Shula Chiat

MMMM
This page intentionally left blank

Some children can hear and can speak, yet have trouble
understanding or producing utterances. In this accessible intro-
duction to childrens language difculties, Shula Chiat explores
the stumbling blocks which lie behind their struggle. The
uniqueness of this book lies in its focus on individual children,
the extensive and intriguing examples which illustrate their
problems, and the step-by-step search for the source of those
problems. Difculties with words, verb structures, function
morphemes and meaning are examined and illustrated by
detailed case studies.
This book provides a clear overview of what children with
language difculties say and do, and introduces a logical
approach to analysing childrens language problems. It will be
welcomed by students of linguistics, psycholinguistics and
speech and language therapy; researchers in the areas of
language development and language disorders; therapists and
teachers working with children with language and communica-
tion impairments; and lay people with an interest in how
children discover their language.
Dr Shula Chiat is Senior Lecturer in Linguistics at the Depart-

ment of Language and Communication Science, City University,
London. She has authored and co-authored papers in Journal of
Child Language, Cognition, Linguistics, British Journal of Disorders
of Communication, Clinical Linguistics and Phonetics, Aphasiology,
Language and Cognitive Processes, Journal of Psycholinguistic Re-
search, Journal of Neurolinguistics and Brain and Language.
MMMM
Understanding Children with Language
Problems
Cambridge Approaches to Linguistics
General editor: Jean Aitchison, Rupert Murdoch
Professor of Language and Communication,
University of Oxford
In the past twenty-ve years, linguistics the systematic study
of language has expanded dramatically. Its ndings are now of
interest to psychologists, sociologists, philosophers, anthropolo-
gists, teachers, speech therapists and numerous others who
have realized that language is of crucial importance in their life
and work. But when newcomers try to discover more about the
subject, a major problem faces them the technical and often
narrow nature of much writing about linguistics.
Cambridge Approaches to Linguistics is an attempt to solve
this problem by presenting current ndings in a lucid and
non-technical way. Its object is twofold. First, it hopes to outline
the state of play in key areas of the subject, concentrating on
what is happening now, rather than on surveying the past.
Secondly, it aims to provide links between branches of linguis-
tics that are traditionally separate.
The series will give readers an understanding of the multi-
faceted nature of language, and its central position in human
affairs, as well as equipping those who wish to nd out more
about linguistics with a basis from which to read some of the
more technical literature in textbooks and journals.
Also in the series

Jean Aitchison: The Seeds of Speech: Language Origin and Evolution
Charles Barber: The English Language: a Historical Introduction
Jean Aitchison: Language Change: Progress or Decay?
Douglas Biber, Susan Conrad and Randi Reppen: Corpus Linguis-
tics
William Downes: Language and Society (second edition)
Loraine K. Obler and Kris Gjerlow: Language and the Brain
Understanding
Children with
Language Problems
S H U L A CH I A T
City University, London

Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo
Cambridge University Press

The Edinburgh Building, Cambridge , United Kingdom
Published in the United States by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521573863
Shula Chiat 2000
This book is in copyright. Subject to statutory exception and to the provision of

relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.
First published in print format 2000
ISBN-13 978-0-511-06788-4 eBook (EBL)

ISBN-10 0-511-06788-7 eBook (EBL)
ISBN-13 978-0-521-57386-3 hardback

ISBN-10 0-521-57386-6 hardback
ISBN-13 978-0-521-57474-7 paperback

ISBN-10 0-521-57474-9 paperback
Cambridge University Press has no responsibility for the persistence or accuracy of

s for external or third-party internet websites referred to in this book, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
Contents
Acknowledgements page ix
Glossary of text conventions and symbols xi
Introduction 1
Part I Problems with words

1 Whats in a word? 13
2 The childs road to words 17
3 Blocks on the road to words 33
4 Exploring the blockage 40
5 Dant always day dings: problems with phonology 66
6 Stip or step or slip or what?: problems with
lexical processing 88
Part II Grappling with verb structure

7 Translating events 113
8 Growing verb structures 127
9 Shortfalls with verbs 144
10 Thing out. Tip in there: problems with verb
processing 163
Part III Missing function morphemes

11 Filling out sentences 183
12 That one not working, see: problems with
auxiliary verb processing 206
viii Contents
Part IV Hidden meanings, bafing
meanings
13 The roots of meaning 227
14 [] you dont tell nobody this?: strengths in
pragmatic processing 243
15 I can speak Chinese. But I cant speak Chinese:
problems in pragmatic processing 251
Endpoint and springboard 262

Further reading 267
References 269
Index 279
Acknowledgements
I would like to thank: the children and young adults who contrib-
uted to this book, as well as parents and teachers, for letting me
into their world so willingly; Sarah Connolly and Anthony Robert-
son, for their readiness to talk about their experience; Elizabeth
Auger, Linda Brett, Dorothea Cave, Rebecca Lacey, Gwen Lancas-
ter, Mary Solomons, Pauline St Leger, Jane Speake, for introducing
me to children in their language units and clinics, and for making
me welcome as an observer and a researcher; Norma Corkish,
Chief Executive of AFASIC, for involving me in AFASIC activities
which gave me the opportunity to meet and talk with many
children and young adults with language disabilities; John
Richards, AFASIC Activity Week Co-ordinator, for putting up with
me on AFASIC trips and for the privilege of learning from his
experience and understanding; Isabelle Barriere, Dorothy Bishop,
Joyce Brown, Alison Constable, Rosemary Emanuel, Peter Hobson,
Susanna Martin and Penny Roy, for their helpful comments on
sections of this book; Bob and Maggie Fawcus, for opening the
doors which led me to children with language disabilities; Sally
Byng, Head of the Department of Clinical Communication Studies,
and colleagues, for their support, particularly for enabling me to
take a sabbatical in order to complete the book. I am indebted to
Jean Aitchison for creating opportunities, and for her active editor-
ial support, careful reading of drafts and perceptive feedback. I am
also indebted to Maria Black and Jane Marshall for extensive and
insightful comments on the text. To both, and to Eirian Jones, I
owe special thanks for sharing in the irresistible pursuit of ques-
tions about language processing and therapy.
ix
x Acknowledgements
Illustrations by Jon Hunt.
Sample cards from the game Guess who? are reproduced with
kind permission of G U E S S WH O ? 1999 Hasbro International Inc.
Shula Chiat
July 1998
Glossary of text conventions and symbols
input Forms in bold highlight technical terms which are accom-

panied by a denition or claried by examples.
cap Forms in italics are used to cite word forms as opposed to
word meanings.
cap Forms between inverted commas are used to cite word
meanings as opposed to word forms.
/kp/ Items between slashes identify forms phonologically, i.e.
in terms of the sound categories of English.
[kp] Items between square brackets identify forms phoneti-
cally, i.e. in terms of their auditory or articulatory charac-
teristics.
Italics are also used to mark emphasis in the text.
An asterisk * before a word or sentence is a convention used by
linguists to indicate that the word or sentence is not well-formed.
Sound symbols
The sounds corresponding to symbols are illustrated by words
as pronounced in Southern British middle-class English. Some
examples would not be appropriate illustrations in other accents of
English.
xi
xii Glossary
Consonant Example Vowel Example
symbols symbols
p pie i seat
t toe tart
k car paw
b bee u moon
d day bird
guy bin
f fur bed
thin bat
s sea cut
shoe hot
v view put
though, mother about
z zoo a pie
beige, measure e hay
t chip boy
d jam hair
m mug ear
n knee poor
sing cow
w way rose
l light
r rock
j yes
# # #
as in to mato or to my house indicates that the following
syllable or word is stressed
Introduction
Some children cant um cant even um they cant even talk
or anything like that and they cant talk properly. And they get
trouble by talking and um things like that.
Ian, aged 9
Ruth: My mum and dad dan me. And my brothers.

Adult: Your mum and dad ?
Ruth: My daddy and mum h they sidan me.
Adult: They send you?
Ruth: No dand me. Beechin.
Adult: What do they do?
Ruth: Nothing!!
This exchange took place between 10-year-old Ruth and myself.
After a precarious start, it appears to have come to a dead-end. But
with a bit of encouragement, Ruth starts again, slowly, weighing
her words:
Ruth: My mum and daddy, those two those stand me.
Adult: Understand?
Ruth: Yes!!
Adult: They understand you?
Ruth: Yeah.
Weve arrived.
Our struggle over words speaks volumes. Ruth has initiated the
conversation. She is quite denite about what she wants to com-
municate. Behind her eventually clear assertion that her family
understand her lies her implicit recognition that other people
dont. I prove her point with my persistent failure to understand
that key word understand. Ruth is ready to give up. It is just too
hard to get her understanding understood.
Damian is 21, and describes himself as dysphasic. He, too, can
have difculty getting his message across. He explains how he
deals with this situation:
1
2 Introduction
Keep going. If it doesnt work, give up. Go on something else. If
someone, like if someone talks to me, and I talk back to them but not in
a nasty sen just ans answer their question and they walk off, I just
look at them and, ah, hell with you, and walk off.
But what does it leave him feeling? Not anger:
Anger is from the your mind
he says, pointing to his head,
A urge is from inside you, in your heart (pointing to his heart). You
want to hit out.
For Damian, it is a relief to talk his own language, which he feels

free to do with his dysphasic friends friends who share his
language disability:
Speaking English, you have to be nice and clear. But if you speak in
dysphasic language it I ca leave all that out and so you can speak
however you want . . . Dysphasics dont bother. Everyone understand.
Some children, as Ian says, get trouble by talking. They may

