Reading in a Second Language

APPLIED LINGUISTICS AND LANGUAGE STUDY

GENERAL EDITOR
CHRISTOPHER N. CANDLIN,
Chair Professor of Applied Linguistics
City University of Hong Kong, Hong Kong
For a complete list of books in this series see pages v-vi.
Reading in a Second
Language: Process,
Product and Practice

A.H. Urquhart and C.J. Weir

First published 1998 by Pearson Education Limited
Published 2013 by Routledge
2 Park Square, Milton Park, Abingdon, Oxon 0X14 4RN
711 Third Avenue, New York, NY 10017, USA
Routledge is an imprint of the Taylor & Francis Group, an informa business
Copyright © 1998, Taylor & Francis.
The right of A.H. Urquhart and C.J. Weir to be identified as the authors of this
Work has been asserted by them in accordance with the Copyright, Designs and
Patents Act 1988.
All rights reserved. No part of this book may be reprinted or reproduced or utilised in any
form or by any electronic, mechanical, or other means, now known or hereafter invented,
including photocopying and recording, or in any information storage or retrieval system,
without permission in writing from the publishers.
Notices
Knowledge and best practice in this field are constantly changing. As new research and ex­
perience broaden our understanding, changes in research methods, professional practices,
or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in
evaluating and using any information, methods, compounds, or experiments described
herein. In using such information or methods they should be mindful of their own safety
and the safety of others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or edi­
tors, assume any liability for any injury and/or damage to persons or property as a matter of
products liability, negligence or otherwise, or from any use or operation of any methods,
products, instructions, or ideas contained in the material herein.
ISBN 13: 978-0-582-29836-1 (pbk)
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging-in-Publication Data
Urquhart, A.H.
Reading in a second language : process, product, and practice /
A.H. Urquhart and C.J. Weir.
p. cm. — (Applied linguistics and language study)
Includes bibliographical references and index.
ISBN (invalid) 0-582-29836-9 (pbk.)
1. Language and languages—Study and teaching. 2. Reading.
3. Language and languages—Ability testing. I. Weir, Cyril J.
II. Title. III. Series.
P53.75.W45 1998
418'.4— dc21 98-7949
CIP
Set by 35 in 10/12pt Baskerville
 APPLIED LINGUISTICS AND LANGUAGE STUDY

GENERAL EDITOR
CHRISTOPHER N. CANDLIN
Chair Professor of Applied Linguistics
City University of Hong Kong
Hong Kong
Error Analysis The Classroom and the Language
Perspectives on Second Language Learner
Acquisition Ethnography and second-language
JACK C. RICHARDS (ED.) classroom research
LEO VAN LIER
Stylistics and the Teaching of
Literature Observation in the Language
H.G. W ID D O W SO N
Classroom
Contrastive Analysis DICK ALL W RIGHT
CARL JAMES
Learning to Write Listening in Language Learning
MICHAEL ROST
First Language/Second Language
AVIVA FREEDMAN, IAN PRINGLE and
JANICE YALDEN (EDS)
Listening to Spoken English
Second Edition
Language and Communication GILLIAN BROWN
JACK C. RICHARDS and RICHARD W.
SCH M IDT (EDS) An Introduction to Second
Reading in a Foreign Language Language Acquisition Research
J. CHARLES ALDERSON and A.H. DIANE LARSEN-FREEMAN and
UR Q UH ART (EDS) MICHAEL H. LONG

An Introduction to Discourse Process and Experience in the

Analysis Language Classroom
Second Edition MICHAEL LEGUTKE and HOWARD
MALCOLM C OULTHARD THOMAS
Bilingualism in Education
Aspects of theory, research Translation and Translating
and practice Theory and Practice
jim c u m m in s and m e r r i l l s w a i n ROGER T. BELL
Second Language Grammar Language Awareness in the
Learning and Teaching Classroom
WILLIAM E. RUTHERFORD CARL JAMES and PETER
Vocabulary and Language GARRETT (EDS)
Teaching
RONALD CARTER and MICHAEL Rediscovering Interlanguage
M CCARTHY LARRY SELINKER
Language and Discrimination Autonomy and Independence in
A Study of Communication in Multi­ Language Learning
ethnic Workplaces PHIL BENSON and PETER
CELIA ROBERTS, EVELYN DAVIES and VOLLER (EDS)
TON JU P P
Language, Literature and
Analysing Genre the Learner
Language Use in Professional Settings Creative Classroom Practice
VIJAY K. BHATIA RONALD CARTER a n d J O H N
MCRAE ( e d s )
Language as Discourse Language and Development
Perspectives for Language Teaching Teachers in a Changing World
MICHAEL MCCARTHY and BRIAN KENNY and WILLIAM
RONALD CARTER SAVAGE (EDS)
Second Language Learning Communication Strategies
Theoretical Foundations Psycholinguistic and Sociolinguistic
MICHAEL SHARW OOD SM ITH Perspectives
GABRIELE KASPER and ERIC
From Testing to Assessment KELLERMAN (EDS)
English as an International Language
CLIFFORD HILL and KATE Teaching and Language Corpora
PARRY (EDS) ANNE W IC H M A N N , STEVEN
FLIGELSTONE, TONY MCENERY and
Interaction in the Language GERRY KNOWLES (EDS)
Curriculum Errors in Language Learning
Awareness, Autonomy and and Use
Authenticity Exploring Error Analysis
LEO VAN LIER CARL JAMES
Phonology in English Language Translation into the Second
Teaching Language
An International Language STUART CAMPBELL
M ARTHA C. PEN N IN G T O N
Strategies in Learning and Using
Measuring Second Language a Second Language
Performance ANDREW D. COHEN
TIM MCNAMARA Managing Evaluation and
Innovation in Language
Literacy in Society Teaching
RUQAIYA HASAN and GEOFF Building Bridges
WILLIAMS (EDS) PAULINE REA-DICKINS and KEVIN
P. GERMAINE (EDS)
Theory and Practice of Writing
An Applied Linguistic Perspective Reading in a Second Language
WILLIAM GRABE and ROBERT Process, Product and Practice
B. KAPLAN A.H. U RQ UH ART and C.J. WEIR
 vii

Contents

Editors ’ acknowledgements xi
Publisher's acknowledgements xii
General Editor’s preface xiii

Introduction 1
Literacy 1
The scope of the term ‘literacy’ 5
Literacy and power 6
Cognitive v Social 8
Organisation of the book 10

1 Prelim inaries 13
What is reading? 13
The primacy of spoken language 24
Writing systems 27
Reading versus listening 31
The L2 reader 33
Summary 34

2 The theory of reading 37

2.1 Models 37
Classes of reading model 39
Process models 39
Componential models 46
2.2 Components in detail 51
Word recognition 51
Language 57
Background knowledge 62
Literacy 73
viii Contents
2.3 Comprehension, skills, strategies and styles 84
Comprehension 85
Skills 88
Strategies 94
Strategies and skills 96
Strategies and styles 98
2.4 Different kinds of reading 100
Expanding a model 105
3 Testing reading com prehension(s) 111
3.1 Introduction: the limits of testing 111
Measuring types of reading 111
Comprehensions and interpretations 112
Focusing on comprehensions 115
What is there to comprehend? 117
3.2 Task-based factors 120
Testing components 120
Deciding on skills and strategies 121
The evidence for a multidivisible view 123
Casting doubt on multidivisibility 125
Casting doubt on unidimensionality 129
The status of expeditious reading 130
Testing at the microlinguistic level 133
Dangerous implications of a unitary view 134
Dangerous implications of a multidivisibility view 135
3.3 Text-based factors 140
Text type 140
Propositional content 142
Topic familiarity 143
Vocabulary 145
Channel 145
Size 145
Difficulty 146
3.4 Test formats 147
Origins 147
Some important facets of reading test design 150
Recent methods for testing reading ability 154
3.5 Conclusions 165
4 The teaching of reading 171
4.1 Teaching and testing 171
Similarities 172
Differences 173
Background issues in the teaching of reading 177
 Contents ix
4.2 Focus on metacognitive strategies 179
Some potentially useful metacognitive strategies 183
Pre-reading activities 184
While-reading strategies 186
Post-reading strategies 187
4.3 Focus on cognitive strategies and skills 188
Introduction: purpose determines choice 188
Individual tasks at the local level 189
Individual tasks at the global level 201
Extensive reading 215
Working together 222
4.4 What should happen in the reading classroom? 224
The way forward: some general considerations 226
Will it work? 227
Remediation 228
Simply the best? Is there a best way to teach reading? 229
Effective schools 230

5 Future research 233

Introduction 233
Two research communities 234
The applied linguistics contribution 237
Causes for the differences 239
The way forward 240
5.1 Models and different types of reading 242
Criteria for judging models of the reading process 242
Investigations of types of reading 244
5.2 Reading and the teaching of grammar 255
Introduction 255
A pedagogical approach 255
Defining syntax 259
A formal model 261
Correlational studies 264
Beyond correlations 266
Refining the grammar syllabus 267
Grammar and texts 268
Grammar and different kinds of reading 269
Conclusions 269
5.3 The use of tests to investigate componentiality in
reading research 270
Specification of operations and performance conditions
in an EAP reading construct 274
A priori validation 275
A posteriori validation 279
x Contents
Using the test to explore the construct reading 294
Conclusion 295
Appendix 1 297
Appendix 2 306
References 308
Author Index 339
Subject Index 344
 xi

Editors’ acknowledgements

The authors would like to thank Chris Candlin for his support
and guidance throughout the writing of this book. His insightful
comments and firm grasp of the whole domain of applied lin­
guistics have been invaluable. They would also like to thank Jin
Yan and Luo Peng from Shanghai Jiaotong University, PRC, and
other members of the reading research group at CALS, University
of Reading, for their constructive feedback and help in clarifying
our thoughts on the new blueprint for reading developed in this
book, and in particular for their help in developing our taxonomy
of reading skills and strategies.
Cyril Weir offers his gratitude to all his colleagues in CALS for
their understanding over the past few years and, in particular, to
Don Porter and Eddie Williams for their support and encourage­
ment. He would also like to thank Lois Archer, Shigeko Amano
and Jessica Wu for their close reading of the manuscript. Ron
White deserves a special acknowledgement for being one of the
rare breed of managers who is both liked and respected by his
colleagues. His view of the role of management as bringing the
best out of his staff rather than trying to get the most out of them
is much appreciated. CALS generously awarded this author sab­
batical leave both to start and finish this book.
Sandy Urquhart would like to thank his family for their support.
xii

Publisher’s acknowledgements

We are indebted to Kluwer Academic Publishers for permission to

reproduce a version of the contribution ‘The Testing of Reading
in a Second Language’ by Cyril Weir in Encyclopedia of Language
and Education edited by David Corson et al. © 1998.
 xiii

General Editor’s preface

The subtitle of this new contribution to the Applied Linguistics

and Language Study Series is significant. Reading is a matter of
interaction. Products for reading in the form of texts are not only
themselves outcomes of writer processes, cognitive, linguistic, and
social, but invite and activate those processes in their readers. Not
that the engagement of writer and reader only begins when the
reader picks up or otherwise accesses a text. Writers always have
readers in mind, and we know that expert writers tailor their
writing to their readers’ presumed cognitive and linguistic capaci­
ties, partly by establishing some common social context for the
interaction. We may trace in writing evidence of the degree to
which this accommodation takes place, and by monitoring read­
ing processes we can appraise its effectiveness. Thus, as readers
evaluate writers, so do writers envisage readers. Reading is in this
way a kind of sympathetic and unscripted writing. Seen this way,
the instructing and appraisal of reading is linked to those of writ­
ing, providing an example of how teaching and testing can go
hand in hand. They do so explicitly in this innovative book by
Sandy Urquhart and Cyril Weir. It is not axiomatic that teaching
and testing enjoy this synergy. Indeed, it is often argued that they
are adversarial. Testing is held by some to inhibit teaching, or
through washback to come to dominate and direct it. Exploring
the interaction between teaching and testing is one of the distinc­
tive features of this book.
There are further interactions to be noted. Reading and writ­
ing invite in their relationship comparisons between listening and
speaking. The authors establish this parallelism in their Preliminar­
ies in the first chapter of the book. It is a compelling association
since it embraces the cognitive, the linguistic and the social, within
xiv General Editor’s preface
the speaking/listening connection and between that of reading
and speaking. Reading, writing, listening and speaking not only
interactively involve these processes, they are also in themselves
social practices. They occur in contexts. The authors are sensitive
to this but they are also cautious. They warn against too easy an
assumption of transfer from one context to another. Reading in
one context may not be the same as in another, and transferabil­
ity and generalisability of effective reading strategies cannot be
taken for granted. Testing is one way in which we may explore
this issue. Here, too, testing and teaching are interlinked.
Reading as a social practice is, however, not only a little ‘c’
context matter. As a social practice it always also invokes the big­
ger ‘C’ Context, what sociologists refer to as the institutional
matrices within which particular reading practices are embedded.
Such ways of going about reading contribute to our understand­
ing of the role reading plays in the definition of the institution as
a particular social and cultural world. The receptive conditions
governing the act of reading are important, as are the productive
conditions governing writing. The experience and expertise of
the authors is crucial at this point, as elsewhere in the book. Their
close familiarity with the contexts and Contexts of reading en­
sures that there is no light assumption of transferability, nor any
ignoring of the impact of the social interactivity of reading on
its cognitive interactivity within the reader. Nonetheless, explor­
ing reading from the perspective of mind and that of social rela­
tionships requires discrete discussion. Here the interplay in the
book among theoretical, instructional and evaluative concepts and
practices allows each its place, but, in the end, they interact. They
do more than that. They are expressly positioned in the book to
engage each other. Theories and models of reading and their
components are tested against the constraints imposed by testing,
and against the exigencies of instruction. Readers of the book are
thus not led up some garden path. Nor are they allowed to forget
that reading, as I say earlier, is essentially an engagement with
text. How one describes such texts, as a reading specialist or as a
testing specialist, will not be the same as if one described the text
from the perspective of a linguist or a cognitive psychologist, or
indeed as a sociologist. This is not always fully appreciated. Gram­
mar is not disciplinarily neutral. The authors recognise this. In­
deed, they are very explicit about boundaries in this case, as the
 General Editor’s preface xv
section on Future Research makes clear. Nevertheless, they do not
isolate these perspectives in practice. How one processes the gram­
mar of a text does have some connection with the way the text
is structured grammatically. Grammar, after all, acts to stimulate
particular processes of exploring meaning potential.
Understanding such processes requires us to accommodate dif­
ferent strategies and styles of reading. There is no one way of
reading a text, and reading variability depends on a range of
factors: what one is reading, who one is as a reader, who or what
one may be reading for and why, what level of linguistic compe­
tence one has, what relationships one may have with the writer,
and for a host of other reasons. Such a range of variables imposes
a challenge to testers, who have their own criteria and standards
to meet, ones which urge convergence and normativeness. Thus,
analysts of text, analysts of mind, and analysts of social context
necessarily interact in the analysis of reading. Adding testing to
this mixture causes a redefinition of the roles and boundaries of
each. So, too does teaching, but not in the same way. Teaching,
after all, is about divergence as well as convergence.
Interaction involves actors. An issue of importance for the book
is the tension that arises between considering the reader as indi­
vidual and the reader as a member of a social group. The authors
believe that a current bias towards the social act of reading needs
to be corrected so that the cognitive capacity of the reading indi­
vidual is not lost, and his or her personal contribution to the
activity of reading thereby underestimated. Here, again, the indi­
vidualised focus of testing provides a useful corrective. Not that
large scale testing of reading, such as that conducted over several
years by expert teams advised by Cyril Weir in the almost un-
graspable context of several million testees in China, does not
provide some valuable collective and aggregative information.
Nonetheless, examination halls, the authors’ default environment,
are but collections of individuals, whose capacities need personal­
ised assessment, and the enhancement of whose strategies and
skills in classrooms through instruction, they conclude, is cur­
rently best assessed cognitively rather than socially. Yet, authenti­
cating reading remains a significant instructional goal. It is a
measure of the wisdom of this clearly argued and sympathetic
book that the authors conclude that it will only be through the
interaction of theory, instruction and assessment that the challenge
xvi General Editor’s preface
of this authentication can be addressed in classrooms, or indeed,
in examination halls, and the legitimate pressures of the textual,
the cognitive and the social can be accommodated by researchers,
teachers and testers.
Professor Christopher N Candlin
General Editor
English Language Education and Communication Research
City University of Hong Kong
 1

Introduction

The book is primarily intended for those professionally concerned

with the teaching and testing of reading in a foreign language.
We take this to include teachers, both practising and in training,
teacher-trainers, materials writers, and those concerned with the
production of reading tests, or with training testers.
Borrowing a term used by anthropologists and semanticists, we
will refer to our prototypical teaching situation as an L2 classroom
in a school, college or university. This does not mean that we are
ignorant of the fact that reading may be taught, or learned, in
other contexts, e.g. the home, the workplace, or that a teacher
will always be present. We adopt this as our prototype simply in
order to establish a point of reference. Similarly, our prototypical
testing situation is the examination hall. Again we are fully aware
that individuals can test themselves, or, perhaps more important,
that classroom teachers can make use of informal tests to check
reading performance. Again, though, we need a reference point
for the discussion below and throughout the book. We hope that
readers will be able to transfer what we say to their own particular
situation.

Literacy
The teacher of reading is in the business of attempting to improve
literacy. Literacy has been the focus of a great deal of work over
the last few decades. By discussing, however briefly, some of this
work, we shall hopefully set this book against a broad and mean­
ingful context.
The definition of literacy is crucial (cf. Baynham, 1995 and
Venezky et al., 1990). One implied definition, touching in its
2 Reading in a Second Language
optimism, can be found in the US Census of 1940, where a per­
son is judged literate if they are 10 years of age or older and has
completed 5 or more years in school (Newman and Beverstock,
1990). For serious discussion nowadays, the starting point must be
UNESCO proposals of the 1950s that literacy should be defined in
terms of minimal and functional literacy (see Venezky et al., 1990).
The former refers to the ability to read and write a simple mes­
sage; the latter to a level of literacy sufficiently high for a person
to function in a social setting. Prior to this, attempts were made to
define what Street (1995: 76) refers to as ‘autonomous literacy’,
that is, literacy divorced from any context. Such a hypothetical
level is impossible to establish and is inappropriate in many situ­
ations. As Venezky et al. (1990: ix) say:
the diverse communities that make up contemporary America are
so variegated that simple dichotomies such as literate-illiterate fail
to capture what are real differences in what people know and how
they behave in certain situations.
The decision to take into account the social relevance of literacy
has been momentous, leading as it has to modern notions of mul­
tiple ‘literacies''. The concept can apply both to attitudes towards
the value of literacy, and to the role of literacy in the society as a
whole. Street (1995) comments on problems caused by a clash in
such values between outside educators and local values in Melan­
esia and among Amish communities in the USA.
In many parts of modern Western society, illiteracy in an adult
is seen as a stigma, which the person concerned is often at pains to
disguise. In a novel by Ruth Rendell, a character becomes alien­
ated from society as a result of such subterfuges, grows up devoid
of normal social emotions, and eventually murders her employers
in order to disguise the fact that she cannot take a message.1 In
contrast, in a well-known article, Fingeret (1983) describes how,
in urban communities in the USA, illiteracy is not a stigma, help
with filling in forms, etc., being exchanged for other skills like car
repairing.2 It is likely that large and important cultures exist which
view literacy in a way very different from our own. Certainly it has
been our experience in the L2 classroom that some students do
not seem to view written material as a potential source of informa­
tion accessible to the individual reader. If this observation is valid,
the students’ behaviour may be a result of views of the role of
literacy in their society. To remedy this, an intensified training in
 Introduction 3
reading skills may not be sufficient; possibly more generalised
approaches to the uses of written material might work.
The different levels of literacy required by different groups can
be seen when one considers different occupations or professions
in the same society. It seems obvious that different occupations
make greater professional demands of the written word than others.
Tudge - arguing in New Scientist that the population divides into
those who like reading and those who do not, and that the educa­
tion system should take this into account - notes that trainee sur­
geons do not need to make much use of the printed word (they
can rely on video-tapes, computer simulations, etc.) whereas lawyers
do (Tudge, 1987). It is also obvious that the required level of pro­
fessional literacy varies radically over time. It seems likely that, in
the last century, plumbers, for example, had comparatively low
professional requirements for literacy. This is not likely to be the
case now.
L2 students in the language class are often preparing to study
on academic or other courses conducted in the L2. Such courses,
e.g. engineering, nursing, can be assumed to have their own literacy
requirements and criterion levels. It is clearly of vital importance
that test and syllabus designers take these different requirements
and levels into account. In Chapter 3 we discuss this topic under
task- and text-based factors. Particularly in LSP, Needs Analysis has
long been part of the design of courses (cf. Munby, 1978; Nunan,
1993). It is our opinion, however, that detailed ethnographic case
studies of the needs of different study areas have not been con­
ducted, or at least published. The difficulty of conducting such
studies should not be underestimated.
Related to this is the notion of transfer of literacy from one area
to another. Implicit in what has been said above, reading skills
acquired in the reading class should be transferred to the student’s
eventual study area. However, the literature on transfer tends to
be pessimistic. Mikulecky (1990) claims that a major misconception
in literacy studies is that ‘mastering literacy in one context sub­
stantially transfers to other contexts’, and adds ‘Transfer of literacy
abilities is severely limited by differences in format, social support
networks, and required background information as one moves
from context to context’ (p. 25). He contrasts ‘literacy in schools’
involving ‘independent reading for answering questions at the end
of the chapter, or, on some occasions, carefully studying material
to remember, synthesize, or evaluate it [what we later refer to as
4 Reading in a Second Language
“careful local” or “careful global” reading] ’ and claims that these
activities differ from ‘those used to read a troubleshooting manual
on the job or gather information to fill in a form ’ (p. 25). Sticht
(1980) has also claimed that there is little transfer from general
reading ability, and reading in specific situations. He contrasts
‘reading to learn’ in school, with ‘reading to do’ in the workplace,
or ‘reading toward a goal of locating information for immediate
use that need not be recalled later’. While transfer is always a
potential problem, we see behind some of this discussion rather
conventional and unimaginative limitations on the types of read­
ing which should be practised in the reading class, and point
below (Chapter 4) to a wider range of tasks and behaviours.
Returning now to multiple literacies, the notion extends to
individuals. A reader may be highly literate in one area but minim­
ally literate in another. The present authors are not functionally
literate with respect to, for example, professional scientific texts,
and only marginally so vis-a-vis teenage fanzines. The L2 teacher
of adults will be familiar with the situation in which students in
the class are superior readers in certain professional or other
areas. In such cases, common in LSP situations, the teacher clearly
must see their role as that of facilitator rather than pedagogue.
This aspect of multiple literacy also throws doubt on the concept
of the ‘good reader’, often referred to in the research literature.
Venezky (1990: 12) argues that ‘Most readers show differing read­
ing abilities across different types of material’, a claim repeated by
Urquhart (1996). An implication of this is that, for the L2 reader,
we may sometimes need not a single test score but a profile com­
posed of performances on different types of text and task. In
Chapter 3 we present evidence from testing to support this view.
What might be viewed as an extreme case of different literacies
within the same individual can be seen when the individual moves
into a different language area. Venezky draws attention to the
immigrant into the USA, illiterate in English but literate in Viet­
namese and French; and the reverse, literate Americans who are
functionally illiterate when they move to other parts of the world.
A large proportion of L2 students in reading classes are already
literate in their LI (but see Wallace, 1988; Williams, 1995). A
number of consequences rise from this differential literacy in dif­
ferent languages. The fact that many learners are literate in their
LI but not functionally literate in the L2 is the basis for the
notion of ‘transfer of reading skills’ discussed in Chapter 2. More
 Introduction 5
generally, LI literacy must be one of the components which readers
bring to the task of L2 reading (cf. Bernhardt, 1991b). Bernhardt
sees ‘literacy’ in this sense as consisting of knowledge of texts, etc.
However, it must be broader than that, including not just, in
many cases, knowledge of script, but also, crucially, knowledge
that written text is language, containing messages from other
language users.

The scope of the term ‘literacy’

When we talked earlier about ‘defining literacy’, we really meant
‘defining levels of literacy’. There is also the question of what
precise activities are to be covered by the term. Of recent years, it
has been given a wide range of referents. Newman and Beverstock
(1990) point out that ‘membership of an interpretive community
is often referred to as a specific literacy, as when we speak of computer
literacy. . .’. They later cite ‘civic literacy, prison literacy, work­
place literacy’ (p. 44). Nor is this just rhetoric. As they point out,
‘in Michigan, the definition of literacy is being broadened through
testing in five areas: written and oral communication, mathematics
and related skills, problem, workplace attitude, and job-seeking
abilities’ (p. 42). They warn that ‘(while) the old definitions of
literacy were too narrow; some of the new definitions may be too
broad. The new and humane . . . definitions of literacy are so pro­
fuse that we may be loading the word with burdens of meaning
that will eventually negate its usefulness as a working term ’ (p. 43).
Venezky, viewing literacy as a complex of skills, mentions read­
ing, writing, numeracy and document processing as four skill areas
lumped under literacy. Venezky argues that ‘reading is clearly
primary to any definition of literacy, and, in some sense, the others
are secondary’ (1990: 9). For numeracy, he suggests that it be
constrained to ‘those basic numeric operations that are critical
for ordinary meaning of print’ (p. 8). It is a problem which often
emerges during reading test construction. On the whole, we tend
to think that numeracy should be treated as a separate skills area,
as it often is in British educational debate. Venezky seems to see
document processing as a separate skill area, dependent on read­
ing. His reason seems to be statistical, based on low correlations
found between performance on ‘documents’ and on ‘prose’. There
are hints throughout the literacy debate that the texts being used
6 Reading in a Second Language
to measure literacy are very traditional - ‘stories’, etc. There is
nothing in the models considered in Chapter 2 to indicate that
‘document reading’ should differ qualitatively from the reading
of prose. We thus have no difficulty in fitting it into what we con­
sider as reading.
Macias (1990) argues against ‘multi-literacies’: \ . . numeracy
and document processing, as well as the uses of literacy (primarily
writing), become secondary aspects of literacy study, not parts of
the definition of literacy.’ With the exception of document pro­
cessing, we go along with this in practice, though perhaps not in
theory. We are concerned in this book with reading, and exclude
writing, the second half of the literacy equation. This is in spite of
the attractiveness of combining the two activities in realistic tasks.
Our reasons are partly pragmatic; it is well known, as is set out in
Chapter 3, that using written work to test reading runs the risk of
confounding two skills areas. In general, we consider that the two
skills areas are distinctive enough to be handled separately, as is
conventionally the practice in L2 testing and teaching. Writing is
well covered in the comprehensive volume by Grabe and Kaplan
(1996). We exclude numeracy, however, because, as is argued in
‘Preliminaries’ (Chapter 1), we consider that the term reading is
best kept for the processing of written language.

Literacy and power

The introduction of cultural context into literacy studies has led
many writers to consider the relationship between literacy and
power. Fingeret (1990) cites Goody and Watt (1968) to the effect
that ‘choices about who reads, what they read, and how they use
what they read always have been connected to the distribution of
power in a society’. This connection is often seen as negative - one
group using literacy to maintain and exercise power over another.
Groups accused of doing so include slave owners in nineteenth-
century America: slaves in some of the southern states were for­
bidden to read (Newman and Beverstock, 1990). Missionaries in
nineteenth-century Melanesia ‘saw literacy practice as a means
simply to conversion and to social control, (and thus) they had no
interest in providing any more than was necessary for this bare
minimum’ (Street, 1995: 84). This form of control is nicely dubbed
‘domestication’ by Mikulecky (1990: 26). The power group in
 Introduction 7
our own culture may be the literate majority, which ‘secure in its
position of dominance, partially attributes its success to literacy and
guards entrance into literate domains’ (Fingeret, 1990: 35). Finally
it may be the centralised state, through its institutions (Gee, 1990).
The antidote to such undesirable aspects of literacy is often
seen in ‘critical literacy’ (cf. Baynam, 1995: 205). Mikulecky cites
Kretovics (1985) to the effect that ‘the teaching of a “critical lit­
eracy” can enable teachers and students to formulate strategies to
change the form, content, and social relations of education with
an interest in freedom and democracy’ (Mikulecky, 1990: 26). This
approach, which is closely related to ‘critical discourse analysis’
(Fairclough, 1995) and ‘critical reading’ (Wallace, 1992a), stresses
not just decoding a text and processing the meaning, but using
the text for the reader’s purposes, analysing its presuppositions,
and how it works on its readers (cf. Freebody and Luke, 1990).
Street (1995) comments that ‘teaching awareness o f . . . the ways
in which literacy practices are sites of ideological contest, is itself
already a challenge to the dominant autonomous model that dis­
guises such processes’ (p. 137).
The motto of the ‘critical’ group would seem to be 'It's best to be
bolshie’ - a position for which we have a certain amount of sym­
pathy. One can, however, give only limited support. It seems to us
that Fairclough’s argument that the discourse analyst must be
aware of the ideological presuppositions contained in a discourse
is essentially valid; we would query, however, whether the aims of
analyst and reader are necessarily identical. Then it seems a little
rich to insist that literacy is primarily a means of socialisation and
acculturation, only to add that the reader should be prepared to
analyse and possibly reject the power structures inherent in the
society or culture. Again, one cannot and should not, in our view,
always be a critical reader: the first year biology student reading
the textbook recommended by the departm ent might do well to
adopt a submissive attitude to its contents (see Chapter 2).3 The
nuclear worker reading safety regulations might be wise to restrict
any criticism. And if this is true for teaching, it is much more true
for testing: the student who rebels against the power structure
behind the test is unlikely to obtain a good grade.
There are further problems. Street remarks that ‘the student is
learning cultural models of identity and personhood, not just how
to decode script or to write a particular hand’ (1995: 140). We are
dubious as to whether the adult students in the L2 classroom are
8 Reading in a Second Language
learning ‘cultural models of identity’, and we are certain that in
most instances they are not learning ‘models of personhood’. They
come into the class bringing their own models of culture, and are
often very capable of identifying ideologies in discourses which
conflict with their own cultural values.
There is also the problem of the teacher or test writer’s own
ideological position. Street remarks that the Freireian approach,
whereby literacy is accompanied by revolt,
is vulnerable to such culturally blind manipulation by activists im­
bued with ideological fervour and believing so strongly that they
are ‘empowering’ ‘ignorant’ peasants that they fail to see their own
cultural and political domination. (p. 138)
There is, it seems to us, a danger that L2 students in the read­
ing class may be subjected to the ideology of teachers, whether
they be radical feminists, environmentalists, vegetarians, atheists,
Piets against Fascism or whatever. In fact, teachers would help
more to empower students if they were prepared to ‘back off’,
stopped imposing ‘correct’ interpretations of texts, let students
choose their own texts and set their own agenda.

Cognitive v Social
According to Bernhardt (1991b: 6), ‘taking a cognitive perspective
means examining the reading process as an intrapersonal problem­
solving task that takes place within the brain’s knowledge struc­
tures’. She notes (p. 8) that the critical element in any cognitive
view of reading is that it is an individual act. As a social process,
she cites Bloome and Greene (1984): ‘reading is used to establish,
structure, and maintain social relationships between and among
peoples’.
Bernhardt doesn’t specify the social aspects except to point out
that the same individual and the same text can vary:
the processing of text can be viewed only within a unique cultural
context. . . there are basically no generic or generalized readers or
reading behaviours . . . there are multiple readers within one person
. . . multiple ‘texts’ within a text. (pp. 10-11)
There are, however, many more factors than this in the social
approach to reading.
No text on reading can ignore the social aspects of the activity.
Reading, as the studies of literacy mentioned above have made
 Introduction 9
clear, is a social activity, related always to particular contexts. How­
ever, there is no doubt that in comparison to a book like Wallace
(1992a) we put more emphasis on the cognitive side. There is
more than one reason for this.
1. We consider that the cognitive aspect is primary. Reading with­
out social aspects might be an odd idea, but something like it
exists in the experiments of cognitive psychologists, some of
whom are mentioned in Chapter 2. Reading without cognitive
activity, on the other hand, is simply an impossibility.
2. Wallace remarks, quite correctly, that ‘as readers we are fre­
quently addressed in our social roles rather than our personal
and individual ones’ (1992a: 18) and gives as an example the
reader of an advertisement aimed at a social group. We would
claim, however, that while this is clearly relevant to an analyst
of the text, and certainly relevant to the production of the text,
it is not necessarily relevant to readers who probably see them­
selves as reading the text as individuals. To this extent, read­
ing is always an individual activity. We should not, of course,
fall into the trap of equating ‘individual’ with ‘cognitive’; the
individuals bring their own societies with them. Nevertheless,
there is an element of truth in it - see Bernhardt above.
3. Wallace claims that: ‘Classrooms are themselves communities
with their own uses of literacy and ascribed roles for teachers
and learners.’ We agree, but are conscious that the L2 reading
class is often a very loose, transitory community, more akin in
some ways to the population of an airport lounge. The mem­
bers have come from different communities and are intending
to join different communities. The communities that they plan
to join must remain in our consciousness. Again, we see classes
as collections of individuals. And if this is true of classrooms,
how much more is it true of our other prototypical situation,
the examination hall?
4. Finally, in Chapter 4, where we review teaching methods, we
try to focus on those areas where empirical evidence exists in
support of particular practices. It is our impression that more
of such evidence at present exists for those methodologies which
have a cognitive, rather than a social focus.
We do, however, try to take the social aspect of reading explicitly
into account in our handling of comprehension in Chapter 2,
and in the importance we ascribe to authentic tasks and texts in
10 Reading in a Second Language
Chapters 3 and 4. A fuller treatment of the social approach to
reading, including activities for the classroom, can be found in
Wallace (1992a, 1996).

Organisation of the book

Reading is surprisingly complex to define. Its precise relation­
ship with language in general, and with listening in particular, are
a cause of controversy or misunderstanding. So we begin with
Chapter 1, ‘Preliminaries’, in which we attempt to deal with these
difficulties.
Chapter 2 is concerned with theoretical issues concerning read­
ing. Here we discuss models of the reading process, including
process models, which attempt to model the actual cognitive pro­
cesses involved in reading, and componential models, which are
content to discuss the likely components of the process. In addi­
tion, we look at the topics of comprehension, reading skills and
strategies.
In Chapters 3 and 4 we proceed to the more applied sections
of the book, and examine the testing and teaching of reading. In
our accounts of these areas, we have three main aims:
1. To provide a reasonably comprehensive account of the develop­
ment and current state of the art. We stress the importance of
considering expeditious reading strategies (selective, quick and
efficient reading activities such as skimming, search reading
and scanning) as well as the more conventional careful reading
activities.
2. To relate this as far as possible to the theoretical background
already discussed. We see this as essentially a two-way issue: one
can examine, for example, the justification for testing or teach­
ing practice in terms of theory; conversely, one can examine
whether the theory needs to be expanded to take account of
some apparendy successful or popular practice. Both aspects lead
to suggestions for new forms of test and of teaching materials.
3. To provide the reader with guidelines for selecting testing
or teaching activities and materials that have some empirical
support.
Finally, in the light of our wish to balance theory and practice, we
conclude with a chapter on reading research (Chapter 5). Our
 Introduction 11
aim here is to acquaint the reader with some of the methodology
and constraints operating in the research area, by suggesting ways
in which readers can construct and carry out their own research
in the broad areas of theory, testing and teaching.

Notes
1. Rendell, R. (1977) A Judgement in Stone.
2. It is perhaps worth noting that, in our own society, highly literate
adults will admit without embarrassment to not being able to ‘read’
music.
3. Wallace (1996) accepts this, suggesting that critical reading is one
strategy available to the reader.
This page intentionally left blank
 13
1

Preliminaries

What is reading?

We all know what reading is. And many of us have suffered, at

some time or another, from the type of bore who stops any argu­
ment or discussion with ‘Ah, it depends on what you mean by . . .’.
So it is with some reluctance that we begin this part with an
attempt to define reading, to say what we mean by the term. Our
excuse is that people do use the term in different ways, and that
while this may be permissible when everyone is conscious of the
differences, on other occasions it can cause real confusion and
difficulty.
Like Bernhardt (1991b) we approach the problem via diction­
ary definitions. The Concise Oxford Dictionary gives thirteen entries
for the word read, of which twelve refer to the verb. Below we give,
in an abbreviated form, the first three:
1. interpret mentally, declare interpretation or coming develop­
m ent of (read dream, riddle, omen, men’s hearts or thoughts orfaces);
2. (to be able to) convert into the intended words or meaning
written or printed or other symbols or things expressed by
such symbols . . . ; reads or can read, hieroglyphics, shorthand, Braille,
Morse, music. . . ;
3. reproduce mentally or . . . vocally, while following their sym­
bols, with eyes or fingers, the words of (author, book, letter,
etc.); read the letter over. . .
We should say right away that none of these, or in fact any of the
entries we mentioned, matches precisely our own definition of
reading, which goes some way towards excusing our spending
14 Reading in a Second Language
time on such definitions. That apart, for our purposes here, we
can from now on reject definition 1 above. It is obviously a legit­
imate use in the English language of the verb read, but we are
using the term in relationship to written texts, and so will not
include uses like reading dreams in our definition. It is worth point­
ing out, however, that this first entry seems to focus on the use of
the verb to mean something like interpret, which is an important
aspect of text reading. Entry 3 looks as if it is largely confined to
the activity often referred to as decoding,, and will be discussed
below. Entry 2 looks the most promising, but even here there is a
problem. For reasons which will be discussed shortly, we will con­
sider it appropriate to refer to reading hieroglyphics, or shorthand,
Braille or Morse. However, we shall not use the term to include
the reading of music, or, for example, maps, or mathematical
symbols, etc. The reason for this apparently arbitrary decision
rests on a prior definition of what is meant by writing.
In his book on writing systems, which will be referred to at
several points in this chapter, Sampson (1985) finds it necessary
to devote several pages to defining ‘writing’, just as we are having
some difficulty defining ‘reading’. One of the main difficulties
facing Sampson is the fact that there exist written or printed sym­
bols which would not normally be referred to as ‘writing’. The
international road sign for ‘bridges ahead’ is one example, as are
the symbols distinguishing m en’s from women’s toilets, and so on.
Sampson distinguishes between glottographic symbols, which repres­
ent language, and semasiographic symbols, which do not. Thus the
printed word three is glottographic: it actually represents the Eng­
lish word pronounced [0ri:]. The figure 3, on the other hand, is
semasiographic. While it can be ‘translated’ by the English word
three, it can also represent trois, or drei, or tatu, or words all mean­
ing ‘three’ in any number of languages. Mathematical symbols,
in fact, represent a very sophisticated semasiographic system, as
does musical notation. Here we follow Sampson in not counting
semasiographic systems as writing; hence the interpreting of mes­
sages in such systems will be considered as being outside the scope
of reading proper. Reading, as used here, will mean dealing with
language messages in written or printed form. That is why, when dis­
cussing the COD entries above, we accepted Braille, hieroglyphics
and Morse as messages suitable for reading, but excluded music.
This restriction on the term may strike some of our readers
as unduly narrow. From a practical point of view, for example,
 Preliminaries 15
written messages often include semasiographic symbols, perhaps
mathematical figures, or maps, and so on. ‘Communicative’ read­
ing material often includes maps and time-tables, which the student
reader is expected to process along with any written material.
Surely it seems pedantic to say that someone reads a message until
they come across a figure, at which point they stop reading and
begin another activity? And if we don’t read bus time-tables, what
do we do with them?
However, there are good reasons for the restriction. Perhaps
the most important, for our purposes, is that reading involves
processing language messages, hence knowledge of language. We
do not want to become embroiled in the controversy as to whether
language competence is separate from, and, in its acquisition, quite
different from other areas of competence such as mathematical
knowledge. We would, however, consider it fairly uncontroversial
that the process of reading, certainly above the level of decoding,
makes demands on linguistic competence which are more imme­
diate and more pervasive than appeals to other competence areas.
Secondly, in the classroom, a distinction between reading and
other symbol-processing activities is made, even though it may not
be made consciously. LI primary school teachers, or EFL teachers,
may incorporate semasiographic material in reading lessons, but
they are not likely to think of themselves as teaching students how
to read, for exam ple, maps. The presence of maps, etc., in reading
lessons is incidental to the main teaching focus, which is on how
to read language messages. Neither of the present authors can
read music, but neither found themselves in a remedial reading
class, on that account at least.
Finally, our use of the term reading seems to agree with the
non-technical, everyday use of the word. Faced with a questionnaire
asking about hobbies, the person who gets a thrill out of perusing
mathematical formulae is not likely, with that in mind, to respond
with ‘reading’. That answer is restricted to those who derive pleas­
ure from newspapers, novels, history books, etc.
Before leaving this point, we need to clarify what we mean by
language. After all, mathematics is often described as being a kind
of language; but so is formal logic, and we often talk of computer
languages. What we mean, here, however, by the term are languages
such as French, Swahili, Mandarin and so on. Sampson refers to
them as ‘oral languages’, but immediately has to qualify this by
saying that an ‘oral’ language is not necessarily ‘spoken’, which
16 Reading in a Second Language
seems rather confusing. We might refer to ‘natural’ languages,
except that this term tends to be used to contrast languages such
as French, Japanese, etc., with ‘artificial’ languages such as Esper­
anto. From our point of view, Esperanto is a language, just as
English is, and can be read as such. Perhaps at this point we just
have to rely on our readers’ knowledge of the world, and assert
that we are using the term ‘language’ to refer to French, Esperanto,
etc., linguistic languages as it were, and not to mathematical or
computer languages.
The fact that the information conveyed by a written message
is encoded in language obviously has major pedagogical conse­
quences. At the moment, we simply point out that for people
learning to read in their native language, it is clearly relevant that
they already know the language, or at least a significant part of it.
As opposed to the learning of mathematics, that is, the learner
does not need to learn both the conceptual system and the con­
ventions of encoding it. And the fact that LI learners of reading
have this advantage is not affected by whether we take the pedago­
gical decision that ‘all’ they need to learn is to decode, or whether
we decide to make use of previously acquired language know­
ledge in the process of learning to read.
As far as the L2 situation is concerned, the fact that reading
involves processing language should perhaps make us a little scep­
tical of dichotomies, found in the literature (e.g. Alderson, 1984),
as to whether L2 reading difficulties are ‘reading’ or ‘language’
problems. It may be difficult to distinguish the two. This question
is addressed in more detail in Chapter 5.

Reading and language

The texts which we read, then, are language texts. So reading is
involved with language texts. But what is the relationship between
reading and language? Roughly speaking, there are two answers
to this in the literature. The first defines reading as decoding, as
Perfetti (1985) glosses it, ‘the skill of transforming printed words
into spoken words’. This decoding definition offers some good
arguments. It delineates a restricted performance and allows a
restricted set of processes to be examined. However, as Perfetti
points out, it has limited popularity partly because it has limited
application to the demands of actual reading. Moreover, it is not
really feasible to view decoding as the initial process which is over
 Preliminaries 17
by the time other cognitive linguistic processes begin. In one of
the best-known papers on reading, Goodman (1967) argues that
syntactic, semantic and pragmatic knowledge are involved in the
decoding process. And this is the position of those advocating an
interactive model of the reading process, discussed in Chapter 2.
Finally, a practical, commonsensical view of certain activities
should persuade us that decoding cannot be equated with read­
ing. Given the regularity of sound/letter correspondences in the
spelling of Spanish, for example, it would be possible, if fairly
pointless, to teach an English speaker who knows no Spanish to
read aloud from a Spanish text with reasonably good pronuncia­
tion, but no comprehension. Few sensible people, however, would
describe such an activity - sometimes referred to as ‘barking at
print’ - as reading, even though activities resembling it are carried
out under the guise of ‘teaching reading’ in many parts of the
world. (In a seminar conducted by one of the authors, a British
teacher remarked thoughtfully, ‘I have a boy in my class who’s a
really good reader. The problem is he doesn’t understand anything
he reads.’) We conclude that decoding must be an important part
but not the whole of the reading process.
The second answer defines reading as the whole parcel of cog­
nitive activities carried out by the reader in contact with a text.
Thus Nuttall (1982), having considered definitions of reading in
terms of reading aloud, or decoding, settles for the extraction of
meaning from written messages. Similarly, Widdowson (1979) has
defined reading as ‘the process of getting linguistic information
via print’. And Perfetti, as the alternative to a definition in terms of
decoding, suggests that reading can be considered as think­
ing guided by print, with reading ability as skill at comprehension
of text.
Perfetti’s second definition, that reading is thinking guided by
print, also has problems. Those who have trouble with reasoning
or fail to learn from reading will be said to have reading prob­
lems. Clearly the difficulty here is one of establishing boundary
lines, and being able to say, ‘Up to here we have reading prob­
lems, beyond . . .’. The problem is present in an ambiguity in
Widdowson’s definition. One could say that ‘linguistic informa­
tion’ was restricted to information about, say, syntax, morphology
and lexis. Perhaps we can emend the definition to read ‘informa­
tion carried by linguistic messages via print’. The open-ended
nature of what some people mean by reading is what Fries (1963)
18 Reading in a Second Language
was commenting on, when he remarked, of ‘broadly based’ edu­
cational approaches to reading:
But we should certainly confuse the issue if we insist that this use of
reading in stimulating and cultivating the techniques of thinking,
evaluating and so on, constitutes the reading process.
There is a real difficulty here, one that has resurfaced in recent
years in the debate about ‘reading skills/strategies’ as opposed to
‘language skills/strategies’. Here, however, in spite of the difficul­
ties associated with defining reading in such an open-ended way,
we consider reading to be a language activity, involving at some
time or another all the cognitive processes related to language
performance. Thus we consider that any valid account of the read­
ing process must consider such cognitive aspects as reading strat­
egies, inferencing, memory, relating text to background knowledge,
as well as decoding, and obvious language aspects as syntax and
lexical knowledge. In this respect, at least, we are in agreement
with the proponents of ‘critical reading’, discussed briefly in the
Introduction.
This would seem, in practice, to be the position most favoured
by teachers and materials writers. Many reading textbooks, for
example, contain practice in such cognitive activities as drawing
conclusions on the evidence of written messages, making infer­
ences, evaluating texts in terms of truth, persuasiveness, beauty,
etc. It would be hard to find an EFL reading textbook which
restricted activities to decoding. In fact, it is quite difficult to find
one that includes any explicit material on decoding at all.
The same is true in the research area. In an editorial in the
Reading Research Quarterly in 1980,1 under the heading ‘Why Com­
prehension?’, the editors comment that an emphasis in the 1960s
on decoding and early reading attracted much less attention in
the 1970s, being replaced by an emphasis on comprehension.
This view can be substantiated by statistics: Figure 1.1 based on
data from ERIC,2 shows the number of articles and other publica­
tions, published between 1966 and 1996, which mention the term
reading in their title, or in ERIC’s index or abstract. The graph
shows a rapid increase in such papers in the latter part of the 1970s,
presumably reflecting a major increase in interest in reading. (It
also, of course, shows a gradual decline from that peak, which we
are unable to account for, although the very rapid decline after
1994 may have something to do with methods of data collection.)
 3500 -
- Reading
3000 -

2500 -

2000 -

1500 -

1000 -

500 -

0 -

1966 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
Preliminaries

Figure 1.1 Number of articles and other publications published between 1966 and 1996 that mention ‘reading’ in
19

their title or in ERIC’s index or abstract (based on data from ERIC).

20 Reading in a Second Language
The rapid rise after 1966 is paralleled by a rise in interest in
comprehension, as shown in Figure 1.2. It is likely, as is suggested
by the editors, that with the move towards research in ‘compre­
hension’ - which has to be taken as including all the cognitive
processes thought to be involved in reading - reading became of
much wider interest than it had been when decoding was the
main interest, becoming attractive to cognitive psychologists and
psycholinguists, as well as educationalists.
All this having been said, there are occasions when we may wish
to take reading as meaning decoding. When, for example, we say
that someone can speak, but not read X language, then we by and
large have in mind an inability to decode the script. With LI
learners in mind, Wardhaugh (1969) has justifiably criticised edu­
cators who confuse knowledge of the script with knowledge of the
language, and make remarks to the effect that, for example, an LI
child does not know the difference between the short and long
vowels in English, when in fact what they mean is that the learner
does not yet know their spelling in the written language.3
This need not be interpreted, however, as meaning that reading
is separate from language; rather that, in teaching reading, the
teacher should keep in mind that a major component involved in
the reading process is already available to the native-speaker and
does not need to be taught (though the knowledge may need to
be activated).
For the purpose of finalising our definition, we shall return to
the Widdowson quotation above, that reading is ‘the process of
getting linguistic information via print’. While broadly agreeing
with this, we want to modify it somewhat. First of all, by talking
about ‘getting information’, Widdowson appears to imply that this
is a fairly one-way process from writer or text to reader: ‘Here is
some information which I am passing on to you. Now take it.’
Widdowson would not, of course, want to appear to be implying
this, since he has been an important advocate of the view that the
reader interprets and contributes to incoming messages. In order
to avoid jargon such as ‘interacting with written data’, we shall settle
for ‘receiving and interpreting information’. Then there is the prob­
lem, already mentioned, of deciding on the limits of the informa­
tion received. After all, some writers would consider ‘linguistic
information’ to be limited to such items as ‘X is the subject of the
main clause’. We shall say that the information can be of any kind,
but that it is encoded in language. Thus our definition has become:
 700 -|
Comprehension
600 -

500 -

400 -

300 -

200 -

100 -

£|
0 -
S'a
1966 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 S.
S
Figure 1.2 Number of articles and other publications published between 1966 and 1996 that mention ‘comprehension’ NO
in their title or in ERIC’s index or abstract (based on data from ERIC). l- J
22 Reading in a Second Language
Reading is the process of receiving and interpreting information encoded in
language form via the medium of print.
This may not be very neat but it suits our purposes.

Reading and listening

When, in an earlier section, we discussed decoding, we deliber­
ately omitted one question: If reading consists of decoding, then
what do we call the other linguistic processes which presumably
follow the decoding stage? From one point of view, the answer to
this would be ‘Listening’. In this view, reading consists of decod­
ing from written symbols to sounds; after that, presumably, ‘normal’
listening processes come into play, readers listening to themselves
decoding. It is noticeable that Perfetti glosses decoding as ‘trans­
forming written words into spoken words’. Gough (1972) is another
who seems to see the reading process as resulting in speech (see
Chapter 2). (The COD definition 3 given above, however, which
clearly refers to decoding, refers either to reproducing words men­
tally or orally, thus allowing for a non-spoken decoding.) Thus read­
ing is seen as parasitic on listening - speech is ‘the primary linguistic
activity; reading a secondary, language-based skill’ (Mattingly and
Kavanagh, 1972). ‘Language is speech, writing a secondary system
- a mnemonic device reminding us of word sounds, intonation,
rhythm, etc.’ (Lefevre, 1964).
The alternative view, represented by, for example, Sticht (1972),
is that speech and writing are alternative manifestations of lan­
guage. Sticht cites Huey (1968):
There do not exist two kinds of language comprehension - only
one holistic ability to comprehend by language.
And in spite of the fact that, as pointed out earlier, Fries wanted
to reserve the term ‘reading’ for the decoding process, he appears
to take a similar view:
Learning to read is a process of transfer from auditory signals
to visual ones. The language signals are the same for both talking
and reading.
It should be noted that by this remark, Fries does not mean that
the visual signals ‘represent’ auditory ones, rather that the same
set of contrasts is made by both sets of signals.
 Preliminaries 23
It is interesting to speculate on what hypotheses follow on from
these two contrasting positions. If one adopts View 1, that reading
is parasitic on listening, then presumably:
(a) Deaf children would not learn to read. Mattingly and Kavanagh
(1972) say they learn very slowly. Rayner and Pollatsek (1989)
agree on the whole, but admit that some deaf children do
learn to read well. Holcomb and Peyton (1992) claim that
most deaf children read at third or fourth grade level. Padden
(1990) concedes that deaf children have difficulty learning to
read and write, but claims that deaf children from middle-
class, educated deaf families perform ‘consistently well’ in read­
ing and writing. There would thus appear to be a marked social
factor in success or failure to read. Attitudes to language also
seem to be important. Ramsey (1990) points out that deaf adults
‘greatly value ASL (American Sign Language) as a marker of
group identity and share a certain proprietary feeling about
the mode of signing itself. We have already mentioned the
importance many writers attach to the role of reading in
socialisation. Rayner and Pollatsek (1989: 210) note that ‘sec­
ond generation profoundly deaf readers seem to consult their
native language (ASL) when reading English’. Cangiano (1992)
is one of a num ber of writers who considers the problem of
deaf children learning to read English as an aspect of L2
reading. Clearly the whole area is complex, involving factors
in the home, attitudes to the LI and L2, and aspects of L2
acquisition. It seems clear to us that reports from the area
do not support the view that reading is inevitably guided by
phonological factors.
(b) At all stages of language learning, both for LI and L2 speakers,
listening would have to precede reading. Menyuk (1984), for
example, appears to believe that if a child cannot understand
a construction in oral form, then that child will not be able
to comprehend it in written form. In EFL contexts, reading
instruction would invariably have to follow a course of oral
language learning.
(c) All other things being equal, listening would presumably be
faster than reading, since reading speed would equal listening
speed plus decoding time.
If, on the other hand, one adopts the view that listening and
reading are parallel activities, then:
24 Reading in a Second Language
(a) Deaf children might be slow to learn, lacking as they do import­
ant sources of data. But once they had learned, their reading
might equal or surpass that of non-deaf children. (This seems
sometimes to be the case, though rarely; see above.)
(b) For the LI learner, listening, and learning through listening,
would normally precede reading in the early stages, but after
reading began, a great deal of language learning - lexis, syn­
tax, rhetorical organisation - would be accomplished via read­
ing. This fits with the commonly expressed view that we acquire
a large part of our adult lexicon via the written language. We
would suggest that we also acquire much of our knowledge of
complex grammatical structure via the written language.
(c) An L2 reading course would not necessarily need to be pre­
ceded by an oral course. A ‘reading to learn language’ stage
would precede a ‘reading to learn’ stage.
(d) It is doubtful whether any hypothesis can be formed about the
relative speed of listening and reading (although see below).
It ought to be obvious that this question is not of purely theoret­
ical interest, but is of real practical concern to those engaged in,
for example, designing language courses. For this reason, we shall
examine the problem further in the following sections.

The primacy of spoken language

There seems little doubt that the position of those who consider
reading a secondary process has in the past been reinforced by
the assertion, virtually an axiom in linguistics, that spoken lan­
guage is primary. Spoken language, the textbooks tell us, is prim­
ary both phylogenetically and ontogenetically, i.e. speech preceded
writing in the history of the human species, and for the normal
individual child, speech also comes first. This is unquestionably
true, and no one nowadays would dispute it. But as Sampson
points out, linguists adopted this position in a programmatic way;
that is, speech was the aspect of language that linguists were sup­
posed to study. And one of the reasons for adopting the position
was as a reaction against a previously dominant view that written
language represented the correct form, as opposed to debased, or
‘careless’ speech forms. As part of this reaction, written language
was held to be derivative, and not worthy of study.
 Preliminaries 25
From a strictly theoretical point of view, this attitude is at least
unnecessary. De Saussure drew a distinction in language between
form and substance. By form he meant the whole structure of rela­
tionships, both contrasts and equivalences, inside the language;
one of the meanings of substance is the medium in which the form
is realised. Just as, in de Saussure’s well-known analogy with chess,
the game can be played with pieces composed of any num ber of
materials or shapes, so language can be ‘realised’ by different
media, in our case by either sounds (spoken language), or shapes
(written language). The implication of this argument is that writ­
ten language can be seen as a realisation of a language parallel to
the spoken form. For de Saussure, language was basically form;
the substance in which the form was realised was immaterial. Based
on this distinction, Lyons (1968: 60-1) argues that:
When we say that [t] is in correspondence with t, [e] with e, and in
general that a particular sound is in correspondence with a particu­
lar letter, and vice versa,, we can interpret this to mean that neither
the sounds nor the letters are primary, but that they are both altern­
ative realizations of the same formal units, which of themselves are
quite abstract elements, independent of the substance in which
they are realized.
Making the same point, Sampson (1980) remarks:
After all, English is still English whether we realise it as spoken
sounds or as ink on paper.
W hether linguists did, in fact, obey the injunction to restrict
themselves to spoken language is doubtful. Brown and Yule (1983)
remark that the language in descriptive linguistic grammars is
often characteristic of the written form. And Lyons (1968) remarks
that in a highly literate society, linguists find it difficult to view
spoken language objectively.
Apart from this, however, the long-standing linguistic prejudice
against written language has recently been eroded. Sampson is
one linguist who considers that written language is a proper area
of study for linguistics. For him the written and spoken languages
are two ‘closely related dialects’, and the influence of one dialect
on the other is not one-way. In an article provocatively titled ‘The
primacy of writing’, Householder (1971) claims that the written
language ‘has probably been the greatest single cause of phono­
logical change in modern English, both British and American’. It
26 Reading in a Second Language
is likely that a considerable amount of the vocabulary, syntax and
knowledge of rhetorical structure of a mature native speaker is
learned via the written language.
If this was simply a dispute within linguistics, then it would not
matter to us here. The study of reading may borrow from lin­
guistics, but does not belong to the area. The linguistic argument
as to relative primacy is in one sense irrelevant. From the point of
view of a language user at a particular point in time, it is ridicu­
lous to argue that speech is always primary. If we meet a neigh­
bour and exchange greetings, we choose speech; anything else
would be strange. On the other hand, when we set out to com­
plete an income tax return, we are more or less obliged to resort
to the written language - a failure to complete the form on the
basis that written language was secondary would not be treated
with much sympathy.
For the purposes of this book, we assume that written and
spoken language are parallel realisations of language. This is not
to deny that, at least for the native speaker, learning to read will
normally involve a transfer from the spoken language to the writ­
ten language. But we are not bound to believe that the written lan­
guage, and hence reading, remains parasitical on the spoken form.
We have not, however, answered the question as to whether the
reader, when performing, is dependent on spoken language. To
some extent, linguists are not well qualified to decide on this
question. Linguistics describes language systems. If the systems
are different, then linguists are well qualified to describe the dif­
ference. However, reading is a process, involving mental activity.
Linguists are not professionally equipped to make judgements
about such processes, although they are as entitled as anyone else
to venture opinions. Thus the view, derived from linguistic theory,
that a written language could exist parallel to a spoken language
as a separate realisation of the same underlying language system
simply allows us to say that such a situation could happen, that is,
that users could use written language independently of the oral
language. It does not entitle us to claim that users do actually do
this, or even that it is practically possible. Thus we are left with the
possibility that readers could be ‘translating’ into the spoken lan­
guage as they read, as those who push the claims of ‘vocalisation’
have asserted.
The debate has taken place in terms of a suggested opposition
between a direct and a phonological route to word recognition
 Preliminaries 27
(see Chapter 2). The reader using a direct route is envisaged as
going straight from written word to meaning, without an interven­
ing ‘spoken’ stage. When using a phonological route, the reader
is seen as going from the written word to its spoken pronuncia­
tion, using phonem e/graphem e correspondences, and finally to
meaning. In the pedagogical literature, the distinction can be
seen in the difference between the look and say method of initial
reading and phonic approaches.
It might at first seem easy to decide whether readers use direct
or phonological routes. One may argue that if the readers can be
shown to vocalise while reading, then they are using the phono­
logical route. However, things, as often, are not so simple. Rayner
and Pollatsek (1989) distinguish between ‘subvocalization’ (mus­
cular activity in the speech tract) and ‘inner speech’ (the ‘voice’
we hear in our heads while reading). They conclude from the
experimental data that subvocalisation is ‘a normal part of natural
silent reading’ (p. 192). However they point out that its function
is obscure. They are more interested in the phenom enon of inner
speech, but this is a more mysterious entity, and its function is
controversial. One might argue that since some L2 readers may
not know how to pronounce certain words, they are precluded
from phonological access. However, a little thought should show
that this is an unjustified conclusion as they may be using a non­
native pronunciation.4 We return to this topic in rather more
detail in Chapter 2.
Some light may be thrown on the problem by an examination
of different writing systems, both alphabetic and other, which we
touch on next. The discussion should be of value in two other
respects: statements about reading by European authors often imply
the use of an alphabetic system of writing. We need to be warned
against such Eurocentrism. Also, the discussion should remind us
of a possibly important factor to be taken into account when we
discuss the diversity of L2 readers.

Writing systems
The account given here is based on Sampson (1985), and we use
his terminology. In particular, we use the term graph to refer to
any sign in a writing system, e.g. for alphabetic letters and for
Chinese ‘characters’.
28 Reading in a Second Language
Sampson divides scripts into two major types. In logographic
scripts, of which his main example is the Chinese writing system, a
graph relates to a meaning unit, either a word or morpheme (in
Chinese, according to Sampson, the distinction between word and
morpheme is not clear, and symbols largely refer to morphem es).
It is as if, in English a single symbol, say C, represented the w ord/
morpheme cat, and another symbol, say M, represented the w ord/
morpheme mat.5
Phonographic scripts, on the other hand, rely on analysing lan­
guage at the sound level. Included under phonographic scripts
are syllabic, consonantal, alphabetic and featural scripts. In syllabic
scripts, a graph represents a syllable. In English, mat would be
spelled with one symbol, but matted would require two. Teachers
who instruct children that the sequence cat ‘spells /k a /, /a /, /to /
’ could be said to be converting English spelling into a syllabic
script. In a system of astonishing complexity, Japanese uses two
syllabaries, hiragana and katakana, to supplement a logographic
script, kanji, derived from Chinese writing.
In consonantal scripts, of which Arabic and Hebrew are two
important modern examples, the syllabic sound unit is segmented,
but only consonants are systematically represented. It is as if in
English, mftkl spelled emphatically. The system suits Semitic lan­
guages such as Arabic, where vowels are largely predictable from
the grammar, but would give clear problems in English.
In alphabetic systems, such as the ones familiar to readers of
English and other European languages, both vowels and con­
sonants are represented. Finally, in what Sampson terms featural
systems, the symbols represent phonological features. In English,
for example, the symbols k and g represent velar plosives distin­
guished phonologically by the fact that the first is voiceless and
the second voiced. We could move English spelling in the direc­
tion of a featural system if, say, we spelled the sounds / k / and / g /
with one graph, but represented the presence of voicing by un­
derlining. Thus / k / would be represented by the graph k, and
/ g / by the graph k. Similarly t could represent the phoneme / 1/
and t represent /d /. Sampson gives two examples of featural sys­
tems, Pitman’s shorthand, and H an’gul, devised, in a remarkable
feat of linguistic analysis, by a fifteenth-century Korean, and used
as a national script by both present Korean republics.
For the purposes of the argument here, the most significant
point is that logographic scripts do not rely on an analysis of the
 Preliminaries 29
sounds of words. Thus, according to Sampson, Chinese script was
composed originally of graphs which each represented a w ord/
morpheme. At this stage, there was no generalised relationship
between graph and sound. The system was extended according to
the rebus principle, whereby a graph for a word could also be used
for another word with the same or similar pronunciation. In Eng­
lish, using this system, a graph representing ‘eye’ could also be
used for the pronoun ‘I’. This does introduce a phonetic element
into the system, though not in any systematic way. The rebus
principle, however, has limitations. Chinese contains, by the stand­
ards of European languages, an astonishing num ber of homo­
phones (words with the same pronunciation) .6 It is as if English
had, say, twelve words all pronounced the same as ‘eye’. In order
to reduce ambiguity, the system was again extended to include
complex graphs. These are composed of at least two elements,
one of which Sampson calls a phonetic, and the other a signific
(other writers use the terms classifier, or radical, for signific). The
phonetic hints at pronunciation, the signific at semantic value. If
we try to construct an English analogy, we might think of a graph,
say M, representing the word meet Then, by the rebus principle,
this can be extended to represent both meat and the archaic adject­
ive mete, which are all homophones. Our graph now represents any
of three homophones. We can reduce ambiguity by producing a
complex graph. By incorporating a signific f, meaning, say, food,
we can produce a complex graph Mf which is interpreted as mean­
ing ‘sounds like meet but has to do with food’. Similarly, having
already a graph representing the word flower, say F, we can write Ff
to indicate ‘sounds like flower but has to do with food’, i.e. flour.
So far, it might look as if Chinese graphs were essentially phono­
graphic, but Sampson denies this on two grounds. Firstly, the
pronunciation of Chinese has changed radically since the ‘spell­
ing’ was stabilised; secondly, the sound correspondence between
the original phonetic and the secondary complex graph was not
always close. To take another example, it is as if in English we had
M (= meet) plus a signific c indicating clothing, giving us the graph
Mc for mitt. The result is, as Sampson (1985: 157) says,
A Chinese-speaker who learns to read and write essentially has
to learn the graphs case by case; both signifies and phonetics will
give him many hints and clues to help him remember, but the
information they supply is far too patchy and unreliable to enable
him to predict what the graph for a given spoken word will be,
30 Reading in a Second Language
or even which spoken word will correspond to a graph that he
encounters for the first time.
It has to be said that Sampson’s account has been vehemently
criticised by DeFrancis (1984) and by Unger and DeFrancis (1995).
They claim that 66 per cent of the graphs used in Mandarin
contain enough phonetic indicators to enable a reader to make a
good guess at the syllable (this figure being based on an analysis
carried out in 1942). They thus claim that Chinese script is basic­
ally a rather odd syllabary. They back this up with the claim that
learning thousands of logograms would be equivalent to learning
thousands of telephone numbers.
To some extent the argument is academic; Sampson does not
argue that the system is purely logographic, merely that it is not
primarily phonographic. We are, however, rather dubious about
some of DeFrancis and Unger’s arguments. Any empirical experi­
m ent devised to test the phonological readability of graphs would
be bafflingly difficult to set up. The subjects would have to be
literate readers of Chinese, for which a knowledge of 5000 graphs
is often suggested as minimal. On the other hand, the graphs they
were shown would have to be unfamiliar to them, otherwise the
subjects’ evidence would be hopelessly unreliable. Therefore we
are justified in doubting the statistics cited by DeFrancis and
Unger.7 As opposed to telephone numbers, Chinese graphs con­
tain a certain amount of iconic information (i.e. some of them
are recognisable in picture terms), and semantic information (the
signifies). Moreover, even if the phonetic information were as
important in Chinese reading as Unger and DeFrancis say, it is
difficult to see how it would remain important when the same
graphs are used to write Japanese, a totally different language.
Sampson’s main response to earlier criticism by DeFrancis
(Sampson, 1994) is that DeFrancis fails to distinguish between the
historical origin of a linguistic system and the nature of the system
as it exists at the present.
However, we are not in a position to make a real judgem ent on
the issue. It is one of these areas in which empirical evidence is
crucial yet seems difficult to collect. We think that there is enough
force in Sampson’s case to suggest that Chinese and Japanese
readers - a large section of the world’s literate population - are
likely on many occasions to make use of direct access to lexical
items. This at least should make us cautious about making Euro­
 Preliminaries 31
centric generalisations on the nature of reading, based on our
experience with an alphabet. As Sampson says (1985: 146):
The European idea that from a knowledge of the pronunciation of
a word one should be able to make at least a good guess at how to
write it would seem bizarre to a Chinese.
In fact, what would appear to be the situation in Chinese reading
is often very strange, if not bizarre, to a European reader. Accord­
ing to Sampson, because of the differences between the written
and spoken languages, a Chinese script can be read aloud but
may very well be incomprehensible as a spoken text.
But we can go further than this. Sampson is not the first to
argue that, for adult readers of English at least, words are treated
as logograms. It has been argued that English orthography is so
irregular in its graphem e/phonem e correspondence that it virtu­
ally forces readers to use a direct lexical access route. But we
should not assume, even if the correspondence is very regular, as
in Spanish, that experienced readers will rely on decoding phono-
logically. It may well be that while this is a useful device for the
learner, it becomes increasingly discarded the more advanced the
reader becomes.
In this book we shall assume the possibility that both direct and
phonological routes may be open to a reader of English, as well as
combinations of the two.

Reading versus listening

We have already mentioned the possible relationship between read­
ing and listening. It is obvious that they are likely to be very
similar language activities. Once we get past the decoding phase
of the reading process, much of what can be said about reading in
terms of the reader trying to make sense of the message by relat­
ing it to background knowledge, etc., can also be said about lis­
tening (and vice versa). Kintsch and van Dijk (1978: 364) consider
that their model of reading applies to listening as well.
However, these similarities, while of huge importance, can be
overstressed. There are also considerable differences between the
two activities, separate from, but related to, the difference in the
substance in which the linguistic form is encoded (Rost, 1990).
32 Reading in a Second Language
The most obvious difference between reading and listening is, of
course, the fact that the input in the listening situation is sound,
whereas that in the reading situation is verbal shapes, writing. Less
obviously, the sound is ‘raw’ data, which has to be structured by
the listener. The verbal message which has to be read is linguistic­
ally structured; it has already been analysed. Hence the interesting
suggestion by Mattingly, among others, that linguistic awareness is
a prerequisite for efficient reading.
In addition, there are the linguistic differences between the
two dialects mentioned by Sampson. Some of these may not seem
to be very important. The fact that moreover is more common in
writing than it is in speech is not likely to give a competent native
speaker much cause for alarm. Differences like this, moreover,
are more likely to occur in certain types of text, e.g. academic
texts like the present one. A more important difference might
seem to be the greater degree of syntactic complexity found in
written texts of certain types. Not only is there a whole range of
structures which are more common in writing than in speech
(and vice versa), but in writing there is likely to be far more
embedding than in the comparatively simple structures of conver­
sational English. On the other hand, for the foreign listener, the
incomplete structures of spoken English may give at least equal
difficulty.
Probably of more importance than any of this, however, are
what might be called the contextual, or situational, differences
between listening and reading. In listening, intonation patterns
are available, as are facial expressions of the speaker. Feedback is
immediately available, and the listener can generally ask for clari­
fication (in this respect, as in others, formal lectures are more
similar to written texts than to conversation).
The reader, while lacking these props, has available the possi­
bility of stopping the inflow of information at any point desired,
while he cogitates or takes notes. He can move forward in the text
skimming in order to get an idea of what problems are likely to be
encountered, or can ‘regress’ to reread a point which is either
puzzling or perhaps particularly meaningful, or even amusing.
(Regression was condemned by specialists in ‘faster reading’; how­
ever, while it may be bad practice if allowed to become a habit, it
is a practice in which we all indulge, probably to our advantage.)
It is true that many of these tactics can be paralleled to some
extent by a listener equipped with a tape-recorder. However,
 Preliminaries 33
fast-forward skimming is not as easy with a tape as it is with printed
material. Nor is locating a required passage. And while the aver­
age tape-recorder is, of course, capable of repeating part of a
message, it cannot easily be slowed down significantly without dis­
tortion, while the reading process can be slowed almost infinitely
at the control of the reader.
It has often been claimed that reading is potentially much faster
than listening, tied as the latter is to the speed of normal speech.
Sticht (1984), however, has denied this, using evidence from experi­
mental work using speeded-up but non-distorted speech to argue
that there is a maximal speed for both reading and listening,
which he estimates as c. 250-300 wpm. However, the num ber of
variables involved in such experiments, the different functions of
spoken and written language, and the different tasks they imply,
should make us cautious of accepting these results. If valid, they
are evidence for the essential sameness of listening and reading.

The L2 reader

Bernhardt (1991b), who distinguishes between already literate

and non-literate L2 learners, is one of the few writers who does
not treat L2 readers as a homogeneous group. The num ber of
variables to be kept in mind may be very large, including the
following:
(a) literacy in the LI: as Bernhardt points out, some readers may
be literate in their LI, others may not;
(b) the experience of previously acquired languages (and liter­
acies): some Arabic readers, for example, may approach Eng­
lish as an L2; others may already be highly literate in another
language, e.g. French, so that English may become an L3,
or L4;
(c) the linguistic relationship between LI and L2: English and
French, for example, are comparatively similar in many re­
spects; English and Turkish are markedly less so;
(d) cultural relationships: readers of English as an L2 from West­
ern Europe are likely to bring far more shared knowledge to
English texts than are readers from other cultures;
(e) the script of LI and L2: English and Swahili are written in the
same alphabet; English and Greek are written in different
34 Reading in a Second Language
alphabets; Chinese is written in a script which at least tends to
the logographic.
If we assume that reading is more or less the ‘same’ activity in
all languages, we shall not pay much attention to such variables.
If, however, we consider that reading is a language activity, invol­
ving at some levels at least factors specific to a particular language,
then these variables, and others, are likely to be given more promin­
ence. Their potential presence should at least make us wary about
postulating generalised ‘L2 reading processes’.

Summary
In this chapter we have tried to deal with a number of possibly
contentious issues. We have argued that reading is a language
skill, an aspect of language performance. It follows that level of
performance is one aspect of a person’s ability to use the language.
We have presented the distinction between reading as decoding
and reading as message interpretation, and come down conven­
tionally enough on the side of the second view. Somewhat more
controversially, we have suggested that reading ability must go
beyond ‘pure’ language skills, and include pragmatic knowledge
and skills, whereby the readers interpret the text in terms of their
knowledge of the world. Finally, we have pointed out the similarities
between reading and listening, both of which are receptive lan­
guage skills, and taken the position that reading is not necessarily
dependent on listening, but may be a parallel mode of language
reception, with implications for models of the reading process:
activity.

Notes
1. Reading Research Quarterly, XV, 2.
2. We should like to thank Ted Brandhorst of ERIC for his help in
compiling this data.
3. Some years ago The Guardian published an article on remedial read­
ing. Discussing her experiences teaching a 10-year-old native speaker,
one ‘reading expert’ remarked: ‘So I taught him the word “me”.’ One
would have imagined that most native speakers know the word ‘me’ by
that age; what they don’t know is its visual representation.
 Preliminaries 35
4. The eastern academic who referred in a lecture to ‘ospices deeties’
knew perfectly well what he meant. His British audience took some
time to work out that the phrase represented ‘auspicious deities’.
5. It is arguable that in a language such as English, numerals are logo­
grams rather than semasiograms. That is, the word ‘three’ is in a one-
to-one relationship to the symbol ‘3’, which thus represents a linguistic
word rather than a non-linguistic concept. The fact that ‘3’ in French
represents a different word, ‘trois’, is of no relevance to English. How­
ever, symbols such as ‘+’ are clearly semasiographic: the string ‘3+3’
could be translated into English as ‘three plus three’, ‘three and three’,
‘adding three and three’ and so on.
6. One of the authors was told that a Chinese student called ‘Yang’ was
due to join a forthcoming course, and that ‘yang’ in Mandarin meant
‘sheep’. When he asked another Chinese student, a university lecturer,
whether this translation was accurate, he was surprised to get the answer,
‘I don’t know’. It emerged that ‘yang’ has about 10 or 12 meanings, so
that without the appropriate Chinese graph, the particular meaning
intended cannot be decided on.
7. Sampson (1994) doubts the statistics for another reason, namely that
the appropriateness of the phonetic element to a graph’s pronuncia­
tion cannot be measured statistically.
This page intentionally left blank
 37
2

The theory of reading

2.1 MODELS
Reading can clearly be viewed as a cognitive activity; it largely
takes place in the mind, and the physical manifestations of the
activity, eye movements, subvocalisation, etc., are comparatively
superficial. As a cognitive activity, reading has, since the 1960s,
been a major interest of cognitive psychologists. In fact, the huge
increase in articles cited in ERIC, which took place in the 1960s, is
probably brought about to a major extent by psychology research.
In this section we begin by looking at some of the psychological
work which we consider to be relevant to our main themes.
Cognitive psychologists who are interested in reading construct
and test hypothetical models of the reading process as it is thought
to take place in the human mind. Some of these models are in
outline familiar to many teachers of reading: the bottom-up and
top-down models have achieved some fame in the teaching world
since the 1970s, as has the later development, the interactive-
compensatory model. These will be examined in some detail be­
low. The first two of these models have inspired recognizable
methodological approaches; the third has, perhaps, yet to make
its mark.
In fact, the contribution of the cognitive psychologists to L2
reading has been a major one, and would in itself be justification
for a review in this chapter. In addition, these psychologists pro­
vide an admirable example of how to formulate and test empirical
hypotheses, activities which have not always been the strong point
of L2 reading. In this chapter, there is not enough space to give a
full account of the experimental methods involved - a num ber of
38 Reading in a Second Language
which will be examined in Chapter 5. We hope, however, that
we have included enough in this part to indicate how hypotheses
are tested.
When one sets out to examine a factor possibly involved in
reading performance - for example, grammatical knowledge, or
knowledge of the world - there are basically two questions to ask.
The first is whether it can be shown that the factor does have a
measurable effect on reading performance. Once that has been
established, one can then ask in what precise way the factor oper­
ates. It is our impression that researchers in L2 reading have
largely been content with examining the first question, and that
for determined attempts to answer the second question we must
turn to the psychologists.
That having been said, the psychologists cannot be expected to
answer all our problems. They tend to be at their most convincing
when reporting work on what are often labelled ‘low level’ activ­
ities, e.g. word recognition, and not to have a great deal of empir­
ical data on ‘higher level’ activities, such as comprehension of
extended text. Then again they have their own professional inter­
ests which are not likely to be identical to those of teachers and
testers of reading.
An example of the sort of explanation psychologists look for
may make this clear. Stanovich (1980), well known for his formu­
lation of the ‘interactive-compensatory m odel’, reports that adults
and children faced with the task of naming a target word, were
faster when the word was preceded by an incomplete sentence
congruent with it (so that, presumably, the completed sentence
made sense). However, the children’s performance was badly
affected when the target word was preceded by a non-congruent
sentence. Adults, on the other hand, showed no difference in the
time they took in this second condition from the time when
the word was presented out of context. Stanovich suggests that
these results may be explained by a theory which postulates that
there are two processes acting, an automatic activation process
and a conscious attention mechanism. Faced with such a discussion,
the L2 researcher may be tempted to think that, whatever the
validity of the theory, it has little to offer the teacher or tester of
L2 reading. Such a reaction may be unwise, but the example
should remind us that, however grateful we are to the psycholo­
gists, we must constantly assess their work for its relevance.
 The theory of reading 39
Classes of reading model

It is useful to distinguish between two different classes of model.

In the first type, of which the models of Gough (1972), Just and
Carpenter (1987) and Rayner and Pollatsek (1989) are examples,
an attempt is made to model the actual process of reading. This
sort of model may include descriptions of how words are recog­
nised, how long they are kept in working memory, when syntactic
processing begins, and so on. We call these process models. In
contrast, componential models merely describe what components
are thought to be involved in the reading process, with little or no
attempt to say how they interact, or how the reading process actu­
ally develops in time. To return to the distinction mentioned above,
componential models limit themselves to arguing that such and
such a factor is actually present in the process, whereas process
models attempt to describe how the factor operates. It might
appear that process models are more desirable; certainly they might
seem braver. However, there is an argument that for the reading
process as a whole, the design of process models may be premature,
and that we should, perhaps, restrict ourselves to componential
ones. We think both are valuable.
We begin below with an account of the better-known process
models, before moving to an account of componential models.

