investigation. It is suggested that future
research using structural equation models (SEM) would provide several advantages
over regression, ANOVA, and other statistical approaches. First, SEM focuses
directly on the analysis of theoretical constructs that are thought to underlie and
cause performance on observed variables. That is, the focus of SEM is on the latent
variable, not on the observed or manifest variable. Second, SEM explicitly models
measurement error and thus achieves a more accurate representation of the latent
variables underlying performance. Third, SEM allows for the inclusion of multiple
indictors of the same underlying latent construct. This permits the control of
measurement error and provides an enhanced understanding of the latent variable.
Finally, SEM allows the researcher to directly investigate the relationship among
multiple latent variables as well as those among observed test scores.
11.5 Conclusion
The results of this research support past findings and generate new directions for
researchers studying the reading and written language of adults with LD and
AD/HD. The populations with disabilities in this study were required to meet
rigorous eligibility criteria for disability categorization (Le., LD, AD/HD, and LD +
AD/HD). Therefore, the outcomes of this research are based on uniform and well-
controlled sets of group criteria. The answers to many of the questions raised here
will require careful control of the populations studied, sophisticated methodology (to
identify latent variables critical to reading and written expression), and use of
reliable and valid measures of cognitive and linguistic constructs.
SUBLEXICAL & LEXICAL PROCESSING OF YOUNG ADULTS WITH LD/ADIHD 355
12. AFFILIATIONS
Dr. Knoel Gregg

Learning Disabilities Center
Regents Center for Learning Disorders
331 Milledge Hall
University of Georgia
Athens, Georgia 30602
knoelgregg@aol.com
13. NOTES
1 Wechsler Adult Intelligence Test - III (Wechsler, 1997).