have difculty understanding words, or they may understand
words in different ways from other children. They may have a
hard time producing words, may leave them out, distort them, or
use them in different ways from other children. In one way or
another, these children will struggle to be understood. What they
encounter is not just misunderstanding of their words. It is also
misunderstanding of their problem with words. All too often,
people hear their halting or unusual speech and jump to the
conclusion that they are lazy, or crazy, or stupid.
For most people, everyday talk is easy and effortless. We have
little awareness of what makes it happen, so we have little idea
what may stop it from happening. The only steps we can register, if
we actively seek to, are that we have something to say and that we
move the tongue and lips to say it. We have no conscious access to
the output processes which go on in-between: the psycholin-
guistic processes which take us from meanings in the mind to
motor movements in the mouth:
Introduction 3
In general, we experience the connections as instantaneous. We

can, by conscious effort, separate out the two ends of output
processing intending a meaning, and making sounds. But how-
ever hard we try, we cannot catch ourselves making the question-
marked connection between these.
The processes by which we understand other peoples utteran-
ces are equally closed to consciousness. We can make ourselves
aware of hearing sound and obtaining meaning from that sound.
But we cannot catch ourselves in the psycholinguistic act of input
processing: of turning the sound which strikes our ears into
meaning.
Since we are normally oblivious to the connections we make
between hearing/speaking and meaning, it is easy to mistake
problems with one for problems with another. Yet obstacles to
speaking or hearing are clearly distinct from obstacles in their
connections to meaning. Children who have problems with the
auditory or motor ends of language processing may have no
problems with language processing itself. They prove the point
when they nd some means of bypassing the barrier to hearing or
speaking and their language processing is freed.
Anthony Robertson, for example, has never been able to pro-
duce speech himself, but this has not stopped him producing
language. He has no problem creating language in his mind, and
with communication aids which provide an alternative outlet for
language, he is able to get that language out. In his early child-
hood, he communicated by using his eyes to point to things in the
environment or to a symbol board. Then at 11, he got one of the
4 Introduction
rst light pointers to indicate symbols. Now in his twenties, An-
thony supplements symbol boards with a sophisticated electronic
device named Liberator, which turns symbols into speech. In
order to communicate with this, Anthony directs his headpointer
at symbols, or icons, corresponding to each word he wants to
produce. An infra-red sensor picks up the combination of icons he
selects, and relays it to the computer. The computer turns it into
synthesised speech.
Using this device, Anthony can tell us in detail about his
experience of communication. The points he makes, and the lan-
guage he uses to make his points, speak for themselves: Anthonys
speech may be blocked, but his language processing is not.
Consider Anthonys explanation of how he uses Liberator:
A: To make a word, it does take two or three icons. Thats how I can
say a lot more than 128 words. [There are only 128 squares on
the keyboard.]
SC: How would you produce the word language?
A: I would use vocabulary because that is programmed.
SC: And how would you produce the word word?
A: Word is phone-God-noun.
SC: Why does God come into it? Just because its one of the icons?
A: The rationale is that in the beginning God communicated the
word.
SC: So you had to learn what the icons were for every word.
A: Yes. Like learning French, I guess.
SC: Its amazing. Each word involves a couple of moves, yet you hold
the whole sentence together.
A: That comes with practice.
SC: So you can hold the sentence in mind as you do each word.
A: Somehow, yes, without thinking hard about it.
SC: You dont take short cuts and miss out the little words?
A: Sometimes at the week-ends I do!
What would happen if Anthony did not have communication

aids to express the words in his mind?
SC: Have you come across people who didnt have your experience?
A: And they are vegetables. Sorry, that is real life if you walk into
most institutions.
Introduction 5
Anthony Robertson with Liberator
When a child cannot produce recognisable speech and does not

have an alternative means to convey words, other people may
wrongly assume that those words are not there. If we think of
language as nothing more than speech, we are bound to see an
absence of recognisable speech as an absence of language. If we
make the further mistake of confusing spoken language with the
meaning intentions it expresses, our judgement of a child with
impaired speech may be even more devastating: the decit in
speech may be taken to reect a decit in thought. Anthony again
articulates this experience:
SC: How do you think people saw you when you were a child?
A: I wasnt thought to be bright till when people sat down and

worked with me.
SC: Was there any period in your life when people didnt nd out what
was in your head?
A: Yes, because of the time factor to be able to communicate. Then
meant getting my board out.
SC: From when you were ve?
6 Introduction
A: Six really, because took some time to learn.
SC:Before you were six, can you remember what it was like?
A: I ought to I used to think like any four years old person.
SC:And could you understand everything?
A: Yes. My understanding is always has been pretty normal.
SC:Did people realise that?
A: Dad and mum knew that, and some people at school.
SC:Although you didnt use speech, when you had thoughts, were
they in words in your head?
A: Yes, and in my dads voice whenever I think of something. You
must understand I didnt get a voice like this [i.e. computer
voice] till eighteen.
So, Anthonys apparent silence does not reect a silence in his

head. A disruption in the ability to move the vocal organs for
speech is not to be mistaken for a decit in language or thought.
Obviously it is vital to distinguish different stages of language
processing and to clarify where disruption occurs.
Deaf children face a different situation. In their case, the impair-
ment will disrupt their processing of language when that language is
spoken. However, if the barrier to auditory input is bypassed, the
childs capacity for acquiring language may be freed. Lip-reading
provides a Deaf person with some limited visual information about
speech. Fingerspelling and writing provide a full visual representa-
tion of words. But the words they spell out are those of a spoken
language which is not directly available. Only Sign language is
entirely independent of the auditory channel, being a language
which connects meaning directly to visual rather than auditory
representations. When Deaf children receive normal input in Sign
language, their acquisition of language proceeds along similar
lines to hearing childrens acquisition of spoken language (Kyle &
Woll 1985).
Sarah Connolly communicates something of her experience as
a Deaf person who uses different modalities to communicate:
SC: Are you completely Deaf or do you hear a little?
S: I am profoundly Deaf 50 per cent hear nothing, 50 per cent can
hear. I hear you with hearing aid.
Introduction 7
Sarah adds the following comments in writing. These are recorded
exactly as she wrote them:
When I was born, I hear nothing which I need to learn the language of
English, most imporation [presumably Sarah means important] is Eng-
lish language around me for life. Very difcult to learn the language, I
am too young, actually my parents, sister and brother teach me. So I
used my eyes to watch the Sign language, lip-reading and writing and
when I getting older I learn the more and more English language and
Sign language.
We continue in speech, supported by some use of Sign:

SC: Is it difcult to lip-read?
S: Sometimes difcult, sometimes easy . . . If Im very tired, I cant
watch lip-read, so I use Sign.
SC: And when youre at home with your husband?
S: Sign all the time.
SC: Why?
S: Because David use Sign all the time. [Sarah adds in writing] David
can do the language because his family are Deaf so they always
used Sign language.
Sarahs experience at her work-place provides an example of what

happens when a Deaf person enters a hearing environment:
SC: Your job was what?
S: Laboratory technician at Wellcome [Sarah ngerspells these
words].
SC: What happened at work?
S: When I met them rst, they dont understand how communicate
with me, so I teach them how communicate write out or
lip-read. Much improve. One person did want learn Sign language
because she like involve Deaf community, I think. I think she want
easy communicate with me, and help us understand language.
SC: When you meet people who dont know you, how do they respond
to you?
S: Two different ways. The rst one when people talk to me I say Im
sorry I am Deaf. Another one when I talk, they know Im Deaf
because my voice or speech different. Or I tell them Im Deaf before
we talk.
SC: Do people ever think youre silly because you talk differently?
8 Introduction
At this point Sarah switches to writing:
Some people think that I am silly when I talked different languge
because they dont know or understand the language what Deaf
people used or Deaf community, some people knows already because
they might have Deaf friends or family learning the language.
SC: What do you feel when people think you are silly?
S: I feel not very happy, so I feel people must learn Deaf community
or the language what Deaf people using so Deaf people do learning
the hearing community or language. So it is fair!
As Sarah says, her speech and voice are different from hearing
peoples; as the transcription of our dialogue shows, the organisa-
tion of her words and sentences also differs from a hearing persons
in certain respects. These differences reect the different routes to
English which Sarah has taken in the face of her lack of hearing.
They do not reect differences in her potential for making connec-
tions between the meanings and forms of language.
Children who cant hear or cant speak can nevertheless ac-
quire language normally. Conversely, some children can hear and
can speak, yet they do not talk like other children. For these
children, problems must be arising at some point beyond the
processes to which we have conscious access, in those hidden
processes which turn heard sound into meaning, or meaning into
the motor movements of speech. This book is about these children,
and the search for the hidden processes which give rise to their
unusual language.
The book
Part I introduces this psycholinguistic enterprise. Its focus is
words. It unpicks the process of word-learning, developing a
theoretical framework and techniques for exploring a childs dif-
culties with words. With these theoretical and methodological
tools in hand, individual childrens difculties with words are
followed up. Preliminary observations of their language give rise
to a hunch about their difculties which leads into detailed investi-
gation.
The inquiry then turns to different types of words, in particular,
Introduction 9
those on which sentences depend. Part II looks at childrens pro-
cessing of verbs and the syntactic structures which they de-
mand. Part III focuses on their processing of function mor-
phemes.
Part IV steps back from connections between sound and mean-
ing, and digs deeper into meaning and function in language and
their relationship to the ways in which human beings process
experience.
Throughout the book, individual children are centre stage. The
point is, rst, to bring their language to life: to show how it is
limited, or odd, or bafing, but also expressive, or moving or,
paradoxically, articulate. Where is it coming from? Focusing on
the individual child, we explore the particular ways in which
language processing is limited or blocked, and the particular ways
in which the child has negotiated the obstacles. This is illustrated
by in-depth case studies in each section of the book.
This book is both introductory and searching. It spells out the
thinking behind much of the research into the nature of speech
and language impairments in children. At the same time, it serves
as a springboard for investigations which delve deep into aspects of
language processing in children who are or are not developing
normally.
The book is for those who want or need to know more about
what is going on in the mind of a child with a language disability.
It is also for those who are fascinated by the language of children,
normally developing or otherwise. They may be practitioners or
students in speech and language therapy, students in linguistics or
psychology, parents, or teachers. Researchers in the eld may nd
fresh insights and questions in the extended evidence and dis-
cussion of specic aspects of language processing.
In a different sense, the book is for children with language
disabilities. It is an invitation to listen to what they tell us and to
understand better what lies behind what we hear, bringing us
closer to childrens individual experience of language. The book
does not deal directly with therapy or therapeutic techniques. But
its insistence on exploring and understanding the individual
childs language is integral to the therapeutic process. The more
sense we can make of the childs language processing, the more we
10 Introduction
can respond and initiate in ways which make language more
comfortable for the child, and optimise her potential for processing
it.
Note: To clarify my use of gender-marked personal pronouns when refer-

ring to the child: I use feminine and masculine forms in alternate
chapters throughout the book.
Part I
Problems with words
MMMM
1 Whats in a word?
Seven-year-old Steven struggles to name a microphone:

S: Oh this is tricky . . . when you on . . . its when youre on
. . . like on a concert . . . and you get this . . . micro . . .
scope . . . no . . . like thing when you speak into it and its
even louder. (From Hayman 1996)
Six-year-old Eamonn requests a favoured toy:

E: I want that . . . deepersiper.
SC: The what?
E: [j] ( = you know) the deeper . . . Where deedeepsiper? I dunno
deedeepsiper is. Wheres the deepbersiper? The deepdeepsiper?
Eamonns search is for the deep-sea diver.

Children who have problems with language have problems
with words. Typical descriptions of such children state that they
are unable to understand words, nd words, or say words, or that
they omit words or use words in odd ways.
Such descriptions are no more than a starting-point for looking
at their difculties. To probe these further, we need to consider
what it is the child cant understand, nd, say or put in the right
place. This means knowing what words are. We can then consider
what is involved in coming to understand, nd, say and combine
them, and what might stop a child from doing one or more of these
things.
While it seems obvious that intact language users know words,
it is by no means obvious what it is that they know. Once we go
beyond the obvious to question what it is that we know, we nd
13
14 Problems with words
that words are not simple things. Each word is a collection of
rather special sorts of information.
Take the word hair. To know the word hair is to have
represented in your mind a connection between the following:
( Phonological information: the sound pattern of the
word. In this case, a single syllable containing the se-
quence of sounds /h/ (in accents where the nal r is not
pronounced).
( Semantic information: the meaning to which the
sound pattern can refer. In the case of /h/, this is the
clump of stuff that grows on the human head or in
varying quantities on other parts of human beings and on
animals.
( Syntactic information: the position which the word
may occupy in relation to other words. This information
would include the fact that hair may be preceded by the
determiner the and by adjectives such as long, and that
it will occur in certain positions within the sentence such
as before a verb (Hair grows), following certain verbs
(He cuts hair) and so on.
Each of these pieces of information is abstract in the sense that
it is picked out or abstracted from another level of information.
The phonology of the word is the sound pattern, and not the sound
signal which occurs when the word is uttered. Nor does it include
every detail of that sound signal. For example, knowing the word
hair does not involve knowing the loudness with which it might
be uttered. It does involve knowing the consonantvowel sequence
/h/. This means picking out those features of the sound signal
that differentiate it from other sequences of consonant and vowel
such as /d/ and // (dare and share). Similarly, the semantics
of the word hair does not include every detail of the thing it might
refer to. Knowledge of the word does not include the number or
shades or length or density of hairs on a particular head.
In summary, words are not concrete objects out there in the
world. Their existence consists in a connection between phono-
logical, semantic and syntactic information, and that information
is picked out from the stream of information we receive.
Whats in a word? 15
Different languages, of course, have different words. Most obvi-
ously, languages differ in the phonological form which they attach
to a meaning. Part of learning a new language is learning the
phonological forms of words. But languages also differ in the exact
semantics which they attach to phonological forms. For example,
the phonological form hair in English refers to the clump of stuff on
the head and elsewhere on the body. In contrast, French uses a
distinct phonological form for hair on the head (cheveux) and hair
on other parts of the body (poil).
Languages also differ in the syntactic properties they attach
to particular phonologicalsemantic pairings. In English, hair
when referring to stuff on the head is not used with plural endings
or with a plural verb: Mona Lisas hair is long is ne, but Mona
Lisas hairs are long is not at least not if the intended reference is
to hair on the head! In French, the converse is true, so that the
French version of Mona Lisas hair is long would make hair
plural, and in fact mark the verb (are) and the adjective (long) as
plural.
Each word, then, is a phonologicalsemanticsyntactic com-
plex. To know words is to have stored such complexes in our
minds, in what is termed our mental vocabulary or lexicon. But
knowing words involves more than knowing each such complex.
Our mental vocabulary is not an undifferentiated collection of
such complexes. Words share properties with other words. They
may be phonologically similar to each other. Hair, for example, is
phonologically similar to fare and chair and bear in sharing the
rhyme // (the variable spelling of that rhyme being irrelevant).
It is phonologically similar to house in sharing the initial conson-
ant /h/. It is phonologically identical to hare, though spelled differ-
ently. Words such as hair and hare, which are phonologically
identical but semantically distinct, are known as homophones.
Phonological relationships between hair and other words
Homophone hare
Shared rhyme fare, chair, bear . . .
Shared initial consonant house, hat, hen . . .
Semantically, hair is most similar to words which refer to parts

of the body such as skin and nail. It has more in common with
words referring to concrete substances such as wire or soil or
bread than to concrete entities such as table or book. It has
even less in common with words referring to abstract things such
as idea or message, or to places such as library or park, or to
states such as joy or poverty, or to events such as revolution or
accident.
Syntactically, though, hair shares properties with all these
words, which belong to the category noun. This means they may
be preceded by a determiner and adjective:
Det Adjective Noun
the famous skin/nail/wire/soil/bread

weird table/book
new idea/message
library/park
joy/poverty
revolution/accident
But of all the above nouns, hair has most in common with skin,
soil and bread, which are distinct from other nouns in certain
respects. These nouns may, for example, occur without a deter-
miner where other nouns may not:
Hair should be kept clean
Skin should be kept clean
*Table should be kept clean
*Library should be kept clean
Nouns like hair (as on the head) form a subcategory known as
mass nouns. They are distinguished from other nouns, known as
count nouns, in a number of ways, one being the possibility of
occurring without a determiner.
Looking at just the word hair illustrates amply the idea that a
word consists of connections between phonological, semantic and
syntactic information, and is connected to other words by virtue of
sharing some of that information.
2 The childs road to words
In acquiring language, children acquire thousands of words.