Process models

Process models may be sequential, that is, they model the reading
process as a series of stages, each of which is complete before the
next stage begins. Alternatively, they are non-sequential, as in the
case of Stanovich’s interactive-compensatory model, where ‘a pat­
tern is synthesised based on information provided simultaneously
from several sources’ (Stanovich, 1980: 35).
The popular view of the development of process models, which
turns up in many article introductions and innumerable PhDs,
goes roughly as follows. First of all came the bottom-up approach,
which was then replaced by the top-down model, which in turn
was replaced by interactive models. In fact, the most frequently
cited example of a bottom-up model, that of Gough, was published
40 Reading in a Second Language
in 1972, whereas the corresponding most frequently cited example
of a so-called top-down theory, that of Goodman, was first pub­
lished in 1967. On the evidence of this paper, Goodman was react­
ing, not against a psychological model, but against a pedagogical
approach to the teaching of initial reading. More seriously, although
Goodman is usually cited as a top-down theorist, there is a good
argument that his theory is an interactive one. Most serious of all,
while the name ‘top-down’ suggests the reverse of a bottom-up
model, no such top-down model exists, nor does it seem likely
that it could ever exist. Finally, while we are all interactive theor­
ists now, the general impression given by some is that a bottom-up
approach, with some modifications, has won the day, e.g. ‘while
all reading may not be characterised as a data driven, bottom-up
process, fluent reading may best be characterised as just such a
process’ (Hoover and Tunmer, 1993: 4). It is worth keeping in
mind that, while this may well be true of word recognition, and
less certainly of syntactic processing, it is certainly not proved to
be the case for other aspects of the reading process.

Bottom-up approaches
Bottom-up analyses begin with the stimulus, i.e. the text, or bits of
the text. In Gough’s (1972) model, the reader begins with letters,
which are recognized by a s c a n n e r . The information thus gained
is passed to a d e c o d e r , which converts the string of letters into a
string of systematic phonemes. This string is then passed to a
l i b r a r i a n , where with the help of the l e x i c o n , it is recognized as
a word. The reader then fixates on the next word, and proceeds
in the same way until all the words in a sentence have been pro­
cessed, at which point they proceed to a component called m e r lin ,
in which syntactic and semantic rules operate to assign a meaning
to the sentence. We should point out that this is only part of the
model. The final stage is that of the Vocal System, where the reader
utters orally what has first been accessed through print. Gough’s
model of the reading process is a model of the reading aloud
process.
We should note that in a model like Gough’s, there are two sets
of entities. First there are text units, arranged more or less in
order of size - that is, the model envisages the reader dealing with
letters, words, then sentences, in that order.1 This in itself would
 The theory of reading 41
entitle it to the term ‘bottom-up\ In addition, though, there are
the processing components, in Gough’s case the scanner, decoder,
librarian, then ‘Merlin’, which are brought to bear on the text
units. More generally, such knowledge/skills components as letter
recognition, lexical access, syntactic parsing, semantic parsing, are
frequently ranked in the literature as ‘lower’ or ‘higher’ skills -
usually, it seems, because of the text components they relate to. In
Gough’s model, textual and processing components operate in
parallel, but this, as we shall see, is not absolutely necessary.
As Rayner and Pollatsek (1989) point out, Gough’s model is
explicit enough to be tested at various points, with the result that
the straightforward bottom-up direction has had to be emended.
For example, in the original model, the letters were seen as being
fed serially into the scanner for recognition. If this were true,
then a word should take longer to recognize than a single letter.
But in fact experiments have shown that this is not the case: words
can be recognized more quickly than individual letters, and even
pseudo-words can be recognized at the same speed as single let­
ters. It appears, then, that at the word-recognition stage, letters
are processed in parallel. More importantly for the debate about
the direction of processing, readers have been shown to use syn­
tactic information to deal with ambiguous words. And Kolers (1969)
found that bilingual readers, reading part French, part English
texts aloud, pronounced words as they would be in the ‘predom­
inant text language’, e.g. ‘murs’ might be read as ‘moors’. Thus it
appears that ‘higher level’ information is being used in word re­
cognition, which conflicts with the unidirectionality of the model.
It is also difficult to see how, as is claimed, one stage of the
process is over before the next stage begins. If all the words in a
sentence had to be recognised before syntactic processing began,
then the model would not appear to have any way of knowing
when to stop processing words and move to processing sentences.
With words, there seem to be few problems: not only is a word
indicated by white space on either side, but it will, if all goes well,
be present for recognition in the lexicon. With sentences, how­
ever, there is no real equivalent of the lexicon, and it is hard to
believe that the reader is entirely dependent on the clues pro­
vided by stops and capital letters.2 On the other hand, if words are
accessed one at a time and fed into the syntactic processor once
recognised, then recognition and syntactic processing are surely
going on at the same time.
42 Reading in a Second Language
Top-down approaches
If, with bottom-up models, it is difficult to see when to stop, with
top-down models, the difficulty is seeing where they should begin.
Bottom-up models start with the smallest text unit, either letters
or letter features. One might expect, then, that top-down models
should begin with the largest unit, the whole text. However, it is
virtually impossible to see how a reader can begin by dealing
with the text as a whole, then proceed to smaller units of the text,
say paragraphs, then down to individual sentences, ending with
single letters. In fact, the term ‘top-down’ is deceptive, appearing
to offer a neat converse to ‘bottom up’, a converse which in reality
does not exist.
In practice, the term is used to refer to approaches in which
the expectations of the reader play a crucial, even dominant, role
in the processing of the text. The reader is seen as bringing hypo­
theses to bear on the text, and using the text data to confirm or
deny the hypotheses. The scope of a hypothesis varies consider­
ably. In the account by Goodman (1967), possibly the best-known
name associated (perhaps wrongly) with top-down approaches,
the hypotheses relate largely to single words. In the applied lin­
guistic and L2 literature, the hypothesis may relate to the whole
text, and be generated by reference to supposed schemata. Given
the somewhat misleading nature of the term ‘top-down’, we sug­
gest that the related terms ‘text(or data)-driven’ and ‘reader-driven’
are more generally useful when describing the contrast between
‘bottom-up’ and ‘top-down’. In the first, the reader processes the
text word for word, accepting the author as the authority. In the
second, the reader comes to the text with a previously formed
plan, and perhaps omits chunks of the text which seem to be
irrelevant to the reader’s purpose.
Goodman is often cited as the representative of the top-down
approach, though he himself has denied the association, and it is
arguable that Frank Smith (1971, 1973) is the more appropriate
choice. As is well known, Goodman views reading as a process of
hypothesis verification, whereby the readers use selected data from
the text to confirm their guesses. Judging by the 1967 paper, it
appears that he developed his position as a reaction, not against
theorists like Gough, but against a pedagogic tradition, which
stressed a fairly strict bottom-up approach to the teaching of read­
ing to young native speakers. Goodman characterizes this approach
 The theory of reading 43
as viewing reading as ‘precise, sequential identification’, with the
consequence that children should be made to be more careful in
their identification first of letters, then of words. From his work
with young native speakers (in the 1967 paper, one of the subjects
is definitely LI, the other appears to be bilingual), Goodman
concluded that this view of reading was wrong, that rather than
painstakingly going from letter to letter, word to word, readers in
fact sampled the text, employing text redundancy to reduce the
amount of data needed and using their language knowledge (syn­
tax and semantics) to guide their guesses. His model of reading,
then, sees the reader as (1) scanning a line of text and fixating at
a point on the line; (2) picking up graphic cues guided by con­
straints set up through prior choices, his language knowledge, his
cognitive styles, and strategies he has learned (p. 270); (3) form­
ing an image which is ‘partly what he sees and partly what he
expected to see’, then making a tentative choice (presumably as
to the identity of the word).
It can be seen from the description above that Goodman’s
model is top-down, to the extent that readers’ expectations are
seen as being brought to the text, i.e. the model is reader-driven.
Secondly, the reading process is seen as cyclical, the reader mov­
ing from hypothesis to text to hypothesis, and so on.
The popularity and influence of Goodman’s first paper was
probably due to a number of reasons. First, it offered an altern­
ative to what might be seen as the grind of moving from letter to
letter, word to word. Learning reading became a more exciting
business. Secondly, it fitted what Chomsky (1965) was saying at
the time about human language users imposing existing ‘rules’ or
expectations on ‘degenerate’ data. Finally, it meshed well with,
although probably pre-dated, notions which became commonplace
about texts always being incomplete and being completed by the
readers by referring to their background knowledge.
The importance which Goodman attributes to hypothesis for­
mation and sampling has had a considerable influence on L2
reading theory: see, for example, Hosenfeld’s claim that the good
reader is a good guesser (e.g. Hosenfeld, 1984). It is also the
aspect which has turned out to be most vulnerable. One criticism
comes from studies of eye movements; Rayner and Pollatsek (1989)
point out that fixations occur on the majority of the words in a
text. While this is only indirect evidence of the process of reading,
it does not conform easily to Goodman’s claims that only part of
44 Reading in a Second Language
the text was sampled. But perhaps the most damaging criticism
concerns the claim by Goodman, Smith and other writers that
good readers guess more, and use the context more than poorer
readers. A great deal of work has shown, quite conclusively, that
while all readers use context, good readers are less dependent on
it than poor ones. In fact, it has been shown that what distin­
guishes good from poor readers, at least among young populations,
is the ability of the members of the first group to decode rapidly
and accurately.
Goodman found that his young subject read words in a sup­
posedly ‘difficult’ text which she failed on when she encountered
them in a more or less meaningless, phonics-focused text. At other
times, he claimed that readers read (i.e. recognized words more
accurately) when faced with the words in a real text than when
the same words were met in a list. However, when Nicholson
(1993a) tested these claims with quite large groups of children,
the results did not seem to support Goodman’s position. While
the results are not always clear cut, it seems that it was the poor
and average readers who benefited from contexts; older and better
readers seem to have been mainly affected by a practice affect, in
that they made fewer errors on the second presentation, whether it
was a list or a text (Nicholson is not clear as to how much time was
allowed between the two trials). In fact it is virtually accepted in
psychology nowadays that, at least at the level of word recognition
and lexical access, some form of bottom-up process is followed.
In spite of this, as has been said above, the assertion by some
that good readers use a bottom-up approach is only really proven
for word recognition. Nicholson, as described above, only par­
tially succeeded in contradicting Goodman’s findings, and it is not
easy to see how a bottom-up approach can account for Goodman’s
original data. It is possible that his model is more appropriate for
L2 readers at certain stages of development than it is for skilled
adult LI readers. Goodman has also been more careful than some
writers in distinguishing between reception and production.

Interactive approaches
Bottom-up models are sequential, in that one stage is completed
before another is begun. In interactive models - one of which was
first credited to Rumelhart (1977) - such a regular sequence does
 The theory of reading 45
not occur. As we noted above for Stanovich (perhaps the best-
known proponent of interactive models), in interactive models, a
pattern is synthesised based on information ‘provided simultaneously
from several sources’ (1980: 35). In Rumelhart’s model, once a
Feature Extraction Device has operated on the Visual Information
Store, it passes the data to a Pattern Synthesiser which receives input
from Syntactical, Semantic, Lexical and Orthographic Knowledge,
all potentially operating at the same point. If one takes Stanovich’s
description as defining interactive models, then Goodman’s is one
such, since, according to him, ‘Readers utilize not one, but three
kinds of information simultaneously’ (Goodman, 1967: 266). The
information is orthographic, syntactic, and semantic.3

Interactive-compensatory approaches
Stanovich calls his model an ‘interactive-compensatory’ one. The
compensatory refers to the idea, intuitively appealing, that a weak­
ness in one area of knowledge or skill, say in Orthographic Know­
ledge, can be compensated for by strength in another area, say
Syntactical Knowledge. At the risk of labouring a point, we might
claim that Goodman’s account contains this notion, since he refers
to weaknesses in the orthographic area being made up for by the
‘strong syntax’ of a real text, meaningful to the young reader. The
notion of compensation has been alluded to in research in L2
reading, for example in Alderson and Urquhart (1985), where it
was hypothesised that background knowledge might make up for
inadequate language skills.
Interactive-compensatory models are very attractive and have
received a great deal of support. Their main weakness, from the
experimental point of view, is that, as Rayner and Pollatsek (1989:
471) point out, they are very good at explaining results but com­
paratively poor at predicting them in advance. To some extent
this is because each reader must be viewed as potentially different,
with different strengths and weaknesses. Hence two readers may
on one occasion arrive at the same level of performance by utilis­
ing different strengths. But this situation, while exasperating for
the model-building psychologist, may simply reflect a widely per­
taining reality.
Before we leave these models, which all to some extent attempt
to mirror the actual process of reading, we shall make some points
about such models in general. First, one consequence of the advent
46 Reading in a Second Language
of interactive models is that an almost infinite variety of models
might seem possible, since one can have all sorts of variations of
interactive top-down and interactive bottom-up models. Thus
Rayner and Pollatsek consider Just and Carpenter’s model to be
basically bottom-up, with interactive elements, while they concede
that Goodman’s model, while basically top-down, also might be
said to have interactive aspects.
Secondly, while there does seem to have been a swing towards
bottom-up models (see the remark by Hoover and Tunm er earl­
ier) , it should be stressed that the empirical evidence in favour of
such models is strongest only in the area of word recognition and
lexical access. Beyond that stage, there is comparatively little agree­
ment, so all sorts of model may be possible.
Thirdly, the psychologists take as a given what is sometimes
referred to as ‘normal reading’; Rayner and Pollatsek narrow this
down to the careful reading of textbooks. But while such a posi­
tion may be convenient for experimenters, it is too narrowly
defined to be acceptable to those interested in the whole range
of reading activities. Once we include other kinds of reading as
legitimate, we may then be tempted to take the view that different
tasks may require different types of reading and different models
of the processes involved. Thus it might seem reasonable to sug­
gest that search reading (see below) is largely reader driven, while
the careful reading of new material is likely to be predominantly
text driven. And investigations by educational psychologists (e.g.
Entwhistle et al., 1979) suggest that either text driven or reader
driven may be the preferred styles of particular classes of reader
(see Section 2.3). While appreciating the cognitive psychologists’
attempts to equip themselves with operational definitions in order
to make testable predictions, we must always keep in mind the
sheer complexity of the activities grouped under the term ‘read­
ing’. Thus Gibson and Levin (1974) deny the possibility of having
models of the reading process, precisely because of different styles
and different responses to different reading tasks.

Componential models
The models we have looked at above attempt to describe the
actual process of reading, a cognitive activity operating in real
time. In fact, Rayner and Pollatsek are critical of some models,
 The theory of reading 47
including those of Goodman and, to a lesser extent, that of
Rumelhart, in that they are insufficiently explicit about the pro­
cess, and hence, in their terms, are not really models at all. The
componential descriptions we now look at do not even begin to
model the process, consisting as they do simply of areas of skills or
knowledge thought to be involved in the process. According to
Hoover and Tunmer (1993), such descriptions try to model read­
ing ability rather than the reading process. The use of such
componential models, again according to Hoover and Tunmer,
is ‘to understand reading as a set of theoretically distinct and
empirically isolable constituents’ (1993: 4). Thus one should be
able to account for different reading performance in terms of
variation in one of the components.

The two-component model

The simplest componential model will consist of two components.
In the case of Hoover and Tunmer, who refer to theirs as ‘the
simple view’, the components are word recognition and linguistic
comprehension. They claim that Fries (1963), Venezky and Calfee
(1970) and Perfetti (1977) hold the same view.
Clearly, provided that it can account for the observed data, a
model with fewer variables is preferable to one with more. So if
the simple two-variable model can do this, then it should be cho­
sen. Hoover and Tunmer claim that it can, and adduce evidence
of different types in support of their claim (see Chapter 3 for
further discussion of the componential issue in relation to testing).
The evidence takes two main forms: first, data showing that the
two variables are separable. Here we might point to the obvious
evidence that LI illiterates, both young and adult, can generally
be assumed to understand language, but cannot read, because
of their lack of decoding skills. Then Hoover and Tunm er cite
evidence from disabled readers. Dyslexics have normal linguistic
competence, but are deficient in decoding skills (1993: 10). Chil­
dren suffering from hyperlexia, on the other hand, show high
levels of decoding skill but poor linguistic comprehension. Fin­
ally, longitudinal studies of normal children have shown that, in
the early grades, the correlations obtained between decoding and
linguistic comprehension are low, but become steadily higher as
the children advance through the school.
48 Reading in a Second Language
The other kind of evidence comes from statistical techniques
measuring the contribution that different factors make to reading
performance. In a typical study by Stanovich et al. (1984), decod­
ing accounted for 38 per cent and linguistic comprehension 13 per
cent of the variance in reading comprehension of fifth graders,
after the effects of non-verbal intelligence had been removed
(p. 12).
However, while this accumulation of evidence is impressive,
doubts must remain as to the simplicity of the ‘simple’ view. Hoo­
ver and Tunmer use the terms ‘decoding’ and ‘word recognition’.
Definitions of the first vary, they admit, with some limiting the
term purely to the ability to replace strings of graphemes (letters)
with corresponding strings of phonemes, and others insisting on
including full lexical access, i.e. recognising what the word means
(Goodman is one of the latter). Hoover and Tunmer use ‘decod­
ing’ for accessing the lexicon by means of a phonological route
(spelling it out) and ‘word recognition’ for the process of access­
ing the lexicon ‘based on graphic information’ (i.e. presumably,
without recourse to the phonological route). This doesn’t appear
to leave any term for the ability to decode pseudo-words which do
not appear in the lexicon (although they at one point say that the
ability to recognize such ‘words’ should be taken as evidence of
decoding). But if ‘word recognition’ involves accessing the mental
lexicon, isn’t this an important part of ‘linguistic comprehension’?
The situation becomes more complex when we consider ‘lin­
guistic comprehension’. Hoover and Tunmer define this opera­
tionally as the ability to answer questions about an oral narrative.
There are two objections to this: one, that it is too narrow; the
other, that it is too broad. The narrowness relates to the obvious
limitation on text-type; the broadness concerns the fact that such
an ability may, and almost certainly does, involve far more than
linguistic competence. For example, Carroll (1972) tries to distin­
guish ‘pure’ or ‘simple’ comprehension from ‘total’ comprehen­
sion. The second includes processes of inference, deduction and
problem solving, the first does not. His description of pure com­
prehension as that which is linguistically committed to the text, is
probably very similar to what Fries had in mind when he limited
comprehension to ‘a grasp of meaning in the form in which it is
presented’ (1963: 115). It is clear, however, both from the way in
which it is assessed, and from descriptions such as ‘the ability to
take lexical information (i.e. semantic information at the word
 The theory of reading 49
level) and derive sentence and discourse interpretations’ (Hoover
and Tunmer, 1993: 8) that Hoover and Tunm er’s view of compre­
hension is closer to Carroll’s ‘total’ comprehension.
The distinction may seem very academic, but it does have quite
immediate consequences, particularly in the L2 area. As we shall
see below, attempts to distinguish between ‘reading skills’ and
‘language skills’ often rely on a definition of language skills sim­
ilar to that implied by Carroll, i.e. a competence in the system.
Thus they depend on splitting ‘language ability’ into ‘language
competence’ and ‘language use’. And it is of great importance to
language teachers to know whether increasing their students’ know­
ledge of part of the linguistic system, e.g. grammar, will help their
reading (see Section 5.2 for how this might be investigated). Hence
it is important to define one’s terms more precisely than Hoover
and Tunm er have done. Even with native speakers, it is perfectly
feasible that aspects of their LI language use may develop dif­
ferentially in different students who all share the same basic
competence. Hence one might expect variation in later perform­
ance which could not be accounted for in Hoover and Tunm er’s
simple model.
The simple view and the L2 reader
For all the attractions of the simple view, there are obvious diffi­
culties in using this model to describe L2 readers. Describing as
they do young LI learners, Hoover and Tunm er can make the
reasonable assumption that their language knowledge can be
tapped by an oral comprehension test. This is unlikely to be a safe
assumption to make with L2 learners, who may well perform bet­
ter on a reading test than on an equivalent oral test. It is not even
clear whether it is a safe assumption to make with adult LI readers.
Faced with this difficulty, the L2 researcher is likely to have to
pay more attention to what is meant by language, and may have to
subdivide this area into components such as syntax, cohesion, text
structure, etc. As has been argued above, this may have the effect
of rendering the simple view no longer very simple.
Three-component models
Coady (1979) and Bernhardt (1991b), both describing L2 reading,
include three variables: in Coady’s case these are Conceptual Abil­
ities, Process Strategies and Background Knowledge. Bernhardt’s
50 Reading in a Second Language
model consists of Language, Literacy and World Knowledge. Tak­
ing Coady’s model first, ‘conceptual abilities’ are equivalent to
intellectual capacity, and Coady (1979: 7) remarks that some ‘adult
foreign students . . . may fail to achieve the competence necessary
for university instruction because they lack intellectual capacity
and not totally or necessarily because they cannot learn English’.
He adds that there is not much to do about this situation. ‘Back­
ground knowledge’ will be dealt with below. It might be pointed
out here, however, that Hoover and Tunmer constructed their
experiments to neutralize the possible effect of background know­
ledge. By ‘process strategies’, Coady means both a knowledge of
the system arid the ability to use the knowledge, e.g. ‘knowledge
of the phonology of a language implies the ability to identify
phonemes and use this knowledge for practical purposes such as
listening’ (p. 7). In other words, Coady’s component consists of
language use, rather than ‘simple’ competence.
One can see Coady’s model as adding more components to the
simple view. Alternatively, one may see it as subdividing Hoover
and Tunm er’s ‘comprehension’ into separate components. Given
that Coady considers the interaction between his components to
result in ‘comprehension’, the latter is the neater solution. In
other words, for Coady, background knowledge is not an addition
to comprehension, it is an actual component of comprehension.
However, there is one important component in Hoover and
Tunm er’s model which is lacking in Coady’s, namely Word Recog­
nition. The only acknowledgement Coady makes of this is to in­
clude phonem e/graphem e correspondences as part of the Process
Strategies component. Of course, it could be argued that Coady’s
is a model of comprehension and not of the reading process. It
seems more likely, however, that this is evidence of the predom­
inance in L2 reading of applied linguists, with a tendency to
emphasize high-level processes.
Bernhardt’s model also lacks a separate word-recognition com­
ponent. Like Coady, she also proposes a model with three vari­
ables, namely World Knowledge, Language and Literacy. World
knowledge equals background knowledge; language includes the
‘seen’ elements of the text, such as word structure, word meaning,
syntax and morphology. It must, however, be seen as including
language use. Literacy equals operational knowledge: knowing how
to approach text, knowing why one approaches it and what to do
with it. How would one go about finding empirical support for
 The theory of reading 51
the separate existence of Bernhardt’s literacy component? Pre­
sumably by finding students who shared equivalent background
knowledge of a topic, and were at the same proficiency level in
the language, but differed in performance on texts in a systematic
way which could be attributed to self-monitoring etc.

2.2 COMPONENTS IN DETAIL

We now proceed to examine in detail some of the components
mentioned, namely word recognition, language (examined in dif­
ferent subcomponents), background/world knowledge and literacy.
In other words, the ‘m odel’ whose components we are going to
discuss is in some ways an amalgam of those of Hoover and Tunmer
(1993), Coady (1979) and Bernhardt (1991b). In many cases, the
research we touch on involves LI subjects. However, when L2 data
exist, we shall concentrate on that.

Word recognition
We should start by saying that even the definition of the meaning
of the term ‘word recognition’ is disputed. Hoover and Tunmer
(1993) mention three interpretations. The most obvious one would
be to have the term mean ‘recognize an English word in print, be
able to pronounce it, and give its meaning’. Some people would
dispute the need to include pronunciation (the phonological part).
Many of the experiments on word recognition have involved the
use of pseudo-words, such as ‘m ard’. So we might have to extend
the term to mean ‘recognition of pronounceable strings of letters
which are not actual words in English’. But some of the experi­
ments involve recognition of ‘unpronounceable’ pseudo-words.
So we might have to add ‘recognition of any letter string with space
boundaries on either side’.
Then the processes of word recognition appear to be extremely
complicated, and not well understood. So we will content our­
selves here with stating what is generally agreed, what facts any
theory has to account for, and the immediate implications for a
reader of a foreign language. In our account we rely mainly on
Rayner and Pollatsek (1989), keeping in mind that most of what
they have to say concerns skilled adult readers.
52 Reading in a Second Language
First, as we reported earlier, letters are not processed serially. If
they were, then the time taken to recognize a word would be
longer than the time needed to recognise a single letter, and the
longer a word, the longer it would take to recognise. Within reason,
this does not seem to be the case. Furthermore, subjects are more
accurate in reporting letters in words, e.g. the ‘d ’ in ‘word’, than
they were in recognizing ‘d ’ by itself.
On the other hand, it cannot be the case that words are recog­
nised as templates or pictures. In the experiments mentioned
above subjects were as quick to report letters in pseudo-words like
‘orwd’ as they were to report the single letter, although it is incon­
ceivable that they had a template for such pseudo-words. And as
Rayner and Pollatsek point out, the notion of templates cannot
account for subjects’ ability to read in a number of different fonts.
So Rayner and Pollatsek settle on a modified version of a model
by Paap et al. (1982), in which the visual input goes first to f e a t u r e
DETECTORS, then to LETTER DETECTORS, and finally to WORD DETEC­
TORS. Since on their own admission the evidence for the viability
of the model is too complex to present in their book, we shall
content ourselves with mentioning it.
Given that in normal reading the purpose of word recognition
is to access the lexicon, it is generally recognised that there are
two routes to this. The first, known as the direct route, goes straight
from the visual input to meaning without recourse to sound; the
other, known as the phonemic or phonological route, goes from
visual input to sound to meaning. Evidence that readers of Eng­
lish use a direct route comes from consideration of the writing
system; the system notoriously contains too many irregularities to
allow total reliance on the phonological route. However, use of
the direct route alone cannot explain subjects’ ability to handle
pseudo-words like ‘m and’, or ‘birn’, which are not likely to be
present, and hence accessible, in the reader’s lexicon. Other evid­
ence for a phonological route comes from phonological influence
on word recognition: recognition of words like ‘touch’ have been
shown to be slowed down when they are preceded by, in this case,
the word ‘couch’. Also recognition of pseudo-homophones (non­
words which share the same pronunciation as real words, e.g.
‘phocks’) have been shown to be slower than recognition of other
pseudo-words. Here, presumably, the activation of the pronunci­
ation collides with the real word in the lexical entry, thus causing
confusion.
 The theory of reading 53
Finally, Rayner and Pollatsek cite as evidence for two routes the
fact that so-called surface dyslexics can pronounce most words,
but regularize irregular ones, while phonemic dyslexics pronounce
most words correcdy but cannot pronounce non-words. This can
be explained by arguing that the first group are limited to the
phonological route, and the second group to the direct route.
Thus Rayner and Pollatsek (1989: 109) conclude that ‘the com­
mon ground for all positions is that direct visual access is import­
ant and that sound encoding plays some part’.
Two other aspects of word recognition seem to be relevant
here. Most of the research appears to have been done with single
morpheme words, such as ‘touch’ and ‘word’. It seems, to say the
least, somewhat odd to suggest that there is a separate lexical
entry for, say, not only ‘expect’ but also ‘expects’, ‘expected’,
‘expecting’, ‘expectation’, ‘unexpected’, etc. Rayner and Pollatsek
report evidence that, in such cases, the root morpheme is accessed
first, together with some evidence that ‘content’ words are stored
separately from ‘function’ words, so that, for example, ‘expect’
would appear in a different lexicon from, say, ‘the’. Function words
would seem, anyway, to pose problems for the ‘phonological route
only’ school: it seems, frankly, incredible that a skilled reader,
reading fast, will distinguish between different phonologically
conditioned pronunciations of ‘the’.

Word recognition in L2
Randall and Meara (1988) remark that most L2 reading research
‘has been centred on the relatively higher-order skills of discourse
organization and the interpretation of continuous text’, and say
that this is ‘for obvious reasons’. Perhaps one obvious reason is
that many of the potential subjects are presumed to be past the
‘simple’ stage of word recognition by the time they become avail­
able to researchers, though such an assumption is by no means
certain. A further reason is that many L2 researchers have a train­
ing in applied linguistics, which has tended to ignore this area.
Whatever the reasons for the neglect, it is clear that word re­
cognition poses intriguing problems for L2 reading researchers.
Take, for example, the question of phonological access to the
lexicon, which presupposes that for a word to be accessed in read­
ing, the lexical entry must contain a phonological component, i.e.
it must contain information as to how the word is pronounced. In
54 Reading in a Second Language
most cases, skilled adult LI readers can be assumed to have this
information, while most difficulties foF young LI learners can be
avoided by careful control of vocabulary. But these assumptions
cannot be made with respect to many L2 readers; in a huge number
of cases they are going to come across words which they have not
heard pronounced. Does this mean they are unable to access them?
Then there is the question of the script with which the L2 learners
are familiar when they begin to read the L2. If the learners come
from a different orthographic tradition, is this likely to affect their
reading in the L2?

Phonological versus direct access

Koda (1987) cites experimental findings which show that English
readers took longer and made more errors judging homophone
sentences than Chinese readers. The claim is that Chinese readers,
brought up on a logographic script, will access words directly.
Suarez and Meara (1989) argued that as Spanish readers are used
to a ‘regular’ script - i.e. one in which there is a regular corres­
pondence between phoneme and grapheme - they may, when
faced with English, behave differently from English readers. One
hypothesis was that they might rely entirely on the phonological
route, and hence make a disproportionally large number of errors
on ‘irregular’ words. Suarez and Meara used a method devised by
Glushko (1979), whereby the subjects were exposed to (a) regular
words (e.g. ‘bleed’), (b) regular pseudo-words (e.g. ‘dreed’),
(c) irregular words (e.g. ‘blood’) and (d) irregular pseudo-words
(e.g. ‘drood’). The authors consider that their results were incon­
clusive, though the pattern of errors seemed to be similar to that
of English-speaking surface dyslexics. We are somewhat dubious
in this case as to the validity of the experimental method, in par­
ticular the status of ‘irregular pseudo-words’, and, more generally,
of ‘regular’ words in the context of L2 research. We deal with
these doubts in Chapter 5.

The effect of script type

Randall and Meara (1988) tested Arabic speakers to see if they
processed letter strings in the same way as English speakers (Ara­
bic, of course, is written from right to left). Studies of English
 The theory of reading 55
speakers had shown that, given a target shape, then presented
with a verification task of locating the target among an array of
five shapes, English speakers displayed a ‘u ’ pattern of response,
i.e. they took longest if the target was either at the beginning or
end of the array. On the other hand, when faced with an array of
five letters or digits, the pattern they displayed resembled a tilted
‘M’ shape; in other words, they located the target quickest if it was
located at the beginning or end of the array, and rather quicker
if it occurred at position 3 than at position 2 or 4. The argument
is that this reflects reading patterns, and that English readers
are used to paying attention to the beginning and end of words
(an argument which seems a little tenuous to us). Arabic readers
responded in a similar way with target shapes, displaying a ‘u ’
response pattern. However they also showed this response pattern
when confronted with arrays of English letters. Moreover, their
responses seemed to show that they were processing the words
from right to left. The study was repeated over a period, during
which the right-to-left processing disappeared, but the same ‘u ’
pattern persisted. Randall and Meara point out that they are not
predicting any direct consequences of this difference in response,
but presumably some consequence may be possible.

The effect of morphological differences

Ryan and Meara (1991) noticed in a pilot experiment that Arabic
speakers, when presented with incomplete words like ‘presrve’
and asked to complete them, were more likely than other L2
subjects to produce different words, e.g. ‘pressure’. In their main
experiment, Arabic speakers, when given an identification task
involving either identical or slightly altered words, e.g. ‘depart­
m ent/dpartm ent’, made far more errors than either LI or other
L2 subjects. Ryan and Meara tentatively suggest that this is the
result partly of the structure of Arabic, in which the ‘root’ of a
family of words is composed of a sequence of three consonants,
and partly of the Arabic orthography, which largely omits vowels
(see Chapter 1). This combination of factors may dispose Arabic
speakers, when reading English, to identify words on the basis of
their consonants, while not paying sufficient attention to vowel
letters.
56 Reading in a Second Language
Learning unknown words
Cognitive psychologists investigating the reading behaviour of
skilled adult LI readers can assume, on the whole, that words
encountered will be in the subject’s lexicon; hence, with the ex­
ception of pseudo-words, what they are studying is word recogni­
tion. In the case of L2 learners, however, such an assumption
cannot be made, so that, in addition to word recognition, there is
also a major area of concern involving the initial handling and
learning of unfamiliar words.
Pitts et al. (1989) showed that L2 readers were able to acquire
a number of unfamiliar lexicon items when reading chapters from
The Clockwork Orange. The words were ‘nadsat’ words, Russian lex­
ical items used as slang in the book, and hence were presumably
unknown to the readers on first encounter. Day et al. (1991)
found that Japanese readers were again able to acquire previously
unknown vocabulary, this time English, from reading an English
text. Both studies showed that readers had correctly assigned mean­
ing to the items; in neither case did the investigators enquire
whether they had also assigned phonological readings to the words.
The English words used by Day et al. were on the whole ortho-
graphically regular, e.g. ‘fire’, ‘clear’, ‘stare’, so a phonological
reading could have been carried out. Unfortunately Pitts et al. do
not list the ‘nadsat’ words used.
Koda (1987) tested Japanese students reading an English text
in which a num ber of previously encountered pseudo-words had
been inserted. Some of the words were ‘pronounceable’, some
‘unpronounceable’. The aim was not, in fact, to examine word
recognition; previous studies had shown that phonological recoding
appears to be important in short-term memory syntactic process­
ing (see below) and Koda wanted to see if Japanese readers, used
to a logographic kanji script, would be affected in this respect.
The subjects reading unpronounceable words, contrary to expec­
tations, and in contrast to the behaviour of LI readers, read the
text faster than those reading the text with pronounceable words.
However, later self-monitoring reports by subjects showed that
they had adopted a range of avoidance strategies, e.g. renaming
the pseudo-words, so the experiment tells us little about how
unknown words are used. In addition, we are dubious about
the general acceptance, in the psychological literature, of ‘unpro­
nounceable’ words (see Chapter 5, Section 5.1).
 The theory of reading 57
In general, the whole topic of the recognition and learning of
lexis in reading seems to us to be one of major importance in L2
reading, and one to which L2 researchers are in a position to
make a valuable contribution.

Language
It has already been pointed out that the distinction between lan­
guage and word recognition, drawn by Hoover and Tunmer, is
not a clear one, since the lexicon is clearly part of our linguistic
competence. However, we are keeping the distinction here for
convenience. We need, however, to break the topic down into
more manageable subtopics. Hence, below, we discuss syntax, then
cohesion, and larger aspects of text structure.

Inner speech
Before we consider syntax, we should look briefly at the notion of
inner speech. This is the Voice in the head’ which many of us are
aware of while we are reading. Rayner and Pollatsek (1989), who
devote considerable space to the issue, try to distinguish inner
speech from subvocalisation. The latter, which involves actual phys­
ical activity in the speech tract, used to attract the attention of
reading teachers, who taught that it should be suppressed. How­
ever, electromyographic recording shows that subvocalisation is a
normal part of silent reading.
Inner speech is more interesting. There is considerable evid­
ence that the sounds of words influence the speed or accuracy
of silent reading. Rayner and Pollatsek report that readers find
strings like
Crude rude fude stewed food
difficult to read silently. They suggest that the effect of inner
speech is post-lexical, i.e. occurs after lexical access, and that its
function is to hold material in the working memory until it is
processed. Perfetti and McCutchen (1982) claim that inner speech
is not a complete representation of every word in the text, but is
biased towards the beginning of words. They suggest that function
words may not require as elaborate a phonetic representation as
content words.
58 Reading in a Second Language
As far as we know, nothing is known about the effect of the
presence or absence or form of inner speech in L2 readers. Koda’s
experiment, reported above, was directed towards this area, but
produced no convincing results. Rayner and Pollatsek (p. 211)
report that inner speech ‘may be somewhat less important in Chi­
nese than in English’. With reference to deaf LI readers, they
report that ‘the comprehension and memory advantages provided
by one’s primary language weigh heavily in the choice of a recoding
system’ (p. 210). One would speculate that if the LI reader of
English relied on a phonological rendering of the message to
assist in processing syntactic units, then the L2 reader is likely to
be doubly handicapped, being uncertain of both the syntax and
the phonology. It is just possible that the finding by Dhaif (1990)
that Arab students’ comprehension of written English was signi­
ficantly improved by the teacher reading aloud while they read
silently in parallel has some relationship to inner speech.

Syntax
In addition to words being recognised, the significance of the
relationships between them (e.g. the syntax) needs to be extracted
by the reader. It would be reasonable that, given the vast amount
of work which has been done in linguistics in the area of syntax,
we would be well informed as to how readers operate. This, how­
ever, is not the case (the reader may remember that the syntactic
and semantic component in Gough’s model was called ‘Merlin’).
Rayner and Pollatsek mention a number of approaches, none of
which seems to have attracted anything like the attention in psy­
chology as have problems of word recognition or eye movements.
They single out for special mention two approaches. The first,
they refer to as the ‘Clausal’ model of processing, developed in
the 1970s. This ‘m odel’ consisted of a number of pragmatic strat­
egies, e.g. ‘take the first clause to be the main clause unless there
is a subordinating conjunction’. Rayner and Pollatsek are fairly
dismissive of this approach, claiming that such pragmatic rules
would form ‘an unsatisfactory hodgepodge’ (p. 246). However, it
seems to us that L2 readers may well build up such a set of strat­
egies, partly derived from their LI, partly constructed to deal
specifically with the L2.
The approach Rayner and Pollatsek favour is the so-called ‘gar­
den path’ approach. This contains two main principles. According
 The theory of reading 59
to the first, known as ‘minimal attachment’, the reader structures
data to try to minimize .-the number of grammatical nodes re­
quired. Thus, a sentence like ‘The girl knew the answer by heart’
is likely to cause fewer processing difficulties than the sentence
‘The girl knew the answer was wrong’ since the latter, in terms of
Phrase Structure Grammar, requires a subordinate sentence node
not required for the former sentence. The second principle, known
as ‘late closure’, claims that, when grammatically possible, readers
will attach new items to preceding items rather than subsequent
ones. Hence, given the two sentences
Since Jay always jogs a mile this seems like a short distance to him.
Since Jay always jogs a mile seems like a long distance to him.
readers are predicted to have fewer problems with the first, since
in both cases they will initially attach ‘a mile’ to ‘j ogs’ - a move
which works with the first sentence but has to be revised in the
case of the second.
Rayner and Pollatsek claim that experiments using observation
of readers’ eye movements support the existence of both prin­
ciples. However, on the examples they produce, it is not clear to
us that there are, in fact, two principles involved. In both sets of
sentences quoted above, the readers’ difficulties might be attrib­
uted to taking ‘knew’ and ‘j ogs’ as transitive verbs requiring an
NP object, then assuming that the first possible NP, ‘the answer’
and ‘a mile’, completes the Verb Phrase; in other words, a version
of ‘late closure’. In fact, one is tempted to agree with Ridgway
(1997) that an approach based on ‘dependency’ grammar would
be fruitful. Such grammars, however, have the disadvantage of
being relatively underdeveloped.

Syntax in L2 reading
When we turn to grammar in L2 reading, we find again a dearth
of data. There are probably at least two factors involved here.
Some years ago it was not uncommon to find EFL books contain­
ing ‘reading passages’ which seemed to have been included mainly
to supply fodder for grammar teaching. If one considers written
text in this way, then it is not likely that one will investigate
the effect of one on the other. After this, the ‘communicative’
approach tended to stress language use, and hence disparage
60 Reading in a Second Language
attention being paid to ‘knowledge’ areas such as syntax. Finally,
as commented on by Randall and Meara (1988), a concentration
on ‘high-level’ factors such as background knowledge, skills and
strategies, led to the comparative neglect of lower-level factors
such as syntax.
We have been taking the conventional position that syntactic
parsing of some kind was necessary in order to impose meaning
on the words recognised. This apparently commonsensical position
has been contradicted by findings of Ulijn and his associates. In
Ulijn and Kempen (1976), Dutch and French speakers read a text
about finding their way around an imaginary town, Beausite. The
text, which was in French, existed in two versions. In one version,
French syntactic structures not found in Dutch were included. In
the other, such structures were avoided. However, there was no
difference in either the Dutch or the French readers’ responses
to the text. Ulijn and Kempen conclude that: ‘under normal con­
ditions reading comprehension is little dependent on a syntactic
analysis of the text’s sentences.’
In later experiments (Strother and Ulijn, 1987) students from
different linguistic backgrounds - English and others - read a text
on an aspect of computer science. One version of the text was the
original; in the other, ten ‘passages’ (i.e. sentences) had been
‘simplified’ in certain specific syntactic ways, e.g. passives were
replaced by active equivalents, nominalisations by expanded N oun-
Verb constructions. Again no significant differences were found
between responses to the original and simplified text. Strother
and Ulijn conclude that readers use a ‘conceptual strategy’, con­
sisting largely of knowledge of word meanings together with know­
ledge of the text’s subject area. Thus, in the model, a syntactic
element could be eliminated.
As far as simplification of text for L2 readers is concerned,
there may well be a case for an emphasis on lexis, as Strother and
Ulijn argue, though whether results based on the ‘simplification’
of ten sentences of a text of unspecified length is good evidence
for this is debatable. To claim, however, as Ulijn seems to do at
times, that syntactic processing is not necessary, is frankly unbe­
lievable. This is easily demonstrated. The following string rep­
resents an English sentence from which most (not all) function
words and all inflectional morphemes have been deleted. More­
over, since ordering plays a major part in English syntax, the
order of the remaining words has been jumbled.
 The theory of reading 61
begin several it recogniser module machine digital pass record speech
We challenge anyone, whether expert in the content area (arti­
ficial language) or not, to process this string. Things begin to be a
bit better if we restore the original ordering:
Machine begin digital record speech pass it several recogniser module
However, it is only when we restore function words and inflec­
tions that the message becomes easy to extract:
The machine begins by digitally recording the speech and passing it to
several recogniser modules.
The subjects used by Ulijn and his associates were comparatively
expert in the L2: for example, the Chinese students used by
Strother had TOEFL scores of 550+ and had been in the USA for
nine months. Ulijn’s Dutch students had studied English at sec­
ondary school for six years and had had ‘considerable exposure to
English’ ( Ulijn and Kempen, 1976: 94). It can be assumed, there­
fore, that the subjects’ syntax was sufficient to cope with whatever
was given them. It might, of course, be reasonably concluded that,
at their level, extra syntactic tuition would give smaller returns
than an emphasis on vocabulary building. But that is a very dif­
ferent thing from claiming that ‘reading comprehension is litde
dependent on a syntactic analysis’.
One point of interest that can be retrieved from the work
of Ulijn and his associates can be found in the remark that ‘a
thorough syntactic analysis is unnecessary’ (our italics). It has
sometimes been claimed that the amount of syntactic knowledge
necessary for reading is less than that required for writing or
speaking. Thus, given that the readers had enough background
knowledge, they might make quite reasonable sense of, say,
machine begin by digitally record speech and pass it to several recogniser
module.
In other words, a successful processing of this text might not
depend on a detailed knowledge of the determiner system, mor­
phological marking of plurals, etc. Thus it might be possible to
distinguish between a receptive and a productive syntactic pro­
cessor (see Section 5.2).
Alderson (1993) has produced evidence of a strong connec­
tion between grammar and reading. During the preparation of
the IELTS test, item writers were instructed to produce a test of
62 Reading in a Second Language
grammar which could be used along with tests of reading, listen­
ing and writing in the total test. After tests had been trialled in
both the UK and Australia, it was found that very high correla­
tions held between the grammar test and different tests of read­
ing. For example, the correlation between the grammar test and
the science and technology reading test was 0.80. This was in spite
of the fact that the grammar test was designed as a test of gram­
mar in general, rather than of structures found in the reading
tests, and that the reading tests did not include any specifically
grammatical item.
While we have some doubts about details of the tests used (see
Part 5), there seems little doubt that Alderson is correct in con­
cluding that ‘it must be the case that, in some intuitive sense, a
reader must process the grammar in a text in order to understand
it’, and that ‘. . . the evidence certainly does not support any claim
that one can successfully understand text without grammatical
abilities’ (p. 219).
However, this is more or less where the case rests. We don’t
know how L2 readers process texts syntactically, though, as men­
tioned above, we may suspect that they apply a collection of prag­
matic strategies, e.g. in English, the first NP is likely to be the
Subject. These strategies are likely to be influenced by their experi­
ences with reading in their LI, as Cowan (1976) has posited. It is
quite likely that more breakdowns occur in processing than are
obvious on the surface; one of the authors discovered that Indo­
nesian students seemed not to be able to assign Subject or Object
roles to nouns in relative clauses. But little is known of this area.

Background knowledge
Both Coady’s and Bernhardt’s models contain a component called
‘background knowledge’ (in Coady’s model) or ‘world knowledge’
(in Bernhardt’s model). Hoover and Tunmer also mention back­
ground knowledge but only to attempt to exclude it. They are
interested in reading ability, rather than reading performance;
presumably also one’s background knowledge can be assumed to
be constant whether one is reading or listening, and therefore
cannot be used to distinguish between the two activities. In L2
reading, however, and in particular in the area of LSP, the con­
cern is more to predict performance on particular reading tasks,
 The theory of reading 63
and for this, the background or world knowledge of L2 readers
may well need to be taken into account.
The theoretical justification for including background know­
ledge as a component of our reading model can be seen as deriv­
ing from two different sources. First, it is part of the theory of
comprehension associated with the notion of ‘schemata’ (see below)
that text is never complete, and that the reader (or listener) must
supply additional material derived from their existing knowledge
of the world. From this point of view, background knowledge
is inevitably present in all kinds of reading, both LI and L2.
The second source is interactive models of the reading process.
Although not by any means constructed with L2 readers in mind,
such theories are good at predicting that L2 readers, with signific­
ant defects in their knowledge of the language, may sometimes
perform as well as LI readers. The theory will predict that, assum­
ing they have the required background knowledge, L2 readers
may use this knowledge to compensate for linguistic shortcomings.
The possibility of such an outcome is of practical importance in
deciding, for example, whether an L2 student is capable of pro­
ceeding to an academic course of study involving reading in their
own speciality. Thus background knowledge has been of particu­
lar interest to those involved in testing and teaching LSP (see
Chapter 3).
There is a considerable amount of experimental evidence in
L2 reading that background knowledge can play the part envi­
sioned for it in the theory. Bernhardt (1991b) gives an extensive
list of studies, to which we refer the reader. The majority of studies
she cites were successful in showing that readers’ familiarity with
content had a significant effect on their performance. However,
in a number of cases no such effect has been found (e.g. Clapham,
1990, who found that ‘subject area had no significant effect on
scores’). Because of this, we shall focus on the conditions required
before the effect of background knowledge becomes evident.
We shall begin by examining two studies - one by Mohammed
and Swales (1984) and one by Alderson and Urquhart (1985) -
which will serve to illustrate some of the relevant factors. Both
these studies belong to the same group in Bernhardt’s classifica­
tion, being concerned with background knowledge of topic rather
than cultural background. However, if we accept the claim by
Widdowson (1978), that science constitutes a culture, this division
becomes somewhat arbitrary.
64 Reading in a Second Language
Mohammed and Swales gave twelve postgraduate students the
tasks of using an instructional pamphlet to (a) set the current
time on a digital clock and (b) set the alarm for a specific time the
next day. Performance was measured in time required to accom­
plish the tasks. The subjects were categorized as (a) Native speaker
scientists (NS), (b) Native speaker arts (NA), (c) Non-native speaker
scientists (NNS), (d) Non-native speaker arts (NNA). The linguistic
proficiency of the non-native speakers was arrived at using teachers’
estimates of the subjects’ potential band scores on the IELTS test:
General and Reading modules.
The subjects were video-taped during the tasks, to investigate
their overall behaviour (reading different parts of the instructions,
manipulation of controls, etc.). While this record is irrelevant
here, such data are clearly important for studies of different read­
ing behaviours and different reading models.
The main measure of performance was the time subjects took
to complete the tasks. They did so in the following order of pro­
ficiency: (1) NS, (2) NNS, (3) NA, (4) NNA. Thus the non-native
speaker scientists performed better than the native speaker arts
subjects, in spite of the fact that their average band score was 5.4,
while the native speakers were assumed to have a band score level
of 9. Moreover, the NNS group also outperformed the NNA group,
in spite of the fact that the latter group had an average band
score of 7.9. Mohammed and Swales (1984) ascribe the difference
between groups to ‘either field-familiarity or, more likely, familiarity
with the genre of technical instructions’ (p. 211), and express
surprise at the strength of the influence of technical experience,
and the ‘apparent unimportance of general English proficiency
above a presumed threshold level’ (p. 216).
Alderson and Urquhart (1984) carried out a series of three
studies using subject-related groups of L2 postgraduate students,
namely Engineers (ENG), Science and Maths (SM), Development
Administration and Finance/Economics (DAFE) and Liberal Arts
(LA). In Studies 1 and 2 there were three groups of texts, aimed
primarily at the ENG, DAFE and LA groups. The tasks were gap-
filling (Studies 1 and 2), and gap-filling plus short-form answers
(Study 2). In Study 3, three modules of the ELTS test, Technology,
Social Sciences, and General Academic were used.
The results were inconsistent. In Study 1, ENG outperformed
DAFE on the engineering texts, as predicted, while DAFE outper­
formed ENG on the DAFE texts. However, in Study 2, while the
DAFE group outperformed ENG on the DAFE texts, there was no
 The theory of reading 65
difference between the two groups on the ENG texts. In Study 3,
in contrast, the SM-ENG group (ENG and SM combined) outper­
formed the DAEE group on the Technology module, while on the
Social Science module, the two groups did not differ significantly.
Alderson and Urquhart concluded that in the case of the Engin­
eering texts, the background knowledge of the engineering stu­
dents was compensating for their comparative low level of language
proficiency.
Thus while the studies provided evidence of an effect of back­
ground knowledge, this effect was not consistent throughout. There
was also evidence of a factor related to language proficiency (the
LA groups, which were all through more proficient on measure­
ments of proficiency than the other groups, in virtually all the
tests either equalled or surpassed the other groups). There was in
addition evidence of both text effect (some texts proving consist­
ently easier than others, though on a Fog Readability Index they
were equivalent) and of method effects.
In spite of such inconclusive evidence, it seems to us undeniable
that background knowledge has an effect on reading. While this is
probably true for all texts, it is most easily comprehended in rela­
tion to what, for some people, are highly specialised texts. Given
such a text on nuclear physics, for example, taken from a profes­
sional journal, it seems undeniable that a professional physicist
will read it differently from most of the readers of this book (as
well as the authors). We can state this in relativistic terms, and say,
as do Harri-Augstein and Thomas (1984), that our comprehensions
will be different, or we can be more absolute and claim that the
physicist’s reading is likely to be better. We think most people, after
comparatively little reflection, will be inclined to agree with this.
If, however, we accept that background knowledge is involved
in all normal reading, then we are obliged to account for the fact
that studies have not always been able to detect a significance dif­
ference brought on by apparent differences in knowledge brought
to the task by experimental subjects. There are at least three factors
to be discussed, and we shall now discuss this factor by factor,
referring to the two studies described above.

Texts
If our aim is to show, as it was in Alderson and Urquhart, that two
or more groups of readers will perform differently as a result of
66 Reading in a Second Language
differences in background knowledge, then it seems obvious that
the texts used should be as specialised as possible. Clearly, a text
which is equally accessible to both groups in terms of the know­
ledge required will not show any difference between the groups.
Clapham’s remark that it is only with highly specific texts that
background knowledge has an effect on student test performance
seems almost too obvious to make (Clapham, 1996a). In the first
two studies by Alderson and Urquhart, the Engineering texts were
chosen with the help and advice of an academic engineer. Even
so, on the whole they failed to discriminate between engineers, on
the one hand, and science and maths students on the other, pre­
sumably because of being insufficiently specialised in the direc­
tion of the engineers. If one uses parts of existing tests aimed at a
wide range of testees, as was the case in Study S of Alderson and
Urquhart, and also in Clapham (1990), then there is a danger of
the texts being insufficiently specialised, having been filtered by
test constructors and editing committees. This was certainly the
case for the IELTS tests used by Clapham. The focus of the re­
search then shifts to the question of whether, in cases like IELTS,
having ‘specialised’ ESP modules is justified. This is a worthwhile
question to ask, but tells us little or nothing about the effect of
background knowledge.
In fact, some of the conflicting findings in the literature may
be traceable back to a difference in focus on the part of the
researchers, leading to a difference in text selection. In the first
two studies by Alderson and Urquhart, the focus was on possible
differences caused by differing background knowledge. In their
third study, and in some at least of Clapham’s work, the focus
changes to whether the texts (and possibly tasks) of existing ESP
tests are successful in discriminating between different groups,
and are therefore worth using. It is perfectly possible to consider
that background knowledge is an important factor in reading
performance while at the same time being of the opinion that
broad-based ESP tests are probably not worth having.
In view of what has been said above, the text used in Moham­
med and Swales - instructions for setting a digital clock - seems
slightly anomalous. After all, it was not presumably aimed at
a specialised audience: digital clocks are widely used by non­
scientists. Presumably we have here a case of a sender of a mes­
sage who has not taken sufficient account of the skills of their
audience.
 The theory of reading 67
Subjects
It follows from what has been said above that if we are to find
significant differences between groups, the groups should be as
different as possible from each other in terms of the relevant
knowledge each group possesses. This is fine in theory but often
difficult in practice. Suppose we try to use two groups, one of post­
graduate management students, the other of engineers. One often
finds (generally after the experiment) that some members of the
management group have prior training as engineers. Mohammed
and Swales operated with two groups, scientists and arts students.
Yet the distinction is a very rough and ready one. Many linguists
would wish to describe themselves as scientists, but Mohammed
and Swales classified students of applied linguistics as arts students.

Tasks
The test tasks used by Alderson and Urquhart in Studies 1 and 2
were of the form of gap-filling and short-form answers. Such tasks
have the advantage that they are easily designed and administered.
They have the disadvantage that they are not particularly appro­
priate in terms of reading either to the texts or to the readers.
Hence, if no difference is found, suspicion may fall on the tasks.
Bernhardt (1991b), for example, has suggested that the frequent
use of cloze procedure may be a factor in obscuring the effect of
background knowledge. An alternative is to use tasks which are
functionally appropriate both to the text and to the readers. Pre­
ferably, they should be composed by members of the discourse
community which uses the text. That is, if the text relates to, say,
architecture, then the task should ideally be devised by architects.
Without having to go through all the difficulties this is likely to
entail, Mohammed and Swales score highly in this respect, since,
as they point out, the appropriate task relating to instructions is
for the readers to carry out these instructions. Otherwise what we
are likely to get is either cloze tasks, which Bernhardt has charac­
terised as ‘a syntactic/productive measure of clausal knowledge’,
or ‘comprehension’ tasks devised by item writers with a training in
EFL or applied linguistics, which may be fine for describing what
a typical EFL teacher may extract from a text, but hardly suitable
to map out what a specialist may learn from it.
68 Reading in a Second Language
Language level
Many of the studies in this area refer to the language proficiency
level of the students. In Alderson and Urquhart’s studies, for
example, though a language factor was not built into the design
of the experiments, it was noted that, according to earlier pro­
ficiency tests, liberal arts students were more proficient than
development and finance students, who were in turn superior to
the engineers. In Mohammed and Swales’ design, language (or
general reading proficiency) was integrated more closely into the
design, though the m ethod of ascertaining it, by asking teachers
to estimate ELTS reading scores, was to say the least subjective.
Ridgway (1997) argues that the level of language proficiency is
crucial, and differences in level may have masked the background
knowledge effect in some cases. Certainly, Alderson and Urquhart’s
LA group equalled the engineers on engineering texts in two out
of three studies, presumably because of their higher language
proficiency. Ridgway, like Mohammed and Swales, argues for a
threshold linguistic level, below which any relevant background
knowledge cannot be brought into play, and this seems reason­
able (see below for threshold effects). We have less sympathy with
his claim that there is also an upper threshold level beyond which
the readers’ language ability is sufficient to allow them to read any
text with equal success. This, we feel, runs counter to our experi­
ence of subject-specialised texts.

Schema theory
Finally, we should touch on the vexed issue of background know­
ledge itself and what it consists of. Fairclough (1995) has criticised
some discourse analysis because of the assumption that back­
ground presuppositions are ‘neutral’, or in accord with some kind
of objective reality. Instead, he points out that such presupposi­
tions may represent the views of ideological groups. While he
seems to us to make a good case for his position, the criticism
cannot be applied in much of the area we are discussing. This is
not because advocates of the role of background knowledge are
not vulnerable to criticisms of the type Fairclough raises. It is
rather that background knowledge is often not specified in suffi­
cient detail to enable the presuppositions to be examined.
 The theory of reading 69
Carrell (1983b) distinguishes between formal and content sche­
mata, i.e. knowledge about (a) the rhetorical structure of texts
and (b) the content Both have been shown to have an effect at
times on reading performance. Mohammed and Swales (1984),
for example, are inclined to attribute their results to ‘familiarity
with the genre of technical instructions’ (p. 211), more or less
equivalent to Carrell’s formal schemata. We, however, prefer to
discuss this aspect of background knowledge under the heading
of ‘Literacy’ (below). Bernhardt, as we mentioned above, divides
studies into those concerned with cultural knowledge, subject-
specific content, and information supplied to readers shortly
before reading. In terms of the individual, we see no harm in
grouping the first two types of knowledge together. We suspect,
however, that information supplied to readers shortly before they
read a text is likely to play a different part in reading from that
played by longer established knowledge. We would like, then, to
consider this under teaching methodology, and concentrate on
well-established knowledge.
There are two related problems here: (a) to define in some way
what it means to say, for example, that someone ‘knows’ chem­
istry; and (b) to test the person’s knowledge. Clearly the answer to
(a) must be more than just a collection of facts: it must include
relationships between ‘facts’, some idea of the purpose of the
pursuit, possibly of the history of the subject, and future applica­
tions. Equally obviously it cannot be determined in terms of vocabu­
lary. As far as (b) is concerned, some form of test, such as the
free-association tests used by Langer (1984), might be considered.
However, while tests might successfully establish whether or not
a reader was already well informed about limited topics such
as cricket or baseball, it is difficult to imagine an easily adminis­
tered test of an adult’s knowledge of, say, production engineering.
Given this problem, one tends to fall back on the sometimes naive
assumption that if readers have already completed several years’
study of an academic subject, then they will possess a store of
background knowledge about it. This is the assumption made by
Alderson and Urquhart. Given the uncertainty of a lack of ability
to define the crucial variable, it is hardly surprising that some
experiments fail to come up with positive results.
In spite of all this, however, there is enough evidence in the
literature to support the theory that background knowledge plays
a crucial part in the reading process.
70 Reading in a Second Language
Studies of the effect of background knowledge, when they find
a positive effect, provide evidence that such knowledge can legit­
imately be considered a component of reading. They tell us noth­
ing of the process, i.e. what is going on to produce this effect. For
any sort of answer to this, we are usually referred to schema theory.
This theory has been extensively described in the reading liter­
ature; in fact it sometimes seems to be obligatory for anyone writ­
ing a thesis on reading to begin with a lengthy description of the
theory, beginning with Kant (1781), moving to Bartlett (1932),
then to Rumelhart (1980). Bartlett found that English subjects,
given a North American folk tale, were unable to comprehend or
remember parts of it. This was in spite of the fact that none of the
words or sentences was linguistically unfamiliar or senseless. It
appeared that for comprehension and remembering to take place,
the linguistic input needed to match existing mental configura­
tions or concepts. Input which did not match the configuration
was not remembered, even though it presented no actual lin­
guistic difficulty.
While the notion of such configurations, or schemata, seems
very attractive, there are huge problems attached. Sadoski et al.
(1991: 466) quote Bartlett as saying:
I strongly dislike the term ‘schema.’ It is at once too definite and
too sketchy.
Below are some reasons for believing that schemata are not very
useful in reading research (or possibly, by the ease with which
they can be invoked in any number of situations, too useful):
1. Schemata are often described as being ‘structures’ or ‘tem­
plates’, and are often seen as being hierarchical (e.g. Collins
and Quillian, 1969). Rumelhart (1980), on the other hand, sees
schemata as being fluid and constantly capable of adapting to
fresh information. Bartlett, also, in the excerpt referred to by
Sadoski et al., refers to the need to invoke ‘active, developing
patterns’. But a constantly changing template is n o t likely to be
a very useful instrument. In fact, the need for schemata to be
structured in advance, yet adaptable to text-driven alterations,
has been a problem for schema theorists from the beginning.
2. It has been argued that the term ‘schema’, as commonly used,
is virtually synonymous with ‘background knowledge’, and hence
is useless (cf. Sadoski et al., 1991).
 The theory of reading 71
3. Related to this is the odd fact that, at least in the L2 research
literature, while schemata are frequently appealed to, they are
seldom described in any detail. Compare the more rigorous
experimental investigations of prototype theory, particularly the
work on the cognitive representations of semantic categories
by psychologists such as Rosch (1975) and Rosch et al. (1976).
Thus L2 researchers invoke experimental subjects’ possession,
or lack of possession, of schemata related to weddings, Christ­
mas, etc., without ever giving a description of what is contained
in such schemata. Given that schemata are simultaneously de­
scribed as ‘structures’, this is very odd indeed. It is not always
the case that such description is missing. In the theoretical
literature we find some illuminating descriptions of hierarch­
ical structures, either of single vocabulary items, e.g. for the
item ‘canary’ in Collins and Quillian (1969), or for an event
such as a ‘ship christening’ in Anderson and Pearson (1988).
But such fairly detailed structures, while admirable and cap­
able of being tested, raise suspicions immediately. For example,
the ‘canary’ schema has, attached to the ‘bird’ node, the fact
that a bird ‘has wings’, ‘can fly’ and ‘has feathers’, but not that
it has a beak or builds nests. The ‘ship christening’ schema,
which is a very loose ‘structure’, and basically in fact is just a set
of unordered components, contains the information that the
christening takes place ‘in dry dock’. But how many readers
are likely to know this?
4. In addition to such lack of explicit description, L2 researchers
entertain remarkably loose notions of the whole concept, so
that schemata can be ‘activated’ or even ‘acquired’ at the drop,
so to speak, of a short passage of introductory reading. But if
the term is to have any use at all, then surely it must describe
mental constructs of some stability, developed over some time
by a sizeable portion of a population.
In the reading literature, different types of schemata have been
suggested. We have already referred to Carrell’s distinction be­
tween: ‘content schem a(ta)’, relating to the content of a text read;
‘formal schemata’, relating to the rhetorical structure of the text;
and ‘cultural schemata’, more general aspects of cultural know­
ledge shared by large sections of a cultural population. Carrell
(1988a) has also added ‘linguistic schemata’. W hether it is in real­
ity useful to apply the same term to notions as different as, say,
72 Reading in a Second Language
our knowledge of the passive voice, of behaviour at a wedding, of
birds, of the meaning and purpose of life, or of newspaper art­
icles, is questionable. Here we prefer to use the term ‘background
knowledge’ for content or cultural schema. Formal schemata we
prefer to deal with under Bernhardt’s ‘literacy’ component, and
linguistic schemata under different areas of language. For a detailed
treatment of schemata, the reader is referred to the work of
Cavalcanti (1983).

Threshold levels
Mohammed and Swales found that one of their NNS/Science
subjects, whose estimated proficiency in English was low, was unable
to perform the tasks, although presumably the subject had the
necessary background knowledge. They put this down to the exist­
ence of a threshold level, which in this instance they locate at
about Band 5 of the IELTS scale. In terms of an interactive model,
what this amounts to is a claim that there is a level below which a
deficit in one component cannot be compensated for by a corres­
ponding strength in another. The term was first used by Clarke
(1979) to account for aspects of data gathered from Spanish speak­
ers reading English. Clarke hypothesized that some of his subjects,
who were good readers in Spanish, were unable to transfer those
reading skills because of inadequate mastery of the L2. Recently
Ridgway (1997) has again invoked the threshold effect, this time,
like Mohammed and Swales, to explain why a group of Turkish
readers were unable to utilize background knowledge in reading
English. Thus the threshold level has been used to explain why
either background knowledge, or ‘reading skills’ (possibly equatable
with Bernhardt’s ‘literacy’ component) are unable to compensate
for a lack of linguistic proficiency (Bernhardt, 1991b).
The notion of a threshold level seems commonsensical: no
matter how good our reading skills are in the LI, or how expert
we are in the content area, we are not likely to make much of a
text in a language which is totally unknown to us. The mistake is
to imply or infer that there is a general linguistic threshold level,
valid for all tasks and all subjects. In fact, it seems obvious that
some tasks will require a higher threshold level than others. It is
probably also true that some subjects are able to make more of
their limited linguistic proficiency than others. Thus the thresh­
old level must be ‘reset’ for each subject or group of subjects, and
 The theory of reading 73
each set of tasks. Given this limitation, it is a constraint which
experimenters (and teachers) should keep in mind.

Literacy
Our final component is again taken from Bernhardt. By ‘literacy’,
she means operational knowledge: knowing how to approach text,
knowing why one approaches it and what to do with it. It includes
the reader’s preferred level of understanding, goal setting and
comprehension monitoring.
Under this heading, we include both ‘cohesion’ and ‘text struc­
ture’. Both decisions may seem slightly controversial: Halliday and
Hasan (1976) would clearly class cohesion as a part of language
knowledge. And Alderson (1993) is not unusual in deciding to
including cohesive items in a test of grammar. De Beaugrande
(1980), however, criticises an exclusive focus on the linguistic ele­
ments because of the lack of consideration paid to ‘the underly­
ing connectivity of text-knowledge and world-knowledge that makes
these (cohesive) devices possible and useful’ (p. 132). The rela­
tionship between cohesive elements and text knowledge seems a
good argument for including cohesion under ‘Literacy’.4
As far as text structure is concerned, Carrell’s labelling of know­
ledge of such structures as ‘formal schemata’ might suggest that
this topic should be included under the general heading of ‘back­
ground knowledge’. However, to return to Bernhardt’s formula­
tion of literacy, ‘knowing how to approach a text’ must surely
include knowledge of what kind of text it is, and hence how it is
likely to be structured.
It would also be in line with Bernhardt’s description to place in
this section an account of readers’ strategies. However, for various
reasons, we have chosen to discuss these in Section 2.3. We should
like the reader to note, however, that strategies clearly, in our
opinion, form part of the literacy component.
Decisions as to where to place various elements are not just
part of authorial housekeeping. We argue below that some of the
discussion of reading skills versus language skills has been vitiated
by vagueness as to what ‘reading skills’ actually consist of. It seems
to us that Bernhardt’s ‘literacy’ component is the best place to
look for distinctively reading skills. Hence, what we decide to in­
clude in this component becomes crucial in an interesting and
important area of research.
 74 Reading in a Second Language
Cohesion
For de Beaugrande (1980), ‘cohesion subsumes procedures where­
by surface elements appear as progressive occurrences such that
their sequential connectivity is maintained and made recoverable’
(p. 19). In the 1970s and early 1980s, there was considerable
interest in the effect of cohesion on L2 reading, and many
books designed for classroom use, such as the Focus series (e.g.
Glendinning, 1974), contained exercises designed to train readers
in responding to cohesive devices in texts. Teachers, in our experi­
ence, have not always been convinced of the usefulness of such
exercises. There is, moreover, comparatively little good research
in this area. The topic, in fact, is rather more difficult and obscure
than is sometimes recognised.
There is a hint in de Beaugrande’s definition that ‘cohesion’ is
a cover term, and this introduces a problem for the researcher.
Different cohesive procedures may have radically different func­
tions. The most obvious difference is that between Conjunction,5
whereby a cohesive device indicates the pragmatic relationship
between two text utterances or blocks, and devices such as Refer­
ence, Substitution, and Ellipsis, where the cohesive item replaces
previously occurring parts of the text. The skills employed in hand­
ling these two groups are likely to be very different, raising the
( question as to whether it is desirable to investigate the effect of
‘cohesion’ seen as a homogeneous entity. The actual function
played in texts by the different elements is not obvious, and it
might be useful to distinguish between writer functions and reader
functions. It might seem that the function of conjunction is not
difficult to account for: by making the relationships between text
units more transparent, the presence of conjunctive items might
be expected to make texts more transparent for the reader, and
hence easier to read. It is true that both Meyer (1975) and
U rquhart (1976) found that, in the case of native speakers, mark­
ing the relationships did not seem to effect recall of a text. On the
other hand, Cohen et al. (1979) found that with a reasonably
extended text, native speakers of English structured their under­
standing in part by depending on conjunctives, whereas the non­
native readers failed to appreciate the relationships signalled by
the conjunctives.
As various writers have pointed out, however, a sequence of
text units may be coherent without the conjunctive item formally
 The theory of reading 75
signalling the relationship. Thus the importance of conjunction
on any particular occasion is open to question, varying as it is
likely to do between readers. Steffensen (1988) argued that if
cohesion was weakly related to coherence, recall of ‘native’ texts,
i.e. relating to the culture of the reader, being more coherent,
should contain more cohesive items than recall of corresponding
‘foreign’ texts. The hypothesis was not confirmed and Steffensen
concludes that the formal teaching of cohesive devices in L2 read­
ing should be treated with caution. We would argue, however,'
that her data suggest that the use of conjunction here is writer-
focused, the writers using conjunctions to try to make sense of the
text they are producing.
Urquhart (1976) found that academically gifted LI teenagers
introduced conjunctives when recalling texts which had not ori­
ginally contained them. This, and Steffensen’s results, then, serve
to remind us that cohesion may be at least as important for the
writing class.
The textual function of the other main group of cohesive
devices, Pronominal Reference, Ellipsis and Substitution, and its
relationship to reading, is perhaps even more problematic. Differ­
ent writers have suggested different functions: continuity (Halliday
and Hasan, 1976); economy (de Beaugrande, 1980); foregrounding
(Chafe, 1972). We find it easier to describe such functions from
the writer’s point of view: ‘Use pronouns to be economical and
avoid repetition’, etc. From the point of view of the reader, the
effect of such cohesion is more difficult to define. With respect to
economy, de Beaugrande refers to the trade-off between compact­
ness and rapid access, i.e. pronominal reference is compact but
ambiguities in reference may confuse and delay access. The final
effect on readers may depend on individual skills, language profi­
ciency, and knowledge of the world. The effect of foregrounding
might seem to be even more difficult to assess.
In some cases, the effect of cohesive items may be very much
on the surface. Cohen et al. (1979) report that in some cases their
subjects simply did not know the meaning of conjunctives such as
‘thus’. Some similar lack of surface familiarity may have been
responsible for the situation, reported by Berman (1984), that
L2 readers preferred texts in which pronominals and substitu­
tion items had been replaced by their lexical equivalents. The
readers may just not have been familiar with the use of the co­
hesive items.
76 Reading in a Second Language
As far as the question of how readers handle cohesion is con­
cerned, psychological research in the LI area has concentrated
on the relative difficulty of identifying antecedents. One hypo­
thesis is that the distance between antecedent and pronoun will
cause processing difficulty (this hypothesis is implicit in Halliday
and Hasan’s description, with its taking account of ‘distance’ in
terms of number of sentences, and ‘mediated ties’ (sequences
of pronouns all with the same antecedent). Rayner and Pollatsek
(1989: 273), reviewing the evidence, conclude that ‘pronoun
reference ... is governed not only by linguistic rules but by a looser
set of discourse guidelines . . . based on the type of verb, parallelism
of form, and whether the noun is still the topic of the discourse’.
Kintsch and van Dijk’s model of reading (see below) relies
heavily on repetition to establish overlap and hence coherence
between propositions (Kintsch and van Dijk, 1978). Depending
on how one defined it, repetition could include, in Haliday and
Hasan’s terms, ‘lexical reiteration , ‘collocation , major aspects of gram­
matical reference and substitution. We are a little cautious about such
a seemingly crude approach to coherence, but there seems little
doubt that repetition must be involved in readers’ perception that
the writer is continuing to talk about ‘the same thing’.
It would thus seem very likely that cohesive procedures on the
part of the reader have effects on reading performance. In both
LI and L2, however, the investigation of these effects would seem
to require more subtlety than has been evident up to now. The
importance of the teaching of cohesive procedures in the writing
class should perhaps be emphasised.