2 Peabody Picture Vocabulary Test - III (Dunn & Dunn, 1997).
3 Woodcock-Johnson Psycho-Educational Battery-Revised (Woodcock-Johnson, 1989).
4 Group 1= LD; Group 2 = AD/HD; Group 3 = LD & AD/HD; Group 4 = Nondisabled.
14. REFERENCES
Adams, M. 1. (1990). Beginning to read: Thinking and learning abut print. Cambridge, MA: MIT Press.
Andrews, S. (1982). Phonological recoding: Is the regularity effect consistent? Memory & Cognition, 10,
565-575.
Bell, L. C., & Perfetti, C. A. (1994). Reading skill: Some adult comparisons. Journal of Educational
Berninger, V. W. (1994). Reading and writing acquisition: A developmental neuropsychological
perspective. Madison, WI: Brown & Benchmark.
Berninger, V. W., & Abbott, R. D. (1994). Multiple orthographic and phonological codes in literacy
acquisition: An evolving research program. In V. W. Berninger (Ed.), The varieties of orthographic
knowledge, 1: Theoretical and developmental issues (pp. 222-319). Netherlands: Kluwer.
Booth,1. R., Perfetti, C. A., MacWhinney, 8., & Hunt, S. 8. (2000). The association of rapid temporal
perception with orthographic and phonological processing in children and adults with reading
impairment. Scientific Studies of Reading, 3, 2, 101-132.
Bradley, L., & Bryant, P. E. (1978). Difficulties in auditory organization as a possible cause of reading
Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning to read: A causal connection.
Nature, 310, 4419-421.
Brown, 1., Fishco, V., & Hanna, G. (1993). The Nelson-Denny Reading Test. Lombard, IL: The Riverside
Publishing Company.
Bruck, M. (1985). The adult functioning of children with specific learning disabilities. In I. Sigel (Ed.),
Advances in applied developmental psychology Vol. 1 (pp. 91-120). Norwood, NJ :Ablex.
Bruck, M. (1990). Word-recognition skills of adults with childhood diagnoses of dyslexia. Developmental
Psychology,26, 3,439-454.
Bruck, M. (1992). Persistence of dyslexics' phonological awareness deficits. Developmental Psychology,
28, 5, 874-886.
Bruck, M. (1993a). Word recognition and component phonological processing skills of adults with
childhood diagnosis of dyslexia. Developmental Review, 13, 258-268.
Bruck, M. (1993b). Component spelling skills of college students with childhood diagnoses of dyslexia.
Journal of Learning Disabilities, 16, 171-183.
Bruck, M., & Waters, G. (1990). An analysis of the component spelling skills of good readers-poor
spellers. Applied Psycholinguistics, 11, 425-437.
Byrne, B., & Ledez, 1. (1983). Phonological awareness in reading-disabled adults. Australian Journal of
Carr, T. H. & Posner, M. I. (1994). The impact of learning to read on the functional anatomy of language
processing. In 8.de Gelder & 1. Morais (Eds.), Language and literacy: Comparative Approaches
(pp.32 ). Cambridge, MA: MIT Press.
Carrow-Woolfolk, E. (1996) The Oral and Written Language Scales. Circle Pines, MN: American
Guidance Service.
Coleman, c., & Nielsen, K (2000). The Orthographic Fluency Measure, Unpublished paper, University
of Georgia, Athens, Georgia.
information processing (pp. 151-216). London: Academic Press.
Cossu, G., Shankweiler, D., Liberman, I. Y., Katz, L., & Tola, S. (1988). Awareness of phonological
segments and reading ability in Italian children. Applied Psycholinguistics, 9, 1-16.
Cunningham, A. E. (1990). Implicit vs. explicit instruction in phonemic awareness. Journal of
Experimental Child Psychology, 50, 429-444.
Cunningham, A. E., & Stanovich, K D. (1990). Assessing print exposure and orthographic processing
skill in children: A quick measure of reading experience. Journal of Educational Psychology, 82, 4,
733-740.
Cunningham, A. E., Stanovich, K E., & Wilson, M. R. (1990) Cognitive variation in adult college
students differing in reading ability. In T. H. Carr, & B. A. Levy (Eds.), Reading and its
development. Component skills approaches (pp. 129-159). San Diego, CA, US: Academic Press.
Danks,1. H., & End, L. 1. (1987). Processing strategies for reading and listening. In R. Horowitz & S. J.
Samuels (Eds.), Comprehending oral and written language (pp. 11-50). San Diego: Academic Press.
DeFries, 1. C., & Baker, L. A. (1983). Colorado Family Reading Study: Longitudinal Analyses. Annals of
Dyslexia, 33, 153-162.
Dunn, L. M., & Dunn, L. M. (1997). Peabody Picture Vocabulary Test-Third Edition, Circle Pines: MN:
American Guidance Company.
Ehri, L. C. (1998). Grapheme-phoneme knowledge is essential for learning to read words in English. In 1.
L. Metsala & L. C. Ehri (Eds.), Word recognition in beginning literacy (pp. 3-39). Mahwah, NJ:
Lawrence Erlbaum Associates.
Ellis, N. c., & Large, B. (1987). The development of reading: As you seek, so shall ye find. British
Journal of Psychology, 78, 1-28.
A review. Psychonomic Bulletin and Review, 2, 460-493.
Finucci,1. M., Gottfredson, L. S., & Childs, B. (1986). A follow-up study of dyslexic boys. Annals of
Dyslexia, 35, 117-136.
Foorman, B. R. (1994). Phonological and orthographic processing: Separate but equal. In V. W.
Berninger (Ed.), The varieties of orthographic knowledge: Theoretical and developmental issues
(pp. 321-357). Dordrecht, The Netherlands: Kluwer Academic Publishers.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K E.Patterson, J. C. Marshall, & M.
Coltheart (Eds.), Surface dyslexia (pp. 301-330). Hillsdale, NJ: Lawrence Erlbaum Associates.
G1ushko, R. (1979). The organization and activation of orthographic knowledge in reading aloud. Journal
of Experimental Psychology: Human Perception and Performance, 5, 674-691.
Gough, P., & Tunmer, W. (1986). Decoding, reading, and reading disability. Remedial and Special
Education,7, 6-10.
Goswami, U. (1998). The role of analogies in the development of word recognition. In. J. L. Metsala & L.
C. Ehri (Eds.), Word recognition in beginning literacy (pp. 41-63). Mahway, NJ: LEA.
Gregg, N., Heggoy, S., Stapleton, M., Jackson, R., & Morris, R. (1994). The Georgia model of eligibility
for postsecondary students. Learning Disabilities: A multi-disciplinary Journal, 29-36.
Holligan, c., & Johnston, R. S. (1988). The use of phonological information by good and poor readers in
memory and reading tasks. Memory and Cognition, 16, 522-532.
Jackson, M. D., & McClelland, 1. L. (1979). Processing determinants of reading speed. Journal of
Experimental Psychology: General, J08, 151-181.
Johnson, D. J., & Blalock, 1. (Eds.). (1987). Young adults with learning disabilities. Orlando, FL: Grune
& Stratton.
Juel, C. (1988). Learning to read and write: A longitudinal study of 54 children from first through fourth
grades. Journal of Educational Psychology, 80, 437-447.
Klicpera, c., & Gasteiger-Klicpera, B. (1993). Lesen und Schreiben: Entwicklung und Schwierigkeiten
[Reading and writing: Development and difficulties]. Bern: Verlag Hans Huber.
Knight, D. (2000). A cognitive and linguistic model of individual differences in the reading
comprehension of college students with and without learning disabilities. Unpublished doctoral
dissertation, University of Georgia, Athens, Georgia.
Lander!, K, Frith, U., & Wimmer, H. (1996). Intrusion of orthographic knowledge on phoneme
awareness: Strong in normal readers, weak in dyslexic readers. Applied Psycholinguistics, 17, 1-14.
Liberman, I. Y., Rubin, H., Duques, S., & Carlisle, 1. (1985). Linguistic abilities and spelling proficiency
in kindergarten and adult poor spellers. In D. Gray & J. Kavanagh (Eds.), Biobehavioral measures of
dyslexia. Parkton, MD: York Press.
Manis, F. (1999). Phonological and surface dyslexia: A two-year longitudinal study. Presented at the
Society for the Scientific Studies of Reading, Montreal, Quebec.
Olson, R. K, Forsberg, H., & Wise, B. (1994). Genes, environment, and the development of orthographic
skills. In V. W. Berninger, (Ed.), The varieties of orthographic knowledge,l: Theoretical and
developmental issues (pp. 27-71). Netherlands: K1uwer.
Palmer, 1., MacLeod, C. M., Hunt, E., & Davidson, 1. E. (1985). Information processing correlates of
reading. Journal of Memory and Language, 24, 59-88.
Pennington, B. F., Van Orden, G. c., Smith, S. D., Green, P. G., & Haith, N. M. (1990). Phonological
processing skills and deficits in adult dyslexics. Child Development, 61, 1753-1778.
Perfetti, C. A, Beck, L., Bell, L., & Hughes, C. (1987). Phonemic knowledge and learning to read are
reciprocal: A longitudinal study of first grade children. Merrill-Palmer Quarterly, 33, 283-319.
Perfetti, C. A, Goldman, S., & Hogaboam,T. (1979). Reading skill and the identification of words in
discourse context. Memory and Cognition, 77, 273-282.
Perfetti, C. A, & Roth, S. (1981). Some of the interactive processes in reading and their role in reading
skill. In A Lesgold & C.Perfetti (Eds.), Interactive processes in reading (pp. 269-297). Hillsdale, N1.
Erlbaum.
Pratt, A c., & Brady, S. (1988). Relation of phonological awareness to reading disability in children and
adults. Journal of Educational Psychology, 80, 319-323.
Rack,1. P., Hulme, c., Snowling, M. J., & Wightman, 1. (1994). The role of phonology in young children
learning to red words: The direct mapping hypothesis. Journal of Experimental Child Psychology,
57,42-71.
Rawson, M. (1968). Developmental language disability. Baltimore, MD: Johns Hopkins Press.
Roberts, R., & Mather, N. (1997). Orthographic dyslexia: The neglected SUbtype. Learning Disabilities
Research and Practice, 12, 4, 236-250.
Rumsey, M. H., B., Donohue, c., Nace, K, Maisog, M., & Andreason, P. (1997). Phonological and
orthographic components of word recognition: A PET -rCBF scan study., Brain, 120, 739-759.
Seidenberg, M. S. (1997). Language acquisition and sue: Learning and applying probabilistic constraints.
Science, 275, 1599-1603.
Seidenberg, M. S., & McClelland, J. L. (1989). A distributed, developmental model of word recognition
and naming. Psychological Review, 96, 447-452.
Seidenberg, M. S., Waters, G. S., Barnes, M. A., & Tanenhaus, M. K (1984). When does irregular
spelling or pronunciation influence word recognition? Journal of Verbal Learning and Behavior, 23,
383-404.
Share, D. L., & Stanovich, K E. (1995). Cognitive processes in early reading development:
Accommodating individual differences into a model of acquisition. Issues in Education, I, 1-57.
Stahl, S. A, & Murray, B. (1998). Issues involved in defining phonological awareness and its relation to
early reading. In J. L. Metsala and L. C. Ehri (Eds.), Word recognition in beginning literacy (pp. 65-
87). Mahway, NJ: LEA
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. & West, R. R. (1989). Exposure to print and orthographic processing. Reading Research
Quarterly, 24, 402-484.
Stanovich, K E., West, R. F., & Cunningham, A E. (1991). Beyond phonological processes: Print
exposure and orthographic processing. In S. A Brady & D. P. Shankweiler (Eds.), Phonological
processes in literacy: A tribute to Isabelle Y. Liberman (pp. 219-235). Hillsdale, NJ: Erlbaum.
Van Orden, G. c., Pennington, B. F., & Stone, G. O. (1990). Word identification in reading and the
promise of subsymbolic psycholinguistics. Psychological Review, 97, 488-522.
Vellutino, F. R., & Scanlon, D. M. (1988). Phonological coding, phonological awareness, and reading
ability: Evidence from a longitudinal and experimental study. In K. E. Stanovich (Ed.), Children's
reading and the development of phonological awareness (pp. 77-119). Detroit, MI: Wayne State
University Press.
Wagner, R. K, Torgesen, 1. K., & Rashotte, C. A (1994). Development of reading-related phonological
processing abilities: New evidence of bi-directional causality from a latent variable longitudinal
study. Developmental Psychology, 30, 73-87.
Waters, G., Seidenberg, M., & Bruck, M. (1984). Children and adults' use of spelling sound information
in three reading tasks. Memory and Cognition, 12, 293-305.
Wechsler, D. (1997). Wechsler Adult Intelligence Scale-Third Edition. San Antonio, TX: The
Psychological Corporation.
Wilkinson, G. S. (1983). Wide Range Achievement Test-third Edition. Wilmington, DE: Jastak
Associates.
Wimmer, H. (1993). Characteristics of developmental dyslexia in a regular writing system. Applied
Wimmer, H. & Goswami, U. (1994). The influence of orthographic consistency on reading development:
Word recognition in English and German children. Cognition, 51, 91-103.
Wolf, M., Pfeil, C., Lotz, R., & Biddle, K. (1994). Towards a more universal understanding of the
developmental dyslexias: The contribution of orthographic factors. In V. W. Berninger (Ed.), The
Varieties of Orthographic Knowledge I: Theoretical and Developmental Issues (pp. 137-171).
Dordrecht, The Netherlands: Kluwer Academic Publishers.
Woodcock, R. W., & Johnson, M. B. (1989). Woodcock-lohnson Psycho-Educational Battery-Revised.
Itasca, IL: Riverside.
T.KUJALA
THE MISMATCH NEGATIVITY AS AN INDEX OF