What they are doing is establishing thousands of connections
between phonological, semantic and syntactic information.
The rst evidence of this process emerges around 9 to 12
months when children make a consistent connection between a
form they hear and a thing or action to which it refers. They may
point to the appropriate creature on hearing the word tiger, or
move their hands appropriately on hearing the word clap. By the
time they understand 50150 words, around 1 to 112 years, they
start producing words (see Ingram 1989). Their lexical acquisition
then proceeds apace. It is estimated that they know some 8,000
root words by the age of 6, which works out at an average rate of
ve new words a day, assuming they kick off at 18 months (see
Carey 1978 and Clark 1993).
Children absorb words like sponges. How are they doing this?
They are not presented with vocabulary lists which pair phono-
logical forms and meanings. They start from the scenes they
observe and participate in, and the stream of speech they hear. In
order to acquire words, they must segment that stream of speech,
separating out the sound pattern of each word. They must seg-
ment the scenes in which the speech occurs, separating out the
aspects of the scene which are picked out by the words. And they
must map the one onto the other, making connections between
the sound patterns they have segmented and the aspects of scenes
they have segmented.
Construed in this way, the childs task looks daunting. Imagine
a child hearing the utterance Theres a tiger as he looks at a
17
striped creature strutting round (see illustration).
How does the child know which aspects of the scene are the crucial
ones, without already having the word to pick these out for him?
How does he know that tiger refers to that type of striped creature
(i.e. tiger) rather than to just any creature (i.e. animal) or that
particular creature (i.e. its name), or something else in the scene
(such as fence), or even the entire event (i.e. strutting, or tiger
strutting or animal strutting)? Equally, how does the child know
which chunk of sound is the crucial one? How does he know that
tiger is a sound pattern with a meaning as a whole, rather than tige
or atiger or theresatiger? And how does he know which chunk of
sound maps onto which chunk of meaning? How does he know
that the chunk of sound corresponding to the striped creature is
tiger rather than theresa? These questions are crucial pieces in the
puzzle of word acquisition: how does the child segment the form
from the stream of speech, segment the category from the scene,
and map the one onto the other?
Perhaps the puzzle isnt that hard if we take into account the
contribution of those who provide the language input to children.
Perhaps, when people talk to children, they smoothe their path to
words by removing some of the complications we have found
The childs road to words 19
along that path. Suppose that, instead of being bathed in streams
of speech, children are presented with isolated word forms and a
clear indication of what they refer to. In this idealised situation,
the child hears just tiger as the speaker points out a tiger. This
would seem to solve the problem of phonological segmentation,
since the sound pattern is separated out for him. It would seem to
solve the semantic segmentation problem, since the relevant crea-
ture has been picked out for him. And it would seem to solve the
mapping problem, the connection of one to the other, since there is
just one form and one meaning to connect up. But how does the
child know that the speaker has helpfully isolated the whole form
of the word, and know not to segment tiger into two words tige and
er, with just tige referring to that striped creature? And how does
he know that the speakers gesture indicates the whole creature,
not some part of the creature, or something it is doing, in which
case tiger might mean black and white stripes on ginger back-
ground, or strut about? Presenting a single word may appear to
hold out a linguistic hand to the child, but it is only helpful if the
child knows that it is a single word. Such knowledge could not
itself come from the environment. Hence it would have to stem
from the child. Through this example, we expose a logic in the
process of word learning: any adaptations in the environment can
only cue the child if he is sensitive to those adaptations. He can
only take advantage of a cue if he notices it (see Newport, Gleitman
& Gleitman 1977 and Gleitman, Newport & Gleitman 1984).
The childs task is still more awesome when it comes to words
encoding abstract things, or events, or relations. Here the segmen-
tation and mapping problems are even more striking. Take words
which refer to abstract notions such as friend or idea, or events
such as chase or persuade, or attributes such as huge or
peculiar, or spatial relations such as at or across. Such words
are rarely uttered in isolation, and cannot be pointed out. Words
which mark time relations such as has in He has escaped or
possessive relations such as my in my tiger or specicity such as
the in the tiger never occur in isolation (unless they are being
cited), and it is impossible to imagine what pointing them out
would mean.
It follows that the majority of those thousands of words which
emerge over a few years of a childs life are acquired from exposure
to streams of speech and streams of events. The puzzle remains. If
the environment is not simplied for the child, the child must come
prepared in some way to deal with the complexities it presents. He
must arrive at the word-learning task with some built-in clues
about what to extract from the environment.
Turning to the child, what sensitivities does he bring to word
learning and what cues might he notice? Research has revealed a
variety of expectations which children have about word semantics
and word phonology, which go some way to explaining how the
child overcomes the semantic and phonological segmentation
problems. They lead the child to focus on certain aspects of the
stream of speech and stream of events, and to lter out innite
other aspects.
Filtering scenes
Children are not blank screens registering every feature of the
scenes which pass them by. They constantly lter the myriad
stimuli they receive, attending to certain aspects rather than
others. This ltering process leads them to make sense of what
goes on around them; it also leads them to respond and interact
with their environment in ways which make sense to others.
The process of ltering scenes is itself quite involved. Even
childrens very early use of words shows them talking about things
from different perspectives (Clark 1997). It is true that their words
seem to favour certain components of scenes, evident from the
order in which they acquire different types of words and the
meaning they attach to words. But they are not conned to these
favoured components.
Dominant amongst their earliest words are labels for familiar
objects, people and animals. This indicates that they have focused
on these entities. It also indicates their preferred assumption that a
word refers to a whole entity, and not to some part or property of it.
Experimental evidence conrms that children make this assump-
tion. Young children who are presented with labels for parts of
unfamiliar objects treat the part-label as a label for the whole
object (see Clark 1993 for evidence and discussion).
It has also been observed that children start out with what have
been identied as basic-level terms (see Markman 1989 for a
review of the evidence). These terms pick out categories which lie
between the very specic and the very general, dog for example
being a basic-level term relative to the more specic spaniel and
the more general animal. These are, of course, the terms that
adults most typically select when talking to children (Brown
1958). We are much more likely to label the thing in the garage as
a car than as a hatchback at one extreme or a vehicle at the
other. But any such bias towards basic-level terms cannot account
for the childs acquisition of these as basic-level terms. Every time
children hear a basic-level term, its reference is ambiguous. They
could justiably conclude that the terms refer to a more general or
a more specic category. And occasionally they do. For example,
they may overgeneralise, using a term such as dog for animals
other than dogs. Or they may undergeneralise, restricting dog
to just one particular creature. However, such misconstruals are
not that frequent, and tend to apply to childrens earliest acquired
words (Rescorla 1980). This shows that children are themselves
prioritising basic-level categories, and assuming that terms used
by adults refer to these.
But biased as they may be towards words which pick out
basic-level categories of things, they are open to other possibilities.
They do not assume that all terms refer to basic-level categories.
Their earliest words typically include names for signicant individ-
uals Mummy or David or Fido. Again, mistakes are not
unknown. For example, children at an early stage may use daddy
to refer to any man. They may utter mummy to express some-
thing like I want . . . as they reach for something. But such
misconstruals are rare and brief. In general, children restrict these
terms to the appropriate individual. This suggests that certain
individuals in certain relationships are as salient to the child as
basic-level categories. The individual mother is not primarily
perceived as one of the category woman, but as an individual in
a specic relation to the child. Children typically focus on things
and people, but they are not limited to one perspective on these.
Nor is their focus entirely restricted to things and people. They
also notice events, states, and relations in which things and people
participate. This is evidenced by their early use of terms such as
there, more, mine, hot, broken, open, gone, down, here
you are. These terms refer to the location, state or possession of an
entity, or to changes in such states of affairs. (See Nelson 1973,
Greeneld & Smith 1976.)
If children are attentive to different aspects of scenes, how do
they know which aspect of a scene is the focus of the words they
hear? Perhaps this is where others step in and help the child, by
highlighting in some way the relevant aspect of the scene. At rst
sight, studies of adultchild interaction in a variety of cultures lend
support to this possibility.
One of the commonly observed features of child-directed lan-
guage is that it relates to the here-and-now. Adults talk to toddlers
about what has just happened, is happening, or is about to happen
in their immediate environment, not about events which are
removed in time or space: Oh no, the towers fallen over, Look
Sam, theres an aeroplane, Were going to have lunch in a
minute. Children are much more likely to attend to what we are
talking about under these circumstances than if we talk about
what will happen when they grow up or what we watched on TV
while they were asleep, and their attendance is crucial. It is a
prerequisite for working out the possible meaning of the utterance
and hence of the words it contains.
But adults might direct children to meaning in more precise and
subtle ways than this. The timing of adult utterances has been put
forward as an important cue to the child. An early assumption was
that simultaneous occurrence of words and what they refer to
would afford the child the most straightforward evidence of what
the words mean. If the child hears Thats a tiger when there is a
tiger present, it should be most obvious what tiger refers to. This
assumption can be investigated by checking out whether children
do acquire words more easily if they are presented in this so-called
ostensive condition (pointing). It turns out that they do if the
words are object labels. An even more effective version of this
learning condition is the joint attention episode, where the
adults label for an object follows the childs focus on that object,
rather than directing the childs attention to a different object. For
example, if the adult says What a beautiful tiger when the child is
already looking at or playing with a tiger, the child is likely to learn
the word more easily than if the adult says Come and see this
tiger when the child is playing with an elephant. (See Tomasello &
Todd 1983, Tomasello & Farrar 1986, Tomasello & Kruger 1992
for evidence and discussion.)
But not all words are learned best under ostensive conditions
where the word is heard and its meaning pointed out at the same
time. Evidence is emerging that verbs may be learned more easily
in non-ostensive conditions, and that the best conditions for learn-
ing verbs may depend on the event to which the verb refers. In an
experiment where children were taught a new verb under different
timing conditions (Tomasello & Kruger 1992), it was found that
the verb was learned better when it was presented before the event
than at the same time as the event. The novel verb in this experi-
ment was plunk, and the novel event was the pushing of a button
to make a doll roll down a ramp and into a car or helicopter seat.
Children who heard an utterance such as Look, Jason, Ill plunk
the man before pushing the button were better able to understand
and produce the new verb plunk than children who heard an
utterance such as Look, Jason, Im plunking the man as the doll
began to roll down the ramp. From this, we might infer that an
utterance heard before a noticeable event occurs will alert the child
to look for a verb referring to the event.
The timing cue may be more complicated than this, however.
Perhaps what is crucial is a combination of the timing of the
utterance and what is salient in the scene. If children hear an
utterance when an event has been initiated and is happening, they
might focus on the event and take little notice of the utterance.
Suppose, though, that the event is such that it produces an effect
which is still observable following the event. Under these condi-
tions, children might take more notice of an utterance which refers
to that event after it is over, when the effect is salient. An investiga-
tion into childrens learning of two new verbs supported this
possibility (Ambalu, Chiat & Pring 1997). Children were presented
with a novel event in which the experimenter used an ofce stamp
to print on different types of paper. The novel verb bock was used
to refer to this event. Children who heard utterances such as Look,
I bocked the paper learned the novel verb bock better than
children who heard utterances such as Look, Im going to bock
the card. The same children were also presented with a novel
event in which the experimenter spun an object on a spinning
wheel, and which was labelled by the novel verb pog. In this case,
where the salient feature of the event was the causing of move-
ment rather than the causing of an effect, children who heard the
verb before the event (Look Im going to pog the ring) learned the
novel verb better than those who heard it afterwards (Look, I
pogged the ower).
Who tunes into who?