Text structure
Brown and Yule (1983) point out that some of the coherence of a
text derives not so much from the presence or absence of surface
cohesive features such as conjunctives, but from underlying text
relationships to which the conjunctives are pointers. From the
1970s onwards, several models have been available which attempt
to map underlying coherence in text. Meyer (1975) points to a
distinction between models which take into account the author’s
organization of the text being analysed, and those which impose
another type of organization. Meyer has in mind analysts like
Crothers (1972) who impose a form of logical structure on texts.
Davies also ignores the author’s organization, basing the analysis
 The theory of reading 77
on types of information found in texts, so that, for example, ‘physi­
cal structure’ texts contain information about part, location, prop­
erty and function (Davies and Greene, 1980; Davies, 1983; Johns
and Davies, 1983). In the discussion below, we shall concentrate
on models of the first type. We have selected descriptions which
are (a) similar enough to each other to make for, hopefully, a
coherent discussion and (b) dissimilar enough to make compari­
son interesting. Of the three selected, the models used by Meyer
(1975) and Kintsch and van Dijk (1978) have been used in read­
ing experiments; the ‘composite’ model formed by the work of
Hoey (1983) and Winter (1994) has been offered as ‘pure’ text
analysis, independent of reading.
Our models typically consist of (a) some form of ‘unit’ out of
which the larger structure is constructed, (b) a set of relationships
between such units and (c) a larger, global structure, to which the
more local structures are in some way related.
Thus in the work of Winter (e.g. Winter 1994) and Hoey (1983)
the units are natural language clauses, and the local relationships
include Generalisation/Exemplification, and Denial/Correction.
In addition to these ‘clause relations’, Winter and Hoey refer to
basic text structures, such as Situation/Problem/Solution. As far as
the relations between clause and text structures are concerned,
Winter is not very explicit. Hoey (1983: 57) provides suggestions
for mapping clause relations on text structure, along the lines of:
If a Cause/Consequence relation consists of a and b, and a is
identified as a Problem, then if b contains the role of agent, b is
Response.
We don’t know whether such mapping rules have ever been sys­
tematically tried on extended texts.
Meyer’s (1975) model is taken from the linguist Grimes (1975).
The basic unit is the proposition, consisting of a predicate and one
or more arguments. There are two kinds of proposition: lexical and
rhetorical. In the first, arguments are related to their predicates by
semantic roles such as agent, patient, range, etc. (cf. Fillmore,
1968). The second kind are rhetorical propositions.
Their main function could be thought of as that of organizing the
content of discourse. They join lexical propositions together, and
they join other rhetorical propositions together.
(Grimes, 1975: 207)
78 Reading in a Second Language
One might reach the conclusion that lexical propositions oper­
ated up to clause level, while rhetorical propositions took over to
link clauses or sentences. This is not strictly true, since rhetorical
predicates can occur within clauses: ‘the rhetorical predicates attri­
bution, specific, collection and equivalence are frequently found
in simple sentences’ (Meyer, 1975: 45). They do, however, tend to
be superordinate in text structure. According to Grimes (p. 207)
In a tree that represents the underlying structure of a discourse
. . . most of the propositions near the root are likely to be rhetorical,
while most of the propositions near the leaves are likely to be lexical.
Rhetorical propositions are divided into three types: paratactic,
hypotactic and neutral In paratactic propositions, both arguments
are at the same ‘level’; a typical paratactic predicate is collection, of
the sort ‘[there is] A and B and C’, in which A, B and C are equal
in rhetorical level. In a hypotactic proposition, one argument is
superordinate to another. Thus in the Evidence predicate, the
evidence argument is subordinate to the argument for which it
supplies evidence. Neutral predicates can be either paratactic or
hypotactic.
Since hypotactic predicates have the effect of subordinating
one argument to another, and, more generally, since an argu­
m ent can consist of a proposition which can include arguments
which . . . etc., the result of the analysis of a text is a ‘hierarchically
arranged tree structure’ called a ‘content structure’.
There are some quite close resemblances between Meyer’s ana­
lysis and that of Winter and Hoey. Relationships such as ‘Cause/
consequence’, ‘General/specific’, occur in both, though with dif­
ferent terminology. ‘Problem /solution’ again is present in both
analyses. However, in Meyer, it is a rhetorical predicate, capable
of appearing at different levels of the content structure; in Hoey it
is a basic text structure, seemingly different from clause relations
In general, in the 1975 account, Meyer does not distinguish
between local and global relations. Meyer and Rice (1984: 326)
refer to three levels, with micropropositions at sentence level,
macropropositions at paragraph level, where ‘the concern is with
the relationship among ideas represented in complexes of ideas
or paragraphs’, and the third level is ‘the overall organising prin­
ciple’ of the text, e.g. causality, problem/solution, etc. (p. 327).
Before begining our account of our third model, that of Kintsch
and van Dijk (1978), we should point out that it is much more
 The theory of reading 79
than a description of text organization. In fact, it sets out to model
a large part of the process of reading and remembering an ex­
tended text. As such, it contains a considerable amount of discus­
sion about how the reader proceeds to take in a certain amount
of information at a time, and operate on this information while it
is in working memory before proceeding to the next chunk of
information. It also deals, in rather less detail, with how the reader,
simultaneous with processing the clauses and sentences of the
text into a coherent whole, the text base, also proceeds to build
up an account of the gist of the text, the macrostructure, or rather,
given the constraints of memory and the cyclical nature of the
processing model, a sequence of macrostructures. To the extent
that Meyer also tested her model empirically, this constitutes a
similarity between her and Kintsch and van Dijk, and a major
difference between these writers, on the one hand, and Winter
and Hoey on the other.
As with Meyer, the basic unit of analysis for Kintsch and van
Dijk is the proposition, consisting of a predicate and arguments.
Predicates ‘may be realized in the surface structure as verbs, adject­
ives, adverbs, and sentence connectives’ (Kintsch and van Dijk,
1978: 367). Thus the distinction drawn by Grimes and Meyer be­
tween lexical and rhetorical predicates is not observed. Proposi­
tions are expressed in the form ( b e t w e e n , e n c o u n t e r , p o l i c e ,
b l a c k p a n t h e r ) , which represents the text ‘encounters between
police and Black Panther Party members’ (p. 377). The system
incorporates a set of semantic role relationships similar to those
described by Grimes, but in the 1978 article, these are not indic­
ated, ‘to keep the notation simple’, and in fact, Kintsch and van
Dijk’s trees are much easier to read than Meyer’s.6
These micropropositions are built up into a structure referred to
as a text base, or microstructure, which can be depicted as a coherence
graph. Such graphs are very similar to Meyer’s content structure.
Coherence is maintained largely by referential coherence, depicted
in the graph as overlap of arguments. Thus the two propositions
‘ABC’ and ‘XDC’ achieve coherence by the presence in both of
the argument ‘C’. Sometimes coherence cannot be detected in
this way, and then the reader may need to generate inferences
to maintain coherence. Thus the total list of propositions may
be longer than that contained in the text. Propositions are also
arranged in terms of ‘level’, i.e. some propositions are superordin­
ate to others. This ordering is done partly in terms of the simplicity
80 Reading in a Second Language
of the resulting graph structure, and partly in terms of coherence
relations. If, for example, proposition 17 has been nominated as
superordinate, and propositions 15, 10 and 13 relate cohesively to
17, then they are all considered as subordinate to 17. If, then,
propositions 20 and 23 relate cohesively to 13, they are in turn
subordinate to it.
The other main organizational component in Kintsch and van
Dijk’s model is the macrostructure. If the microstructure is ‘the
local level of the discourse, that is, the structure of the individual
propositions and their relations’, the macrostructure is ‘of a more
global nature, characterising the discourse as a whole’ (p. 365).
Thus Kintsch and van Dijk, like Winter and Hoey, have two sep­
arate, though connected levels of organization. Kintsch and van
Dijk see the macrostructure as being built up at the same time as
the microstructure; in other words, the former is not a summary of
the latter. The macrostructure is formed partly by the application
of macrorules, which operate on the microstructure, for example
deleting irrelevant propositions, or substituting generalisations for
sequences of detailed propositions. The propositions so derived
are organised by a schema, which is brought by the reader into
contact with the microstructure.
On the whole, Kintsch and van Dijk deal with schemata as ‘con­
ventional schematic structures of discourse’ (p. 366), equivalent
to Carrell’s ‘formal schemata’. However, they do make allowances
for a different form of schema.
The reader’s goals in reading control the application of the macro­
operators. The formal representation of these goals is the schema.
(p. 373)
In other words, the schema is produced in accordance with the
reader’s goals in reading. Given this view of schema, Kintsch and
van Dijk (1978) envisage three situations. In the first, ‘a reader’s
goals are vague, and the text that he or she reads lacks a conven­
tional structure’ (p. 373). In this case, the schema invoked and
the macrostructure would be unpredictable. This is the form of
reading which we shall later be discussing under the term ‘brows­
ing’ (p. 103). In the second situation, the text type is highly con­
ventional, and this in turn sets clear goals. In the third situ­
ation, the goals are again clear, but are set by the reader who has
a special purpose in mind, which may override the text structure.
Such a reader is discussed below as a ‘dom inant’ reader.
 The theory of reading 81
Empirical validation of the text structure models
Various aspects of the models above have received some measure
of experimental validation. Kintsch and Keenan (1973) found that
sentences became more difficult to read and understand in rela­
tion to the num ber of propositions they contained: the more pro­
positions, the more difficulty (see also Weaver and Kintsch, 1991).
Meyer (1975) used her model to investigate the type of informa­
tion recalled by subjects after reading an extended prose text.
Information which the model showed to be at a higher level in
the text was recalled better than lower level information. At the
overall organization level, Stanley (1984) found that both natives
and non-natives preferred summaries organised according to a Prob­
lem /Solution model, as opposed to texts of the same length and
linguistic difficulty which deviated from the model. Carrell’s findings
(Carrell, 1984) that familiar rhetorical organisation appeared to
help readers, can be added here. Rayner and Pollatsek (1989) again
cite evidence that readers’ recall of the gist of texts tended to
resemble the macrostructure.
Conclusion
Clearly an organized text is more than a string of clauses or sen­
tences or propositions. Equally clearly, this fact is likely to be
relevant to the reading process. All the descriptions of text struc­
ture above seem to us to have something to contribute in this area
and the most useful will be discussed further in Part 3 in relation
to choice of texts for testing. However, before we leave the topic,
some criticisms remain to be dealt with.
The use of propositional analysis by Kintsch (1974) and van
Dijk (1977) has been criticised by Brown and Yule (1983) on the
grounds that, in spite of its appearance of formalism and objectiv­
ity, it is fundamentally subjective. While this is probably true, it
does not seem to us to be exceptional. Most interesting analysis of
natural language rests on consensus; the most formalized genera­
tive grammar is supposed to be constructed on the basis of native
speakers agreeing on whether such and such a sentence is accept­
able or not. The fact that they are seldom consulted is beside the
point. It is fairly easy to check the extent to which two or more
analysts agree.
Brown and Yule also criticize such analyses for concentrating
on content, and for being unable to deal with staging. This again
82 Reading in a Second Language
is probably true; Meyer (1975) comments on the inability of
Grimes’ analysis to incorporate staging successfully. From our point
of view, given that our main focus is on students’ learning from
texts, it is possible that staging is of minor importance in compar­
ison with conventional text structure. This would not be sur­
prising: we have already noted how another linguistic aspect of
writing, namely signalling, seems to have comparatively little effect
on the reader.
Rayner and Pollatsek query whether the elaborate apparatus
used by Kintsch and van Dijk is justifiable, or whether it could be
replaced by something less formal, available to the ‘intelligent lay
reader’. This certainly seems to be a valid criticism of an analysis
such as Meyer’s, which is difficult to read, made scoring of sub­
jects’ recall scripts, on her own admission, an ‘extremely tedious’
task (p. 101), and which is made complex by the inclusion of case
roles which could not be correlated with any effects in the recalls.
The complex formality is particularly dubious, given the compar­
atively crude description of how to arrive at the analysis, e.g.
The topic sentence of the third paragraph of this passage states
that the breeder reactor is the solution to the previously stated
problems. (p. 54)
In Section 5.3 (pp. 275-7), we put forward less formal ways of
analysing texts for our particular purposes.
Finally, we should like to make a comment on specific patterns
of organization in texts, referred to and to some extent critical for
all the analyses above. The question is: How finite are these differ­
ent patterns of organization? We have seen that Kintsch and van
Dijk’s macrostructures seem in part to depend on what they refer
to as ‘highly conventionalized text types’. Such text types unques­
tionably occur. It is our experience, however, based on teaching
the analysis of written texts, either that the number of such text
types is very large indeed, or that a number of the texts encoun­
tered are indeterminate as to overall text structure. Hoey (1983:
34) argues that ‘the number of discourse patterns that can be
built out of a finite set of relations signalled in a finite num ber of
ways is indefinitely large’. W hether the first part of this claim is
true (Hoey appears to be rather coy about listing interclausal
relations), we tend to agree with the second part.
 The theory of reading 83
Text types
Text types have been referred to above in relation to Kintsch and
van Dijk’s model. With their reference to ‘highly conventionalised
text types’, it seems that these authors have in mind something
like Swales’ genres (Swales, 1990). Just because they are so spe­
cific, Swales’ genres seem to us more useful in accounts of writing.
In our discussion here, we concentrate on an older tradition of
describing rather generalised types which, in our opinion, are of
more general relevance for reading.
De Beaugrande (1981: 307) asserts that ‘. . . reading models
will have to find control points in the reading process where text-
type priorities can be inserted and respected’. The implication is
that the different textual and communicative demands of differ­
ent text types will affect reading performance, and further that
some readers may be limited with regard to the types they can
handle.
De Beaugrande mentions Narrative, Descriptive and Argument­
ative types. Calfee and Curley (1984) have Object, Sequence and Idea,
in which Sequence relates to Narrative, Idea to Argument, and Object
is loosely related to Description. Moore (1980) adds Exposition and
Enquiry, and Brooks and Warren (1952) have Exposition, Narrative,
Description and Argument as the four ‘basic types of writing’.
An examination of these writers suggests that there are at least
four criteria involved in defining text types.
■Communicative intent. Thus Brooks and Warren define Argu­
m ents ‘the kind of discourse used to make the audience . . . think
or act as the arguer desires’.
■Content. Calfee and Curley’s Object category is defined as ‘dis­
cussions of things, persons and even ideas’.
■Structure. Calfee and Curley define Sequence as ‘to do with any
account in which progression is the key to the structure’.
■Status of the information. Moore argues that Exposition ‘presents
knowledge already established’, while Enquiry is concerned ‘to
raise questions, . . . and express doubts and possibilities’.
There is an interesting attempt at a taxonomy using communicat­
ive intent, status of information, and expected response from reader
in Baten and Cornu (1984). A development of their taxonomy
relating to Expository texts is presented in Urquhart (1996).
84 Reading in a Second Language
There appears to be little work done on the possible differ­
ential effects of text types on readers. However, individuals often
report that they prefer fiction (imaginary narrative) or are poor at
reading instructions. McCormick (1992) hypothesized that nar­
ratives should be easier than expository texts. This was not con­
firmed but she considers that background knowledge is more
important in the case of expository texts. Reading tests, such as
that contained in IELTS, use a fairly informal categorisation of
text types. A combination of different text types with tasks suitably
tailored for particular types (Narrative would seem to invoke a
different set of responses from Exposition) has been put forward
by Urquhart (1996) as a potentially rewarding area of research
(cf. Kobayashi, 1995). We shall revisit the issue of text types in
Chapter 3 when it will be considered as a performance condition
in testing activities.
Having examined in some detail the components of the various
models we need to explore how they might be operationalised by
readers interacting with text(s) for different purposes. As well as a
concern with the nature of text readers may engage with, we need
to look in more detail at the process and product involved in and
resulting from such interactions.

2.3 COMPREHENSION, SKILLS, STRATEGIES

AND STYLES
The concepts that we discuss in this section are rather more dis­
parate than might appear at first sight. Comprehension is frequently
mentioned in cognitive and educational psychology, as well as, of
course, the pedagogical literature. There is often an assumption
in the literature that it is the goal of the reading process. As such,
it does not appear in the models discussed above. On the other
hand, much of what has been said about the components of
models could be translated into terms of skills: decoding is a skill,
accessing the lexicon is a skill, parsing syntactic structures is a
skill, and so on. The focus on Skills can be seen historically partly
as a development of interest in comprehension, an attempt to
break down that rather vague and undifferentiated concept into
more accessible chunks. Educationally it may be seen, at least in
part, as an attempt to give teachers a structure for both syllabuses
and for materials. Strategies appear to have come into Reading
 The theory of reading 85
Research via psychology, where they were used to describe how an
organism sought to attain its goals. In both reading research and
practice, a focus on strategies has had the effect of making the
whole operation more learner-centred. Finally, what we are call­
ing here Styles, also known as Strategies, have for some time been
an im portant topic in Educational Psychology. All four terms re­
late, in different ways, to reading activities and outcomes. The first
three, in particular, have been important foci of attention in
pedagogically related studies of reading, while the fourth is likely
to become of greater importance in future research.

Comprehension
As noted in Chapter 1 (Preliminaries), the focus of attention among
people concerned with reading in education was ‘decoding’,
whereas in the 1970s the focus moved to ‘comprehension’. This
can be seen clearly in Figure 1.2 above, based on ERIC data,
which shows comprehension studies taking off from 1966 onwards.
The switch of attention from decoding to comprehension must
have been very liberating for teachers and researchers, since,
although we have stressed the importance of decoding, there seems
little doubt that a focus on the information being communicated
by texts has more potential for interest. A focus on comprehen­
sion is in line with our feeling that this is what reading is ‘about’,
i.e. getting information from written texts. And there is no doubt
that our monitoring of our own reading comprehension is of
major importance. A judgem ent that we have not understood a
text may well leave us unsatisfied, or lead us to re-read it, or
perhaps reject it in disgust.
In spite of this, however, comprehension in some areas re­
mains a somewhat elusive entity. Rayner and Pollatsek (1989), for
example, give neither definition nor description of comprehen­
sion itself, though, according to their index, the larger part of the
chapter dealing with ‘Representation of Discourse’ is concerned
with ‘comprehension processes’. From the first part of their chap­
ter, one might gather that, for them, comprehension equals ‘ “the
meaning of the text” that is being read’ (p. 264).
It is, in fact, our contention that in the teaching and testing of
reading, ‘comprehension’, as generally defined, has been either not
very helpful or positively dangerous. Urquhart (1987) summarises
86 Reading in a Second Language
common assumptions behind the pedagogical view of compre­
hension as follows:
Assumption 1. There is such a thing as ‘total’ or ‘perfect’ compre­
hension of a text.
Assumption 2. Careful reading, which aims to extract perfect com­
prehension, is superior to any other kind of read­
ing, e.g. skimming, and is, in fact, the only kind of
reading which deserves the name.
Evidence of the existence of these assumptions is pervasive. As
evidence for Assumption 1 we may quote Fry (1963) as saying that
100 per cent on his comprehension exercises equals ‘perfect com­
prehension’; Sticht (1984) argues that claims for the possibility of
reading much faster than listening rest on a confusion between
skimming and scanning on the one hand and reading on the
other. Hence skimming and scanning, which can accept lower
levels of comprehension, are not really ‘reading’ at all.
W hether or not ‘perfect comprehension’ is a feasible goal, we
should reject the assertion by Fry and others that it is equivalent
to a 100 per cent score on comprehension questions. Even with a
short text, it is usually possible to devise a large num ber of ques­
tions. The conventional ten questions, often multiple-choice, which
pass as a comprehension test, represent at best a sampling of
information gained by reading. As Lunzer et al. (1979: 66) put it,
How a student completes a test is an INDEX of his capacity to
comprehend; it is not the capacity itself and still less is it the com­
prehension itself.
We have said that the typical pedagogical view is ‘dangerous’.
Firstly, by largely insisting on the superiority of one type of read­
ing at the expense of all others, it has the effect of disparaging
perfectly normal types of reading behaviour. This does not just
apply to reading types such as skimming and scanning (see Sec­
tion 2.4 below), where the view allows a drop in comprehension
in return for an increase in speed. By giving preference to what
we call ‘careful’ reading - i.e. the type normally associated with
study - it also effectively downgrades the value of the type of
reading behaviour many people will adopt when reading, say, de­
tective novels for enjoyment, where the reader’s monitor is likely
to accept lower standards of comprehension. Classroom reading
becomes almost exclusively ‘intensive’ reading (see Chapter 4),
 The theory of reading 87
and if classroom tasks have any influence on students’ behaviour
outside the classroom, this may well result in slow, laborious read­
ing when this is not, in fact, necessary.
The other danger lies in the assumption that a text contains
a finite am ount of information, accessible to all readers. The in­
formation is, in other words, ‘on the page’. This clashes with the
currently widely accepted view that the reader interacts with the
text in order to obtain a message, e.g.
Thus, contrary to conventional wisdom, which states that comprehen­
sion is the process of getting meaning from a page, comprehension
is . . . the process of bringing meaning to a text.
(Samuels and Kamil, 1988: 206)
This view has serious consequences both for the teaching and
testing of reading. If each reader brings meaning to a text, then
each comprehension is likely to be different. The notion of a
‘right’ answer has now to be treated with care. Variations in com­
prehension are likely to come from different background know­
ledge brought to the text (though this is not the only possible
source). In a classroom where teacher and students share the
same culture, such variations may not be very large. In EFL or ESP
classrooms in the English-speaking world, however, where teacher
and students may come from a wide range of backgrounds and
cultures, the possibility of varying comprehensions may become a
major problem. And if this is true for the classroom, it is even
more true for international EFL reading tests.
Urquhart (1987) distinguishes between ‘comprehensions’, re­
ferring to differences brought about by readers setting themselves
different levels of acceptable comprehension (i.e. between reading
a book for an examination and reading it for light amusement),
and ‘interpretations’, referring to differences resulting either from
different readers bringing different information to a text, or the
same reader at different times, bringing a different mind-set. While
the terms may not be perfectly chosen for keeping the different
factors apart, the distinction should serve as a reminder of the
number of variables likely to be present in many teaching or test­
ing situations (discussed in Chapters 3 and 4 below).
We should mention here the notion that the ‘ideal’ compre­
hension consists of the recovery of ‘author’s m eaning’. We do not
think that it can be doubted that readers often strive to do this; it
is an im portant aspect of careful reading, and, since it involves
88 Reading in a Second Language
close attention to textual features such as use of conjuncts (‘how­
ever’, etc.), headings, the ordering of information, and so on, it is
something that can partly be taught. We have only two doubts
about it being used as the ‘ideal’ comprehension. First, it can
never be fully achieved. We can never be sure that we have totally
entered the writer’s mind. It could be said, however, that it is in
the nature of all good ideals never to be achieved. Secondly, a
careful attempt to recover author’s meaning is not characteristic
of all reading; the reader engaged in scanning, for example, may
pay little attention to author’s intentions. Such attention is, in
fact, characteristic of careful reading, particularly where this is
submissive (see below). As such, it is important, but cannot be a
definition of comprehension in general.
We have just argued that ‘author’s meaning’ can never be recap­
tured in its entirety. It can be argued, however, that just recapturing
author’s meaning is not enough. Advocates of ‘critical reading’
(cf. Fairclough, 1995; Wallace, 1992a&b) point out that texts are
dependent on presuppositions stemming from their authors’ own
particular world view, their ‘ideology’. It then becomes the duty of
the critical reader, by spotting such ideological presuppositions,
to evaluate a text in its cultural context.
It is clear that comprehension cannot be viewed simply as the
product of any reading activity. Rather, in any reading situation,
comprehension will vary according to the reader’s background
knowledge, goals, interaction with the writer, etc. Comprehension is
a useful term to contrast with decoding, otherwise it is best perhaps
taken as the product resulting from a particular reading task, and
evaluated as such.

Skills
A reading skill can be described roughly as a cognitive ability
which a person is able to use when interacting with written texts.
Thus, unlike comprehension, which can be viewed as the product
of reading a particular text, skills are seen as part of the general­
ized reading process.
Skills have been a major area of reading research over recent
years, as can be seen in Figure 2.1, based on data from ERIC.
Skills have been recommended by Lunzer et al. (1979) and
Vincent (1985) as a means of structuring reading syllabi, and are
 3500-
- Skills
3000 ■ ■Strategies

2500 ■

2000 -

1500 -

1000 -

500-

n J
The theory of reading

1966 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
Figure 2.1 Number of articles and other publications published between 1966 and 1996 that mention ‘skills’ and
89

‘strategies’ in their title or in ERIC’s index or abstract (based on data from ERIC).
90 Reading in a Second Language
probably still the best framework for doing this. They have also
been used for test construction, notably in the ELTS test and
TEAP (Weir 1983a, 1990). Useful as the concept of skills has been,
there are considerable problems attached to it. Williams and Moran
(1989: 223) point out that while a number of skills taxonomies
exist, there is little consensus concerning the content of the
taxonomies or in the terminology used to describe them. Below we
give a selection of fairly typical taxonomies, which serve, inciden­
tally, to justify Williams and Moran’s comment.
1. Davis (1968):
■Identifying word meanings.
■Drawing inferences.
■Identifying writer’s technique and recognising the mood of
the passage.
■Finding answers to questions.
2. Lunzer et al. (1979):
■Word meaning.
■Words in context.
■Literal comprehension.
■Drawing inferences from single strings.
■Drawing inferences from multiple strings.
■Interpretation of metaphor.
■Finding salient or main ideas.
■Forming judgements.
3. Munby (1978):
■Recognising the script of a language.
■Deducing the meaning and use of unfamiliar lexical items.
■Understanding explicitly stated information.
■Understanding information when not explicitly stated.
■Understanding conceptual meaning.
■Understanding the communicative value of sentences.
■Understanding the relations within the sentence.
■Understanding relations between parts of text through lex­
ical cohesion devices.
■Interpreting text by going outside it.
■Recognising indicators in discourse.
■Identifying the main point of information in discourse.
■Distinguishing the main idea from supporting detail.
■Extracting salient points to summarise (the text, an idea)
■Selective extraction of relevant points from a text.
 The theory of reading 91
■Basic reference skills.
■Skimming.
■Scanning to locate specifically required information.
■Transcoding information to diagrammatic display.
4. Grabe (1991: 377):
■Automatic recognition skills.
■Vocabulary and structural knowledge.
■Formal discourse structure knowledge.
■Content/world background knowledge.
■Synthesis and evaluation skills/strategies.
■Metacognitive knowledge and skills monitoring.
It is comparatively easy to criticise some of these taxonomies,
even at first sight. Davis’s ‘Finding the answers to questions’ seems
to include all the others, and prompts the query: ‘Which ques­
tions?’ It is hard to believe that the assignment of separate status
to ‘Drawing inferences from single strings’, and ‘Drawing infer­
ences from multiple strings’ in Lunzer et al.’s taxonomy is really
justified. There are, however, wider questions to ask about the
taxonomies.
How inclusive is a skill? Clearly, in the taxonomies above, some
skills seem more inclusive than others; Grabe’s taxonomy, for
example, uses very general categories, virtually equivalent to know­
ledge areas. Rayner and Pollatsek (1989) begin their preface,
‘Reading is a highly complex skill. . .’ (p. ix). Clearly, if reading
itself is a skill, it must be possible to break this down into different
levels of component skills categories. Williams and Moran (1989)
suggest a rough distinction between ‘language related’ skills and
‘reason related’ skills.
Various attempts have been made to arrange skills into hier­
archies. Of the taxonomies above, that of Lunzer et al. is so
arranged, with the ‘lowest level’ skill at the top. Munby’s tax­
onomy was not intended to be hierarchically arranged, though in
his review of the work Mead (1982) argued that it should have
been, on the grounds that some skills seem to presuppose the
learning of other skills. Some possible criteria for ranking skills
are as follows:
(a) Logical implication. One component in the system can logically
be considered to presuppose all components below it. This is
the criterion used by Bloom et al. (1956, 1974).
92 Reading in a Second Language
(b) Pragmatic implication. A reader displaying one skill in the sys­
tem can be assumed to possess all the ‘lower’ skills.
(c) Difficulty. The components are arranged in order of increas­
ing difficulty.
(d) Developmental. Some skills are acquired earlier than others.
Some syllabi, rather unwisely, in our view, assume that readers
pass through a period of comprehending ‘explicitly stated’
information before they arrive at the stage of inferencing.
(e) Discourse level A skill is ordered with respect to the size or
level of the discourse unit it relates to. We have not found
explicit mention of this criterion in the literature, but suspect
it is commonly used by teachers and applied linguists. It would
explain a tendency to rate ‘thematic’ questions, aimed at the
whole of a text, as being ‘high level’.
There have been various attempts to investigate the psycholo­
gical reality (or separateness) of different skills. On the basis of
tests based on the taxonomy given above, Lunzer and Gardner
(1979) concluded that there was no evidence for the separate
existence of the skills, and that:
reading comprehension should not be thought of in terms of a
multiplicity of specialized aptitudes. To all intents and purposes such
differences reflect only one general aptitude: this being the pupil’s
ability and willingness to reflect on whatever it is he is reading.
(p. 64)
This is the so-called ‘Unitary Hypothesis’, as opposed to the ‘Multi-
divisible Hypothesis’. Lunzer et al. also investigated the pragmatic
validity of their hierarchy of skills, testing the hypothesis that:
there exists an identifiable group of pupils whose performance on
higher-level tasks is defective to a degree which would not be pre­
dicted on the basis of their performance on lower-level tasks.
(p. 61)
Again they found no evidence for this, although they did find
some evidence of a difficulty hierarchy. Further research in these
areas is discussed in Part 3 in relation to testing.
Before we leave the topic of skills, we can make the following
general comments:
1. The possession of a specialised comprehension skill, if we hypo­
thesise such a thing, does not guarantee success in completing
 The theory of reading 93
a particular task. We are all capable of making mistakes. This
argues for the need for a longitudinal study into the reality or
otherwise of such skills.
2. The conclusions by Lunzer et al., cited above, can only apply to
comprehension. Virtually everyone concedes that decoding is a
separate component/skill, since, as pointed out earlier, normal
young children can comprehend, without being able to de­
code, while some disabled individuals can decode without com­
prehending. Hence we have to accept at least a twin-skills model
(this discussion is taken further in Section 3.2 (p. 120).
3. Pragmatic validation is not the only form of justification for
skills taxonomies. Difficulty has been mentioned as a criterion
for a hierarchy, and most of us would presumably agree that,
on average, ‘Understanding the relations within the sentence’
is an easier skill than ‘Extracting salient points to summarise
(the text, an idea)’. The L2 learner, forced back to an earlier
developmental stage by the difficulties of unfamiliar syntax or
lexis, may well have to wait before being able to summarise
a text.
Finally, as said before, skills are useful tools for the development
of both teaching materials and tests. In spite of the doubts that
have been raised, we shall continue to make use of the taxonomies.

Transfer of reading skills

During the 1980s, there was considerable debate about the amount
of transfer of reading skills from reading in one language to read­
ing in another. The focus of the debate is encapsulated in the title
of Alderson’s article, ‘Reading in a foreign language: a reading
problem or a language problem?’ (Alderson, 1984). Since we are
taking the view here that reading can be defined as the receptive
handling of a language in written form, the dichotomy implied by
Alderson’s title does not really exist. Moreover, much of the dis­
cussion was poorly focused, since the ‘reading skills’ which were
supposed to be transferred across languages were seldom well
defined. This is discussed further in Section 5.2. In the light of
the discussion above, however, it would appear that ‘reading skills’
are best seen as Bernhardt’s ‘literacy’ component. The extent to
which the ‘literacy’ skills that she describes are transferable re­
mains to be examined. Perhaps the single most important literacy
94 Reading in a Second Language
skill, surely likely to be transferred, is the ability to view a printed
text as a piece of language discourse, to be viewed in a similar way
as an oral discourse.

Strategies
Reference to Figure 2.1 shows that strategy research is a later
development than skills research, only becoming popular in the
1980s. The research methodology is very different from that asso­
ciated with skills. In much skills research, the investigator begins
with a taxonomy of skills, arrived at, perhaps, by means of text
analysis. The psychological validity of this taxonomy may then be
empirically checked. In strategies research, on the other hand,
the researcher begins by having subjects (often divided into ‘good’
and ‘poor’ readers) read a text and, either retrospectively at the
end of reading or at points during reading, report on what they
are doing. The strategies revealed by these reports are then cat­
egorized. If a prior division of subjects has been made, an attempt
is often made to equate some aspect of strategy use to the ‘good’
or to the ‘poor’ group.
Because of the time-consuming methodology, the num ber of
subjects tends to be small, and there is an emphasis on qualitative
rather than quantitative results. However, it should be noted that
the uncovering of strategies is pragmatic; hence, unlike skills, their
psychological validity does not need to be investigated.
Two fairly representative examples of strategy research are the
investigations of Olshavsky (1977), who did her work with English-
speaking readers, and Sarig (1987), who worked with bilingual
Hebrew- and English-speaking subjects. Since the focus here is on
the actual strategies used by readers, together with definitions of
what constitutes a strategy, no details are given as to the hypo­
theses examined, or the discussion of strategy use.
The strategies detected and categorised by Olshavsky were as
follows:
■Word related: Use of context to define a word, synonym substitu­
tion, stated failure to understand a word.
■Clause related: Re-reading, inferences, addition of information,
personal identification, hypothesis, stated failure to understand
a clause.
■Story related: Use of information in story to solve a problem.
 The theory of reading 95
Sarig used a similar technique to investigate the behaviour of
Hebrew-speaking students reading in English and in Hebrew. Again
a ‘think-aloud-when-reading’ technique was use, though it is not
clear from Sarig’s account precisely when they verbalised. Sarig
refers to responses to any particular problem as ‘moves’, and strat­
egies as being combinations of moves. To avoid confusion, how­
ever, her ‘moves’ will be referred to as strategies. The strategies
she uncovered and categorised were as follows:
■Technical aid: Skimming, scanning, skipping.
■Coherence detecting: Identification of macroframe, use of content
schemata, identification of key information in text, etc.
■Clarification and simplification monitoring: Syntactic simplification;
using synonyms, circumlocutions, etc. Change of planning, mis­
take correction, ongoing self-evaluation, controlled skipping,
repeated reading.
Even given the ten years between the two papers, it is striking
just how different are the two lists of strategies. While, as claimed
above, it can be argued that the detection of strategies, unlike
that of skills, is pragmatic, there is clearly an element of subjectiv­
ity both in identifying and in categorising them. This subjectivity
may be reduced if we can agree on a definition of strategies.

Definitions
Both Olshavsky and Sarig view reading as ‘a problem-solving pro­
cess’. Admittedly there may be some problems defining ‘problem ’,
but, in commonsense terms, we can regard strategies as ways of
getting round difficulties encountered while reading. Thus, initially
at least, strategies can be seen as responses to local problems in a
text. We should also include in our definition a reference to the fact
that the response must be a conscious one. Olshavsky claims that
a strategy is ‘a purposeful means of comprehending the author’s
message’ (p. 656; our emphasis). Pritchard (1990: 275) defines a
strategy as ‘a deliberate action that readers take voluntarily to*
develop an understanding of what they read’. Cohen (1998: 5)
points out that the question is controversial but comes down firmly
on the side of conscious choice:
In my view, the element of consciousness is what distinguishes strat­
egies from those processes that are not strategic.
96 Reading in a Second Language
Given a definition along the lines we have indicated above, it is
hard to accept some of Olshavsky’s strategies as such; in par­
ticular, ‘stated failure to understand a word’ or ‘stated failure to
understand a clause’ as strategies. Olshavsky emends this by dividing
her strategies into ‘problem identification’ (i.e. monitoring), and
‘problem solving’. Another of her ‘strategies’ involves the reader
substituting synonyms during recall. But this is a general feature
of readers’ recall of text (cf. Steffensen and Joag-Dev, 1984).

Strategies and skills

There is a fair amount of confusion in the literature as to what
distinguishes a skill from a strategy. Some writers (e.g. Nuttall,
Grabe) refer to ‘skills/strategies’ as if the two were interchangeable.
Table 2.1 illustrates the amount of overlap. Admittedly, some of
this apparent confusion may be due simply to the fact that the
skills proponents did not attempt to separate skills from strategies
(we have already noted the practice of referring to ‘skills/strat­
egies’ as if the two were interchangeable). However, it would
be satisfying, for the sake of clarity, to arrive at some generally
accepted distinction. The following are possible differences:
■Strategies are reader-oriented, skills are text-oriented. It is cer­
tainly true that skills taxonomies tend to focus on text. Munby’s
taxonomy of skills is overwhelmingly text based. In the list of 19
reading skills cited by Criper and Davies (1988), only 7 can be
said unambiguously to take the reader into account, e.g.
Interpreting text by going outside it.
Selective extraction of relevant points from a text.
The other 12 are text based, e.g.
Understanding conceptual meaning.
Understanding the communicative value of sentences.
Understanding the relations within the sentence.
Understanding relations between parts of text through lexical
cohesion devices.
In fact, Munby’s skills represent a passive reader, typical verbs
used being ‘understand’, ‘recognise’. Anything which the reader
actively contributes to the text, such as Olshavsky’s ‘Personal iden­
tification’, or any behaviour which the reader, as an individual,
 The theory of reading 97
Table 2.1 Distinguishing features between ‘skills’
and ‘strategies’
STRATEGIES SKILLS
Olshavsky Sarig Munby, Nuttall,
Lunzer et al.
Use of context to Deducing the meaning
define a word or and use of familiar lexis
synonym
Synonym substitution Synonyms
Stated failure to Ongoing self-
understand a word evaluation; mistake
correction
Re-reading Repeated reading Repeating reading
Skimming or scanning
Inference Paraphrasing Understanding
information in the text,
not explicitly stated
Addition of Extra-textual content Interpreting text by
information going outside it
Stated failure to Deserting a hopeless
understand a clause utterance
Flexibility of Variations in
reading rate reading rate

manifests towards the text, e.g. skipping, are likely not to appear
on Munby’s list. Obviously, since monitoring is a reader-directed
activity, many of its manifestations will not appear, e.g. self-
evaluation in general, admitting failure to understand part of
the text, are again unlikely to appear. The fact that some such
activities do appear - e.g. scanning and skimming - may be an
indication that Munby’s ‘skills’ do, in fact, include a num ber of
‘strategies’.
■Strategies represent conscious decisions taken by the reader,
skills are deployed unconsciously. Another way of phrasing this
is that skills have reached the level of automaticity. Certainly
many of Munby’s skills, such as lexical recognition and syntactic
98 Reading in a Second Language
parsing, can be assumed to have reached automatic levels in LI
or advanced L2 readers, and hence would not be reported in
strategy research. There are difficulties associated with this cri­
terion; all the descriptions detailed above include ‘re-reading’.
The regressions reported from eye-movement research could be
considered a type of re-reading, and it might be difficult to
decide just how conscious readers were of regressing. However,
the criterion of ‘conscious v automatic’ seems a good one to us.
■Strategies, unlike skills, represent a response to a problem, e.g.
failure to understand a word or the significance of a proposi­
tion, failure to find the information one was looking for, etc.
This criterion is closely related to our first one: at a local level,
something only becomes a problem if one becomes aware of it.
It is true that the term ‘problem ’ poses some difficulties in
itself. We have been using the term to refer to local difficulties
encountered when reading a text. However, in Newell and
Simon’s theory, used by Olshavsky, a problem may be anything
in the task environment which stands between the organism
and its goal. In this wider sense, if we decide to read a book for
information on a particular subject, the whole text of the book
becomes a problem. We discuss what we call ‘global’ strategies
in the next section.
On the whole, however, we agree with the distinction drawn by
Williams and Moran (1989: 223):
A skill is an ability which has been automatised and operates largely
subconsciously, whereas a strategy is a conscious procedure carried
out in order to solve a problem.

Strategies and styles

A rather different approach to strategy research can be found in
the work of Pask (1976) and Entwhistle etal. (1979) in Britain,
and Marton (1976), Saljo (1975) and Fransson (1984) in Sweden.
Widdowson (1984) should be included in this group, as should
Hartman (1992) in the USA. These researchers are usually con­
cerned with effective learning, rather than reading as such, but
since their subjects are usually students in tertiary education, much
of whose learning is achieved through reading, their work is rel­
evant here.
 The theory of reading 99
Table 2.2 Different reading styles
Pask Holistic: Characterised by Serialist: Learners attempt
the attempt to relate text to build understanding
content to what is already out of the component
known or to current details, logical steps and
concerns. The reader uses operations taken strictly
personal analogies, in a linear sequence.
illustrations, and concrete Relationships are sought
examples which, while entirely within the context
formally incorrect, serve as of the task.
temporary pegs on which to
hang partially understood
abstract concepts.
Mar ton, Deep processors: Aim at a Surface processors:
Saljo, thorough understanding of Attempt rote learning
Fransson the author’s main message of important pieces
or argument (Marton); of information on
relate ideas of the text to which questions might
their own personal subsequently be asked
experience (Fransson). (Marton); direct their
learning toward learning
the text itself (Fransson).
Hartman Logocentric: The reader is Resistant: The reader tries
buried in a passage, his or to absent the author’s
her ‘textual world’ being meaning by asserting his
largely defined by the or her own.
author’s meaning.
Widdowson Dominant: The reader Submissive: The reader
asserts the primacy of his adjusts in a submissive
or her conceptual pattern, manner to the writer’s
fitting textual information scheme, following the
into it directly and short- discourse development the
circuiting the discourse writer has plotted.
process (p. 223).

The different reading strategies suggested by these authors,

and the oppositions between them may be roughly summarised in
Table 2.2.
There are clear similarities between some of the global strat­
egies given in Table 2.2. Pask’s ‘holistic’ reader resembles Fransson’s
‘deep processor’, while both have resemblances to Hartm an’s
100 Reading in a Second Language
‘logocentric’ reader. Hartm an’s ‘resistant’ reader seems to be an
extreme example of Widdowson’s ‘dom inant’ reader. However,
the parallels must not be pushed too hard. Both the ‘holistic’ and
‘serialist’ readers are being submissive with respect to the author’s
message, just employing different local strategies to achieve their
aims. The ‘surface processors’, as described, are different in re­
sponding, not so much to what they perceive as the author’s mes­
sage as to the task.
As to which is the ‘best’ global strategy, the Swedes appear to
be most prescriptive, in favouring deep processing at all times.
Fransson, in fact, cites Dewey to the effect that everyone is a deep
processor when not inhibited by nervousness, task demands, etc.
Widdowson, on the other hand, along with Pask, sees the differ­
ent strategies as being appropriate to different reading contexts.
Pask, in fact, refers to an exclusive reliance on one strategy as a
‘pathology’.
Global strategies, as we have been calling them, are important
for a number of reasons, in spite of differences in the descrip­
tions. They can be related to different contexts and outcomes; in
practical terms, it is difficult to see how anyone doing a test can be
anything but a submissive reader, but teachers should keep in
mind that students should be given the opportunity to be domin­
ant in appropriate contexts. Finally, the definition provided by
Entwhistle et al., that a strategy is ‘a description of the way a
student chooses to tackle a specific learning task in the light of its
perceived demands’, makes it possible for us to incorporate differ­
ent reading types such as skimming, search reading and scanning,
as strategies.

2.4 DIFFERENT KINDS OF READING

All the models of reading that have been looked at so far have
been designed with careful reading in mind. Hoover and Tunmer
(1993), for example, consider that their notion of the simple view
‘assumes careful comprehension: comprehension that is intended
to extract complete meanings from presented material as opposed
to comprehension aimed at only extracting main ideas, skimming,
or searching for particular details’ (p. 8). In fact many of the
models of reading that have surfaced in the literature to date have
been mainly concerned with careful reading; Rayner and Pollatsek
 The theory of reading 101
(1989: 439) state that for most of their account of the reading
process they are focusing on the skilled, adult reader reading
material of the textbook variety. They point out that careful read­
ing models have little to tell us about how skilled readers can cope
with other types of reading such as skimming for gist (Rayner and
Pollatsek, 1989: 477-8).
However, while such a restriction of focus is perfectly legitim­
ate for psychologists attempting to establish precise experimental
data about the reading process, it is a luxury which we cannot
allow ourselves, since the reading needs of students, and hence
the teaching and testing of reading, requires a wider range of
reading behaviours. Thus, if possible, we must expand the model
in order to accommodate this wider range.
The overriding attention paid to careful reading in the theor­
etical literature has meant that, in Britain at least (see Section 3.1
below), we have somewhat ignored expeditious reading behavi­
ours such as skimming, search reading and scanning in both LI
and L2 teaching of reading. We have theories of careful reading
but very little on how readers process texts quickly and selectively,
i.e. expeditiously, to extract important information in line with in­
tended purpose (s). Given the value of these types of reading to
the work forces of states in the northern hemisphere, let alone
those of emerging nations, it is time more attention was paid to
them in the professional and ‘academic’ literature.
In addition, because of the focus on the local level, e.g. word
recognition or syntactic parsing, the psychological literature has
paid only limited attention to careful reading at the global level,
i.e. comprehension of the main ideas in a text or of the discourse
topic; the macropropositional as against the micropropositional
level of text.
We feel, therefore, that in addition to careful reading at the
local level, it is important to discuss a further five kinds of read­
ing: Search reading, Skimming, Scanning, Careful reading (at the
global level) and Browsing. We choose these as our main illustra­
tions, with no assumption that the list is exhaustive, or even that it
includes all the most important kinds of reading (irrespective of
how one would set about judging that). We do know, however,
that the first four are the kinds of reading that appeared most
frequently in our analysis of textbooks for the teaching of reading
and in published and available tests of reading, and that they will
be familiar to teachers of L2 reading at least.
102 Reading in a Second Language
One problem which confronts us immediately is that, while at
least the first five terms for different types of reading are often
used in the literature, they often appear to be used in different
ways. For example, in the remark by Hoover and Tunm er cited
above, they appear to distinguish between ‘extracting main ideas’
and ‘skimming’, although many people would consider these two
to represent the same behaviour. The IRA dictionary of reading
terms (Harris and Hodges, 1981) defines scanning as ‘to read some­
thing quickly but selectively’, but then equates it with skimming.
As examples of scanning, they give ‘to scan an article for the
general idea, scan a directory for a telephone num ber’. Many
people would consider that reading ‘for the general idea’ consti­
tuted skimming. Under skimming, they say that the term ‘shares
only the first of two primary meanings of “scan”, to read rapidly
and selectively, but purposely, rather than to read carefully’. It is
difficult to justify the apparent contrast between reading ‘select­
ively’ and ‘purposely’.
In Appendix 1 (pp. 297-304) we try to make clear the distinction
between the different types of reading by providing an example of
a specification produced by the Testing and Evaluation Unit at
CALS, University of Reading, in collaboration with colleagues in
Shanghai Jiatong University, PRC, for use in the development of
the Advanced English Reading Test (AERT). These types of read­
ing are used for different purposes. The specification sets out a
number of reading behaviours categorized along the two axes of
(a) Local v Global, and (b) Careful v Expeditious. To help clarify
the distinctions between the different types, we take each type and
examine it in terms of Purpose, Operationalisations, Comprehension
Focus, Text Coverage, Rate of Reading, Direction of Processing and Rela­
tionship with Underlying Process.
In Chapter 3, on the testing of reading, and in Chapter 4, on
the teaching of reading, we will use this taxonomy as a basis for
discussion as we feel it provides a simple but coherent route
through the vast literature in this area. It should help teachers
and researchers make sense of a difficult and often overcomplex
field. In Chapter 3 we will examine these reading types (operations)
more fully and also discuss the performance conditions that affect
them, such as length of text, topic, etc. For the moment we put
forward the following rough working definitions:
Skimming: Reading for gist. The reader asks: What is this text as a
whole about?, while avoiding anything which looks like detail.
 The theory of reading 103
Reading schemes like SQ3R recommend starting the reading to
learn process with skimming, so that the reader has a frame­
work to accommodate the whole text. While Hoover and Tunmer
appear to distinguish skimming and ‘reading for main ideas’, these
appear to us to be the same thing. The defining characteristics
are (a) the reading is selective, with sections of the text either
omitted or given very little attention; (b) an attempt is made to
build up a macrostructure (the gist) on the basis of as few details
from the text as possible.
Search reading: Locating information on predeterm ined topics.
The reader wants information to answer set questions or to pro­
vide data, for example, in completing assignments. It differs from
skimming in that the search for information is guided by predeter­
mined topics so the reader does not necessarily have to establish a
macropropositional structure for the whole of the text.
Scanning: Reading selectively, to achieve very specific reading
goals, e.g. finding the number in a directory, finding the capital of
Bavaria. The main feature of scanning is that any part of the text
which does not contain the preselected symbol(s) is dismissed. It
may involve looking for specific words/phrases, figures/percent-
ages, names, dates of particular events or specific items in an index.
Careful reading: This is the kind of reading favoured by many
educationalists and psychologists to the exclusion of all other types.
It is associated with reading to learn, hence with the reading of
textbooks. The defining features are (a) that the reader attempts
to handle the majority of information in the text, that is, the
process is not selective; (b) that the reader adopts a submissive
role (see Section 2.3 above), and accepts the writer’s organiza­
tion, including what the writer appears to consider the important
parts; and (c) that the reader attempts to build up a macrostructure
on the basis of the majority of the information in the text.
Browsing: We have to say that we do not know what ‘norm al’
reading consists of; there are far too many reading behaviours to
allow us to be confident about selecting any of them as predom­
inant. ‘Careful reading’ may be the preferred mode among students
faced with textbooks, and hence may be of particular concern to
us here, but it is not something that the majority of people are
likely to engage in for a large part of the time. So we have added
a type ‘browsing’ to describe the sort of reading where goals are
not well defined, parts of a text may be skipped fairly randomly,
104 Reading in a Second Language
and there is little attempt to integrate the information into a
macrostructure, except, perhaps, for a topic structure like ‘This
text seems to be ab o u t. . As noted above, Kintsch and van Dijk
(1978) refer to the cases where ‘people read loosely structured
texts with no clear goals in mind. The outcome of such com­
prehension processes, as far as the resulting macrostructure is
concerned, is indeterm inate.’ This is what we are referring to as
‘browsing’. We should add, however, that it is not a requirem ent
for browsing that the text be ‘loosely structured’; we can ‘browse’
through virtually any text, given only that it consists of more than
a few words.

Note
We can expand this to point out that there is no necessary cor­
relation between a particular reading behaviour and a particular
genre of text. We might assume that people are more likely to
apply their ‘careful reading’ processes to a study text, but text­
books can be read for amusement, not learning, while texts taken
from tabloid newspapers may be scrutinised with great care. We
should also point out that readers are under no obligation to
maintain a particular reading behaviour throughout the length of
a text; they may switch from careful reading to skimming to search
reading to scanning and back to careful reading over a small
number of pages.
We can now compare and contrast the five provisional types,
in order to examine the factors which are involved in differenti­
ating them.
1. Skimming, search reading, scanning and careful reading are
distinguished from browsing by the presence in the first four
of a clearly defined goal. The goals, of course, differ widely,
but in all four cases, the reader can be assumed to know before
reading what it is they want from the text.
2. Search reading, scanning, skimming, and possibly browsing are
distinguished from careful reading by the factor of selectivity.
In the last type, all the text can be presumed to be examined;
in at least the first two types, and probably the third, the reader
will deliberately either avoid, or pay minimum attention to
some parts of the text. In scanning, the parts ignored may
constitute a majority of the text.
 The theory of reading 105
3. In careful reading, and in skimming, the reader makes a con­
scious effort to construct a macrostructure, the gist of the text.
In careful reading, this is likely to be done by reference to the
whole text, in skimming from parts of the text. In scanning,
there is no attempt to construct a macrostructure; in browsing,
some vague notion of the topic may be built up, but without
any attempt to retain it; in search reading it is probable that
only certain key ideas in the macrostructure will be sought.

Expanding a model

We shall now try to illustrate this by referring our descriptions to

a generalised representation of aspects of the reading process.
Figure 2.2 is derived from Just and Carpenter (1980, 1987) with
additions from Kintsch and van Dijk (1978). However, we should
stress that it is for discussion purposes only and not intended to
be taken seriously as a model of the reading process.
Since the discussion concerns a num ber of different reading
behaviours, we have included in the figure a Goalsetter, which, by
deciding on the overall goal of the reading, also selects the type of
reading which is likely to achieve that goal, and a Monitor; which
provides the reader with feedback about the success of the par­
ticular reading process. The monitor is controlled by the goalsetter,
since, as will be seen, the monitoring procedure depends crucially
on the type of reading.
Comparisons we have recently made with the ‘blueprint for the
speaker’ developed by Levelt (1989) are encouraging. He portrays
talking as an intentional activity and his ‘conceptualizer’ involves
both message generation and monitoring:
Talking as an intentional activity involves conceiving of an inten­
tion, selecting the relevant information to be expressed for the
realization of this purpose,. . . These activities require the speaker’s
constant attention. The speaker will, moreover, attend to his own
productions, monitoring what he is saying and how ... (p. 9)
In our ‘blueprint’ for reading, skills presumably should be seen
as automatised aspects of the long-term memory. Local strategies
emerge in response to the monitor, which alerts the reader that a
problem has emerged, and sets the reader to choosing and em­
ploying a suitable strategy. Global strategies can be seen as being
106 Reading in a Second Language
GOALSETTER
Get next input: Choose type
move eyes o f reading
WORKING MEMORY
Microstructure:
propositions a, c, e, f, g
Extract physical
features

Encode word MONITOR BACKGROUND

and access lexicon ^ KNOWLEDGE

Parse syntactic
structures
LONG-TERM MEMORY^
Macrostructure:
Integrate with propositions A, C y
representation
of previous text

No End of
sentence?

Sentence
wrap-up
Figure 2.2 A development of Just and Carpenter’s model of
the reading process.

activated by the goalsetter. The reader may choose to scan, skim,

search read, or read carefully, all in response to the perceived
demands of the learning task.
We shall begin with Careful Global Reading, since it shows
all components of the model in use. The reader decides to read
the text carefully, presumably for study or some similar purpose.
In terms of the figure, the goalsetter is set to careful reading. This
in turn gives the monitor its instructions. The reader begins
each sentence at the beginning and continues through to the
end, following the processes detailed, i.e. extracting features,
encoding words, parsing, etc. If we are to follow Kintsch and van
Dijk’s account, microstructures are processed, converted into
semantic propositions, and stored in the working memory, while
 The theory of reading 107
the coherence between them is established. As the process moves
on, a macrostructure has begun to be built up. Background know­
ledge, actually stored in long-term memory, is utilised to supply
an appropriate schema for the macrostructure, as well as to aid
coherence detection in the construction of the microstructure.
While all this is going on, the monitor is at a high level of
attention, checking, among other matters, whether meanings
accessed from the lexicon are appropriate, whether parsing results
in an acceptable sentence structure, whether the micropropositions
appear to cohere, whether an appropriate provisional schema has
been located, and so on.
Compare this now with Scanning. Here far fewer components
of the model are involved. Suppose at the lowest level, the goal
has been set as scanning a text to find a name. We don’t know
what processes go on when the goalsetter is set to ‘scan’. Pre­
sumably, however, little or no syntactic processing needs to be
involved, no checking of coherence, and no attempt to build up a
macrostructure. In fact, it is arguable that only a limited amount
of lexical access is required; the reader might scan by decoding
alone, without accessing the meaning and phonological repres­
entation of the words. Whatever is the case, the monitor can be
envisaged as set at a simple Yes/No level, checking only to see if
the word or words being scanned fitted the search description or
not. If it did, then the search would be over; if not, the word
would be rejected, and the search continued. There is no need to
complete the reading of the sentence, or to integrate the word
into the structure of preceding text. Checking the coherence of
micropropositions would seem to be redundant, and there seems
little use for any sort of macrostructure to be built up.
Skimming is more complex. Like Scanning, it is selective, but
we are unsure how much data the reader decides to process at any
point. They may access words, and possibly process entire sen­
tences. However, since the process is selective, some material must
be skipped. Presumably, here, as with scanning, the monitor checks
as to whether the material surveyed is appropriate or not; how­
ever, in this case, the amount processed may be quite substantial.
If skimming is equivalent to gist extraction, then presumably propo­
sitions are committed to the long-term memory on the hypothesis
that they represent the macrostructure. That is, the process of
debating whether a proposition is part of the macrostructure or
not, which is assumed to take place during careful reading, is here
108 Reading in a Second Language
replaced by a guess that it is. The macrostructure will then have
more components contributed by the reader, making skimming a
more reader-driven activity.
Search reading seems at different points to share affinities with
both scanning and skimming. The search will primarily involve
keeping alert for words in the same or related semantic field and
operationalisations such as using titles, subtitles and other dis­
course clues will be employed (see van Dijk 1977: 79). The reader
may contribute formal knowledge of text structure to assist in this
search for information on prespecified macropropositions. As with
skimming, where selected text is identified as being important, it
will be read more carefully and the process identified above for
careful reading will kick in.
Finally, we come to Browsing. This is the least well defined of
our reading types. The goals are vague, the monitor operates at
low attention, and the macrostructure, if built up in any sense, is
probably at a level of ‘topic’. We will pay little further attention to
this reading type as, given the difficulty of defining its purpose, it
is difficult to operationalise for teaching and testing purposes.
Most of what has been said above is speculative, and quite
possibly wildly inaccurate. The important point is that in our view
future models will have to be able to explain at least five types of
reading: reading carefully at the local level; reading carefully at the
global level; skimming; search reading; and scanning.
In Chapters 3 and 4 we shall explore the extent to which these
types of reading can be tested and taught. In this way we hope to
make a contribution to the development of a framework of read­
ing that goes beyond decoding and the microlinguistic level to
take account of reading for different purposes.

Notes
1. Gough’s ‘decoder’ converts letters into phonemes. This is in line with
his apparent assumption that the end-product of reading is a spoken
sentence. Since we don’t share that assumption, the decoder has been
left out.
2. We do not know of experiments testing the effect of lack of punctua­
tion on reading. However, in Britain at least, the fact that teachers can
read students’ essays seems to be evidence that it is possible.
3. Rayner and Pollatsek admit that Goodman’s account can be viewed as
interactive, but consider it as basically top-down because ‘bottom-up
 The theory of reading 109
processing plays such a minor role’ and there are ‘so little constraints
on the interactions’. In a sense, though, this is a repetition of their
criticism that his account lacks precision, a criticism earlier made by
Gibson and Levin (1974).
4. De Beaugrande’s reference to background knowledge here should
remind us of the crucial role of such knowledge in, for example, iden­
tifying pronominal reference. This role suggests that cohesion tests
may be a useful way of assessing the effect of background knowledge
on reading.
5. We are using the terms found in Halliday and Hasan (1976). Other
writers on cohesion, e.g. de Beaugrande, and Quirk et al. (1973) use
similar though not identical terminology.
6. It should be remembered that Kintsch and van Dijk’s model sets out
to describe the actual reading process, in which propositions are pro­
cessed in cycles, generally about 4 at a time, with selected propositions
being carried over to the following cycle.
This page intentionally left blank
 111
3

Testing reading
comprehension(s)

3.1 INTRODUCTION: THE LIMITS OF TESTING

We begin this chapter on the testing of reading with the question
as to whether reading can be tested. This may seem a strange pro­
cedure; however, it is important to address this issue because we
need to be aware of the parameters we can work within. We need
to be sure that we can actually test what we want to test (validity)
and that we can depend on the results our tests provide us with
(reliability).

Measuring types of reading

Conventional testing involves measuring subjects’ performance
against an agreed standard. For any reading test, this means that
there must be a consensus as to what constitutes good reading in
the context established by the test. Given this, if the purpose of
reading is for amusement, entertainment, aesthetic satisfaction, to
gain a general background in an area, or merely browsing, then
it is difficult to see how reading performance can meaningfully
be tested. For example, enjoyment derived from reading a novel
may be investigated, even measured, but since we do not have
explicit standards of enjoyment, we cannot test it. We can certainly
test retention of a plot, or of details, or of characterisation, etc.,
but this is quite different.
However, many language tests do not set out to measure stu­
dents’ aesthetic reactions to texts, pleasure and so on, so it might
be argued that the above remarks are irrelevant. EAP reading
tests, it might be claimed, measure learning in the sense of adding
112 Reading in a Second Language
to our knowledge base, and this can be tested. However, even with
EAP reading we need to be careful about the statements we make
on the basis of our tests. While the claim for addition to our know­
ledge is to some extent true, a closer examination of this claim also
leaves it in some doubt. Neville and Pugh (1982) comment that:
‘The output of reading is . . . difficult to capture, since what is
achieved from (real life) reading with comprehension is often a
modification of the conceptual system.’ Furthermore, the reading
modules of EAP tests cannot normally be said to measure learning
per se; rather they measure behaviour that we hope is associated
with successful learning.
As we argue in Chapter 2 and Section 5.2, it is imperative that
we specify as clearly as possible what it is we are attempting to test,
teach or research. The more explicit we can be, the more we are
likely to understand the nature of reading or, more precisely, the
limits of such understanding.

Comprehensions and interpretations

A further limitation arises when we consider the interaction be­
tween readers and texts that makes up the reading process. As
we noted in Chapter 2, there has been for some time a consensus
in the field of reading theory about the general outlines of the
relationship between reader, text and ‘comprehension’, though,
of course, details vary. Thus there is agreement among reading
theorists that every text is incomplete, and has to be converted
into meaningful discourse by the reader. Candlin (1984: x)
remarks:
Texts do not have unitary meanings potentially accessible to all,
they rather allow for variety in interpretation by different readers,
governed by factors such as purpose, background knowledge, and
the relationship established in the act of reading between the reader
and the writer.
This in turn entails the likelihood, in fact necessity, of different
readings of the same text.
As we noted in Section 2.1, Urquhart (1987) sees the variation
in product as operating in two dimensions. Readings may differ in
the case of readers from different cultures, either ethnic or pro­
fessional, or in the case of the same reader at different times, with
 Testing reading comprehension (s) 113
different knowledge or different preoccupations. Such differing
readings, which are generally not under control of the readers,
Urquhart terms ‘interpretations' .
The reading product may also vary according to a dimension
controlled by readers’ purposes, As Candlin points out (op. cit.),
the reader ‘may decide to glimpse at the text, extracting “gist”, or
work conscientiously through it, satisfying himself that he has
made sense of all of it’. In Urquhart (1987) such variations are
labelled ‘comprehensions1.l Books teaching reading, and many read­
ing tests, tend to assign special importance to ‘deep’ or ‘intensive’
careful reading, but there is no particular justification for this
and, as we argued in Chapter 2, expeditious reading skills may be
just as important.
On the whole, variations in the comprehension dimension
should not cause testers much difficulty, once the validity of dif­
ferent purposes for and different types of reading is accepted. In
testing terms, different purposes can be translated into responses
to different tasks. Pugh (1978: 78) argues:
comprehension is best regarded as a state of achievement, rather
than as an activity or a skill which is applied to texts. By stressing
the different strategies and styles of reading which are appropriate
for achieving various goals, one breaks away from the apparent
impasse into which many of the discussions of comprehension lead.
‘Interpretations’, on the other hand - variations brought about
in the reading product on account of different schemata - are a
major problem for conventional testing. In fact, this is perhaps
the greatest area of potential dispute between theories of reading
and reading tests. It is of crucial importance, yet it is seldom
discussed. The theory insists that the good reader makes sense of
the text by supplying knowledge based on his or her own unique
experience. The testers, on the other hand, are obliged to look
for ‘correct’ answers. Candlin (1984: xi-xii) remarks that conclu­
sions regarding different readings
put at question . . . how we make use of tests to assess reading skill;
the viability of much product-oriented testing must be in doubt,
especially if the results of such tests are deemed to be revelatory of
reader strategy.
We have claimed that there is a consensus among people
concerned with the theory of reading about the all-pervasive
114 Reading in a Second Language
importance of background knowledge and of different interpreta­
tions of the same text. However, outside the circle of reading
theorists, it is by no means the case that this importance is either
fully accepted (although it may be paid lip-service) or accepted
at all. This lack of appreciation may take two forms. In the first,
background knowledge is seen as a resource available to readers
when, as it were, the going gets tough. Some taxonomies of read­
ing strategies (see discussion in Chapter 2, and, for example,
Pritchard, 1990) mention recourse to background knowledge as
an appropriate means of overcoming a reading difficulty.
As is often the case, however, strategy research, by concentrat­
ing on readers consciously problem-solving, tends to deal with the
tip of the iceberg. It must be stressed that if the theory is correct,
then background knowledge is not just an extra resource, it is, as
it were, a filter through which we view all texts.
Secondly, the importance, or even possibility of differing inter­
pretations is simply not considered. For example, when studying
adult native speakers’ confidence in their answers to ‘thematic’
questions on ‘challenging’ texts, Pressley et al. (1990) conclude
that their subjects suffered from ‘gross comprehension prob­
lems’ (p. 247) and showed ‘high confidence in the majority of the
incorrect answers’ (p. 245). The authors put this down to poor
monitoring, and do not seem to have considered the possibility
that the readers’ interpretations may have had at least as much
validity as their own (cf. McCormick, 1992).
Beyond the editing stage, questions aimed at inferential process­
ing which often seem to produce a scattering of responses between
different multiple-choice options are often dropped as ‘poor’ items.
Yet it is possible that the responses reflect differing, but possibly
legitimate, responses on the part of different testees relating the
item to different backgrounds.
Of greater concern in testing reading proficiency is emerging
empirical evidence (Perkins and Brutten, 1988) that inference
questions are poor discriminators and perhaps should be avoided
for that reason alone in testing. If good students (as defined by
overall performance) are getting items wrong and poor students
such items right, this does not appear to be telling us much of use.
It is likely that conventional tests simply cannot make allow­
ance for individual interpretations. This is not sufficient reason
to abandon reading tests, but it is a necessary caution against
extrapolating too much from them. We cannot emphasise too
 Testing reading comprehension (s) 115
much the need to be explicit about exactly what it is reading tests
can tell us. We may never be able to detail precisely the complex­
ity of individual differences in reading ability (Spiro and Myers,
1984). However, we will argue that by selecting appropriate texts,
tasks and formats we can still generate useful performance test
data on a person’s ability to read for certain purposes under cer­
tain performance conditions.

Focusing on comprehensions

Despite the limitations discussed in the previous section we are

still left with the need to make statements about reading profi­
ciency for many areas of human activity, e.g. for academic study,
business or technological purposes. As Farr et al. (1986: 135) point
out: ‘tests . . . are, above all, a political reality. They are a constant
with which all students, teachers, and administrators know they
will have to contend.’
As we have indicated above, we do not believe that it is possible
to incorporate ‘interpretations' into testing as it is practised at the
moment. The answer must be to concentrate on ‘comprehensions1,
that is, variations resulting from different and consciously adopted
purposes. This does not mean, of course, that we can eliminate
different interpretations from testees’ performance; it means that
we cannot formally take account of them in awarding grades.
One of the main causes of differing interpretations is back­
ground knowledge, and the elimination of this variable would seem
an obvious step.2 For most practical purposes, however, this is likely
to be an impossibility; the theory would suggest that background
knowledge is always present. All we can do is attempt to minimise
the effect of the variable. As Farr et al. (1986: 140) point out:
Factoring background knowledge out of reading assessment is next
to impossible; yet background knowledge must be controlled so
that it will not account for an indeterminate amount of assessment
results ... to factor reader background out of assessment... is to
assess something other than reading comprehension.
Two methods for minimising the effect of background know­
ledge suggest themselves, one involving the choice of text, the other
the choice of task. As far as texts are concerned, three solutions
are suggested in the literature. First, we might employ a variety of
116 Reading in a Second Language
short texts, covering a wide range of topics. This appears to be the
method chosen by the TOEFL examination. Disadvantages include
the large chance factor, if the hope is that all testees will find some
texts familiar and others not. Also, the method appears to involve
the use of short texts, which may be considered undesirable (see
below). Secondly, texts might be selected which are unfamiliar to
all the candidates so that, as Roller (1990) argues, text variables
rather than background knowledge have the most influence. This
was the policy adopted by the Joint Matriculation Board in the UK
in its Test in English for Overseas Students where deliberately
obscure texts were chosen so as to be equally unfamiliar to all
candidates.3 Thirdly, an appropriate level of familiarity with the
topic of the text across the test population might be established
in advance through group interview or controlled questionnaire
survey (see discussion of background knowledge in the section on
performance conditions below).
The other main method of minimising the effect of background
knowledge is through the tasks demanded. Ideally we would want
to test only what is retrievable from within the text itself irrespect­
ive of differing states of background knowledge relating to the
focus of information retrieval. This attempt has a fairly long history.
For example, Carroll attempted to distinguish what was ‘committed’
to the text, the retrieval of which constituted ‘adequate compre­
hension’ (see discussion in Section 5.3 of ways of establishing
this empirically), from what could be built up from the test as a
whole together with background knowledge, retrieval of which
constituted ‘total comprehension’ (Carroll, 1972). Others have
attempted to distinguish between what Chikalanga (1992) refers
to as ‘propositional inferences’, which are text-based, and ‘prag­
matic inferences’, which are based on information outside the
text (see Section 4.2 for a discussion of these in teaching). Similar
distinctions can be found in Crothers (1978), Farr et al. (1986),
Hughes (1993), Weaver and Kintsch (1991) and Pearson and
Johnson (1978). An example should make the distinction clear.
Chikalanga used a text in which a violent father asks his wife to
confirm one of his stories. She gives an evasive answer, ‘flashing
her bony hand under her chin’ (1993: 946). Students were asked
to supply a pragmatic inference as an answer to the question:
‘Why, do you think, did Jojo’s mother flash her hand under her
chin?’ Clearly such questions must be answered on the basis of
information outside the text.
 Testing reading comprehension (s) 117
We must admit, however, that the distinction, while valuable,
can only be pushed so far. Given a text like:
He woke up feeling depressed. He still had not finished editing the
testing chapter.
and asked the question, ‘Why did he feel depressed?’, we can
answer by relating two parts of the text. But the relationship is a
pragmatic one, and the location of an appropriate answer depends,
unless the testee is trusting to luck, on background knowledge.
In general, however, the principle is a useful one; we exclude
from testing, though not from teaching (see Section 4.2), items
which depend solely on information supplied from outside the
text, and include items which, though still depending on the reader
supplying background knowledge, can be made to relate to each
other in the text. The extent to which such items can be reliable
testing items is a matter for empirical research.

What is there to comprehend?

With this in mind we need to examine in more detail what com­
prehension of a text might involve. When constructing test tasks,
testers need to operate with a consensus as to what information
readers may be expected to extract from a text, given the con­
straints of the kind of reading required, e.g. careful or expedi­
tious reading. As we saw in Chapter 2, most attention has been
paid by theorists to decoding carefully at the word level. We noted
that in contrast Kintsch and van Dijk (1978) offered a model which
concentrates on comprehension above word level. As Samuels and
Kamil (1988: 188) say, Kintsch and van Dijk:
have built a model emphasising comprehension to the exclusion of
word identification (most other models including Rumelhart’s, seem
to have a bias for explaining word identification).4
A focus on comprehension is important in the real world where
we have to develop tests and teaching exercises to account for
global comprehension as well as local. If we focus on a text as a
whole, for the purpose of, say, careful reading, or possibly skim­
ming, we can use Kintsch and van Dijk’s model to describe the
elements and relationships which can be taken as constituting
comprehension, at least as far as careful reading is concerned.
Such research into global theories of text comprehension (Kintsch
and van Dijk, 1978; Just and Carpenter, 1980; Meyer, 1975; Meyer
118 Reading in a Second Language
and Rice, 1984* Perfetti, 1985; van Dijk and Kintsch, 1983) offer
us some hope that we can develop test tasks for expository text
comprehension, i.e. measure the extent to which information com­
municated by the writer is understood at both micropropositional
and macropropositional levels.
We noted in Chapter 2 that micropropositions refer to pro­
positions which are the smallest units of text that can logically
be proven false (Kintsch and van Dijk, 1978; van Dijk, 1980).
McNamara et al. (1991: 491) similarly define propositions as: ‘the
smallest units of knowledge that can stand as separate assertions;
the smallest units that can be true or false’. According to Kintsch
and van Dijk, micropropositions may be linked meaningfully to
each other either structurally, in sentences, or by cohesion or
inferences between sentences. The microstructure consists of a
structured network of all the propositions in a text, shown at dif­
ferent levels of importance, thus representing the common view that
some propositions (main ideas) are more important than others.
Finally, certain of the most important, highest level propositions
are mapped onto a schema, e.g. a problem /solution structure, to
form the macrostructure.
In the discussion of what to test in terms of skills and strategies
(Section 3.2 below) we shall return to the distinction between
micro- and macropropositions. It is useful in distinguishing be­
tween decoding at the local and comprehension at the global
level. An item which tests skimming might be seen as an attempt
to expeditiously establish a superordinate macroproposition; a
discourse topic for a text. Global careful and global expeditious
reading (including both skimming and search reading strategies)
could be seen as the means of establishing the macropropositions
in a text (i.e. main ideas comprehension; cf. Aulls, 1986). Tests
such as scanning for specific detail or gap filling and cloze could
be seen as expeditious and careful reading strategies and skills
that focus on the micropropositional local level (decoding).
Armed with such a structured description of the content of a
text, test developers are in a better position to construct valid
items aimed at different types or different levels of information in
a text. An understanding of the relationship between micro- and
macropropositions is important to the test developer because the
level of focus of items may have important consequences for our
estimates of candidates’ abilities in our tests (see Section 3.2 below
for a discussion of this critical issue).
 Testing reading comprehension(s) 119
A full analysis of a text, as demanded by models such as those
of Kintsch and van Dijk or Meyer (1975), is probably too complex
and time-consuming for use in test construction. Harri-Augstein
and Thomas (1984) suggest more rule-of-thumb apparatus for
arriving at a consensus on, say, main as opposed to minor ideas
in a text and the relationship between them. In Appendix 2
below (pp. 305-6), we propose a form of text-mapping, designed
to serve the same purpose.
In constructing tests it is important to include texts and activit­
ies which mirror as closely as possible those which students have
been exposed to an d/or are likely to meet in their future target
situations. In our view tests should, as far as possible, attempt to
activate real-life reading operations performed under appropriate
performance conditions. We cannot, admittedly, fully replicate
reality in language tests, for the reasons discussed in the opening
remarks in this chapter of the book. Full genuineness of text or
authenticity of task is unlikely to be attainable in the tests we write
(see Lewkowicz, 1997). However, we are still left with the need to
make our tests as valid as possible if we are to measure anything of
value and generalise beyond the test situation to a candidate’s
ability to perform specified operations under certain conditions
to a certain level in real life (Weir, 1993). Our tests must be valid
if we are to relate test performance to real-life reading ability
(B ronfenbrenner, 1976: 5—15; McNamara, 1996; Venezky, 1984:
27; Weir, 1990 and 1993). The solution must lie in our ability to
select appropriate texts, to be read for valid purposes, from which
the reader is able to select an agreed level of meaning under
certain conditions.
The purposes for reading in the test should, wherever possible,
match the purpose (s) for reading those texts in real life. Farr
et al. (1986: 141) point out that ‘A reader’s purpose determines
whether and how a reader interacts with a particular text’. The
purposes for reading in test tasks should be as appropriate as we
can make them (Davies, 1995: 133-4; Grabe, 1991: 378; Weir,
1993). In Section 3.2 we examine in detail the nature of the strat­
egies and skills we might wish to include in our reading tests. It is
im portant that such activities are premised on a strong theoretical
basis, validated as far as possible by systematic empirical research.
We need also to implement faithfully any im portant perform­
ance conditions in developing a reading test. Ensuring that these
conditions are taken account of in text selection will help us to
120 Reading in a Second Language
generalise more directly to performance in the real world on the
basis of data provided by reading tests. We need to consider what
performance conditions - i.e. contextual features - we need to
embody in our reading tests. Weir (1990, 1993) provides an initial
taxonomy of such features and these are developed further in
Section 3.3 (see also Bachman, 1990).
In choosing texts, and in developing activities for readers to
perform on those texts, we need to be as explicit as possible about
our criteria for selection. Such explicitness will help establish a
descriptive framework for generalisations that can be made on
the basis of test performance. Such tests will not tell us everything
about a candidate’s reading ability but they should provide us
with a certain amount of useful, usable information. The value of
such information to the people who use it will be the ultimate
validation of such tests. This approach to testing might be termed
utilisation focused.
The discussion of tasks, texts and formats below will draw upon
the theoretical discussion in Chapters 1 and 2 and is of obvious
relevance to the teaching of reading discussed in Chapter 4.

3.2 TASK-BASED FACTORS

We have already discussed tasks in Chapter 2. At this point we
need merely repeat that testing, as we know it, is all about tasks.
As pointed out in Section 3.1, we have to reject whole areas
of reading as being outside the scope of testing, either because
no consensus in general exists as to relevant tasks or what consti­
tutes ‘successful’ performance on tasks (as in, e.g., reading for
enjoyment), or because readers’ performance on tasks is so un­
predictable that the reliability of items becomes suspect (as with
questions aimed at pragmatic inferences).

Testing components
Until the early 1980s the major concern in the testing of reading
was with the issue of methodology. However, in the 1980s one
noted a switch in concern away from how to test reading towards
a concern with what we were trying to test; a concern with the
nature of the reading construct itself - in broad terms, a move
 Testing reading comprehension (s) 121
away from a focus on method to a focus on the content of reading
tests. The next section examines the componentiality of reading
and the implications of this for the testing of reading.
The issue of how to test, i.e. deciding which are the most
suitable test formats, is still a key decision and we will return to
this in Section 3.4. Once test developers have a clear idea of the
performance conditions that need to be built into a test (see
Section 3.3 below) and the reading skills and strategies that are to
be tested, only then can they make decisions on formats. In the
past the methods’ tail has often wagged the testing dog and for­
mats have been chosen without sufficient thought a priori to what
is being tested.

Deciding on skills and strategies

To satisfy the need for information on an individual’s or a group’s
reading ability, the language tester would normally try to reduce
the possibility of muddied measurement. This might be achieved
by eliminating the influence of reading-irrelevant factors associ­
ated with the test method, and by focusing on a maximally clear
characterisation of the construct of reading per se. The tester might
therefore avoid tasks such as selective summary based on prior
reading of texts - where the extended writing involved in task
completion might interfere with extrapolations we might wish to
make concerning candidates’ reading abilities alone.
However, in certain contexts, such as academic study, such in­
tegration might on the contrary be considered desirable (see Weir,
1983a). Additionally, if reading was the msgor focus the actual
writing could be marked at a low level of expertise with the em­
phasis on successful extraction of content. Whatever decision is
taken with regard to reading and writing, first and foremost, the
tester is obliged to be as explicit as possible concerning the nature
of the ability about which the test is designed to be providing
information.
As was seen in Chapter 2, the current consensus is that reading
is an interactive process involving both bottom-up and top-down
processes. The strength of the interactive model lies in its com­
prehensiveness (particularly with regard to careful reading) and
its applicability to different readers in different reading contexts.
In the operationalisation of such models in teaching and testing
122 Reading in a Second Language
materials there is an attempt to give students the opportunity to
practise across the range of skills and strategies (see Appendix 1
for details of the specification used for an Advanced English Read­
ing Test in China).
Williams and Moran (1989: 224) refer to an apparent current
consensus among writers of teaching materials on the nature of
reading comprehension, stating that:
While materials writers may disagree on the emphasis to be devoted
to any particular skill, there seems to be substantial agreement on
the importance of such skills as guessing the meaning of unknown
words, identifying anaphoric reference, identifying the main idea,
and inference.
In a less pedagogically oriented discussion, Grabe (1991: 379-82)
may also be seen as representative of current views. He comments
on the importance of automaticity in reading, particularly in word
identification, and also identifies as components of skilled read­
ing: syntactic knowledge; knowledge of formal discourse structure
(formal schemata); content and background knowledge (content
schemata); and metacognitive knowledge and skill monitoring,
e.g. recognising the more important information in a text, skim­
ming, and searching for specific information. Grabe (1991: 382)
concludes:
A ‘reading components’ perspective is an appropriate research
direction to the extent that such an approach leads to important
insights into the reading process. In this respect, i t ... is indeed a
useful approach.
Carr and Levy (1990: 5) comment: ‘this approach attempts to
understand reading as the product of a complex but decompos­
able information processing system’. They argue that the mental
operations are distinguishable and empirically separable from each
other (see also Chapter 2 for an extensive discussion of these
components).
Such a focus would seem to accept that reading can be broken
down into underlying skills and strategies for the purposes of
teaching and testing - a view shared by many language teachers
and testers, though opposed by others. A wealth of taxonomies is
available in the literature (see Section 2.3 above). Here, for pur­
poses of reference, and to help move the discussion forward, we
provide in Table 3.1 a recent four-level version of such a break­
down based on Weir (1993) and Pugh (1978).
 Testing reading comprehensions) 123
Table 3.1 Matrix of reading types
Global Local
Expeditious A. Skimming quickly B. Scanning to locate
to establish discourse specific information;
topic and main ideas. symbol or group of
Search reading symbols; names, dates,
to locate quickly figures or words.
and understand
information relevant to
predetermined needs.
Careful C. Reading carefully to D. Understanding syntactic
establish accurate structure of sentence and
comprehension of the clause. Understanding
explicitly stated main lexical and/or
ideas the author wishes grammatical cohesion.
to convey; propositional Understanding lexis/
inferencing. deducing meaning of
lexical items from
morphology and
context.

The reader will note that Williams and Moran (1989) and Grabe
(1991) identify reading components from all four levels A, B, C
and D in Table 3.1. Their inclusion of word-level components
relating to more specifically linguistic comprehension is indicative
that this is seen by many people as an important part of reading,
not as something separate. The contribution of the latter to tests of
reading ability is an important issue that will be taken up below.
Global comprehension, which can be related to Kintsch and
van Dijk’s macrostructure (Kintsch and van Dijk, 1978) normally
refers to comprehension beyond the level of micropropositions -
from macropropositions to discourse topic. Local comprehension
refers to the decoding of micropropositions and the relations
between them.5

The evidence for a multidivisible view

It is held by many teachers, textbook writers, and the constructors
of language tests that reading is made up of different skill and
124 Reading in a Second Language
strategy components, and that students may exhibit differences
in level of proficiency across these. It is moreover often claimed
by practitioners that sets of reading components provide useful
frameworks on which to base course design, teaching, and test
and materials development (see Lumley, 1993).
There is research evidence in the literature, albeit limited, that
reading is not simply a general factor dating back to Gates (1926)
(see also Carnine and Silbert, 1979; Davis, 1968, 1972; Hillocks
and Ludlow, 1984.)
Dubin et al. (1986: 163) argue that:
interactive models (of the reading process) suggest a need to test
skills at many levels, since all these skills are assumed to play a
significant role in the reading process. They include everything
from rapid identification of vocabulary and syntactic structures,
to the interpretation of larger discourse patterns, the making of
inferences etc.
Thus processing at the level of word recognition and lexical access
as well as at the level of integration of textual information and
resolution of ambiguity are seen as important. The focus of both
linear and interactive models discussed in Chapter 2 was largely
on decoding rather than on comprehension and they do not in
themselves account for higher level text comprehension. As we
noted earlier, Kintsch and van Dijk (1978) are among the few
researchers who focus on comprehension and offer a theory of
text structure (see also Just and Carpenter, 1987).
Perhaps as a result of the lack of any clear-cut theoretical model,
it is widespread practice among test developers to focus on these
reading components, either singly or in combination, when con­
structing test items relating to a text, even though the sum of
these parts - the answers to these test items - might not necessarily
equate fully with what the reader would normally take away from
the text. Indeed, whatever theoretical position the test developer
takes, the need to construct individual test items will exert strong
pressure to attempt to measure individual reading components
and strategies, or combinations of them. If specific components
or strategies could be clearly identified as making an important
contribution to the reading process, then it would of course be at
least possible, if not necessary, to test these and to use the com­
posite results for reporting on the reading proficiency revealed. It
would moreover be necessary, once hypothesised skill components
 Testing reading comprehension (s) 125
and strategies were substantiated, to determine the nature and
strength of interactions between them. It is, however, as yet too
early to explore such interactions; thus far we can have but little
idea whether any components may be superordinate to others,
or the extent to which ‘higher order’ operations within reading
depend on ‘lower order’ operations.

Casting doubt on multidivisibility

Despite the widespread influence of a multidivisible view of read­

ing on current practice, caution needs to be exercised. Accept­
ance of such a view by practitioners tends to be anecdotal, and
based on pedagogical ‘experience’. Fyfe and Mitchell (1985: 4)
comment:
no analysis of reading skills or the reading process has yet achieved
general acceptance. There is agreement that word recognition can
be distinguished from reading comprehension, although the dis­
tinction is funny at the boundary, but there has been argument for
decades over the existence of different comprehension skills.
In opposition to a multidivisible view of reading, a substantial
number of studies have found that it is not possible to differenti­
ate between reading components, either through empirical demon­
stration of the separate functioning of such components when
these are operationalised in language test items, or through the
judgem ent of experts on what the focus of such test items actually
is (see, e.g., Alderson, 1990a; Alderson and Lukmani, 1989; Carver,
1992; Rosenshine, 1980; Rost, 1993).
This, however, seems at odds with the componential views of
reading (see Sections 2.1 and 2.2 above) that are suggested in the
theoretical literature. The two- or three-component models we
mapped out in Chapter 2 conflict with a unidimensional view of
reading. First we shall examine in detail further evidence from
language testing that, at the very least, there are two components
in L2 reading which we might characterise broadly as global and
local. We then consider recent test data that suggest a further
distinction between expeditious and careful reading, which is con­
nected to the third component of literacy discussed in Section 2.2
above. As we emphasise later in Section 5.3, language testing is
126 Reading in a Second Language
the methodological procedure that is most likely to provide valid
and reliable information on this issue.