AUDITORY DYSFUNCTION IN DYSLEXIA
Abstract. It is widely accepted that various deficits may underlie dyslexia. However, in the majority of
the cases dyslexia is currently thought to primarily result from a dysfunction of the phonological or the
auditory system. In identifying phonological and auditory dysfunctions in dyslexia a brain response called
the mismatch negativity (MMN) might be a useful tool since it reflects sound-discrimination accuracy. It
is elicited without the subject's attention which makes it well-suited even for studying patients and
children that are not able or willing to perform tasks. The MMN might be used in the identification of
auditory-system deficits early in life. This is especially welcome in developmental impairments like
dyslexia since early definition would make it possible to start the remediation before major learning
delays occur.
1. INTRODUCTION
It has been estimated that 5-15 % of children have problems in learning to read
despite of normal intelligence, no obvious sensory deficits, and adequate educational
resources. These learning difficulties often cause delayed cognitive development,
and are reflected also in motivational problems and low self esteem affecting the
individual's whole life.
It is evident that there are various deficits that may underlie dyslexia. The debate
on the nature and origin of dyslexia and factors underlying it has been going on for
recent decades, resulting, however, in no agreement. In dyslexic individuals
problems have been reported in the visual and auditory domain (for a review see
Farmer & Klein, 1995) as well as in the somatosensory one (Grant, Zangaladze,
Thiagarajah, & Sathian, 1999). In the visual modality, it was shown, for example,
that dyslexic individuals have a slower rate of visual processing than non-dyslexic
individuals (Brannan & Williams, 1988; Di Lollo, Hanson, & McIntyre, 1983;).
Recently, modem brain-imaging methods were applied to study visual functions in
dyslexia. Measurements with magnetoencephalography indicated that passive
viewing of words, which normally elicits activity in the left inferior occipital region,
did not activate this area in dyslexic subjects or the activity was later than in control
subjects (Salmelin, Service, Kiesilii, Uutela, & Salonen, 1996). Functional magnetic
resonance imaging was used to study possible magnocellular-system deficit in
dyslexia (Demp, Boynton, & Heeger, 1998). It was found that moving sine-wave
gratings at a low mean luminance elicited lower visual-cortex activation in the
dyslexic than control subjects. The dyslexic subjects were also less accurate than the
control subjects in discriminating speed of moving visual stimuli. Moreover, there
was a correlation between the visual-cortex activation and speed discrimination:
subjects with stronger fMRI responses in visual areas had lower speed
discrimination thresholds.

and Reading Disability, 35~368.
360 T. KUJALA
It has been suggested that in most dyslexic individuals problems in phonological

processing are the major factor that underlies the reading difficulties (Liberman &
Shankweiler, 1985; Mody, Studdert-Kennedy, & Brady, 1998; Rayner, Pollatsek, &
Bilsky, 1995; Schulte-Kame, Deimel, Bartling, & Remschmidt, 1998; Studdert-
Kennedy & Mody, 1995). Phoneme segmentation and awareness tasks as well as
rhyming skills differentiate good and poor readers and serve as good predictors of
future reading abilities (for a review see Farmer & Klein, 1995). It has been
suggested that dyslexic individuals might actually have a more general auditory
dysfunction underlying their phonological difficulties (Baldeweg, Richardson,
Watkins, Foale, & Gruzelier, 1999; Farmer & Klein, 1995; Hari & KiesiHi, 1996;.
McGivern, Berka, Languis, & Chapman, 1991; Reed, 1989; Talla11980; Wright et
aI., 1997). According to this proposal, dyslexic individuals have problems in
discriminating rapid temporal changes that are present in speech, which affects
correct speech perception and language development. It was proposed that if these
fine discriminations can not be made the development of veridical phonological
code is disturbed, which causes various problems in language perception and
development (Merzenich et aI., 1996; Tallal et aI., 1996).
The fact that dysfunctions in the various sensory systems have been observed in
dyslexia point to a general sensory deficit in this disorder. Furthermore, dyslexic
individuals have problems in the different sensory systems both in processing
languege related and unrelated material (e.g., visually or auditorily presented words
and visual sinusoidal gratings or tonal patterns) and in discriminating sensory
transients (Stein & Talcott, 1999; see also Farmer & Klein, 1995). It has been,
therefore, proposed that a general magnocellular deficit underlies reading difficulties
(Stein, this volume).
The fact that dyslexia has a high incidence and is difficult to remediate since
there may be various underlying causes calls for better means to diagnose its
subtypes as early as possible in life. Measurement of task performance, a traditional
way of assesment, is often problematic especially if the dysfunction should be
identified in children. One possible means that might be well-suited even for very
early identification of auditory-system dysfunctions is an electric brain response
called the mismatch negativity (MMN).
2. THE MISMATCH NEGATIVITY RESPONSE OF THE BRAIN
The MMN of the event-related brain potentials (ERPs), discovered by Nlilitlinen and
colleagues in 1978 (Niilitiinen, Gaillard, & Miintysalo, 1978), can be elicited by
presenting the subject with a block of several hundred identical stimuli
("standards"), which are occasionally replaced by acoustically "deviant" stimuli
(Niilitiinen, 1992). When the brain's response to the standard stimuli is subtracted
from that to the deviant stimuli, the MMN can be seen approximately 100-300 ms
from the stimulus-change onset. The MMN has its major generator source in the
auditory cortex (Figure 1) but frontal-lobe sources have also been reported (Giard et
aI., 1995; Hari et aI., 1984; Rinne et aI., 1999; Rinne, Alho, Ilmoniemi, Virtanen, &
Nliiitiinen, 2000; for a review see Alho, 1995). The MMN reflects the functioning of
the auditory sensory memory (Naatanen, 1992). The repetitive standard stimulus
MISMATCH NEGATIVITY AND DYSLEXIA 361
forms a neural representation, "memory trace", of the stimulus features. If a new

stimulus does not match this representation in some wayan MMN is elicited.
The MMN can be elicited by any perceptible physical change in an auditory
stimulus sequence. In addition to simple stimulus features, such as pitch, duration, or
intensity of a sound (for a review see Niiiitiinen, 1992), even more complex and
abstract sound changes elicit it. For example, it was shown to be elicited by a
phonetic change of a stimulus (Aaltonen, Niemi, Nyrke & Tuhkanen, 1987;
Niiiitiinen et aI., 1997) or by an order reversal of a tone pair (Tervaniemi, Radii,
RadiIova, Kujala, & Niiiitiinen, 1999) even, when the tone pairs are presented
randomly over a wide frequency range (PaaviIainen, Jaramillo, & Niiiitiinen, 1998).
The duration of the sensory-memory trace can be measured with the MMN by
prolonging the inter-stimulus intervals (Miintysalo & Niiiitiinen, 1987).
The Mismatch Negativity (MMN)
Responses Subtraction waves