It seems that variations in the timing of utterances relative to
events can make it more or less easy for children to nd out the
connection between the two. This nding is just one in a rapidly
growing jigsaw of ndings regarding the cues children exploit to
discover the reference of words. The picture which emerges from
these ndings is that children are doing much more than simply
connecting words they hear to what they are currently looking at.
A key piece of evidence is the way they react when they are
focused on one novel object, and the speaker focuses on and labels
a different novel object (Baldwin 1993). In this situation, children
of 1;61;7 who heard the speaker use the novel label peri took this
to refer to the object of the speakers focus rather than their own:
this was the one they selected when they were confronted with
both objects and asked to point to the peri. This shows that they
knew to use the speakers visual focus as a cue to the reference of a
novel word the speaker produced.
But children do not use this cue indiscriminately. They can pick
up cues that the object the speaker is looking at is not the intended
focus, and not what a word refers to (Tomasello, Strosberg &
Akhtar 1996). They did just this in an experiment where they
heard the experimenter announce Lets nd the gazzer, then saw
her pick up an object, frown, and reject it, before picking up the
intended object with glee. Children of 1;6 learned the word gazzer
just as well in this situation as when the experimenter found the
target object straightaway. Two-year-olds picked up similar cues
that an action they observed was not the speakers intended focus
(Tomasello & Barton 1994). Here, the experimenter announced
Lets hoist Big Bird, performed an accidental action and said
Woops, then performed the target action and said There. Again,
children recognised that the new verb hoist referred to the second
action rather than the rst. They learned the verb just as well
when they had to pass over the rst accidental action as when
they saw the intended action straightaway. These ndings indi-
cate that very young children can use non-verbal expressions of
disappointment and glee and verbal expressions such as Woops
as cues to a speakers intended focus.
These studies show children picking up a variety of clues to the
speakers focus of attention. They leave us in no doubt that very
young children know a word refers to what the speaker is focusing
on, regardless of their own focus. They nd out the meanings of
words by identifying what the speaker is focused on at the time of
using them. They will do this most easily if they already share the
speakers focus. This explains why it is easiest for them to learn an
object name when they and the speaker are jointly attending to
the object. In the absence of such joint attention, it is they who
tune into the speakers attention and thereby nd out what the
speaker means. (See chapter 14 and Tomasello 1995 for further
discussion.)
Filtering speech
Filtering scenes to pick out likely word meanings is not enough
to acquire words. Children face the further challenge of picking out
speech forms. Words usually occur in a stream of speech rather
than in isolation. Children are no more blank tapes registering
every aspect of that stream of speech than they are blank screens
registering scenes. Again, their earliest words provide evidence of
the ways in which they lter the spoken input they meet.
Children do sometimes produce a chunk of words with no
indication that they have separated out the words it contains
(Peters 1983). These are rather like adult routines such as Thank
you, How do you do, Good morning. The evidence that they are
chunks is their rigid use. The child attaches them to a scenario as a
whole, never varying the words inside them. This can result in
inappropriate use:
Sit my knee (child wanting to sit on adults knee)
I carry you (child wanting to be carried) (Clark 1974)
Good girl (said to father as he lays down to sleep)
(A. Hirson, personal communication)
But as they break chunks down, which they rapidly do, children
prove very adept at segmenting words (Brown 1973). The vast
majority of their early words are appropriately segmented. Almost
without exception, the forms they produce correspond to the
phonology of just one word, no more and no less. Children might
hear the form tiger in something like the following contexts:
Look, a tiger
Wheres your tiger?
Theres a tiger and an elephant
as well as in isolation. This does not lead them to connect atiger or
yourtiger or tigerand to the striped creature they pick out. We know
that they segment just the chunk tiger, since this is the chunk
they attempt to produce when they refer to that creature (even if
they do not produce it perfectly).
A striking characteristic of childrens early words in English is
that they are stress-carrying words. They are words which hold
prominent positions in the rhythm of an utterance and which
must contain full vowels, rather than unstressed words whose
vowels can reduce (be shortened). They are words like teddy,
car, stop, broken, there, mine, up, off, rather than words
like a, the, he, at, from, must which can be reduced giving
[], [], [], [t], [frm], [ms]. In addition, these words frequent-
ly occur as the last item of a phrase or sentence, or in isolation, as
in
Thats mine
Do you want to get down?
Take it off
Oh, its broken
There
In these positions, they are most likely to carry the greatest stress
in the utterance. So, it looks as if children are initially picking out
words which are themselves stressed and which are likely to be the
most stressed word in utterances they hear. This opens up the
possibility that children are sensitised to stress, and that stress
serves as a cue to key phonological units. If they home in on the
most prominent part of the utterance, and treat this as the start of
a word unit, they will successfully break into the stream of speech
and segment their rst words.
A variety of empirical evidence supports the inference that
stress is a signicant cue in childrens early speech processing. For
example, when children imitate sequences of syllables or words,
they tend to imitate selectively, and it is the stressed items which
they reproduce (see Chiat 1979 for examples and discussion).
Evidence also comes from the errors children make. When they
omit or group syllables incorrectly, this generally involves an
unstressed syllable. For example, when children meet words
which begin with an unstressed syllable in English, they are prone
to produce the word from the stressed syllable onwards, omitting
that initial unstressed syllable. This gives rise to familiar childhood
# # # # # #
forms such as mato, raffe, nana, jamas for to mato, gi raffe,
# #
ba nana, py jamas.
Though errors in word segmentation are rare, where they do
occur, they appear to involve the attachment of an unstressed
syllable. In some cases, the child fails to separate out an unstressed
syllable which is a separate word and incorrectly treats it as part of
a word. For example, a child is observed to use that a as a unit,
treating a as part of a word which it frequently follows. As a
result, a occurs in contexts where it should not:
That a dog
That a book
That a my book
That a Uncle Clyde
That a screws (Brown 1973)
Conversely, an unstressed syllable which is part of a word may be

treated as separate from it:
Adult: Behave!
Child: I am being have
We can be in little doubt that children are biased to notice stress
in utterances. But this is just the tip of an iceberg which stands out
sufciently in the childs own speech to be readily noticed by
observers. The iceberg is much more than this visible tip. When we
look below the surface it becomes apparent that children are able
to register whole rhythmic patterns and to discriminate ne dif-
ferences in rhythmic timing which go well beyond the discrimina-
tion of the most prominent element. Such subtle discrimination is
necessary to distinguish between single words and groups of
words, reected in contrasts such as:
tomato vs. to my toe
agree vs. a tree
forget vs. for good
idea vs. I do
packet vs. pack it
bacon roll vs. they can roll vs. steak and roll
While children are obviously not set the task of sorting out pairs of
utterances as similar as these, they are required to separate out
word units which occur in less prominent positions and rarely in
isolation. They rapidly accumulate such words, for example verbs
which they will necessarily hear with other constituents:
You spilled your juice
Im cooking your dinner
Will you put your toys away
and prepositions which they will hear not only with other consti-
tuents, but unstressed and often in reduced form:
Your teas on the table
Im going to the shops
Keep away from the re
Research is beginning to produce evidence that children are as
nely tuned to rhythmic patterns as they would need to be to
grapple with these less salient elements of speech. Ingenious ex-
periments have been carried out to see how very young infants
respond to subtle distortions of rhythm. These rely on a preferen-
tial listening technique which checks whether children prefer
to hear speech with normal rhythmic patterns preserved, rather
than speech whose rhythm is disrupted. Childrens preference is
measured by the length of time they focus on loudspeakers produc-
ing normal or disrupted rhythms. In this way, it has been shown
that children as young as 412 months prefer speech with pauses at
clause boundaries to speech with pauses in the middle of clauses
(Jusczyk & Kemler Nelson 1996). By 9 months, they show sensitiv-
ity to units within clauses as well, preferring to listen to speech
where pauses occur at phrase boundaries rather than within
phrases (Jusczyk & Kemler Nelson 1996, Hirsh-Pasek & Golinkoff
1996).
Infants of 712 months have been found to pick out even smaller
units: words. These infants were familiarised with certain words
by hearing them repeated, and were subsequently played passages
which either did or did not contain the familiarised words. They
were found to listen longer to passages which did contain the
familiarised words than passages which did not (Jusczyk & Kemler
Nelson 1996). The same thing happened when the sequence of
presentation was reversed, with words played rst in passages and
then in isolation. Infants listened longer to words which had
occurred in the passages than to those which had not. The indica-
tions are that very young infants can pick out rhythmic units, and
smaller rhythmic units within these, and can recognise further
details within these units.
A complementary nding is that 9-month-old (American) in-
fants listened longer to words which followed the predominant
stress pattern of English and began with a stressed syllable than to
words which began with an unstressed syllable (Cutler 1996).
Six-month-olds did not show this difference. This nding is par-
ticularly signicant. It indicates that by 9 months infants are
sensitive to the stress patterns in their language which they must
notice if as we have suggested they are to use stress as a cue to
word units.
Although the segmentation of words depends crucially on the
child, the process may be facilitated by highlighting the cues to
which the child tunes in. Studies of adultchild interaction in
many (though not all) cultures have found that adults use higher
pitch and more marked stress when talking to children. They have
also found that children attend more to speech with these charac-
teristics (see Snow 1995 for a review of the evidence). Attendance
to speech is clearly a prerequisite to segmenting it. But heightened
pitch and stress can aid the segmentation process in more subtle
ways. If children use stress as a signpost to word units, magnifying
stress differences will magnify the signpost. So, while adults do not
actually separate out the words they do not, for example, pause
between words they seem to extend the cue which children use
to separate out words themselves.
Connecting speech and scenes

So far, we have looked at the childs ltering of scenes and of
speech as if these were independent of each other. Up to a point,
they must be. No amount of information about the structure of the
scene could reveal the structure of the speech relating to that
scene, and vice versa. This means that the child must have the
capacity to interpret each in its own right. It does not mean,
though, that the interpretation of each is wholly independent of
the other. The childs target is, after all, to discover the connection
between the two, rather than to process speech or scenes as ends
in themselves. To make these connections, children must have
access to possible units of speech and possible units of meaning. But
their search for connections may aid their discovery of what the
actual units are, through what have commonly come to be known
as bootstrapping processes. These are processes whereby one
type of information acts as a bootstrap which helps children get
hold of another type of information.
Childrens identication of a particular focus in a scene may
lead them to look for a phonological form which accompanies that
focus. Suppose, for example, that they have experienced scenes
where the focus is a small very cold transparent block which can
go in the mouth. This prompts them to search for the common
phonological form in utterances such as
Ooh, feel the ice!