Quantitative research
A number of empirical test-based studies, typically using factor
analysis, have cast doubt on the multidivisible nature of reading
(e.g. Lunzer et al., 1979; Rosenshine, 1980; Rost, 1993). Factor
analysis is a statistical procedure for extracting the extent to
which putatively different variables - in our case the so-called
‘skills and strategies’ in reading; reading types - in fact function
in a similar manner. If a number of putatively different skills and
strategies function in a very similar manner it is said that they
‘load on the same factor’ and we have at least to entertain the
possibility that they are not different at all, only a single construct
in different guises. If all conceivably different skills and strategies
load on a single factor, we have to consider the strong possibility
that there are in fact no skills and strategies at all, only* a single
undifferentiated ability: reading. If some putative skills and strat­
egies function in a statistically similar manner and load fairly heavily
on one factor, while other putative skills and strategies function
statistically in another manner and load on a second factor, then
this is evidence that reading is at least bi-divisible.
For example, the work of Lunzer et al. (1979) is often cited by
reading specialists as evidence that it is not possible through test
data to differentiate between the so-called subskills and strategies
in reading. This study is said to show that reading (at least as defined
by completing their reading tests) is a single undifferentiated abil­
ity. However, it is interesting to note that while only one principal
factor - presumably undifferentiated reading - is identified in this
study through factor analysis, there does appear to be some doubt
(Lunzer et al., 1979: 55-7) concerning the strength of the loading
of test items testing word-meaning on that principal factor.
The reader must also remember that Lunzer et al.’s study
(as with many of those finding no evidence of multidivisibility)
was conducted on native speakers of English; in fact primary school
pupils, presumably still largely free of the specific linguistic prob­
lems experienced by some non-native speakers. Our particular
concern in this book is different from that of Lunzer et al. in that
we are primarily interested in testing adult L2 readers who will
tend to be spread out across the language ability range.
 Testing reading comprehension (s) 127
The most recent investigation conducted by Rost (1993), again
on native speakers, found strong evidence of unidimensionality,
leading Rost to warn against differential skill component inter­
pretation for all available reading comprehension tests (1993: 88).
However, once again, it is important to note that, in the reported
factor analysis, a second factor that Rost believes to be vocabulary
did emerge when the factors were rotated. Rost (1993: 80) indeed
cites earlier research where ‘two factors of reading comprehen­
sion, namely “vocabulary” or “literal reading” on the one hand,
and “general reading comprehension” or “inferential reading” on
the other’ emerged from the data (Johnson and Reynolds, 1941;
Stoker and Kropp, 1960; Vernon, 1962; Pettit and Cockriel, 1974;
Steinert, 1978).
There is further evidence in the literature that the phenom­
enon of vocabulary loading on a separate factor is not uncommon.
Davis (1944) identified two important separate factors in read­
ing as ‘memory for word meanings’ and ‘reasoning in reading’
(a combination of weaving ideas together and drawing inferences
from them). Similarly, in a later study (Davis, 1968) a recogni­
tion vocabulary test accounted uniquely for a sizeable proportion
(32 per cent) of the non-error variance. There is also evidence in
Spearritt’s reanalysis (1972) of Davis’s earlier data that vocabulary
tests are differentiable from the single basic ability ‘reasoning in
reading’ measured by other labelled reading components in the
reading comprehension tests used in the study. Spearritt (1972:
110) concluded:
Vocabulary is the best differentiated, as in both the Davis and
Thorndike analyses ... it could not in fact be subsumed under one
general factor with the other three skills.
Similarly, Rosenshine (1980: 543) admits to the fact that in three
out of the four analyses done on Davis’s data the one unique factor
that emerged as separate from the others was vocabulary (‘remem­
bering word meanings’), the only exception being Thorndike’s
(1973) analysis which he categorises as being less sophisticated than
Spearitt’s later study. Rosenshine cites data from Berg in support of
the non-divisibility position, but in four out of the five studies sum­
marised by Berg (see Rosenshine, 1980: 544) lexical competence
appears as a separate factor. Farr (1968) found two factors - one
clearly vocabulary, which loaded heavily on the three vocabulary
measures, and one that could be labelled as comprehension.
128 Reading in a Second Language
Though the quantitative studies reported above seem to suggest
that it may not be consistently possible to identify multiple, separate
reading components, there does seem to be a strong case for con­
sidering vocabulary as a component separate from reading com­
prehension in general. Given that most factor analyses in the studies
reported above produced more than one factor, it would be difficult
to maintain that reading is a unitary ability. Furthermore, even if
the components which load more heavily on a second factor also
load on the first general reading factor, it might be appropriate
only to select test items which load heavily on the first factor when
developing a measure of general reading ability. Alternatively, if
vocabulary is considered to be part of reading, a bi-divisible view
of reading would seem to be more appropriate.
Qualitative research
As part of a new wave of qualitative investigation in language
testing studies, Alderson (1990a; see also Alderson and Lukmani,
1989) investigated the reading component question through the
judgem ent of experts on what reading test items actually test. In
this study, groups of experts —usually students on MA courses —
were presented with a long list of posited reading components,
and asked to identify cold (‘heuristically’) what items in a pilot
version of an EAP reading test were measuring in terms of the list.
The resulting lack of agreement on assigning particular ‘skills and
strategies’ to particular test items, i.e. on agreeing what an item
was testing, and even whether an item was testing a ‘higher level’
or ‘lower level’ component, could be taken as evidence of the
indivisibility of the reading ‘skill’, or at the very least could be
seen as casting doubt on the feasibility of distinguishing reading
components. Nevertheless, these conclusions need to be subjected
to scrutiny.
The authors of the reading test items used in the Alderson
(1990a) study were aware of the possible overlap between compon­
ents tested by individual items. At the time, Weir (1983a: 346)
had summarised the approach to the design of the reading com­
ponent in the TEEP as follows (for ‘skills’ in this quotation, read
‘skill components’):
... we aimed to cover as many of the enabling skills in each of the
reading subtests ... as was feasible ... we indicate opposite each
item in the reading sub tests what the Project Working Party and
 Testing reading comprehensions) 129
other experts in the field considered to be the major focus of that
item. We were aware that though an item might be seen to be
dependent on a particular enabling skill for successful completion,
other skills and strategies might be contributing to getting the answer
right. We realised that the skills and strategies we were sampling
were not necessarily discrete.
Thus any conclusions regarding the feasibility of distinguishing
separate components, based on the inability of judges in the
Alderson (1990a) study to agree on what single component was
tested by individual items, must necessarily be open to question.
(For a discussion of further weaknesses in this study, see Weir et
al., 1990 and Matthews, 1990). Furthermore, any similar investiga­
tions in this area should ensure that the experts involved share a
common understanding of the categories of description employed
in the study. Lumley (1993) emphasises the need for clear defini­
tions and a common understanding of the terms employed, in
particular ‘higher level’ and ‘lower level’ components, if the attempt
to assign components to test items is to be meaningful.
In contrast to Alderson’s findings, there is an alternative liter­
ature which suggests that it is possible with clear specification of
terms and appropriate methodology for testers to reach closer
agreement on what skills and strategies are being tested (Anderson
et al., 1991; Brutten et al., 1991; Kobayashi, 1995; Lumley, 1993;
Weakley, 1993; Weir et al., 1990).

Casting doubt on unidimensionality

It is im portant to note that, despite the variability in teachers’
judgements in Alderson’s study, Alderson (1990b: 465) describes
how ‘there was more or less agreement’ when an item was con­
cerned with deducing the meaning and use of unfamiliar lexical
items, or involved understanding relations between parts of a text
through cohesion devices. If we re analyse Alderson’s data in terms
of our broader categories listed in Table 3.1 above, in particular,
types C (careful global) and D (careful local), it turns out that
there is a majority agreement among the judges on which of
these types the items would fall in nine out of the ten cases selected
by Alderson (1990b: 466). If one were to discount the scanning
felt necessary to locate one of the items, then there would also
be agreement that this item would fall into D-type operations.
130 Reading in a Second Language
It therefore seems likely that even untrained judges are able to
determine when an item is dependent upon specifically linguistic
knowledge (D-type operations) as against focusing on more global
comprehension (A- and C-type operations). Alderson himself would
seem to make this distinction when he classifies without demur
gap-filling items in the TEEP battery as focusing on local D-type
reading (Alderson 1990a: 433).
So, perhaps not too surprisingly, it does appear that judges are
able to distinguish items which focus upon specifically linguistic
knowledge at the word level. Additionally, there is the quantitative
evidence from a number of the empirical studies referred to earlier
which seems to suggest that these items may well load on a factor
separate from that on which other more global items load in
reading comprehension tests. Here are two sets of evidence which
seem to suggest a bi-divisible view of reading, at least as far as word
meanings and reading comprehension in general are concerned.
There is thus both qualitative and quantitative evidence for con­
sidering specifically linguistic elements as potentially separable
from global comprehension. This is a long way from multidivis­
ibility, however.

The status of expeditious reading

There has for a long time been evidence from survey data that L2
readers found particular difficulty in reading quickly and efficiently
in the target language (Weir, 1983a). Slow careful reading also
poses problems but the difference between LI and L2 readers is
most marked in expeditious reading. For reasons which are diffi­
cult to explain, dedicated tests of the latter ability have not fea­
tured in examinations with the notable exception of the Test for
English Majors in the People’s Republic of China which has had a
separate section on this since 1990. Data from this test support
the view that candidates perform differentially in this section as
against the careful reading section (see Shen et al., 1998).
There is further empirical evidence emerging of differences
in performance on level A operations (reading expeditiously for
global comprehension) and level C operations (reading carefully
for global comprehension) in the new Advanced English Reading
Test being developed for undergraduates also in the PRC (see also
Carver, 1992; Guthrie and Kirsch, 1987; Weir, 1983a).
 Testing reading comprehension (s) 131
Why have such differences not emerged in almost fifty years of
research on this issue? The answer is depressingly obvious. Given
that the research instruments used in the studies reported above
almost universally failed to include items testing expeditious read­
ing (skimming, search reading or scanning) then their claims to
have demonstrated that reading is a unitary ability would seem
to be questionable. If one does not take the time and trouble to
operationalise these strategies carefully in tests, then one should
not be surprised that careful reading tests are just that - tests of
careful reading with a possible division between global and local.
Given the stranglehold this view of reading has had on research
due to the different agenda of psychologists, it is perhaps not
surprising that, with a few exceptions (Pugh, 1978), little attention
has been paid to expeditious reading.
Careful reading models have little to tell us about what happens
in expeditious reading. We noted in Chapter 2 that Rayner and
Pollatsek (1989: 477-8), who provide one of the clearest accounts
of the reading process, had to admit that there is little hard infor­
mation on this. Paris et al. (1991: 633) confirm this:
Testing is a mainstay of US education, and students endure a wide
variety of criterion-referenced and norm-referenced tests every
year. But educational tests of reading have not changed to conform
with our notion of strategic reading. Instead, they are surprisingly
uniform. The common format of most reading tests requires the
students to read brief paragraphs and answer multiple-choice ques­
tions about them. Although decoding, vocabulary, syntax and other
features of language are often tested, comprehension scores are
usually derived from reading several short paragraphs. Most of these
paragraphs are disembodied prose - they do not have titles, pic­
tures or structures like the selections used in basal readers or text
encountered in content areas.
De Leeuw and de Leeuw (1965) were among early proponents
of the need (p. 10) ‘to use a strategy of reading’, i.e. read in different
ways, at different speeds for different purposes. They argued the
importance of efficient reading involving selectivity and the use of
expeditious reading strategies (see also Pugh, 1976, 1978). Expedi­
tious reading involves the conscious use of strategies to sample a
text in the most efficient fashion in line with a particular purpose
(see Paris et al., 1991, for a theoretical description of strategies as
actions selected deliberately to achieve particular goals).
132 Reading in a Second Language
Top-down processing may help to determine the sampling strat­
egy with a reversion to conventional bottom-up processing when
closer attention is paid to the selected parts of the text. The process
may well differ from that of careful reading in important aspects
(Rayner and Pollatsek, 1989: 477-8).
Van Dijk (1977: 79) suggests:
... we also have explicit discourse cues to select correct macro
structures. Such clues include titles, initial summaries and declara­
tion of content/intention. The reader by convention interprets such
properties as approximate indications of the global meaning of the
discourse. Additional cues are provided by expressions that indicate
the relative importance of certain thematic propositions, expres­
sions such as The crucial point is . . . And then the most important thing
happened. . . , etc. They are textual warrants for the plausibility of
some hypothesis concerning the macro-structural. . . relevance of
some proposition.
These clues may assume great importance in skimming, in par­
ticular, where we are attempting to establish expeditiously a macro­
structure and, to a lesser extent, in search reading where we are
seeking the answers to predetermined questions on the macro­
structure. They are the means whereby we can ‘graze’ the text to
select parts which are more important for establishing a macro­
structure quickly. Careful reading of text will embrace these but
at the same time the whole of the microstructure is taken in as
macro-rules are applied to get at an underlying macrostructure in
the text.
All this is not to deny that different readers would appear to
use different strategies/skills to extract the author’s message in a
text (Alderson, 1990a: 436, 1990b; Storey, 1995; Weir, 1981: 34
and 1983a: 346). There may also be a degree of overlap in some
hypothesised components (Weir, 1983a: 346); for example, it is
possible that the act of quickly skimming through a passage to get
an overall idea of the content will activate various schemata and
establish the basis for closer reading by drawing on both the main
ideas and important details (see Lee and Musumeci, 1988). Thus
the interactive parallel processing Buck (1990) talks about in rela­
tion to listening may also make it difficult in reading to establish
the completely separate existence of the types of reading A, B and
C above. This is particularly the case where questions from differ­
ent levels are set on the same passage. It may be necessary, there­
fore, to use different texts to test the separability of each of these
 Testing reading comprehension(s) 133
skills and strategies. Basing tests on a single passage may have
been a confounding variable for much of the earlier research in
this areal At the very least, independence of items must be in
doubt in a num ber of the studies reported above. Strict time
controls will also need to be imposed on the passages within these
separate tests to ensure that each passage that is meant to be read
expeditiously is in fact processed in that manner. Conversely, care­
ful reading tests should allow candidates more time than they will
need, as in real life such reading is not normally constrained by
time conditions.

Testing at the microlinguistic level

Research often ‘testifies’ to the importance of word recognition.
Stanovich (1991: 423) in a recent survey notes that research studies
have indicated that ‘word recognition is a fundamental component
of reading comprehension’.
The ability to perform reading types A-C above is obviously
dependent on a certain level of competence in the microlinguistic
skills detailed as type D. Alderson (1984: 19) termed this ‘the
threshold level’ (see also Carrell, 1991; Coady, 1979; Hudson, 1988;
Samuels and Kamil, 1988; Stanovich, 1980 and the discussion in
Section 2.2).
It does seem improbable that students would be able to work
out the main ideas of a text without some baseline competence
in the microlinguistic skills, without understanding some of the
relations within at least some sentences of that text (Alderson
and Urquhart, 1984; Alderson and Lukmani, 1989; Carrell, 1991;
Clarke, 1980/1988; Devine et al., 1987; Eskey, 1988; Grabe, 1991:
391; Stanovich, 1980; Storey, 1995; Weir et al., 1990). However,
the degree to which the reader needs these lower order abilities is
not yet clear and may prove difficult to quantify (Weir et al., 1990:
508). An interactive or interactive-compensatory view of read­
ing (Rumelhart, 1977; Stanovich, 1980) would seem to imply that
readers can make differential use of a range of components which
we might loosely label ‘specifically language related’ and ‘reason
related’ (Grabe, 1991; Williams and Moran, 1989).
The simple answer is that we cannot clearly state what the value
of the contribution of skills and strategies at the local level is. In
those countries such as China where the aim is to encourage
134 Reading in a Second Language
learning to bring students up to and beyond this threshold level,
there is naturally some concern that solid foundations in lexical/
syntactic knowledge need to be laid. If it is in the test, it will happen.
If it is not a recognisable component of the new Advanced English
Reading Test (AERT), then it might not.
Given the evidence against both fully unitary views of reading
ability on the one hand, and multidivisible views on the other, it is
important for those concerned with language test development
to reflect critically on the ramifications of operationalising either
in reading tests.

Dangerous implications of a unitary view

There is a disturbing corollary of the fully unitary argument that
deserves serious attention from all involved in developing language
tests. If there are no discernible components in reading - i.e. if it
is unidimensional - it should not really matter how we test it, or
what operations we try to assess. The inability to provide consistent
or conclusive empirical evidence (either quantitative or qualitative)
for the separability of components might well encourage us to utilise
test formats with a specifically linguistic focus (D-type operations
in Table 3.1 above). These are often relatively easy to construct,
administer and mark, frequently have respectable psychometric
properties and reliability estimates (usually high internal consist­
ency estimates), and frequently correlate fairly highly with more
global tests of reading comprehension.
Given that the current consensus on the nature of reading
includes microlinguistic elements as important components within
reading (Grabe, 1991; Williams and Moran, 1989), and given the
evidence for a unitary view of reading (Lunzer et al., 1979; Rosen­
shine, 1980; Rost, 1993), it is perhaps not surprising that many
current reading tests have in fact ventured down this micro-
linguistic road (quite often for reasons of practical expediency
rather than from a principled view of unidimensionality). In the
British Council/UCLES ELTS test, one-third of the items in some
of the reading modules appeared to be testing at the microlinguistic
level, and in the G1 General Reading Comprehension Module,
Criper and Davies (1988: 89-97) indicated that over 50 per cent
of the items focused on such lower order elements. The more
recent IELTS test has a number of items at this level (see Weir,
 Testing reading comprehension(s) 135
1990); so do the TEEP test (see Weir, 1990), TOEFL, and many
other tests of international standing and good repute. It is thus
fairly urgent that the status of such lower order elements in the
measurement of reading ability be investigated.
There are serious question marks against the value of testing
directly the specifically linguistic D-type operations listed in Table
3.1 above. The evidence from the literature and our own initial
investigations throw some doubt on the value of including any
items which focus on specific linguistic elements (e.g. individual
words or cohesive devices) in tests which purport to make direct
statements about a candidate’s reading ability. Tests focusing on
level D microlinguistic elements may well correlate quite highly
with tests that attempt to tap into global facets of that ability, but
they do appear to load on a different factor from general reading
comprehension in many studies. It is, in our opinion, unarguable
that level D elements contribute to level A and C activities, but
on their own they may not constitute an adequate predictor of
that ability.
The limited data reported below are intended to raise aware­
ness of the possible danger that some candidates might be disad­
vantaged by the inclusion of such discrete linguistic items in tests
of reading comprehension (level D operations in Table 3.1 above)
where the purpose is to indicate whether a candidate has under­
stood the main ideas and important detail provided by the writer,
i.e. the reader has understood the text (A and C reading types in
Table 3.1). Further data sets are also cited which cast doubt on
the use of specifically linguistic test items and the formats with
which these are commonly associated to test reading ability. We
are aware that our argument smacks of the assertive at present
and that further data will need to be generated to prove or falsify
our fears. It is a purpose of this chapter to generate interest in this
issue so that a concerted body of empirical research might throw
further light on it.

Dangerous implications of a multidivisibility view

The dangers of a unitary view of reading have been outlined above,
in the light of evidence concerning the status of items testing at
the specifically linguistic level. Equally serious problems may arise
in accepting unthinkingly a multidivisibility view. This ‘scattergun’
136 Reading in a Second Language
approach to the testing of components in reading, as referred to
above in connection with the original TEEP test (Weir, 1983a,
1991) likewise needs to come under close scrutiny. In the past,
broadly sampling components across the four types (A-D) seemed
a sensible course of action for assessing reading comprehension.
By attempting to sample across the range of hypothesised ‘skills
and strategies’, the intention was to take an adequate sample of a
construct that could be labelled reading (Weir, 1983a, 1990).

The University of Reading data

Doubts now arise concerning the relative contribution of the vari­
ous components to measure the construct of reading, at least in
terms of the relationship between type D and types A and C. In
the placement assessment (broadly styled on the TEEP approach
described above) administered to students recently entering the
presessional EAP courses at Reading University in July and August
(candidates arriving range from roughly around band 5 on the
IELTS scale to band 6.5), it was noted that there were a number
of students who might be able to cope quite well with reading
passages and questions at the global level, but that this was not
matched by their performance on test items focusing on more
microlinguistic items: cohesion markers, lexis and structural ele­
ments. Regardless of the item-type and regardless of their score
on the microlinguistic items, students scored overwhelmingly in
the top 50 per cent of the available score range on global items;
scores on microlinguistic items, on the other hand, were more
evenly distributed on either side of the 50 per cent line, with
a slight tendency to fall below rather than above the line. This
situation initially gave pause for thought as to how to allocate
such students to classes.
However, of far greater concern were the implications of using
specifically linguistic test items later on in the programme for
assessing proficiency in reading prior to entrance into the univer­
sity proper. Here there was a serious dilemma in interpreting test
results. Doubts arose about the wisdom of using tests or items which
appeared to focus on microlinguistic elements. To be more precise,
there was major concern about the fairness of such tests for the
subset of students who did ‘well’ on the global comprehension
items in our reading tests (at the moment for descriptive purposes
arbitrarily defined as more than half the items right) but did not
 Testing reading comprehension(s) 137
do so well on the microlinguistic items (less than half the items
right). Such a situation may occur for a variety of reasons, e.g.
successful application of background knowledge to the text under
review an d/or transfer of higher level processing and strategies
from the LI which compensate for deficiencies in lower level
linguistic abilities (Coady, 1979; Goodman et al., 1979; Hudson,
1988). This group of candidates must in any case have passed a
threshold level of language proficiency necessary for this transfer
to take place in order for global questions to be correcdy answered
(Alderson, 1984; Clarke, 1980/1988; Cziko, 1980; Devine, 1987,
1988).
The low loading of vocabulary on a general reading compre­
hension first factor in some of the analyses reported earlier in this
chapter is consistent with these findings, in that such results may
reflect that candidates can comprehend the overall ideas in a
passage irrespective of specific linguistic shortcomings. If we try
to relate this to the research on the organisation of prose and its
effect on memory (see Section 2.2 above), the work of Meyer
(1975: 165, 167) shows that:
the height of information in the content structure influences its
recall. . . ideas located high in the content structure are better re­
membered than ideas low in the structure.
She posits that this may be because central ideas high in the content
structure subsume ideas located low in the structure; ideas in high
structural positions are more retrievable from memory or readers
most likely select ideas high in the content structure to process
for long-term storage rather than low level ideas. This may help to
explain why we have found that, in many cases in our tests, global
comprehension of a text is superior to local comprehension.
The evidence will need to be supplemented by further, more
detailed studies. What an adequate performance on the two item
types is will need to be more precisely operationally defined. Simply
taking a 50 per cent cut-off point on each test, while useful for
descriptive purposes at the early stage of our inquiry, is rather
crude. Furthermore, it is possible that a group of candidates might
perform worse on microlinguistic items as a normal feature of
score distribution. Moreover, it needs to be investigated whether
the same candidates are similarly distributed over a num ber of
measures of each of the bi-divisible elements identified. Recent
data (Khalifa, 1997) suggests this to be the case.
138 Reading in a Second Language
There is also a possibility that the data we present may have
been unduly influenced by the formats employed; some candidates
may have reacted adversely to the gap-filling format used to assess
microlinguistic knowledge. However, synthesis of information avail­
able from the variety of sources reported below would seem to
suggest that format effect would be unlikely to account for our
results. Evidence from short-answer and multiple-choice formats
indicates that the items in reading comprehension tests which
focus on microlinguistic elements such as lexis or cohesion do
not necessarily contribute to the overall measurement of reading
in ways similar to items which test more global comprehension, in
that they can be shown to correlate more highly with their own
subtest or with other microlinguistic elements in a test battery
(such as separate subtests of vocabulary and structure).

,
Data from ESP Centre Alexandria9 Egypt
Similar differential performance is emerging in data from a bat­
tery of EAP tests under development by the Testing and Evalua­
tion Unit at the ESP Centre in Alexandria, Egypt. These are
short-answer question tests designed to test, separately, reading
comprehension types A, B, C and D in Table 3.1 above. Point
biserial correlations show that items in section 1 of the battery
testing at levels A and C correlate more with their own subtest
than they do with level D (the microlinguistic items), and vice
versa. Similarly, there is a small subset of students who, while
coping well with A and C operations, experience more difficulty
with the specifically microlinguistic items. So here again there is
the same phenom enon of differential performance on global as
against specifically microlinguistic items. It should be noted that,
in the Alexandrian case, the reading test format differed from
that used in the Reading study.

Data from College English Test (CET), China

Evidence of differential performance appears in further test con­
texts. Items which focus on cohesion or working out the meaning
of words in context appear to be out of place in a recent version
of a reading component of the College English Test (CET) used to
test over 1.8 million undergraduates across China. Such items, accor­
ding to item /part correlational data (Pearson Product Moment),
 Testing reading comprehension (s) 139
would appear to relate more closely to a later section in the test
which focuses on vocabulary and structure than they do to the
section designed to test reading comprehension. The format used
in this test is multiple-choice, so the phenom enon would appear
to repeat itself here in a third format. This is further evidence that
the results obtained on the microlinguistic as against global focus
are probably not the effect of item type.

Data from the Advanced English Reading

Test (AERT), China
A new dedicated reading test designed for undergraduates at Chi­
nese universities is also producing data which suggest that all 15
items in the part of the test focusing on careful global compre­
hension load on a different factor than do all the 15 items in the
part testing careful local comprehension at the word level (see
Chapter 5, Table 5.2). Additionally items in the skimming and
search reading part testing expeditious global comprehension tend
to load on a different factor than those items testing slower, careful
reading. In fact, varimax rotation of the data generated by the
second pilot on 1100 candidates suggests a four-factor structure in
line with the four-part division outlined in Section 3.1 above.

Data from Alderson and Lukmani

Alderson and Lukmani (1989: 269) found that:
. . . what seems to be happening is that weaker students overall do
somewhat better on higher order questions than lower order ques­
tions . . . perhaps lower order questions measure language ability
whereas higher order questions might be said to measure something
like cognitive skills, logic, reasoning ability and so on ... it might
perhaps be possible to get a better estimate of a student’s reading
abilities . . . from higher order questions rather than from lower
order questions . . . one should not perhaps believe that students
with lower language levels are incapable of answering higher order
questions. In other words, one should not be inferring from poor
performance on lower order questions an inability to perform well
on higher order questions.
In proficiency tests, the issue of validity is crucial, given the use
to which the results of such tests are normally put. If items with a
specific linguistic focus are used in the measurement of reading
140 Reading in a Second Language
ability, and if the results are then used in taking decisions on
entrance to further study or to the professions, unfairness might
result. Tests including such items might discriminate against
the linguistically disadvantaged but otherwise competent reader.
Those students who can understand almost all of the main ideas
and important information in a passage - i.e. those who can clearly
establish the macrostructure, but are unsure about the meaning
of particular lexical items or cohesion devices, such test items
selected for inclusion perhaps subjectively or idiosyncratically -
may not pass through the entrance gate which the test embodies.
To this extent the idea of reading comprehension as either purely
unitary or multidivisible would appear to be dangerous.
The only sensible solution until research reveals a clearer pic­
ture of this issue is to profile abilities in each of the areas of our
framework. In this way both strengths and deficiencies might be
revealed rather than being hidden behind a potentially mislead­
ing single score or grade. Different candidates will exhibit differ­
ent strengths and weaknesses, so, in the interests of fairness alone,
different sections of the test should be weighted equally. Profiling
of ability in the four levels of our taxonomy of tasks will remove
the need for weighting. The users of the profile are in the best
position to determine how performance at the different levels is
relevant to their needs.
As well as clearly establishing the operations we wish our test
tasks to involve, there are a number of additional considerations
relating to test task development. We turn next to the issue of the
conditions under which the operations discussed above might need
to be performed.

3.3 TEXT-BASED FACTORS

There has often been a tendency to focus on the reader rather
than the text in reading research. With advances in text linguistics
over the last twenty years, text should have an equal status in test
development with reader, task and output.

Text type
Unfortunately, as we noted in Chapter 2, many text analysis pro­
cedures are so detailed and produce so much data that they can
 Testing reading comprehension (s) 141
be of little value to testers in making decisions on whether or not
to select a text for a test. Deciding what are appropriate text types
for the test population is a crucial step in test development. This
decision is currently best informed by needs analysis of the students’
target situations and by careful examination of the texts (and tasks)
used in other tests and teaching materials aimed at the particular
test population. (See Section 5.3 on page 274 for exemplification
from the development of the Advanced English Reading Test
(AERT) project in China.)
There has been a consensus for a num ber of years that texts
used both for teaching and testing should be ‘authentic’, though
this requirem ent has become more a matter of common sense
than, as originally, of almost missionary dogma. There has, however,
been a suggestion that in dealing with heterogeneous test popula­
tions fully genuine texts are not essential. This is a view supported
by the work of Lewkowicz (1997) which indicates that texts might
only need to resemble those that the candidates will process in
the future in terms of salient ‘authentic features’. She argues that
full authenticity of text may not be necessary, attainable or desir­
able. However, until such findings are substantiated by stronger
evidence than her initial pilot studies, we would be best served
by selecting texts which exhibit as many salient features of target
situation texts for the population as is possible.
Obviously the skills and strategies it is wished to test will also
influence selection: problem/solution, causative or comparison
texts from journals or textbooks may well lend themselves better
to testing reading carefully for main idea(s) comprehension than
more descriptive texts with lots of detailed information. Though,
as Carrell (1984: 464) points out, this might be unfair to certain
native language groups such as Arabs for whom it is a preferred
rhetorical pattern and more facilitative of recall than other patterns.
In careful reading the texts may not necessarily have clear main
ideas for selection and main ideas might have to be constructed
through propositional inferencing, whereas in skimming and search
reading they should be explicit.
Where candidates are expected to skim or search read lengthier
texts, these would ideally have a clear overt structure and be clearly
sequenced with a clear line of argument running through them.
A journal article or chapter from a textbook with clear sections
and headings, and where paragraphs contain topic sentences in
initial position which signal the information to be presented, may
142 Reading in a Second Language
prove suitable for testing these expeditious reading strategies. Prob­
lem and solution, causative and comparison texts may have the
clearest, tightly organised structures (Carrell, 1984; Meyer and
Freedle, 1984; Meyer, 1975). One might also look for texts which
are overtly organised into sections. Texts without a clear structure
may well be authentic but they do not lend themselves easily to use
in testing expeditious reading, just as in real life they are difficult
to follow quickly, to summarise or to make notes on. A collection
of description texts (Carrell, 1984; Meyer and Freedle, 1984) may
be the best vehicle for testing scanning for specific detail.
We realise that the guidelines we have presented are, at best,
rather skeletal. Lewkowicz (1997) points out that a key area for
future research is in determining the text types that allow the best
testing of the various skills and strategies. Kobayashi (1995) has in
fact made a good start in this area. She found (pp. 266-7) that:
there seems to be a close relationship between text type and question
type. For example, tightly organised texts tended to produce more
questions on main ideas than less organised texts. At least, it seemed
easier to generate a variety of questions when texts were highly
organised. If texts were loosely organised, on the other hand, ques­
tions tended to focus on details or literal understanding.
At lower levels, texts employed in tests are often artificially con­
structed or simplified because of the restrictions imposed by the
structures and lexis available to the students. This may seriously
constrain the range of strategies and skills that can be tested and
it may be that expeditious strategies are simply not testable at this
level because of length constraints.
In Section 4.3 we return to the issue of text selection for teaching
purposes and a num ber of criteria for selecting text are explored
in detail.

Propositional content

In most test populations we are dealing with a heterogeneous

body of people in terms of their interests, backgrounds and occu­
pational, academic or professional situations. Thus we need to
select texts which are within the experience of the whole test
population. They should be at an appropriate level in terms of
propositional content and candidates should be similar to the
 Testing reading comprehension^) 143
audience originally addressed by the writer. Weir (1993) emphas­
ises that the relationship between the writer of the text and
the reader - e.g. expert : layperson - needs to be considered at
the selection stage. A text written for a different audience than the
candidates may not be appropriate.

Topic familiarity
Topic familiarity is increasingly seen as one of the criterial deter­
minants of performance in reading tests (Khalifa, 1997; Aulls, 1986:
124-5). This obviously overlaps with the nature of the existing
schemata candidates possess (see Section 2.2 for a full discussion
of this). Weir (1990, 1993) points out that the topic should be
selected from a suitable genre, at an appropriate level of specificity,
and should not be culturally biased or favour any section of the
test population. The issue of what is a generally accessible text
remains with us. In those situations where we are writing tests for
heterogeneous groups of students, we are by necessity forced to
select texts with a wider appeal than is the case when we have a
more homogeneous group. Clapham (1996a/b) suggests that in
her research it is only with more specific texts that background
knowledge has a significant effect on text comprehension. We
might also need to consider the effect of interestingness. The
work of Spilich et al. (1979) suggests that this is an important
influence on the interaction between readers and text.
The content of a text should be sufficiently familiar to candid­
ates so that candidates of a requisite level of ability have sufficient
existing schemata to enable them to deploy appropriate skills and
strategies to understand the text.
Royer and Cunningham (1978) suggest that texts selected should
be within the knowledge base of the candidates; they introduce
the concept of ‘tailoring’ texts for specific audiences. Such match­
ing may not be available in tests with large heterogeneous popula­
tions and a cruder more general strategy may be unavoidable.
As part of the a priori validation process the familiarity of the
text can be established through survey and we would want to
avoid texts at the extremes of a familiarity continuum (Khalifa,
1997). In general, a text should not be so unfamiliar that it can­
not be mapped onto a reader’s existing schemata. Conversely, the
content should not be so familiar that any question set can be
144 Reading in a Second Language
answered without recourse to the text itself (Roller, 1990). This
should be checked rigorously, whichever of the formats for testing
reading are employed. A key pretesting check is to determine if
any of the questions are answerable without recourse to the text,
and any such questions should be removed. The reader is referred
to the discussion of schemata in Section 2.2.
Texts currently employed in testing reading tend to be genuine
and undoctored and, as far as possible, are selected with appropriacy
for the target situation needs of the test takers in mind (West,
1991). However, it is not always easy to determine which texts are
most appropriate for which test takers; a postgraduate in business
studies may well have come from a science or engineering academic
background; an undergraduate may be studying a variety of sub­
jects. The practicalities of constructing multiple forms for an EAP
proficiency test for different subject areas are intimidating, and
though there is some evidence that performance is enhanced by
background knowledge in the content area of a reading compre­
hension passage (Alderson and Urquhart, 1985) the evidence is
not conclusive (see Clapham, 1994;Ja’far, 1992; Koh, 1985). How­
ever, it does appear that the more specific a text the more import­
ant the contribution of background knowledge to comprehension
(Clapham, 1994: 281-2), the less specific a text the more important
the contribution of language proficiency. This would encourage
us to select texts with a preponderance of semi-technical as against
technical vocabulary. The development of computer programs for
concordancing of texts is already in use in the People’s Republic
of China for checking this facet of texts in test design.
As far as a canonical culture is concerned, students sitting an
EAP test, for example, should not, if possible, be faced with texts
which are too far outside their academic culture. If the texts are
selected well, testees should be inside what Swales (1990) has termed
the ‘discourse community’. There remains the problem of the
extent to which the test developers also belong to the appropriate
community.
If general texts are to be selected in Academic Purpose tests, it
appears that the non-science texts may be the most suitable as,
although non-science students seem to be adversely affected by
science texts in tests, the reverse does not appear to be the case.
Most science students appear not to be adversely affected by non­
science texts in tests as they are familiar with these areas in their
own reading (Clapham, 1994: 277).
 Testing reading comprehension (s) 145
Grabe (1991) found that a major implication of research in
this area is that students need to activate prior knowledge of a
topic before they begin to read. If this is absent then they should
be given ‘at least minimal background knowledge from which to
interpret the text’. It is interesting to note that, in our survey of
teaching tasks used in coursebooks, prediction was seen as a useful
pre-reading strategy. In contrast, in the analysis of testing tasks we
carried out, a pre-reading activity was seldom built in.

Vocabulary
Researchers have attempted to differentiate three levels of vocabu­
lary : common core, subtechnical and technical (Inman, 1978;
King, 1989). In tests for heterogeneous populations care should
be taken to avoid technical terms (Robinson, 1991: 28). For higher
level students in particular, we need to examine whether the lex­
ical range is appropriate in terms of common core, technical and
subtechnical vocabulary. In EAP tests, where the focus is on lexis,
there is a preference for testing subtechnical words which Cowan
(1974: 391) defines as: ‘Context independent words which occur
with high frequency across disciplines’ (see also Yang, 1986: King,
1989). Marton (1976: 92) sees subtechnical words as academic
vocabulary, ‘the words have in common a focus on research, analysis
and evaluation - those activities which characterise academic work’.
In general, this seems sensible advice but it is not always easy to
determine the level of a word unequivocally and reliably.

Channel
Particularly in science texts, diagrams are extensively used to con­
vey information (Ferguson, 1977; Shepherd, 1978). The presence
of diagrams affects the way text is written and processed (Hegarty
et al., 1991; Koran and Koran, 1980). Test developers need to
decide on the nature and amount of non-verbal information that
is desirable, e.g. graphs, charts, diagrams, etc.5

Size
Johnston (1984: 151) notes that currently texts used in reading
comprehension tests tend to be many and brief. The length of
146 Reading in a Second Language
text(s) that candidates are exposed to will influence the strategies
and skills that the candidate may be asked to deploy. If texts
are too short it may not be possible to test expeditious reading
strategies (search reading, skimming and scanning), only careful
reading.

Difficulty

Here the concern is with establishing the difficulty of the text/

tasks students are expected to handle. The difficulty of the text
will be largely determined by its linguistic, organisational, pro-
positional and discoursal attributes (see above) and is obviously
relative to individual variation across a number of parameters (e.g.
background knowledge, purpose(s) for reading, etc.).
The literature abounds with warnings against reliance on read­
ability formulas for estimating text difficulty (Weaver and Kintsch,
1991). But as these authors note, and accepting the complexity of
the conditions noted above, a simple formulaic approach which
takes account of these other conditions is not available and is not
likely to be. Miller and Kintsch (1980: 348) present some direc­
tions in which we need to go:
. . . the best index of readability is a measure that takes both factors
into account, that is, reading time per unit recalled . . . the read­
ability of a text is determined by the ways that certain text properties
- primarily the arrangement of the propositions in the text base,
but also word frequency and sentence length - interact with the
reader’s processing strategies and resources.
but even this analysis did not take account of inferencing ability
or macroprocessing.
The amount of work in conducting such analyses is usually
beyond the time and resources available in most teaching/testing
situations. At best we need to be aware of these factors and a panel
of experts familiar with the use of texts for the specified context
are the most likely to be able to make sensible judgements about
the inclusion or non-inclusion of texts for testing purposes.
The levels of text that candidates will have to process in the
target situation are perhaps still the best guide to the types of text
that might be included. A priori needs analyses may help here.
This should be supported by other validation procedures such as
 Testing reading comprehension (s) 147
eliciting reactions from future candidates, people preparing can­
didates and future end users of test results to establish, among
other things, appropriate levels of difficulty of texts. Farr et al.’s
suggestions (1986) for collecting data on functional literacies is in
line with this.
Appendix 1 also contains a summary checklist of the conditions
appropriate to the assessment of reading ability that a test writer
might want to take into consideration. The list is not exhaustive
(see Bachman, 1990, for additional categories) and conversely not
all the components will necessarily be relevant in all situations.
It is only when decisions have been taken in relation to selection
of texts and tasks for a test that the issue of format can be settled.
In the next section we briefly review the use of a number of reading
test formats from a historical perspective, consider a number of
important facets in test design and then examine in detail the
advantages and disadvantages of the most commonly employed
formats used for testing reading comprehension.

3.4 TEST FORMATS

Origins
Pugh (1978) suggests that silent reading replaced a long tradition
of reading aloud only in the nineteenth century, possibly as the
conditions affecting reading rooms in libraries supplanted those
of the medieval monastery. However, a more probable explana­
tion is that the growth of literacy in that century reduced the
numbers of those needing to listen to texts read aloud. Venezky
(1984: 21) dates a change in instructional emphasis from oral to
silent reading in the USA at around the end of the first decade of
the twentieth century as a result of an increased emphasis on
meaning in reading instruction accompanying a social need for
literacy, matched by an increased availability of cheap paper and
high-volume printing techniques. The chief complaints against oral
reading being (a) it did not deal with understanding and (b) it
was not the most important kind of reading (Johnston 1984: 152).
Venezky (1984: 3) details how Romanes reported the first study
of reading comprehension in 1884 (ibid.: 13) in which rapid readers
recalled the most. However, he notes that in the seventy-five years
which followed, ‘the investigation of cognition in reading compre­
hension was never pursued as systematically as word recognition
148 Reading in a Second Language
was . . . the present day importance given to comprehension in
reading research is a phenomenon of the last two or three decades’.
Historically, then, the focus has been essentially on vocabulary
rather than comprehension in general in both the teaching and
testing of reading (op. cit.: 14).
Venezky (1984: 16) describes the work of Thorndike (1917) on
comprehension processes based on data gathered in field testing
a reading comprehension test. He also provides fascinating details
of the use of reading tests in school surveys (evaluations) in the USA
in 1915-16 (p. 18), the use of survey data to improve teaching
methodology (Neal and Foster, 1926) and a model for planned
instructional improvement and its evaluation in Chicago schools.
Venezky traces the development of standardised reading tests to
the early part of this century and details how they became ‘an
indispensable component of the schooling ritual, a position they
still hold today’ (p. 19).
However, Johnston (1984: 149 et seq.) sounds a more cautionary
note in pointing out that support for testing reading came largely
from administrators because of its accountability and gatekeeping
function, and met considerable opposition from teachers. Thus
from the beginning of this century the standardised, group silent
reading product-oriented model for testing was institutionalised.
An individualised, descriptive process-oriented model has not
emerged (ibid.: 168 et seq.) and a concern for summative as against
formative assessment has maintained its dominance (see Calfee
and Hiebert, 1991: 291-301, for the argument for more teacher
internal assessment to provide the data which will guide instruc­
tional decisions).
It should be noted that such summative tests tend to be unidi­
mensional as reliability is a major criterion for inclusion (Calfee
and Hiebert, 1991: 286) and the final versions of tests are very
consistent internally and, consequendy, provide little information
about student strengths and weaknesses. Expeditious reading
subtests seem to have lower internal consistency figures associated
with them, whereas items focusing on the microlinguistic have much
higher. One might speculate that this may in part explain why
attempts to develop models for expeditious reading, or to include
such reading types in research, tests or teaching, has suffered
relative to a traditional focus on careful reading, particularly at
the local level. Pugh (1978: 20) notes that from about 1910 there
was an increased interest in comprehension exercises on texts
 Testing reading comprehension(s) 149
requiring close careful reading. He contrasts this with an American
interest in the speed of silent reading, which is noticeable by its
absence in the British literature in the first half of this century.
Slow silent reading features in many of the second language
tests produced in the earlier part of the twentieth century (Spolsky,
1995: 41; Venezky, 1984: 13), although from a current perspective
it is easy to be dismissive about these early efforts (West, 1991).
Generally they did not select texts with the target situation of
candidates in mind - although in 1934James Conant is quoted by
Spolsky (1995) as arguing for testing reading knowledge in the
candidate’s own field. Nor did they meet the demand for ‘authen­
ticity’ in text and task selection; they do not appear to have been
based on any theoretical view of underlying reading process; test
items do not appear to have been designed in any principled fash­
ion; and no need was felt to demonstrate their statistical properties,
such as their reliability.
Farr et al. (1986: 62) note:
. . . the passage was seen simply as a stretch of prose providing
language for comprehension . . . the concept that readers read in
different ways according to their purpose and the type of text was
as yet unrecognised.
Spolsky (1995) describes how such traditional descriptive-
humanitarian tests, usually short-answer questions (SAQ) or trans­
lation, were largely supplanted by rational-empiricist objective
measures, most commonly multiple-choice questions (MCQ), on
the basis of the superior reliability and psychometric qualities of
the latter. Tests produced within this new ‘m odern’ paradigm did
little to embrace the criterion of validity, however.
We would now seem to be entering a ‘post-modern age’ as con­
cern mounts that, though objective reading tests may be reliable,
they may not be delivering valid information on the abilities we
seek to measure. We describe below how the wheel now appears
to have turned full circle. We appear to be moving from the
‘rationalist-empiricist’ measurement era (Spolsky, 1995: 1) which
attempted to make the testing of reading more objective and more
reliable through the use of multiple-choice questions (MCQ) and
cloze procedure (Weir, 1990). The short-answer (SAQ) format
employed in the first reading tests is now regaining currency as it
attempts to rectify its earlier deficiencies using techniques such as
mindmapping (Sarig, 1989) and introspection (Faerch and Kasper,
150 Reading in a Second Language
1987; Storey, 1995) to establish what is to be tested and to ensure
that what is being tested is what was intended. Statistical validation
through internal consistency estimates, correlational data and fac­
tor analysis has also proved useful in the validation of such tests.
The humanistic-descriptive approach they embody, with its concern
for what we are measuring, is now supported and enhanced by an
additional concern for the accuracy and replicability of measure­
m ent (see Spolsky, 1995: 5 and Section 5.3 below).
In the TESOL field much of the discussion about the testing of
reading until the early 1980s focused on how we were going to test
it. The concern was largely with the formats we might employ to
do so and with the reliability of the measurement involved. The
format employed as the vehicle for assessing reading ability will
certainly constrain the operations and performance conditions
we attempt to include. We examine below some of the principal
vehicles for testing various types of reading in the latter half of
this century.
In the current consensus it is felt that a valid direct reading test
should reflect as closely as possible the interaction that takes place
between a reader and a text in the equivalent real-life reading
activity, where background knowledge, formal knowledge, and
various types of language knowledge may all interact with informa­
tion in the text to contribute to text comprehension in line with
the intended purpose(s) and type of reading.
Developers of reading tests need to make sure they do not
artificially constrain the processing options available to the test
taker either through inappropriate performance conditions aris­
ing from texts employed in tests (e.g. topic, text length, discoursal
structure, etc., discussed in Section 3.3 above) or through restrict­
ive, insufficient or ill-conceived sampling of the operations (the
activities/reading types discussed in Section 3.2 above) the reader
is asked to perform on the text(s) selected (Weir, 1993). Lastly,
when all of these are sorted, only then can we decide which test
formats are most consonant with our specification. For this reason
we have left discussion of test facets and formats until last.

Some important facets of reading test design

Speed
The literature abounds with investigations of the question of read­
ing speed (Fry, 1963; Taylor, 1965; Carver, 1985; Haynes and Carr,
 Testing reading comprehension (s) 151
1990). Most make suggestions that good readers read at about
300-350 words per minute but there is wide variation in these
studies down to 140 and up to 800 (see also Nuttall, 1996: 56).
However, these estimates are really meaningless without consid­
eration of the purpose for reading and other considerations such
as text difficulty, etc.
If too much time is given, candidates may simply read a passage
slowly and carefully, and questions designed to test ability to process
text expeditiously to elicit specified information may no longer
activate such operations. Establishing time available for read­
ing for various purposes is difficult to establish and is best done by
empirical trialling on an appropriate test population with clearly
established criteria for successful completion of tasks.

Time control
If the test does not control how much time students spend on
certain items/sections this may change the operations that are
needed to answer them. Too much time spent on a search read­
ing question may change it into one that only requires a slow
careful reading. A similar problem might arise in careful reading
tests where too much time on earlier items means that subsequent
questions which demand careful scrutiny have to be answered
more hurriedly, and a candidate is forced into constructing invalid
test-taking strategies to come up with an answer in the restricted
time left.
Careful thought needs to be given to grouping questions into
sections (and most probably using different texts for different
skills and strategies); empirically determining time necessary to
deploy the required skills/strategies; and carefully structuring the
test through rubrics and invigilation so that timings are adhered
to. If more than one passage is used within a section, then time
controls need to be applied here as well or there is a tendency to
spend more time on the earlier passage especially in expeditious
reading. Thus earlier passages might become tests of careful read­
ing and only the final passage is processed quickly in the absence
of strict time controls.

User friendliness
How much help is given? A number of factors need to be taken
into account, such as the clarity of the rubrics and whether the
152 Reading in a Second Language
rubrics are in the First Language (LI) or the Target Language
(TL). Shohamy (1984) goes further and suggests that questions
set in LI are easier than in TL and the latter may not give as
accurate a measure of comprehension. In monolingual contexts it
seems logical that candidates might be permitted to write their
answers in their m other tongue as well as having the rubrics and
questions in LI. Lee (1986) suggests that recall of a text was
significantly better when done in LI rather than TL.

Format familiarity
Weir (1993) advises that every attempt should be made to ensure
that candidates are familiar with the task type and other environ­
ment features before sitting a test. Sample tests or examples in
test manuals should be available for national examinations, and in
the school context similar formats should have been practised in
class beforehand. Where such help is not available in the pre-test
situation, thought might be given to providing examples at the
start of the test paper if item types are not familiar to candidates.
Anderson and Armbruster (1984: 659) gave support to this when
they argued that ‘performance on the criterion task is a function
of knowledge of the task’.

Issues in question design

Fillmore and Kay (1983) provide a useful set of guidelines for
setting questions (see also Weir, 1993):
■Questions should not contain harder vocabulary than the text.
■Questions should have only one unequivocal answer.
■If the candidate understands the text they should be able to
answer the question.
■Rejection of alternatives on grammatical grounds should not be
allowed.
■Skills not related to reading, e.g. mathematics, should not be
tested.
■Incidental insignificant information should not be tested.
■Questions that require stylistic or other ambiguous judgements
should be avoided.
Further consideration might also be given to the following
questions:
 Testing reading comprehension (s) 153
■How are the questions to be ordered?
■Does the ordering help bring the process of taking the test
closer to the way readers would normally process that particular
text, or at least satisfy a critical audience that it promotes one
reasonable way of doing so?
■Should skimming and search reading questions occur first and
be separated from those questions requiring closer reading of
the text?
It would seem to be consistent with the literature on processing
that candidates may find it helpful to gain an overview of the
passage first. Bransford et al. (1984) showed how performance is
affected by activation of background knowledge. Graves and Cook
(1980) and Graves et al. (1983) found that previewing text helped
increase comprehension of explicit and implicit information. Paris
et al. (1991) provide further evidence of empirical support in
their review of the literature on strategies.
Some writers on the teaching of reading, such as Nuttall (1996)
and Grellet (1981: 6), advocate that activities which lead to an
overall global view of the text should precede more local detailed
understanding. This would suggest that the first activity in tests
might be a skimming task that helps them develop a macrostructure
for the passage. In teaching materials a prediction activity normally
precedes the reading exercises but it is rarely present in test tasks.
In test tasks students should be advised in advance of the pur­
pose of the item(s). This should contribute to the authenticity of
the test task by helping students to adopt appropriate strategies/
skills in task completion. It is important that students should see
the questions to which they are to find answers before they actu­
ally read the text (skimming may be an exception to this as the
search for overall meaning is not usually predeterm ined in any
way and the elaboration of a question would affect this). In this
way they will be able to approach the text in the desired manner.
Therefore, if the questions only demand specific pieces of infor­
mation, the answers can be sought by quickly scanning the text
for these specifics rather than reading it through very carefully
line by line.
We need to consider whether to set questions testing different
skills and strategies on the same passage. It may be beneficial to
separate those items focusing on expeditious reading from those
catering for more intensive reading ability by employing different
154 Reading in a Second Language
texts for each. Within the expeditious reading section we might
also wish to separate skimming (Pugh, 1978: 54) for the discourse
topic, search reading for main ideas (see Pugh, 1978: 53; Guthrie
and Mosenthal, 1987) and scanning for specific information (Pugh,
1978: 53; Rosenshine, 1980).
The order of questions should be the order in which informa­
tion occurs in the text for all test items with the possible exception
of scanning (Hughes, 1989: 130). This is consistent with work on
text processing (Kintsch and van Dijk, 1978; Just and Carpenter,
1980, 1987) which indicates that readers construct referential rep­
resentation of a text incrementally. The sequential ordering of
questions (for search and careful reading) will help candidates
mirror this cumulative process.

Recent methods for testing reading ability

It is necessary to be sure that the methods used do not unduly
constrain the range of skills and strategies we might want to test.
It is likely that different test formats will permit the measurement
of different aspects of reading ability (Reder and Anderson, 1980;
Kintsch and Yarborough, 1982; Kobayashi, 1995). The limitations
in the cloze procedure and multiple-choice questions (MCQ) need
to be examined in this light.
Formats should not in themselves adversely affect performance
- e.g. do multiple-choice tests test reading and/or the ability to do
multiple-choice tests? The measurement of the reading trait should
not be contaminated by the method employed.
Another form of muddied measurement might occur if lan­
guage abilities additional to those of reading per se are involved. If
short-answer questions are used, to what extent is the ability to
write part of the measurement?
Formats should be familiar to the candidates and, if they are
not, a practice test should be given to familiarise them. If it is an
achievement test, it should reflect the types of text and associated
activities that have been practised in class. If it is a proficiency test
it should mirror future activities and text types.
The main proviso for testing within a communicative frame­
work is that the test tasks should, as far as possible, reflect realistic
discourse processing and cover the range of contributory skills
and strategies, purposes for reading, that have been identified as
appropriate for the target population.
 Testing reading comprehension (s) 155
Selective deletion gap filling
Gap filling is indirect in that it seems to measure only a limited
part of our construct of reading proficiency, namely microlinguistic
contributory skills (see Table 3.1 above) and normally would not
seem to provide any evidence on a candidate’s ability to extract
information expeditiously by search reading or skimming a text
or to read it carefully to understand its main ideas (reading types
A and C in Table 3.1 above). To the extent that it does not match
the range of operations and the conditions under which reading
tasks might normally be performed by the intended target group,
then the task is indirect. Weir (1993) points out that the more
indirect the tasks, the more difficult it is to generalise from scores
on this test to statements about students’ reading ability for par­
ticular purposes under specified conditions. How much would the
student have to score to be deemed to have met the pass grade in
reading and be deemed a competent reader?
Weir (1993) suggests that it is probable that this technique
largely restricts one to sampling a much more limited range of
enabling skills and strategies (i.e. a limited set of those abilities
which collectively represent the overall ability of reading) than do
short-answer questions on a text. Whereas short-answer questions
allow the sampling of a range of reading skills or strategies, gap
filling is much more restrictive where only single words are deleted.
He notes that there is even some difference of opinion on what
is being tested where only single lexical items are deleted. Is it
testing the ability to recognise which form of the word is required
an d/or lexical knowledge? On its own it is an insufficient indic­
ator of a candidate’s reading ability. If the purpose of a test is to
sample the range of enabling skills and expeditious strategies, then
an additional format to gap filling is essential.
The format is not likely to have a positive washback effect on
learning as it is not in itself a direct measure of the reading con­
struct, and it is difficult to see how performing this test or prac­
tising for it relates to a normal reading process or provides
information that has a broad diagnostic potential. Tests of this
type may be more happily described as tests of general proficiency
rather than tests of reading and, as such, may have a role to play
in placement testing where an accurate description of ability is
less important than quickly putting students into roughly similar
levels of ‘general ability’.
156 Reading in a Second Language
However, if the focus of attention in the reading class is at the
microlinguistic level, selective deletion enables the test construc­
tor to determine where deletions are to be made and to focus on
those items which have been selected a priori as being important
to a particular target audience. For example, in EAP contexts
computerised concordancing of a wide range of academic texts
can identify common and frequent academic semi-technical lexis
that occurs in these types of text.

Cloze
Weir (1990, 1993) describes how the term ‘cloze’ was first popu­
larised by Taylor (1953) who took it from the gestalt concept of
‘closure’, which refers to the tendency of individuals to complete
a pattern once they have grasped its overall significance (Alderson,
1978). Johnston (1984: 151) details how it had its earlier roots in
the completion task used by Ebbinghaus in 1897 in his efforts to
find a measure of mental fatigue, for which it proved unsatisfact­
ory though its value as a measure of intellectual ability was noted.
In cloze the reader comprehends the mutilated sentence as a
whole and completes the pattern. Words are deleted from a text
after allowing a few sentences of introduction. The deletion rate is
mechanically set, usually between every 5th and 11th word. Can­
didates have to fill each gap by supplying the word they think has
been deleted.
Alderson (1978: 39) described how
The general consensus of studies into and with cloze procedure for
the last twenty years has been that it is a reliable and valid measure
of readability and reading comprehension, for native speakers of
English ... As a measure of the comprehension of text, cloze has
been shown to correlate well with other types of test on the same
text and also with standardised tests of reading comprehension.
Cloze tests are easy to construct and easily scored if the exact
word-scoring procedure is adopted. With a 5th-word deletion rate
a large number of items can be set on a relatively short text and
these can exhibit a high degree of internal consistency, in terms
of Kuder-Richardson coefficients. This consistency may vary con­
siderably, dependent on the text selected, the starting point for
deletions and the deletion rate employed (Alderson, 1978).
 Testing reading comprehension (s) 157
Some doubts have been expressed, especially concerning its
validity as a device for testing global comprehension of a text.
One of its main flaws is that it seems to produce more successful
tests of syntax and lexis at sentence level, comprehension of the
immediate local environment, than of reading comprehension in
general or of inferential or deductive abilities (Alderson, 1978;
Chihara et al., 1977; Kintsch and Yarborough, 1982; Kobayashi,
1995; Markham, 1985). Alderson (1978: 99) found that
. . . cloze is essentially sentence bound . . . Clearly the fact that cloze
procedure deletes words rather than phrases or clauses must limit
its ability to test comprehension of more than the immediate envir­
onment, since individual words do not usually carry textual cohesion
and discourse coherence (with the obvious exception of cohesive
devices like anaphora, lexical repetition and logical connectors).
The process underlying successful completion appears to be largely
bottom-up with an emphasis on careful passive decoding at the
word or immediate constituent level. The focus appears to be on
local comprehension at the microlinguistic level rather than global
comprehension of ideas encoded by the writer across the text as a
whole. Bernhardt (1991b: 198) comments:
... it focuses a reader’s attention on individual words to the detri­
ment of a global understanding of the tex t... It clearly has little if
anything to do with a reader’s understanding of a piece of connected
discourse.
In reading for academic study purposes it is difficult to see how
it can test the ability to read through a text expeditiously or care­
fully to extract main ideas and important detail. Cloze appears to
have little to do with a reader’s understanding of a piece of con­
nected discourse (Markham, 1985), measures information only
within clause boundaries (Kamil et al., 1986; Shanahan et al., 1982)
and focuses attention on individual words to the detriment of
global understanding of a text. That such decoding is seen as the
hallmark of a poor reader (automaticity; rapid context free word
and phrase recognition being the hallmark of the fluent reader
according to Carrell et al., 1988: 94-5 and Stanovich, 1981: 262)
may lead us to question its place in either teaching or testing.
Bernhardt (1991b: 197) argues that as far as the construct
validity of cloze as a test of reading is concerned, ‘cloze testing is
profoundly inadequate’.
158 Reading in a Second Language
Multiple-choice questions
In reaction to the earlier ‘pre-scientific’ SAQ tests, an interest
developed in using the more objective MCQ format which still
appears in major international second language tests, e.g. TOEFL
and UCLES examinations, to this day.
Weir (1993) details how multiple-choice tests exhibit almost
complete marker reliability as well as being rapid and often more
cost effective to mark than other forms of written test. The marking
process is totally objective because the marker is not permitted to
exercise judgem ent when marking the candidate’s answer; agree­
ment has already been reached as to the correct answer for each
item. The format allows scripts to be machine marked in large-
scale examinations, as in China, where over 2 million take the
College English Test (CET). Selecting and setting items are, how­
ever, subjective processes (Meyer, 1985) and the decision about
which is the correct answer can be a matter of subjective judge­
m ent on the part of the item writer or moderating committee.
However, even for experienced examiners it is extremely difficult
and time consuming to develop a sufficient number of decent items
on a passage. Items need to be validated through trialling before
we can be confident of their statistical properties, e.g. facility and
discrimination. The development of items for the CET in China
go through a number of rigorous trialling phases, and even so it
takes the national moderating committee ten days to finalise the
papers each year. Given how difficult it is to write such items,
there must be a serious question mark against teachers using this
format to test reading for practical reasons alone.
In more open-ended formats for testing reading comprehension,
e.g. short-answer questions, the candidate has to deploy the skill
of writing. The extent to which this affects accurate measurement
of the trait being assessed has not been established. Multiple-choice
tests avoid this particular difficulty.
With the growth of interest in overall text comprehension as
against decoding in the 1970s, an interest in top-down processing
as against bottom-up decoding, testers also became more aware of
assessing comprehension of text at the global level. A comparison
of the ELBA test with the ELTS test makes this distinction clear. It
is, however, extremely time consuming and demanding to get the
requisite num ber of satisfactory items for a passage, especially for
testing strategies such as skimming in the MCQ format. A particular
 Testing reading comprehension^) 159
problem appears to lie in devising suitable distracters for items
testing the more extensive receptive strategies. West (1991: 63)
comments:
. . . while multiple-choice reading items are well able to test isolated
details or ‘fragmentary’ comprehension, they are not very suitable
for more global tests of reading. By ‘global’ reading is meant some
broader response to the text - either comprehension across the
text as a whole (or at least a considerable portion of it) or an
understanding of the text as a text: an appreciation of the charac­
teristics of the text type, the intended audience, the writer’s inten­
tion, the overall message, or the structure of the text.
Although MCQ items could (albeit with some difficulty) be
written in these areas, such items would seem to inhibit the use of
top-down strategies (skimming, predicting), not least because it
is likely to encourage test takers to try to match the stem and
options with words in the text. Heaton (1988) had noted earlier
that, for global comprehension activities, it is more helpful to set
simple open-ended questions rather than multiple-choice items;
otherwise students will find it necessary to keep in mind four or
five options for each item while they are trying to process the text.
There must be some doubt about the validity of MCQ tests as
measures of reading ability. Answering multiple-choice items is an
unreal task, as in real life one is rarely presented with four altern­
atives from which to make a choice to signal understanding. In a
multiple-choice test the distracters present choices that otherwise
might not have been thought of. In MCQ tests we do not know
whether a candidate’s failure is due to lack of comprehension of
the text or lack of comprehension of the question. A candidate
might get an item right by eliminating wrong answers, a different
skill from being able to choose the right answer in the first place.
Nevo (1989) details how test-taking strategies can lead to right
answers for some candidates and reading strategies to incorrect
ones on an MCQ test.
Bernhardt (1991b: 198) raises the issue of passage independ­
ence in such tests and cites evidence of candidates being able to
determine answers without recourse to the passage (see also
Pyrczak, 1975; Jarvis and Jensen, 1982; Barnett, 1986). Evidence of
this being a problem is also presented in Katz et al. (1990) .
There is also some concern that students’ scores on multiple-
choice tests can be improved by training in test-taking techniques
and that such improvement reflects an enhanced ability to do
160 Reading in a Second Language
multiple-choice tests rather than any increase in language ability.
This is a matter which is in need of serious investigation.
Carrell et al. (1989) found that the effectiveness of training in
reading strategies varied according to the test format employed.
Metacognitive training led to improvement in the sample whose
ability was measured by open-ended questions but not for those
on MCQ tests.
Weir (1993) draws attention to the danger of the format having
an undue effect on measurement of the trait. There is some evid­
ence that multiple-choice format is particularly problematic in
this respect. This has been evidenced by low correlations both with
alternative reading measures and with other concurrent external
validity data on candidates’ reading abilities (see Weir 1983a).
The scores obtained by candidates might have been affected by
the method used. This is not a problem with direct measures of
language ability.
Reading tests in this approach were more concerned with the
psychometric properties of the test than with the nature of the
construct being measured. Thus, in earlier versions of the TOEFL
reading test one could find a section of decontextualised vocabulary
items being used as indicators of reading ability; this is very much
a bottom-up approach to reading and a very limited part of it at
that. As Spolsky (1995: 4) points out: ‘what can be measured reli­
ably is not necessarily the same as the ability one is interested in’.
A more recent variant of this technique termed ‘multiple match­
ing’ (where the answers to all questions plus a number of distracters
are all provided in the same list for the candidate to select from)
appears in a num ber of recent ELT exams produced by UCLES.
Its advantages are enumerated by West (1991). However, its pro­
ponents still do not explain how the underlying processes that
help select the right answer from the many available equates with
normal processing for the reader. It is nevertheless an improve­
m ent on traditional MCQ and in those situations where tests need
to be machine scored because of huge populations its potential
should be investigated.

Information transfer tasks

Direct language tests which aim to measure ability to perform
authentic tasks (e.g. extracting main ideas quickly) under real-life
performance conditions (e.g. reading undoctored texts from the
 Testing reading comprehension(s) 161
real world) provide the tester with information which can more
effectively be translated into statements concerning the candidate’s
ability to cope with reading in the target language situation.
All of the conditions identified in Appendix 1 can be taken
account of in this format and in contrast to the indirect formats
discussed above, the normal purposes for which people read can
be more easily accommodated.
In testing reading comprehension there is a potential problem
of the measurement being ‘m uddied’ by having to employ writing
to record answers. Weir (1993) provides examples of how, in an
attempt to avoid this contamination of scores, several examina­
tion boards in Britain have included tasks that require informa­
tion from written texts to be translated into a non-verbal form,
e.g. by labelling a diagram, completing a chart or numbering a
sequence of events.
However, a good deal of care needs to be taken that the non­
verbal task the students have to complete does not itself complic­
ate the process or detract from the authenticity of the experience.
In some of the more sophisticated tasks using this format, there is
sometimes a danger that students may be able to understand the
text but not be totally clear what is expected of them in the transfer
phase. There is also a danger that in order to fit more neatly into
such formats, texts are sometimes expressly written for this purpose.
Weir (1993) details how the questions set in this format can
cover the im portant information in a text (overall gist, main ideas
and important details) and understanding of structures and lexis
that convey this. The nature of the format and the limitations
imposed by capacity at lower levels of ability will constrain the types
of things students might be expected to read for in the early stages
of language learning. In the first year of learning a language,
reading will normally be limited to comparison and contrast of the
personal features of individuals and extracting specific informa­
tion from short non-personal texts. At higher levels more complex
texts will be used and most of the skills and strategies in Table 3.1
above - involving search reading and reading carefully to under­
stand the main ideas and important detail - can be assessed.
Weir (1993) describes information transfer tasks as a useful
variant of short-answer questions which require candidates to write
down answers in spaces provided on the question paper. These
answers are normally limited in length either by using short lines to
indicate the num ber of words, restricting the space made available
162 Reading in a Second Language
to candidates through boxes, or by controlling the amount that
can be written by deleting words in an answer that is provided for
the candidate. All of these techniques help keep the answers brief
and reduce writing to a minimum in an effort to avoid possible
contamination from students having to write answers out in full.
Short-answer questions
This was one of the earliest forms of testing and predates MCQ
and Cloze. In its earlier incarnation it suffered from the fact that
the texts were often short and in many cases artificially constructed
(see West, 1991). Texts now tend to be genuine and undoctored
(see Weir, 1990, for examples) and as far as possible are selected
with appropriacy for the target situation needs of the test takers
in mind.
Developments in test validation techniques have also contributed
to making better use of the SAQ format. Through meaning con­
sensus techniques (Sarig, 1989) a more principled basis for arriving
at test items is now available. If the focus for such questions is
determined in a principled fashion, for example, by textmapping
to form a consensus framework of the main ideas and important
details a reader might be expected to extract from a text (Sarig,
1989), idiosyncrasy in content selection can largely be avoided
(see Section 5.3 for details of this procedure). The immediate
recall protocols advocated by Bernhardt (1991b: 200-10) for testing
individuals might alternatively be used for helping to decide what
a candidate should be able to take away from a text.
Expert judgem ent and introspection (Storey, 1995) by appro­
priate readers enhance the probability that the required operations
are being tested, and introspection in particular can illuminate
the extent to which the behaviour that test items produce equates
with the behaviour identified in the theory-based model as being
representative of the real reading which occurs outside the test
itself (see Section 5.3 for details of how these procedures were
used in the development of a recent EAP reading test). Storey
(1995) shows how introspective validation can help identify the
proportions of construct-relevant and construct-irrelevant variance
generated by test items.
Dubin et al. (1986: 163) point out:
. . . interactive models (of the reading process) suggest a need to
test skills at many levels, since all these skills are assumed to play a
 Testing reading comprehension(s) 163
significant role in the reading process. They include everything
from rapid identification of vocabulary and syntactic structures, to
the interpretation of larger discourse patterns, the making of infer­
ences, etc. In addition, since interactive models are defined by the
interaction of various skills and strategies, ideal assessment strat­
egies would involve combining skills.
Weir (1993) describes how, with careful formulation of the ques­
tions, a candidate’s response can be brief and thus a large number
of questions may be set in this format, enabling a wide coverage.
The SAQ format lends itself to testing all the types of reading
identified earlier in Chapter 2 and discussed above in Section 3.2
(search reading, skimming for gist, scanning for specific informa­
tion and reading carefully to extract the main ideas and important
details from a text). Activities such as understanding implicidy stated
ideas, recognition of a sequence, comparison and establishing the
main idea of a text, often require sentences in a text to be related
with other elements which may be some distance away in the text.
This can be done effectively through short-answer questions where
the answer has to be sought rather than being one of those pro­
vided as in multiple-choice. Consequentiy, correct answers are more
likely to result from comprehension of the text rather than from
test-taking strategies such as guessing, matching, etc.
The main disadvantage to this technique is that it involves the
candidate in writing, and where candidates are required to use
their own words rather than language supplied in the text, there
is some concern that a significant amount of difficulty may be
added (Pollitt and Hutchinson, 1986; Bensoussan and Kreindler,
1990: 59), and there is accordingly interference with the measure­
ment of the intended construct. Care is also needed in the setting
of items to limit the range of possible acceptable responses.
Weir (1993) claims that if the num ber of acceptable answers to
a question is limited, it is possible to give fairly precise instructions in
the mark scheme as to the range of semantically acceptable answers.
In those cases where there is more debate over the acceptability of
an answer, in questions requiring inferencing for example, there
is a possibility that the variability of answers might lead to marker
unreliability. However, careful moderation and standardisation of
examiners should help to reduce this. Mechanical accuracy criteria
(grammar, spelling, punctuation) should never feature in the scor­
ing system as this affects the accuracy of the measurement of the
reading construct.
164 Reading in a Second Language
The attraction of SAQ over all other formats is that texts can be
selected to match performance conditions appropriate to any level
of student, and the format allows the testing of all the operations
that might be required in a test of reading.

Recall measures
The measures we have discussed so far for assessing comprehension
have been the choice of educators: write questions about informa­
tion in a passage and then evaluate readers’ responses to them.
Meyer and Rice (1984: 320) describe how psychologists have tended
to get readers to write down all they can remember from texts.
They point to difficulties in marking such protocols, not least the
difficulties of establishing marking frames, especially where infer­
ences are made.
Kobayashi (1995: 111), in reviewing recall protocols, comments:
Recall protocols can be classified as either oral or written in terms
of the language mode, or either immediate or delayed in terms of
time of recall, or either free or probed, i.e. with or without cues for
recalls. First a text is analysed in terms of idea units (or propositions)
and this analysis becomes a template for scoring recalls. The number
of propositions recalled after listening or reading will be counted
as scores.
She notes that the method has not yet gained much ground in
testing but is increasingly common in second language research
studies (see also Lee, 1986, and Lund, 1991). She points to the
difficulty in establishing propositions and a hierarchy of relative
importance within these. The difference between reproduction
and comprehension is also noted. More mature readers who integ­
rate ideas and synthesise may be penalised because their protocols
lack details and they may have used their own words.
Bernhardt (1991b: 200-10) proposes immediate recall as an
alternative to traditional testing measures, drawing on her experi­
ence in cognitive psychology and LI reading research. However,
at present the analysis of protocols necessary to determine an
estimate of performance would take far too long for this technique
to be feasible for either classroom or large-scale testing, particularly
if the recall protocols are to be written in the native language of
each author.
Developing a marking scheme even on a relatively small passage
can take up to 50 hours (Bernhardt 1991b: 202) and marking
 Testing reading comprehension (s) 165
individual protocols, an hour for relatively short passages. She
does, however, suggest a revised and more efficient scheme, based
on breaking the propositions into pausal units, which is quicker
but there must still be some concern that scoring is based on very
small units of information and often single lexical items; this raises
again the central issue of construct validity.
As well as problems in efficiency there is a serious question
mark against the validity of this procedure for testing reading
comprehension. Kobayashi (1995: 113) draws attention to the fact
that readers may not be able to remember all they have understood.
Comprehension is not necessarily equatable with remembering.

3.5 CONCLUSIONS
In Chapter 2 we looked at the theory relating to componential
and process models of reading. We noted that a comprehensive
model of the processing involved in different types of reading is
not yet available and that, for the present, L2 researchers, teachers
and testers might be better served by focusing on the components
of reading ability. In Section 2.2 an a priori case was made for there
being more than one component in reading and a preference was
stated for a three, as against a two, component model. In Section
2.3 we translated these components into terms of skills and strat­
egies which are more familiar to testers and teachers.
In Chapter 3 we have examined test-driven research for empir­
ical evidence relating to the componentiality issue in terms of what
strategies and skills can and should be assessed. The data emer­
ging from such studies offer some tentative but encouraging support
for the theoretical view of the components favoured in Chapter 2.
The rigorous requirements of validation in language testing
have necessitated a closer examination of the parameters of texts
than in Chapter 2, Section 2.2. The need for explicit specification
in testing means that we also have to establish any performance
conditions that may affect the product of reading comprehension
and perhaps even processing itself. We believe that these text-
based facets must form part of any definition of reading objectives
or any definition of reading proficiency. All skills and strategies
are performed under certain performance conditions and not in
the vacuum or text neutral position purely theoretical work some­
times assumes. We do not just read expeditiously but rather we
166 Reading in a Second Language
skim/search read/scan a certain type of text, of a certain length,
of a certain degree of familiarity, under certain time constraints,
etc. (see Section 3.3 above). Altering the conditions will alter
performance in comprehending a text and possibly the way we
process the given text.
In terms of the test data presented in this chapter, the argument
as to whether reading is multidivisible, consisting of a num ber of
components which can be identified clearly, or whether it is an
indivisible, unitary process, is still not fully resolved. The ubiquitous
call for further research is necessary. If a unitary view is to be
convincingly rejected, future research will need to demonstrate the
consistent presence of at least a second component in repeated
analyses across a range of samples of ESOL candidates. Secondly,
future research will need to investigate whether such components
are identifiable. Finally, it will have to establish the extent to which
each component has a meaningful effect on the measurement of
reading comprehension. How much of the overall variance does
each component explain in a reading test? It will be important to
use more exigent statistical techniques to test whether the presence
of each component is statistically significant.
There is cause for immediate concern that wholesale adherence
to either the unitary or the multidivisible view in language testing
may be problematic. As a matter of urgency we need to develop
tests which are maximally valid tests of the skill components
at levels A, B, C and D in Table 3.1. If, in constructing tests of
expeditious and careful reading strategies, test constructors faith­
fully mirror these in the mapping of texts for testing purposes,
then we might be able to make a stronger case for suggesting
that we are actually testing these skills or strategies. Student intro­
spection at the piloting stage might lend further credence to our
efforts. Statistical analysis of data from a normally distributed test
population, in particular principal components analysis, might add
further weight to the success of operationalising the constructs.
A full account of a systematic and principled methodology for
researching the construct of reading through language test data is
presented in Section 5.3.
Such research may not, of course, run as smoothly as we would
like. It may prove impossible to operationalise the posited four
types of reading separately in a test. It may be that reading is such
a massively parallel interactive process that we will not be able to
distinguish clearly between such components. It may be that, at
 Testing reading comprehension(s) 167
certain levels of ability - for example, weak and strong readers -
reading is indeed unitary; divisibility may be a function of the
level of student being tested. For readers linguistically proficient
in the target language and already competent readers in their LI,
reading in the target language may well be uni-componential,
whereas this may not be the case where either of these conditions
is not met (see Downing and Leong, 1982).
However, Johnston (1984) rightly emphasises the need to view
validity as the interpretation to be made from test results rather
than residing in the test itself. We must not lose sight of the
emerging evidence that there is doubt about the status of items
that focus on specifically linguistic operations at level D as part of
the assessment of a candidate’s general reading ability. As a matter
of urgency, it is necessary to investigate whether testing D-type
reading does in fact give us sufficient information about a candid­
ate’s ability to handle global comprehension tasks A and activities
C. We must address the implications of evidence that there may
be groups of candidates who are capable of type A and C reading
but who are severely challenged by type D test items. There must
be serious concern that test items which focus on the specifically
linguistic/individual word level may not be good predictors of
general reading ability, i.e. they do not give us an accurate picture
of the reading ability of all the individuals who sit a test.
We believe that utilisation-focused tests of reading need to be
based on a clear specification of the target situation needs of
candidates and an attempt will have to be made to identify the
skills/strategies which are needed to carry out their future activities.
A representative sample of those reading types should be incor­
porated into the test in a number of different sections with their
own specific configurations of performance conditions. Kintsch
and Yarborough (1982: 834) argue in a similar vein:
It is clearly false to assume that comprehension is an ability that
can be measured once and for all, if only we had the right test.
Instead, ‘comprehension’ is a common sense term for a whole
bundle of psychological processes, each of which must be evaluated
separately. Only a collection of different tests, each tuned to some
specific aspect of the total process, will provide adequate results.
Crucially, a profile of appropriate abilities would indicate whether
or not the candidate is likely to be able to function effectively in
the target language situation in respect of each of the identified
168 Reading in a Second Language
skills/strategies. Everything we have said in our earlier review of
models in Chapter 2 and in the empirical review of test-based
research in Chapter 3 supports the case for profiling. Spolsky
(1994) succinctly adumbrates the complex and multidimensional
nature of comprehension and stresses the need for full description
in reporting results as against a single grade or score. He argues
(1994: 151):
... we will need to design and use a variety of reading assessment
procedures (not only tests) to allow us to report on a variety of
aspects of the student’s ability to understand, and to establish some
systematic way of reporting the results on all of them. The differ­
ences the student shows across this range of results will inform us at
least as much as will the result of adding them together. However
good our tests are, a single score will always mislead.
Given the distinct possibility that different skills and strategies
can be taught and tested, and an acceptance that it is worth while
investigating these, then some form of profiling of these abilities
is essential rather than collapsing scores into a single score or grade
for reporting purposes.
Lastly, we have argued that, as well as carefully considering
operations and the conditions under which they are performed,
the test developer must pay due attention to selecting appropriate
formats for assessing performance. Kobayashi (1995) has clearly
demonstrated that the formats used for testing reading com­
prehension may well influence performance. Such method effect
should be limited, as far as possible, by the inclusion of a range of
task types which replicate real-life performance (Johnston, 1983)
and which have been shown to be suitable instruments (valid,
reliable and utilisation focused) for measuring posited reading
skills and strategies.

Notes
1. We must be careful not to fall into the trap of thinking that these
two types of variation are separable. Clearly the dimensions intersect:
readers’ careful reading may be expected to produce a different product
from their expeditious reading, but their careful reading may also
produce different interpretations from other readers’ careful reading.
2. It will be recalled that Hoover and Tunmer adopt this course when
describing the components of reading ability.
 Testing reading comprehension (s) 169
3. Testers’ assumptions can so easily be thwarted. One of the authors
once set a JMB test apparently requiring a detailed background know­
ledge of horse-shoes to a group of British adults. One student performed
brilliantly, only to confess that in his youth he had passed a lot of time
in a farrier’s workshop.
4. As pointed out in Chapter 2, Kintsch and van Dijk’s description is best
suited to careful reading.
5. The provision of such a categorisation should not be taken as pre­
empting the question as to whether multicomponent categorisations
are valid.
This page intentionally left blank
 171
4

The teaching of reading

4.1 TEACHING AND TESTING

A few years ago it was a commonplace to assert that a great deal of
testing reading went on but little teaching. With a lot of justifica­
tion, it was held that a typical reading lesson consisted of the
teacher, with little or no prior discussion or any other kind of
preparation, presenting the students with a text, which the stu­
dents then read. After this stage, the teacher asked questions and
the students answered. Feedback was limited to the students being
told that their answers were right or wrong. Variations existed: the
teacher might read the text aloud, while the students followed it
on the page; or the students m ight be asked to read it aloud. The
teacher might use a textbook which contained the questions, and
sometimes the answers.
If such a depiction of the typical reading lesson is valid, and it
is our experience that not only was it virtually ubiquitous some
years ago but it is still prevalent in many parts of the world, then
the accusation that reading was tested but not taught is justified.
The sequence of activities described above amounts to an informal
assessment of students’ reading performance, i.e. a test Teaching
would involve, among other aspects, more structured feedback
as to why responses were acceptable or not, some instruction in
how to arrive at the desired response, in addition to more global
instruction in how to set about the task, alternative strategies which
might prove helpful, etc.
In the past few years, a great deal of work has been done on
these and other aspects of teaching, as opposed to testing of read­
ing. This work will be reviewed below. Before we proceed to it,
172 Reading in a Second Language
however, we shall look more closely at similarities and differences
between testing and teaching.