DlNiant
MMN
~~!.SbrwJard
400... """"A4.-t-f-.-
<? rN
~
Figure 1. A schematic illustration of the mismatch negativity (MMN). In the upper panel,
responses recorded at the franta-central scalp (Fz) and in the lower panel, responses
recorded at mastoid (Rm) are presented. Left column: the response elicited by the repetitive
standard stimulus (thin line) and that elicited by the occasional deviant stimulus (thick line).
The MMN is marked with shading. In column "Subtraction waves" the response elicited by
the standard stimuli is subtracted from the response elicited by the deviant stimuli.
It has been shown that there is a strong correlation between the MMN elicitation and
behavioral discrimination accuracy, and that stimulus discriminability is reflected in
MMN amplitude and latency (Tiitinen, May, Reinikainen, & Niiiitiinen, 1994). For
example, in backward-masking studies (Winkler & Niiiitiinen, 1992) it was shown
that if the masking stimulus was presented shortly after each auditory stimulus (20-
50 ms), no MMN was elicited and the subjects could not detect deviant stimuli in a
behavioral session. However, when this interval was prolonged to 150 ms, the MMN
was obtained and deviant stimuli were behaviorally detected by the subjects.
Niiiitiinen, Schrager, Karakas, Tervaniemi and PaaviIainen (1993), in tum, showed
that the MMN was elicited in subjects who were able to discriminate complex
spectrotemporal sound patterns but not in subjects who could not discriminate them.
The study of Niiiitiinen et ai. (1993) also showed that the MMN can serve as an
index of learning-associated neural plasticity. Subjects who could not discriminate
the complex sound patterns were given discrimination training. It was found that in
those subjects who learned to discriminate the sound patterns also the MMN
362 T. KUJALA
emerged. Kraus, McGee, Carrell, King, and Tremblay (1995) and Tremblay, Kraus,
and McGee (1998) found corroborating results by investigating discrimination
learning of speech contrasts that were initially impossible to discriminate. They
found that the MMN was elicited when subjects learned to discriminate the speech
contrasts. In fact, in the study of Tremblay et al. the MMN often emerged before the
subject was able to behaviorally discriminate the contrasts.
Winkler et aI. (1999) used the MMN in determining how a new phonetic
category is formed as a result of language learning. They compared the processing
of a Finnish phonetic contrast in Finns, Hungarians who did not know Finnish and
Hungarians who had a good command of Finnish. It was found that the Finnish
phonetic contrast, which is not present in the Hungarian language, was not
discriminated by Hungarians who did not know Finnish, whereas it was
discriminated by Finns and Hungarians who had a good command of Finnish.
Moreover, in these latter two groups this contrast elicited very similar MMNs
whereas it did not elicit MMN in Hungarians who did not know Finnish.
Importantly, the MMN can be obtained even without the subject's or patient's
attention to the auditory stimuli (Alho, Woods, Algazi, & Niiiitiinen, 1992;
Niiiitiinen, 1991). Usually it is measured when the subject is engaged in some other
activity than one relating to the auditory stimuli, for example, in reading book or
watching movies. This makes it especially attractive in studying infants or patients
who are unable or unwilling to communicate or perform task. For example, the
MMN has been obtained from comatose patients (Kane, Butler, & Cummins, 1993),
and aphasic patients (Alain, Woods, & Knight, 1998; Ilvonen et aI. , 2000) as well as
from children and infants (Alho, Sainio, Sajaniemi, Reinikainen, & Niiiitiinen, 1990;
Cheour et aI., 1998; Cheour-Luhtanen et aI., 1996; Kraus et aI., 1996).
3. THE MMN AND DYSLEXIA

Since the MMN is an index of auditory discrimination accuracy and it can be
elicited even without the subject's attention it has gained popularity during the
recent years in investigations on various clinical groups. It has also been applied in
studying phonological and auditory dysfunctions in dyslexia, and the results are very
promising.
Schulte-Kame et aI. (1998) used the MMN to compare the discrimination of
speech and non-speech stimuli in dyslexic and control adolescents. As speech
stimuli they had syllables Ida! and /bal, and as non-speech stimuli sine wave tones of
1000 Hz and 1050 Hz in frequency (with a 90-ms stimulus duration). It was found
that the MMNs elicited by the tone stimuli did not differ between the groups
whereas the syllables elicited a smaller MMN in dyslexic than in control
adolescents. According to the authors the result reflects a deficit specific to the
phonological system rather than a general failure in processing auditory information
in dyslexia.
Subsequent MMN studies provided somewhat contradicting results to those
obtained by Schulte-Kame and colleagues. Baldeweg et aI. (1999) used tone stimuli
in examining pitch and duration discrimination in dyslexic adults by the means of
the MMN and behavioral discrimination measurements. They had four different
deviants for each feature. In the duration condition the standard stimulus was 200 ms
long, and the deviant stimuli 160, 120, 80, and 40 ms long (the pitch of each
stimulus being 1000 Hz) and in the pitch condition the frequency of the standard
stimulus was 1000 Hz and those of the deviant stimuli 1015, 1030, 1060, and 1090
Hz (the duration of the stimuli being 50 ms). They found no group differences in
MMNs elicited by duration deviants nor in their active discrimination. However, the
MMNs elicited by the pitch changes were smaller in the dyslexic than in the control
subjects especially for smaller deviances, and they were less efficiently
discriminated by the dyslexic than control sUbjects. The MMNs for the 90-Hz
difference between the standard and deviant stimulus were fairly similar in the two
groups whereas they were diminished for the 60-, 30-, and, especially, IS-Hz
differences between the standard and deviant stimulus in the dyslexic subjects.
Taken these two studies together, the 1000-Hz vs. 1050-Hz stimulus difference
in the study of Schulte-Kame and colleagues might not have been small enough to
show differences between the dyslexic and control subjects. In the study of
Baldeweg et ai. group average MMNs suggested differences between the groups for
the 1000-Hz vs 1060-Hz stimuli but not for the 1000-Hz vs. 1090-Hz stimuli. It
should be noted that the pitch changes in the Baldeweg et ai. study were presumably
generally more difficult to discriminate because of a shorter stimulus duration (50
ms) than that used in the Schulte-Kame et ai. study (90 ms). The finding of impaired
syllable discrimination in dyslexic subjects of Schulte-Kame and colleagues, in turn,
corresponds well to previous literature suggesting that the discrimination of
consonants, which include rapid acoustic changes is especially difficult for dyslexic
individuals (see for example, Steffens, Eilers, Gross-Glenn, & Jallad, 1992; Reed,
1989).
There is a plenty of evidence obtained in studies using behavioral measures that
rapidly successive sound elements are deficiently processed by dyslexic individuals
(Tallal, 1980; Farmer & Klein, 1995). In addition, it has been suggested that their
auditory system is especially vulnerable to masking (Wright et aI., 1997). It could be
expected that the correct perception of long sound patterns, such as words or tone
patterns imitating words, might be difficult for dyslexic individuals. Sound-pattern
discrimination in dyslexia was recently studied with the MMN and behavioral
discrimination task (Kujala et aI., 2000). There were two stimulus conditions, one
including rhythmic patterns of four tones and the other one tone pairs, all tones
being equal in physical properties (see Figure 2). In the condition including patterns
of four tones, the deviant pattern was formed by presenting the 3rd tone of the
standard pattern earlier. In the tone-pair condition, the tone pairs were produced so
that the first and last tones of the tone patterns were removed.
In the behavioral session, dyslexic adults detected significantly less deviant tone
patterns than did control subjects (Figure 3). However, the tone pairs were equally
well discriminated by both groups. The MMN results provided further information
on what was problematic in the processing of the tone patterns in the dyslexic
subjects. The deviant tone pattern, which, in fact, differed from the standard pattern
twice (the too early tone is equivalent first to an addition and then to an omission of
a tone), elicited two MMNs in the control subjects (Figure 2). This suggests that
their auditory system reacted to the two deviations (first to an addition and then an
omission of a tone) in the pattern. In the dyslexic subjects, however, only the latter
MMN was elicited. This suggests that their auditory system could detect only the
change that occurred in the end of the pattern. This obviously was not enough to
364 T. KUJALA
normally perceive the difference between the patterns since the dyslexic subjects
performed poorly in the attentive deviant-detection task. These results indicate that
with the MMN one could determine which segments of auditory patterns are
deficiently discriminated.
Brain Responses
Control subjects
- - Dyslexic subjects
Tone-Pattern Condition Tone-Pair Condition
"vL .. ~. It
loom~
• • •• Standard Standard
•
~
•• • Deviant ... Deviant
Figure 2. MMNs (subtraction waves obtained by subtracting ERPs to standard stimuli from
those to deviant stimuli) in control (thick line) and dyslexic (thin line) subjects from one
franta -central EEG channel. Under the brain responses shown, schematic illustrations of the
standard and deviant tone patterns and pairs are presented with a time scale corresponding
to that of the brain responses. Figure adapted from Kujala et al. (2000).
Behavioral Responses
• Conlrol subjects
o Dyslexic subjects
Tone- Pattern Condition Tone-Pair Condition