Do you want some ice?
The ice is melting
I dropped the ice
Here, we may view the child as using semantic bootstrapping to
reach for phonology: the picking out of the object helps them to
notice the form ice. Conversely, childrens encounter with an
unfamiliar phonological form will alert them to nd out what is
focused when they hear that form. If they hear the new form in the
presence of an object for which they do not yet have a label, they
will assume it refers to that object. If, on the other hand, they
already have a label for that object, they will assume that the new
form refers to something different. This will push them to look for a
different aspect of or angle on the object as the meaning behind the
form (Clark 1993, 1997). Meeting the forms mug or handle or china
in the presence of something they call a cup will trigger a search for
a different and relevant perspective on cups, and lead them sooner
or later to focus on its shape, its parts, what it is made of. When
phonology drives a search for meaning, the child is exploiting
phonological bootstraps.
These bootstrapping processes can only operate if children have
a powerful capacity for holding in mind scenes and speech simul-
taneously. It is only by holding the two simultaneously that shared
or distinct speechscene combinations can be noted. It seems that
children are highly sensitised not only to certain aspects of scenes
and certain aspects of speech, but also to connections between
these.
The further we search for cues in the input which guide childrens
discovery of words, the more we reveal how rich and subtle those
cues are, and how nely tuned children must be to pick them up.
Children end up as the key players, demonstrating their extraordi-
nary sensitivities. These sensitivities are wide-ranging yet highly
specic, targeting: the rhythmic patterns of speech; components of
scenes such as entities, events and states; the speakers focus
within the scene; and, crucially, co-occurrences between these:
between a focus on a component of a scene, and a unit within a
rhythmic pattern.
When adults talk to children, they may heighten the rhythmic
contrasts in their speech and so highlight the word units. They
may heighten their use of gesture, eye gaze and facial expression
and time their utterances in ways which highlight the aspects of
scenes focused by their words. By exaggerating the cues to which
children are sensitive, they play into those sensitivities. But chil-
dren cannot attain normal control of words without those sensi-
tivities. This will become increasingly apparent as we explore
childrens negotiation of different kinds of words.
3 Blocks on the road to words
What of children who are not developing language normally? For

such children, words are in some way a problem. They do not
come to understand and produce words as rapidly or reliably as
their normally developing peers.
Their difculties may show up in tests of word comprehension
such as the British Picture Vocabulary Scales (Dunn, Dunn, Whet-
ton & Pintilie 1982) or Peabody Picture Vocabulary Test (Dunn &
Dunn 1981). These tests present the child with a set of pictures
from which she must select the one which matches the word that
she hears. A child of 5 who makes the number of correct choices
typically made by a 3-year-old demonstrates a problem with
understanding words. Similarly, a child may perform at a level
lower than expected on a test of word production, such as the
Word-nding Vocabulary Scale (Renfrew 1980) or Test of Word
Finding (German 1986). Some children may be up to the mark on
tests of word comprehension and production, but fall down on an
assessment of the way they say words, for example on the Edin-
burgh Articulation Test (Anthony, Bogle, Ingram & McIsaac
1971). They may also score below par on a test such as Wepmans
Auditory Discrimination Test (Wepman & Reynolds 1987), in
which they have to tell similar-sounding words apart. Shortfalls in
any of these formal tests suggest the child has some sort of dif-
culty with words, and may also suggest what sort of difculty. (For
further examples and details of such clinical assessments, see Lees
& Urwin 1997, Stackhouse & Wells 1997.)
Difculties with words are also revealed in experiments where
children are taught new words and subsequently tested on their
33
comprehension or production of the words. Such experiments
explore childrens ability to form a representation of a word after
limited exposure to it, an ability which has come to be known as
fast mapping. In one such experiment (Rice, Buhr & Nemeth
1990), children were shown a video which incorporated unfamil-
iar words within a narrative, and then tested for their understand-
ing of the new words. For example, the children heard one story
which included the words viola and trudge in utterances such
as He takes the viola and trudges down the road . . . Billy keeps
trudging down the road . . . After they had watched the video, the
new words were presented to them in arrays of four pictures, and
they were asked to point to the one named. One of the groups of
children tested had been identied as language-delayed. This
group came out with lower scores on the comprehension test than
a group of younger children who were at a similar level of utter-
ance production (being matched on the average length of their
utterances) but were developing language normally. This lan-
guage-matched group in turn achieved lower scores than a group
of children older than them but matched in age to the language-
delayed group. Clearly, the language-delayed children had greater
difculty in acquiring the new words than their younger lan-
guage-matched and their age-matched counterparts.
Reduced speed and scale of word learning in tests or experi-
ments provide some indication of problems with words. There may
be other more overt evidence. As they talk, children may show
signs of struggling for words. They may stall, or grope for a word,
or make repeated attempts at it (see German 1987, German &
Simon 1991 on word-nding characteristics). They may resort to
a roundabout form of words, a circumlocution, to express their
meaning. Michael, aged 7, exhibits all these signs of struggling for
words in his response to a picture of handcuffs:
key . . . oh what do you call them . . . oh yeah . . . you put . . . you put
. . . with your . . . with your . . . oh . . . with your . . . when you . . .
when someones stole something . . . and . . . what do you call them
. . . necklace? . . . no . . . I just dont know the word
(Constable, Stackhouse & Wells 1997)
In attempting to produce words, children may make actual errors.

Blocks on the road to words 35
They may substitute a different word, or produce the appropriate
word but mispronounce it, or produce an unrecognisable form.
Eamonn, Joseph and Ruth provide examples:
Substitution for target
Eamonn: I got this motor bike, er, motor bike, motor bike.
SC: A motor bike?
Eamonn: Motor bike, er, boat.
Mispronunciation of target
# # #
Joseph: [ a ws az]
= I got a rabbit but it died
# # #
[ s w ws ]
= Because it got something wrong with it
# #
[ wn u zn]
= Went up to heaven
Unrecognisable realisation of target
#
Ruth: There is [ bz] bar.
SC: Huh?
#
Ruth: [ mz] bar there.
SC: Bar?
Ruth: Yeah me tal bar.
SC: Oh, metal bar.
All of these children are clearly attempting to produce words and

know something about those words, but are having a problem
somewhere along the way. What is stopping them from hitting the
target?
Where the blocks may lie

Our consideration of what goes into word learning provides a
starting-point for thinking about what might go wrong with it.
We are talking about a child who can hear and produce sounds
crisply. The problem with words must then lie somewhere beyond
hearing and articulation, in the processes which connect these
to meaning. We start with a preliminary analysis of what these
processes are. We will rene this analysis as we venture into
childrens processing of different sorts of words.
Input processing
Having heard the input, the child may have difculties:
( picking out those features of the speech signal which are
crucial in segmenting the stream of speech into word
units
( storing the sound patterns of words which have been
identied
( picking out those aspects of scenes which are crucial to
the meanings of words
( mapping sound patterns onto meanings
Problems could arise in just one of these components of process-
ing or in more than one. Where one is necessarily dependent on
another, we would expect problems in one to affect the other. For
example, a problem in processing the stream of speech must surely
affect the establishment of phonological representations. It is hard
to see how you could identify the sound patterns of words if you
could not process the sound patterns of speech in the rst place.
We might also expect problems in phonological processing to
affect the identication of word meaning, since this involves the
mapping of phonology onto aspects of scenes. As we investigate
childrens difculties, we will be considering possible interdepen-
dencies between different components of processing, and possible
effects of one on another.
Figure 3.1 represents components of input processing which
may be impaired, and expected or possible relationships between
these.
Output processing
It may be that input processing is intact, and problems arise
only in output. The child may have difculties in:
( picking out relevant aspects of scenes
( selecting aspects of scenes corresponding to word seman-
tics
( accessing word phonology corresponding to word seman-
tics
( planning articulation of word phonology
Discrimination of relevant Discrimination of relevant

features of speech features of scenes
Identification of word Identification of word

phonology semantics
Storage of word Storage of word

phonology semantics
Figure 3.1 Components of input processing
As with input, problems could conceivably arise in one or more

of these components, and problems in certain components would
be expected to affect others. It is hard to see how you could nd the
phonology for what you want to say if you cannot work out what
you want to say. If you did nd word phonology without rst
selecting word meaning, it could not be the phonology for a
meaning you intended. We would therefore expect a problem in
selecting word semantics to affect accessing of phonology. But
accessing of phonology could be an obstacle even if selection of
word meaning is not. You might know what you want to say, but
hit problems in nding the phonology to say it. If you did have
difculties in accessing word phonology, we would expect these to
disrupt your articulatory planning, since you can only articulate
what you have accessed. The reverse effects such as the effects of
difculties with word phonology on word semantics are less
predictable.
Figure 3.2 represents components of output processing which
may be impaired, and expected or possible relationships between
these.
Discrimination of relevant
features of scenes
Selection of word Accessing of word

semantics phonology
Planning of articulation
Articulatory execution
Figure 3.2 Components of output processing
Repercussions for lexical development

The processes we have been considering are on-line: they are
processes which take place as the child receives an utterance or
produces one. Through these on-line processes the child discovers
words, stores them, and has them available for her own use. They
are therefore necessary stepping stones in the childs acquisition of
new words and development of her mental lexicon. As the child
acquires words, the relationship between on-line processing and
the mental lexicon reverses. Instead of processing utterances to get
to words, the child can draw on her previously established knowl-
edge of words to process utterances. She can pick out familiar
phonological chunks within the utterances she hears, and retrieve
the meaning she has already attached to these chunks. She can
use her knowledge of words to package her thoughts into word-
size meanings, and access the phonology corresponding to these.
Where the stepping stones to words are lacking, there will be
repercussions for lexical development. The child who has yet to
learn words is in a different position from someone who has
already established a mental lexicon when she acquires a decit in
on-line processing, as a result of, say, a stroke. Such a person may
have stored information about words despite severe difculties
with recognising words in input or accessing words for output. Not
so the word-learning child. (See Chiat, Law & Marshall 1997 for
further discussion of differences in the effects of developmental and
acquired processing decits.) As we saw above, if a child has
difculty extracting information from the stream of speech, this
will inevitably affect her establishment of word phonology, and
may also affect her establishment of word meaning. The represen-
tation she stores in her mental lexicon will be correspondingly
limited. We may expect the child to learn new words slowly, and
even when words make it into her lexicon, we may expect them to
be unreliably specied. The limitations in her lexical development
will rebound on her on-line processing. She will have less ready-
made information about words to help her pick out words from
incoming utterances and select words for her own utterances. So
we may expect to see a spiral from inadequate on-line processing
giving rise to inadequate lexical representations giving rise to
inadequate on-line processing.
Depending where a childs problems arise, the repercussions
between on-line processing and representations will vary, with
varying effects on the childs lexical development. This will become
apparent when we look at children whose problems arise at very
different points in processing.
4 Exploring the blockage
Thinking about where problems could arise begs the question of