Similarities
Both testing and teaching involve students being given a written
text or texts, and being required to read it. Usually, they are also
expected to respond overtly to some task requirem ent (though in
classroom ‘silent reading’, this requirement may be dropped).
Apart from such an obvious overlap, the following similarities can
be discerned.
We noted in Chapter 3 that a feature of the British approach to
communicative testing is that the tasks and the conditions under
which they are performed should approximate to real life per­
formance as closely as possible (McNamara, 1996; Weir, 1993).
We feel that the same should apply to the teaching of reading, at
least as far as comprehension activities are concerned.
Consideration of the types of task discussed for testing reading
in Chapter 3 and the performance conditions under which they
are performed, such as
■purpose
■nature of the texts
■length of text
■rhetorical structure
■topic area
■background knowledge
■w riter/reader relationship
■speed of processing
■range of vocabulary
■grammatical complexity
are of equal relevance in the reading classroom in determining
activities and selecting texts.
For example, as far as purpose and texts are concerned, read­
ing genuine texts for authentic purposes is held to be crucial in
motivating learners to read (Nuttall, 1996). In both teaching and
testing, coping with genuine text is likely to be an important
objective at a certain stage for many learners. A key objective for
both the materials writer and the test developer should be to give
the students a realistic purpose for every reading activity (see Moran
 The teaching of reading 173
and Williams, 1993: 68). Personal interest may be difficult to cater
for in a course book or a test, but instrumental purposes are
relatively easy to simulate in information-giving texts (see Paran,
1991 and 1993, for examples of this). Such purposes enable the
activity to move beyond the ubiquitous post-text comprehension
question(s) demanding a ritual show of understanding (ibid.).
A feature of modern textbooks on reading is to try to provide a
clear purpose for reading a text and Paran (1993) provides some
nice examples of how this might be set up through pre-reading
activities such as: an initial questionnaire followed by reading to
compare findings; a quiz and a reading to check how much you
know; prediction of content from title, words, illustrations, etc.,
and checking text to see how right you are; discussing own opin­
ions and comparing with opinions in the text.
The question of text length is another aspect shared by testing
and teaching. It is clear that people learn to read by reading not
just by doing exercises. Learners must therefore read enough in a
programme for it to make a difference (see Mahon, 1986). So just
as longer texts are necessary in tests where the interest is in ex­
peditious reading, so too in learning to read. The tendency to
employ short texts in tests and course books meant to cover the
range of reading skills and strategies is questionable (for example,
note the limited length of texts in McGovern et al., 1994). Obvi­
ously, in a teaching situation the length of texts that it may be
possible to use is far greater, especially if out-of-class work can be
given. At this point differences between teachers and testing
begin to emerge.
Activities and performance conditions discussed in Chapter 3
are criterial for both teaching and testing; it is in the use made of
the data generated by the reader interacting with these that differ­
ences emerge.

Differences
It is universally accepted that, in testing, reliability of measurement
is of crucial importance. Any factor that reduces reliability must
be isolated and, if possible, eliminated or at least minimised. In
the teaching situation, however, reliability of measurement is far
less important. This is the major difference between testing and
teaching, from which all the other differences, set out below, derive.
174 Reading in a Second Language
Interpretations
We have argued that traditional testing aimed at forming a reli­
able estimate of an individual reader’s performance. In order to
achieve this, a consensus as to what constituted an agreed stand­
ard performance in a particular context had to be agreed in
advance by the testers. Alternatives to this standard had to be
eliminated. In other words, the test had to be constructed in such
a way that for each task there was a right answer. Scenarios, even
if reflecting situations in the real world, where either there was no
agreed right answer (as in reading for enjoyment) or where differ­
ent answers were arguably equally valid, had to be avoided.
In Chapter 2 we discussed different styles of reading and con­
trasted the submissive versus the dominant reader. We argued in
Chapter 3 that in testing there was little room for dominant read­
ing or challenging the texts as this would certainly defeat any
attempts to assess such activities reliably. In addition, no account
could be taken of the way in which the response offered had been
arrived at. In other words, testing at present is concerned with
product, and not process. Teaching, on the other hand, since
reliability of measurement is not as important, can take differing
products, interpretations, into account. Also, since it must be con­
cerned with how one goes about solving a reading task, it is also
permitted to encompass process.
Thus, reader-specific responses to text are possible and to be
encouraged in the teaching situation, whereas for reasons of test
reliability they were excluded from consideration in testing (see
Section 3.1). In learning to read, activation of relevant schemata
is seen as a key part of the reading process. Pragmatic inferencing
(Chikalanga, 1992) is also viewed as an important aspect of the
reading classroom, particularly at more advanced levels (see Sec­
tion 4.2 below for a discussion of this type of reading, and Nuttall
(1996: 121 and 167) for discussion of classroom procedures and
exemplification). Inferencing may indeed promote effective learn­
ing (Pearson and Fielding, 1991).
Training
Recent textbooks (Paran, 1991, 1993) build in cognitive and meta-
cognitive training as a central part of each unit. Their exercises
go beyond mere answering of comprehension questions and
attempt to teach strategies for coping with texts at the pre-reading,
 The teaching of reading 175
while-reading and post-reading stages. Sometimes learning is paid
lip service in testing tasks, but this is not common. It is this devel­
opmental rather than accountability function which distinguishes
summative testing from teaching.
Formative testing in the classroom is different, however, in that
the purpose, the use to which the results are put, should be dia­
gnostic. Teaching cannot proceed without reliable information
on what students can or cannot do. Formative testing has a crucial
role in providing such data for developmental purposes and, as
such, the distinction between teaching and testing tasks is blurred
in such formative use of tests.
Tasks
We have already said that testing is fundamentally concerned with
tasks. The testees must perform overtly, in response to a task set,
and their performance is assessed. Tasks, of course, occur in the
teaching classroom too, but there are differences. While there are
numerous definitions of tasks in teaching (Nunan, 1989, 1993;
Candlin and Murphy, 1987; Crookes and Gass, 1993a and b; Skehan,
1996; Skehan and Foster, 1995) we are happy to take Williams and
Burden’s (1997: 167) simple description as our working definition
for teaching purposes:
Basically, a task is anything that learners are given to do (or choose to do)
in the language classroom to further the process of language learning.
We would want to extend tasks to include such activities outside
the classroom and limit it for our purposes to reading, e.g. read­
ing a book at home for pleasure. We would also see formative test
tasks as falling under this umbrella. We also accept that N unan’s
elements of a task - input data, activities, goals, role of learners
and role of teachers - are all important and all interact with each
other (Nunan, 1993). Consideration of one will necessarily involve
some consideration of the others. For ease of description the role
of the learner and goals are considered in Section 4.2; input data,
i.e. texts and activities, are discussed in Section 4.3; and we examine
the role of the teacher in Section 4.4.
Staging tasks
Unlike the test, in the reading classroom tasks may well be broken
down, staged or scaffolded to help the less able reader. The teacher
176 Reading in a Second Language
provides help to enable students to complete tasks they would not
be able to do on their own. In contrast to the driving test itself, in
learning to drive one would not expect to do everything in a
single lesson, at least not in the early stages. So too with reading;
at some stage it might be necessary to focus on certain strategies
or skills, or analysis of sentence functions or text structures (see
Nuttall, 1996: 100-24).

Microskills
There may well be a necessity to bring some students up to a
threshold level of linguistic ability whereby they are enabled to
establish, expeditiously or carefully, the macrostructure of a text.
Activities promoting global comprehension may not be sufficient
for this. Learning the important skills of word recognition and
decoding may involve less direct, less global activities. Whereas in
testing we have suggested that successful expeditious and careful
reading for global comprehension must indicate a minimally
adequate knowledge of lexis and structure, and thus make testing
at the word level unnecessary except for diagnostic/placement
purposes, the teaching situation is different. We may have to pro­
vide the opportunities to practise activities promoting word recog­
nition and decoding skills in reading classes although, as such,
they may never appear in proficiency or achievement tests of
comprehension.

Co-operation
We noted above that in summative testing the teacher and fellow
students are removed from the interaction, and the help that can
be provided by both in learning to read is not available to the
student. The interest in the test situation is in what the student is
capable of comprehending unaided. In testing, some students
might be expected to fail; whereas in teaching, the agenda is to
try to ensure that nobody does.
In the test situation the reader is isolated from contaminating
sources, such as help from other students or from the teacher, in
an attempt to measure his or her ability in a construct unmuddied
by other influences. In the reading classroom pedagogical input
exists in terms of instruction and mediation that is absent from
the test context: i.e. advice on strategies and skills, practice in
 The teaching of reading 177
their use and discussion of their value. The aim here is to bring
about understanding rather than just measuring it in a statistically
reliable fashion.
The agenda of the classroom is more formative, co-operative
and developmental. Thus the methodology, the activity of the
reading classroom, can be wider and richer. The tasks available
for learning to read are more diverse and may involve working
with others, students or teachers, in both pairs and groups. The
learners are not being asked to demonstrate how well they can
use strategies or skills but rather to develop and improve their use
of these. The tasks employed may be similar in the reading lesson
to those of the test but differ in the way they are used.
In the classroom, comprehension questions set on texts are
often done in groups which promotes effective discussion con­
cerning how the answer was arrived together with feedback for
individuals in a non-summative manner. By verbalising about their
own reading they come to understand better the processes involved.
Nuttall extols the value of buzz groups (1996: 201): small groups
work on the task for a short period, and report back in plenary
followed by whole group discussion.
Conclusion
While there are a num ber of similarities between teaching and
testing there are also marked differences which necessitate con­
sidering teaching separately.

Background issues in the teaching of reading

Bernhardt (1991a: 173), in a broad survey of L2 reading research,
argues that
Research has not yet firmly established how to teach compre­
hension (or for that matter whether it is teachable). Neither has
research provided substantial insights into the process of second
language learning.
Research into L2 reading is certainly a relatively new field. As we
saw in Chapter 2, most of the serial models of reading were devel­
oped for LI learners by psychologists. In L2 it has mainly been
applied linguists who have been responsible for research, and this
is clear from their focus on componential models with a direct
178 Reading in a Second Language
link to pedagogy, as against serial models which are more con­
cerned with internal workings of the brain.
Given the limitations of our understanding in this field we con­
centrate below on reading activities and accompanying cognitive
behaviour which research has indicated as leading to measured
improvement in comprehension. Obviously any activity in the right
hands in the right circumstance may result in better performance.
However, we are interested in consistency of empirical evidence
and we have used Barr et al. (1991) and Bernhardt (1991a) to
guide our selection, as their reviews of reading research in LI and
L2 provide a wide empirical and theoretical base for selecting
teaching/learning activities for the reading classroom.
The comparative lack of research in L2 reading has encour­
aged us to draw on research in LI settings where this seems appro­
priate. In addition, we have had to rely on examining methods
and classroom textbooks to inform us about what goes on in the
L2/EFL classroom in the comparative absence of observational
studies of L2 reading training (see Siedow et al., 1985).

To teach or not
A car sticker in Britain carries the message: 'If you can read this,
thank a teacherThe implication is clear: if you are not taught to
read, you will not learn. It is, however, perfectly likely that some
LI children, at least, learn to read with little or no formal teach­
ing, and it is certainly the case that L2 readers learn to read the
second language without formal instruction. W hether explicit
instruction is any more effective than simply encouraging students
to read and form their own rules remains unproven. We would,
however, agree with Pearson and Fielding (1991), in their excel­
lent review of comprehension instruction in LI, that the danger
of the non-interventionist approach is that the good readers get
better and the poor do not. The gap widens.
In some classes students will not lack ability in the skills and
strategies discussed below, and until problems become evident in
these areas they may well be avoided. The time to practise these
skills and strategies is in response to needs or lacks that become
evident.
What is also clear is that comprehension teaching effectiveness
may differ from context to context and there are no generic class­
rooms (Bernhardt 1991a: 173); those of young bilinguals will be
 The teaching of reading 179
very different from adult second language readers preparing for
postgraduate study in an overseas context. Because of the great
diversity of reading contexts we shall not be providing multiple
examples of reading activities and texts; there are stimulating col­
lections of these already available to which we would direct the
reader (Grellet, 1981; Nuttall, 1996; Wallace, 1988; Williams, 1984)
and a plethora of reading textbooks aimed at specific audiences.
We will instead focus on the evidence from principled reading
research and instruction to formulate generic suggestions for teach­
ing. We will attempt below to determine what research indicates
as being the most productive activities across diverse situations.
The discussion below is therefore for the most part in terms of
principles rather than commentary on specific examples of class­
room tasks. Reference will, however, be made to sources the reader
can consult for practical exemplification.

4.2 FOCUS ON METACOGNITIVE STRATEGIES

As we noted above, one of the major differences between teach­
ing and testing is the presence in the former of teacher input in
the form of training, in which strategy training has become of
increasing importance in recent years (see discussion in Sections
2.3, 2.4 and 3.2 on strategies and skills). Cohen (1998) provides a
comprehensive and clear account of strategies for using and learn­
ing a second language. A useful distinction can be made between
cognitive and metacognitive strategies.
Cognitive strategies are the more familiar mental processes that
enable us to read, ranging from working out the meaning of words
in context through to skimming a whole text quickly to extract
the gist. Metacognitive strategies are more concerned with thinking
about the reading experience itself and are seen to involve
. . . learners stepping outside their learning, as it were, and looking
at it from outside. Such strategies include an awareness of what one
is doing and the strategies one is employing, as well as knowledge
about the actual process of learning. They also include an ability to
manage and regulate consciously the use of appropriate learning
strategies for different situations. They involve an awareness of one’s
own mental processes and an ability to reflect on how one learns,
in other words, knowing about one’s knowing.
(Williams and Burden, 1997: 148)
180 Reading in a Second Language
Cohen (1998: 7) describes metacognidve strategies as dealing with:
pre-assessment and pre-planning, on-line planning and evaluation,
and post-evaluation of language learning activities and of language
use events. Such strategies allow learners to control their own cog­
nition by coordinating the planning, organizing, and evaluating of
the learning process. There is a rather extensive literature demon­
strating that the higher proficiency students are more likely to use
metacognitive strategies than the lower-proficiency ones and to use
them more effectively as well.
He details distinctions that may be made between general approa­
ches and specific techniques or actions. He suggests (1998: 10):
A solution to the problem would be to refer to all of these simply as
strategies, while still acknowledging that there is a continuum from
the broadest categories to the most specific or low level.
He also acknowledges that the distinction between metacognitive
and cognitive strategies may not always be clear cut and there may
on occasion be some overlap (op. cit.: 12) and both may be legit­
imate interpretations of some actions.
Williams and Burden (1997: 155-6) draw attention to the import­
ance of knowledge of tasks (see also Cohen, 1998: 14):
Knowledge about task refers to an awareness of the purpose and
demands of the task, as well as an ability to assess the information
provided, and to select what is relevant from what is irrelevant.
Knowledge of strategy involves an understanding of what strategies
should be used for different types of task.
They also detail (1997: 145-6, 156) a num ber of preplanning and
planning-in-action metacognitive strategies drawing on the work
of Nisbet and Shucksmith (1991) and Wenden (1987b), namely:
■determining objectives
■selecting methods
■predicting difficulties
■asking questions
■planning
■monitoring
■checking
■revising plans
■evaluating outcomes.
 The teaching of reading 181
Cohen (1998: 14) outlines the areas of metacognitive strategy use
in a similar fashion:
(1) a goal setting component where the respondents identify the tasks
and what they are going to do;
(2) an assessment component, whereby the speakers (listeners, readers
or writers) determine what is needed, what one has to work with,
and how well one has done; and
(3) a planning component, whereby the respondents decide how to
use their knowledge of the topic and their language knowledge.
From their perspective of education as a life-long process, Wil­
liams and Burden see one of the main aims of the educator as being
to help students cope with self-directed, autonomous learning.
Teaching students to read effectively unaided would seem to be
potentially a powerful contribution to this, if not the single most
important. If we can help students to read carefully and expedi­
tiously on their own for their own purposes, then this would be
success indeed.
There is evidence to support such metacognitive training.
Alvermann and Moore (1991) cite research studies which demon­
strate the value of learning strategies initiated and directed by
students. They detail how, through metacognitive training and
self-questioning, students can be taught to monitor comprehension
(see also Carrell et al., 1989; Cohen, 1998; Williams and Burden,
1997: 144-66). Oxford and Crookall (1989) report on a number
of studies where the experimental group received training in
metacognitive strategies and subsequently outperformed a control
group who did not.
As far as strategy training is concerned, Williams and Burden
(1997: 156-66) offer sound advice on how best to go about it (see
also Cohen, 1998: ch. 4). They present (pp. 158-9) a strategic
teaching model developed by Jones et al. (1987), which has the
following guidelines:
■Assess strategy use (through think aloud, interview, questionnaire).
■Explain strategy by naming or telling how to use it, step by step.
■Model strategy by demonstration or verbalisation of own thought
processes while doing it.
■Scaffold instruction by providing support while students practise;
adjusting support to suit students’ needs; phasing out support
to encourage autonomous strategy use.
182 Reading in a Second Language
■Develop motivation by providing successful experiences; relat­
ing strategy use to improved performance.
Williams and Burden (1997: 162) suggest that most of the pro­
cedures for strategy training they surveyed:
generally involve a sequence of first helping students identify or
become aware of strategies they are already using, then presenting
and explaining a new strategy, with a rationale for using it. At this
stage the teacher might model the strategy. This is followed by
practising it, at first with substantial support or ‘scaffolding’ but
gradually reducing this to encourage autonomous use. Finally, stu­
dents are helped to evaluate their success.
This would seem appropriate also for training in the cognitive
strategies such as skimming, search reading and scanning, dis­
cussed in Section 4.3 below.
Cohen (1998: ch. 4) offers a comprehensive discussion of the
various forms that strategy training can take, and the reader is
referred to this for a full discussion of this area. He also provides
an informed consideration of extending the role that teachers can
play. In this he develops a view of a role for teachers as learner
trainers as well as language instructors (p. 97):
One potentially beneficial shift in teacher roles is from that of
being exclusively the manager, controller, and instructor to that of
being a change agent - a facilitator of learning, whose role is to help
their students to become more independent, more responsible for
their own learning. In this role, the teachers become partners in
the learning process.
The teacher’s role is seen as one of gradual withdrawal with the
objective that, by the end of the course, the student should be
able to apply the strategies independently. The difficulty seems
to be deciding when to start turning over more responsibility to
the students, for example, by reducing coaching or scaffolding.
Palincsar and Brown (1984) describe a technique, called recipro­
cal teaching, which they used with retarded readers where initially
the teacher does a lot of the work although still in a co-operative
format and gradually learners assume more responsibility as they
improve until they are able, finally, to work independently.
As yet there is little evidence on the long-term effects of strat­
egy training or transferability to novel contexts. However, there is
some support for their immediate impact on reading performance
 The teaching of reading 183
(see Cohen 1998: 483 et seq. and ch. 5). Different readers will
have different perceptions of their usefulness and part of the train­
ing would be enabling learners to explore this relevance to them.
As Williams and Burden (1997: 164) point out:
they will employ particular strategies if they have a sense of owner­
ship or choice in the strategies used, they are clear why they are
using them, and they want to complete a task to achieve a goal that
they have identified as worthwhile.
Cohen (1998: 12) rightly warns that:
. . . the effectiveness of a strategy may depend largely on the char­
acteristics of a given learner, the given language structure (s), the
given context, or the interaction of these,
and suggests that a strategy may work in one particular case but
not necessarily in all cases. He also indicates that strategies may
need to be related to learning styles and other personality related
variables (op. cit.: 15-16).

Some potentially useful metacognitive strategies

Having briefly surveyed the nature of metacognitive strategies and

how, in general, to present them in training we would now like to
examine in closer detail a number of such strategies that appear
to be of some value. We adopt the now conventional distinction
between pre-reading (planning) strategies, while-reading (monitor­
ing) strategies and post-reading (evaluation) strategies.
Pearson and Fielding (1991: 836-9), on the basis of a wide
survey of the LI reading research literature, identify two generic
while-reading strategies and practices to support students to engage
independently and actively with text. These are self-questioning
and self-monitoring.
We would like to add two pre-reading strategies: namely preview­
ing and prediction which relate to Pearson and Fielding’s gener­
ative learning. Generative learning describes how comprehension
occurs when learners build relationships between parts of the text
and between the text and their background knowledge: associations
improve comprehension. The two pre-reading strategies discussed
below can help activate schemata prior to the reading process and
can contribute to this process. We also look at post-reading activities
184 Reading in a Second Language
which involve evaluation of the text, relating it to one’s own experi­
ences or context.

Pre-reading activities

Previewing
Previewing can be used to make a decision whether to read a
book, an article or a text. Where appropriate to text type it might
involve:
■thinking about the title
■checking the edition and date of publication
■reading the table of contents quickly
■reading appendices quickly
■reading indices quickly
■reading the abstract carefully
■reading the preface, the foreword and the blurb carefully.
Hamp-Lyons (1984: 305) adds that previewing helps students re­
cognise the difficulty level of a text and comparative difficulty
with other texts in the same field, helps them judge the relevance/
irrelevance of a text for a particular topic, and helps them decide
which book from a set of possibilities would be more appropriate
to read for a specific purpose. Its value for teaching is the amount
of time it might save if it prevents prolonged reading of some­
thing of no value (see Nuttall, 1996: 45-8).
It is of particular use in deciding whether textbooks or parts of
a textbook are of value, though browsing through a novel at the
airport bookshop before deciding to purchase is another manifesta­
tion. The reason that it seldom features in tests is on the grounds
of efficiency and reliability. It is difficult in the exam situation to
provide the same textbook(s) for large numbers of candidates.
Additionally, the num ber of items that can be usefully written are
often limited with implications for test reliability. Similarly, how
would one evaluate xs decision to purchase a particular book?
However, in the classroom context, previewing may be very
useful, particularly for English for Academic Purposes Students.
Previewing has obvious links with expeditious reading strategies,
particularly skimming for gist, discussed in Section 4.3 on cognit­
ive strategies.
 The teaching of reading 185
For exemplification of previewing, see Grellet (1981: 58-61),
Hamp-Lyons (1984) and Trzeciak and Mackay (1994: 5-10).
Prediction
After taking the decision to read a text, this strategy is used to
anticipate the content of a text; to make hypotheses about the
macropropositions it might contain. It is a form of psychological
sensitising, thinking about the subject and asking oneself related
questions.
In theoretical terms it accords with the hypothesis advanced
earlier in Section 3.3 that establishing a macrostructure for a text
is an aid to more detailed comprehension. One might also hypo­
thesise that the activation of relevant schemata should facilitate
the reader’s interaction with a text (see Section 2.2). Finally, this
activity has the potential to clarify for the reader what the pur­
poses for reading the particular text might be.
It is often a case of supplying or activating appropriate back­
ground knowledge, and this might best be done through pre-
reading activities: lectures, discussion, debate, real-life experiences,
text previewing or introduction of vocabulary. It makes use of
top-down processing to activate different kinds of schemata in
common with many pre-reading activities.
Haines (1988) uses surveys and questionnaires to encourage
discussion and activate and build up background knowledge pre-
reading, and Paran (1993) uses surveys to similar effect. Tomlinson
and Ellis (1988) offer a range of pre-reading activities aimed at
activating formal knowledge of text.
Williams and Moran (1993: 66) suggest:
Perhaps the most effective of these activities are those which elicit
factual information or a personal response and ask the students to
pool such information in pair or group work. Preferably, this is
followed by a task which relates the discussion to the first reading
of the passage.
In comparing teaching and testing tasks currently in use it is notice­
able that, whereas prediction activities are now a common feature
in textbooks on the teaching of reading, they seldom feature in
tests. Part of the reason for this is presumably that such data do
not lend themselves easily to assessment. The open-ended and
idiosyncratic nature of such prediction, based as it is on already
existing constructs, is obvious.
186 Reading in a Second Language
Williams (1984: 36-51) provides useful advice and examples of
pre-reading activities that can be employed in the language class­
room. Swaffar (1981) and Carrell and Eisterhold (1988) describe key
word and key concept pre-reading activities. See also Glendinning
and Holmstrom (1992: 20-4), Langer (1981) and McGovern et al.
(1994: 11-12) for further exemplification.

While-reading strategies

Self-questioning
This is identified by research as a characteristic of good reading
when it promotes cognitive processes such as inferencing, monitor­
ing understanding and attending to structure. Alvermann and
Moore (1991: 961) detail how ‘generally, instruction in self ques­
tioning improves student processing of text’ and note that ‘poorer
readers tend to benefit most from such training. Scaffolding of
instruction leading to gradual control appears to be beneficial.’
Nuttall (1996: 37) describes this activity as interrogating texts; text
talk. For students unfamiliar with this activity the teacher inter­
rogating the text aloud can provide a valuable example, particularly
where the focus is put on important problematic aspects of a text.
Organised methods involving self-questioning have been in use
for some time (see Nuttall, 1996: 129; and Richards, 1989 for de­
tails of SQ3R). Palincsar and Brown (1984) focused on teaching
summarising, questioning, clarifying and predicting skills, arguing
that these activities, if engaged in while reading, enhanced compre­
hension, and, at the same time, gave the student the opportunity
to monitor whether comprehension was succeeding.

Selfmonitoring
Monitoring one’s own comprehension - checking that compre­
hension is taking place and adopting repair strategies when it
isn’t - is seen as a hallmark of skilled reading. It is important that
students are aware of how various strategies will help them. Self­
verbalisation was also seen as important (Pearson and Fielding,
1991: 838).
The connection with schema theory is clear: by asking them­
selves whether they understand, learners are asking whether it fits
 The teaching of reading 187
in with what they know already. Thus they learn how to under­
stand what they read in the process of learning how to monitor
their comprehension (Pearson and Fielding, 1991: 847).
Alvermann and Moore (1991: 962 et seq.) sound a note of
caution in that many of these studies were achieved under experi­
mental rather than field conditions with consequent threats to
their ecological validity: they were decontextualised; an experi­
m enter rather than the normal classroom teacher introduced the
intervention; texts were specially prepared and were often shorter
than normally met texts; and students were often not prepared
in the use of the strategy before the intervention. They argue
(p. 974) that we need to develop a research methodology which
would actively involve teachers and carefully document baseline
data of the situation preceding the decision to implement an
innovation and also collect data to monitor its effect.

Post-reading strategies

Evaluation and personal response

Questions of evaluation and personal response are also seen by
teachers and course book writers as a valuable post-reading activ­
ity, relating the text to the outside world (Nuttall, 1996: 167, 188-
9). The work may be done either orally or in writing, though
Nuttall (p. 167) favours the former because of the importance of
discussion and exchange of views. Research suggests that in learn­
ing to make the text their own the readers will better comprehend
it. Readers can be encouraged to relate content to their existing
schemata and to evaluate it in the light of their own knowledge
and experiences. This promotes greater interaction with text and
may lead to more successful reading encounters. Orlek (1996)
provides a thorough and stimulating case for teaching readers
to challenge text rather than submissively trying to recreate the
author’s intended message.
Training in metacognitive strategies is a relatively new depar­
ture in L2 teaching of reading but it seems to have potential cash-
in value for enhancing the reading of particular texts and, more
importantly, carrying over to future reading experiences. Cohen
(1998: ch. 4) provides a detailed and useful review of strategy
training, focusing primarily on strategies-based instruction aimed
188 Reading in a Second Language
at increasing learners’ awareness of the benefits that such strat­
egies might have for them. More familiar to teachers is training in
cognitive strategies and skills and most of the textbooks we surveyed
attempt to cover a range of these with varying degrees of success.

4.3 FOCUS ON COGNITIVE STRATEGIES

AND SKILLS

Introduction: purpose determines choice

An awareness of top-down (reader-driven) and bottom-up (text-

driven) processing strategies can benefit teachers. It should help
them to incorporate classroom activities which encourage L2 read­
ers to use appropriate combinations of such strategies in reading
for different purposes (Carrell, 1988; Richards, 1989). Informa­
tion on effective and ineffective reading strategies can help inform
and improve students’ reading efficiency. For example, we will argue
below that some of the time devoted in class to working out the
meaning of words in context might be better spent on activities
promoting automaticity. An excessive focus on the former might
actually impede developing fluency!
Over-reliance on either top-down or bottom-up processing to the
neglect of the other may hamper readers; efficient and effective
second language reading requires both top-down and bottom-up
strategies in different combinations for different purposes (Car­
rell, 1988: 240-1; Rumelhart, 1977, 1980). For example, it would
be mistaken only to rely on word by word bottom up processing in
a skimming exercise designed to extract quickly the gist from a
text. However, some careful processing of textual clues to overall
meaning might be necessary once these are located through more
expeditious strategies.
Cognitive strategies may range from macro activities, such as
skimming a text quickly for gist, down to micro activities, such as
working out the meaning of a word in context through cognates,
translation, etc. (see Jimenez et al., 1996 and Hosenfeld, 1984). As
we saw in Section 2.3, strategies have the following characteristics:
■they are essentially problem solving on whatever level, macro
or micro
■they are goal oriented
 The teaching of reading 189
■they are purposeful
■they involve efficiency and selection
■in the case of expeditious strategies, they also involve speed
■they are consciously adopted (in contrast to subconscious use
of skills)
■they are, by implication, directly teachable.
Williams and Moran (1993) point to an increasing eclecticism
in the strategies in the course books they surveyed. A similar pic­
ture emerges in the more recent textbooks we surveyed. This has
not always been the case. Following the introduction of top-down
approaches there has been a tendency to overlook the crucial
importance of word recognition and decoding in the early stages
of learning to read and in training those not brought up on the
Roman script (see Paran, 1996).
Williams and Moran (1993: 67), in their review of reading course
books, argue:
It would in most second language teaching situations clearly be
misguided for teaching to focus exclusively on top down processes.
As Eskey (1988: 98) puts it, ‘If we can no longer afford to teach
reading as a kind of linguistic analysis, we also cannot afford to
teach it as a kind of cued speculation.’

Individual tasks at the local level

Word recognition: a neglected area
Bernhardt (1991a: 174) points out that an assumption is often
made in L2/EFL classrooms that students can already read and that
. . . reading is just a slower form of first language reading, reading
instruction per se really does not exist.
One of the problems, especially in the L2 teaching of reading, is
that we are accustomed to teaching students who will already have
learned to read in their LI. As a consequence, the skill of word
recognition may not receive as much attention as has global com­
prehension of text for example. This may mean that second lan­
guage reading students who do not share a common orthographic
script may be placed at a disadvantage (see p. 54). The additional
processing time they require for word recognition may have a
knock-on effect in other areas of the reading process.
190 Reading in a Second Language
Additionally we tend to have little experience of emergent lit­
eracy, early reading and writing development (Sulzby and Teale,
1991) as many of our students in the UK tend to be adults rather
than very young children and an assumption is made either that
they will have gone through this stage, or that acceptance proced­
ures of overseas students on courses preclude such recognition
problems. Williams and Moran, in their review of course books,
call for more research into initial reading in a foreign language
and refer to Wallace (1988) as one of the few serious pieces of
work in this area.
Juel (1991: 759) argues that we know little about the transition
from emergent to beginning LI reading largely due to a focus on
‘higher order’ reading processes by the research community aris­
ing in the USA from a concern over the poor performances of
older students on national assessments of higher order reading
skills. She details how they already felt they were doing an adequ­
ate job teaching lower order decoding skills. She then queries
how good a start children in the USA are actually given with the
dire consequences this may have for later development as the LI
research evidence mounts that there is a strong link between early
progress in reading and later reading ability and comprehension.
She also notes that most of the models we have are for the skilled
reader and little attention has been paid to developing a compre­
hensive model of the reading acquisition process.

Automaticity
We pointed out in Chapter 2 that a strong criticism of so-called
top-down models was that they attributed too much importance to
hypothesising, or guessing, whether of lexical items or larger units.
Stanovich (1980) points to the implausibility of hypothesis testing
being of much value to the skilled reader as it take so much less
time to recognise a word than to go through a complex guessing
game. In one of the most important contributions by cognitive
psychologists to reading research (and potentially the teaching of
reading) in recent years, it was repeatedly found that good readers
used context much less often than poor readers when recognising
printed words. In fact they appear to be able to recognise words
without any conscious thought, i.e. at the automatic level.
Juel (1991: 771) cites important evidence that early attainment
of decoding skill/word recognition is a very accurate predictor of
 The teaching of reading 191
later reading comprehension in LI children. Those who do poorly
in the first year of learning to read are unlikely to improve their
position as compared to those who do well. Poor decoding skill
may delimit what the child can read and the differences are further
compounded by out of school experiences.

Automaticity in L2
The increased importance attributed to automatic word recogni­
tion in LI reading has been extended, though with less empirical
support, to the L2 reading area. Previously in L2, a great deal of
faith had been placed on decoding by means of context. Haynes
(1984), however, points out that we need to get the level of
automaticised vocabulary up rather than focusing on decoding in
context. Haynes points out (48):
Rapid precise recognition of letters and words, that is, bottom up,
more input constrained processing, must be mastered before fluent
reading can take place.
She cites evidence from LI studies that fluent reading is achieved
by increasing one’s bottom-up processing of print and decreasing
semantic and syntactic guesswork, though this is not as yet proven
for L2. She questions the emphasis given in textbooks to guessing
the meaning of unknown words from surrounding context as
the main approach to learning vocabulary. Context often proved
inadequate to support accurate inferencing and encouraging the
guessing from context strategy might well lead to frustration in
these cases.
There is some negative evidence for this position in the L2
area. Bensoussan and Laufer (1984) found no evidence that bet­
ter readers are able to use context more effectively for lexical
guessing than less proficient students. More crucially, they argue
that in many cases only a minority of word meanings can be recov­
ered from the context. Working out the meaning of words in
context is only a part of the vocabulary skills needed for fluent
reading and it appears that it may actually interfere if a student
over-relies on this strategy.
Beck (1981) argues that ‘basic recognition exercises to improve
speed and accuracy of perception may constitute an important
component of an effective second language reading programme’.
192 Reading in a Second Language
If, as appears likely, automatic word recognition is more import­
ant to fluent processing of text than context clues, the large-scale
development of recognition vocabulary may be crucial to reading
development (van Dijk and Kintsch, 1988; Perfetti, 1985). Poor
readers have simply not acquired automatic decoding skills. Poor
readers spend too much processing time thinking about words
and relating them to the surrounding context, rather than auto­
matically recognising them.
Bernhardt (1991a: 235-6) argues that the ultimate goal is auto-
maticity. Good LI readers process language in the form of written
text without thinking consciously about it, and good L2 readers
must also learn to do so. It is only this kind of automatic process­
ing which allows the good reader to think instead about the larger
meaning of the discourse - on the one hand, to recover, the
message that the author intended to convey and, on the other, to
relate that new information to what the reader knows and feels
about the subject, and to his or her reasons for reading about it.
In short, it is only this kind of local processing that allows for
global meaning with true comprehension.

L2 reading speed
Bernhardt (1991a: 234) argues that a major bottom-up skill is
reading as fast in that language as their knowledge of it will allow,
in relation to their reading purposes. Where appropriate we need
to dedicate some time for rapid identification of lexical and gram­
matical form. Juel (1991: 771) quotes Chall (1979): ‘. . . (learners)
have to know enough about the print in order to leave the print.’
It is noted in the literature that L2 readers often read texts
more slowly than LI readers. One of the most striking differences
between LI and L2 readers of English texts is their speed. Haynes
(1984: 50) and others have identified the root of the problem as
the length of the fixation slowing down the reading rather than
num ber of fixations or regressions. Haynes (1984: 50) notes:
There is no clear experimental evidence explaining these longer
visual fixation times, but a strong possibility involves the time re­
quired for lexical access, that is the time it takes for a reader to
match the printed word to a word meaning in memory.
It seems likely that it takes longer to access lexical meanings,
remember what a word means, in L2 than it does in LI. L2 readers
 The teaching of reading 193
of English do not have large well-practised vocabularies and years
of experience of recognising words in print. Hence, it takes them
longer to decide whether a word is known or unknown and, in
the latter case, whether to skip it or not.

Developing automaticity in L2 reading

Although the importance of automaticity in decoding is gaining
general recognition, there is less agreement on how to achieve
the goal in L2 reading. Haynes (1984: 59) argues that:
The importance of word unit processing needs to be recognised in
ESL teaching. First, precision of encoding spelling and pronunci­
ation can be increased through oral and written practice of import­
ant vocabulary from reading.
Haynes also advocates dictionary work as a method of separating
words which look familiar. The aim is to help the student distin­
guish new words efficiently in lexical memory. Nuttall (1996: 62)
is less sanguine about the use of dictionaries and warns that it can
slow down reading considerably and reduce effectiveness in read­
ing. Occasional rather than constant use is advocated. She also
cautions that all of this does not mean that guessing from context
is not essential but should be seen as one of a num ber of bottom-
up as well as top-down strategies that can help the student to reach
a meaning.
Juel (1991: 783) points out that automaticity in most skills comes
from overlearning:
repeated practice frees up one’s attention so that it does have to be
focused on the mechanics involved in the specific activity. . .
though she does admit that we do not know exactly what it is
about word recognition that becomes automatic, i.e. is it recogni­
tion of common sound spelling patterns, recognition of high fre­
quency words, etc?
Paran (1996), in a recent article, makes a strong case for devel­
oping exercises to help EFL students recognise vocabulary more
automatically. He puts the case for a greater focus on bottom-up
processing (p. 25): ‘good readers do not rely on hypothesis forma­
tion and prediction as much as is commonly thought. Visual input
and bottom up processing during reading are of great import­
ance.’ He argues that the top-down approach (Goodman, 1967:
194 Reading in a Second Language
75, 78) has permeated ELT reading materials and teacher train­
ing despite the advent of more comprehensive interactive models
and cites Grellet (1981), Nuttall (1982) as important books illus­
trating this bias. While accepting the value of interactive compens­
atory approaches which allow learners to compensate for lack of
developed automaticised skills, Paran feels we need to go further.
Though providing context or activating background knowledge is
important to compensate for poor automaticity, readers need to
be weaned off this support and be trained to do without it.
Paran (1996: 29) claims:
. . . one of the goals of L2 reading instruction is to make readers
less reliant on top down processing and help them progress towards
greater reliance on bottom up strategies as they become more
proficient.
Guessing may well take place on a global level, but Paran argues
that this is not an appropriate strategy for word or phrase recogni­
tion; the lower levels of processing (p. 30). Top-down approaches
would seem to be more suited to expeditious reading such as
skimming or search reading where the focus is on the macro­
structure of the text. We note along with Paran the absence of
such activities from most course books on the teaching of reading.
Paran also points out the limited variety in the exercise types
for automaticity; basically matching identical words in lists and
suggests that they might not go down well at advanced level. Bloor
(1985: 345) sees little use for it beyond an elementary level. How­
ever, if we relate it to the activity of scanning discussed later under
expeditious reading, it may well be that practising this strategy has
important spin offs for automaticity and widens the base of exer­
cise types that concerns Paran and Bloor. Scanning is an activity
suitable for advanced readers as well as elementary!
It may also be the case that one major value of extensive read­
ing (see p. 215 below) lies in its value for developing automaticity
of word recognition. Familiarisation with a large number of words
through extensive practice is likely to free up processing time.
For further exemplification the reader is advised to look at
Mahon (1986) and Nuttall (1996: 54-61) for useful suggestions
concerning building L2 reading rates and de Leeuw and de Leeuw
(1990) for LI. For further discussion on automaticity in L2 read­
ing teaching, see Nuttall (1996: 45-61), Dubin et al. (1986) and
Silberstein (1994).
 The teaching of reading 195
Vocabulary acquisition
As we have discussed so far, automaticity has largely been con­
cerned with decoding. The LI readers, given that they are read­
ing appropriate material, can be assumed to have lexical entries
previously established for the items encountered. In other words,
they ‘know’ the words; they just have to make automatic recogni­
tion of the orthographic rendering of these words. This is particu­
larly the case with adult or near adult readers. This situation cannot,
of course, be assumed in the case of L2 learners. Here the readers
may well frequently face items which they not only have difficulty
decoding, but with which they are simply unfamiliar; the item may
not yet have a place in their lexicon. Thus, except in fairly unusual
cases where the L2 readers are very fluent in the spoken language
before they are introduced to the written form, L2 readers have
to acquire vocabulary items before they can begin to make recog­
nition of these items automatic.
How best to acquire a large stock of vocabulary is a problem. It
has often been argued that vocabulary is best acquired through
reading. Thus Nuttall (1996: 62) claims: ‘An extensive reading
programme is the single most effective way of improving vocabu­
lary.’ However, according to Nation (1997), extensive reading is
unlikely to lead to large increases in our vocabulary knowledge
unless we read a very large amount so that new words are re­
peated sufficiently in context for us to learn them (apparently 12
times is the recommended dose, based on Saragi et al., 1978).
Thus extensive reading may not be the panacea for vocabulary
acquisition that it is often thought to be. For acquiring vocabulary
we may have to rely also on instruction in intensive mode.
Many teachers see one of their key roles in the reading class­
room as expanding vocabulary knowledge and developing learners’
ability to continue to increase their vocabulary. It is generally
accepted (see Nuttall, 1996: 63) that a vocabulary of 5000 words
is needed to start independent reading, although the empirical
evidence for this is slight (see Laufer, 1989). If this figure were
accurate it would seem to be beyond many of the students we
teach and we would need to be extremely careful in selecting
texts for the majority of our students (i.e. ensuring the lexical
load is below their level) an d/or focus on language work for a
longer period with those only possessing a restricted vocabulary
of, say, less than 2000 words.
196 Reading in a Second Language
The desire to expand student vocabulary is supported by the
LI literature on the effects of vocabulary knowledge on compre­
hension. Correlations between knowledge of word meanings and
ability to comprehend passages containing these words are high
and well established in LI reading studies (Anderson and Freebody,
1981) and in the L2 literature (Laufer, 1989). Such correlational
evidence, of course, fails to establish knowledge of word meanings
as a cause of comprehension but suggests that the two are distrib­
uted in a similar manner in the population. Stahl and Fairbanks
(1986) provide a meta-analysis of studies concerned with effects of
vocabulary instruction on comprehension and on learning of word
meanings.
Clarke (1979, 1980/1988) and Cziko (1978) hypothesise that
competence in the second language (grammar and vocabulary)
may place a ceiling on second language reading ability. Implications
are that ‘good reader’ top-down reading skills may be parasitic on
language to a larger degree than in first language reading. How­
ever, Hudson (1982: ch. 2) has found that schema production,
top-down processing, is very much implicated in the so-called short
circuit of second language reading, and that schemata can over­
ride language proficiency as a factor in comprehension. This would
accord with data we put forward from language testing studies in
Chapter 3.
Beck and McKeown (1991) argue that before we can talk of the
value of instruction we need to have a clear idea of what it means
to know a word and how we are to measure vocabulary size and
development. Like many before them, they suggest that know­
ledge of a word is not an absolute but rather a continuum from
not knowing to ‘rich decontextualised knowledge of a word’s
meaning, its relationship to other words, and its extension to
metaphorical uses’ (p. 792). The distinction between receptive
and productive vocabulary has a long history. We need to recognise
that there is a difference between becoming aware of a new word
and learning it. Obviously size of vocabulary is determined by how
a word is defined; what knowing it means and what corpus (and
sample from that corpus) of words are used to test it (p. 793).
Despite these problems it is clear that high-ability students have
much larger vocabularies than lower ability students.
Much of the research on vocabulary knowledge is based on
multiple-choice question tests; however, this format obviously suf­
fers from the drawbacks listed in Chapter 3 and raises a question
 The teaching of reading 197
mark about much of the research. In particular, such tests do not
distinguish between words that are known well and those which
are only vaguely familiar, etc.
We need to develop measures/tasks which reflect the point
along the continuum of word knowledge at which we wish to elicit
data and determine the measures to be introduced optimally at
each point. Do we want to test minimal degree of familiarity or
discrimination among related concepts or completeness of word
understanding as in accurate definition? It also seems likely,
according to Beck and McKeown (1991: 808), that the effect of
vocabulary instruction on comprehension may well be qualitative
rather than quantitative. Studies focusing on qualitative under­
standing may well demonstrate clearer effects than the latter.

A case for direct instruction

Direct instruction cannot account for all the vocabulary acquired
by learners and so it is often argued that learning vocabulary from
context in written text is likely to promote vocabulary learning.
We expressed some reservations concerning this in the previous
section. The research evidence on the value of attempting to teach
the ability to use context is mixed (Beck and McKeown, 1991:
802-3) but the ubiquitous finding is that ‘learning word meanings
from context does not seem to occur with particular ease’.
Finding ways to encourage learners inside and outside the class­
room to extend their reading is another contender, although as
we pointed out above (Nation, 1997) this may not be the panacea
for vocabulary deprivation it was once thought. Getting students
to read more is not likely to be a problem with the skilled reader
as nothing succeeds like success. The problem lies in encouraging
less able students to do so, and in recognising that wider reading
may simply not be enough for them to get their vocabulary know­
ledge up to a critical mass. For these students direct instruction
may be necessary.
This may involve classroom activities where word meaning
information is intentionally given to students either through de­
finition from recourse to a dictionary (Nuttall, 1996: 62-3) or
explanation in context. More recently, the key word method,
semantic mapping and semantic feature analysis have appeared in
the literature (Beck and McKeown, 1991: 804). Again research
has not indicated a best method across studies. Part of the problem
198 Reading in a Second Language
lies in the way methods are labelled but there is also some degree
of overlap which may interfere with experimental enquiry. How­
ever, what is clear is that instruction does have an effect (Beck
and McKeown, 1991: 805) on word learning and ‘there is advant­
age from methods that use a variety of techniques . . . there is
advantage from repeated exposures to the words to be learned’.
There is little indication that vocabulary instruction leads
to increased comprehension, i.e. little evidence of a causal link,
despite strong reported correlations between the two. Beck and
McKeown (p. 806) argue that this may be because the semantic
processes in reading comprehension required fluency of access to
word meaning and richness of semantic network connections, in
addition to accuracy of word meaning knowledge.
To have an effect, research indicates that successful instruction
‘provided more than one or two exposures to each word, pre­
sented both definitional and contextual information, and engaged
students in deeper processing’ (ibid.: 806). Semantic features ana­
lysis and semantic mapping involve active processing by getting
students to examine the relationship between words. Both involve
active processing, e.g. comparing and combining with known in­
formation to establish meaning. They go beyond simply entering
new information in memory as per use of a dictionary.
Rudzka-Ostyn (1986) introduces students to a range of tech­
niques, including semantic fields, componential and collocational
analysis, and semantic relationships such as synonymy. Similarly,
Beck and McKeown (1991) advocate the use of semantic feature
grids to discuss the relationships between items. For a discussion
of semantic-based vocabulary teaching methods see also Pearson
and Johnson (1978).
Such direct instruction cannot obviously teach all words. A strat­
egy which focused on high-frequency words in a mature vocabu­
lary which also have a high utility across discipline areas, e.g.
semi-technical vocabulary, might prove to be the most effective
and efficient strategy.
Nuttall (1996: 64 et seq.) offers useful advice on how to get stu­
dents to ignore unknown words and provides a series of practical
techniques for helping in this, e.g. understanding gapped texts:
shows students they can get the gist of a text without understand­
ing every word. Nuttall also provides a num ber of valuable activit­
ies for developing word attack skills for dealing with words which
are important for comprehension (pp. 69-77):
 The teaching of reading 199
■looking for structural clues
■grammatical function: its place in the sentence
■morphology: its internal structure
■inferencing its meaning from context
■using a dictionary for unknown key words not accessible by
other means.
Paran (1991, 1993) offers a wide variety of exercises for helping
learners to understand strategies they might use for acquiring the
meanings of words, for example:
■relating a new word to a known word
■contrasting a word with another word in the sentence
■understanding the sentence as a whole
■employing knowledge of the world
■relating a word to a word in LI.
It is noticeable that top-down techniques are being used here on
occasion.
As in all exercises we must be aware that we can never be sure
that, on the basis of meeting a word once, we know what it means.
Bright (1965) stresses to students the multiple meanings of com­
mon words: ‘The way a word fits into the language and life of the
people who use it is not a simple thing. Hardly any words except
technical ones have one simple meaning.’ Like many others, he
recommends extensive reading as a method, not of acquiring new
words, but of broadening one’s knowledge of the multiple mean­
ings of already known items. While seemingly sensible advice, there
is little empirical support for this in the literature, as is the case
for the other techniques mentioned here.

The place of vocabulary teaching in the reading lesson

Reading classes should principally be for reading and not language
work. We recognise there is a paradox that enhanced vocabulary
will aid reading, so where should it occur? A conventional solu­
tion is offered by Paran in his Burlington Proficiency Series books.
He saves the first reading of text for comprehension activities.
Towards the end of a reading unit, after the second reading, he
considers it legitimate to focus on linguistic features such as lexis
or cohesion as well as further away activities such as appreciation
or evaluation.
200 Reading in a Second Language
For further reading on vocabulary in EFL the reader is referred
to Carter and McCarthy (1988), McCarthy (1990) and Nation
(1990, 1997).
Grammatical skills
Surprisingly in the vast literature on reading there is relatively
little on the relationship between grammatical knowledge and
reading ability and we will make some suggestions in the next
section on research as to how this might be rectified (see Section
5.2, p. 255). Understanding syntax can help the L2 readers to
comprehend text more readily. Increasing syntactic knowledge
may help them to deal with more complex sentences and increas­
ing their automaticity in recognising syntactic structures should
free up processing time.
Nuttall (1996: 78-99) argues the case for paying close attention
to language, particularly syntax and cohesion in order to interpret
difficult text. She sets this in a wider context of trying first with
top-down strategies to establish meaning and, if this does not
prove sufficient, suggests resorting to the additional information
such as examining the syntax and matching this with top-down
insights to consider differing interpretations. Its problem-solving
nature might make this a candidate for the term strategy rather
than skill.
Nuttall suggests a num ber of fairly complex exercises designed
to help students to
■remove optional elements from sentences systematically until
only the core remains and the bare structure of a sentence is
clear;
■paraphrase optional elements of complex sentences one by one,
and fit them into the whole structure to make sense of them.
Cohesion
Chapman (1979a) found a relation between reading ability and
ability to complete anaphoric relations in a cloze test and con­
cluded that mastery of such textual features is a central factor in
fluent reading and reading comprehension. Weir (1983a) found
that gap-filling tests including items on cohesion, discourse markers
and structure correlated well with general reading comprehension
tests. Mackay et al. (1979) and Cowan (1976) have argued that
 The teaching of reading 201
recognition of conjunctions and other intersentential linguistic
devices is crucial to the information-gathering skills of second
language readers.
Williams (1984) discusses the importance of recognising co­
hesive ties and has suggested teaching materials for this. Nuttall
(1982/1996: 86-98) provides a number of useful exercises for recog­
nising and interpreting cohesive devices and discourse markers.
For exemplification of grammatical work, see Morrow (1980:
63-4), Nuttall (1996: 79-86), Silberstein (1994: 63-6) and Sim and
Laufer-Dvorkin (1982: 59-61). For work on cohesion, see Williams
(1983).
Work on the place of microlinguistic elements, in particular
the role of grammar, has an important role, which has tended to
be neglected in the recent past. In class, we must of course distin­
guish between vocabulary and grammar work, which is aimed at
improving reading, and the use of texts to teach vocabulary and
grammar. Both are important aspects of teaching an L2, but we
have to concentrate on the first. What we turn to now - the recog­
nition of main ideas and important information in a text through
either careful or expeditious reading strategies - is more exclus­
ively a part of reading activities in the classroom.

Individual tasks at the global level

Comprehension instruction: the process of
reading text carefully
In Chapter 3 we considered how we might test some aspects of
reading a text carefully. In testing we felt constrained to limit
ourselves to careful reading for explicitly stated main ideas and to
propositional inferencing. We felt that pragmatic inferencing was
beyond the remit of the tester as it goes beyond text-based com­
prehension. However, a strong case can be made for the value of
the latter in the classroom.
We briefly expand on our earlier descriptions of these three
aspects of careful reading below as they are all criterial for de­
veloping reading ability in the classroom. We describe them in
accordance with a taxonomy developed in the reading research
group of the TEU at CALS Reading University. Practice in these
activities should form the backbone of any careful reading pro­
gramme and should be particularly effective when supplemented
202 Reading in a Second Language
by activities for activating background knowledge and summarisa­
tion skills discussed in the following section. We then focus on text
selection and a number of activities which we see as important for
promoting attention to comprehension instruction in the reading
classroom.

Reading carefully for explicitly stated main ideas

Careful and thorough reading of text for explicitly stated main
ideas and important information is an important purpose for read­
ing. We often need to decode the whole of a text to understand it
or to establish its macrostructure. In this mode the reader has to
read a text at a careful rate from beginning to end in a linear and
sequential fashion with regressions as necessary. This will mainly
be a bottom-up sequential process with some limited top-down
processing, and might involve:
■separating explicitly stated main ideas from supporting detail by
recognising topic sentences or by recognising lexical indicators
of importance
■generating a representation of the text as a whole
■understanding the development of an argument an d/or logical
organisation.
Further exemplification is provided in Glendinning and Holmstrom
(1992: 52-3), Grellet (1981) and Nuttall (1996: ch. 9).

Reading carefully for implicitly stated main ideas

In some texts the ideas may not be explicitly stated and students
can be alerted to the nature of propositional inferences. These
are made when the reader uses explicit statements in the text to
form an inference without recourse to knowledge from outside
the text (Chikalanga, 1990). This might involve making
■propositional informational inferences which are either refer­
ential, typically answering questions beginning with what and
which, or spatiotemporal, typically answering questions begin­
ning with where and when;
■propositional explanatory inferences which are concerned with
motivation, cause, consequence and enablement and will often
answer questions beginning with why and how.
 The teaching of reading 203
All the information required to make such propositional inferences
is recoverable from the text. Readers’ activities might include
■discovering writer’s intention
■understanding writer’s attitude to the topic
■identifying the addressee
■distinguishing fact from fiction.
For exemplification see Ellis and Tomlinson (1988: 71-2, 86-8),
Glendinning and Holmstrom (1992: 54-6), Grellet (1981: 98, 239-
44) and Paran (1991: 3-4, 44).

Inferring pragmatic meaning related to a text

Pragmatic inferencing takes place when readers rely mainly on their
own schemata an d/or opinions to interpret a text (Chikalanga,
1992). This might involve making
■pragmatic informational inferences which are either referential,
typically answering questions beginning with what and which, or
spatiotemporal, typically answering questions beginning with
where and when;
■pragmatic explanatory inferences which are concerned with
motivation, cause, consequence and enablement and will often
answer questions beginning with why and how;
■pragmatic evaluative inferences where the reader makes an evalu­
ation on the basis of the content of a text
- applying the main idea(s) in the text into other contexts
- evaluating a point of view
- expressing own opinion on the subject.
With reference to their own background knowledge and experi­
ence, the readers would try to interpret, respond to, evaluate and
possibly apply the writer’s message (s) contained in the text.
A recent story had the headline ‘Who baas wins’. It was a story
of how sheep had crossed a cattle grid (a device for preventing
animals from crossing, consisting of horizontal metal bars with
gaps between) by one of the flock lying across it and the rest
walking over it to greener pastures. The title would be unintelligible
to anyone who did not know the motto of the Special Armed
Services (SAS) ‘who dares wins’.
204 Reading in a Second Language
For further exemplification see Glendinning and Holms trom
(1992: 25-7), Grellet (1981: 28-9, 34-8, 41-2, 45-6, 245-9) and
McGovern et al. (1994: 12-13).

Promoting careful reading skills in the ESOL classroom

First, we need to avoid specially written texts constructed to illus­
trate specific language points. The reading text is frequently seen
as the conduit for lexical and syntactic learning rather than the
source of new content information for the student.

Text selection
Williams (1984: 15) discusses the shortcomings of the types of text
used solely for learning language and sees, as its key failing:
There is litde attention to reading as a skill in its own right that
might need to be developed in different ways for different purposes.
He concludes (p. 125):
. . . although it is very tempting to use written text as a basis for
the learning and teaching of language, an approach that goes no
further not only neglects reading as a skill but also neglects the
ultimate purpose of learning a language which must surely be to
use that language. Being able to read skilfully and flexibly is an
important use of language.
Nuttall (1996: 30) argues in a similar vein for focusing on using
texts to convey meaning rather than as a convenient vehicle for
conveying language:
partly because this is often neglected in the language classroom,
partly because treating texts as if they meant something is more
effective in motivating students and promoting learning.
She offers a number of criteria for text selection including:
■suitability of con :ent: it is essential that text should interest the
reader;
■exploitability: facilitation of learning. How well can it help de­
velop reading ability; this is not a language lesson or a content
lesson but rather ‘how language is used in conveying content
for a purpose’ (p. 172).
 The teaching of reading 205
Appropriate texts, for example in terms of:
■intended audience
■intended purpose
■source
■length
■lexical range
■rhetorical structure
■topic familiarity
■relationship to background knowledge
■channel of presentation
should be chosen to enable students to practise careful reading.
The same selection criteria apply to the other strategies and skills
we discuss below. It seems that careful reading can accommodate
implicit text structure and ideas, whereas expeditious reading is
more dependent on explicitness in text structure or ideas. This is
what Hamp-Lyons (1984) called a ‘text strategic approach’. The
focus here is on exploiting generalisable features of text ‘in order
to help learners develop skills for approaching any text’.
This is an area in critical need of attention by reading teachers.
What are the salient features of text selection which will facilitate
selecting texts to best practise appropriate activities? What is a
principled set of procedures to determine whether texts appropri­
ate in terms of the above conditions actually allow the practising
of intended activities, purposes for reading. An attempt to draw
up a specification for text selection for an advanced reading test
in China is presented in Appendix 1. The same categories of
description are applicable to text selection for teaching purposes.
In addition, in Section 5.3 a text-mapping procedure is described
which offers a systematic method for the development of tasks
once a text appropriate for the intended purposes of reading has
been identified.
It is also important that students are exposed to the range
of materials they might later have to cope with for either in­
formational or entertainment purposes. For example, it is no
use basing EAP reading materials solely on texts taken from
newspapers, though obviously the introduction of target texts
will only occur when it is appropriate to do so both in terms of
background knowledge and linguistic readiness. Hamp-Lyons
(1984: 308) cautions:
206 Reading in a Second Language
. . . our readings in schema theory. . . convinced us of the need to
choose texts which the students would be easily able to integrate
with their own prior experience and knowledge of the world.
Authenticity was also discussed in Chapter 3 with regard to test­
ing. Nuttall (1996: 177) argues:
To pursue the crucial text attack skill we need texts which exhibit
the characteristics of true discourse: having something to say, being
coherent and clearly organised. Composed (i.e. specially written)
or simplified texts do not always have these qualities.
This is not to say that texts may not be modified with due caution
(see Lewkowicz, 1996). For example, difficult words can be substi­
tuted or complex syntax unravelled. Williams and Moran (1993:
66) note that the claims for authenticity are not taken as literally
as they once were and simplified or specially written texts have a
place in the reading course books they reviewed. Lewkowicz (1997)
makes the point that as long as salient performance conditions,
e.g. appropriate rhetorical structure, are present full authenticity
may not be essential in the texts employed for teaching or testing
specified skills and strategies.
Williams (1984: 18-19) makes a number of points about the
linguistic difficulty of the text selected:
... it should not contain a large amount of language that is too
difficult for most of the class ... if too difficult, then either the
pace of the lesson will be slow, and boredom will set in, or the pace
will be too fast, and the learner will not understand enough, and
frustration will result.
Nuttall (1996: 174-6) deals with this under the heading of read­
ability (see also discussion of text difficulty in Section 3.3). She
sees it as a combination of structural and lexical difficulty though
recognising the influence of conceptual difficulty and interest.
Texts selected should take the level of the students in terms of
vocabulary and structure into account. In multi-level classes, self-
access work at different levels may be an essential supplement if
the provision of differentiated reading materials is not available
for regular classroom instruction (p. 174).
Nuttall (1996: 36) talks of the ‘next step’ level, i.e. one step
further than where the student currently is, but no more, as the
target for pushing them on. The teacher provides ‘scaffolding’ to
help them take this extra step. Nuttall describes this as never
 The teaching of reading 207
doing anything for them that they can do themselves with a little
support. This is discussed by Williams and Burden (1997: 65-6) as
the zone of proximal development from the field of educational
psychology:
... it suggests that the teacher should set tasks that are at a level
just beyond that at which the learners are currently capable of
functioning, and teach principles that will enable them to make
the next step unassisted. Bruner and others have used the term
‘laddering’ to refer to this process.
Williams (1984: 34) also advocates using a range of materials,
selecting texts ‘that deal with the same topic or theme, since this
will result in consolidation and extension of language and lan­
guage use in a way that is comprehensible to the general learner’.
Vygotsky (1962: 78) is one of the earliest writers to deal with
mediation in the sense of using tools to achieve goals, and his
work previews much of the current discussion in this area.
If learners can choose their own texts, this is likely to be highly
motivating; but in those cases where textbooks are prescribed, this
may not be possible, and how the teacher uses texts becomes
crucial. Walker (1987) offers a proforma set of activities for stu­
dents who bring their own texts to the classroom so that even
though the instruction is individualised in terms of text the activ­
ities being practised are common.
In addition to work focusing on the careful reading activities
detailed at the start of this section, there are a num ber of other
interventions which should help to ground these skills properly.

Interventions designed to build or activate

background knozvledge
The role of background knowledge in reading was discussed in
Section 2.2 and its relationship to testing was considered in Sec­
tions 3.1 and 3.3. Whereas in testing we argued for trying to min­
imise its influence, the reverse might be said to be true in teaching.
In the section on metacognitive strategies above we considered ways
in which the students might consider the relationship between
their existing state of knowledge and information to be found in a
text at the pre-reading, while-reading and post-reading stages. Here
we look more closely at how understanding the structure of a text
can aid main ideas comprehension in careful reading mode.
208 Reading in a Second Language
We focus on two activities in the classroom, understanding text
organisation and writing summaries of main ideas, which appear to be
particularly beneficial for successful careful reading.

Text organisation
Students with varied profiles appear to benefit when teachers help
them activate or build formal knowledge of text structure: struc­
tural relations between main ideas in a text. Pearson and Fielding
(1991) provide examples of story structure and expository text
structure instruction. They describe ways of activating knowledge
of the structure of the text itself, e.g. of a story grammar. This
might involve consideration of its abstract hierarchical structure:
setting, problem, goal, action, outcome; and giving practice in
identifying category relevant information.
Comprehension, particularly inferential comprehension, is also
helped when connections are made between readers’ background
knowledge and experience and the content of the text under
review (Pearson and Fielding, 1991: 847). This may happen prior
to reading. Invoking knowledge structures aids comprehension.
Making predictions before reading and confirming them during
reading, and asking inference questions during and after reading,
improves comprehension - particularly inferential comprehension
(ibid.).
As well as formal knowledge of text structure, knowledge about
specific topics and themes related to a story is important. The
role of pre-reading discussions to generate expectations in this
respect has been shown to be effective (Pearson and Fielding,
1991: 822). Other methods include using writing to anticipate
story information and developing a short list of key words, and
such a cognitive engagement has been found to help poor readers
(p. 823).
Longer texts, or a number of texts on the same theme, are seen
by Williams and Moran (1993) as another way in which authors
have tried to build up background knowledge in a certain area
(see Haines, 1987, and Tomlinson and Ellis, 1988, for examples
of these).
Additionally, inferential questions and prediction questions - a
focus on important ideas (central events in a story), on construct­
ing an interpretation and on summary - are seen by these reviewers
as useful techniques for improving the understanding of a story.
 The teaching of reading 209
As far as expository text structure is concerned, it has been
suggested that visual summary is a useful tool. What was said earl­
ier in Sections 2.2 and 3.3 about the organisational structure of a
text is relevant here. The importance of summary (see below),
schematic representation of a text, and rating the importance of
ideas related to the text to comprehension, learning and remem­
bering, are noted by Pearson and Fielding (1991: 827) as they
promote attention to text structure. The effectiveness of teaching
students to use text structure to identify main ideas is confirmed
by Alvermann and Moore (1991: 960), though they point out that
students’ familiarity with the topic appears to mediate instructional
effectiveness.
It is clear that readers who are knowledgeable about, and who
can follow the author’s text structure, recall more of a text than
those who lack these attributes (Pearson and Fielding, 1991: 827)
and they note that more good readers than poor follow the writer’s
structure in recall of texts. Hierarchical summaries using discourse
clues and visual representations (networking, flowcharting, con­
ceptual frames) are also seen as useful in helping recall text infor­
mation better and improving comprehension, particularly for lower
ability students who need more help in developing strategies.
Nuttall (1996: ch. 12) offers a variety of information transfer task
examples that might be used in the reading class. Also the section
on information transfer in Section 3.4 on testing offers advice on
the use of this task type.
Pearson and Fielding (1991: 832) conclude:
It appears that any sort of systematic attention to clues that reveal
how authors attempt to relate ideas to one another or any sort of
systematic attempt to impose structure upon a text, especially in
some sort of visual representation of the relationships among key
ideas, facilitates comprehension as well as both short term and
long term memory for the text.
It appears that while most strategies are of value across the ability
range:
. . . the more able readers benefit the m ost. . . (but) regardless of
ability level, the teaching strategies have their greatest effect when
students are actively involved in manipulating conceptual relation­
ships and integrating new information with old knowledge.
(Alvermann and Moore, 1991: 960)
210 Reading in a Second Language
Nuttall (1996: 100-24) provides useful advice and sound practical
exemplification of a range of text attack skills: recognising func­
tional value of sentences; recognising text organisation; recognis­
ing presuppositions underlying a text; recognising implications;
and making inferences. A particularly interesting example is where
students are given parts of a chapter or of a text and they have to
put the parts in the right order. This is best done in groups. It
involves an integration of many of the skills and strategies dis­
cussed in this chapter.
Creating text diagrams to illustrate the way ideas and informa­
tion are presented in a text is probably best done by the students
working in groups with classroom discussion later. Not all texts
lend themselves to this technique, so input texts need to be cho­
sen carefully. The section on mindmapping in Section 5.3 offers
some insights into how students might go about this process and
learn something more about skills and strategies at the same time.
Nuttall notes of text diagrams that
Their great advantage which outweighs the disadvantages ... is that
they demand close study of the way the text is put together and
promote text focused discussion. They are useful either to display
common patterns of paragraph organisation or to elucidate the
structure of complex text. (p. 109)
Useful discussion of networking can be found in Danserau et al.
(1979); flowcharting in Geva (1980, 1983) and for work on top
level rhetorical structures see Meyer (1975) and Bartlett (1978),
who show how, through diagram, ideas and their relationships are
represented within the text. Williams (1984) provides a useful
basic survey of text structure and some ways of introducing it to
students.
Careful reading into writing,: a product from
the reading process
Summarisation is perhaps the verbal equivalent of the visual dia­
grammatic representation of text structure discussed in the previ­
ous section, which could easily be subsumed under the broad
umbrella of summary. In contrast to earlier work on summarisation,
where research results were confounded by the use of low-level
multiple-choice items as criterion scores, Pearson and Fielding cite
positive support for summarising including improved comprehen­
sion on the texts involved, increased recall and even improvement
 The teaching of reading 211
on standardised reading test scores by students involved in this
activity (1991: 833). There is also evidence that summarisation
training transfers to new texts. They argue for its value as a broad-
based comprehension training strategy.
Students understand and remember ideas better when they have to
transform those ideas from one form to another. Apparently it is
in this transformation process that author’s ideas become reader’s
ideas, rendering them more memorable. (p. 847)
In Chapter 3 we discussed the aim in testing of measuring reading
unmuddied by the contaminating influence of other variables,
e.g. writing. Measurement considerations such as this do not loom
as large in the teaching situation. A good case can be made for
the fruitful interaction between reading and writing in the lan­
guage classroom, both activities being seen as potentially comple­
mentary to each other. Zamel (1992) argues that reading and
writing instruction benefit each other in an integrated approach
and argues for ‘writing one’s way into reading’. Silberstein (1994:
70-1) argues that by integrating instruction students come to un­
derstand the way in which both readers and writers compose text.
Smith (1988: 277) comments that ‘writing is one way of promot­
ing engagement with a text which leads to better comprehension’.
The student has to establish the main ideas in a text, extract
them and reduce to note form and then rewrite the notes in a
coherent m anner in their own words. Brown and Day (1983) iden­
tified a num ber of rules for summarising which match the rules
of Kintsch and van Dijk for establishing macropropositions (see
Chapter 2, p. 80):
■delete trivial information
■delete redundant information
■provide a superordinate term for members of a category
■find and use any main ideas you can
■create your own main ideas when missing from the text.
Pearson and Fielding (1991: 834-5) report that such training
enhanced summarisation and increased scores on reading tests
when compared with control groups. Exemplification of summary
tasks can be found in Grellet (1981: 233-6), Paran (1991: 5, 27,
41) and Trzeciak and Mackay (1994: 26-8, 33-55).
We expressed concern about what actually happened as regards
reading in the L2 classroom at the start of this section. The evidence
212 Reading in a Second Language
suggests that little attention is devoted even to teaching the skills
and strategies necessary for successful careful reading for global
comprehension of text. The situation may be even worse as regards
expeditious reading strategies. Leaving aside the prevalence of short
texts in most course books, less attention is devoted to these strat­
egies in comparison with careful reading.

Teaching expeditious reading skills

We noted in Section 2.4 and Chapter 3 that, in contrast to in­
tensive careful reading, expeditious reading activities such as
skimming, search reading and scanning were often overlooked
in the past, certainly by testers, researchers and reading model
builders. Nuttall (1996: 39) makes a similar point in relation to
teaching materials:
Moreover, longer texts are liable to get forgotten in the classroom,
since it is easier to handle short texts which can be studied in a
lesson or two. But the whole is not just the sum of its parts, and
there are reading strategies which can only be trained by practice
on longer texts. Scanning and skimming, the use of a contents list,
an index and similar apparatus are obvious ones. More complex
and arguably more important are the ability to discern relation­
ships between the various parts of a longer text, the contribution
made by each to the plot or the argument, the accumulating evid­
ence of a writer’s point of view, and so on.
and devotes an entire chapter to efficient reading in the 1996
revised edition of her book on the teaching of reading skills.
Williams and Moran (1993) make a similar point in relation to
the length of reading texts in course books.
Hamp-Lyons (1984: 305) notes the paucity of materials avail­
able in teaching at the macro-level in English for academic pur­
poses reading. Needs analyses have made it clear (see Weir, 1983a)
that one of the biggest problems for overseas students studying in
a second language context is the sheer volume of reading that is
required (see also Hamp-Lyons, 1984; Robb and Susser, 1989).
Thus, as well as needing to read texts carefully to understand all
in the text, they also have to read expeditiously to extract main
ideas quickly or to establish whether it is relevant to their pur­
poses (see Sections 2.4 and 3.2 for earlier discussion of these
expeditious strategies, and also Nuttall, 1996: 39, 44).
 The teaching of reading 213
Previewing and predicting were dealt with earlier in this part of
the book in the section on generic metacognitive activities and
practices. We addressed the expeditious strategies of skimming,
scanning and search reading in Chapter 3 and the tasks used to
test these are equally suitable as teaching tasks though, in the latter
case, they may be mediated by teacher or students co-operating
on them and their results put to a different use.
These strategies were introduced in Chapter 2 and the reader
is asked to forgive a limited amount of repetition necessary to
contextualise the taxonomy under development in the TEU at
CALS Reading University (see Appendix 1). In this section we will
expand our definitions of these strategies by exploring the pur­
poses for which they might be employed and the actual opera­
tionalisations that might be focused on in classroom activities.
These are the key areas that future research will need to focus on
as such mediation by the teacher may help students to read more
efficiently.

Skimming
This involves processing a text selectively to get the main idea(s)
and the discourse topic as efficiently as possible, which might
involve both expeditious and careful reading and both bottom-up
and top-down processing. The focus may be global or local and
the rate of reading is likely to be rapid, but with some care. The
text is processed quickly to locate important information which
then may be read more carefully. Purposes for using this strategy
might include:
■to establish a general sense of the text
■to quickly establish a macropropositional structure as an outline
summary
■to decide the relevance of texts to established needs.
Where appropriate to text type it might involve one or more of
the following operationalisations:
■identifying the source
■reading titles and subtitles
■reading the abstract carefully
■reading the introductory and concluding paragraphs carefully
■reading the first and last sentence of each paragraph carefully
214 Reading in a Second Language
■identifying discourse markers
■noting repeated key content words
■identifying markers of importance
■skipping clusters of detail
■glancing at any non-verbal information.
Readers would be taught to be flexible as not all strategies would
work with all texts. Also, some attention might usefully be paid to
metacognitive strategies discussed above such as prediction and
monitoring; the former to facilitate the use of existing knowledge,
the latter to help separate less important detail from main ideas.
Practical exemplification can be found in Grellet (1981: 71, 73-5,
81-2), Paran (1991: 79-81) and van Dijk (1977: 79).

Search reading
This differs from skimming in that the purpose is to locate informa­
tion on predetermined topic (s), for example, in selective reading
for writing purposes. It is often an essential strategy for complet­
ing written assignments.
The process, like skimming, is rapid and selective and is likely
to involve careful reading once the relevant information has been
located. Like skimming, bottom-up and top-down processing are
therefore involved. Unlike skimming, sequencing is not always
observed in the processing of the text although it is likely to be
more linear than scanning. The periods of closer attention to the
text tend to be more frequent and longer than in scanning. It
normally goes well beyond the mere matching of words to be
found in scanning activities, and might include the following
operationalisations where appropriate:
■keeping alert for words in the same or related semantic field
(unlike scanning, the precise form of these words is not certain)
■using formal knowledge of text structure for locating information
■using titles and subtitles
■reading abstracts where appropriate
■glancing at words and phrases.
Examples of search reading activities can be found in Ellis and
Tomlinson (1988: 86-7), McGovern et al. (1994: 12), Morrow
(1980: 15, 17, 37, 39) and Paran (1991: 55).
 The teaching of reading 215
Scanning
This involves looking quickly through a text to locate a specific
symbol or group of symbols, e.g. a particular word, phrase, name,
figure or date. The focus here is on local comprehension and
most of the text will be ignored. The rate of reading is rapid and
sequencing is not usually observed. It is surface level rather than
deep processing of text and is mainly reader-driven processing.
There is a rapid inspection of text with occasional closer inspec­
tion. Pugh (1978: 53) describes it as:
finding a match between what is sought and what is given in a text,
very little information processed for long term retention or even
for immediate understanding.
The operationalisations involved might include looking for/
matching
■specific words/phrase
■figures/percentages
■dates of particular events
■specific items in an index/directory.
The Crescent Series, designed for use in the school systems in
the Middle East, offers a useful general procedure for scanning
(O’Neill et al., 1996; Teacher’s Book 8, xxii). For further exempli­
fication of scanning, see Grellet (1981: 83), Morrow (1980: 18)
and Nuttall (1996: 49-51). Nuttall (pp. 51-3) also provides some
interesting ideas and examples on how graphic conventions -
print size and style, layout, spacing, indentation - help the reader
navigate a text and sometimes can signal, e.g. through different
type faces, how the text is structured.

Extensive reading
The distinction we have been drawing between careful and expe­
ditious reading can easily be confused with another, earlier, dis­
tinction between intensive and extensive reading. While there is
undoubtedly an overlap, there are significant differences between
the two dichotomies. The careful/expeditious distinction, taken
together with the distinction between local and global, results in a
num ber of different reading styles, or strategies, which can be em­
ployed either alone or, more often, in conjunction to accomplish
216 Reading in a Second Language
a range of reading tasks. While expeditious reading is likely to be
directed at lengthy texts, there is no reason why careful reading
must be restricted to short texts. In fact, there are cogent reasons
in academic contexts as to why it should not be.
The intensive/extensive distinction, on the other hand, is largely
a pedagogical construct. Bright and McGregor (1970), for ex­
ample, see them as being distinguished in terms of the num ber of
questions the teacher decides to ask about a text:
For the sake of convenience we shall discuss and exemplify exten­
sive and intensive reading as though they were opposites. This will,
however, be misleading unless we think of them as lying at opposite
ends of a scale determined by question density. The point on the
scale at which we decide to work will depend on:
(i) how much there is in the passage waiting to be discovered.
Not all passages are worth meticulous attention.
(ii) how much time is available. By no means all the passages
worth serious attention can be tackled.
(iii) how much the class is capable of seeing and how well they
respond.
(iv) how much is essential to a minimum worth-while response.
(v) how hot the afternoon is - and so on.
(Bright and McGregor, 1970: 65)
Bright and McGregor (p. 80) remark that
... it is not whole lessons but parts of lessons that may properly be
so divided. In the middle of a chapter, we may stop to dwell on one
word. This is intensive study.
However, our experience in a wide range of countries suggests
that the distinction has become fossilised, with intensive reading
being confined to the classroom, where it involves the teacher
asking a large number of questions about a short text, while extensive
reading refers to either ‘silent reading’ in the classroom, or reading
done unsupervised in the library or at home, the aim being pleasure
or practice, or both.
Hafiz and Tudor (1989: 1-2) see the goal of this type of extens­
ive reading as ‘to “flood” learners with large quantities of L2 input
with few or possibly no specific tasks to perform on this material’.
Nuttall (1996: 127) describes it as ‘the private world of reading
for our own interest’ and offers some valuable suggestions for
organising such activities. She argues that reading extensively is
 The teaching of reading 217
the easiest and most effective way to improve reading and it is
easier to teach in a climate where people enjoy the activity as well
as value it for pragmatic reasons.
Davis (1995: 329) defines an extensive reading programme
(ERP) as:
... a supplementary class library scheme, attached to an English
course, in which pupils are given the time, encouragement, and
materials to read pleasurably, at their own level, as many books as
they can, without the pressures of testing or marks. Thus, pupils
are only competing against themselves, and it is up to the teacher
to provide the motivation and monitoring to ensure that the max­
imum number of books is being read in the time available.
Williams (1984: 10) sees extensive reading as the ‘relatively
rapid reading of long texts’ (see Hedge, 1985) and emphasises
that it should normally be at the level of the student’s reading or
below it. This contrasts with careful intensive reading where the
aim is often to stretch the student slightly.
Bamford (1984: 218) claims that ‘for all but advanced learners,
the best way to promote extensive reading is by graded readers’.