100% 1500ms 100 ... 1500ms
T T
75 T T 75
1000 1000
T
50 50
1 500
25
500 T
25
T
0 0 0 0
HIT FA RT HIT FA RT
Figure 3. Performance in the behavioral target-detection task by control (black bars) and
dyslexic (grey bars) subjects. The subjects were instructed to detect occasional deviant tone-
patterns and tone-pairs (in separate blocks). The bars show mean percentages of hit and false
alarm (FA) rates and mean hit reaction times (RT) (with standard errors of mean) in tone-
pattern discrimination (left) and in tone-pair discrimination (right). Figure adapted from
Kujala et al. (2000).
This result might reflect a greater vulnerability of the dyslexic individuals'

nervous system to masking, as suggested previously (Wright et al., 1997). In order
to clarify further whether it is the interference caused by the preceding or following
auditory input that complicates the discrimination of auditory events in dyslexic
individuals, the MMN was recorded in a subsequent study to tone pairs (the two
tones differing in pitch) alone and to tone pairs with a preceding or a following tone
(Kujala, Belitz, Tervaniemi, & Naatanen, in preparation). It was found that the
dyslexic subjects had a diminished MMN to tone-order reversals. However, even
larger difference between the groups was obtained in the condition in which the tone
pairs were followed by a tone, whereas no group differences were found for the
MMNs elicited by the tone pairs that were preceded by a tone. These results suggest
that the auditory system of the dyslexic individuals is especially vulnerable to
backward masking effects.
An important application of MMN measurements in developmental language
impairments like dyslexia is to detect the deficit early in life. The MMN might offer
a means to identify children at risk early enough in order to start intervention before
any major developmental delays occur. It has been proposed that there is a strong
genetic component in dyslexia (for a review see Pennington, 1995). It would be
valuable to try to find early markers of the disorder in infants whose close relatives
have dyslexia.
By electrophysiological recordings it was recently shown that auditory
processing in infants who are at high genetical risk for dyslexia differed from that in
control children. Leppanen and colleagues (Leppanen, Pihko, Eklund, & Lyytinen,
1999; Leppanen et aI., in press) used an oddball paradigm, usually applied in
obtaining MMN, to compare auditory discrimination in infants at high risk for
dyslexia and in infants without such risk. As the repetitive standard stimulus they
used pseudoword fatal and as deviant stimuli two words /atta/, one having an
intermediate and the other a long silent gap corresponding the occlusion stage of the
stop consonant. These stimuli are relevant in the Finnish language since a change in
consonant or vowel duration may change the meaning of a word.
Leppanen and colleagues found several differences in the electrophysiological
responses of these two groups of children. First, there were group differences
already in the basic processing of the standard stimulus, as reflected in the P1
response of the infants. Second, the smaller duration change elicited no MMN in the
at high-risk group whereas it did in the control children and, in addition, the larger
duration change, which elicited MMNs in both groups, was diminished over the left
hemisphere in the at high-risk infants. This is consistent with previous reports
suggesting specifically left-hemisphere dysfunction in dyslexia (Galaburda, Menard,
& Rosen, 1994; Galaburda, this volume; Hughdahl, Helland, Faerevaag, Lyssand, &
Asbjoernsen, 1995; Witelson, 1977). Leppanen et al. (in press) also evaluated how
well the infants of the at high-risk group could be identified on the basis of
electrophysiological responses. It was found that 85% of the control infants and 68%
of the at-risk infants could be correctly classified on the basis of brain responses.
366 T. KUJALA
4. REMARKS
Dyslexia seems to be a disorder in which there are dysfunctions of different

modalities. However, currently an abnormal functioning of the phonological or the
auditory system in general is thought to be the major underlying factor in dyslexia.
The MMN, with which the auditory-system functioning can be probed even in
childhood, might offer a means to identify such subtypes of dyslexia in which the
dysfunction relates to the phonological or auditory system.
A means with which the discrimination of auditory stimuli or sensory memory
functions can be studied without such confounding factors as attention or motivation
is especially valuable when one wishes to detect developmental deficits or learning-
related problems. These dysfunctions should be identified as early as possible so that
the remediation could be started before major learning delays occur. Yet, their early
identification has been problematic since with the traditional way of measuring task
performance it is difficult to study perceptual functions of children and infants.
There are already several reports suggesting the usefulness of electrophysiological
methods in showing overall differences between clinical populations and control
healthy subjects. The MMN response is very promising in this approach since it can
be obtained even from a patient or child that is not willing or able to communicate or
concentrate in a task. It is, therefore, important that means to obtain the MMN and
other electrophysiological responses reliably from individual subjects would be
developed further in the future.
5. AFFILIATIONS
Dr. Teija Kujala