how we nd out where they do arise. Not only are there many
possible stages at which breakdown may occur. There are also
many possible effects of breakdown at one stage on another stage.
It is no mean task to identify stages at which breakdown occurs,
and disentangle the source of the breakdown from its effects.
Consider the following difculties exhibited by Eamonn:
1. E: Daddy [fin] the cat (describing picture of a man feeding a
horse)
SC: Whos hungry?
E: The [k] the horsie.
2. E: [ju] pick him up, he get [fi ns] ( = if you pick him up, he
gets very nervous, talking about a robot).
The rst appears to be a semantic confusion, with Eamonn select-

ing the word cat which is related in meaning to his target horse.
Here we might infer a problem in semantic processing. The second
appears to be a phonological confusion, with Eamonn producing
the forms [fi] and [ns] related to the sound of his targets very
and nervous. Here we might infer a problem in phonological
processing.
However, neither confusion is transparent: both are open to
alternative accounts. In the rst case, it may be that Eamonn could
not access the phonology for horse, picked a word close to his
intended meaning, and ended up with a different animal name. In
the second case, it is conceivable that Eamonn was hazy about the
semantics of the words very and nervous and this made his
40
Exploring the blockage 41
access to phonology precarious. To get below the surface of these
sorts of problems with words, we will need tools for digging deeper.
The purpose of those tools will be to tap particular stages of
processing, separating these out from other stages.
The starting-point for probing a childs difculties is a hunch.
This will not be just any hunch, but a well-informed hunch based
on initial observation of the child. The richest source of informa-
tion at this point is the childs spontaneous output. This can be
tapped by obtaining a sample of language the child produces in a
naturalistic situation, where he is at ease and talking as he nor-
mally does. If any tests of word comprehension or production have
been carried out, these may also generate a rst hunch.
The key to that hunch is noticing what is or is not all right
about the childs words. What sort of words are produced appro-
priately, and what sort of words are omitted or produced inappro-
priately? If words are produced inappropriately, which aspects of
those words are nevertheless appropriate, and which aspects are
responsible for their inappropriateness? The aim at this stage is to
identify a possible pattern. First observations may suggest that the
childs words are semantically ne, but phonologically distorted.
Or the initial impression may be the reverse, that the childs words
are phonologically on track but that they are not quite right
semantically, or even more strikingly, that they are completely
inappropriate for the context. The initial hunch about the area of
difculty will drive further investigation. That investigation will
check out whether the hunch is right: whether the child does have
a problem with the aspect of words which is suspected. It will go on
to explore where that problem arises in input and output process-
ing, and how it might relate to other aspects of the childs process-
ing.
At this point, we need more rened tools of investigation. The
tools available to us are a range of techniques for tapping particu-
lar aspects of processing. These techniques have endless variations
and continue to develop as new questions are posed about chil-
drens language processing and novel methods are devised to
address these. However, they have certain things in common.
They all present the child with some deliberately selected input,
such as pictures, or actions, or combinations of speech sounds, or
words, or a mix of these. They all elicit a response to that input
which may involve producing speech or words; making a judge-
ment of right/wrong or same/different; pointing to a picture; or
carrying out an action. What is presented to the child and what is
required in response crucially determine what aspect of word
processing is being tapped.
Such psycholinguistic investigation starts with and is led by the
child. It draws on an understanding of what is involved in lan-
guage processing in order to focus on and delve into a particular
childs comprehension and production of language. The child is
not a passive recipient, subjected to a gamut of predetermined
psycholinguistic tests. The aim is rather to pursue questions which
arise from observations of the child, where it is appropriate to
pursue those questions, using activities which are responsive to
the general concerns, needs and interests of the child. With these
assumptions as a starting-point, we turn to some examples of the
tasks which may tap different stages of processing.
From hearing speech to phonology: phonological

discrimination tasks
To nd out whether children are able to pick out relevant
aspects of speech, we need a task which taps just their discrimina-
tion of speech, and does not involve knowledge of words. This
means asking children to make judgements about speech. Eliciting
such judgements might run along the following lines:
Im going to say two words, and I want you to tell me whether they
sound the same or different. Here they are . . . cap tap. Do they sound
the same or different? . . . Now here are two more. See if they sound the
same or different. tap tap . . .
The only way the listener can make this judgement is to notice the
sound difference between tap and cap. You may notice that tap and
cap are different words with different meanings, while tap and tap
are the same word if your language processing is intact you
probably will. This may make the judgement easier. But the point
is that you do not need to notice this. You could make the judge-
ment just on the basis of the sound of each pair, and this is
something you must anyway do in order to register whether the
words are different or the same. It is also what you must do if you
are presented with pairs of non-words rather than words, as in
the following:
Im going to say some funny words, and I want you to tell me if they
sound the same or different. tep kep . . .
Here, the input consists of phonological forms which you have not
met or stored, and which have no meaning. By using non-words in
a discrimination task, we can ensure that the judgement is made
on the basis of sound only.
Such tasks can be used to check whether a child is able to
recognise particular phonological distinctions. The distinctions
one chooses to investigate will depend on the initial hunch which
motivates the investigation. This may be an observation that the
child is not making certain distinctions in his own speech. Suppose
that words beginning with /t/ and /k/ both sound as if they begin
with /t/ in the childs speech, and words beginning with /f/ and
/v/ both sound as if they begin with /b/. In this childs speech,
members of word pairs such as tap and cap, tea and key, t and bit,
vase and bars would sound the same. In this case, the point would
be to check whether the child perceives the differences he is not
making. A discrimination task will check this. Can he tell that
tap/cap, tea/key, t/bit, vase/bars sound different?
In this example, certain initial consonants were at stake. A
discrimination task can be used to check other aspects of speech
suspected to present difculty, and quite subtle ones. The child
may produce particular consonants differently depending where
the consonant occurs within the word. For example, he may
produce /k/ and /'/ correctly at the ends of words, but not at the
beginnings of words. Or he may leave out consonants at the ends
of words, but not at the beginnings. If it looks as if the position of a
consonant affects the way it is produced, it will be important to
control the position of the consonant in the words used in a
discrimination task in order to see if this affects discrimination of
the consonant as well. We might want to present the child with
judgements about consonants in nal position as well as initial
position:
Discrimination Discrimination Discrimination Discrimination

of /k//t/ in of /k//t/ in of consonant in of consonant in
word-initial word-nal word-initial word-nal
position position position position
keytea sitsick eatbeat beebeat

teatea sitsit eateat beatbeat
Children may make errors in sequencing consonants, in which

case the task could present pairs such as lost/lots which differ in
this respect.
More unusually, a child may show problems with producing
vowels, in which case a discrimination task would present pairs
differing in the vowel they contain:
Vowel discrimination
bitbet
bitbit
carcore
carcar
Problems may affect whole syllables rather than just sound seg-
ments. In this case a discrimination task could present pairs differ-
ing by a whole syllable:
Discrimination of syllable structure
butbutton
winwindow
All these examples illustrate how discrimination tasks may be
used to nd out whether problems in producing speech are due to
problems in picking out the relevant features of speech. This
means checking out discrimination of not just any speech distinc-
tions, but the particular distinctions which are affected in the
childs output.
On the other hand, initial observations of a childs speech may
point to difculties with word phonology but not with any specic
phonological distinctions. We would still want to know if the
childs errors in production may be related to his discrimination of
speech. The most revealing way of checking this would be to feed
back the childs own incorrect forms paired up with the correct
forms to see if the child can discriminate these.
What can such discrimination tasks tell us? If the child is able to
make the judgement, a great deal. It is only possible to differentiate
same pairs from different pairs if you can discriminate the
feature which distinguishes these. So, if the child can judge captap
and keptep to be different, and taptap and kepkep to be the same,
he must have the /t//k/ distinction at least in initial position
within words.
But suppose the child does not make correct judgements, re-
sponding randomly to all pairs, or always giving the same re-
sponse whether the pairs match or not. Can we conclude that the
child is unable to discriminate between the relevant sounds? Not
necessarily. Any discrimination task involves more than just pho-
nological discrimination. For example, it entails an understanding
of the task itself, which means understanding what is meant by
same and different. It entails a willingness to listen to the input,
and to attempt a judgement. These demands are taxing for
children under about 4. So, if the child appears to fail at the task,
this could be due to these other requirements. He may lack the
same/different concepts. Or he may simply hate the task and refuse
to co-operate. In either case, we have no further information about
the childs phonological discrimination.
This brings out a general point about the use of tasks designed
to tap a specic aspect of language processing. If someone succeeds
on these tasks, we can be sure that they can process the relevant
distinction. They could not respond reliably unless they appreci-
ated the specic aspect of language they are being required to
process. If, on the other hand, someone does not respond reliably,
we cannot draw rm conclusions. All tasks involve more than just
the language processing under investigation, and other task vari-
ables may be responsible for the persons difculty with the task.
We must therefore be cautious about interpreting failure on a
task such as discrimination. We need to explore further what may
lie behind the childs responses. Here, the key will be any pattern in
the childs responses: any evidence that the child can do the task in
some cases but not in others. Suppose, for example, that the child
can make reliable judgements on /p//k/ pairs but not on /t//k/
pairs:
/p//k/: judged correctly /t//k/: judged incorrectly
patcat tapcap
peakey teakey
peskes tepkep
This would be proof enough that the child understands the task,
including the concepts same/different, and that he can discrimi-
nate certain speech differences. We could then infer that there is a
specic problem with discriminating /t//k/. In the absence of
such evidence, it may be important to establish that the child can
make same/different judgements using input of a different kind.
For example, if we can establish that a child can make a same/
different judgement about visual stimuli such as shapes and sizes,
we at least know that the concepts same/different are under-
stood.
From phonological discrimination to word