Graded readers
In terms of contributions to the teaching of reading in this cen­
tury, Howatt (1984: 245) singles out the work of West in Bengal in
the 1920s who argued against the prevailing orthodoxy for the
greater surrender value of basic literacy as against training in spo­
ken language. West developed a system of readers using the prin­
ciples of lexical selection and lexical distribution, the latter giving
the reader more practice material between the introduction of
new words - a distinct problem with earlier primers which intro­
duced too many new words too quickly.
Hill and Reid-Thomas (1988a: 44) describe a graded reader as
follows:
A graded reader may be either a simplified version of an original
work or a ‘simple original’, i.e. an original work written in simple
English. In either case it is written to a grading scheme which may
be set out in terms of vocabulary, sentence structure, and, in some
cases, content.
Nuttall makes the distinction between needing to read, which
can be instigated in the classroom, and wanting to read (1996:
218 Reading in a Second Language
130), which is a greater incentive to more people. She provides
helpful advice on how to promote this, including choosing appro­
priate suitable books at the right level, short, appealing, varied
and easy; on how to organise a library (pp. 133-41); on how to
organise an extensive reading programme (pp. 141-4) by creating
interest, developing incentives to read; and devising appropriate
monitoring and assessment procedures.
Survey reviews by the Edinburgh Project on Extensive Reading
(EPER) staff in E LT Journal (Hill, 1992, Hill and Reid Thomas,
1988a, 1988b, 1989, 1993) give advice on graded readers in terms
of levels, readability level, appearance, text subject matter, aids to
reading, recommended reader age, and a quality rating on a scale
of 1 to 5 (an interest rating in the 1988 reviews). Hill and Reid-
Thomas (1989) note a trend to shorter books and advise publishers
that these books may be too short and possess insufficient meat
to be used as class readers. They make a plea for some longer
readers (more than 72 pages at the upper level and more than 40
at the lower).
Davis (1995: 331-5) provides some useful advice for develop­
m ent of extensive reading programmes:
■the watchwords are quantity and variety, rather than quality
■books should be attractive and relevant to students’ lives
■books should be more than sufficient in number
■books should be graded and colour coded by reading level but
students should be encouraged to move between levels
■try to make ERP school policy
■try to get it built into the timetable
■try to get financial support at least for a book box
■integrate with library studies where appropriate, for example,
through parallel grading of fiction books in the library
■develop a simple system for using the books
■develop a quick and painless system for monitoring the use of
the books.
Very similar advice can be found in Bright and McGregor (1970).

The class reader: sustained silent reading

A similar but slightly different form of extensive reading is the
silent sustained reading which may take place in class with a class
set of books. Nuttall (1996: 145-8) describes how to select class
 The teaching of reading 219
readers and outlines a number of useful activities that can be
done in class and outside class. What is clear is that it is fine if
the students enjoy such extensive reading; however, if they do not
enjoy a particular story, drop it immediately.
Davis (1995) suggests that the class reader may be seen by
administrators as a cheaper alternative to ERPs and one easier to
handle in the classroom.

The teacher’s role

Williams (1984: 19) points out that extensive reading in class is
important in giving learners uninterrupted quality time to read
and reflect and, just as important, it gives teachers tim e/breath­
ing space to reflect on their teaching. He comments: ‘Teachers
should not feel that they are neglecting their job if they are not
constantly explaining something or organising classroom activities.’

Does extensive reading work?

In common with the rest of this chapter, we are concerned that
any advice we give should where possible be supported by evid­
ence of positive impact. Surprisingly little negative comment
appears in the literature concerning extensive reading. On the
downside are:
■the cost of extended reading programmes
■the amount of time setting them up and running them efficiently
■the fact that curriculum time is required for private reading.
Nuttall (1982: 62) asserts that
an extensive reading programme ... is the single most effective way
of improving vocabulary and reading skills in general. You can
organise a programme of vocabulary building, but greatly increas­
ing the amount of material they read is the best way of all.
She repeats the same in the 1996 revised edition of her book but
limits the effect to improving vocabulary in the revised edition.
Rodrigo (1995) argues that there is some evidence for enhanced
progress in reading comprehension and vocabulary development
as a result of extended reading courses (see also Day et al., 1991,
and Pitts et al., 1989). This is a view supported by Krashen (1993)
220 Reading in a Second Language
for whom the dominant mode of language learning is acquisition.
Hafiz and Tudor (1989: 5) argue that comprehensible input in
large quantities in a tension-free environment satisfies a key con­
dition for acquisition to take place. Extensive reading programmes
can provide large amounts of reading in L2 for personal interest
reading. They quote the claim by Krashen and Terrell (1983) that
comprehensible input gained in reading may contribute to a gen­
eral language competence that underlies both spoken and written
performance.
In Rodrigo’s study students were encouraged to read as much
as they could that was comprehensible during a semester, with a
chance to choose what interested them and focus on meaning
rather than linguistic form. Certain readings were done in class,
often in the library, and others were personal readings. As well as
the experimental reading course there was a discussion group
where readings were discussed and presented orally. The control
group focused on grammar and vocabulary. The reading material
was divided into three levels:
■Graded books: ‘edited, simplified an d/or abridged so that they
can be used at certain levels . . . excellent for competence and
confidence building’ (Rodrigo, 1995: 6).
■Light reading or easy ungraded native reading: ‘unabridged not
classical literature or technical’.
■Literature or information texts: ‘more complex unabridged read­
ing material, such as literature, adult books, bibliography, his­
tory and technical and informational books’ for those who had
succeeded at the stage of light reading.
Rodrigo (1995) compared the two groups using performance on
a vocabulary recognition test and the experimental group did
outperform the control, but the results were not significant, prob­
ably because of low sample size. Qualitative feedback from the
students clearly indicated that they had enjoyed the extensive read­
ing course.
Robb and Susser (1989) report on the value of extensive read­
ing as against a skills focus in experimental work they conducted
in Japan. Elley and Mangubhai (1983) offer evidence of the im­
pact on reading of an ERP used in Fijian primary schools. They
showed that subjects receiving extensive reading improved mean­
ingfully in reading and word recognition at the end of the first
 The teaching of reading 221
year of the study. Hafiz and Tudor (1989) report how an extens­
ive reading programme using graded readers in a control group
experimental design resulted in a statistically significant improve­
ment in the experimental (the extensive) group, especially in terms
of their writing skills. The improvements in reading scores, though
significant, were small.
Hamp-Lyons (1985) offers some evidence that students taught
by a text-strategic approach evidenced higher gains in a pre-test/
post-test design than students who followed a traditional approach.
Nation (1997) cites empirical evidence to show that there are many
benefits for extensive reading ‘in quality of language use, language
knowledge and general academic success’.
The value of extensive reading receives some support in the
literature (Davis, 1995; Elley, 1991; Hamp-Lyons, 1985; Krashen,
1993; Nation, 1997; Robb and Susser, 1989; Tudor and Hafiz,
1989). However much more research needs to be done to provide
a firm empirical base to convince a wider audience of its value.
Lunzer et al.’s massive LI study (1979) quoted in Nuttall (1996)
showed that in British classrooms right across the curriculum little
sustained reading occurred and it accounted for less than 15 per
cent of class time. It involved more writing than reading work.
Nuttall (p. 128) suggests that the situation had not improved by
1989 according to inspectorate reports. Williams and Moran (1993)
comment:
. . . the rather curious situation has arisen, whereby, despite univer­
sal acceptance of the view that one becomes a good reader through
reading, reading lessons where most time is actually spent on
reading (as opposed to discussion, answering questions, etc.) are
relatively rare.
It is clear that what is important is how much learners read and
the degree of enjoyment derived. The challenge is to access stu­
dents to such literature and provide the m eans/incentives/oppor-
tunities for them to read it and to carry on reading.
So from the research we have a picture of what might be useful
in terms of reading strategies. How teachers take this and make
use of it in the classroom is the challenge for the next decade.
What is clear from the research is that these strategies as yet do
not play a large part in th j LI reading classrooms in the USA
(Alvermann and Moore, 1991: 974). The picture is brighter, how­
ever, in the L2 situation in the UK
222 Reading in a Second Language
Working together
We have indicated above that an obvious difference between
testing and teaching is that the teaching situation allows for co­
operation between students, whereas testing virtually never does
(not officially, at least). We also referred to a survey of US second­
ary classrooms by Alvermann and Moore (1991) which found little
group work in evidence. By contrast, we next examine some ways
in which students have been encouraged to work co-operatively.

Peer interaction
Research supports the view that working together co-operatively
benefits all levels of students in mixed ability groups or pairs. It
may be used as follow up to teacher-directed activities (Pearson
and Fielding, 1991: 839). They comment:
In general, successful groups work towards group goals while mon­
itoring the success of each individual’s learning as a criterion of
group success; also associated with positive growth are peer interac­
tions that emphasise offering explanations rather than right answers.
It might, however, be the novelty of working with each other,
rather than the activities engendered, that has produced the results.
Nuttall (1996: 161-6) offers advice on how students might be
guided during the reading process itself, and while emphasising
the value of individuals reading in silence on their own (this is
what reading actually is), perhaps in a self-access system, she points
to some advantages of working in a group: motivation, individuals
participate more actively because it is less threatening than whole
class activity and partly because of reciprocity, the recognition that
everyone in the group should contribute. It all makes students
aware how others read and promotes thoughtful discussion on
reading strategies and skills.
What is clear is that motivation has a very strong influence on
strategy use. Williams and Burden (1997: 154) argue ‘increased
motivation and self-esteem lead to more effective use of appropri­
ate strategies and vice versa’. Motivation appears to be enhanced
by co-operative learning experiences (Roehler and Duffy, 1991:
867).
Jigsaw reading is a good example of the more innovative devel­
opments that have taken place in the EFL classroom (see Grellet,
 The teaching of reading 223
1981; Geddes and Sturtridge, 1982; and Nuttall, 1996: 209 and
257). The class is split into groups and given only partial informa­
tion on a topic situation or story. The groups are then reorgan­
ised so each member has a different piece of information from
which the new group has to reassemble the whole. Unless you
have information from all of the texts you cannot understand
something im portant in the story or situation or perform a key
task. Williams (1984: 115-11) provides some good examples of
this technique and also of what he terms enquiry strategy, where the
groups decide what information they would like to find out as a
pre-reading task.

Grouping students for reading instruction: does it work?

For a long time it has been accepted wisdom to group students on
the basis of ability for reading classes. Barr and Dreeben (1991),
however, on the basis of some recent LI studies, cast some doubt
on this: according to them, in general the results are ‘equivocal
and inconsistent’ (p. 895). The authors feel that the research in
this area has neglected the properties of ability groups, the prin­
ciples underlying their formation, the nature of instruction received
and the connection between grouping and other aspects of the
educational enterprise. The consequence has been that studies
have produced results which are inconclusive, inconsistent, and
not particularly interesting. They argue that we need to focus on
how knowledge is imparted through instruction to various popu­
lations of students, e.g. what the nature is of tasks and instruction
provided to similar and different groups or tracks across schools;
how these differences influence attitudes and learning.
Nuttall (1996: 164-5) offers sound practical advice for this
activity. She suggests that groups should not be more than five in
number; non-participators might be put together; members of the
group should face each other; mixing or streaming groups can be
varied from activity to activity; the teacher should take a consultant’s
role and not interfere; tasks must be explicit as only one group can
be attended to at a time; tasks should aim to involve all members
of the group.
We have now looked at texts, goals, activities, and roles of stu­
dents in the learning to read process. We finish by considering
what should happen in the reading classroom in general and what
actually happens.
224 Reading in a Second Language
4.4 WHAT SHOULD HAPPEN IN THE
READING CLASSROOM?
What actually happens in the reading class? Alvermann and Moore
(1991: 964) detail the more usual LI classroom practices in Amer­
ican schools as lecture, textbook assignment, and classroom recita­
tion. They found that reading activities tended to occur only in
short bursts of about 15 minutes and were connected with other
forms of classroom communication: listening, writing and speak­
ing in a supportive role; ‘continuous reading is rare’ (p. 965).
They found that research suggests considerable variation in practice
across teachers with regard to activities, textbook use and academic
tracks.
In their survey they also discovered the following aspects of the
reading classroom. These are for the most part traditional and
date back to the turn of the century (p. 969):
■predominance of textbooks
■emphasis on factual textual information
■teachers in control of students’ encounters with print; student
initiated comments or questions are rare
■very little work in groups
■little planned reading instruction in terms of teaching skills and
strategies.
They argue that this approach is maintained for institutional rea­
sons such as maintenance of order, accountability, socialisation
(conveying a sanctioned body of knowledge) and resources such
as time (to prepare and deliver and own expertise) and materials
(pp. 979-82). It is an environment that rewards control. In the
end teachers may well teach in the way they have been taught
themselves as this is what they know best.
Bernhardt (1991a: 175-6) presents a similar picture for the L2
EFL reading lesson. Here she suggests that texts are often spe­
cially written to illustrate use of lexis, structures and syntactic fea­
tures rather than for teaching comprehension. The shortness of
the texts employed restrict the reading styles that can be applied
and encourage a focus on lexis and structure. There is usually
some pre-teaching of vocabulary and the reading selection is usu­
ally assigned for homework with a commitment to answering the
comprehension questions set by the author. Follow-up in class
often takes the form of oral reading plus answering the questions.
 The teaching of reading 225
Some additional focus might be given to vocabulary or pronunci­
ation errors. Exercises on grammatical elements of text occur either
before or after the reading text is studied and, similarly, encourage
a microlinguistic focus.
Bernhardt argues that this picture is based firmly on the text­
books which teach teachers how to teach. She notes how little
time or attention is available for teachers on how to teach reading
in these methodology books (1991a: 177).
Hamp-Lyons (1984: 307) notes:
. . . the traditional approach to reading comprehension in ESL class­
rooms, in which students are asked to read short passages very
carefully and expected to understand them in minute detail, being
tested on this comprehension by questions mainly at the lexical
level, reinforces concrete level process strategies ... it also rein­
forces the tendency to read slowly and discourages the develop­
ment of sufficient reading speed to synthesise meaning from the
passage (Smith 1971) . . . ESL readers spend much longer on each
fixation than native language readers do.
Robb and Susser (1989) present a similar picture of the situ­
ation in Japan, with reading largely taught by the translation pro­
cedure so that many are only able to decode at the sentence level
and are denied reading books for pleasure.
Pearson and Fielding (1991: 815) note that there are now a
plethora of reviews of research about LI instruction intentionally
designed to improve reading comprehension. They point out that
much that passes as comprehension instruction is little more than
doing tests or answering questions on passages, with little advice on
how to perform such tasks, i.e. instruction intentionally designed
to improve reading comprehension.
Often it is a case of ‘procedural display’ - getting the lesson
done, rather than substantive engagement in some academic con­
tent (Bloom, 1985, quoted in Bernhardt, 1991a: 181). Hoffman
(1991: 939-41) details a num ber of studies which produced sim­
ilar findings of very little comprehension instruction happening
and also an absence of such instructions in manuals and textbooks
in use.
Bernhardt (1991b) describes how L2 students learn to ‘look
literate’: putting hands up to answer questions, looking at the
book while another reads aloud but (p. 182) ‘they did not learn
how to use or interpret printed discourse for meaning . . . are rarely
226 Reading in a Second Language
asked to display nonteacher-mediated understanding. . She
quotes Duffy and Anderson (1981):
. . . the primary concern of teachers is maintaining the flow of
ongoing classroom activities rather than thinking about students’
instructional needs and adapting lessons accordingly.
Getting through the lesson (Bloom, 1985) and avoiding lesson
breakdown often compel teachers to mediate learning through
conversational interaction, often supplying the answers themselves.
Students who threaten to impede the flow of a lesson, say, by not
understanding, are ignored because they slow it down. Thus,
Bernhardt notes, it is the learners who have the most difficulties
who often receive the least attention. Those who help promote
the successful achievement of the teacher-decided activities receive
the most reward.

The way forward: some general considerations

Drill and practise were common procedures and they still have a
value in the training of automaticised outcomes. This traditional
approach of the teacher giving controlled guidance and then stu­
dents practising the various skills appears to receive support from
high student performance on standardised tests in research stud­
ies (Roehler and Duffy, 1991: 861). However, LI research now
suggests the value of moving beyond the exercise model with its
emphasis on repetition. Roehler and Duffy (1991) point out that
cognitive psychology and information processing have demon­
strated the importance of organisation, coherence and connec­
tedness in transmission of knowledge. To reach long-term memory,
information needs to be transformed ‘into meaningful concepts
that can be referenced and stored in an organised way’ (p. 861).
Secondly has come an understanding of the importance of
metacognition; conscious control of cognitive processing with self­
regulation and a focus on understanding (p. 862). As a result,
drill and practice are no longer seen as adequate and a more
co-operative and cognitively based approach is advocated ‘where
teachers provide information and mediate student mental pro­
cessing’ (p. 863). So, as well as providing information, importance
is also now placed on mediating student learning; helping students
to learn about learning (see Section 4.2 above).
 The teaching of reading 227
The value of explicit instruction on how to comprehend has
long been recognised, but now it is felt that it is the nature and
content of these interactions in the reading classroom that count
(Pearson and Fielding, 1991: 841). Pearson and Fielding’s review
of LI research suggests that activity and involvement on the part
of the students is as important as instruction; shared reciprocity
for interactions and negotiation of meaning rather than mono­
logue are to be encouraged (see Bernhardt, 1991a: 181-9).
Williams and Burden (1997: ch. 4) advocate a social con­
structivist approach and see mediation by the teacher as central to
the learning process; ‘(its) influence is both powerful and pro­
found’ (p. 84). Their focus is on what teachers can do to help their
learners become effective and independent. They see mediation
as being concerned with empowering (p. 68):
. . . with helping learners to acquire the knowledge, skills and strat­
egies they will need in order to progress, to learn more, to tackle
problems ... to meet new and unpredictable demands. It is also
concerned with helping learners to become autonomous, to take
control of their own learning, with the fundamental aim of enab­
ling them to become independent thinkers and problem-solvers.

Will it work?
Where teachers’ messages convey how to construct meaning in
reading, a positive effect has been found. Pearson and Fielding
(1991: 848) detail how this might involve focus on text structure;
encouraging students to connect background knowledge to text
ideas to make inferences, predictions, and elaborations; or prompt
students to ask their own questions about the text. In their wide
review of the research, Pearson and Fielding (1991) found such
interventions to be at least moderately successful although trans­
ferability was not tested for. Hoffman (1991: 942-3) also provides
data which show a positive result on achievement measures for
explicit teacher explanation.
The effective teacher will help students develop an awareness
of reading strategies necessary for successful encounters with text.
Explicit instruction in strategies and skills discussed in Sections
4.2 and 4.3, and in the section on testing, were found by Pearson
and Fielding (1991: 849) to be helpful, particularly where the
focus is on ensuring that students understand when and why they
228 Reading in a Second Language
might be employed (see also Roehler and Duffy, 1991: 867 et seq.
and Paris et al., 1991). Cohen (1998: 19) reports:
The findings of the study would suggest that explicitly describing,
discussing and reinforcing strategies in the classroom - and thus
raising them to the level of conscious awareness - can have a direct
payoff in student outcomes.
Following instruction by questions that help mediate and build
up student understanding was also found to be of value by Roehler
and Duffy (1991: 872) in their review of the LI research literat­
ure. Nuttall (1996: 181-91) also examines the value of question­
ing in the classroom and argues that it provides a window into the
students’ mental processes; especially where answers are wrong,
opportunities arise for learning through ‘thoughtful searching’.
Questions which make the reader work are advocated as these
focus attention on difficult elements of the text, especially where
follow-up questions get students to reveal how they arrived at their
answers. The use of MCQ format may have a place here, especially
if the distracters perform this function and useful discussion may
result. Any questions which promote discussion have a valuable
role to play as they get learners thinking about reading and devel­
oping interpretative strategies (see Paran, 1993, for examples of
these in teaching materials).
Roehler and Duffy (1991: 864-6) outline the importance of
planning by teachers to identify critical features for their students,
to simplify the tasks and create effective examples. The selection
of tasks students are asked to do will constrain the operations
students acquire and how they interpret learning experiences. To
motivate students, tasks should encourage students to engage
in cognitive activity appropriate for the intended outcome and
students should be aware of the purposes behind these activity
structures. Teachers should specify clearly how the learning experi­
ence is intended to be useful, and the expectancies they have of
their students. Hoffman (1991: 923) provides some empirical evid­
ence for the positive effect of the latter on student performance.

Remediation
What happens when they do not learn? Hoffman (1991: 915)
refers to a frequently cited frustration for teachers as dealing with
and meeting the needs of students experiencing difficulty in the
 The teaching of reading 229
reading classroom. He presents data to suggest that the slow pac­
ing in low ability groups ‘does not appear to hold any promise or
pay off in terms of successful reading development’ (p. 936).
The high incidence of teacher correction often at the point of
error in reading aloud in low-ability groups is also seen as debil­
itating and helps create an even wider gap between high- and
low-ability groups (pp. 937-8).
Johnston and Allington (1991: 985-6) feel that the very use of
the term ‘rem ediation’, with its connotation of sickness of the
child, creates an unfortunate role structure for the children tagged
in this way. They suggest that we would look at the situation differ­
ently if we used the terms ‘children with different schedules for
reading acquisition’ or ‘children we have failed to teach’. They
question taking students out of mainstream programmes and
show how those in many remedial programmes often receive less
reading instruction than those in the classes they have been taken
from; read less text and spend less time reading any text. In such
programmes teachers’ expectations of students are lowered with
consequent effects on the way teachers interact with the students
and the results obtained. Those who get off schedule in reme­
diation hardly ever get back on (p. 998) and it may be the nature
of the instruction they receive, e.g. a focus on decoding, rather
than meaning which keeps them that way (p. 999).
Johnston and Allington (1991) argue that the only way to deal
with it is to eliminate the need for it in the first place and most
effectively by early intervention before problems are compounded.
Class sizes might be reduced but the solution is likely to lie in
higher quality instruction and the creation of non-competitive
tasks involving concepts to increase involvement and co-operation
- in fact, through many of the co-operative procedures discussed
above and below. In this way remediation might become interces­
sion or friendly intervention by consent or invitation (p. 1005).
Nuttall (1996: 144-5) offers some sound advice for those who
simply read too slowly in the EFL classroom. She also argues that
special attention may be needed and the provision of lots of easy
readers is insufficient on its own.

Simply the best? Is there a best way to teach reading?

This is a question that is always asked but which is almost imposs­
ible to answer despite a predom inant accountability orientation in
230 Reading in a Second Language
school evaluation. Hoffman (1991: 917 et seq.) provides an extens­
ive review of the LI literature, comparing different methods of
teaching reading, and no one method of teaching reading showed
itself to be superior to others. Perhaps looking for answers to such
big questions is always likely to be unrequited. A particular prob­
lem with these earlier studies was that they did not collect any
process data and so faithfulness of implementation of a particular
m ethod was seldom available in the interpretation of product re­
sults; for example, no data on actual time spent in reading were
available. Hoffman (1991: 921) cites later studies which showed
the importance of this independent variable.
Hoffman’s impressive survey of effective teaching studies (1991:
921-30) suggested
. . . the important role teachers play in organising and managing
the instructional environment in a way that serves to maximise stu­
dent engagement in academics . . . the important role teachers play
in presenting academic content in a way that promotes learning.
He also provides a model for effective academic teaching (p. 930).
Useful taxonomies of advice for teaching reading are also
provided by Hamp-Lyons (1984: 308), Hoffman (1991: 921-30),
Bernhardt (1991a: 186-7) and Nuttall (1996: 32-3). Similar lists
are suggested by Grabe (1991), Richards (1989), Williams (1996)
and Williams and Burden (1997: 69).
However it is clear from the evaluation literature (Weir and
Roberts, 1994) that any attribution of causality will be difficult. It
may prove impossible to single out any of the features of medi­
ation or activities listed by these authors and directly trace its
impact on reading in the classroom.

Effective schools
Successful reading instruction is more than just an interaction
between teacher, students and materials, however. The context in
which learning takes place can have an important effect and it is
necessary to consider this wider environment for learning as well,
even if the teacher can do little about it outside the walls of his or
her own classroom.
Hoffman (1991) surveyed research into effective schools, as
defined by performance on standardised reading test scores, and
 The teaching of reading 231
identified a num ber of features that are potential contributors
though reservations are expressed about methodology, generalis-
ability and validity. They included:
■clear school mission
■strong curriculum leadership usually from a head
■instructional efficiency: ‘utilisation of resources to achieve max­
imal student outcomes’
■high expectations for students
■good atmosphere; safe, orderly and positive
■commitment to improvement
■individualisation
■careful evaluation of student progress
■reading identified as an important instructional goal
■breadth of material available
■attention to basic skills
■communication of ideas across teachers.

Three steps forward

In the end all of this is an empirical issue, and in the field of L2
reading such data are noticeable by their absence. As we men­
tioned above, the reflective teacher is constantly alert for what
works and what does not in the reading classroom. What we do
not have is broad-based evidence of what works and what does not
work in the teaching of reading. In L2 reading we have only a
handful of studies which have attempted to investigate empirically
the effectiveness of teaching various strategies or adopting a par­
ticular approach in the classroom.
The situation as regards L2 testing of reading is similar. What is
the status of the various strategies and skills we have considered?
Can we demonstrate their separate existence in the tests we write?
What methods will enable us to do this?
We also need a serial model or a number of models which will
provide a clear theoretical base for L2 reading for different pur­
poses, especially one for expeditious reading. What are some ways
in which we might investigate this?
In the next chapter we will make some suggestions as to how
these areas might be investigated. Chapter 5 is titled ‘future re­
search’. We shall not attempt to survey the field in the manner of
Bernhardt (1991a) for L2 reading or Barr et al. (1991) for LI
232 Reading in a Second Language
reading, but we have attempted to incorporate their research
findings and those of others where available in our overview of
theory, testing and teaching of reading in the earlier chapters of
this book. The literature on reading is so vast that we are sure to
have missed a number of important pieces of research in the
welter of references available on LI and L2 reading in general.
We would be glad to be made aware of any omissions.
Our aim in Chapter 5 is to identify some key areas from our
review of the field in Chapters 1-4 and suggest how these might
be investigated through small-scale research studies to help shed
light on areas badly in need of sustained attention.
 253
5

Future research

INTRODUCTION
Throughout this book so far, we have tried to indicate our prefer­
ence for claims and conclusions based on empirical data, rather
than rhetoric and good-hearted sentiment. We finish, therefore,
with a chapter devoted to considerations of some future research
directions for teaching and testing reading in an L2.
The relationship between research and teaching is complex,
and worthy of a study by itself. A single publication by Goodman
(1967) had what some consider to be a disproportionate amount
of influence on the teaching of reading both in the LI and L2. It
remains to be seen whether the work of Stanovich and others has
anything like the same effect. The huge change in L2 from a
focus on linguistic structures to emphasis on communicative use
was not motivated by empirical data as is normally understood by
the term; it seems to have occurred at least in part by an upswelling
of dissatisfaction among the teaching community with the previ­
ous paradigm.
The characteristics of the participants involved in teaching and
testing on the one hand, and research on the other, are again of
interest. Both sides have their virtues and vices. Teachers and
testers are usually in contact with real learners, and often form
hunches regarding these learners based on extended experience,
which may well be valid, even if difficult to substantiate. They
often lack, however, an explicit theoretical framework against which
to relate these hunches, and thus evaluate them and extend them.
Researchers, on the other hand, usually operate within a theoret­
ical framework. However, they may not bring experience to bear
on this framework. Moreover, they sometimes rely uncritically on
234 Reading in a Second Language
not well-substantiated or repeated experimental results. Here we
make an attempt to bring the two worlds together, to evaluate teach­
ing and testing methods in the light of theories, and vice versa.
It is sometimes fashionable in teaching circles to sneer at ‘theory’
(which sometimes seems to extend to everything except method­
ology). To the extent that some teacher trainers, having escaped
from the classroom themselves, have appeared to exclude ‘prac­
tical’ considerations from what they tell students, this reaction is
understandable. We, however, do not go very far in our sympathy.
Teachers and testers need both a sound grasp of practical mat­
ters, and an educated framework on which to base and to evaluate
their methods. We firmly believe that without such a framework,
the teacher or tester is trapped in a particular set of practices,
with no motivated criteria for making alterations. Often they are
reduced to evaluating teaching practices in terms of whether the
students enjoyed them or not.

Two research communities

Many of the readers for whom this book is intended, as well as the
authors themselves, can be said to belong to a professional group
termed ‘applied linguists’. We shall attempt to define this group
and establish some boundaries though we are aware that many
might disagree with our efforts. Our definitions, like all others,
are partial. The group is rather amorphous: a recent letter from
the Secretary-Treasurer of the American Association for Applied
Linguistics describes the association’s ‘1200+ members’ as work­
ing in ‘many language-related areas, including language educa­
tion, second language acquisition and loss, bilingualism, discourse
analysis, literacy, rhetoric and stylistics, language for special pur­
poses, psycholinguistics, second and foreign language pedagogy,
language assessment, and language policy and planning’.1 The
section of the group with which we are particularly concerned
consists of people who are involved in some way with the teaching
and testing of an L2. They are quite a large group; Sampson
(1997) attributes the rise in importance of linguistics in the 1960s
and 1970s to an influx of such people.
While the book is directed towards this group, we have very
frequently referred to another academic group concerned with
reading - namely, cognitive psychologists. The two groups are
indirectly related, both owing their existence, or preoccupations,
 Future research 235
at least in part to the reaction against Behaviourism in the 1960s
(see Gibson and Levin, 1974). They are, however, fairly distinct.
Applied linguists, as indicated above, are fairly widely spread, both
in terms of professional interests and places of work; cognitive
psychologists seem on the whole to operate in psychology depart­
ments in universities.
In this section, we shall briefly compare the contribution made
by the two groups to research into reading behaviour. The com­
parison is not, on the whole, flattering to the applied linguistic com­
munity, ourselves included. We shall suggest some reasons related
to our respective work situations which serve to some extent to
excuse this. However, the main reason we have for making the
comparison is to suggest ways in which applied linguistic research
into reading may improve by study and use of the psychologists.
The first thing to be said about the contribution of the cog­
nitive psychologists is that the database it has to refer to is very
large. Rayner and Pollatsek (1989) list around 850 entries in their
bibliography, the vast majority taken from their own field. By con­
trast, surveying the L2 reading literature, Bernhardt (1991a) could
find only 121 articles concerned with empirical research into the
reading process. But it is not a matter just of number of articles:
their research tends to be cumulative, one experiment building
on another, proceeding by criticism and extension. For example,
early work on word and letter recognition was criticised, the
method refined, then the work repeated until the results are able
to stand up to critical scrutiny (Rayner and Pollatsek, 1989: 76ff).
The general approach seems to be refreshingly empirical, with a
strong drive to discover details of what is happening at different
points in the reading process. For example, Rayner and Pollatsek
(p. 265) comment that ‘everyone agrees that real-world know­
ledge has to actively intervene in reading for comprehension to
take place; the question is how’.
In addition, there is considerable emphasis on method. Ray­
ner and Pollatsek frequently introduce a topic by describing the
methods used to investigate it, e.g. the methods used to measure
the time involved in word recognition (pp. 63ff). The same methods
can be used by successive researchers, enabling results to be com­
pared. The range of methods is impressive, and the details are
often ingenious (e.g. some of the work on phonological decoding)
and even entertaining: ‘Crude Jude chews food’ rings in the mind
(see Section 2.2).
236 Reading in a Second Language
Inevitably, there are aspects of the cognitive psychologists’ work
about which we have reservations, and which we shall mention
here, to show we are not totally overawed. There is, for example,
the use in some experiments of so-called ‘unpronounceable’ words,
used to determine, for example, whether a reader is attempting a
phonological or direct route in accessing the lexicon. Gibson and
Levin (1974), discussing the use of ‘unpronounceable’ words with
young LI (English) readers, point out that ‘unpronounceability’
and orthographic ‘illegality’ are often confused (pp. 201ff). With
L2 readers, another factor is likely to emerge, the phonological
rules of their LI. For example, Koda (1987) used two sets of
pseudo-words, pronounceable and unpronounceable, with Japa­
nese students learning English. Strings like ‘mastib’ were classed
as ‘pronounceable’, and strings like ‘msatbi’ as ‘unpronounce­
able’. The dangers seem clear: an initial sequence of [ms] would
be perfectly possible and pronounceable for a speaker of Swahili,
and many other languages. Consonant sequences which might be
unpronounceable for speakers of Japanese might seem normal
for, say, a Polish speaker. There is, it seems to us, no such thing as
‘unpronounceability’.2
Then there is the problem about so-called ‘regular’ versus ‘irre­
gular’ words, i.e. words which are spelled ‘regularly’, e.g. ‘doom ’
[du:m] versus ‘blood’ [bL\d] (cf. Rayner and Pollatsek, 1989: 87;
Suarez and Meara, 1989: 352). The argument is presumably that
‘oo’ is the regular spelling of the phoneme [u:] in monosyllabic
words of the structure CVC. What, however, is the ‘regular’ pro­
nunciation of the spelling ‘-ough’? A concordancer check of a
5(1,000 word text found 19 occurrences of words pronounced [ou],
(although, though), 10 of words pronounced [Af] (rough, tough,
enough) and 31 of [u:] (through). It would be hard on this evi­
dence to establish which the ‘regular’ pronunciation was. In fact,
claims as to ‘regularity’ are fundamentally statistical, based (prob­
ably intuitively) on a large number of words from the total Eng­
lish lexicon. But the L2 learner is operating with a much smaller
base. What the learner thinks of as ‘regular’ may be affected by the
actual words in his or her lexicon.
A more serious question concerns the ‘naturalness’ of many
of the experiments carried out in psychology laboratories. The
laboratory allows careful control of the experiment, and the use
of special equipment, for, say, timing responses to the millisecond.
It does, however, leave open the possibility that the results are
 Future research 237
restricted to the laboratory, and cannot be carried over to ‘real
reading’. The fact that experimental subjects can be shown to
vocalise single words flashed on a screen is not, it seems to us, very
good evidence that they do the same thing when reading a long
text quickly. It is arguable that, in the laboratory situation, vocalisa­
tion is the only response subjects have open to them. Rayner and
Pollatsek are very aware of this problem; it is why they attach such
importance to using eye movements as evidence of reading activity,
since they do not disturb normal activity. On the whole, although
laboratories are ‘unnatural’, so too are classrooms and examination
halls, and as long as we are aware of the problem, this is something
we have to live with.
Finally, it could be argued that the cognitive psychologists
haven’t got very far. They know a lot about word recognition but
not much else. This is very unfair, but there is an element of truth
in it. As we noted in the discussion on bottom-up models in Sec­
tion 2.1, Gough’s model, first proposed in 1972, contains a com­
ponent called ‘Merlin’, which handles syntax and semantics in a
suitably mysterious way. Rayner and Pollatsek (1989: 263) admit
that ‘our present knowledge of how sentences are read is meager’.
However, they are optimistic about future developments, and it
has to be said that if cognitive psychologists know very little about
how higher level processing occurs, they know more than anyone
else.
All in all, from consideration of the contribution to reading
research by the cognitive psychologists one gets the impression of
a lot of data, obtained by well-established research methodologies,
and organised by means of models, which not only make sense of
the existing data but point to further areas of research.

The applied linguistics contribution

W hen we look in turn at the experimental data gathered for L2
reading, usually by researchers who would be classed as applied
linguists, the following aspects emerge:
■The database, as Bernhardt (1991a) points out, is small. As an
example partly separate from Bernhardt’s analysis, Reading in a
Foreign Language classifies articles into Reading Process, Text
Analysis, Course Design, Graded Readers, Materials and Meth­
odology, Lexis, and Testing. Up to Volume 11 (1), there have in
238 Reading in a Second Language
all been 91 articles. Suppose we class as experimental data any
paper containing a description of a controlled experiment.3 Of
our total of 91 papers, roughly 30 per cent can be classified as
experimental, with most of these being found in Reading Pro­
cess (14/18) and Testing (6/12).
■The research is often not cumulative; i.e. as Bernhardt points
out, many of the experiments are one-offs. For example, Pegolo
(1985) and Dhaif (1990) both published in Reading in a Foreign
Language. Both papers reported experiments in which reading
silently was accompanied by an oral input, in Pegolo’s case,
prosodic information only, in Dhaif’s case, the text read by a
teacher. The language of the first experiment was French; of
the second, English. The comprehension of the experimental
groups improved significantly in both cases. These results are
interesting and of apparent pedagogical significance but, at least
as far as that publication is concerned, were not repeated. It is
interesting that when an experiment is repeated, as in Nesi and
Meara’s repetition (Nesi and Meara, 1991) of a study done by
Bensoussan and Laufer (1984), one of the authors, Meara, has
training in psychology. Even when a topic is repeated, as with,
say, Background Knowledge, the different methodologies em­
ployed by different experimenters, as Bernhardt says, make
comparisons difficult.
■There is no established body of experimental methods which
the applied linguists can fall back on. In addition, the methodo­
logy used tends to be uninspiring. We cannot all be Bransford
and Johnsons, but there is more to experimental design than
matching two groups reading different texts. The methodology
used by Meyer (1975) to investigate the effect of rhetorical organ­
isation on recall, is elegant - pairs of texts being produced by
embedding the same target paragraph in two different organ­
isational contexts. One has to assume that she, again, has a
background in psychology. Many applied linguists, setting out
to experiment, obviously believe that when one can handle
ANOVA on a PC statistical package, one has completed one’s
apprenticeship.
■There seems to be a tendency for applied linguistic researchers
to accept a theoretical construct without further analysis, defini­
tion, or even empirical investigation. Cohesion and schemata come
to mind, while we have commented in Section 2.3,on the way in
which ‘a reading problem or a language problem ’ was taken up
 Future research 239
without much discussion or definition of what constituted ‘read­
ing skills’ or ‘language ability’. As far as sustained empirical
investigation is concerned, it is striking how Goodman’s claim
about ‘a psycholinguistic guessing game’ was accepted by applied
linguists, without very much testing of the claim with L2 readers
(see, however, Rigg, 1988).

Causes for the differences

When we look for reasons why the applied linguists, including
ourselves, do so comparatively poorly, one is tempted to fall back
on the excuse that their attention is diffused over a number of
loosely related areas. If we revert to our initial description of the
work situations of the members, we find a rather wide range of
occupations and pre-occupations, from language planning to styl­
istics to second language acquisition study. It is a very broad area,
with a potentially misleading subject name, and a lack of core focus.
In addition, many members spend a considerable proportion of
their time actually teaching language, or training language teachers.
Profit or perish is often a more immediate imperative than publish
or perish.
Even when we restrict the focus to reading, the spread of inter­
est for applied linguists can be wide, extending from studies of
the effect of different scripts to the effect of dictionary use to the
influence of different text types. Reference to the quick count
from Reading in a Foreign Language above suggests that some of the
areas in which applied linguists operate are less oriented towards
experiments than others. For example, while Reading Process con­
tains 18 papers, of which 14 contain experiments (78 per cent),
Text Analysis contains none. Research need not be restricted to
experiments.
However, we must be wary of any rather self-pitying excuses for
ourselves. Cognitive psychologists have interests other than read­
ing. The reasons for the poor database must be found elsewhere.
The principal reason seems to be that while cognitive psycho­
logy is an experiment-centred study area, applied linguistics is
not. The typical applied linguist will not have come formally into
contact with experiments during his or her undergraduate train­
ing, and while an applied linguistics course in Britain, at least,
may include a small research design component, it is seldom a
240 Reading in a Second Language
major part of the course. Hence it tends to be the case that when
graduates of such courses design experiments, they lack the ground­
ing the psychologists have had in method, particularly the depth
provided by knowledge of conventional types of experiments which
have already been used in investigating similar areas.
A secondary reason is that many experiments in applied linguis­
tics have been done, as it were, on the side, so that, for example, a
teacher teaching reading may set up a reading experiment. It is not,
in other words, a mainline part of the teacher’s professional duties.
A concentration on experiments in psychology is accompanied
by a major interest in theory and in models of the reading pro­
cess. Until recently, L2 reading research has, as Bernhardt says,
been atheoretical. Applied linguists have paid comparatively little
attention to formal models, particularly process models. Bernhardt
comments that a disproportionate amount of the references in
the applied linguistics area have been to Goodman, at the expense
of, say, Gough, or Just and Carpenter, or Kintsch and van Dijk. She
has a point, though in defence of the applied linguists, they are
often personally involved in teaching reading, or feel a responsib­
ility to the L2 teaching community. Neither Rayner and Pollatsek
nor even Just and Carpenter have much immediate relevance for
the classroom teacher. Goodman had a major effect partly because
it was easy to apply his main precepts.
The value of a model, for researchers, is that it has the ability
to gather together previously diverse experiments and can focus
attention on different aspects of a problem, and prevents the
emergence of a rash of experiments without any organising prin­
ciple. If we look, for example, at Gough’s bottom-up model, we
can see how it invites experiments in areas such as feature recog­
nition, phonological processing, and grammar. The fact that the
psychologists have made little headway in experiments with gram­
mar does not invalidate this claim.

The way forward

1. We think applied linguists should become more familiar with
the cognitive psychology literature.
2. In particular, we should concentrate on experimental methods,
and on a wider range of models. To this extent we agree with
Bernhardt.
 Future research 241
3. With all respect to Ridgway (1996), we do not think it is useful
for applied linguists to put forward rival models to account for
low-level processes. On the whole we have neither the equip­
m ent nor the training.
4. We think we would be justified in proposing new models, if
necessary, for higher level operations. We are better versed in
the methodological procedures appropriate for such strategy
research.
5. We should concentrate on maximising our advantages in the
strategies field: access to readers in real learning situations,
familiarity, even expertise, in certain areas of use.
This chapter contains three sections. In Section 5.1, we start by
suggesting a set of possible criteria for judging models of the
reading process from a utilisation-focused perspective, i.e. what
value do they have for improving teaching, testing and research­
ing reading? We then look at how one could obtain empirical
data concerning the types of reading we have described earlier in
the book. We discuss this in terms of the ad hoc model we adopted
in Section 2.4, which includes a goalsetter, a monitor, as well as a
sentence processor.
In Section 5.2 we suggest research that applied linguists and
classroom teachers can conduct on the relationship between read­
ing and grammar. We have criticised applied linguists for not
defining their terms with sufficient rigour. In response to this -
and also in response to ‘grammar’ tests which contain material we
do not think is appropriate - we devote a considerable amount of
time in this section to defining the scope of syntax.
Finally, in Section 5.3 we focus on the use of language testing
to investigate the componentiality of reading in a second lan­
guage. In Section 3.2 above we voiced our concerns about the
methodological shortcomings of a lot of previous research in this
area. In particular, we were troubled by the narrow careful view of
reading most earlier studies were premised on. It is the belief of
the authors that testing procedures are a valid and reliable way
of improving our understanding of the components of reading in
L2 and the relationships between them in terms of both construct
overlap and level of ability. The findings of such test-driven re­
search should enhance both the teaching and the formative and
summative evaluation of these abilities. A principled set of proced­
ures is offered for such investigation.
242 Reading in a Second Language
5.1 MODELS AND DIFFERENT TYPES
OF READING
We have mentioned in the introduction to this chapter the desir­
ability of operating with models of reading. A very ad hoc model
has been referred to in Section 2.4, and use will be made of it in
the second part of this section. We begin, however, with a discus­
sion of the criteria for evaluating models, since not all are suitable
for research in the areas of teaching and testing. In the second
part of the section, we look at possible research aimed at discover­
ing what reading behaviours are available to L2 readers.

Criteria for judging models of the reading process

The criteria we propose here do not constitute an exhaustive list
for judging how perfect a model is in describing the complex
process of reading, but serve to judge the operationalisability of
the models of the reading process for the purpose of improving
the teaching and testing of reading.
One of the valuable sources for our discussion of the judgem ent
criteria is from Samuels and Kamil (1988) who raise a num ber of
questions that should be considered in evaluating models of the
reading process. Their criteria emphasise the generalisability of
the models: ‘Because of the interactive nature of the variables in a
study, we must attempt to evaluate the different models in terms
of their generalisability’ (Samuels and Kamil, 1988: 26).
De Beaugrande (1981) also proposed sixteen criteria for model
designing. Using these criteria, he characterised ten alternative
models of reading or understanding, including Chomsky, the
Clarks, Gibson, Kintsch, Meyer, Frederiksen, Schank, Rumelhart,
Woods, and the author’s model. But he laid emphasis on proposing
a set of design criteria that can serve to construct comparative
profiles of process models for reading research . . . they are useful
in promoting functional consensus by allowing diverse domains to
be characterised in a common descriptive idiom.
(de Beaugrande, 1981: 263)
We are concerned with the implications of the process models
for the practice of teaching and testing. We shall therefore focus
on those criteria that help us judge if a particular model can serve
 Future research 243
the purpose. However, some of the ideas in de Beaugrande’s de­
sign criteria and in Samuels and Kamil’s evaluation criteria are
valuable for our discussion and are thus incorporated into the
three broad categories of judgem ent criteria we propose, that is,
accessibility, comprehensiveness and falsifiability, which are sum­
marised in the following list, developed with Jin Yan of Shanghai
Jiaotong University, PRC.
CRITERIA FOR JUDGING MODELS OF THE READING PROCESS

(A) Criteria of accessibility

The accessibility criteria evaluate the practical value of the reading pro­
cess model. In the study of teaching and testing of reading, the first and
foremost concern is that the model should be accessible to teachers,
teaching material developers, syllabus designers, and test developers.

(A1) Clarity of the process

Does the model unpack the complex process of reading in terms of the
practical concerns of teachers and testers such as language skills,
processing strategies and knowledge sources so that it is within the
grasp of the majority of teachers and testers?

(A2) Implications for teaching and testing

Can the model directly shape the teaching and testing objectives and
provide guidance on teaching task design, teaching methodology, test
formats, item types and text selection?

(B) Criteria of Comprehensiveness

Comprehensiveness judges the extent that a model can be applied to
various targeted processes of reading, that is, different readers in dif­
ferent reading contexts. It addresses the problem unsolved in many
models, that is, for different types of readers, for different purposes of
reading, and for different types of texts, the processes of reading in
terms of strategies and skills may vary.

(B1) Types of reader

Does the model adequately describe both fluent readers and beginning
readers?

(B2) Purposes of reading

Does the model describe reading across a variety of tasks and purposes?

(B3) Types of text

Does the model describe the reading process for different types of
reading materials?
244 Reading in a Second Language
(B4) Scale of processing
Does the model describe the word recognition process as well as higher
level comprehension processes?
(B5) Transferability
Is the model of the first language reading process generalised enough
so that it can be transferred to the second language or foreign language
reading process?

(C) Criteria of falsifiability

Falsifiability refers to the degree that a model can be subjected to empir­
ical tests and thus be supported, proved or improved by the evidence.
(C1) Empirical evidence
Has the model been substantiated by empirical evidence?
(C2) Predictability
How well can the model predict and explain the reading behaviour in
psychological experiments?
(C3) Replicability
Are the experiments on which the model is based replicable by other
researchers?
(C4) Openness
Can the model accommodate new insights accrued from empirical stud­
ies without requiring a fundamental change of the design?

Investigations of types of reading

Here we research into the different styles of reading available to
L2 readers. We use as a reference point the model we proposed
in Section 2.4, and the classification of reading styles put for­
ward there. We see this research as taking place largely in the L2
reading classroom; this affects the resources and equipment sug­
gested, since any experimental techniques must be confined to
those practical in the normal classroom. The focus, however, is not
on teaching methodology, but on the application of a description
of different reading styles, our ‘model’, in an L2 context. In general,
we are concerned with such questions as:
■Do the learners have access to all the different styles mentioned
in the model?
■If not, do they have access to some, or are they restricted to one
style?
■Are some styles more accessible than others?
 Future research 245
In addition, in light of what was said in the introduction to this
research section, we are concerned with possible methodologies
for investigating these questions.
We should reiterate that our model presented in Section 2.4 is
not intended to be a ‘proper’ process model of reading - that is,
it is not fundamentally based on empirical findings about the
reading process. To some extent, it is in part indirectly based on
such findings, since it is derived from Just and Carpenter. The
intent of setting it out, however, is primarily to direct attention
towards certain aspects of reading behaviour which we consider
fundamental for our readers, and to focus attention on possible
areas of empirical investigation.
As we m entioned in Section 2.4, we have added to the Just and
Carpenter model two additional components, namely a goalsetter;
and a monitor. We have included the goalsetter to emphasise that
there are different purposes for reading, and that different pur­
poses are likely to result in different reading behaviours. Many
writers agree in principle that readers should be flexible about
their reading behaviour. Thus Gibson and Levin (1974: 548) argue
that ‘flexibility of reading style is of the greatest im portance’ and
Rayner and Pollatsek (1989: 452) say that it is ‘generally accepted
. . . that the ability to be flexible in reading is a characteristic of
better, more mature readers’. The decision by the latter writers to
concentrate on careful reading was taken largely because that sort
of reading is best researched, and also, possibly, because it is the
easiest to observe. However, statements such as ‘When you read a
passage of text, your primary aim is to comprehend the passage’
(Rayner and Pollatsek, p. 449) tend to obscure the differences
between different reading behaviours, and hence, we think, justify
our emphasis.
While agreeing with the view that readers should be flexible,
Baker and Brown (1984: 30) suggest that ‘there are students who
still fail to set their own purposes, reading everything at the same
rate’. It is our experience, and that of many L2 teachers, that L2
readers often exhibit this sort of reading deficit. In Section 2.4 we
suggested five fairly conventional types of reading - namely, scan­
ning, search reading, skimming, careful reading at a local level,
and careful reading at a global level - and the research proposed
below is designed to find out how many of these types are avail­
able to L2 readers (see also Sections 3.2 and 4.3 for an extended
discussion of these).
246 Reading in a Second Language
The goalsetter
The goalsetter represents the readers’ purpose. In life outside the
classroom and examination hall, readers choose their own pur­
pose in reading. In the examination hall, in particular, however,
the readers have little choice but to conform to the wishes of the
examiners. Hence we can relate purpose to test task. This near
equation of goal and task is of particular importance to us in that
we consider the importance of task to have been underrated in
aspects of the literature, e.g. in discussion of skills transfer. For us,
‘comprehension’ can best be described as the successful accom­
plishment of a particular reading goal or task. If there is no real
purpose in reading, then the outcome cannot be measured. In
this we agree with Kintsch and van Dijk (1978: 374):
In many cases, of course, people read loosely structured texts
with no clear goals in mind. The outcome of such comprehension
processes, as far as the resulting macrostructure is concerned, is
indeterminate.
We would add, however, that the text need not be ‘loosely struc­
tured’, and that it is not just the macrostructure which is likely
to be indeterminate. As said before, it is for this reason that
we eliminated from our list this reading style, which we termed
‘browsing’.

The monitor
The notion of cognitive monitoring has been part of the reading
construct for more than a decade. Baker and Brown (1984: 22)
define it as:
The ability to use self-regulatory mechanisms to ensure the success­
ful completion of the task, such as checking the outcome of any
attempt to solve the problem, planning one’s next move, evaluating
the effectiveness of any attempted action, testing, and revising one’s
strategies for learning, and remediating any difficulties encoun­
tered by using compensatory strategies.
They add that:
Since most of the cognitive activities involved in reading have as
their goal successful comprehension, a large part of cognitive mon­
itoring in reading is actually comprehension monitoring.
 Future research 247
In our model, the monitor is closely connected to the goalsetter.
Its principal task is to keep a running check on whether the goal(s)
are being achieved. Thus, it is important to realise that the mon­
itor varies in what it considers successful activity, in terms of the
task set by the goalsetter. To drop for a moment into a fashion­
able computer analogy, the monitor may be ‘set’ at different val­
ues.4 Thus the monitor should demand different standards of
behaviour when the reader is, for example, scanning, as com­
pared to when the reader is engaged in careful global reading. If
readers do not vary the ‘setting’ of their monitor, then in many
cases their goal will not be achieved.

Modes of measuring
In theory it should be possible to investigate the reading process
at all the relevant points along the line. We could, for example,
investigate the activity of the goalsetter, of the monitor, the suc­
cess or failure of the sentence by sentence processing activity, as
set out in our model, or the product. In practice, it is likely to be
difficult to distinguish the contribution of the different compon­
ents: the product, for example, may be deficient either because
the initial formulation of the task was defective, or because the
monitor did not operate efficiently while reading was going on.
However, we can suggest general areas of investigation, and be
more specific when we discuss particular reading styles.
The goalsetter may be investigated by interviewing readers as
to how they interpreted the task. Alternatively, the relationship
between goalsetter and product may be examined. There is ample
evidence in the literature of different tasks resulting in different
products (Thomas and Augstein, 1972; Rayner and Pollatsek, 1989:
452-3).
Baker and Brown (1984: 23ff) suggest various ways of invest­
igating the monitor, including:
■observation of readers
■analysis of oral reading errors
■assessing certainty concerning responses
■cloze procedure
■text disruption.
We are happier with some of these than with others: oral reading
has, in our opinion, to be handled very carefully, if at all, in the
248 Reading in a Second Language
L2 classroom. Nevertheless, it is useful to know that some pro­
gress has been made towards an established range of investigatory
methods.
Since we are distinguishing between expeditious and careful
reading, an obvious area of observation is speed of reading. In the
absence of any more sophisticated method of assessing reading
speed, there are two general ways of going about this in the class­
room: either the students read a text and are timed doing it, then
reading speed is measured in terms of words read per minute (cf.
Fry, 1963), or the students are given a reading task and a fixed
time in which to accomplish it. On the whole, in the investigations
suggested below, we favour the first method. It brings with it,
however, the problem of whether to include some measure of
success on a reading task, as well as simply speed of reading. Fry,
for example, uses multiple-choice comprehension questions to
be answered after reading. Since no reference back to the text is
allowed, this introduces a memory factor. It may be best simply
to instruct readers to read a text at their ‘norm al’ speed, and
time their performance. However it is done, when we go on to
compare, say, scanning speed with a more careful reading speed,
we need some estimate of ‘norm al’ reading speed to use as a
base line.
We now proceed to look at the different reading styles, with
a view to making suggestions as to how their existence or non­
existence may be investigated among L2 readers. We begin with
scanning, which might be considered an extreme case.

Scanning
We have already defined scanning as ‘reading selectively, to achieve
very specific goals’, e.g. finding the num ber in a directory, finding
the capital of Bavaria (in a geography or history book). Nuttall
(1982) defines it as:
glancing rapidly through a text either to search for a specific piece
of information (e.g. a name, a date) or to get an initial impression of
whether the text is suitable for a given purpose (e.g. whether a book
on gardening deals with the cultivation of a particular vegetable).
Consideration of these definitions suggests that scanning is not
quite the simple concept that we originally thought. The definition
we had originally in mind covered instances like finding a solitary
 Future research 249
numerically expressed date, or finding the word ‘Munich’. Neither
of these activities, however, seems very natural. A more natural
event, such as being asked ‘Find the date of the battle of Water­
loo’, might involve a search for the collocation of Waterloo and a
date, but once the reader had found the general area, it also
involves a scan towards the beginning, where such introductory
information would be likely to occur. A similar situation would
arise if one were given a book on the geography of Germany and
asked to find the capital of Bavaria.
It looks as if scanning merges with what we have called ‘search
reading’. In what follows, however, we will stick to examples of
what one might call ‘extreme’ scanning, i.e. activities similar to
those carried out by the computer on the instruction ‘Find’. We
might add that such activities are very different from ‘normal
reading’. Rayner and Pollatsek, discussing ‘proof reading’, query
whether the activity has implications for normal reading. We think
that scanning, as defined above, may have such implications in an
L2 context - a point we discuss below.
When we consult our model, it might seem that all that is
involved in scanning, as defined above, is word recognition; there
is no need in the cases above for processing the syntax or seman­
tics of the sentence containing the search item, and no need,
apparently, for the reader to bring background knowledge into
play. In fact, one might conceivably argue that the reader does
not even need to access the lexicon, since it would presumably be
possible to ‘scan’ a text for a nonsense word. Certainly, numbers
in telephone directories do not require access to the lexicon. We
referred above to the facility most popular word processor programs
contain for ‘finding’ particular words. The computer accomplishes
this by a process of running through the text, matching each
word it comes across with the search item. There is no need for
meaning. We don’t know of any research involving readers scan­
ning in a language unknown to them; it seems likely that it would
be an exhausting experience.
Our principal research aim in this case is to determine whether
our L2 readers can scan as well as read ‘normally’, in other words
whether they have access to more than one strategy. A secondary
aim, given that the answer to the first question is ‘Yes’, is to invest­
igate how they are doing it.
The obvious way to investigate the first question would seem to
involve comparison between scanning and, say, normal reading. If
250 Reading in a Second Language
one just gave the students a text, and told them ‘Find the follow­
ing words in the text’, there is nothing to prevent them from
plodding through at normal speed, identifying each target word
in the course of normal reading. In fact, if they have only one
strategy, then this is what they will do. According to our classifica­
tion of reading behaviours, scanning belongs to the ‘expeditious’
group, i.e. it should be carried out at a faster speed than normal
reading. Hence one way of establishing whether students can scan
is first to find out their normal reading speed (which is best done
over a number of trials), then find out whether their scanning
speed is faster. There are several elementary precautions to take.
If we already have a measurement of each student’s normal read­
ing speed, then we do not have to worry too much about the
length of the text(s) used for scanning. Scanning may seem to
require texts of substantial length, which beginning students might
find exhausting to read carefully. However, texts used for scan­
ning and careful reading should probably be similar in terms of
familiarity and difficulty.
The number of items to be looked for in such a test is presum­
ably a matter of experience. If dates are used (they might seem
suitable as a practice to familiarise students with the activity), then
only a small number, i.e. 1 to 3, would seem enough. If the in­
structions are ‘Find the following words or phrases’, then 1 or 2
would probably be too small, and easily missed. Ten would seem a
possible compromise.
We are not restricted to comparative reading speeds as a means
for assessing whether scanning or some other reading behaviour
has taken place. In a scanning operation, we see the monitor as
set at a simple Yes/No standard; i.e. is x the word the reader is
looking for or not? This is the case in a computer Find operation.
If the answer is ‘No’, then the search moves on and the last item
examined is dismissed from attention. This means that if a scan­
ning operation has been carried out successfully, not only will all
the items requested have been identified, but, in theory at least,
nothing else from the text has been recovered, i.e. committed to
memory. Gibson and Levin (1974: 539) remark that: ‘One can
scan for a graphic symbol or a word target very fast, but the scan­
ner remembers almost nothing of what he saw except the target.’
An examination of the reader’s knowledge of the contents of the
text just scanned, carried out, perhaps, a few minutes after the
scanning period is complete, should reveal that the reader retains
 Future research 251
little or nothing of the text. Either an interview or a test can be
used for this. We have here the odd case when the lower the
reader scores on this test, the better.

Pedagogical implications
We have already indicated some doubts as to whether scanning as
we have defined it is reading at all. As stated above, Rayner and
Pollatsek had doubts as to whether editing tasks had ‘implications
for normal reading’, i.e. whether conclusions based on behaviour
on such tasks had any relevance to discussions of normal reading.
We think that scanning is relevant to L2 reading in two ways.
The first way is partly methodological. In our experience, some
L2 readers insist on one style of reading - a relentless, slow plod
through the text, beginning at the top left-hand corner, and con­
tinuing to the end, the process only broken up in some cases by
frequent recourse to a dictionary or to the teacher as a dictionary
equivalent. This form of reading behaviour, which can be quite
hard to break, may be an epiphenomenon, i.e. a product of previ­
ous experience in the classroom. A ruthless insistence on scan­
ning on the part of the teacher may help break this pattern.
The second way is more basic. We have already seen in Section
2.2 the importance that is now attached to automatic word recog­
nition in reading. It is arguable that what makes scanning easy for
a good LI reader is just such recognition; the monitor is able, very
quickly, to provide its ‘Yes/No’ answer. The L2 reader, on the
other hand, being less able to distinguish between, say, blip, flip,
bleet, fleet, etc., is likely to find scanning much more difficult. In
fact, it is again arguable that such readers will have to consult the
context of a target word in order to identify it. Thus scanning can,
at the very least, be used as a useful test of word recognition, not
only of the target items but the surrounding items in the text.

Skimming
The reader is referred back to the relevant text of Sections 2.4, 3.2
and 4.3 where we have discussed these strategies in some detail. In
Section 2.4 we provided initial working definitions which were
then developed in Section 3.2 on testing. In Section 4.3 we tried
to break each down further into its constituent enabling opera­
tions for use in classroom learning tasks. We define skimming as
252 Reading in a Second Language
‘expeditious reading carried out for the purpose of extracting
gist’. It thus contrasts strongly with our description of scanning
(see also Appendix 1 for working definitions used in the Chinese
Advanced English Reading Test).
Authors are in agreement as to the value of skimming. Nuttall
considers that skimming (together with scanning) enables the
reader ‘to select the texts, or the portions of a text, that are worth
spending time on’, thus suggesting that skimming is a preliminary
to careful reading. Rayner and Pollatsek (1989: 447) remark that:
Skimming ... is a very important skill in our society. In careers that
depend on the written word, there is simply too much information
to be assimilated thoroughly, and we are constantly forced to select
what we look at. Those unable to skim material would find they
spend their entire day reading.
Rayner and Pollatsek equate ‘speed reading’ with skimming, and
say that:
speed readers appear to be intelligent individuals who already know
a great deal about the topic they are reading and are able to suc­
cessfully skim the material at rapid rates and accept the lowered
comprehension that accompanies skimming. (p. 448)
The gist which we mentioned above should be something like a
reduced form of Kintsch and van Dijk’s macrostructure. A look at
the model will suggest that, in the process of skimming, even if an
entire sentence is processed, the reader will not necessarily proceed
to the next text sentence. The process is therefore selective, some
parts of the text being omitted. In turn, and keeping in mind
Kintsch and van Dijk’s model, this means that coherence relations
between different propositions in the text are likely to be sacrificed
in the act of skimming. It follows from this that, if the product
of the skimming is to be coherent, then background knowledge is
going to have to play an increased role in the build up of the
macrostructure. Rayner and Pollatsek consider that the successful
speed reader (i.e. skimmer) already knows a lot about the topic.
It seems to us, then, that the efficiency with which L2 readers
skim a text is likely to depend crucially on their knowledge, either
of the topic of the text being skimmed, or the structure of the
text, or both, and that this is likely to be even more the case than
with careful reading. This familiarity may come either from previ­
ous reading in the L2, or from their previously acquired literacy
in the LI.
 Future research 253
Given that tasks aiming to induce skimming can be framed
with sufficient clarity, and that the readers are familiar with what
is expected of them, more than one hypothesis can be derived
from the above discussion.
■On the same text, or on texts of equivalent length and difficulty,
reading speed when in skimming mode will be significantly faster
than when in careful reading mode.
■Macrostructure built up on the basis of skimming should be
significantly less detailed than that acquired through careful
reading. If the skimming is efficient, however, the propositions
omitted from the macrostructure should be the lower level, more
detailed ones.
■Skimming performance should be significantly reduced if the
readers are exposed to texts of unfamiliar structure.
Obviously experimental work designed to test these hypotheses
would incorporate timed reading, as well as recall protocols eli­
cited after the reading was complete. A useful controlled exercise
in teaching skimming consists of students being given very gen­
eral, high-level questions before they read, for example ‘give a
title for the passage’. This method could be used together with
while-reading observation of students to find out how they pro­
ceeded to answer the questions. Signs of skimming might be rapid
movement between pages, possible regression back across pages,
as well as rapid completion of the exercise. Self-report protocols
might also provide insightful data (see Cohen, 1998).

Careful reading
With careful reading, we are in a somewhat different position
concerning speculation as to what strategies are available to stu­
dents. It would be the view of many teachers of reading that stu­
dents are definitely able to read carefully both locally and globally,
whereas this might not be the case for our other posited types of
reading. This may or may not be a valid assumption. The hypo­
thesis would be that students would perform significantly better on
questions aimed at careful reading than on other types of reading
(see Section 5.3 below). Also, their speed of reading in the former
would be similar to their speed when timed during ‘normal read­
ing’ sessions.
254 Reading in a Second Language
Kintsch and van Dijk (1978: 371) speculate that:
If a long text is read with attention focused mainly on gist compre­
hension, the probability of storing individual propositions of the
text base should be considerably lower than when a short paragraph
is read with immediate recall instructions.
Their first case concerns skimming, in our terms; the second is an
instance of what we would refer to as careful local reading.
When comparing our two types of careful reading (global and
local), it is obvious that speed is not going to be an issue. We can
concentrate on other aspects of the reading process, an important
one being performance on tasks (see Section 5.3 below). In Sec­
tion 3.2 we reported evidence of L2 students performing better
on global as opposed to local questions. It looks as if some L2
readers, at least, are using background knowledge to compensate
for linguistic deficiencies when reading for global meaning. An
attractive area of research here would be to locate such students,
then investigate in detail their reading behaviour at low levels as
described in the model, i.e. lexical access, syntactic and semantic
processing, and establishment of cohesive links.
Conclusion
The pedagogic literature frequently refers to careful reading, skim­
ming, scanning, as well as ‘intensive’ and ‘extensive’ reading. We
have indicated that we do not think these types have been defined
with sufficient clarity in the literature. We have attempted in Chap­
ters 2, 3 and 4 to clarify the differences and establish operational
definitions for the five ‘m ain’ skill and strategy groupings. It is our
hope that if a num ber of researchers can be persuaded to work in
the areas discussed above, and if they publish their results with
due attention paid to texts and methods used, together with full
description of the students involved, we may acquire a solid set of
empirical data support and help to develop pedagogical practice.
We now turn from general considerations of theory and how this
might be developed to a particular area in the pedagogy of read­
ing that we feel is in urgent need of investigation by teachers,
testers and researchers. In our discussion of how to investigate the
relationship between the teaching of grammar and its impact on
reading ability, close attention is paid to the definition of terms as
the authors feel this to be the sine qua non of acceptable research.
 Future research 255
5.2 READING AND THE TEACHING
OF GRAMMAR
Introduction
As the research topic related to the teaching of reading, we are
suggesting an investigation of the relationship between the teach­
ing of reading in the classroom and the teaching of grammar. We
have chosen this topic for a number of reasons. First, we do not
believe it has ever been systematically investigated. We share
Bernhardt’s surprise that there has been so little work in this area
(Bernhardt, 1991a). Secondly, it would probably be generally
agreed that the processing of syntax, as part of the wider pro­
cessing of written information, is a ‘low level’ skill, and, as Eskey
(1988) has argued, such skills have tended to be neglected in the
teaching of reading in a foreign language. Thirdly, it seems to us
that many FL teachers have an interest in and knowledge of gram­
mar, so that the topic should be accessible to them.

A pedagogical approach
We should make clear that, in proposing this topic, we are not
querying the view that knowledge of grammar plays a part in the
reading process. All the models of reading reviewed in Section 2.1
either assert or assume that syntactic knowledge is a component,
and we see no reason to question this. What we are proposing is
an investigation into whether the teaching of grammar in the L2
classroom has a discernible effect on the students’ ability to read
the L2. The research is thus pedagogically focused.
Having chosen this research topic, the most obvious focus of
interest concerns whether a conscious, taught knowledge of gram­
mar has an effect, hopefully beneficial, on students’ reading per­
formance. There are, however, other areas of interest involving
grammar which it would be worth while to examine, e.g. whether
there are particular areas of grammatical knowledge which seem
to correlate positively with reading performance.
What follows is not intended to be an outline of a specific
research programme, rather an indication of how we think re­
search might be carried out in a particular area. Because we have
criticised other writers for lack of clarity in the use of terms, we
256 Reading in a Second Language
shall spend rather longer than perhaps expected in discussing
terms, particularly the term ‘grammar’. We are strongly of the
opinion that, unless the researcher has a clear idea of what is
meant by this term, any results obtained will be immediately open
to criticism.

Grammar and the reading syllabus

Over the past fifty years, the relationship between grammar and
reading in the EFL syllabus, at least, has tended to resemble the
man and woman in the old-fashioned devices forecasting weather:
when one is in, the other is out, and vice versa. In the era when
structural linguistics was an important influence on L2 theory,
skills such as reading were hardly considered to form part of the
syllabus. The assumption was that students should be taught the
structures of the language, or at least the most important of them.
Then they would, of their own accord, presumably, use their know­
ledge of these structures, together with their knowledge of vocabu­
lary (which was also a rather down-played area) to speak, write,
read and listen. Reading, if considered at all as an issue, was seen
as a matter of transfer, whether in the LI or L2 domain. This is
the position adopted by Fries (see p. 22). When a structuralist
such as Lado approached the question of language testing, his
first concern was to divide the language tested into three linguis­
tic areas - phonological, lexical and syntactic - together with a
catch-all area referred to as ‘cultural’ (Lado, 1961, cited in Baker,
1989). And while Lado then proceeded to describe four areas of
language use - namely, reading, writing, listening and speaking -
structuralism had little or nothing to contribute to use areas. It is
true that Fries did make a contribution, by describing the parallel­
ism between phonemic and graphemic contrasts; this is, however,
basically all the contribution that he could make which was par­
ticular to reading. Structuralism had (and has) little or nothing to
say about such topics as levels of information, inferencing, strat­
egies, or the part played by background knowledge.
As Baker (1989: 37) points out, tests in the structuralist tradi­
tion gradually developed into
a ‘diluted’ version of the original in which the rigid and system­
atic testing of elements at each level has given way to a more
homely collection of categories: ‘reading comprehension’, ‘grammar’,
‘vocabulary’, etc.
 Future research 257
The same development was true of the teaching syllabus, with
‘structures’ forming the core, and other elements such as ‘read­
ing comprehension’ added without any theoretical justification in
terms of linguistics. Much the same might be said of teaching mater­
ial which remained uninfluenced by the structuralist tradition; we
often find in textbooks passages labelled ‘reading comprehen­
sion’, along with some test questions, accompanied by grammatical
and vocabulary exercises. There is often little obvious relationship
between the text and grammar. At best, and this is by no means
always the case, the text can be seen as supplying data for the
grammar.
When communicative language teaching became fashionable,
the study of syntax as a separate component was downgraded for
at least two reasons: (a) with a concentration on skills, syntax, not
being classed as a communicative skill, tended to be omitted; (b)
it was argued that since all the skills incorporated syntax, a separ­
ate course com ponent was unnecessary. Talking about Commun­
icative Language Teaching (CLT) in general, James, in Johnson
and Porter (1983: 110), says:
The teaching of structure was condemned early on. Indeed, many
of the CLT persuasion saw it as axiomatic that they should not
teach the structures of the target language (TL) but something
else, something rather less tangible, usually its ‘functions’. If you
think this an exaggeration, recall that an early CLT course, Abbs
et al. (1975), was condemned as being ‘cryptostructural’ simply
because it contained discernible structural content.
For syllabuses focusing exclusively on macro-skills areas such as
reading and writing, the position taken by those criticised by James
is intellectually tenable. Alderson (1993) claims that there has
recently been a move back to teaching grammar, but the best way
of combining grammatical elements with reading in a syllabus still
needs considerable thought.
The meaning of ‘grammar’
We have been talking as if ‘grammar’ was an unambiguous term.
This is not, however, the case, as will be discussed below. The
problem can be illustrated initially by reference to an example of
a grammatical test item from Heaton (1988: 9):
A: .................... does Victor L u o ..........?
B: I think his flat is on the outskirts of Kuala Lumpur.
258 Reading in a Second Language
There are a num ber of problems about this item viewed as a test
of grammatical knowledge. The obvious way of completing A’s
utterance is:
‘Where (does Victor Luo) live(?)’
and no doubt this is the response favoured by the tester. But the
question arises as to what kind of knowledge is involved in pro­
ducing this answer. Even if we accept the item in the spirit in
which it is offered, it is hard to see how a response like ‘Why does
Victor Luo commute?’ could be rejected. A much more general
objection is that the semantics of the lexical items plays an import­
ant part in the formulation of the ‘correct’ answer: one could
argue that on strictly ‘grammatical’ grounds the selection of any
‘wh-’ interrogative word to fill the first gap, an d/or any intransit­
ive verb, e.g. eat, drink, cope, ululate, to fill the second gap, would
be difficult to reject. Probably most important of all, any appropri­
ate completion of A’s utterance involves recognition that the two
utterances form an adjacency pair, and that the first part must be
completed in such a way that it forms an appropriate question to
which the second utterance is the answer. This, however, is not
what is normally considered to be ‘grammar’. If we used a number
of such items as a ‘grammar’ test in our research, we would be
very vulnerable to the charge that what we were testing was a lot
more than just grammar’.
There are two well-known uses of the term ‘grammar’ in mod­
ern linguistics. The first use is the traditional one adopted by
Crystal (1997), namely ‘j ust one branch of language structure,
distinct from phonology and semantics’ (p. 89). This is the use
probably most familiar to teachers in L2 and is the one adopted
here. The other use is that of Chomsky and other writers in the
transformational tradition, where it means all the systematic pat­
terns of language, e.g. Cook (1988: 29):
. . . the grammar must show how the sentence is pronounced - the
sequence of sounds, the stress patterns, the intonation, and so on;
what it actually means - the individual words, the syntactic struc­
tures, and so on; and how these are related via various syntactic
devices.
Such writers use the term ‘syntax’ as their equivalent to the tradi­
tional use of ‘grammar’ Here we shall use ‘grammar’, in the tradi­
tional sense, to refer either to syntax or to syntax and grammatical
morphology, what Quirk et al. (1972) refer to as ‘morpho-syntax’.
 Future research 259
We still are faced with the problem of deciding on the scope of
syntax, i.e. what comes under ‘syntax’ and is therefore to be in­
cluded in our research, and what is outside, a different language
level. Here we find problems since, while many linguists will agree
that language can be divided into different components, and even
on the components themselves in general, the boundaries be­
tween them vary between different models and between different
versions of the same model over time. Horrocks (1987: 24) points
out that ‘there is no general agreement about the location of the
boundary between morphology and syntax’. Since we are includ­
ing both under ‘grammar’, this may not be a major problem, but
where to place the boundary between syntax and semantics defin­
itely is one. Again, Horrocks (1987: 30) remarks:
There is thus every reason to expect that the ‘borders’ between
different components will be drawn in different places at different
times and that a phenomenon confidently described as syntactic
will come to be regarded as semantic or lexical at a later date.
Examples include the inclusion in the ‘Aspects’ model of trans­
formational grammar (Chomsky, 1965) of selectional restrictions,
so that a sentence like ‘*John elapsed’ would be considered as
grammatically - as opposed to semantically or lexically - ill-formed;
case grammars contain far more ‘semantic’ information than other,
more syntactically based models; a trend over a num ber of years,
in several models, has been the inclusion in the lexicon of opera­
tions which would previously have been left to the syntactic com­
ponent (see Brown, 1984).

Defining syntax
Such divergences can also be found in pedagogical texts. Some
pedagogical grammars, e.g. Allen ( 1974), Berman ( 1979), Shep­
herd et al. ( 1984) include ‘Reported Speech’ among obviously syn­
tactic structures such as the Passive, Defining/Non-defining clauses
(Berman), Infinitive and Gerund (Allen). By contrast, in Morgan
and Batchelor ( 1959) ‘Direct and Indirect Speech’ is a separate com­
ponent from ‘Grammatical Study , being placed beside ‘Punctua­
tion and ‘Style and Vocabulary . Thus in the early textbook, Reported
Speech is clearly (and correctly) not being viewed as a strictly syn­
tactic phenom enon. More generally, the grammar by Downing
and Locke ( 1992) contains far more ‘communicative’ information
260 Reading in a Second Language
than does that of Quirk and Greenbaum (1973) which, while
eclectic, is more determinedly structural.
Definitions
Sampson (1975) defines syntax as ‘. . . how words are put together
to form sentences’. According to Horrocks (1987: 24):
Syntax is concerned with the principles according to which words
can be combined to form larger meaningful units, and by which
larger units can be combined to form sentences.
For Crystal (1997) syntax is ‘the way in which words are arranged
to show relationships of meaning within (and sometimes between)
sentences’. As aspects of sentence syntax, he mentions hierarchy,
grammatical function, concord, and transformations (pp. 94-7).
Even in these brief definitions there appear to be some significant
divergences. For example, Horrocks, to some extent, and Crystal,
in particular, seem to attach much more importance to meaning
than does Sampson. Crystal’s inclusion of relationships between
sentences would be rejected by many. Hence it may be useful
to examine, rather than definitions, writers’ accounts of general
aspects of syntax.
General principles
Bolinger (1975) considers that ‘the first rule of syntax. . . is that
things belonging together will be together’. As well as this prin­
ciple (which is certainly true for English, but seems less true for,
say, classical Latin), he cites, as coming inside the scope of syntax,
operators, both grammatical morphemes and function words, the
structure of phrasal constituents, word classes (even though he
claims the classes are ‘basically semantic’), grammatical, logical
and psychological functions (e.g. ‘subjects’, ‘objects’, etc.), the
grammatical functions of sentences (e.g. declarative, imperative,
transformations), and ‘higher sentences’ (e.g. performatives).
Horrocks and Crystal both refer explicitly to meaningful syn­
tactic relationships. As will be seen later, just how much meaning
is involved in syntax is an important question for the type of
research we are investigating, and will be reviewed later. In Halliday
and Hasan (1976) there is a suggestion that the difference between
grammatical and lexical meaning is a question of generality, gram­
matical items being more general than lexical ones, e.g.
 Future research 261
The general words (e.g. ‘thing’, ‘person’, ‘idea’) . . . are on the bor­
derline between lexical items and substitutes.
(p. 280; our parentheses)
Bolinger returns on more than one occasion to the notion that
grammar (syntax and morphology) is primarily concerned with
intra-language relationships (an example is the role of the com-
plementiser ‘that’ which functions to show that a following clause
is embedded), whereas semantics is concerned with ‘real-world’
relationships. A similar notion can be found in Halliday and Hasan,
where reference relations are ‘semantico-pragmatic’ while substitution
acts at the Texico-grammatical’ level (pp. 88ff). Bolinger, however,
points out that the distinction is difficult to maintain. In the phrase,
‘J ane’s house’, the possessive morpheme refers to a relationship
in the real world: Jane either owns or occupies the house. In the
phrase ‘J ane’s cooking’, on the other hand, the same morpheme
serves to show that ‘J ane’ is the grammatical subject of the verb
‘cooking’. Hence the distinction, while undeniably there, is of
little use in determining what is syntactic and what is semantic.
General principles, while interesting, appear to be insufficient
to decide with some degree of precision what does or does not
constitute ‘grammar’ or ‘syntax’. Faced with this problem, we have
two choices: either we can define, say, ‘syntax’ operationally, saying,
in effect, ‘this is what we consider to be syntax for the purposes of
this research’, or we can adopt a particular model, saying, in effect,
‘this is a model of syntax; anything described by the model there­
fore comes within the bounds of syntax’.

A formal model
At this point we ought to come clean and state that what we are
really looking for is a ‘formal’, ‘structuralist’ model, with as little
recourse to ‘meaning’ or ‘communicative value’ as possible. Given
the general popularity in recent language teaching of approaches
stressing meaning or communication, this would seem a rather
strange, even perverse, approach to take. And in taking it, we are
not, of course, decrying the importance of meaning or of com­
munication. We have, in fact, two reasons for our preference. The
first is related to what we discussed earlier with reference to the
Heaton example: unless we are very clear about what we mean
about grammar, our research will be open to criticism that it
262 Reading in a Second Language
incorporated a lot more than just grammar. We believe that adopt­
ing a formal approach minimises this danger.
Our second reason will become clearer as we progress, but
concerns the fact that we see one aspect of the research as cor­
relating performance on grammar/syntax tests with performance
on reading tests. It seems clear that the more text-focused or
‘communicative’ our grammar model is, the closer our grammar
tests based on this model will be to tests of low-level reading skills.
There is, however, little point in correlating such tests with read­
ing tests: little can be learned by correlating A with A. Therefore,
we would argue, we should begin, at least, with seeing whether we
can find a correlation between linguistic competence, measured
by a test of formal syntax, and linguistic and communicative per­
formance, measured by a reading test.