Cognitive Brain Research Unit, Department of Psychology,
P.O.Box 13, FIN-00014 University of Helsinki,
Finland.
e-mail: teija.m.kujala@helsinki.fi
6. REFERENCES
AaItonen, 0., Niemi, P., Nyrke, T., & Tuhkanen, 1. M. (1987). Event-related brain potentials and the
perception of a phonetic continuum. Biological Psychology, 24, 197-207.
Alain, c., Woods, D. L., & Knight, R. T. (1998). A distributed cortical network for auditory sensory
memory in humans. Brain Research, 812, 23-37.
A1ho, K. (1995). Cerebral generators of mismatch negativity (MMN) and its magnetic counterpart
(MMNm) elicited by sound changes. Ear and Hearing, 16, 38-51.
A1ho, K., Sainio, K., Sajaniemi, N., Reinikainen, K., & Naatanen, R. (1990). Event-related brain potential
of human newborns to pitch change of an acoustic stimulus. Electroencephalography and Clinical
Neurophysiology, 77, 151-155.
A1ho, K., Woods, D. L., A1gazi, A., & Naatanen, R. (1992). Intermodal selective attention. II. Effects of
attentional load on processing of auditory and visual stimuli in central space.
Electroencephalography and Clinical Neurophysiology, 82, 356-368.
Baldeweg, T., Richardson, A., Watkins, S., Foale, c., & Gruzelier, 1. (1999). Impaired auditory frequency
discrimination in dyslexia detected with mismatch evoked potentials. Annals of Neurology, 45, 495-
503.
Brannan, 1. R., & Williams, M. C. (1988). Developmental versus sensory deficit effects on perceptual
processing in the reading disabled. Perception and Psychophysics, 44, 437-444.
Cheour-Luhtanen, M., A1ho, K., Sainio, K., Rinne, T., Reinikainen, K., Pohjavuori, M., Renlund, M.,
Aaltonen, 0., Eerola, 0., & Niiiitiinen, R. (1996). The ontogenetically earliest discriminative response
of the human brain. Psychophysiology, Special Report, 33, 478-481.
Cheour, M., Ceponiene, R., Lehtokoski, A, Luuk, A, AIlik, J., A1ho, K., & Niiiitiinen, R. (1998).
Development of language-specific phoneme representations in the infant brain. Nature Neuroscience,
1,351-353.
Demp, J. B., Boynton, G. M., & Heeger, D. J. (1998). Functional magnetic resonance imaging of early
visual pathways in dyslexia. The Journal of Neuroscience, 18, 6939-6951.
Di Lollo, V., Hanson, D., & McIntyre, J. S. (1983). Initial stages of visual information processing in
dyslexia. Journal of Experimental Psychology, 9, 923-935.
A review. Psychonomic Bulletin and Review, 2, 460-493.
developmental dyslexia. Proceedings of the National Academy of Sciences, 91, 8010-8013.
Giard, M. H., Lavikainen, J., Reinikainen, K., Perrin, F., Bertrand, 0., Pernier, J., & Niiiitiinen, R. (1995).
Separate representation of stimulus frequency, intensity, and duration in auditory sensory memory:
an event-related potential and dipole-model analysis. Journal of Cognitive Neuroscience, 7, 133-143.
Grant, A. c., Zangaladze, A., Thiagarajah, M. C., & Sathian, K. (1999). Tactile perception in
developmental dyslexia: a psychophysical study using gratings. Neuropsychologia, 37,1201-1211.
Hari, R., Hiimiiliiinen, M., I1moniemi, R. J., Kaukoranta, E., Reinikainen, K., Salminen, J., A1ho, K.,
Niiiitiinen, R., & Sams, M. (1984). Responses of the primary auditory cortex to pitch changes in a
sequence of tone pips: Neuromagnetic recordings in man. Neuroscience Letters, 50, 127-132.
Hari, R., & Kiesilii, P. (19%). Deficit of temporal processing in dyslexic adults. Neuroscience Letters,
205, 138-140.
Hughdahl, K., Helland, T., Faerevaag, M. K., Lyssand, E. T., & Asbjoernsen, A (1995). Absence of ear
advantage on the consonant-vowel dichotic listening test in adolescent and adult dyslexics: specific
auditory-phonetic dysfunction. Journal of Clinical and Experimental Neuropsychology, 17, 833-840.
I1vonen, T.-M., Kujala, T., Tervaniemi, M., Salonen, 0., Niiiitiinen, R., & Pekkonen, E. (2001).The
processing of sound duration after left hemisphere stroke: event-related potential and behavioral
evidence. Psychophysiology, 38, 622-628.
Kane, N. M., Butler, S. R., & Cummins, B. H. (1993). Electrophysiological indicator of awakening from
coma. The Lancet, 341, 688.
Kujala, T., Belitz, S., Tervaniemi, M., & Niiiitiinen, R. (in preparation). Auditory sensory memory
interference in dyslexic adults.
Kujala, T., Myllyviita, K., Tervaniemi, M., A1ho, K., Kallio, J., & Niiiitiinen, R. (2000). Basic auditory
dysfunction in dyslexia as demonstrated by brain-activity measurements. Psychophysiology, Special
Report, 37, 262-266.
Kraus, N., McGee, T. J., Carrell, T. D., King, C., & Tremblay, K. (1995). Central auditory system
plasticity associated with speech discrimination learning. Journal of Cognitive Neuroscience, 7, 27-
34.
Kraus, N., McGee, T. J., Carrell, T. D., Zecker, S. G., Nicol, T. G., & Koch, D. B. (19%). Auditory
neurophysiologic responses and discrimination deficits in children with learning problems. Science,
273, 971-973.
Leppiinen, P. H. T., Pihko, E., Eklund, K. M., Guttorm, T. K., Aro, M., Richardson, V., & Lyytinen, H.
(in press). Brain responses reveal temporal processing differences in infants at a risk for dyslexia.
Developmental Neuropsychology.
Leppanen, P. H. T., Pihko, E., Eklund, K. M., & Lyytinen, H. (1999). Cortical responses of infants with
and without dyslexia: II group effects. NeuroReport, 10, %9-973.
Liberman, I. Y., & Shankweiler, D. (1985). Phonology and the problems to learn to read and write.
Remedial and special education, 6, 8-17.
Miintysalo, S., & Niiiitiinen, R. (1987). The duration of a neuronal trace of an auditory stimulus as
indicated by event-related potentials. Biological Psychology, 24, 183-195.
McGivern, R. F., Berka, C., Languis, M. L., & Chapman, S. (1991). Detection of deficits in temporal
pattern discrimination using the Seashore rhythm test in young children with reading impairments.
Journal of Learning Disabilities, 24, 58-62.
Merzenich, M. M., Jenkins, W. M., Johnston, P., Schreiner, c., Miller, S. L., & Tallal, P. (1996).
Temporal processing deficits of language-learning impaired children ameliorated by training.
Science, 271, 77-81.
368 T. KUJALA
auditory processing or phonological coding? Journal of Experimental Child Psychology, 64, 199-
23l.
Niiiitiinen, R. (1991). Mismatch negativity (MMN) outside strong attentional focus: A commentary on
Woldorff et al. Psychophysiology, 28, 478-484.
Niiiitanen, R. (1992). Attention and brain function. Hillsdale, N.J.: Lawrence Erlbaum.
Naiitanen, R., Gaillard, A. W. K., & Mantysalo, S. (1978). Early selective-attention effect on evoked
potential reinterpreted. Acta Psychologica, 42, 313-329.
Niiatiinen, R., Lehtokoski, A., Lennes, M., Cheour, M., Huotilainen, M., Iivonen, A., Vainio, M., Alku,
P., I1moniemi, R.J., Luuk, A., Allik, J., Sinkkonen, J., & Alho, K. (1997). Language-specific
phoneme representations revealed by electric and magnetic brain responses. Nature, 385, 432-434.
Niiiitanen, R., Schroger, E., Karakas, S., Tervaniemi, M., & Paavilainen, P. (1993). Development of a
memory trace for a complex sound in the human brain. NeuroReport, 4, 503-506.
Paavilainen, P., Jaramillo, M., & Naatiinen, R. (1998). Binaural information can converge in abstract
memory traces. Psychophysiology, 35, 483-487.
Pennington, B. F. (1995). Genetics of learning disabilities. Journal of Child Neurology, 10, 69-77.
Rayner, K., Pollatsek, A., & Bilsky, A. B. (1995). Can temporal processing deficit account for dyslexia?
Psychonomic Bulletin and Review, 2, 501 -507.
Reed, M. A. (1989). Speech perception and the discrimination of brief auditory cues in reading disabled
children. Journal of Experimental Child Psychology, 48, 270-292.
Rinne, T., Alho, K., I1moniemi, R. J., Virtanen, J., & Naatanen, R. (2000). Separate time behaviors of the
temporal and frontal MMN sources. Neurolmage, 12, 14-19.
Rinne, T., Gratton, G., Fabiani, M., Cowan, N., Maclin, E., Stinard, A., Sinkkonen, J., Alho, K., &
Niiatanen, R. (1999). Scalp-recorded optical signals make sound processing of the auditory cortex
visible. Neurolmage, 10, 620-624.
Salmelin, R., Service, E., Kiesila, P., Uutela, K., & Salonen, O. (1996). Impaired visual word processing
in dyslexia revealed with magnetoencephalography. Annals of Neurology, 40, 157-162.
Schulte-Korne, G., Deimel, W., Bartling, J., & Remschmidt, H. (1998). Auditory processing and
dyslexia: evidence for a specific speech processing deficit. NeuroReport, 9, 337-340.
Steffens, M. L., Eilers, R. E., Gross-Glenn, K., & Jallad, B. (1992). Speech perception in adult subjects
with familial dyslexia. Journal of Speech and Hearing Research, 35, 192-200.
Stein, J., & Talcott, J. (1999). Impaired neuronal timing in developmental dyslexia - the magnocellular
hypothesis. Dyslexia, 5, 59-77.
impaired: A critical review of the evidence. Psychonomic Bulletin and Review, 2, 508-514.
Tallal, P. (1980). Auditory temporal perception, phonics, and reading disabilities in children. Brain and
Language, 9, 182-198.
Tallal, P., Miller, S. L., Bedi, G., Byma, G., Wang, X., Nagarajan, S. S., Schreiner, C., Jenkins, W. M., &
Merzenich, M. M. (1996). Language comprehension in language-learning impaired children
improved with acoustically modified speech. Science, 271, 81-84.
Tervaniemi, M., Radii, T., Radilova, J., Kujala, T., & Naatanen, R. (1999). Pre-attentive discriminability
of sound order as a function of tone duration and interstimulus interval: a mismatch negativity study.
Audiology and Neuro-otology, 4, 303-310.
Tiitinen, H., May, P, Reinikainen, K., & Naatanen, R. (1994). Attentive novelty detection in humans is
governed by pre-attentive sensory memory. Nature, 327,90-92.
Tremblay, K., Kraus, N., & McGee, T. (1998). The time-course of auditory perceptual learning: which
comes first, the chicken or the egg. NeuroReport, 9, 3557-3560.
Winkler, I., Kujala, T., Tiitinen, H., Sivonen, P., Alku, P., Lehtokoski, A., Czigler, I., Csepe, V.,
I1moniemi, R. 1., & Naatiinen, R. (1999). Brain responses reveal the learning of foreign language
phonemes. Psychophysiology, 36, 638-642.
Winkler, I., & Niiatanen, R. (1992). Event-related potentials in auditory backward recognition masking: a
new way to study the neurophysiological basis of sensory memory in humans. Neuroscience Letters,
140, 239-242.
Witelson, S. F. (1977). Developmental dyslexia: two right hemispheres and none left. Science, 195, 309-
311.
Wright, B. A., Lombardino, L. J., King, W. M., Puranik, C. S., Leonard, C. M., & Merzenich, M. M.
(1997). Deficits in auditory temporal and spectral resolution in language-impaired children. Nature,
387, 176-178.
INDEX
INDEX
A • cognitive
ability, 185-186
development, 359
• acoustic processing, 205-,
processes, 12, 124-,
215,217
254-257,293-294
• agrammatism, 62-
• coherence, 45-
• amalgams, 107-110
• coherent motion, 201-, 225-
• amplitude modulation, 218
234
• animal models, 241, 243-
• competence difference
245
hypothesis, 84-85
• aphasia, 30-, 64, 74-
• concatenation, 105-
• articulatory gestures, 275
• context modes of
• attention
processing, 25-, 36-37
selective, 157-
-al gradient, 147-148
• contiuity, 84-85
• auditory
• contrastive signal coherence
(CSC), 41, 45-57
discrimination, 362-365
dysfunction, 359-360,
• cortical dominance, 166,
172,179-,186
366
processes, 223-224 • cortical malformation, 245-
246
• cylinder entropies, 45-
B
• bilingual, 110
D
• binarity, 113
• binocular instability, 203- • dendritic spread, 200
204 • desymbolization, 11, 181-
182
• bottom up, 177-180
• brain
• developmental
delay, 182, 217, 330-
injury, 166, 246
dyslexia, 173-174, 184-,
201-,215-,242-247,
254-, 265-, 283-, 330-,
C 359-
impairment, 180
• cell assembly, 165, 181 individual development,
• cerebellum, 204, 208-, 265, 182
270-276 word blindness, 166-
• Chinese language, 280 168
• closed class, 30, 62, 64-77 • discrepancy criterion, 16,
184-186, 188,254-,350
372 INDEX
• dual-route model, 253-254 linkage, 167,207,246,

• Dutch language, 79-, 107- 284
110 mechanisms, 207
• dyslexia, cf. develop. • German language, 36-, 79-,
dyslexia 105-,214,286-294,331
• grammar, 79-, 105-
E
H
• education, 168, 184, 254,
337 • head turn preference
• EEG (electro paradigm, 79, 92-
encephalogram), 43-, 362- • Hebrew language, 283
• English language, 251-252, • hemispheric dominance, cf.
62-, 105-,214-215,331 cart. dominance
• Event Related Potentials
(ERP), 31, 43-57
• evoked potential I
visual, 47-52
auditive, 218- • immunological,207-
• evolutionl81-182 • information integration, 47,
• eye movements, 121-, 138- 174-
148,203,226- • intelligence, 165-166, 184-
186, 221, 254-257
• IQ, cf. intelligence
F
• fixation, 122-133, 138-149, J
175,225-226
• FMRI (Functional Magnetic • Japanese language, 157
Response Imaging), 241,
359
• focusing effect, 156
K
• frequency modulation, 216-
• functional
coordination, 174-
• kana, 156-157
fragmentation, 176 • kanji, 156-157
projections, 87, 108- • kindergarten, 302-
G
• ganglion cell bodies, 200
L
• gaze duration, 126-
• genetic • lateral geniculate nucleus
(LGN), 200-208, 231
INDEX 373
• learning to read, 5,12, 171, sensitivity, 214-216,

179,182,218,300 229-232
• lexeme frequency, 132-133 skills, 202-203, 208-
• lexical 209, 214-215
ambiguity, 28-, 130-131 system, 331
processing, 12-13,25-, units, 251, 257-258,
129, 140-141, 144- • orthography, 129-130, 157,
• low level, 15, 124-, 148-149, 202-,251-,286,301-,331
207,225
P
M
• para foveal processing, 123-
• magnocellular,200-209, 129,141-
241- • parallel processing, 148-150
• mirror-image generalization, • parameter setting, 85, 96-
12,179- • parvocellular, 200-201, 284-
• mismatch negativity 285
(MMN),359- • phonemic awareness, 215,
• morphological complexity, 258-261
131-132 • phonological
• morphology, 88,126 awareness, 176, 215-
• motion sensitivity, 200-206, 224, 246, 261, 266, 299-,
225- 316-325,332-333,350-
• motor system, 241 352
• multicausal, 15, 184-189 neighborhood, 129-130
processes, 12,214,261,
300-
processing, 10-11, 176,
N 216-, 246, 253-, 266,
307,333-,360
• neighborhood size, 129-130 route, 200, 252-
• neural activity, 65, 165 turnaround, 14, 173
• neurophysiological • phonology, 129-130,214-,
processes, 199-,271-272, 251,301-302
360- • phrase structure, 44, 85,
106-
• pivots, 111
0 • psycholinguistic, 172-, 252-
• object constancy, 165, 178