Choosing a model
We are looking, then, for a formal model of grammar, one which
sets out to describe the permissible (grammatical) sequences of
words or formatives in the sentences of whichever language we
are concerned with. However, we quickly run into a major prob­
lem: while there is no shortage of theoretical formal grammars,
particularly in the Chomskyan tradition, they have evolved to a point
where they have major disadvantages for the classroom experi­
menter, being either very difficult and abstract, and hence inac­
cessible, or insufficiently developed at a descriptive level, or both.
Pedagogical grammars, on the other hand, are predictably less
abstract, and are far more accessible to teachers and researchers.
However, for understandable pedagogical reasons, they do not
limit their description to syntactic structures. For example, in A
University Grammar of English, Quirk and Greenbaum (1973) pro­
vide a quite detailed description of formal aspects of the English
verb phrase, but accompany this description with information about
the use of verb forms, e.g.:
In indicating that the action is viewed as in process and of limited
duration, the progressive can express incompleteness even with a
verb like stop whose action cannot in reality have duration; thus the
bus is stopping means that it is slowing down but has not yet stopped.
The progressive (usually with an adverb of high frequency) can
also be used of habitual action, conveying an emotional colouring
such as irritation. (p. 41)
 Future research 263
Similarly, the Cobuild English Grammar (Sinclair, 1990), and related
volumes, combines very formal lists of verb patterns with distinctly
non-formal information such as:
You can also use ‘be possible to’ with ‘it’ as the subject to say that
something is possible. You usually use this expression to say that
something is possible for people in general, rather than for an
individual person. (p. 239)
Thus theoretical grammars are formal, but inaccessible, while ped­
agogical grammars are accessible but contain much non-structural
information.
As a compromise, we propose that researchers allow a theor­
etical grammar to set the limits of formal syntax, while filling in
the descriptive details from pedagogical grammars. If we choose
Government/Binding Theory, for example (see Horrocks, 1987),
the Base Component, which includes the Lexicon, will define as
syntax the following:
(i) The structure of lexical phrases, and of the sentence, includ­
ing embedded sentences.
(ii) Transformations relating, say, Interrogative and other Wh-
constructions to declarative structures.
(iii) A huge variety of syntactic restrictions included in the Lex­
icon. In addition, the morphological component will include
information about the parallelism of structures such as ‘New­
ton developed the theory’ and ‘The development of the theory
by Newton’.
It is, of course, a major feature of G/B Theory that it contains a
num ber of subtheories designed to filter out deviant sentences
which the Base Component has allowed to be generated. On the
whole we do not think these subtheories are relevant to our needs,
so we do not detail them.
With the fundamentals of syntax thus outlined by the theory,
we can now go to the pedagogical grammars to flesh out the des­
criptive detail. A University Grammar of English appears to be very
suitable for details of lexical phrases and embeddings, while the
Cobuild English Grammar, or the Cobuild Grammar Patterns 1: Verbs
(Francis et al., 1996) appears highly suitable for verb patterns.
This combination description can then be used, at least initially,
as the basis for grammar tests in the research outlined below.
264 Reading in a Second Language
Correlational studies
The easiest way of investigating whether a relationship exists in L2
between reading and grammar is to measure students’ performance
on tests of grammar and reading and then correlate the results.
Alderson (1993) did this with a specially constructed ‘grammar’
test and various modules of the IELTS test then being constructed.
He found, in general, high correlations between the grammar test
and the different modules. In fact, correlations were high between
virtually all the tests:
The relationship between Reading and Listening is as close as or
closer than the relationship between one reading test and other
reading tests! (p. 213)
He concludes that ‘the results, then, appear to show that a (vaguely
defined) generalized grammatical ability is an im portant compon­
ent in reading in a foreign language’ (p. 218). It is im portant to
stress that correlational studies of this sort do not point to a causal
relationship. Alderson’s results, as he himself makes clear, could
be interpreted as meaning either that grammatical ability improves
reading, or that reading ability improves performance on a gram­
mar test, or that the relationship is the result of a third factor,
which he terms ‘language proficiency’. Indeed, given the high
correlations between all his tests, this last might seem to be the
most likely explanation. Moreover, from the point of view of the
research being suggested here, correlational studies involve test­
ing but not necessarily teaching. Alderson’s tests are proficiency
tests, unrelated to any particular teaching syllabus. Nevertheless
such studies do highlight certain aspects of the problem which we
shall examine before moving on to more pedagogically oriented
studies.
First, the test of grammar used must, as far as possible, be just
that, i.e. it must relate to a clear definition of what constitutes
grammar; hence the extended discussion above. Alderson’s test
consisted of six subsections: (i) vocabulary; (ii) morphology; (iii)
prepositions, pronouns, etc., along with rather vaguely termed
‘lexical sets’; (iv) verb forms, etc.; (v) transformations; and (vi)
‘reference and cohesion’. In our definition of ‘grammar’, subsec­
tion (i) must be eliminated; the ‘lexical sets’ of subsection (iii)
are doubtful. As far as subsection (vi) is concerned, ‘reference’ is
normally treated as a form of cohesion; the latter, if it concerns
 Future research 265
relationships outside the sentence, will not be classed as syntactic,
while identifying pronominal links between cohesive items and
their referents, etc., has more to do with pragmatics than syntax.
Secondly, as Alderson makes clear, the grammar and reading
tests should be as separate from each other as possible. Since most
grammar tests involve the students in reading, this is not an easy
task. A grammar test can be viewed as a specialised reading test.
However, there are a num ber of steps we can take to reduce the
resemblance. We have already said that syntax, in most defini­
tions, is sentence-bound. Comparatively little written text is sim­
ilarly sentence-bound. Therefore we can achieve some measure of
difference by making our grammar test consist of decontextualised
sentences or phrases. This will tend to rule out the use of continu­
ous text, as in cloze procedure, or the gap-filling of continuous
text recommended by Heaton (1988) and used by Alderson.
It was largely this wish to have the widest possible distance be­
tween grammar and reading that influenced the discussion above
in the direction of strictly formal, as opposed to more functional
or semantic, grammars. Alderson classes the test he used as ‘com­
municative’ on the grounds that ‘we wished to test a student’s
ability to process and produce appropriate and accurate forms in
meaningful contexts’, justifying this on the grounds that ‘the ability
to manipulate form without attention to meaning is of limited
value and probably rather rare’ (p. 218). In general we would agree,
but consider that this argument is irrelevant for our research; in
reading we would require a student to process appropriate forms
in meaningful contexts. What we are investigating is whether a
knowledge of formal syntax is of help in this activity.
Alderson refers to ‘meaningful contexts’ and the undesirability
of teaching ‘form without attention to m eaning’. The question of
meaning is a particular problem in the context we are discussing.
Since reading involves the extraction of meaning, it is clear that
any grammar test involving a heavy emphasis on meaning is likely
to overlap with reading tests. When listing a num ber of grammar
test-types, Heaton remarks that they test ‘the ability to recognize
or produce correct forms of language rather than the ability to
use language to express meaning, attitude, emotions etc.’ (p. 34).
However, one of his types involves matching sentences like ‘Tom
ought not to have told m e’ with possible paraphrases. From the
viewpoint of G/B Theory, this looks like the interface between
syntax and semantics. Significantly, Heaton (1988: 35) remarks
266 Reading in a Second Language
that ‘such an item may be included either in a test of reading
comprehension or in a test of grammar’. We would like to restrict
such items either to grammar tests or to reading tests, and would
suggest that if they are included in grammar tests, care should be
taken to exclude inferences, together with the referential and
sense meanings of lexical items.

Beyond correlations
Correlations, if they exist, between performance on syntactic and
reading tests are interesting and worth while examining. How­
ever, the research design touched on above involves what have
traditionally been termed proficiency tests, i.e. they are not con­
structed with reference to a particular syllabus. This conflicts with
our stated aim at the beginning of this section, namely to investig­
ate the relationship between reading and a taught syntactic com­
ponent. In addition, as pointed out above, correlations do not
establish causes, only relationships. If we want to test whether a
taught grammar course has a measurable effect on subsequent
reading performance, we need a different research design.
A pedagogically more relevant research design would involve
comparison between one group, the experimental group, and an
equivalent group, the control group. Both groups are given a read­
ing pre-test at the beginning of the experimental period. Ideally
there should not be a significant difference between the means
of the two groups, but this is not strictly necessary. During the
experimental period, which would probably extend over a term or
semester, the experimental group is taught grammar, according
to a grammar syllabus based on the topics outlined above. Normal
reading instruction could continue, but should be kept separate
from the grammar instruction. During this time the control group
gets the same reading instruction, but is not given the grammar
classes, having some other activity, preferably not related to read­
ing, put in their place. At the end of the experimental period,
both groups are given a reading post-test. The null-hypotheses are:
(a) if there was no significant difference between the groups on
the pre-test, there should be no significant difference on the post­
test; (b) neither group should show a significant improvement in
scores on the post-test as compared to the pre-test. If, on the other
hand, either of the null-hypotheses was overturned, this would be
 Future research 267
evidence that grammar teaching had the effect of improving read­
ing performance. We should perhaps add, as a third possibility,
that a significant effect may be found, but will be negative, i.e.
that the grammar teaching had a detrimental effect on reading.
Care would have to be taken that the grammar component was
taught as formal grammar, i.e. that ‘communicative’ elements were
excluded as much as possible. Given the emphasis placed now­
adays (probably rightly) on the value of such communicative em­
phasis, this might be a problem for some teachers. However, our
own experience has been that students are quite receptive to formal
grammar, so this may not be a major problem. It should be noted
that in this research design, a grammar test is not strictly necessary,
though it could be included.
A number of such experiments, conducted with different groups
of students in different places, should establish whether the teach­
ing of formal syntax, at least, had a beneficial effect on reading
performance.

Refilling the grammar syllabus

So far, we have not attempted to distinguish different items of
what we earlier defined as syntax. In the research projects out­
lined above, we have assumed that any item on our list is a valid
one for inclusion in a syntactic test or syllabus. However, it may be
that, for reading, certain parts of the syntax component are more
important than others. Discussing the relationship between lan­
guage elements and communication, Heaton (1988: 10) comments:
There is also at present insufficient knowledge about the weight­
ing which ought to be given to specific language elements. How
important are articles, for example, in relation to prepositions
or pronouns?
We can focus Heaton’s general query on reading in particular.
For adequate comprehension of the quotation above, how import­
ant is the presence of the single article ‘the’? In fact, how import­
ant are articles at all for reading comprehension? If, following
H eaton’s line, we were to eliminate prepositions from the quota­
tion, then it becomes much more difficult to read (admittedly the
original quotation contains six prepositions as opposed to one
article). To pursue this argument would be to attempt to devise a
268 Reading in a Second Language
grammar syllabus focused on reading, as opposed to, say, one
focused on writing or speaking, a receptive as opposed to productive
syllabus.
It has to be admitted that producing such a syllabus would be a
fairly impressionistic endeavour. (It is worth while also pointing
out that the grammar syllabuses to which we are accustomed in
textbooks may well be biased towards written language already
(cf. Brown and Yule, 1983), and that, in terms of language teach­
ing in general, a really innovative aim would be to produce a
grammar dedicated to spoken language.) However, it may well be
worth the attempt to devise a grammar syllabus specifically with
reading in mind. First thoughts might include decisions to give
priority to prepositions over articles, declaratives over interrogat-
ives, and simple as opposed to continuous verb forms. More gen­
erally, such a syllabus might include a heavy focus on constituent
structure of phrases, clauses and sentences. Given the suggestion,
often made, that a primary task in sentence-processing is to find
subject and verb, exercises practising this and similar tasks might
well be made prominent. An alternative, or additional, focus might
be placed on distinguishing heads from modifiers and comple­
ments. Given what has been said above about the role of syntactic
restrictions in the lexicon, yet another focus might be to focus
more on vocabulary from a syntactic perspective, directing the
students’ attention to, for example, the types of complement taken
by a wide range of different verbs.

Grammar and texts

The suggestion made above that declaratives and simple verb forms
are more relevant than interrogatives and continuous forms is
based on the fact that, in the case of, say, expository study texts,
interrogatives and continuous forms are comparatively rare. This
raises the question of whether a grammar course could be focused
on a particular written genre. We claimed in Chapter 3 that the
grammar exercises commonly found accompanying texts in conven­
tional textbooks often showed little sign of having been designed
to help the students read the texts. It remains to be seen whether
a more focused approach would have a wider and transferable
effect.
 Future research 269
Grammar and different kinds of reading
Inspection of the reading model we put forward in Chapter 2
suggests that grammatical competence will be involved in differ­
ent degrees in different kinds of reading. In the Just and Carpen­
ter model on which our model is based, each sentence is processed
grammatically before the next sentence is accessed. In our specu­
lative description of different reading behaviours, we suggested
that this need not always be the case. With scanning, for example,
it seems possible that no grammatical processing need take place
at all. All that is necessary for success in extreme cases is for a
word or phrase to be recognised. At the other end of the scale,
careful local reading would seem to depend heavily on successful
syntactic processing. Careful global reading might seem to come in
between these extremes, since the larger part played by back­
ground knowledge might compensate for at least some local gram­
matical processing.
A teacher or researcher could investigate these hypotheses by
organising a syntax course, followed by a series of tests aimed at
testing students’ performance on different kinds of reading. It is
suggested at the level of test design that different tests should take
place at different times; the use of a single test requiring a number
of different reading types would probably result in contamination
of results. The aim would be to measure correlations between a
syntax test and tests of scanning, careful local reading, etc. The
use of a control group, i.e. another class or group of students,
would not be necessary here since the main focus of interest would
be on differences between correlations. If our hypothesis is cor­
rect, and grammar affects different types of reading differentially,
then we would expect a score on scanning to correlate least with a
score on the syntax text, etc.

Conclusions
Grammar is a component of reading which has been almost
ignored in the research. It seems to us that this is an interesting
and potentially valuable research area which L2 teachers and
applied linguists are in a good position to investigate.
270 Reading in a Second Language
5.3 THE USE OF TESTS TO INVESTIGATE
COMPONENTIALITY IN READING RESEARCH
In the previous section we have stressed the importance of defin­
ing terms clearly and adequately as the basis for valid and reliable
research. In addition to a clear idea of what is to be investigated,
research needs to be credible to an outside audience in terms of
design, sampling, methodological procedures, analysis and report­
ing. It should be logical with clear progression from research
questions through data collection, analysis, conclusions and recom­
mendations. Crucially it should be systematic with clearly stated
procedural rules not only to guard against threats to validity and
reliability but also to allow replicability by other researchers.
Davies (1990) has described language testing as the cutting
edge of applied linguistics. He supports this argument by suggest­
ing that one of the single most effective measurement tools for
exploring the nature of language proficiency or language acquisi­
tion is the language test. We share the view and believe that lan­
guage tests offer a reliable and rigorous means for exploring the
componentiality of reading though we are also well aware that
there are a num ber of qualitative procedures that can produce
data that complement those generated by language tests.
Language tests tell us little of the processes that underly read­
ing and we need to employ different methodological procedures
to investigate these. In particular, introspective methods may help
shed light on underlying thought process (see Cohen, 1987; Cohen
et al., 1979; and Rankin, 1988). Olson et al. (1984) point out a
num ber of problems associated with the method, such as the time
taken to administer and analyse, limited sampling and sensitivity
to instructional variables. However, methods such as introspection
and retrospection may offer insights into the perceived processes
that take place during different types of reading and help us
understand the nature of the differences in processing as well as
the existence of such differences.
In this section we are limited by space. Our concern is thus
with the latter, i.e. componentiality, the issue of the divisibility of
the construct, rather than unpacking the mental processes. The
discussion below is, accordingly, for the most part limited to test­
ing. This should not be taken to signify that we consider other
methodological procedures, or the investigation of process, as being
any less important.
 Future research 271
Research is only as good as the tools that are used to operation­
alise constructs. Inadequacies or limitations in these will constrain
the value of any research. In order to investigate the componential-
ity of reading systematically, we need to develop maximally valid
operationalisations of what we believe to be the im portant ele­
ments of that construct in the form of texts and associated tasks.
This is not just the concern of researchers. Anyone who teaches
reading in the language classroom is putting into practice his or
her view of the construct of reading every time reading-connected
activities are carried out by their students. Any of these activities
are open to investigation to evaluate their worth in relation to
impact on students’ reading abilities.
We propose below a number of a priori and a posteriori proced­
ures, which constitute a systematic approach to investigating the
componentiality of the reading ability of students principally
through testing. They should help illuminate whether reading is a
unitary activity or whether it is made up of separable components,
for example: expeditious types of reading as in search reading,
skimming, scanning for specifics, and careful reading at the global
and local levels. They will tell us about the relative contribution of
the posited skills and strategies to the overall picture of a student’s
reading ability. They will tell us about the relationship between
these components and inform us of the relative weaknesses and
strengths of our students. W hether for formative or summative
purposes, such evaluation can impact on whole educational systems
as well as individual classrooms.
These procedures (mutatis mutandis) should be applicable to
the development of any reading test from national to classroom
level. Within the constraints of the classroom all of the proced­
ures may not be practical at one particular point in time, but
every teacher should be aware that, to produce the most accurate
picture of a student’s transitional performance in reading, all have
a contribution to make. The data from these procedures are all
grist to the construct validity mill. They can all shed light on what
it is we are measuring and how well we are doing this. The more
of these we can embrace in our research investigations into read­
ing the better founded might be our findings.
There are no short cuts in rigorous research. This does not
mean it is the preserve of the few or the well resourced. Small-
scale research systematically carried out can be synthesised to pro­
vide real advances. What is required, however, is a comprehensive
272 Reading in a Second Language
but common framework of description of what is to be invest­
igated available to researchers, teachers and testers as well as the
development of systematic procedures that allow full comparison
across studies. Language testing encourages explicitness in speci­
fication and through its potential systematicity offers the possibility
of generalising beyond a particular study.
An overview of the research methodology to investigate the com-
ponentiality of reading is presented below. The exemplification
is from our investigations into EAP reading in China (see Appen­
dix 1) and in Egypt (see Khalifa, 1997). However, the methods
and approach are generic and should for the most part apply to
all reading situations.
A METHODOLOGY FOR INVESTIGATING THE
EAP READING CONSTRUCT

Stage 1: Specification of the construct

a n d s k il l s and the c o n d it io n s under which these activities are

S t r a t e g ie s
performed might be established through:

■ Target situation analysis

■ Theoretical literature
■ Research literature
■ Document analysis: course-books/tests

Stage 2: Development of pilot tests to operationalise EAP reading

specification

2.1 A priori validation

2.1.1 Systematic mindmapping of appropriate texts
2.1.2 Produce pilot version of test(s)
■ Decide format
■ Allow for attrition in texts and items
■ Ensure intelligibility of rubrics
■ Empirically establish timing
■ Consider order of questions/process dimension
■ Check layout
■ Trial on small samples
■ Produce first draft of mark scheme
■ Moderate tasks and mark scheme in committee
 Future research 273
2.2 A posteriori validation
2.2.1 Trial on reasonable sample
2.2.2 Item analysis
■ Facility values
■ Discrimination
2.2.3 Estimates of reliability
■ Marker reliability
2.2.4 Estimates of internal validity
■ Internal consistency
■ Correlations
■ Principal component analyses
■ Level: Means, Mests and cross-tabulation
2.2.5 Estimates of external validity
■ Qualitative expert judgement of items
■ Qualitative introspection/retrospection by test-takers
■ Feedback from test-takers (interview/questionnaire)
2.2.6 Revise
■ Administrator’s instructions
■ Items
■ Timing
■ Rubrics
■ Mark schemes
■ Re-trial any new items

Stage 3: Research study proper

Administration of Revised Version of test(s) to a representative sample
of intended population.
3.1 Item analysis
■ Facility values
■ Discrimination
■ Internal consistency
■ Descriptive statistics, Mests, cross-tabulations
3.2 Estimates of internal validity
■ Correlations
■ Principal component analyses
■ Measures of level
3.3 Estimates of external validity
■ Correlation with other established measures of the construct
(including teachers’ estimates)
■ Feedback from test takers.
274 Reading in a Second Language
Specification of operations and performance
conditions in an EAP reading construct
There are a number of steps that we can take to try to ensure that
we come closer to achieving construct validity in our measuring
instruments in terms of mirroring the skills/strategies underlying
the various types of reading whose empirical existence we wish to
establish.
■We need to be clear about the nature of such type(s) of reading
in terms of a framework of operations and performance con­
ditions established from the theoretical and research-based
literature and document analysis.
■We have to establish, for selected text(s) we are considering
using in a test, what would constitute an understanding of that
text given closely defined purposes for reading it, arrived at on
the basis of target situation analysis.
■Finally we have to establish appropriate test development pro­
cedures to establish test items that were most likely to elicit
those reading behaviours in a testing mode.

Example
To establish a specification of operations and performance con­
ditions to be tested we pursued a num ber of avenues in the de­
velopment of the Advanced English Reading Test (AERT) for
undergraduates in China (see Appendix 1 for background to the
project and for details of the specification). The following tasks
were carried out:
■A needs analysis of reading in the EAP context through docu­
m ent inspection, interviews, group discussion and questionnaire
demonstrated the need for expeditious reading strategies/skills
as well as careful reading; for coping with longer (1000-3000 word
texts) as well as shorter texts (<1000). It emphasised the need to
select texts from journals and books rather than newspapers.
■A review of theories of the reading process and available research
findings (together with the needs analysis) showed the need for
embracing a wider view of reading which would take into account
expeditious reading as well as the more usual careful reading
and for considering comprehension at the global as well as the
local level.
 Future research 275
■An analysis of current learning tasks used in teaching materials
in reading English for Academic Purposes demonstrated the
need to go beyond the traditional concern with slow careful
reading to include tasks focusing on quick, efficient, selective
reading and raised the issue of prediction activities in relation
to activating existing schemata.
■An analysis of test tasks used in assessing reading English for
Academic Purposes showed, among other things, the importance
of providing a purpose for each reading activity; the importance
of controlling time spent on each activity; the importance of
establishing a minimal and maximal level of topic familiarity;
the types of format that lent themselves to the testing of the
various operations.
These data clarified the nature of reading operations across dis­
cipline areas in the Chinese academic context and led to the spe­
cification in Appendix 1. These investigations also provided data
on the conditions for reading activities in EAP. The specifications
for performance conditions in the Chinese AERT are also listed
in Appendix 1. Once appropriate operations and conditions are
established these have to be implemented in a test.

A priori validation
Textmap content of texts to establish content to be extracted
according to purpose for reading
Potentially appropriate texts for a test population should initially
be selected by a moderating committee from a bank of such texts
on the basis of as close a match as possible with the performance
conditions laid out in the specifications (see Section 3.3 for a
discussion of these conditions). The committee would at the same
time need to confirm that the texts selected allowed the testing of
the intended operations.
A practical method of doing this is to establish the content that
might be extracted from a text in line with the established pur­
pose for reading it. Various systems of text analysis are proposed
in the literature (see Section 2.2, text structure) but, though im­
pressive in their attention to detail and replicability/reliability, they
consume inordinate amounts of time and the end results do not
necessarily enable the researcher or test developer to decide which
276 Reading in a Second Language
are the important ideas in a text for testing. A more utilisation-
focused procedure is to try to establish the main ideas of a passage
through expeditious or careful ‘textmapping’ procedures (see
Buzan, 1974; Geva, 1980, 1983; Nuttall, 1996; Pearson and Field­
ing, 1991).
In each textmapping procedure an attempt should be made to
replicate a single type of reading on a single text, e.g. reading a text
slowly and carefully to establish the main ideas. The product of the
particular reading of a text can be compiled in the form of a
spidergram or as a linear summary. This is first done individually
and then, the extent of consensus with colleagues who have fol­
lowed the same procedure is established. The objective of the pro­
cedure is to examine whether what we have decided is important,
is in line with the specified type of reading activity and matches what
colleagues consider important (see Sarig, 1989, for an interesting
empirical investigation of this procedure).
This is a crucial first step in trying to ensure the validity of our
tests. We would be concerned that the answers to the questions we
then wrote revealed the important information in the text that
could be extracted by the particular type of reading being assessed.
An ability to answer the items should indicate that the candidate
has understood the passage in terms of successful performance of
the specified operation (s).
To illustrate the technique of textmapping and to demonstrate
how this can help to summarise in note form the products of
reading a text for a variety of purposes, we give an example in
Appendix 2 of the procedure for textmapping a text for developing
a test of careful reading.
The procedure would remain the same for other skills and
strategies but the key conditions of
■time allowed for the textmapping
■length of text
would alter in line with discussion on performance conditions
presented in Section 3.3 above.
The parts of the EAP reading construct in the exemplification
below are expeditious and careful reading at both global and
local levels (see Section 3.2 and 4.3 above for a discussion of these
issues). We would, therefore, want to test these strategies/skills
using different passages (to ensure the independence of items
and to avoid the possibility of muddied measurement). For this
 Future research 277
reason a different set of procedures is necessary for each to reflect
as closely as possible the processing behind each of those skills or
strategies.
It is im portant to note that the time for the textmapping task
provides a benchmark for the actual test time. All too often test
constructors take considerable periods of time reading and re­
reading texts and they peel off deeper and deeper levels of mean­
ing. They then give candidates 20 minutes or so to reach the same
depth of understanding under exam conditions. This is obviously
a nonsense. The candidates would not normally be expected to
find answers to questions in a shorter period than it has taken the
test setter. Conversely, if one wishes to test expeditious strategies
then the tim e/text length ratio should not allow test takers to
process in a careful non-selective fashion.
The textmapping procedure represents a principled way of
avoiding this particular threat to the test’s validity. If it can be
done with students who are at a suitable level of reading ability
drawn from the population who will eventually take the test, this
may be even more valid than using language or subject specialists
to perform this activity.

Produce pilot version of tests (s)

Decide format
Procedures such as textmapping should enable us to determine in
a principled fashion what we might wish to test. However, the
format, which acts as the vehicle for testing reading activities, may
constrain the operations and conditions we attempt to include.
Therefore, as well as carefully specifying the latter we need to con­
sider carefully the method we are going to use so as to minimise
the influence of method on measurement of the trait (see Section
3.4 above for a detailed discussion of methods available). The
cardinal rule remains, however. We must first decide what types
of reading we want to test and develop systematic procedures for
deciding the micropropositions an d/or macropropositions that
we would expect candidates to extract from texts in performing
these types of reading. Only then do we consider test formats and
decide which will most faithfully mirror the procedures and allow
the appropriate propositions to be extracted. No matter how neat
or efficient the format, if it compromises what we are trying to test
278 Reading in a Second Language
it must be rejected. Any threat to the validity of the measurement
must be resisted.

Write more questions, and use more texts, than you will need
The textmapping procedure will provide the content for each
section of the test. It will also show whether the passage is suitable
for its intended reading purpose. Where it is possible to produce
a consensus textmap this, then, needs to be converted into appro­
priate test items in the format selected. Where consensus is not
achieved or the textmapping produces too few items, these texts
must be rejected!
Wherever possible it is advisable to write more items than are
needed in case some of them do not work. One cannot tell in
advance of empirical trialling those items that will work and those
that will not. It is best to trial items on small numbers initially,
because it may well take two or three attempts before problems in
wording are resolved. Try the test on colleagues or a few students
at a time. If the test is piloted on all immediately accessible can­
didates to begin with, then this could be problematic. The import­
ant trialling on larger numbers should not take place until it is
reasonably certain that the items are working well and any obvious
problems have been eliminated.
Careful attention needs to be given to:
■Rubric: Are the instructions clear, accessible and unequivocal?
■Timing: Is the timing for each reading type on each passage
appropriate given the length of the text and the activities we are
expecting test takers to perform on it?
■Order of questions/process dimension: Do the order of the questions
set on a particular passage and the order of the reading types in
the test as a whole encourage the reading behaviours we are
hoping to test?
■Layout: Does the layout help the students to work efficiently
through each subtest; does it appear elegant and neat?

Produce first draft of mark scheme

It is im portant in the piloting to draw up a comprehensive mark
scheme of acceptable responses to questions, particularly if a short-
answer question format is employed, where variation of wording
 Future research 279
used in answers has to be accommodated. It is frustrating when
later administering the test to the sample proper if an acceptable
response arises which necessitates remarking hundreds/thousands
of scripts because it has not been included in the mark scheme.

Moderate tasks and mark scheme in committee

It is crucial to set aside time and elicit the co-operation of informed
colleagues in closely moderating the test and the mark schemes.
This is a key stage of the test development process and failure to
go through this procedure is a serious threat to a test’s validity.
We are too close to the tests, which we develop for research or
assessment purposes, and after a while we may not be able to see
the flaws.

A posteriori validation
Trial on reasonable sample
Once the necessary development preliminaries described above
are completed, it is important to trial the test on as broad a sample
of the intended population, in terms of ability, as possible and then
subject the results to statistical analysis to establish the test’s value
as a measuring instrument.
It is im portant at the trialling stage to administer the research
instrument to as normally distributed a sample as possible. This
might mean purposefully sampling from top, medium and lower
universities, institutions, schools and classes within these. Normally
distributed data allow the researcher to apply the statistical ana­
lysis recommended below to establish how the items in the subtests
are functioning. A skewed sample, where the majority of students
are too strong or too weak, will not allow the researcher to do
this. This is why samples of less than 30 are normally not recom­
mended. The distribution of scores achieved by the sample on the
test should allow two standard deviations to fit in either side of
the mean. In Figure 5.1 the test is out of 60, the mean is 32.9 and
the standard deviation is 9.91. So in this data set we can get two
standard deviations (2 x 9.91) easily either side of the mean. This
tells us that we have a distribution approximating to normal and
we can continue with the further analysis discussed below.
280 Reading in a Second Language
70-
60 ■

50-

40 ■

SO-

20 ■

Std. Dev = 9.91

10- Mean = 32.9
N = 303.00
o-
5.0 15.0 25.0 35.0 45.0 55.0
10,0 20.0 30i) 40.0 50.0
TOTSCORE
Figure 5.1 Histogram of a reading test.
Statistical analysis
Facility values (mean)
This is based on the percentage of people getting an item right
and is usually reported on a scale between 0 and 1 (see column
labelled ‘Mean’ in Table 5.1 (a)). If an item is too easy most people
will get it right, and if it is too difficult most people will get it
wrong. The variance in performance on these items is usually less.
This is indicated by the ‘SD’ column. Item 06 is too easy with 87
per cent of candidates getting it right and there is far less variance
associated with this item than the others (SD = 0.33). These items
will not usually contribute a great deal to the overall reliability of
the test. Item 06 in Table 5.1(a) would appear to require some
attention.
In test development it is conventional to use items whose facil­
ity values lie between 0.2 and 0.8. Having established the facility
value of each item the next step is to examine whether the items
discriminate between stronger and weaker candidates.
Discrimination
A discrimination index (see ‘Corrected item - total correlations’
in Table 5.1(b)) shows how well an item discriminates between
the strong candidates (based on overall performance) and weak
candidates on each item. The index is based on a scale from 0 to
 Future research 281
Table 5.1(a) Descriptives of AERT Section 1 Careful reading
Number of valid observations (listwise) = 303.00
Number of items = 15
Variable Mean SD Minimum Maximum Valid N
ITEM 01 0.73 0.45 0.00 1.00 303
ITEM 02 0.72 0.45 0.00 1.00 303
ITEM 03 0.36 0.48 0.00 1.00 303
ITEM 04 0.67 0.47 0.00 1.00 303
ITEM 05 0.57 0.50 0.00 1.00 303
ITEM 06 0.87 0.33 0.00 1.00 303
ITEM 07 0.71 0.45 0.00 1.00 303
ITEM 08 0.71 0.45 0.00 1.00 303
ITEM 09 0.51 0.50 0.00 1.00 303
ITEM 10 0.48 0.50 0.00 1.00 303
ITEM 11 0.64 0.48 0.00 1.00 303
ITEM 12 0.61 0.49 0.00 1.00 303
ITEM 13 0.26 0.44 0.00 1.00 303
ITEM 14 0.25 0.43 0.00 1.00 303
ITEM 15 0.69 0.46 0.00 1.00 303

1 and 0.2 is often taken as the cut-off point for acceptable dis­
crimination. Items 02 and 06 are not discriminating very well and
both would need to be considered carefully for exclusion.
Discrimination is important in developing the reading test bat­
tery because it demonstrates that an item can reliably discriminate
between a person who has the ability and a person who does not.
It may well be that, on samples of students who take the test later
because they are all good or weak, such discrimination is lacking.
This does not matter if the item has been shown to discriminate
in the piloting on a normally distributed representative sample of
the large potential population.
There may be cases where an item tests comprehension of an
im portant idea and we have to decide whether we can accept a
low facility value and a lower positive index of discrimination. A
negative discrimination index is never acceptable. The validity of
what we are testing must always come first. If the main idea of a
passage proves easy to extract, so be it. The text has been selected
as representative of the domain the respondent has to cope with
on a principled basis.
282 Reading in a Second Language
Table 5.1(b) Reliability analysis: scale (alpha)
Statistics for scale
Mean Variance SD No. of variables
8.7888 8.8095 2.9681 15
Item - total statistics
Scale mean Scale Corrected Alpha
if item variance item - total if item
deleted if item correlation deleted
deleted
ITEM 01 8.0627 7.9331 0.2690 0.6711
ITEM 02 8.0693 8.1442 0.1802 0.6820
ITEM 03 8.4257 7.8148 0.2832 0.6695
ITEM 04 8.1155 7.8177 0.2935 0.6681
ITEM 05 8.2145 7.6128 0.3481 0.6605
ITEM 06 7.9142 8.3039 0.2069 0.6776
ITEM 07 8.0792 7.8215 0.3072 0.6663
ITEM 08 8.0792 7.7354 0.3429 0.6618
ITEM 09 8.2739 7.8684 0.2459 0.6748
ITEM 10 8.3069 7.5843 0.3536 0.6597
ITEM 11 8.1452 7.7868 0.2960 0.6677
ITEM 12 8.1815 7.8643 0.2573 0.6730
ITEM 13 8.5314 7.8459 0.3146 0.6656
ITEM 14 8.5413 7.9776 0.2642 0.6717
ITEM 15 8.1023 7.6021 0.3869 0.6558
Reliability coefficients'. No. of cases = 303.0; Alpha = 0.6836;
No. of items = 15

Internal consistency
These estimates are often cited as indicators of reliability, but they
are just as useful in exploring the construct validity of a subtest.
They evidence that items within a subtest are measuring a con­
struct in a similar fashion. Table 5.1(b) provides some data on
the internal consistency of a pilot version of a subtest designed
to measure careful reading for main idea extraction. You can see
the internal consistency data in the column headed ‘Alpha if
item deleted’. The alpha for the data is given at the bottom of
Table 5.1 (b) at 0.6836. The ‘Alpha if item deleted’ column tells us
whether an item is contributing to the overall internal consistency
 Future research 283
of the test. In this case the overall alpha for this component would
not be improved by removing any item although it is notice­
able that items 02 and 06 are contributing the least among the
15 items.
The data on facility value, discrimination and internal consist­
ency can all help the researcher to take decisions on how to select
items which provide the best measure of the construct they wish
to investigate through the test(s).

Estimate of reliability
Marker reliability
It is also important to demonstrate that the data have been marked
reliably, otherwise the reliability of the results themselves will be
affected. It is usual to xerox a num ber of answer sheets, 30 plus,
and to have these marked by all the markers involved in the study.
Ideally markers should themselves receive the same set of sheets
at a later stage for remarking to establish intra-marker reliability.
The reliability of a test is a combination of its internal consistency,
inter- and intra-marker reliability. Formulae exist for combining
these to provide an overall reliability estimate which will help the
reader of the research to understand the extent to which one can
depend on the results.

Estimates of internal validity

Correlations
We have suggested that each part of the test should be designed
to operationalise a component of the reading process. When the
data are available from trialling, the totals for each subtest can be
correlated with each other to see the extent to which they overlap.
One might also expect that individual items should correlate more
with their own subtest than with other subtests. If subtests correl­
ate highly with each other (0.8 and above) this might offer an
initial indication that the constructs are overlapping.
To be sure of the relationship between items and components,
factor analysis provides a further method for giving a clearer pic­
ture of the structure of the underlying data. A commonly used
form of this is principal components analysis (PCA).
284 Reading in a Second Language
Principal components analysis
In Table 5.2 you will see an example of PCA that was carried out
on data on two reading subtests. Items 1-15 are careful reading
items testing global understanding in three passages (we looked
at the item analyses of these above) and items 46-60 are careful
reading items testing local comprehension in three passages.
The discerning reader will note that items 1-15 load positively
on the second factor, whereas items 46-60 all load negatively.
This is partial evidence supporting the view that the reading con­
struct may well differ in respect of these two components. Such
data indicate that the two types of reading the test designers in­
tended to test do appear to load on different factors, and this is
contributory evidence for the separability of these two skills.

Level: means, f-tests and cross-tabulation

Correlations tell us about the relationship, the degree of ‘over­
lap’, between constructs but they do not tell us about levels of
ability in the respective constructs. Two abilities may be distrib­
uted in a similar fashion across a population but proficiency in
one may be much higher or lower than in another.
Analysis does not always have to be highly sophisticated. The
purpose of statistics is to lay out the data set so as to facilitate
interpretation. We could examine the descriptive statistics related
to one subset of items and compare it with another. The mean
scores would tell us the levels obtained in both and t-tests would
tell us whether the differences were significant, i.e. if they had
been achieved by chance or whether they represented a real dif­
ference in performance.
In Table 5.3 two subtests (TOTCARE, items 1-15, and
TOTLEXI, items 46-60) are compared in terms of the means of
the whole sample (303). In this example the candidates are per­
forming significantly better on the global (TOTCARE) as against
the local comprehension items (TOTLEXI) by almost one and
a half points. Such overall means tell us how the population is
behaving.
In Figure 5.2 (p. 287) there is a cross-tabulation of the per­
formance at the individual level on TOTCARE, the careful read­
ing global items (1-15), as against TOTLEXI, the careful reading
local items (46-60), which clearly illustrates that respondents had
 Future research 285
Table 5.2 Principle component analysis
Careful global (Section 1) vs Careful
local (Section 5)
Extraction 1 for analysis: 1, Principal
Components Analysis (PCA)
PC extracted. 3 factors
Factor matrix:
Factor 1 Factor 2 Factor 3
Item 01 0.35419 0.07147 0.24566
Item 02 0.25198 0.02516 0.32407
Item 03 0.39938 0.04971 0.32812
Item 04 0.36344 0.19864 0.09796
Item 05 0.35125 0.41042 0.43283
Item 06 0.22666 0.13507 0.46428
Item 07 0.38603 0.19247 0.20150
Item 08 0.37549 0.34387 -0.09751
Item 09 0.20756 0.48369 0.19068
Item 10 0.47791 0.09945 -0.01514
Item 11 0.41513 0.03979 -0.05306
Item 12 0.35698 0.20657 -0.57466
Item 13 0.42686 0.12114 -0.16013
Item 14 0.31132 0.29858 -0.26518
Item 15 0.42710 0.42185 -0.42105
Item 46 0.43179 -0.33806 0.11272
Item 47 0.50645 -0.31203 0.11268
Item 48 0.30775 -0.09041 0.09677
Item 49 0.24747 -0.07246 0.26668
Item 50 0.33496 -0.23607 -0.01828
Item 51 0.42613 -0.13725 0.04570
Item 52 0.32176 -0.34385 -0.03447
Item 53 0.27668 -0.41138 -0.06872
Item 54 0.24263 -0.14900 0.06010
Item 55 0.37222 -0.31271 -0.14617
Item 56 0.47017 -0.11028 -0.10293
Item 57 0.38027 -0.03741 -0.18228
Item 58 0.38288 -0.30800 -0.01056
Item 59 0.51289 -0.01229 -0.19782
Item 60 0.41676 -0.02467 -0.07180
Final statistics:
Factor Eigenvalue % of Var. Cum. %
1 4.19728 14.0 14.0
2 1.77580 5.9 19.9
3 1.60594 5.4 25.3
286 Reading in a Second Language
Table 5.3 TOTCARE and TOTLEXI compared through a t-test
Variable No. of Corr 2-tail Mean SD SE of
pairs Sig Mean
TOTCARE 8.7888 2.968 0.171
303 0.442 0.000
TOTLEXI 7.3696 3.094 0.178
Paired differences
Mean SD SE of Mean t-value df 2-tail Sig
1.4191 3.203 0.184 7.71 302 0.000
95% Cl (1.057, 1.781)

greater difficulty with the latter than with the former. These cross­
tabulated data are summarised in Figure 5.3 which provides an
even clearer view of the potentially differing performance abilities
in these two areas. With a notional pass mark of 60 per cent (9/15)
substantially more would fail the local but pass the global careful
reading test (85) than vice versa (31).
If these data reported in the analyses above were to be repeated
using differing measures and different samples, then we might
begin to synthesise an argument for the divisible nature of the read­
ing construct.

Estimates of external validity

Internal statistical measures are necessary but not sufficient to
establish the nature of the reading abilities under investigation.
We need to get a closer idea of what is actually happening during
the test experience to accumulate evidence that the test is per­
forming in an ecologically valid fashion, i.e. in answering the items
the students are processing as the test developer/researcher in­
tends them to. For example, if test takers were using test-taking
strategies to avoid skimming or search reading through faulty item
construction, the test statistics would not necessarily tell us this.
We need to generate data on the process as well as the product.
We shall examine a number of ways in which this might be done
including survey, introspection and retrospection.
 TOTLEXI
TOTCARE 0.00 I 1.00 I 2.00 I 3.00 I 4.00 I 5.00 I 6.00 I 7.00 I 8.00 I 9.00 I 10.00 I 11.00 I 12.00 I 13.00 I 14.00 I 15.00 I Total

0.00
1.00 1 2 3
2.00 1 1 1 3
3.00 2 3 2 1 8
4.00 1 2 2 4 i 1 11
5.00 1 1 2 2 1 2 3 4 1 1 18
6.00 1 1 1 4 1 1 3 5 2 1 1 26
7.00 1 1 6 5 2 -2 5 2 1 1 28
8.00 1 1 2 2 4 2 5 7 1 1 40
9.00 1 3 5 *| 4 > V 3 4 2 38
10.00 1 L* 1 1 5 | j f?i 9 4 2 1 40
11.00 J If! ■ III 2 4 5 3 2 26
12.00 2 Ill 2i 4 3 3 1 3 2 25
13.00 I j II i 2 5 5 1 2 23
14.00 i -i 111 2 2 2 2 13
15.00 1 1
Total 2 4 10 24 22 24 28 41 36 33 29 24 12 5 9 0 303
Future research

Figure 5.2 Cross-tabulation of total careful reading and total lexical items.
287
288 Reading in a Second Language

TOTLEXI
TOTCARE 0 .0 0 ---------►8.00 9 .00 ------- ► 15.00

0.00
106 ■
1
8.00 I
9.00
i 81
15.00

Figure 5.3 Summary of cross-tabulation in Figure 5.2.

Structured feedback from test takers

The intentions of the test developer are always mediated by the
response of the test takers. Their attitude to facets of the test are
as important as the evidence arising from the statistical data as it
can often explain why things have happened in a certain way.
In a feedback questionnaire one might wish to establish test
takers’ perceptions of, for example,
■topic familiarity
■difficulties with language in a text
■domain specificity of content
■domain accessibility
■familiarity with formats
■value of formats
■sufficiency of time
■clarity of rubrics
■length of test
■whether test likely to achieve its objectives.
An extract of a pilot questionnaire used in the Advanced English
Reading Test (AERT) project in China is shown in Figure 5.4 (pp.
290-1).
Data from structured questionnaires of this type are import­
ant because they give us a broad based view of how the sample is
responding to the test. A lot of the features examined through
these sample questionnaires might impact adversely on the meas­
urem ent of the construct if we have not done a proper job at the
development stage. They act as a check on our ability to faithfully
implement the test specification. As well as this broad-spectrum
data we also need more in-depth information on our test items.
 Future research 289
This is provided by qualitative research procedures such as intro­
spection, retrospection and expert judgement. As well as data relat­
ing to students’ perceptions of the instruments, texts and tasks, we
are also interested in their views on what they thought the items
in the test were actually getting them to perform in terms of
the skills and strategies in our posited construct. Qualitative data
obtained from introspection, retrospection and questionnaire
survey would provide us with process information on what the test
takers thought the test was testing. This would usefully comple­
m ent the quantitative data obtained through test administration.

Introspection
An introspection study into the students’ process of reading texts
and answering the questions should be carried out to find out
what skills and strategies students are using in completing each
section of the test. A procedure for this was developed with Shang­
hai Jiaotong University staff in the AERT project, PRC. The students
were trained to think aloud onto tapes in a language laboratory
while taking the test. Before the test, a training session was necessary
to demonstrate what they were expected to do during the test.
Students should be allowed to use LI if appropriate if necessary in
their verbal reports. The data are then transcribed and content
analysed in terms of the test operations.

Retrospection
A separate retrospection study enables the researchers to obtain
a larger data set (than is possible through the time-consuming
spoken protocols) to establish student perceptions of the skills
and strategies used in the process of taking the test. This can be
carried out in the large-scale trialling of the test. It can be incor­
porated into the process of doing the test by providing a checklist
for candidates to tick after they finish each section of the test (see
Table 5.4).

Experts’judgements
Apart from students’ introspection and retrospection, language
testing experts and reading experts should be asked to give their
professional opinion of the constructs being tested. Table 5.4 is
 290

Figure 5.4 Extract from the test feedback questionnaire used in the AERT project in China.
QUESTIONNAIRE TO CANDIDATES

Nam e:_________________________________________________ University:___

Discipline:_____________________________________________ Candidate No:
1. About the passages
Passages !*aȣtu4*u<
&ti(k life Topic familiarit) Subject wpcsfK m Appropriate discipline
Reading in a Second Language

H M | L I M j H M L N ft M f • L j K \ AH ST ML

2
3

H = high; M = medium; L = low; N = no

AH = arts and humanities; ST = science and technology; ML = medical and life science
Figure 5.4 (Cont’d)
2. About time limits
Section More than enough Just enough Not quite enough Not enough at all
1 Careful reading
2 Skimming
3 Search reading
4 Scanning
5 Lexical items

3. General satisfaction with the test

High Medium Low No
Formats
Content
Length
Time
Future research

Design
Rubrics
291
Table 5.4 Student retrospection/expert judgem ent sheet
292

What skill/strategy do you think is tested in each section?

■Please double tick ( / / ) for the primary focus of the skill/strategy tested.
■Please tick ( /) for the secondary focus of the skill/strategy tested if you think there is one.
Section 1 Section 2 Section 3 Section 4
Primary Secondary Primary Secondary Primary Secondary Primary Secondary
Reading carefully
for main ideas
Reading in a Second Language

Reading quickly to
get the overall idea
of a text

Reading quickly to
search for information
on main ideas
Table 5.4 (Cont’d)
Reading quickly to find
specific information:
words/numbers/
symbols

Carefully working out

meaning of words
from context

Others (please specify)

Future research
293
294 Reading in a Second Language
an example of the type of proforma that might be used to collect
their opinion on the construct validity of the test (designed with
Jin Yan and Luo Peng for the AERT in China).

Revision
As a result of the qualitative and quantitative investigations de­
scribed above the researcher is well equipped to make any neces­
sary amendments to the pilot version of the research instruments
to make them more valid operationalisations of the intended
construct.
On the basis of the procedures discussed above we would have
sufficient data to help us revise our test instruments to ensure that
they come closer to performing the job intended.

Using the test to explore the construct reading

Stage 3: Research study proper
When the test instrument(s) have gone through the rigorous devel­
opm ent phase described above, we can use them to investigate the
nature of reading for the purposes and audience we have in mind.
We would administer the Revised Version of test(s) to a repre­
sentative sample of intended population and then subject the data
to the same procedures outlined above in connection with the earl­
ier trialling:
3.1 Item analysis
■Facility values
■Discrimination
■Internal consistency
■Descriptive statistics, Kests and cross-tabulations.
3.2 Estimates of internal validity
■Correlations
■Principal component analyses
■Measures of level.
3 .3 Estimates of external validity
■Correlations with other established measures of the con­
struct (including teachers’ estimates)
■Feedback from test takers.
 Future research 295
These investigations would help provide insight into the nature of
the reading construct as defined by the specification and opera­
tionalisation in the test(s). They should tell us about
■the unidimensionality or multidivisibility of the reading construct
under investigation
■the relative contribution the different parts of the test were
making to the measurement of an individual’s reading ability
■the relative strengths of the sampled population in the different
parts of the reading test
■the nature of individual differences in performance on each of
the components.

Conclusion
None of this is easy to follow through. Not all of these things are
possible for the teacher in the classroom except over an extended
period of time. What is clear, however, is that the more rigorous
and comprehensive we can be in our investigations, the clearer
the account that is likely to emerge of the nature of reading.
Clear specification of what we are trying to teach or test, and
soundly conceived methodologies for investigating components
and processes, are essential. It is hoped that this book is of some
value in this endeavour.

Notes
1. Richard Joung, November 1997.
2. One of the authors teaches British undergraduates, who regularly
assure him that the initial sequence [kn] is impossible to pronounce,
a judgement that might have seemed strange to their ancestors.
3. This is a broader definition than that used by Bernhardt, who restricts
her attention to examination of the reading process.
4. Only for a moment, though. The effect of such analogies is to give
a spurious impression of precision regarding fairly vague and ill-
understood processes.
This page intentionally left blank
 297

Appendix 1

THE DEVELOPMENT OF A SPECIFICATION

FOR THE ADVANCED ENGLISH READING TEST
(AERT) IN THE PEOPLE’S REPUBLIC OF CHINA
The Advanced English Reading Test (AERT) project was conceived
in the context of the teaching of reading in English for Academic
Purposes carried out in the third year of Chinese universities. It
was clear that the EAP reading was not receiving sufficient atten­
tion from university authorities, teachers and students as required
by the National College English Teaching Syllabus (NCETS).
The NCETS promulgated in 1985 on the basis of a nationwide
comprehensive needs analysis stipulates that the ultimate goal of
Chinese EFL learners at the undergraduate level is to gain access
to scientific and technical literature through the medium of Eng­
lish. To achieve the goal, an EAP reading course is required for
the third-year students who have completed the foundation stage
study. However, the course has never been accorded sufficient
attention and is sometimes neglected. The lack of an adequate
and appropriate assessment tool was seen as a major contributory
factor in this neglect.
The aim of the project is to develop an EAP reading test for
Chinese third- and fourth-year undergraduates. This should en­
able universities to monitor and evaluate students’ performance
in EAP reading over a long period of time. Furthermore, it is
hoped that it will bring about a much-needed beneficial backwash
effect on the teaching of EAP reading in China.
As we described in Section 5.3 (p. 274), the first step in test
development was to investigate the construct of EAP reading
through:
298 Appendix 1
1. a literature review of
(a) the componentiality of reading
(b) the ESP issue in reading;
2. a survey of Chinese undergraduates’ EAP reading needs; and
3. an analysis of EAP reading-teaching materials and test papers
in terms of task types and text types.
On the basis of these investigations, a taxonomy of EAP reading
skills and strategies, and the conditions under which these are
performed was specified. In Table A l.l, the conditions part of the
specification considered facets such as topic familiarity, text length,
propositional content, time constraints, and rhetorical structure of
texts; the operationalisations part of the specification in Table A1.2
comprised the expeditious reading strategies of skimming, search
reading and scanning and the skills of careful reading for local
comprehension and careful reading for global comprehension.
TEXT SELECTION: SPECIFICATION OF THE CONDITIONS FOR AERT
Here we are concerned with the conditions under which reading activities should be carried out. The needs
analysis, the survey of the literature, previous empirical studies, the analysis of existing teaching and testing tasks
for EAP reading, and discussions with the AERT management team and reading specialists in the UK have
indicated that the following performance conditions would need to be built into the test:

Table A l.l Requirements for selecting texts for AERT

Conditions Descriptions
Purpose (s) of reading To test students’ ability to comprehend academic texts and to extract important information
from those texts.
Nature of texts Texts written for a non-specialist audience with informative and interesting ideas.
Source of texts Chapters from textbooks, journal articles, abstracts.
Rhetorical organisation Mainly expository texts with rhetorical organisations of comparison, collection of descriptions,
problem/solution, and causation.
Propositional features
Lexical range Normally no technical jargon: approximately 7000 words (root forms; functional and subtechnical
lexis); academic semi-technical words defined in the syllabus for Section 5.
Topic area Familiar to students: humanity and management/science and technology/biology and medicine.
Background knowledge Within students’ background knowledge but not totally given; students should not be able to answer
Appendix 1

test questions from background knowledge without recourse to the text.

Elocutionary features To inform, to explain, to describe, to advise, to persuade.
299

Channel of presentation Normally textual. Some texts might contain graphics.

Table A l.l (Cont’d)
300

Conditions Descriptions
Size of input/length of text 3 short passages (approx. 600-900) for careful reading (global) 15 items
3 short passages (approx. 250-500) for careful reading (local) 15 items
3 long passages (approx. 1000-1800) for expeditious reading (global) 15 items
3 long passages (approx. 1000) for expeditious reading (local) 15 items
Speed of processing 144 minutes for a total of 12 passages: about 60-90 wpm for careful reading; 100-150 wpm for
Appendix 1

expeditious reading.
Control over skills/strategies Three passages for each skill/strategy, one from arts and humanities, one from science and technology,
one from life and medical science. For careful reading, passages are short and may sometimes have
relatively implicit text structure. For expeditious reading, passages are long and may sometimes have
relatively explicit text structures.
Control over time spent Time is strictly controlled both for each section and for each passage within the section.
Careful reading (global): 60 minutes, 20 for each passage;
Expeditious reading (skimming): 15 minutes, 5 for each passage;
Expeditious reading (search reading): 21 minutes, 7 for each passage;
Expeditious reading (scanning): 18 minutes, 6 for each passage;
Careful reading (local): 30 minutes, 10 for each passage.
Amount of help General instructions (in Chinese) to candidates are provided 15 minutes before the test. Instructions
for each section are clearly written on a separate page in the question booklet and students are
reminded to read instructions before texts. Example provided for the truth/false/justification items
since candidates are not familiar with format.
Number and ordering of tasks Order for the five sections: careful reading (global), skimming, search reading, scanning, careful
reading (local).
Method factor/response mode Formats include: SAQ, true/false, table/flow chart/sentence/text completion.
Question/answer in Ll/TL Mainly in English but could be in Chinese if necessary.
Receptive/productive Mainly receptive, some limited writing involved in SAQ but only brief answers will be required; no
more than 10 words.
Explicitness of weighting All items equally weighted
Table A1.2 A taxonomy of skills and strategies in reading for academic purposes
Types of reading Expeditious reading strategies
strategies and skills
Skimming Search reading Scanning
Purpose Processing a text selectively to get Locating information on Looking quickly through
the main idea(s) and the discourse predetermined topic(s) (e.g., a text - not necessarily following
topic as efficiently as possible in the form of questions set the linearity of the text - to locate a
- which might involve both on main idea(s) in a text). specific symbol or group of symbols:
expeditious and careful reading. This normally goes beyond e.g., a particular word, phrase,
■To establish a general sense of mere matching of words (as name, figure, or date.
the text; in scanning). The process is
■To quickly establish a selective but is likely to involve
macropropositional structure careful reading once relevant
as outline summary without information has been located.
decoding all the text;
■To read more efficiently;
■To decide the relevance of texts
to established needs.
Operationalisations Where appropriate to text-type: Keeping alert for words in the Looking for (matching):
■Reading title and subtitles same or related semantic field ■specific words/phrases
quickly. (not certain of precise form of ■figures, percentages
■Reading abstract carefully. these words). ■dates of particular events
■Reading introductory and Using formal knowledge for ■specific items in index
concluding paragraph carefully. locating information.
Appendix 1

■Reading first and last sentence Using titles and subtitles.

of each paragraph carefully. Reading abstract where
■Glancing at words or phrases. appropriate.
301

Glancing at words or phrases.

Table A1.2 (Cont’d)
302

Types of reading Expeditious reading strategies

strategies and skills
Skimming Search reading Scanning
Focus Both global and local. Both global and local. Local.
Text coverage Selective reading to establish Selecting information relevant to Ignoring most of the text.
Appendix 1

important propositions of a text. predetermined topic (s).

Rate of reading Rapid with some careful reading. Rapid with careful reading when Rapid with some careful reading.
information is located.
Direction of processing Sequencing observed. Sequencing not always observed. Sequencing not observed.
Relationship with the Interactive process involving both Interactive process. Surface level rather than deep
underlying process (es) top-down and bottom-up ■There is more observance of processing of a text. Mainly a
processing. the linearity and sequencing bottom-up process:
as compared with scanning. ■exhibiting a mixture of rapid
■Involves some top-down inspection of the text with an
processing, i.e., using formal occasional closer inspection;
knowledge. The periods of ■finding a match between what is
close attention to the text sought and what is given in a text,
tend to be more frequent very little information [being]
and longer than in scanning. processed for long-term retention
■Also bottom-up involved or even for immediate
because of close attention understanding. (Pugh, 1978: 53)
to the selected part(s) of the But for some items the top-down
text. The periods of close process may be involved (i.e. using*
attention to the text may be formal knowledge to look for
less frequent than skimming specific information in a fixed-
because of the predetermined pattern text, or to draw upon
searching. previous understanding established
through skimming).
Types of reading Careful reading skills
strategies and skills ----------------------------------------------------------------------------------------------------------------------------------------------------------
Understanding a text Understanding lexis Understanding syntax
Purpose Processing a text carefully and thoroughly in order to Lexical inferences are When faced with a text
comprehend main idea(s) and supporting information. of three kinds: the whose meaning they
resolution of lexical cannot untangle,
To decode all in order to comprehend all. ambiguity; the [readers] may be able
prediction of the to identify the
To carefully establish a macro structure for the text. meaning of unknown constituents of its
words; the identification sentences, e.g. the
of pronominal subject, the verb, and to
reference. (Chikalanga, analyse these if they are
1991) complex. (Nuttall, 1996:
78)
Operationalisations ■Separating explicidy ■Making propositional ■Resolving lexical ■Removing all the
stated main idea(s) from inferences. Propositional ambiguity: the reader optional elements of
supporting details inference is made when makes a choice complex sentences
through the reader uses explicit between two or more systematically until
(a) recognising topic statements in the text to meanings of a lexical only the essentials
sentence (s) come to a conclusion item. remain and the bare
(b) recognising lexical that is not explicitly ■Predicting the structure of the
indicator (s) stated, without recourse meaning sentence is clear.
■Generating a to knowledge from of unknown words: ■Paraphrasing optional
representation of a text outside the text. the reader infers the elements of complex
as a whole (Chikalanga, 1991) meaning of unknown sentences one by one,
■Understanding the words from the and fitting them into
Complex Pre

development of an context in a text. the whole structure to

argument and/or logical make sense of them.
organisation
302
Table A1.2 (Cont’d)
304

Types of reading Careful reading skills

strategies and skills
Understanding a text Understanding lexis Understanding syntax
■Distinguishing Where appropriate to the Identifying
generalisations and text-type: pronominal reference:
Appendix 1

examples ■Making propositional the reader identifies

informational inferences the pronominal
which are either anaphoric and
referential typically cataphoric links
answering questions within a text.
beginning with what
and which, or spatio -
temporal (typically
answering questions
beginning with where and
when).
■Making propositional
explanatory inferences
which are concerned
with motivation, cause
and consequences, and
enablement, and will
often answer questions
beginning with why and
how. All the information
needed to make
propositional inferences
is recoverable from the
text.
Types of reading Careful reading skills
strategies and skills ----------------------------------------------------------------------------------------------------------------------------------------------------------
Understanding a text Understanding lexis Understanding syntax
These may involve:
■discovering writer’s
intention
■understanding writer’s
attitude to the topic
■identifying the addressee
■distinguishing fact from
opinion.
Focus Both global and local. Both global and local. Mainly local, Local.
occasionally global.
Text coverage Reading from beginning Will vary but will be Normally use of Use of immediate
to end. selective. immediate context and context.
on occasion wider
context.
Rate of reading Reading the whole Reading the selected parts Reading the selected Reading the selected
text carefully. carefully. part(s) carefully. part carefully.
Direction of processing Linear and sequential, with Not necessarily sequential. Sequencing not Sequential with
regressions if needed. observed. some regression.
Relationship with the Mainly text-based bottom- Text based. Initially use Mainly bottom-up Bottom-up process.
underlying process(es) up sequential process with bottom-up process to process. Occasionally
limited top-down process. identify information from with the help of top-
the text and then top-down down process.
Appendix 1

process kicks in as needed

to activate knowledge
schemata to help make
inferences based on the
305

text.
306

Appendix 2

MAPPING A TEXT - I
Time for the mapping activity should not exceed the actual test
time. The aim is to identify items that can be used to test careful
reading at the global level, that is those which test main ideas.
Text length, c.600 words

(A) Individual work

Stage I: 10-15 minutes
■Read the text carefully to establish the main ideas. Notes may
NOT be taken (see below).
Stage II: 5 minutes
■Without looking back at the text, write out these ideas as clearly
as you can on a sheet of paper.
■Try as far as possible to organise the points in the order you
think they appeared in the text.
■Number the points you have made.

(B) Group work

Stage I
■On a master sheet of paper, list the main ideas on which the
group members agree.
 Appendix 2 307
■Write the num ber of people who agreed on them, e.g. 4/5
(4 out of 5 agreed).
■Normally agreement of N - l is necessary, i.e. if there are 5
people in the group, at least 4 must have included the point
for a consensus.
Notes may not be taken: The aim is to extract only the main ideas/
the macropropositions and to avoid jotting down micropropositions
or minor detail. If careful reading is an incremental process then
arguably we cannot establish the macrostructure until we have
read all the text. By not allowing notes to be taken we are tiding to
avoid experiences in the past where mindmappers have written
down a lot of peripheral detail. If we only transfer important
information from working memory to long-term storage then this
might be a way of achieving that. Mental rehearsal/m onitoring of
what is im portant is an indicator of good reading.
508

References

Abbs, B.A., Ayton, A. and Freebairn, I. 1975. Strategies. London:

Longman.
Alderson, J.C. 1978. A study of the cloze procedure with native and non­
native speakers of English. PhD Thesis. University of Edinburgh.
Alderson, J.C. 1984. Reading in a foreign language: a reading
problem or a language problem? In Alderson, J.C. and Urquhart,
A.H. (eds), pp. 1-24.
Alderson, J.C. 1990a. Testing reading comprehension skills (Part
one). Reading in a Foreign Language, 6, 425-38.
Alderson, J.C. 1990b. Testing reading comprehension skills: getting
students to talk about taking a reading test (Part two). Reading
in a Foreign Language, 7, 465-503.
Alderson, J.C. 1993. The relationship between grammar and read­
ing in an English for academic purposes test battery. In Douglas,
D. and Chappelle, C. (eds), pp. 203-14.
Alderson, J.C. and Hughes, A. (eds) 1981. Issues in Language Testing
ELT Documents 111. London: The British Council.
Alderson, J.C. and Lukmani, Y. 1989. Cognition and reading: cog­
nitive levels as embodied in test questions. Reading in a Foreign
Language, 5, 253-70.
Alderson, J.C. and Urquhart, A.H. (eds) 1984. Reading in a Foreign
Language. London: Longman.
Alderson, J.C. and Urquhart, A.H. 1985. The effect of students’
academic discipline on their performance on ESP reading tests.
Language Testing, 2, 192-204.
Allen, J.P.B. and H.G. Widdowson (eds) 1973. English in Focus.
London: OUP.
Allen, W.S. 1974. Living English Structure (5th edn). London:
Longman.
Alvermann, D.E. and Moore D.W. 1991. Secondary school read­
ing. In Barr, R. et al. (eds), pp. 951-83.
 References 309
Anderson, N.J., Bachman, L., Perkins, K. and Cohen, A. 1991. An
exploratory study into the construct validity of a reading com­
prehension test: triangulation of data sources. Language Testing
8, 41-66.
Anderson, R.C. and Freebody, P. 1981. Vocabulary knowledge tech­
nical. Report No 136. Urbana: University of Illinois, Centre for
the Study of Reading.
Anderson, R.C. and Pearson, P.D. 1988. A schema-theoretic view
of basic processes in reading comprehension. In Carrell, P.L.
et al. (eds), pp. 36-55.
Anderson, T.H. and Armbruster, B.B. 1984. Studying. In Pearson,
P.D. et al. (eds), pp. 657-79.
Aulls, M.W. 1986. Actively teaching main idea skills. In Baumann,
J.F. (ed.), pp. 96-132.
Bachman, L.F. 1990. Fundamental Considerations in Language Test­
ing. Oxford: Oxford University Press.
Bachman, L.F. and Palmer, A.S. 1981. The construct validity of
the FSI oral interview. Language Learning, 31, 67-86.
Back, E. and Harms, R.T. (eds) 1968. Universals in Linguistic Theory.
New York: Holt.
Baker, D. 1989. Language Testing: A Critical Survey and Practical
Guide. London: Edward Arnold.
Baker, L. and Brown, A.L. 1984. Cognitive monitoring in reading.
In Flood, J. (ed.), pp. 21-44.
Bamford, J. 1984. Extensive reading by means of graded readers.
Reading in a Foreign Language, 2, 218-60.
Barnett, M. 1986. Syntactic and lexical/semantic skill in foreign
language reading: importance and interaction. Modem Language
Journal, 70, 343-9.
Barr, R. and Dreeben, R. 1991. Grouping students for reading
instruction. In Barr, R. et al. (eds), pp. 885-910.
Barr, R., Kamil, M.L., Mosenthal, P. and Pearson, P.D. (eds) 1991.
Handbook of Reading Research, Vol II. New York: Longman.
Bartlett, B.J. 1978. Top Level Structure as an Organizational Strategy
for Recall of Classroom Text. Unpublished Ph.D. Dissertation. Ari­
zona State University.
Bartlett, F.C. 1932. Remembering: A Study in Experimental and Social
Psychology. Cambridge: Cambridge University Press.
Baten, L. and Cornu, A.M. 1984. Reading strategies for LSP texts:
a theoretical outline on the basis of text function, with practical
application. In Pugh, A.K. and Ulijn, J.M. (eds), pp. 190-201.
310 References
Baumann, J.F. (ed.) 1986. Teaching Main Idea Comprehension.
Newark, Delware: IRA.
Baynam, M. 1995. Literacy Practices. London: Longman.
Beck, I. 1981. Reading problems and instructional practices. In
Mackinnon, G.E. and Waller, T.G. (eds), pp. 55-99.
Beck, I. and McKeown, M. 1991. Conditions of vocabulary acquisi­
tion. In Barr, R. et al. (eds), pp. 789-814.
Bensoussan, M. and Kreindler, I. 1990. Improving advanced read­
ing comprehension in a foreign language: summaries versus
short-answer questions. Journal of Research in Reading, 13, 55-68.
Bensoussan, M. and Laufer, B. 1984. Lexical guessing in context
in EFL reading comprehension. Journal of Research in Reading,
7, 15-31.
Berg, P.C. 1973. Evaluating reading abilities. In McGinitie, W.H.
(ed.), pp. 27-34.
Berman, M. 1979. Advanced Language Practice for EFL. London:
Hodder & Stoughton.
Berman, R.A. 1984. Syntactic components of the foreign language
reading process. In Alderson, J.C. and Urquhart, A.H. (eds),
pp. 139-59.
Bernhardt, E.B. 1991a. Reading Development in Second Language:
Theoretical, Empirical and Classroom Perspectives. New Jersey: Ablex
Publishing Corporation.
Bernhardt, E.B. 1991b. A psycholinguistic perspective on sec­
ond language literacy. In Hulstijn, J.H. and Matter, J.F. (eds),
pp. 31-44.
Bloom, B.S., Engelhart, M.D., Furst, E.J., Hill, W.H. and Kratwohl,
D.R. (eds) 1956. Taxonomy of Educational Objectives: Cognitive
Domain. New York: David McKay. (See also Bloom, B.S. et al.
(eds), Taxonomy of educational objectives. Handbook I: Cog­
nitive Domain. London: Longman, 1974.)
Bloome, D. and Greene, J. 1984. Directions in the sociolinguistic
study of reading. In Pearson, P.D. et al. (eds), pp. 395-421.
Bloom, D. 1985. Procedural Display. Unpublished manuscript.
University of Massachusetts. Amherst MA.
Bloor, M. 1985. Some approaches to the design of reading courses
in English as a foreign language. Reading in a Foreign Language,
3: 341-61.
Bolinger, D. 1975. Aspects of Language (2nd edn). New York:
Harcourt Brace Jovanovich.
 References 311
Bransford, J., Stein, B. and Shelton, T. 1984. Learning from the
perspective of the comprehender. In Alderson, J.C. and
Urquhart, A.H. (eds), pp. 28-44.
Bright, J.A. 1965. Patterns and Skills in English, Books 1 and 3.
Nairobi: Longman.
Bright, J.A. and McGregor, G.P. 1970. Teaching English as a Second
Language. London: Longman.
Britton, B.K. and Black, J.B. (eds) 1985. Understanding Expository
Text: A Theoretical and Practical Handbook for Analyzing Explana­
tory Text Hillsdale, NJ: LE.
Bronfenbrenner, U. 1976. The experimental ecology of educa­
tion. Educational Researcher, 5, 5-15.
Brooks, C. and Warren, R.P. 1952. Fundamentals of Good Writing.
London: Dennis Dobson.
Brown, A.L. and Day, J.D. 1983. Macrorules for summarizing texts:
the development of expertise. Journal of Verbal Learning and
Verbal Behavior, 22, 1-14.
Brown, G. and Yule, G. 1983. Discourse Analysis. Cambridge: Cam­
bridge University Press.
Brown, K. (1984) Linguistics Today. Fontana.
Brown, P.J. and Hirst, S.B. (1983) Writing reading courses: the
interrelationship of theory and practice. In Brumfit, C.J. (ed.),
pp. 135-51.
Brumfit, C.J. (ed.) 1983. Language Teaching Projects for the Third
World. ELT Documents 116. Oxford: Pergamon Press, in asso­
ciation with The British Council.
Brutten, S.R., Perkins, K. and Upshur, J.A. 1991. Measuring growth
in ESL reading. Paper presented at the 13th Annual Language
Testing Research Colloquium. Princeton, New Jersey.
Buck, G. 1990. The testing of second language listening comprehension.
Unpublished Ph.D. Thesis. University of Lancaster.
Buzan, T. 1974. Use Your Head. London: BBC.
Calfee, R.C. and Curley, R. 1984. Structures of prose in content
areas. In Flood, J.S. (ed.), pp. 161-80.
Calfee, R. and Hiebert, E. 1991. Classroom assessment of reading.
In Barr, R. et al. (eds), pp. 281-310.
Candlin, C. 1984. Preface. In Alderson, J.C. and Urquhart, A.H.
(eds), pp. ix-xiii.
Candlin, C. and Murphy, D. (eds) 1987. Language Learning Tasks.
Englewood Cliffs, NJ: Prentice Hall.
312 References
Cangiano, V.J. 1992. Bilingual and ESL approaches to deaf education:
perspectives on the reading process. Master’s Thesis. Hunter College,
City University of New York.
Carnine, D. and Silbert,J. 1979. Direct Instruction Reading. Columbus,
Ohio: Merrill.
Carr, T.H. 1985. The development of reading skills. New Directions
for Child Development, No. 27. San Francisco: Jossey-Bass.
Carr, T.H. and Levy, B.A. (eds) 1990. Reading and Its Development:
Component Skills Approaches. San Diego: Academic Press.
Carrell, P.L. 1983a. Three components of background knowledge
in reading comprehension. Language Learning, 33, 183-207.
Carrell, P.L. 1983b. Some issues in studying the role of schemata,
or background knowledge in second language comprehension.
Reading in a Foreign Language, 1, 81-92.
Carrell, P.L. 1983c. Some classroom implications and applications
of recent research into schema theory and EFL/ESL reading.
First Rough Draft of a Paper to be Presented at the Colloquium
on Research in Reading in a Second Language at the 1983
TESOL Convention. Toronto, March 17.
Carrell, P.L. 1983d. Background knowledge in second language
comprehension. Language Learning and Communication, 2, 25-34.
Carrell, P.L. 1984. The effects of rhetorical organisation on ESL
readers. TESOL Quarterly, 18, 441-69.
Carrell, P.L. 1988. Interactive text processing: implications for
ESL/second language classrooms. In Carrell, P.L. et al. (eds),
pp. 239-59.
Carrell, P.L. 1988a. Introduction, in Carrell, P.L. et al. (eds),
pp- 1 -P.L.
Carrell, 7-
1991. Second language reading: reading ability or
language proficiency. Applied Linguistics, 12, 159-79.
Carrell, P.L. and Eisterhold, J.C. 1988. Schema theory and ESL
reading pedagogy. In Carrell, P.L. et al. (eds), pp. 73-92.
Carrell, P.L., Devine, J. and Eskey, D.E. (eds) 1988. Interactive
Approaches to Second Language Reading. Cambridge: Cambridge
University Press.
Carrell, P.L., Pharis, G.B. and Liberto, J.C. 1989. Metacognitive
strategy training for ESL reading. TESOL Quarterly, 23, 647-75.
Carroll, J.B. 1972. Defining language comprehension: some specu­
lations. In Freedle, R.O. and Carroll, J.B. (eds), pp. 1-29.
Carroll, J.B. and Freedle, R.O. (eds) 1972. Language Comprehension
and the Acquisition of Knowledge. Washington: Wiley.
 References 313
Carter, R. and McCarthy, M. (eds) 1988. Vocabulary and Language
Teaching. London: Longman.
Carver, R.P. 1985. How good are some of the world’s best readers?
Reading Research Quarterly, 20, 389-419.
Carver, R.P. 1992. What do standardized tests of reading compre­
hension measure in terms of efficiency, accuracy and rate? Read­
ing Research Quarterly, 27, 347-59.
Cavalcanti, M.C. 1983. The pragmatics of FL reader-text interaction.
Key lexical items as source of potential reading problems. Unpub­
lished PhD Thesis, University of Lancaster.
Chackray, C. 1979. Growth in Reading. London: Ward Lock
Educational.
Chafe, W.L. 1972. Discourse structure and human knowledge. In
J.B. Carroll and R.O. Freedle (eds), pp. 41-69.
Chall, J.S. 1979. The great debate. Ten years later, with a modest
proposal for reading stages. In Resnick, L.B. and Weaver, P.A.
(eds), pp. 29-55.
Chall, J.S. 1983. Stages of Reading Development. New York: McGraw-
Hill.
Chall, J.S. 1984. Readability and prose comprehension continuit­
ies and discontinuities. In Flood, J.S. (ed.), pp. 233-46.
Chapman, J. (ed.) 1981. The Reader and the Text. London:
Heinemann.
Chapman, L.J. 1979a. Confirming use of cohesion ties in text:
pronouns. The Reading Teacher, 33, 317-22.
Chapman, L.J. 1979b. Pedagogical strategies for fluent reading.
In Chackray, C., pp. 147-154.
Chihara, T., Oiler, J. and Chavez-Oller, M.A. 1977. Are cloze tests
sensitive to constraints across sentences? Language Learning, 27,
143-51.
Chikalanga I.W. 1990. Inferencingin the reading process. Unpublished
PhD Thesis. University of Reading.
Chikalanga, I.W. 1992. A Suggested taxonomy of inferences
for the reading teacher. Reading in a Foreign Language, 8,
697-710.
Chikalanga, I.W. 1993. Exploring inferencing ability of ESL readers.
Reading in a Foreign Language, 10, 932-52.
Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.:
MIT Press.
Clapham, C. 1990. Is ESP testing justified? Paper Given at the
Language Testing Research Colloquium, San Francisco.
314 References
Clapham, C. 1994. The effect of background knowledge on EAP reading
test performance. Unpublished PhD Thesis. Lancaster University.
Clapham, C. 1996a. The Development of IELTS: A Study of the Effect
of Background Knowledge on Reading Comprehension. Cambridge:
University of Cambridge Press.
Clapham, C. 1996b. What makes an ESP reading test appropriate
for its candidates? In Cumming, A. and Berwich, R. (eds),
pp. 171-94.
Clarke, D.F. and Nation, I.S.P. 1980. Guessing the meaning of words
from context: strategy and techniques. System, 8, 211-220.
Clarke, M. 1979. Reading in Spanish and English: evidence from
adult ESL students. Language Learning, 29, 121-50.
Clarke, M.A. 1980/1988. The short circuit hypothesis of ESL read­
ing - or when language competence interferes with reading
performance. In Carrell, P.L. et al. (eds), 1988, pp. 114-124.
Clymer, T. (ed.) 1968. Innovation and Change in Reading Instruc­
tion. Chicago: University of Chicago Press.
Coady, J. 1979. A psycholinguistic model of the ESL reader. In
Mackay, R. et al. (eds), pp. 5-12.
Cohen, A. 1987. Using verbal reports in research on language
learning. In Faerch, C. and Kasper, G. (eds), pp. 82-96.
Cohen, A. 1998. Strategies in Learning and Using a Second Language.
London: Longman.
Cohen, A.D., Glasman, H., Rosenbaum-Cohen, P.R., Ferrara, J.
and Fine, J. 1979. Discourse analysis and the use of student
informants. TESOL Quarterly. 13, 551-64.
Cohen, A.D., Glasman, H., Rosenbaum-Cohen, P.R., Ferrara, J.
and Fine, J. 1988. Reading English for specialized purposes:
discourse analysis and the use of student informant. In Carrell,
P.L. et al. (eds), pp. 152-67.
Collins, A.M. and Quillian, M.R. 1969. Retrieval time from sem­
antic memory. Journal of Verbal Learning and Verbal Behavior, 8,
240-7.
Conant, J.B. 1934. Notes and news: President Conant speaks again.
Modern Language Journal, 19, 465-6.
Cook, V.J. 1988. Chomsky's Universal Grammar. Oxford: Blackwell.
Cook, V.J. 1992. Chomsky’s Universal Grammar. Oxford: Blackwell.
Cornu, A.M., van Parijs, J., Delahye, M. and Baten, L. (eds) 1986.
Beads or Bracelet: How Shall We Approach ESP? Leuven: OUP.
Coulthard, M. (ed.) 1994. Advances in Written Text Analysis. Lon­
don: Routledge.
 References 315
Cowan, J.R. 1974. Lexical and syntactic research for the design of
EFL reading materials. TESOL Quarterly, 8, 388-99.
Cowan, J.R. 1976. Reading, perceptual strategies and contrastive
analysis. Language Learning, 26, 95-109.
Criper, C. and Davies, A. 1988. English Language Testing Service.
Research Report 1 (i) ELTS Validation Project Report. The
British Council and the University Examinations Syndicate.
Crookes, G. and Gass, S.M. 1993a. Tasks and Language Learning
Clevedon: Multilingual Matters.
Crookes, G. and Gass, S.M. (eds) 1993b. Tasks in a Pedagogical
Context. Clevedon: Multilingual Matters.
Crothers, E.J. 1972. Memory structure and the recall of discourse.
In Freedle, R.O. and Carroll, J.B. (eds), pp. 247-83.
Crothers, E.J. 1978. Inference and coherence. Discourse Processes,
1, 51-71.
Crystal, D. 1997. The Cambridge Encyclopedia of Language (2nd edn).
Cambridge: Cambridge University Press.
Cumming, A. and Berwich, R. (eds) 1996. Validation in Language
Testing. Clevedon: Multilingual Matters.
Cziko, G.A. 1978. Differences in first and second language read­
ing: the use of syntactic, semantic and discourse constraints.
Canadian Modem Languages Review, 34, 473-89.
Cziko, G.A. 1980. Language competence and reading strategies: a
comparison of first and second language oral reading errors.
Language Learning, 30, 101-14.
Dansereau, D.F., Collins, K.W., McDonald, B.A., Holley, C.D., Gar­
land, J., Dickhoff, G. and Evans, S.H. 1979. Development and
evaluation of a learning strategy program. Journal of Educational
Psychology, 71, 64-73.
Davies, A. 1983. The validity of concurrent validation. In Hughes,
A. and Porter, D. (eds), pp. 141-5.
Davies, A. 1990. Principles of Language Testing. Oxford: Blackwells.
Davies, F. 1995. Introducing Reading. London: Penguin Group.
Davies, F. and Greene, T. 1980. Reading for Learning in the Sciences.
Edinburgh: Oliver & Boyd.
Davis, C. 1995. Extensive reading: an expensive extravagance? ELT
Journal, 49, 329-36.
Davis, F.B. 1944. Fundamental factors of comprehension in read­
ing. Psychometrika, 9, 185-97.
Davis, F.B. 1968. Research in comprehension in reading. Reading
Research Quarterly, 3, 499-545.
316 References
Davis, F.B. 1972. Psychometric research on comprehension in read­
ing. Reading Research Quarterly, 7, 628-78.
Day, R., Omura, C. and Hiramatsu, M. 1991. Incidental vocabulary
learning and reading. Reading in a Foreign Language, 7, 541-53.
de Beaugrande, R. 1980. Text, Discourse and Process: Toward a Multi­
disciplinary Science of Texts. London: Longman.
de Beaugrande, R. 1981. Design criteria for process models of
reading. Reading Research Quarterly, 16, 261-315.
DeFrancis,J. 1984. Visible Speech: The Diverse Oneness of Writing Systems.
Honolulu: University of Hawaii Press.
de Leeuw, M. and de Leeuw, E. 1965. Read Better, Read Faster.
Harmondsworth: Penguin.
de Leeuw, M. and de Leeuw, E. 1990. Read Better, Read Faster (new
edition). Harmondsworth: Penguin
Devine, J. 1987. General language competence and adult second
language reading. In Devine, J. et al., pp. 73-87.
Devine, J. 1988. The relationship between general language com­
petence and second language reading proficiency: implications
for teaching. In Carrell, P.L. et al. (eds), pp. 260-77.
Devine, J., Carrell, P. and Eskey, D.E. (eds) 1987. Research in Read­
ing in English as a Second Language. Washington: TESOL.
Dhaif, H. 1990. Reading aloud for comprehension. Reading in a
Foreign Language, 7, 457-64.
Dornic, S. (ed.) 1977. Attention and Performance VI. New York: Aca­
demic Press.
Douglas, D. and Chappelle, C. (eds) 1993. A New Decade of Lan­
guage Testing Research. Washington: TESOL.
Downing, A. and Locke, P. 1992. A University Course in English
Grammar. New York: Prentice Hall.
Downing, J. and Leong, K.L. 1982. Psychology of Reading. New York:
Macmillan.
Dubin, F., Eskey, D.E. and Grabe, W. (eds) 1986. Teaching Second
Language Readingfor Academic Purposes. Reading, Mass.: Addison-
Wesley.
Duffy, G. and Anderson, L. 1981. Final Report: Conceptions of Reading
Project. Unpublished report. Institute for Research on Teaching,
Michigan State University, East Lansing, MI.
Elley, W. 1991. Acquiring literacy in a second language: the effect
of book based programmes. Language Learning, 41, 375-411.
Elley, W.B. and Mangubhai, F. 1983. The impact of reading on
second language learning. Reading Research Quarterly, 19, 53-67.
 References 317
Ellis, G. and Mcrae, J. (eds) 1991. Extensive Reading Handbook for
Secondary Teachers. London: Penguin.
Ellis, R. and Tomlinson, B. 1988. Reading Advanced. Oxford: Oxford
University Press.
Entwhisde, N., Hanley, M. and Hounsell, D. 1979. Identifying dis­
tinctive approaches to studying. Higher Education, 8, 365-80.
Eskey, D.E. 1988. Holding in the bottom: an interactive approach
to the language problems of second language readers. In Carrell,
P.L. et al. (eds), pp. 93-100.
Faerch, C. and Kasper, G. (eds) 1986. The role of comprehension
in foreign language learning. Applied Linguistics, 7, 257-74.
Faerch, C. and Kasper, G. (eds) 1987. Introspection in Second Lan­
guage Research. Clevedon: Multilingual Matters.
Fairclough, N. (ed.) 1992. Critical Language Awareness, No. 16. Lon­
don: Longman.
Fairclough, N. 1995. Critical Discourse Analysis. London: Longman.
Fanselow, J.H. and Crymes, R.H. (eds) 1976. On TESOL 16. Wash­
ington: TESOL.
Farr, R.C. 1968. The convergent and discriminant validity of sev­
eral upper level reading tests. Yearbook of the National Reading
Conference, 17, 181-91.
Farr, R.C., Carey, R. and Tone, B. 1986. Recent theory and research
into the reading process: implications for reading assessment.
In Orasanu, J. (ed.), pp. 134-49.
Ferguson, N. 1977. The m ind’s eye: non-verbal thought in techno­
logy. Science, 197, 827-36.
Feuerstein, R.P., Klein, P.S. and Tannenbaum, A.J. 1991. Mediated
Learning Experience: Theoretical, Psychological and Learning Implica­
tions. London: Freund.
Fillmore C.J. 1968. The case for case. In Back, E. and Harms, R.T.
(eds), pp. 1-88.
Fillmore, C.J. and Kay, P. 1983. Text Semantic Analysis of Reading
Comprehension Tests (Final report, NIE). Berkeley: University of
California, Institute of Human Learning.
Fingeret, A. 1983. Social network: a new perspective on inde­
pendence and illiterate adults. Adult Education Quarterly, 33,
133-46.
Fingeret, A. 1990. Literacy for what purpose? A response. In
Venezky, R.L. et al. (eds), pp. 35-8.
Flood, J.S. (ed.) 1984. Understanding Reading Comprehension: Cogni­
tion, Language, and the Structure of Prose. Newark, DE: IRA.
318 References
Francis, G., Hunston, S. and Manning, E. (eds) 1996. Cobuild Gram­
mar Patterns 1: Verbs. London: Harper Collins.
Fransson, A. 1984. Cramming or understanding? Effects of intrinsic
and extrinsic motivation on approach to learning and test perfor­
mance. In Alderson, J.C. and Urquhart, A.H. (eds), pp. 86-121.
Freebody, P. and Luke, A. 1990. Literacies, programs: debates and
demands in cultural context. Prospect, 5, 7-16
Freedle, R.O. and Carroll, J.B. (eds) 1972. Language Comprehension
and the Acquisition of Knowledge. Washington: Winston, V.H. 8c
Sons.
Fries, C.C. 1963. Linguistics and Reading. New York: Holt, Rinehart
& Winston.
Fry, E. 1963. Teaching Faster Reading: A Manual. Cambridge: Cam­
bridge University Press.
Fyfe, R. and Mitchell, E. 1985. Reading Strategies and their Assess­
ment. London: NFER Nelson.
Gates, A.I. 1926. A study of the role of visual perception, intelli­
gence and certain associative processes in reading and spelling.
Journal of Educational Psychology, 17, 433-45.
Geddes, M. and Sturtridge, G. 1982. Reading Links. London:
Heinemann Educational.
Gee, J. 1990. Social Linguistics and Literacies: Ideology in Discourses.
Brighton: Falmer Press.
Geva, E. 1980. Meta textual notions and reading comprehension. Unpub­
lished PhD Thesis. University of Toronto.
Geva, E. 1983. Facilitating reading comprehension through flow­
charting. Reading Research Quarterly, 18, 384-405.
Gibson, E.J. and Levin, H. 1974. The Psychology of Reading. Cam­
bridge, Mass.: MIT Press.
Glendinning, E. 1974. English in Mechanical Engineering. London:
Oxford University Press.
Glendinning, E. and Holmstrom, B. 1992. Study Reading-A Course
in Reading Skills for Academic Purpose. Cambridge: Cambridge
University Press.
Glushko, R. 1979. The organization and activation of orthographic
knowledge in reading aloud. Journal of Experimental Psychology:
Human Perception and Performance, 5, 674-91.
Goodman, K.S. 1967. Reading: a psycholinguistic guessing game.
Journal of the Reading Specialist, 6, 126-35.
Goodman, KS., Goodman, Y. and Flores, B. 1979. Reading in the
Bilingual Classroom: Literacy and Biliteracy. Rosslyn, Va.: National
Clearinghouse for Bilingual Education.
 References 319
Goodman, K.S. and Fleming, J.T. (eds) 1969. Psycholinguistics and
the Teaching of Reading. Newark, DE: IRA.
Goody, J. (ed.) 1968. Literacy in Traditional Societies. Cambridge:
Cambridge University Press.
Goody, J.R. and Watt, I. 1968. The consequences of literacy. In
Goody, J. (ed.), pp. 27-68.
Gough, P.B. 1972. One second of reading. In Kavanagh, F.J. and
Mattingly, G. (eds), pp. 331-58.
Grabe, W. 1988. Reassessing the term ‘interactive’. In Carrell, P.L.
et al. (eds), pp. 56-70.
Grabe, W. 1991. Current developments in second language read­
ing research. TESOL Quarterly, 25, 375-406.
Grabe, W. and Kaplan, R.B. 1996. The Theory and Practice of Writing:
An Applied Linguistic Perspective. London: Longman.
Graves, M.F. and Cook, C.L. 1980. Effects of previewing difficult
short stories for high school students. Research on Reading in
Secondary Schools, 6, 38-54.
Graves, M.F., Cook, C.L. and LaBerge, H.J. 1983. Effects of pre­
viewing difficult short stories on low-ability junior high school
students’ comprehension, recall and attitude. Reading Research
Quarterly, 18, 262-76.
Grellet, F. 1981. Developing Reading Skills. Cambridge: Cambridge
University Press.
Griffin, P. 1990. Adult Literacy and Numeracy Competency Scales. Vic­
toria, Australia: Assessment Research Centre, Phillip Institute
of Technology.
Grimes, J.E. 1975. The Thread of Discourse. The Hague: Mouton.
Grundin, E.H. and Grundin, H.U. 1984. Reading: Implementing the
Bullock Report. London: Ward Lock International.
Guthrie, J.T. (ed.) 1977. Cognition, Curriculum, and Comprehension.
Newark, DE: International Reading Association.
Guthrie, J.T. and Kirsch, I.S. 1987. Distinctions between reading
comprehension and locating information in text. Journal of Edu­
cational Psychology, 79, 220-97.
Guthrie, J.T. and Mosenthal, P. 1987. Literacy as multidimensional:
locating information and reading comprehension. Educational
Psychologist, 22, 279-97.
Hafiz, F.M. and Tudor, I. 1989. Extensive reading and the develop­
m ent of language skills. ELT Journal, 43, 4-13.
Hafiz, F.M. and Tudor, I. 1990. Graded readers as an input medium
in L2 learning. System, 18, 31-42.
Haines, S. 1987. Reading 3. London: Cassell.
320 References
Haines, S. 1988. Reading 4. London: Cassell.
Halliday, M.A.K. and Hasan, R. 1976. Cohesion in English. London:
Longman.
Hamp-Lyons, L. 1984. Developing a course to teach extensive read­
ing skills to university-bound learners. System, 11, 303-12.
Hamp-Lyons, L. 1985. Two approaches to teaching reading: a class­
room based study. Reading in a Foreign Language, 3, 363-73.
Handscombe, J., Orem, R.A. and Taylor, B.P. (eds) 1984. On TESOL
83: The Question of Control. Washington, D.C.
Harri-Augstein, S. and Thomas, L. 1984. Is comprehension the
purpose of reading? In Grundin, E.H. and Grundin, H.U.,
pp. 250-76.
Harris, T.L. and Hodges, R.E. 1981. A Dictionary of Reading and
Related Terms. Newark: IRA.
Hartman, D. 1992. Eight readers reading: the intertextual links of
able readers using multiple passages. Reading Research Quarterly,
27, 122-3.
Hayes, J.R. (ed.) 1970. Cognition and the Development of Language.
New York: Wiley.
Haynes, M. 1984. Patterns and perils of guessing in second lan­
guage reading. In Handscombe, J. et al. (eds), pp. 163-76.
Haynes, M. 1993. Patterns and perils of guessing in second lan­
guage reading. In Huckin, T. et al., pp. 46-62.
Haynes, M. and Carr, T.H. 1990. Writing system background
and second language reading: a component skills analysis of
English reading by native speakers of Chinese. In Carr, T.H.
and Levy, B.A. (eds), pp. 375-421.
Heaton, J.B. 1988. Writing English Language Tests (2nd edn). Lon­
don: Longman.
Hedge, T. 1985. Using Graded Readers in Language Teaching. Lon­
don: Macmillan.
Hegarty, M., Carpenter, P.A. and Just, M.A. 1991. Diagrams in
the comprehension of scientific texts. In Barr, R. et al. (eds),
pp. 641-68.
Hill, D.R. 1992. The EPER Guide to Organising Programmes of Extens­
ive Reading. Edinburgh: The Institute of Applied Language Stud­
ies, University of Edinburgh.
Hill, D.R. and Reid-Thomas, H. 1988a. Survey review: graded
readers (Part 1). English Language Teaching Journal, 42, 44-82.
Hill, D.R. and Reid-Thomas, H. 1988b. Survey review: graded
readers (Part 2). English Language TeachingJournal, 42, 124-36.
 References 321
Hill, D.R. and Reid-Thomas, H. 1989. Survey review: seven series
of graded readers. English Language Teaching Journal, 43,
221-31.
Hill, D.R. and Reid-Thomas, H. 1993. Seventeen series of graded
readers. English Language Teaching Journal, 48, 250-6.
Hillocks, G. and Ludlow, L.H. 1984. A taxonomy of skills in reading
and interpreting fiction. American Educational Research Journal,
21, 7-24.
Hoey, M. 1983. On the Surface of Discourse. London: George Allen &
Unwin.
Hoffman, J.V. 1991. Teacher and school effects in learning to
read. In Barr, R. et al., pp. 911-50.
Holcomb, T. and Peyton, J. 1992. ESL Literacy for a Linguistic Minority:
The Deaf Experience. Washington, DC: National Clearinghouse
on Literacy Education.
Hoover, W.A. and Tunmer, W.E. 1993. The components of reading.
In Thompson, J.B. et al., pp. 1-19.
Horrocks, G. 1987. Generative Grammar. London: Longman.
Hosenfeld, C. 1984. Case studies of ninth grade readers. In Alder­
son, J.C. and Urquhart, A.H. (eds), pp. 231-49.
Householder, F. 1971. Linguistic Speculations. Cambridge: Cam­
bridge University Press.
Howatt, A.P.R. 1984. A History of English Language Teaching. Oxford:
Oxford University Press.
Huckin, T., Haynes, M. and Coady, J. 1993. Second Language Read­
ing and Vocabulary Learning. Norwood, NJ: Ablex.
Hudson, T. 1982. The effects of induced schemata on the ‘short
circuit’ in L2 reading: non-decoding factors in L2 reading per­
formance. Language Learning, 32, 3-31.
Hudson, T. 1988. The effects of induced schemata on the ‘short-
circuit’ in 12 reading: non-decoding factors in L2 reading per­
formance. In Carrell, P.L. et al. (eds), pp. 183-205.
Huey, E.B. 1968. The Psychology and Pedagogy of Reading. Cambridge,
Mass.: MIT Press.
Hughes, A. (ed.) 1988. Testing. English for University Study. ELT
Documents 127. Oxford: Modern English Press.
Hughes, A. 1989. Testing for Language Teachers. Cambridge: Cam­
bridge University Press.
Hughes, A. 1993. Testing the ability to infer when reading in a
second or foreign language. Journal of English and Foreign Lan­
guages, 10/11, 13-20.
322 References
Hughes, A. and Porter, D. (eds) 1983. Current Developments in Lan­
guage Testing. London: Academic Press.
Hulstijn, J.H. and Matter, J.F. (eds) 1991. Reading in two languages.
ALLA Review, 8 (Amsterdam).
Inman, M. 1978. Lexical analysis of scientific and technical prose.
In Todd, M. et al. (eds), pp. 242-56.
Ja’far, W.M. 1992. The Interactive Effects of background knowledge
on ESP reading comprehension proficiency tests. Unpublished PhD
Thesis. University of Reading.
James, C. 1983. A two stage approach to language teaching. In
Johnson, K. and Porter, D. (eds), pp. 109-25.
Jarvis, D.K and Jensen, D.C. 1982. The effect of parallel transla­
tions on second language reading and syntax acquisition. Modem
Language Journal, 66, 18-23.
Jimenez, R.T., Garcia, E. and Pearson, P.D. 1996. The reading
strategies of bilingual Latina/o students who are successful Eng­
lish readers: opportunities and obstacles. Reading Research Quar­
terly, 31, 90-112.
Johns, T. and Davies, F. 1983. Text as a vehicle for information.
Reading in a Foreign Language, 1, 1-19.
Johnson, D.M. and Reynolds, F. 1941. A factor analysis of verbal
ability. Psychological Record, 4, 183-95.
Johnson, K. and Porter, D. (eds) 1983. Perspectives in Communicative
Language Teaching. London: Academic Press.
Johnston, P.H. 1983. Reading Comprehension Assessment: A Cognitive
Basis. Newark, DE: International Reading Association.
Johnston, P.H. 1984. Assessment in reading. In Pearson, P.D. et al.
(eds), pp. 147-85.
Johnston, P. and Allington, R. 1991. Remediation. In Barr, R.
et al., pp. 984-1012.
Jones, B., Palincsar, A., Ogle, D. and Carr, E. 1987. Strategic Teach­
ing and Learning: Cognitive Instruction in the Content Area.
Alexandria, VA: Association of Supervision and Curriculum
Development.
Juel, C. 1991. Beginning reading. In Barr, R. et al. (eds), pp. 759-88.
Just, M. and Carpenter, P.A. (eds) 1977. Cognitive Processes in Com­
prehension. Hillsdale, NJ: Erlbaum.
Just, M.A. and Carpenter, P.A. 1980. A theory of reading: from eye
fixation to comprehension. Psychological Review, 87, 329-54.
Just, M.A. and Carpenter, P.A. 1987. The Psychology of Reading and
Language Comprehension. Boston, Mass.: Allyn 8c Bacon.
 References 323
Kamil, M.L., Smith-Burke, M. and Rodrigue-Brown, F. 1986. The
sensitivity of cloze to intersentential integration of information
in Spanish bilingual populations. In Niles, J.A. and Lalik, R.V.
(eds), pp. 334-8.
Kant, I. 1781/1787/1963. Critique of Pure Reason (1st edn; 2nd
edn; translated by N. Kemp Smith). London: Macmillan.
Katz, S., Lautenschlager, G., Blackburn, A. and Harris, F. 1990.
Answering reading comprehension items without passages on
SAT. Psychological Science, 1, 122-7.
Kavanagh, F.J. and Mattingly, G. (eds) 1972. Language by Ear and
by Eye. Cambridge, MA: MIT Press.
Khalifa, H. 1997. A study in the construct validation of the reading
module of an EAP proficiency test battery: validation from a variety of
perspectives. PhD Thesis. University of Reading.
Kieras, D.E. and Just, M.A. (eds) 1984. New Methods of Reading
Comprehension Research. Hillsdale, NJ: Lawrence Erlbaum.
King, P. 1989. The uncommon core: some discourse features of
students’ writing. System, 17, 13-20.
Kintsch, W. 1974. The Representation of Meaning in Memory. New
York: Wiley.
Kintsch, W. and Keenan, J.M. 1973. Reading rate and retention as
a function of the number of propositions in the base structure
of sentences. Cognitive Psychology, 5, 257-79.
Kintsch, W. and van Dijk, T.A. 1978. Toward a model of text
comprehension and production. Psychological Review, 85, 363-
94.
Kintsch, W. and Yarborough, J.C. 1982. Role of rhetorical struc­
ture in text comprehension. Journal of Educational Psychology, 74,
828-34.
Klare, G.M. 1984. Readability. In Pearson, P.D. et al. (eds),
pp. 681-744.
Kobayashi, M. 1995. Effects of text organisation and test format on
reading comprehension test performance. PhD Thesis. Thames Valley
University.
Koda, K 1987. Cognitive strategy transfer in second language read­
ing. In Devine, J. et al., pp. 125-44.
Koh, M.Y. 1985. The role of prior knowledge in reading compre­
hension. Reading in a Foreign Language, 3, 375-80.
Kolers, P.A. 1969. Reading is only incidentally visual. In Goodman,
KS. and Fleming, J.T. (eds) Psycholinguistics and the Teaching of
Reading. Newark, DE: IRA, pp. 8-16.
324 References
Koran, M.L. and Koran, J. 1980. Interaction of learner character­
istics with pictorial adjuncts in learning from science text. Journal
of Research in Science Teaching, 17, 477-83.
Krashen, S. 1993. The Power of Reading. Englewood, CO: Libraries
Unlimited.
Krashen, S.D. and Terrell, T.D. 1983. The Natural Approach. Oxford:
Pergamon.
Kretovics, J.R. 1985. Critical theory: Challenging the assumptions of
mainstream educational theory. Journal of Education, 2, 50-62.
Lado, R. 1961. Language Testing. London: Longman.
Langer, J.A. 1981. Pre-reading plan (PReP): Facilitating text com­
prehension. In Chapman, J. (ed.), pp. 125-30
Langer, J.A. 1984. Examining background knowledge and text
comprehension. Reading Research Quarterly, 19, 468-81.
Lass, N. (ed.) 1982. Speech and Language: Advances in Basic Research
and Practice, Vol. 7. New York: Academic Press.
Laufer, B. 1989. What percentage of text lexis is essential for
comprehension? In Lauren, L. and Nordman, M. (eds), pp.
316-23.
Lauren, L. and Nordman, M. (eds) 1989. Special Language: From
Human Thinking to Thinking Machines. Clevedon: Multilingual
Matters.
Lee, J.F. 1986. On the use of the recall task to measure L2 read­
ing comprehension. Studies in Second Language Acquisition, 8,
201- 1 1 .
Lee, J.F. and Musumeci, D. 1988. On hierarchies of reading skills
and text types. The Modem Language Journal, 72, 173-87.
Lefevre, C.A. 1964. Linguistics and the Teaching of Reading. New
York: McGraw-Hill.
Legg, S. and Algina,J. (eds) 1990. Cognitive Assessment of Language
and Math Outcomes. Norwood, NJ: Ablex.
Lesgold, A., Pelligreno, J., Fokkema, S. and Glaser, R. (eds) 1978.
Cognitive Psychology and Instruction. New York: Plenum.
Lesgold, A. and Perfetti, C. (eds) 1981. Interactive Processes in Read­
ing. Hillsdale, NJ: Erlbaum.
Levelt, W.J.M. 1989. Speaking: From Intention to Articulation. Cam­
bridge, Mass.: MIT Press.
Lewkowicz, J.A. 1997. Investigating authenticity in language testing.
Unpublished PhD Thesis. University of Lancaster.
Lucas, C. (ed.) 1990. The Sociolinguistics of the Deaf Community. San
Diego: Academic Press.
 References 325
Lumley, T.J.N. 1993. Reading comprehension sub-skills: teachers’
perceptions of content in an EAP test. Melbourne Papers in Applied
Linguistics, 2, 25-55.
Lund, R.J. 1991. A comparison of second language listening and
reading comprehension. Modem Language Journal,, 75, 196-204.
Lunzer, E. and Gardner, K (eds) 1979. The Effective Use of Reading.
London: Heinemann Educational.
Lunzer, E., Waite, M. and Dolan, T. 1979. Comprehension
and comprehension tests. In Lunzer, E. and Gardner, K (eds),
pp. 37-71.
Lyons, J. 1968. Introduction to Theoretical Linguistics. Cambridge:
Cambridge University Press.
Macias, R.F. 1990. Definitions of literacy: a response. In Venezky
et al. (eds), pp. 17-22.
McCarthy, M.J. 1990. Vocabulary. Oxford: Oxford University Press.
McCormick, S. 1992. Disabled readers’ erroneous responses to
inferential comprehension questions: description and analysis.
Reading Research Quarterly, 27, 54-77.
McGinitie, W.H. (ed.) 1973. Assessment Problems in Reading. Newark,
DE: International Reading Association.
McGovern, D., Mathews, M. and Mackay, S.E. 1994. English for
Academic Study: Reading. Prentice Hall International (UK) Ltd.
Mackay, R., Barhman, B. and Jordan, R.R. (eds) 1979. Reading in a
Second Language. Rowley, Mass.: Newbury House.
Mackay, R. and Mountford, A. 1979. Reading for information. In
Mackay, R. et al. (eds), pp. 106-41.
Mackinnon, G.E. and Waller, T.G. (eds) 1981. Reading Research:
Advances in Theory and Practice, Vol 2. London: Academic Press.
McNamara, T. 1996. Measuring Second Language Performance. Long­
man: London.
McNamara, T.P., Miller, D.L. and Bransford, J.D. 1991. Mental
models and reading comprehension. In Barr, R. et al. (eds),
pp. 490-511.
Mahon, D. 1986. Intermediate skills: focusing on reading rate
development. In Dubin, F. et al. (eds), pp. 77-102.
Markham, P. 1985. The rational deletion cloze and global com­
prehension in German. Language Learning, 35, 423-30.
Marslen-Wilson, W. (ed.) 1989. Lexical Representation and Process.
Cambridge, Mass.: MIT Press.
Marton, J.R. 1976. Teaching academic vocabulary to foreign gradu­
ate students. TESOL Quarterly, 10, 91-9.
326 References
Matthews, M. 1990. Skill taxonomies and problems for the testing
of reading. Reading in a Foreign Language, 7, 511-17.
Mattingly, G. and Kavanagh, F.J. 1972. The relationships between
speech and reading. The Linguistic Reporter, 14, 1-4.
Mead, R. 1982. Review of Munby: communicative syllabus design.
Applied Linguistics, 3, 70-8.
Menyuk, P. 1984. Language development and reading. In Flood,
J.S. (ed.), pp. 101-19.
Meyer, B.J.F. 1975. The Organization of Prose and its Effect on Memory.
Amsterdam: North-Holland.
Meyer, B.J.F. 1985. Prose analysis: purposes, procedures, and prob­
lems. In Britton, B.K. and Black, J.B. (eds), pp. 11-64.
Meyer, B.J.F. and Freedle, R.O. 1984. The effects of different dis­
course types on recall. American Educational Research Journal, 21,
121-43.
Meyer, B.J.F. and Rice, G.E. 1982. The interaction of reader strat­
egies and the organisation of text. Text 2, 1, 155-92.
Meyer, B.J.F. and Rice, G.E. 1984. The structure of text. In Pearson,
P.D. et al. (eds), pp. 319-52.
Mikulecky, L. 1990. Literacy for what purpose? In Venezky et al.
(eds), pp. 24-34.
Miller, J.R. and Kintsch, W. 1980. Readability and recall for short
passages: a theoretical analysis. Journal of Experimental Psychology:
Human Learning and Memory, 6, 335-54.
Mohammed, M.A.H. and Swales, J.M. 1984. Factors affecting the
successful reading of technical instructions. Reading in a Foreign
Language, 2, 206-17.
Moore, J. 1980. Reading and Thinking in English. Oxford: Oxford
University Press.
Moran, C. and Williams, E. 1993. Survey review: recent materials
for the teaching of reading at intermediate level and above.
ELTJournal, 47, 64-84.
Morgan, J. O ’C and Batchelor, E.M. 1959. The Approach to School
Certificate English. London: Academic Press.
Morrow, K.E. 1980. Skillsfor Reading. Oxford: Oxford University Press.
Munby, J. 1978. Communicative Syllabus Design. Cambridge: Cam­
bridge University Press.
Nation, I.S.P. 1990. Learning and Teaching Vocabulary. New York:
Newbury House.
Nation, P. 1997. The language learning benefits of extensive read­
ing. The Language Teacher, 21, 13-16.
 References 327
Neal, E.A. and Foster, I. 1926. A program of silent reading. El­
ementary School Journal, 27, 275-80.
Nesi, H. and Meara, P. 1991. How using dictionaries affects per­
formance in multiple choice EFL tests. Reading in a Foreign
Language, 8, 631-43.
Neville, M.H. and Pugh, A.K. 1982. Towards Independent Reading.
London: Heinemann Educational.
Nevo, N. 1989. Test-taking strategies on a multiple-choice test of
reading comprehension. Language Testing, 6, 199-215.
Newman, A.P. and Beverstock, C. 1990. Adult Literacy: Contexts and
Challenges. Newark: IRA and Bloomington.
Nicholson, T. 1993a. The case against context. In Thompson, G.B.
et al. (eds), pp. 91-104.
Nicholson, T. 1993b. Reading without context. In Thompson, G.B.
et al. (eds), pp. 105-22.
Nickel, G. (ed.) 1976. Proceedings of the 4th International Congress of
Applied Linguistics, Vol 1. Stuttgart Hochschulverlag.
Niles, J.A. and Lalik, R.V. (eds) 1986. Thirty-fifth Yearbook of the
National Reading Conference. Rochester, NY: National Reading
Conference.
Nisbet, J. and Shucksmith, J. 1991. Learning Strategies. New York:
Routledge.
Nunan, D. 1985. Content familiarity and the perception of textual
relationship in second language reading. RELCJournal, 16, 34-51.
Nunan, D. 1989. Designing Tasks for the Communicative Classroom.
Cambridge: Cambridge University Press.
Nunan, D. 1993. Task based syllabus design: selecting, grading and
sequencing tasks. In Crookes, G. and Gass, S.M. (eds), pp. 55-68.
Nuttall, C. 1982. Teaching Reading Skills in a Foreign Language. Lon­
don: Heinemann Educational.
Nuttall, C. 1985. Recent materials for the teaching of reading.
ELTJournal, 39, 198-207.
Nuttall, C. 1996. Teaching Reading Skills in a Foreign Language (rev.
edn). London: Heinemann Educational.
Olshavsky, J.E. 1977. Reading as problem solving: an investigation
of strategies. Reading Research Quarterly, 12, 654-74.
Olson, G.M., Duffy, S.A. and Mack, R.L. 1984. Thinking-out-loud
as a method for studying real-time comprehension processes.
In Kieras, D.E. and Just, M.A. (eds), pp. 253-86.
O ’Neill, T., Snow, P. and Webb, R. 1996. Crescent English Course:
Teacher’s Book. Beirut: OUP.
328 References
Orasanu, J. (ed.) 1986 Reading Comprehension: From Research to Prac­
tice. Hillsdale, NJ: Erlbaum.
Orlek, J. 1996. A theoretical and empirical exploration of main ideas in
reading. Unpublished MA TEFL Dissertation. CALS, University
of Reading.
Oxford, R. and Crookall, D. 1989. Research on language learning
strategies: methods, findings, and instructional issues. The Modem
Language Journal, 73, 404-19.
Paap, K.R., Newsome, S.L., McDonald, J.E. and Schaneveldt, R.W.
1982. An activation-verification model for letter and word re­
cognition: the word superiority effect. Psychological Review, 89,
573-94.
Padden, C.A. 1990. Deaf .Children and Literacy. Literacy Lessons.
Geneva: International Bureau of Education.
Palincsar, A. and Brown, A. 1984. Reciprocal teaching of compre­
hension fostering and comprehension monitoring activities.
Cognition and Instruction, 1, 117-75.
Paran, A. 1991. Reading Comprehension. Limassol, Cyprus: Burlington
Books.
Paran, A. 1993. Points of Departure. Ra’anana: Eric Cohen Books.
Paran, A. 1996. Reading in EFL: facts and fictions. ELT Journal,
50, 25-34.
Paris, S.G., Wasik, B.A. and Turner, J.C. 1991. The develop­
m ent of strategic readers. In Pearson, P.D. et al. (eds), pp. 609-
40.
Pask, G. 1976. Styles and strategies of learning. British Journal of
Educational Psychology, 46, 12-48.
Pearson, P.D. and Fielding, L. 1991. Comprehension instruction.
In Barr, R. et al. (eds), pp. 815-60.
Pearson, P.D. and Johnson, D.D. 1978. Teaching Reading Compre­
hension. New York: Holt, Rinehart 8c Winston.
Pearson, P.D. et al. (eds) 1984. Handbook of Reading Research. New
York: Longman.
Pegolo, C. 1985. The role of rhythm and intonation in the silent
reading of French as a foreign language. Reading in a Foreign
Language, 3, 313-27.
Perfetti, C.A. 1977. Language comprehension and fast decoding:
some psycholinguistic prerequisites for skilled reading compre­
hension. In Guthrie, J.T. (ed.), pp. 141-83.
Perfetti, C.A. 1985. Reading Ability. New York: Oxford University
Press.
 References 329
Perfetti, C.A. and McCutchen, D. 1982. Speech processes in read­
ing. In Lass, N. (ed.), pp. 237-69.
Perkins, K and Brutten, S. 1988. An item discriminality study of
textually explicit, textually implicit and scriptally implicit ques­
tions. RELC Journal, 19, 1-11.
Pettit, N.T. and Cockriel, I.W. 1974. A factor study on the literal
reading comprehension test and the inferential reading com­
prehension test. Journal of Reading Behavior, 6, 63-75.
Pitts, M., White, H. and Krashen, S. 1989. Acquiring second lan­
guage vocabulary through reading: a replication of the clock­
work orange study using second language acquirers. Reading in
a Foreign Language, 5, 271-6.
Pollitt, A. and Hutchinson, C. 1986. The validity of comprehen­
sion tests: what makes questions difficult?. In Vincent, D. et al.
(eds), pp. 41-61.
Posner, M.I. 1986. Chronometric Explorations of Mind. Hillsdale, NJ:
Erlbaum.
Pressley, M., Ghatala, E.S., Woloshyn, V. and Pirie, J. 1990. Some­
times adults miss the main ideas and do not realize it: confidence
in responses to short-answer and multiple choice comprehen­
sion questions. Reading Research Quarterly, 25, 232-49.
Pritchard, R. 1990. The effects of cultural schemata on reading
processing strategies. Reading Research Quarterly, 25, 273-95.
Pugh, A.K. 1976. Implications of problems of language testing for
the validity of speed reading courses. System, 4, 29-39.
Pugh, A.K. 1978. Silent Reading. London: Heinemann Educational.
Pugh, A.K and Ulijn, J.M. (eds) 1984. Reading for Professional Pur­
poses: Studies and Practices in Native and Foreign Languages. Lon­
don: Heinemann Educational.
Pyrczak, F. 1975. Passage dependence of reading comprehension
questions: examples. Journal of Reading, 18, 308-11.
Quirk, R. and Greenbaum, S. 1973. A University Grammar of Eng­
lish. London: Longman.
Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1972. A Gram­
mar of Contemporary English. London: Longman
Ramsey, C.L. 1990. Language planning in deaf education. In Lucas,
C. (ed.), pp. 123-46.
Randall, M. and Meara, P. 1988. How Arabs read Roman letters.
Reading in a Foreign Language, 4, 133-45.
Rankin, M.J. 1988. Designing think-aloud studies in ESL reading.
Reading in a Foreign Language, 4, 119-32.
330 References
Rayner, K. and Pollatsek, A. 1989. The Psychology of Reading. Engle­
wood Cliffs, NJ: Prentice Hall.
Reder, L.M. and Anderson, J.R. 1980. A comparison of texts and
their summaries: memorial consequences. Journal of Verbal Learn­
ing and Verbal Behavior, 19, 121-34.
Rendell, R. 1977. A Judgement in Stone. London: Hutchinson.
Resnick, L.B. and P.A. Weaver (eds) 1979. Theory and Practice of
Early Reading. Hillsdale, NJ: Erlbaum.
Richards, J.C. 1989. Profile of an effective L2 reading teacher.
Prospect, 4, 13-29.
Ridgway, T. 1996. Reading theory and foreign language reading
comprehension. Reading in a Foreign Language, 10, 55-76.
Ridgway, T. 1997. Thresholds of the background knowledge effect
in foreign language reading. Reading in a Foreign Language, 11,
151-68.
Rigg, P. 1988. The Miscue-ESL project. In Carrell, P.L. et al. (eds),
pp. 206-20.
Robb, T.N. and Susser, B. 1989. Extensive reading versus skills
building in an EAP context. Reading in a Foreign Language, 5,
239-51.
Roberts, J.R. 1997. Language Teacher Education. London: Arnold.
Robinson, H.M. (ed.) 1960. Sequential Development of Reading Abil­
ity. Chicago: University of Chicago Press.
Robinson, P. 1991. ESP Today: A Practitioner's Guide. London:
Prentice Hall.
Rodrigo, V. 1995. Does a reading program work in a foreign lan­
guage classroom? Paper presented in the Extensive Reading
Colloquium at the American Association of Applied Linguistics.
Long Beach, CA.
Roehler L.R. and Duffy, G.G. 1991. Teachers’ instructional ac­
tions. In Barr, R. et al. (eds), pp. 861-84.
Roller, C.M. 1990. The interaction between knowledge and struc­
ture variables in the processing of expository prose. Reading
Research Quarterly, 25, 79-89.
Rosch, E. 1975. Cognitive representations of semantic categories.
Journal of Experimental Pschology (General), 104, 192-233.
Rosch, E., Mervis, C.B., Gray, D., Johnson, D.M. and Boyes-Braem,
P. 1976. Basic objects in natural categories. Cognitive Psychology,
8, 383-439.
Rosenshine, B.V. 1980. Skill hierarchies in reading comprehen­
sion. In Spiro, R.J. et al. (eds), pp. 535-54.
 References 331
Rost, D.H. 1993. Assessing the different components of reading
comprehension: fact or fiction. Language Testing, 10, 79-92.
Rost, M. 1990. Listening in Language Learning. London: Longman.
Royer, J.M. and Cunningham, D.J. 1978. On the Theory and Meas­
urement of Reading Comprehension (Tech. Rep. No. 91). Urbana:
University of Illinois, Center for the Study of Reading.
Rudzka-Ostyn, B. 1986. Vocabulary teaching: a cognitive approach.
In Cornu, A.M. et al. (eds), pp. 98-106.
Rumelhart, D.E. 1977. Toward an interactive model of reading. In
Dornic, S. (ed.), pp. 573-603.
Rumelhart, D.E. 1980. Schemata: the building blocks of cogni­
tion. In Spiro, R.J. et al. (eds), pp. 123-56.
Ryan, A. and Meara, P. 1991. The case of the invisible vowels: Arabs
reading English words. Reading in a Foreign Language, 7, 531-40.
Sadoski, M., Paivio, A. and Goetz, E.T. 1991. A critique of schema
theory in reading and a dual coding alternative. Reading Research
Quarterly, 27, 463-89.
Saljo, R. 1975. Qualitative differences in learning as a func­
tion of the learner’s conception of the task. Acta Universitatis
Gothoburgensis. Goteborg.
Sampson, G. 1975. The Form of Language. London: Weidenfeld and
Nicholson.
Sampson, G. 1980. Schools of Linguistics. London: Hutchinson.
Sampson, G. 1985. Writing Systems. London: Hutchinson.
Sampson, G. 1994. Chinese script and the diversity of writing sys­
tems. Linguistics, 32, 117-32.
Sampson, G. 1997. Educating Rita. London: Cassell.
Samuels, S.J. and Kamil, M.L. 1988. Models of the reading pro­
cess. In Carrell, P.L. et al. (eds), pp. 22-37.
Saragi, T., Nation, I.S.P. and Meister, G.F. 1978. Vocabulary learn­
ing and reading. System, 6, 72-8.
Sarig, G. 1987. High-level reading in the first and in the foreign
language: some comparative process data. In Devine, J. et al.
(eds), pp. 105-20.
Sarig, G. 1989. Testing meaning construction: can we do it fairly?
Language Testing, 6, 77-94.
Shanahan, T., Kamil, M. and Tobin, A. 1982. Cloze as a measure
of intersentential comprehension. Reading Research Quarterly, 17,
229-55.
Shen, Z., Weir, C.J. and Green, R. 1998. The Test for English Majors
Validation Project. Shanghai: Foreign Languages Education Press.
332 References
Shepherd, D. 1978. Comprehensive High School Reading Methods (2nd
edn). Columbus, OH: Charles E. Merrill.
Shepherd, J., Rossner, R. and Taylor, J. 1984. Ways to Grammar:
a Modem English Practice Book. London: Macmillan.
Shohamy, E. 1984. Does the testing method make a difference? The
case of reading comprehension. Language Testing, 1, 147-70.
Siedow, M.D., Memory, D.M. and Bristow, P.S. 1985. Inservice Edu­
cation for Content Area Teachers. Newark, DE: International Read­
ing Association.
Silberstein, S. 1994. Techniques and Resources in Teaching Reading.
New York: Oxford University Press.
Sim, D.D. and Laufer-Dvorkin, D. 1982. Reading Comprehension
Course: Selected Strategies, Collins Study Skills in English. Glasgow:
Collins.
Sinclair, J. (ed.) 1990. Collins Cobuild English Grammar. London:
Collins.
Singer, H. and Ruddell, R.B. (eds) 1970. Theoretical Models and Pro­
cesses of Reading. Newark, DE: International Reading Association.
Skehan, P. 1996. A framework for the implementation of task
based instruction. Journal of Applied Linguistics, 17, 38-62.
Skehan, P and Foster, P. 1995. Task type and task processing con­
ditions as influences on foreign language performance. Thames
Valley University Working Papers in English Language Teaching, 3,
139-89.
Smith, C.B. 1988. Does it help to write about your reading? Journal
of Reading, 32, 276-85.
Smith, F. 1971. Understanding Reading: A Psycholinguistic Analysis
of Reading and Learning to Read. New York: Holt, Rinehart &
Winston.
Smith, F. 1973. Psycholinguistics and Reading. New York: Holt,
Rinehart 8c Winston.
Spearritt, D. 1972. Identification of subskills in reading compre­
hension by maximum likelihood factor analysis. Reading Research
Quarterly, 8, 92-111.
Spiegel, D.L. and Fitzgerald, J. 1990. Textual cohesion and coher­
ence in children’s writing revisited, Research in the Teaching of
English, 24, 48-64.
Spilich, G.J., Vesonder, G.T., Chiesi, H.L. and Voss, J.F. 1979. Text
processing of domain-related information for individuals with
high and low domain knowledge. Journal of Verbal Learning and
Verbal Behavior, 18, 275-90.
 References 333
Spiro, R.J., Bruce, B.C. and Brewer, W.F. (eds) 1980. Theoretical
Issues in Reading Comprehension. Hillsdale, NJ: Erlbaum.
Spiro, R.J. and Myers, A. 1984. Individual differences and underly­
ing cognitive processes in reading. In Pearson, P.D. et al. (eds),
pp. 471-504.
Spolsky, B. 1995. Measured Words. Oxford: Oxford University
Press.
Stahl, S.A. and Fairbanks, M.M. 1986. The effects of vocabulary
instruction: a model based meta-analysis. Review of Educational
Research, 56, 72-210.
Stanley, R.M. 1984. The recognition of macrostructure: a pilot
study. Reading in a Foreign Language, 2, 148-53.
Stanovich, K.E. 1980. Toward an interactive compensatory model
of individual differences in the development of reading fluency.
Reading Research Quarterly, 16; 32-71.
Stanovich, ICE. 1981. Attentional and automatic context effects in
reading. In Lesgold, A. and Perfetti, C. (eds), pp. 241-67.
Stanovich, K.E. 1991. Word recognition: changing perspectives. In
Barr, R. et al. (eds), pp. 418-52.
Stanovich, K.E., Cunningham, A. and Feeman, D.J. 1984. Intel­
ligence, cognitive skills, and early reading progress. Reading
Research Quarterly, 19, 278-303.
Steffensen, M.S. 1988. Changes in cohesion in the recall of native
and foreign texts. In Carrell, P.L. et al. (eds), pp. 140-51.
Steffensen, M.S. and Joag-Dev, C. 1984. Cultural knowledge and
reading. In Alderson, J.C. and Urquhart, A.H. (eds), pp. 48-
61.
Steffensen, M.S., Joag-Dev, C. and Anderson, R.C. 1979. A cross-
cultural perspective on reading comprehension. Reading Research
Quarterly, 15, 10-29.
Steinert, J. 1978. Allgemeiner deutscher Sprachtest (German gen­
eral language test). Braunschweig/Gottingen: W estermann/
Hogrefe.
Sticht, T.G. 1972. Learning by listening. In Freedle, R.O. and
Carroll, J.B. (eds), pp. 285-314.
Sticht, T.G. 1980. Literacy and Human Resource Development at Work.
Alexandria, VA.: Human Resources Research Organization.
Sticht, T.G. 1984. Rate of comprehending by listening or reading.
In Flood, J.S. (ed.), pp. 140-60.
Sticht, T.G. and James, J.H. 1984. Listening and reading. In
Pearson, P.D. et al. (eds), pp. 293-318.
334 References
Stoker, H.W. and Kropp, R.P. 1960. The predictive validities and
factorial content of the Florida state-wide ninth grade testing
program. Florida Journal of Educational Research, 2, 105-14.
Storey, P. 1995. A process approach to reading test validation. Unpub­
lished PhD Thesis. University of Reading.
Street, B.V. 1995. Social Literacies: Critical Approaches to Literacy in
Development, Ethnography and Education. London: Longman.
Strother, J.B. and Ulijn, J.A. 1987. Does syntactic rewriting affect
ESL for science and technology (EST) text comprehension? In
Devine, J. et al. (eds), pp. 89-101.
Suarez, A. and Meara, P. 1989. The effects of irregular ortho­
graphy on the processing of words in a foreign language. Reading
in a Foreign Language, 6, 349-56.
Sulzby, E. and Teale, W. 1991. Emergent literacy. In Barr, R. et al.
(eds), pp. 727-58.
Swaffar, J.IL 1981. Reading in a foreign language classroom: focus
on process. Unterrichtspraxis, 14, 176-94.
Swales, J.M. 1990. Genre Analysis: English in Academic and Research
Settings. Cambridge: Cambridge University Press.
Sweet, H. 1954. Anglo-Saxon Primer (9th edn; rev. Davis, N.). Oxford:
Oxford University Press.
Taylor, I. and Olson, D.R. (eds) 1995. Scripts and Literacy: Read­
ing and Learning to Read Alphabets, Syllabaries and Characters.
Dordrecht: Kluwer Academic Publishers.
Taylor, W.L. 1953. Cloze procedure: a new tool for measuring
readability. Journalism Quarterly, 9, 206-23.
Taylor, S.E. 1965. Eye movements in reading: facts and fallacies.
American Educational Research Journal, 2, 187-202.
Thomas, L.F. and Augstein, E.S. 1972. An experimental approach
to the study of reading as a learning skill. Research in Education,
8, 28-46.
Thompson, G.B., Tunmer, W.E. and Nicholson, T. 1993. Reading
Acquisition Processes. Clevedon: Multilingual Matters.
Thorndike, E.L. 1917. Reading as reasoning: a study of mistakes
in paragraph reading. Journal of Educational Psychology, 8,
323-32.
Thorndike, R.L. 1973. Reading Comprehension Education in Fifteen
Countries. New York: Wiley.
Todd, M. et al. (eds) 1978. English for Specific Purposes: Science and
Technology. Corvallis, OR: English Language Institute, Oregon
State University.
 References 335
Tomlinson, B. and Ellis, R. 1988. Reading Advanced. Oxford: Oxford
University Press.
Trzeciak, J. and Mackay, S.E. 1994. Study Skills for Academic Writing.
Hemel Hempstead: Prentice Hall International.
Tudge, C. 1987. To read or not to read. New Scientist, 1569, 70-1.
Ulijn, J.M. and Kempen, G.A.M. 1976. The role of the first language
in second language reading comprehension: some experimental
evidence. In Nickel, G. (ed.), pp. 495-507.
Unger, J.M. and J. DeFrancis 1995. Logographic and semasio-
graphic writing systems: a critique of Sampson’s classification.
In Taylor, I. and Olson, D.R. (eds), pp. 45-58.
Urquhart, A.H. 1976. The Effect of Rhetorical Organization in the Read­
ability of Study Texts. Unpublished PhD Thesis. University of
Edinburgh.
Urquhart, A.H. 1987. Comprehensions and interpretations. Reading
in a Foreign Language, 3, 387-409.
Urquhart, A.H. 1996. Identifying the good reader. Triangle 14: To
Read in a Foreign Language. Paris: Didier-Erudition, pp. 17-34.
Van Dijk, T.A. 1977. Semantic macrostructure and knowledge
frames in discourse comprehension. In Just, M. and Carpenter,
P.A. (eds), pp. 1-32.
Van Dijk, T.A. 1980. Macrostructures. Hillsdale, NJ: Erlbaum.
Van Dijk, T.A. and Kintsch, W. 1983. Strategies of Discourse Compre­
hension. New York: Academic Press.
Venezky, R.L. 1984. The history of reading research. In Pearson,
P.D. et al. (eds), pp. 3-38.
Venezky, R.L. 1990. Definitions of Literacy. In Venezky R.L. et al.
(eds), pp. 2-16.
Venezky, R.L. and Calfee, R.C. 1970. The reading competency
model. In Singer, H. and Ruddell, R.B. (eds), pp. 273-91.
Venezky, R.L., Wagner, D.A. and Cilberti, D.S. 1990. Towards De­
fining Literacy. Newark, DE: International Reading Association.
Vernon, P.E. 1962. The determinants of reading comprehension.
Educational and Psychological Measurement, 22, 269-86.
Vincent, D. 1985. Reading Tests in the Classroom: An Introduction.
Windsor: NFER Nelson.
Vincent, D., Pugh, A.K. and Brooks, G. (eds) 1986. Assessing Read­
ing: Proceedings of the UKRA Colloquium on the Testing and Assess­
ment of Reading. London: Macmillan Educational Books.
Vygotsky, L.S. 1962. Thought and Language. Cambridge, Mass.: MIT
Press.
336 References
Vygotsky, L.S. 1978. Mind in Society. Cambridge, Mass.: MIT
Press.
Walker, C. 1987. Individualising reading. ELT Journal, 46, 46-
50.
Wallace, C. 1988. Learning to Read in a Multicultural Society. New
York: Prentice Hall.
Wallace, C. 1992a. Reading. Oxford: Oxford University Press.
Wallace, C. 1992b. Critical literacy awareness in the ESL class­
room. In Fairclough, N. (ed.), pp. 59-92.
Wallace, C. 1996. Critical reading in the foreign language class­
room. Triangle, 14, 80-101.
Wardhaugh, R. 1969. The teaching of phonics and comprehen­
sion: a linguistic evaluation. In Goodman, K.S. and Fleming,
J.T. (eds), pp. 79-90.
Weaver, C.A. and Kintsch, W. 1991. Expository text. In Barr, R.
et al. (eds), pp. 230-45.
Weakley, S. 1993. Procedures in the content validation of an EAP pro­
ficiency test of reading comprehension. Unpublished MATEFL Dis­
sertation. CALS, University of Reading.
Weir, C.J. 1981. A reply to Morrow. In Alderson, J.C. and Hughes,
A. (eds), pp. 26-37.
Weir, C.J. 1983a. Identifying the language needs of overseas students in
tertiary education in the United Kingdom. PhD Thesis. Univerity of
London, Institute of Education.
Weir, C.J. 1983b. The associated examining board’s test in English
for academic purposes: an exercise in content validation. In
Hughes, A. and Porter, D. (eds), pp. 147-53.
Weir, C.J. 1990. Communicative Language Testing. London: Prentice
Hall.
Weir, C.J. 1993. Understanding and Developing Language Tests. Lon­
don: Prentice Hall.
Weir, C.J. 1994. Reading as multi-divisible or unitary: between Scylla
and Charybdis. Paper presented at RELC, Singapore, April.
Weir, C.J. and Roberts, J.R. 1994. Evaluation in ELT. Oxford:
Blackwell.
Weir, C.J., Hughes, A. and Porter, D. 1990. Reading skills: hierar­
chies, implicational relationships and identifiability. Reading in
a Foreign Language, 7, 505-10.
Weir, C.J. and Porter, D. 1996. The multi-divisible or unitary nature
of reading: the language tester between Scylla and Charybdis.
Reading in a Foreign Language, 10, 1-19.
 References 337
Wenden, A. 1987a. Conceptual background and utility. In Wenden,
A. and Rubin, J. (eds), pp. 3-13.
Wenden, A. 1987b. Metacognition: an expanded view of the cog­
nitive abilities of L2 learners. Language Learning, 37, 573-97.
Wenden, A. and Rubin, J. (eds) 1987. Learner Strategies in Language
Learning Hemel Hempstead: Prentice Hall.
West, R. 1991. Development in the testing of reading. English as a
World Language, 1, 60-70.
Widdowson, H.G. 1976. The authenticity of language data. In
Fanselow, J.H. and Crymes, R.H. (eds), pp. 261-70.
Widdowson, H.G. 1978. Teaching Language as Communication.
Oxford: Oxford University Press.
Widdowson, H.G. 1984. Reading and communication. In Alderson,
J.C. and Urquhart, A.H. (eds), pp. 213-30.
Widdowson, H.G. 1979. Explorations in Applied Linguistics. Oxford:
Oxford University Press.
Williams, E. 1984. Reading in the Language Classroom. London:
Macmillan.
Williams, E. 1987. Classroom reading through activating content-
based schemata. Reading in a Foreign Language, 4, 1-7.
Williams, E. 1994. Reading in Two Languages in African Schools. ODA
Project Report.
Williams, E. 1995. First and second language reading proficiency
of year 3, 4 and 6 children in Malawi and Zambia. Reading in a
Foreign Language, 10, 915-29.
Williams, E. 1996. Reading in two languages at year five in African
primary schools. Applied Linguistics, 17, 182-209.
Williams, E. and Moran, C. 1989. Reading in a foreign language at
intermediate and advanced levels with particular reference to
English. Language Teaching, 22, 217-28.
Williams, M. and Burden, R.L. 1997. Psychology for Language Teachers:
A Social Constructivist Approach. Cambridge: Cambridge University
Press.
Williams, R. 1983. Teaching the recognition of cohesive ties in
reading a foreign language. Reading in a Foreign Language, 1,
35-53.
Wilson, P.T. and Anderson, R.C. 1986. What they don’t know will
hurt them: the role of prior knowledge in comprehension. In
Orasanu, J. (ed.), pp. 31-48.
Winter, E, 1994. Clause relations as information structure: two basic
text structures in English. In Coulthard, M. (ed.), pp. 46-68.
338 References
Yang, H. 1986. A new technique for identifying science/techno­
logy terms and describing science texts. Literary and Linguistic
Computing, 1, 93-103.
Zakaluk, B.L. and Samuels, S.J. 1988. Readability: Its Past, Present,
and Future. Newark, DE: International Reading Association.
Zamel, V. 1992. Writing one’s way into reading. TESOL Quarterly,
26, 463-81.
 339

Author Index

Abbs, B.A. 257 Bloor, M. 194

Alderson, J.C. 16, 45, 61-9, 73, 93, Bolinger, D. 260-1
125, 128-30, 132-3, 137, 139, 144, Bransford, J.D. 153
156-7, 257, 264-5 Bright, J.A. 199, 216, 218
Allen, W.S. 259 Bronfenbrenner, U. 119
Allington, R. 229 Brooks, C. 83
Alvermann, D.E. 180, 186-7, 221-2, Brooks, G. 268
224 Brown, A. 245-7
Anderson, J.R. 154 Brown, A.L. 182, 186, 211
Anderson, L. 226 Brown, G. 25, 76, 81
Anderson, N.J. 129 Brown, K. 259
Anderson, R.C. 71, 196 Brutten, S.R. 114, 129
Anderson, T.H. 152 Buck, G. 132
Armbruster, B.B. 152 Burden, R.L. 175, 179-83, 207, 227, 230
Aulls, M.W. 118, 143 Buzan, T. 276
Bachman, L. 120, 147 Calfee, R.C. 47, 83, 148
Baker, D. 145, 256 Candlin, C. 112-13, 175
Baker, L. 245-7 Carnine, D. 124
Bamford,J. 217 Carpenter, P.A. 39, 46, 105-6, 117,
Barnett, M. 159 124, 154, 240, 245
Barr, R. 178, 223, 231 Carr, T.H. 122, 150
Bartlett, B.J. 210 Carrell, P.L. 69, 71, 73, 80-1, 133,
Bartlett, F.C. 70 141-2, 157, 160, 181, 186, 188
Batchelor, E.M. 259 Carroll, J.B. 48-9, 116
Batten, L. 83 Carter, R. 200
Baynham, M. 1, 7 Carver, R.P. 125, 130, 150
Beck, I. 191, 196-8 Cavalcanti, M.C. 72
Bensoussan, M. 163, 191, 238 Chafe, W. 75
Berg, P.C. 127 Chall, J.S. 192
Berman, M. 259 Chapman, L.J. 200
Berman, R.A. 75 Chihara, T. 157
Bernhardt/E.B. 5, 8-9, 13, 33, 49-51, Chikalanga, I.W. 116, 174, 202-3
62-3, 67, 69, 72-3, 93, 157, 159, Chomsky, N. 43, 258-9
162, 164, 177-8, 189, 192, 224-6, Clapham, C. 63, 66, 143-4
230-1, 235, 237-8, 240, 255 Clarke, M. 72, 133, 137, 196
Beverstock, C. 2, 5 -6 Coady, J. 49, 50-1, 62, 133, 137
Bloom, B.S. 91 Cockriel, I.W. 127
340 Author Index
Cohen, A. 74-5, 95, 179-83, 187, 228, Fransson, A. 98-100
253, 270 Freebody, P. 7, 196
Collins, A.M. 70-1 Freedle, R.O. 142
Conant, J.B. 149 Fries, C.C. 17, 22, 47-8, 256
Cook, C.L. 153 Fry, E. 86, 150, 248
Cook, V.J. 258 Fyfe, R. 125
Comu, A.M. 83
Cowan, J.R. 62, 145, 200 Gass, S.M. 175
Criper, C. 96, 134 Gates, A.I. 124
Crookall, D. 181 Geddes, M. 223
Crookes, B. 175 Gee, J. 7
Crothers, E.J. 76, 116 Geva, E. 210, 276
Crystal, D. 258, 260 Gibson, E.J. 46, 109 (n), 235-6, 245, 250
Cunningham, D.J. 143 Glendinning, E. 74, 186, 202-4
Curley, R. 83 Glushko, R. 54
Cziko, G.A. 137, 196 Goodman, K.S. 17, 40, 42-7, 108(n),
137, 193, 239-40
Dansereau, D.F. 210 Goody, J.R. 6
Davies, A. 96, 134, 270 Gough, P.B. 22, 39-41, 58, 108(n)
Davies, F. 76-7, 119 Grabe, W. 6, 91, 119, 122-3, 237, 240
Davis, C. 217-19, 221 Graves, M.F. 153
Davis, F.B. 90-1, 124, 127 Greenbaum, S. 259, 262
Day, J.D. 211 Greene, T. 77
Day, R.C. 56, 219 Grellet, F. 153, 179, 185, 194, 202-4,
de Beaugrande 73-5, 83, 109 (n), 242 211, 214-15, 222
de Leeuw, E. 131, 194 Grimes, J.E. 77-9, 81
de Leeuw, M. 131, 194 Guthrie, J.T. 130, 154
DeFrancis, J. 30
Devine, J. 133, 137 Hafiz, F.M. 216, 220-1
Dhaif, H. 58, 238 Halliday, M.A.K. 73, 75-6, 109(n),
Downing, A. 259 260-1
Downing, J. 167 Hamp-Lyons, L. 184-5, 205, 212, 225,
Dreeben, R. 223 230
Dubin, F. 124, 162, 194 Harri-Augstein, S. 65, 119, 247
Duffy, G. 222, 226, 228 Harris, T.L. 102
Hartman, D. 98-100
Eisterhold, J.C. 186 Hasan, R. 73, 75-6, 109 (n), 260-1
Elley, W. 221 Haynes, M. 150, 191-3
Elley, W.B. 220 Heaton, J.B. 159, 257, 261, 265, 267
Ellis, R. 185, 203, 208, 214 Hedge, T. 217
Entwhistle, N. 46, 98, 100 Hegarty, M. 145
Eskey, D.E. 133, 189, 255 Hiebert, E. 148
Hill, D.R. 217-18
Faerch, C. 149 Hillocks, G. 124
Fairbanks, M.M. 196 Hodges, R.E. 102
Fairclough, N. 7, 68, 88 Hoey, M. 77-80, 82
Farr, R.C. 115, 119, 127, 147, 149 Hoffman, J.V. 225, 227-8, 230
Ferguson, N. 145 Holcomb, T. 23
Fielding, L. 174, 178, 183, 186-7, Holmstrom, B. 186, 202-4
208-11, 222, 225, 227, 276 Hoover, W.A. 40, 46-51, 57, 62, 100,
Fillmore, C.J. 77, 152 102-3, 168(n)
Fingeret, A. 2, 6, 7 Horrocks, G. 259-60, 263
Foster, I. 148, 175 Hosenfeld, C. 43, 188
Francis, G. 263 Householder, F. 25
 Author Index 341
Howatt, A.P.R. 217 Lado, R. 256
Hudson, T. 133, 137, 196 Langer, J.A. 69, 186
Huey, E.B. 22 Laufer, B. 191, 195-6, 238
Hughes, A. 116, 154 Laufer-Dvorkin, D. 201
Hutchinson, C. 163 Lee, J.F. 132, 152, 164
Lefevre, C.A. 22
Inman, M. 145 Leong, K.L. 167
Levelt, W.J.M. 105
Ja’far, W.M. 144 Levin, H. 46, 109(n), 235-6, 245, 250
James, C. 257 Lewkowicz, J.A. 119, 141-2, 206
Jarvis, D.K. 159 Locke, P. 259
Jensen, D.C. 159 Ludlow, L.H. 124
Jimenez, R.T. 188 Luke, A. 7
Joag-Dev, C. 96 Lukmani, Y. 125, 128, 133, 139
Johns, T. 77 Lumley, T.J.N. 124, 129
Johnson, D.D. 116 Lund, R.J. 164
Johnson, D.M. 127 Lunzer, E. 86, 88, 90-3, 97, 125, 134,
Johnson, K. 257 221
Johnston, P. 229 Lyons, J. 25
Johnston, P.H. 145, 147-8, 156,
167-8, 198 Macias, R.F. 6
Jones, B. 181 Mackay, R. 200
Juel, C. 190, 192-3 Mackay, S.E. 185, 211
Just, M.A. 39, 46, 105-6, 117, 124, Mahon, D. 173, 194
154, 240, 245 Mangubhai, F. 220
Markham, P. 157
Kamil, M.L. 87, 117, 133, 157, 242-3 Martin, J.R. 98-9, 145
Kant, I. 70 Matthews, M. 129
Kaplan, R.B. 6 Mattingly, G. 22-3, 32
Kasper, G. 149 McCarthy, MJ. 200
Katz, S. 159 McCormick, S. 84, 114
Kavanagh, FJ. 22—3 McCutchen, D. 57
Kay, P. 152 McGovern, D. 173, 186, 204, 214
Keenan, J.M. 81 McGregor, G.P. 216, 218
Kempen, G.A.M. 60 McKeown, M. 196-8
Khalifa, H. 139, 143, 272 McNamara, T.P. 118-19, 172
King, P. 145 Mead, R. 91
Kintsch, W. 31, 76-83, 104-6, 109 (n), Meara, P. 53-5, 60, 236, 238
116-19, 123-4, 146, 154, 157, Menjuk, P. 23
167, 169(n), 192, 211, 240, 246, Meyer, B.J.F. 74, 77-9, 81-2, 117, 119,
252, 254 137, 142, 158, 164, 210, 238
Kirsch, I.S. 130 Mikulecky, L. 3, 6, 7
Klare, G.M. 146 Miller, J.R. 146
Kobayashi, M. 84, 129, 142, 154, 157, Mitchell, E. 125
164-5, 168 Mohammed, M.A.H. 63-4, 66-9, 72
Koda, K. 54, 56, 58, 236 Moore, D.W. 181, 186-7, 209, 221-2,
Koh, M.Y. 144 224
Kolers, P.A. 41 Moore, J. 83
Koran, J. 145 Moran, C. 90-1, 98, 122-3, 133-4, 172,
Koran, M.L. 145 185, 189-90, 206, 208, 212, 221
Krashen, S.D. 219-21 Morgan, J.O’C 259
Kreindler, I. 163 Morrow, K.E. 201, 214-15
Kretovics, J.R. 7 Mosenthal, P. 154
Kropp, R.P. 127 Munby, J. 3, 90-1, 96-7
342 Author Index
Murphy, D. 175 Ramsey, C.L. 23
Musumeci, D. 132 Randall, M. 53-5, 60
Myers, A. 115 Rankin, M.J. 270
Rayner, K. 23, 27, 39, 41, 43, 45-6,
Nation, I.S.P. 195, 197, 200, 221 51-3, 57-9, 76, 81-2, 85, 91,
Neal, E.A. 148 100-1, 108, 131-2, 235-7, 240,
Nesi, H. 238 245, 247, 249, 251-2
Neville, I. 112 Reder, L.M. 154
Nevo, N. 159 Reid-Thomas, H. 217-18
Newman, A.P. 2, 5, 6 Rendel, R. 2, ll(n )
Nicholson, T. 44 Reynolds, F. 127
Nisbet, J. 180 Rice, G.E. 78, 118, 164
Nunan, D. 3, 175 Richards, J.C. 186, 188, 230
Nuttall, C. 17, 96-7, 151, 153, 172, Ridgway, T. 59, 68, 72, 241
174, 176-7, 179, 184, 186-7, Rigg, P. 239
193-5, 197-8, 200-2, 204, 206, Robb, T.N. 212, 220-1, 225
209-10, 212, 215-19, 221-3, Roberts, J.R. 230
228-30, 248, 252, 276 Robinson, P. 145
Rodrigo, V. 219-20
Olshavsky, J.E. 94-8 Roehler, L.R. 222, 226, 228
Olson, G.M. 270 Roller, C.M. 116, 144
O’Neill, T. 215 Rosenshine, B.V. 125-7, 134, 154
Orlek, J. 187 Rost, D.H. 31, 125-7, 134
Oxford, R. 181 Royer, J. 143
Rudzka-Ostyn, B. 198
Paap, K.R. 52 Rumelhart, D.E. 44-5, 47, 70, 133,
Palinscar, A. 182, 186 188
Paran, A. 173-4, 185, 193-4, 199, 203, Ryan, A. 55
211, 214, 228
Paris, S.G. 131, 228 Sadoski, M. 70
Pask, G. 98-100 Saljo, R. 98-9
Pearson, P.D. 71, 116, 174, 178, 183, Sampson, G. 14, 15, 24-5, 27-32,
186-7, 198, 208-11, 222, 225, 35 (n), 234, 260
227, 276 Samuels, S.J. 87, 117, 133, 242-3
Pegolo, C. 238 Saragi, T. 195
Perfetti, C.A. 16-17, 22, 47, 57, Sarig, G. 94-5, 97, 149, 162, 276
117-18, 192 Shanahan, T. 157
Perkins, K. 114 Shen, Z.C.J. 130
Pettit, N.T. 127 Shepherd, D. 145, 259
Pitts, M. 56, 219 Shohamy, E. 152
Pollatsek, A. 23, 27, 39, 41, 43, 45-6, Shucksmith, J. 180
51-3, 59, 76, 81-2, 85, 91, 100-1, Siedow, M.D. 178
108, 131-2, 235-7, 240, 245, 247, Silberstein, S. 194, 200, 211
249, 251-2 Silbert, J. 124
Pollitt, A. 163 Sim, D.D. 201
Porter, D. 257 Skehan, P. 175
Pressley, M. 114 Smith, C.B. 211
Pritchard, R. 95, 114 Smith, F. 42, 44
Pugh, A.K. 112-13, 122, 131, 147-8, Spearritt, D. 127
154, 215 Spillich, G.J. 143
Pyrczak, F. 159 Spiro, R.J. 115
Spolsky, B. 149-50, 160, 168
Quillian, M.R. 70-1 Stahl, S.A. 196
Quirk, R. 109 (n), 258-9, 262 Stanley, R.M. 81
 Author Index 343
Stanovich, K.E. 38-9, 45, 48, 133, 157, 168(n), 192, 211, 214, 240, 246,
190 252, 254
Steffensen, M.S. 75, 96 Venezky, R.L. 1, 2, 4, 5, 47, 119, 147-9
Steinert, J. 127 Vernon, P.E. 127
Sticht, T.G. 4, 22, 33 Vincent, D. 88
Stoker, H.W. 127 Vygotsky, L.S. 207
Storey, P. 132-3, 149, 162
Street, B.V. 2, 6-8 Walker, C. 207
Strother, J.B. 60-1 Wallace, C. 4, 7, 9, 10, 11 (n), 88, 179,
Sturtridge, G. 223 190
Suarez, A. 54, 236 Wardhaugh, R. 20
Sulzby, E. 190 Warren, R.P. 83
Susser, B. 212, 220-1, 225 Watt, I. 6
Swaffar, J.K. 186 Weakley, S. 129
Swales, J.M. 63-4, 66-9, 72, 83, 144 Weaver, C.A. 81, 116
Weir, C.J. 90, 119-22, 128-30, 132-6,
Taylor, W.L. 156 143, 149-50, 152, 155-6, 158,
Teale, W. 190 160-3, 172, 200, 212, 230
Terrell, T.D. 220 Wenden, A. 180
Thomas, L. 65, 119, 247 West, R. 144, 149, 159-60, 162
Thorndike, E.L. 127, 148 Widdowson, H.G. 17, 20, 63, 98-100
Thorndike, R.L. 127 Williams, E. 4, 90-1, 98, 122-3, 133-4,
Tomlinson, B. 185, 203, 208, 214 173, 179, 185-6, 189-90, 201,
Trzeciak, J. 185, 211 204, 206-8, 210, 212, 217, 219,
Tudge, C. 3 221, 223, 230
Tudor, I. 216, 220-1 Williams, M. 175, 179-83, 207, 222,
Tunmer, W.E. 40, 46-51, 62, 100, 227, 230
102-3, 168(n) Williams, R. 201
Wilson, P.T.
Ulijn, J.A. 60-1 Winter, E. 77-80
Unger, J.M. 30
Urquhart, A.H. 4, 45, 63-9, 74-5, Yang, H. 145
83-5, 87, 112-13, 133, 144 Yarborough, J.C. 154
Yule, G. 25, 76, 81
van Dijk, T.A. 31, 76-83, 104-6, 108,
109(n), 117-19, 123-4, 132, 154, Zamel, V. 211
344

Subject Index

applied linguists 50, 53, 177, 234-5, teaching of 224-5

237, 241, 269 testing of 138, 148, 156, 158-9
authenticity 119, 141, 149, 160-1, 206 correlations 138, 150, 160, 196, 262,
automaticity 157, 188, 190-5, 226, 264-6, 269, 283-4
251
decoding 14-18, 20, 22, 44, 47-8, 88,
background knowledge 18, 45, 49, 93,107-8,117-18,176,189-92, 229
62-6, 72-3, 109, 114-17, 143-4, dictionaries 193, 197-9
185, 207-8, 252, 254, 269
browsing 80, 101, 103-4, 108, 246 expeditious reading 101-2, 118,
130-1, 139, 148, 153-4, 194, 212,
careful reading 46, 86-7, 100-1, 215, 251, 274, 277
103-8, 117, 133, 141, 201-14, eye movements 43, 58-9, 98, 237
225, 245, 248, 281-2
global 106, 118, 123, 125, 129-30, factor analysis 126-8, 130, 137, 139,
139, 253-4, 269, 284-7 166, 283-5
local 123, 125, 129, 139, 148, 157,
176, 253-4, 269, 284-7 genres see text type
classrooms 211, 216, 218, 221-31, 244, gist see also main ideas 107, 163, 179,
248, 255, 271, 295 184, 188, 198, 252, 254
cognitive, goal setter 105-7, 181, 241, 245-7
psychologists 37-8, 56, 190, 234-7, goals 80, 98, 103-4, 119, 131, 175, 189,
239-40 205, 212-14, 243, 245-8, 274-5
training 174-5, 179-80 graded readers 217, 221, 237
coherence 75-6, 79-80, 107, 252 grammar see syntax
cohesion 49, 73-76, 90, 109, 118, grapheme/phoneme correspondence
122-3, 129, 157, 201, 238, 254, 31, 50, 54, 236, 256
264-5 guessing see hypothesis formation
communicative (approaches, etc.) 59,
154, 171, 233, 257, 259, 261, 262, high/low (level, etc.) 50, 53, 60, 118,
265, 267 125, 128-9, 133, 137, 139, 210,
comprehension 18-20, 47-50, 84, 88, 237, 244, 253, 255
100, 117-18, 147, 168, 186-7, 191, hypothesis formation 42-3, 94, 185,
198, 225, 246 190-1, 193, 239
comprehensions 112-16
global 123, 130, 137-9, 157, 167, inferences 114, 127, 208, 266
192, 207-10, 212, 215, 274, 284 pragmatic 116, 120, 174, 201, 203
local 123, 137-9, 215, 274, 284 propositional 116, 123, 201-3
 Subject Index 345
inner speech 27, 57-8 problem-solving 95-6, 98, 114, 200,
interpretations 87, 112-15, 174 227
introspection 270, 273, 286, 289 propositions 77-81
macropropositions 78, 108, 118,
language, 123, 185, 211
and reading 16, 20, 93 micropropositions 78-9, 118, 123
competence 20, 47-51, 134, 262 pseudo-words 41, 48, 51-2, 54, 56,
proficiency 64-5, 68, 72, 137, 161, 236, 249
167, 196 purposes see goals
skills 91, 133-8, 148, 156-7, 201
use 50, 59, 204, 221, 256, 265 reader-driven see top-down
lexical access 48, 124, 192, 254 readers,
direct route 26-7, 31, 52, 54, 236 Arabic 54-5, 141
phonological route 26-7, 31, 48, bilingual 41, 178
52-4, 56, 58, 107 Chinese 29-31, 54
lexicon 56-7, 134, 195, 259, 268 deaf 23, 24, 58
lexis 24, 93, 196, 199, 219-20, 224-5, Japanese 30, 56
266 LI 56, 58, 64, 75, 195
unfamiliar 195, 217 L2 33-4, 53, 56, 58, 63-4, 75, 93,
listening 22-3, 31-3, 62, 132, 264 189, 192, 195, 225, 251-2, 254
literacy 1-9, 217 Spanish 54
as a component of reading 50-1, styles of 46, 98-100, 174
72-3, 93, 125 reading,
aloud 40, 147, 224, 238
macrostructure 80-2, 103-8, 123, 132, and language 18, 34, 49
213, 252-3 cognitive v social approaches to
main ideas 102-3, 118-19, 122-3, 155, 8-10
201-3, 207-9, 281-2 components of 47-51, 123-5
memory 18, 56-7, 79, 105-7, 226, 248 critical 7, 88
metacognitive (knowledge, strategies) definitions of 13, 17, 20
122, 179-87, 214, 226 disorders 47, 53-4
training 160, 174 global 101-2, 125, 136, 194, 201
method effect 65, 121, 139, 160, 277 intensive/extensive 86, 153, 194-5,
microlinguistic skills see language skills 199, 212, 215-22, 254
microstructure 79-80, 106, 118, 132 local 102, 125, 139, 148, 153, 157,
models of reading 194, 240-4 194, 213, 215, 159
bottom-up 39, 41, 46, 240 silent 147-9
compensatory 37, 38-9, 45, 63, 194 speed 86, 151, 192, 194, 248, 252-3
componential 39, 46-51, 165, 177 reading processes, bottom-up and top-
interactive 17, 39-40, 44-6, 63, 121 down 132, 157-60, 188-94, 196,
process 39-40, 79, 105-6, 124 199-200, 202, 213-15
sequential 39, 177-8 recall 96, 141, 152, 162, 164-5,
top-down 40, 42-4, 46 209-10, 253-4
monitor 105-8, 246, 251, 246-7 reliability 111, 120, 148-9, 158, 163,
monitoring 73, 85-6, 91, 114, 183, 173-4, 273, 282-3
186-7, 214, 246-7 rhetorical organization see text
morphology 50, 258-9, 261, 263-4 structure
performance conditions 115, 119-21, scanning 100-5, 107-8, 118, 123, 142,
150, 160, 164-5, 206, 274-6 154, 194, 214-15, 248-9, 251, 269
post-reading 175, 188, 207 schema theory 68-70, 186
pre-reading (previewing, prediction) schemata 70-2
145, 153, 173-4, 183-6, 207-8, content- 69, 71-2, 143
213-14, 223 formal- 69, 71, 73, 80
346 Subject Index
scripts 33 tasks 67, 72-3, 152-3, 175-7, 180,
glottographic 14, 28-31, 34-5, 228-9
54-6, 189 cloze 67, 149, 156-7, 265
semasiographic 14-15, 35 gap-filling 67, 130, 138, 155-6, 265
search reading 100, 103-4, 108, 132, information transfer 160-1, 209
139, 141, 154-5, 214, 249 MCQ 138-9, 149, 158-60, 196,
skills, 228
language 49, 72-3, 93-4, 201-2 SAQ 67, 138, 149, 155, 161-4, 278
multi-divisible hypothesis 92, 126-8, summary 121, 201, 209-11
130, 134-6, 166-7, 270-1, 286, teaching of,
295 skills and strategies 173, 176-7,
reading 49, 72-3, 84, 88-90, 105, 222-7
12 1-8 structures and lexis 176, 191-2
taxonomies 90-4, 96 testing 270-2
unitary hypothesis 92, 125-30, skills and strategies 151, 153-4,
134-5, 166-7, 271, 295 165-8, 231, 271-2, 274, 274-7
v strategies 96-8 structure and lexis 145, 157, 256
skimming 86, 100-5, 107-8, 118, 123, tests (function),
132, 139, 153-5, 179, 213-14, achievement 154, 176, 227
251-4 comprehension 127
speech/speaking 22, 61, 256, 268 EAP 111-12, 128, 138, 144-5
strategies 18, 73, 84-5, 94-100, 105-6, formative and summative 175-6,
121-3, 126, 129, 131, 184, 186, 271
188, 194, 198, 200, 215, 221 language 128
learning 179, 182, 246 placement 155, 176
test-taking 151, 159, 163, 286 proficiency 139, 154-5, 176, 264,
training 181-3, 186-7, 209, 211, 266
228, 231 reading 211, 226, 230, 261, 265-6
subvocalisation 27, 37, 57
syntax 32, 40-1, 45, 58-62, 138-9, writing 14, 26, 75-6, 83, 121, 163,
200-1, 206, 255, 266-9 210-11, 268