• oculomotor control, 137-
• open class, 64-76,111-112 R
• orthographic
processing, 301-, 330- • RAN, 300- 312
350
374 INDEX
• rapid • syntax
serial naming, 299- aquisition, 105-
tapping, 241 objects, 105-
• reaction time (RT), 28-, 180, processing, 257
256-,364
• reading
acquisition, 182, 266, T
299
disability, cf. dyslexia • temporal
instruction, 184, 302 distribution, 41-, 75
level match design, 261 processing, 227, 241-,
normal-, 11-, 123, 141, 266-
148, 169, 293-, processing deficit, 266-
silent -, 199,304-
• thalamus, 200-, 242-
• reading span test (RST), • three stage model, 41-
155-
• top down, 177,230-
• recurrent network, 61, 64-
• topographical distribution,
• reversal errors, 15, 165, 172- 41-
• transient, 14-,207-,225-
285
S • transparent orthography,
251-
• saccade programming, 141,
147-
• saliency map, 149 V
• sensory -linguistic approach,
232- • verb placement, 79-
• sentence comprehension, 12, • ventrolateral stream, 200
25-,66-67,156,161
• visual
• severe speech impairment, cortex, 12, 33, 170, 183,
315- 200-,228-
• short term memory, 62-, deficit, 14, 169-,207,
176-,316-, 225-,284-
• simulation, 66-, 141 object recognition, 11,
• sound discrimination, 360- 165-166
• Spanish language, 214, 251-, processing, 16, 138,
331 173,181-,225-,266,
• specifier, 106- 359
• speech perception, 216-, 351 route, 253, 333
• spelling, 129, 202, 224, 229-, system, 14, 182, 200-,
255 f., 269-, 299-, 315-, 226,231,241-,284
329- word center, 167-
• strephosymbolia, 14-, 169- • visuomotor, 124-
• sustained, 14-,226 • vocabulary, 109,217,303-
• symmetry generalization,
15,178-,
INDEX 375
w
• well-formed ness condition,
86
• word
blindness, 14, 166-
identification, 252-253,
301-,352
recognition, 129-,221-,
252-, 282-, 329-
• working memory
deficit, 284-
visual-, 283-
NEUROPSYCHOLOGY AND COGNITION
The purpose of the Neuropsychology and Cognition series is to bring out volumes that promote
understanding in topics relating brain and behavior. It is intended for use by both clinicians and research
scientists in the fields of neuropsychology, cognitive psychology, psycholinguistics, speech and hearing,
as well as education. Examples of topics to be covered in this series would relate to memory, language
acquisition and breakdown, reading, attention, developing and aging brain. By addressing the theoretical,
empirical, and applied aspects of brain-behavior relationships, this series will try to present the
information in the files of neuropsychology and cognition in a coherent manner.
1. P.G. Aaron: Dyslexia and Hyperlexia. 1989 ISBN 1-55608-079-4; Pb. ISBN 0-7923-3155-9
2. R.M. Joshi (ed.): Written Language Disorders. 1991 ISBN 0-7923-0902-2
3. A. Caramazza: Issues in Reading, Writing and Speaking: A.
Neuropsychological Perspective. 1991 ISBN 0-7923-0996-0
4. B.F. Pennington (ed.): Reading Disabilities: Genetic and Neurological
Irifluences. 1991 ISBN 0-7923-1606-1
5. N.H. Hadley: Elective Mutism: A Handbook/or Educators, Counsellors
and Health Care Pro/essionals. 1994 ISBN 0-7923-2418-8
6. W.C. Watt (ed.): Writing Systems and Cognition: Perspectives from
Psychology, Physiology, Linguistics, and Semiotics. 1994 ISBN 0-7923-2592-3
7. I. Taylor and D.R. Olson (eds.): Scripts and Literacy: Reading and
Learning to Read Alphabets, Syllabaries and Characters. 1994 ISBN 0-7923-2912-0
8. V.W. Berninger (ed.): The Varieties o/Orthographic Knowledge: I:
Theoretical and Developmental Issues. 1994. ISBN 0-7923-3080-3
9. C.K. Leong and R.M. Joshi (eds.): Developmental and Acquired
Dyslexia: Neuropsychological and Neurolinguistic Perspectives. 1995 ISBN 0-7923-3166-4
10. N. Gregg: Written Expression Disorders. 1995 ISBN 0-7923-3355-1
11. V.W. Berninger (ed.): The Varieties o/Orthographic Knowledge: II:
Relationships to Phonology, Reading, and Writing. 1995 ISBN 0-7923-3641-0
Set (Vols. 8 & 11) ISBN 0-7923-3081-1
12. Y. Lebrun (ed.): From the Brain to the Mouth: Acquired Dysarthria and
Dysfluency in Adults. 1997 ISBN 0-7923-4427-8
13. J. Rispens, T.A. van Yperen and W. Yule (eds.): Perspectives o/the
Classification o/Speclfic Developmental Disorders. 1998 ISBN 0-7923-4871-0
14. C.K. Leong and K. Tamaoka (eds.): Cognitive Processing o/the Chinese
and the Japanese Languages. 1998 ISBN 0-7923-5479-6
15. P. Reitsma and L. Verhoeven (eds.): Problems and Interventions in
Literacy Development. 1998 ISBN 0-7923-5557-1
16. I. Lundberg, F.E. T0IInessen and I. Austad (eds.): Dyslexia: Advances in
Theory and Practice. 1999 ISBN 0-7923-5837-6
17. T. Nunes (ed.): Learning to Read: An Integrated View from Research and
Practice. 1999 ISBN 0-7923-5513-X
18. T. Heien and J. Lundberg: Dyslexia: From Theory to Intervention. 2000 ISBN 0-7923-6309-4
19. H.J. Hartman (ed.): Metacognition in Learning and Instruction: Theory,
Research and Practice. 2001 ISBN 0-7923-6838-X
20. E. Witruk, A.D. Friederici and T. Larchann (eds.): Basic Functions 0/
Language, Reading and Reading Disability. 2002 ISBN 1-4020-7027-6

Basic Functions of Language

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Basic Functions of Language

Uploaded by

Copyright:

Available Formats

BASIC FUNCTIONS OF LANGUAGE,

R. Malatesha Joshi, College of Education, Texas A&M University, U.S.A.

George Hynd, University of Georgia, U.S.A.

C.K. Leong, University of Saskatchewan, Canada

John Marshall, University of Oxford, U.K.

Gabriele Miceli, Universita Cattolica del Sacro Cuore, Italy

Loraine Obler, City University ofNew York, U.S.A.

Pieter Reitsma, Paedologisch Instituut Amsterdam, The Netherlands

SPRINGER SCIENCE+BUSINESS MEDIA, LLC

Copyright © 2002 by Springer Science+Business Media New York

Printed on acid-free paper.

INTRODUCTION TO BASIC FUNCTIONS OF LANGUAGE, 3

FROM LANGUAGE TO READING AND READING DISABILITY: 9

BASIC FUNCTIONS OF LANGUAGE ACQUISITION

CONTEXT EFFECTS ON LEXICAL PROCESSING DURING 25

MEASURING THE NEURAL DYNAMICS OF LANGUAGE 41

A MODEL OF LEARNING SYNTACTIC COMPREHENSION 61

THE ACQUISITION OF VERB PLACEMENT IN GERMAN: 79

MERGE AS A BASIC MECHANISM OF LANGUAGE: 105

COGNITIVE PROCESSES AND EYE MOVEMENTS DURING 121

THE ROLE OF ATTENTION AND SPATIAL SELECTION 137

THE EFFECT OF FOCUSING ON A SENTENCE IN JAPANESE 155

BASIC FUNCTIONS OF READING DISABILITY

READING DISABILITY AS A DEFICIT IN FUNCTIONAL 165

THE NEUROBIOLOGY OF READING DIFFICULTIES 199

A SENSORY-LINGUISTIC APPROACH TO NORMAL AND 213

READING DISABILITIES IN A LANGUAGE WITH TRANSPARENT 251

DYSLEXIA, THE CEREBELLUM AND PHONOLOGICAL SKILL 265

WORKING MEMORY IN DYSLEXIC CHILDREN: 281

PHONOLOGICAL RECODING PROBLEMS IN CHILDREN WITH 315

SUBLEXICAL AND LEXICAL PROCESSING OF YOUNG ADULTS 329

THE MISMATCH NEGATIVITY AS AN INDEX OF AUDITORY 359

• BASIC FUNCTIONS OF LANGUAGE ACQUISITION AND

• BASIC FUNCTIONS OF READING

• BASIC FUNCTIONS OF READING DISABILITY

Leipzig, January 2002

INTRODUCTION TO BASIC FUNCTIONS OF

1. THE STRUCTURE OF THE BOOK

E. Witruk, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading,

2. BASIC FUNCTIONS OF LANGUAGE ACQUISITION AND LANGUAGE

This section contains a selection of compelling papers which investigate basic

3. BASIC FUNCTIONS OF READING

1980) applied in numerous studies on sentence processing. They show that a

4. BASIC FUNCTIONS OF READING DISABILITY

4.1 Theoretical Approaches

4.2 Empirical Studies

The chapter by Witruk, Ho and Schuster discusses three experiments conducted

The present collections of papers provide an interesting view on a human ability

a surprise that recent brain imaging studies have demonstrated an increased

FROM LANGUAGE TO READING AND READING

E. Witruk, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading,

2. BASIC FUNCTIONS OF LANGUAGE COMPREHENSION

3. BASIC FUNCTIONS OF READING

In most general terms reading can be characterized as language comprehension,

4. ASPECTS OF MEMORY AND ATTENTION DURING READING

Besides these fundamental processes of reading and language comprehension,

5. READING DISABILITY RESEARCH

Dr. Angela D. Friederici

Carpenter, P. A, Miyake, A, & Just, M. A (1994). Working memory contraints in comprehension:

BASIC FUNCTIONS OF LANGUAGE

CONTEXT EFFECTS ON LEXICAL PROCESSING

2. SCOPE OF REVIEW / ANALYSIS

E. Witruk, A. D. Friederici, & T. Lachmann (Eds.), Basic Functions of Language, Reading,