phonology: lexical decision tasks
This detailed introduction to discrimination tasks sets the scene
for tasks tapping other stages of processing.
Once relevant features of speech are discriminated, the child
must pick out and store the chunks of these features which consti-
tute words. How can we check whether the child is storing words
appropriately? One way to go about this is to see if the child can
decide whether a string of sounds is a word or not, using what is
known as a lexical decision task. Here, the child is presented
with phonological forms, some of which are words, and some not.
The child must indicate whether these forms sound right or not.
One way of eliciting such a judgement is with the help of a puppet:
Ive got a puppet. Looks like a clown, doesnt he . . . This clown is
having a bit of a problem. Sometimes he says words and they sound
right. But sometimes they sound silly. I wonder if you can help him.
Can you tell him when he says something right and when he says
something funny? . . .
Alternatively, the childs interest might be engaged by inviting
him to judge the speaker:
Im going to say some words now. Sometimes Ill get them right, but
sometimes Ill get them wrong. When I get them right you can give me
a tick, and when I get them wrong you can give me a cross. Then at the
end we can see how many I got right . . .
The materials used in the task would be determined by the

observations of the child which motivated the task. Suppose the
child is producing incorrect forms of words, and we want to check
whether he has stored the forms correctly. The ideal check is to
feed back the childs own errors to see if he can recognise them as
such, and distinguish them from the appropriate forms. For
example, a child who was producing /k/ and /'/ at the beginning
of words as [t] and [d] might be presented with the correct forms
and the incorrect forms:
Words Non-words
card tard
carrot tarrot
carpet tarpet
girl dirl
garden darden
Another child may produce variable and more distorted forms for
a word. With such a child, we might feed back all these forms
along with the target word in a lexical decision task.
A rather different case for using a lexical decision task would be
with a child who is producing correct forms of words, but whose
range of words is unusually limited. We may want to nd out
whether the child recognises a wider range of word forms than he
is using. This could be done by presenting the child with words and
non-words, to see how much the child knows about the phonology
of words.
This is precisely what a lexical decision task is getting at. It need
not involve the meaning of words at all. If you do know the
meanings of the words in a lexical decision task, this may help you
to recognise and approve them as words. But the lexical decision
task does not require this. The minimum you have to do is to
discriminate the phonological forms and compare them with pho-
nological forms you have stored. Suppose a child can make lexical
judgements reliably, accepting the words and rejecting the non-
words. We can then be sure that the child has discriminated and
stored the phonological form of the words correctly. If the child
can reject non-words such as tarrot and darden which we hear him
produce, then he must have not only noticed the difference be-
tween /t/ and /k/ and between /d/ and /'/, but also stored the
words carrot and garden with the correct initial consonant.
What if the child cannot make reliable judgements? As we saw
with discrimination tasks, apparent failure is not clear-cut. There
are many reasons why a child may have difculty with this task. A
problem with storage of word forms is just one possible reason. The
child may actually have the correct forms stored, and still make
errors in lexical decision. The task is, after all, very taxing in other
ways. In order to accomplish the judgement, the child must have
grasped the notion of something sounding right. He must hold
onto the form he has heard (e.g. carrot) and must check to see
whether he has this form in his mental lexicon. A tall order. One
would not expect a child under 4 to cope with these demands.
Even at a later age, we would need to consider whether the child
was having difculty with the task demands rather than the words
themselves. Again, a pattern of response would be the key. For
example, suppose we can show that the child is able to reject
non-words except where the difference from a word involves /k/ or
/'/. This suggests that the child has grasped the concept of judge-
ment, and that he has correctly represented most sounds in words.
We then have stronger evidence that there is a problem in the
storage of words with /k/ and /'/ assuming that we already
know the difculty is not due to discrimination.
The lexical decision task may be easier for the child if meaning
is brought into the picture, and the judgement is not a purely
phonological one. Instead of just asking the child to judge whether
a form sounds right, you present him with a picture of the target
word and ask him to judge whether a form is right for the picture:
Is this a carrot?
vs.
Is this a tarrot?
The presence of the picture may provide a more meaningful con-

text for making phonological judgements. Instead of just judging
the sound of the word the child must think if it sounds like the
word for that picture. Yet the task still requires a phonological
judgement. If the child is able to make consistent judgements that
carrot is right for the picture but tarrot is not, we can be sure the
forms are correctly stored.
However, because the task asks the child to make a phonologi-
cal judgement in relation to a picture, any problem in responding
could be due to the picture and its meaning rather than the
phonology. For a start, the child may be distracted by the seman-
tics. He may look at the picture, recognise the vegetable, and
decide that tarrot is clearly aiming for the right word semantically
and so accept it as appropriate, even though it doesnt sound
exactly right. Or the picture may be more than a distraction from
the phonology. It may itself present a problem. Suppose the childs
difculties lie in semantics, or the connection between semantics
and phonology. Such a child may have difculty deciding whether
carrot or tarrot are right for the picture, even if he knows that carrot
is a word while tarrot is not. This knowledge might show up on a
purely phonological judgement task, and be obscured once mean-
ing is introduced.
From phonology to semantics: word-to-picture

matching tasks
Making the leap from phonology to semantics might be seen as
the key stage of word processing. The main function of words, after
all, is to act as a vehicle for conveying a meaning. Not surprisingly,
this is the aspect of word processing thats focused in traditional
approaches to childrens problems with words. It is also the aspect
thats targeted in traditional methods for assessing childrens word
processing: picture-pointing tasks. In such tasks, the child is pres-
ented with a set of pictures and asked to point to one of them. For
example, given pictures of a bear, spoon and dress, he might be
asked to show me the bear. Here, the child must discriminate the
phonology of bear, recognise it, and connect it to the appropriate
animal. What will we nd out about the childs processing from
this traditional sort of assessment?
This will depend crucially on the content of the task. If we
present the child with a random set of pictures to choose from, the
task will not tell us much. On the one hand, it involves all aspects
of input processing, from phonological discrimination to meaning.
On the other hand, that processing need not be very precise. Since
bear sounds nothing like spoon or dress, you could pick out the right
picture even if you registered its phonology only loosely. Similarly,
since spoon and dress belong to quite different semantic catego-
ries, if you knew nothing more about bear than that it was an
animal, you could pick out the right picture.
But suppose the selection of pictures is more calculated. Sup-
pose the pictures include one which is phonologically similar to
the target, such as pear, which acts as a phonological distractor.
The child must then discriminate the phonology quite precisely if
he is to pick the pear rather than the bear picture. Similarly, the
pictures might include a semantic distractor which is semanti-
cally related to the target, for example a picture of a wolf. This ups
the semantic stakes for the child: his semantic information about
bear must be precise enough to rule out other animals. Where
both phonological and semantic distractors are combined with the
target, the child must have phonological and semantic informa-
tion well specied if he is to home in on the correct picture.
So, with careful selection of the pictures, a picture-pointing task
can reveal just what a child knows about the phonology and
semantics of the target words. If the child is consistently able to
pick the right picture, we can be sure he has at least the phonologi-
cal and semantic information about the target word needed to
distinguish it from the phonological and semantic distractors. If he
sometimes picks the phonological distractor, but not the semantic
one, this would suggest some difculty with phonological process-
ing. We could follow this up by looking at his phonological dis-
crimination and representation in more detail. If he sometimes
picks the semantic distractor, but not the phonological one, we
might infer that his semantic representations are not fully speci-
ed, and follow this up by looking at his semantics more closely.
From perception of the world to semantics:

picture-sorting tasks
One means of tapping semantics, as we have seen, is a word-to-
picture matching task with semantic distractors. This necessarily
taps semantics via phonology. If the child has a problem with the
recognition of a words phonology, this could block access to its
semantics even if the semantics would not itself be a problem. How
can we tap the childs semantic categories without involving
phonology?
One way would be to rely entirely on visual materials, and to
ask the child to do something which involves picking out some
aspect of the visual materials but not saying anything about them.
We could, for example, elicit judgements about the relationship
between a set of pictures. The pictures would represent meanings
which were or were not related in particular ways. For example,
we might want to know if the child categorises objects in semanti-
cally relevant ways, that is, in ways which correspond to words.
Does the child group together different items which would be
identied by the same word? We could check this by presenting the
child with sets of pictures to decide which ones went together, or
which one was the odd one out. Each set of pictures would
represent distinct examples of the item along with an item which
was different (even if it looked quite similar):
Sets of pictures could be selected to tap all manner of distinctions,

from distinctions between different members of a general category
such as clothes:
to distinctions between members of different general level catego-

ries such as vehicles and buildings:
They might represent properties such as shape, or colour, or

temperature, or emotion:
This sort of task could even be used to look at the categorisation of

events. The child might be presented with a picture of one event,
then given pictures of two more events to choose the one which
matches the rst event:
Exploring the blockage
55
First published in J. Marshall, S. Chiat & T. Pring, Event Perception Test, in Marshall et al. 1999, reproduced by permission of Speechmark/Winslow.
(see Marshall, Pring & Chiat 1993, Marshall et al. 1999 for further
explanation and examples). Or the task might look at the childs
recognition of what sorts of things are involved in different events.
You might present him with a picture of an event, such as cutting,
and a set of pictures of objects which can or cant participate in the
event, such as paper, cloth, bread, pencil, iron, milk.
An alternative to picture materials is acting out. This is par-
ticularly appropriate for investigating semantic aspects of events,
to see if the child is picking out those aspects of events which are
encoded by words. Because events are dynamic and occur over
time, their features are more easily picked out from a dynamic
representation, such as an acted-out scene, than from a still pic-
ture. Imagine, for example, representing a girl pushing a car. If this
event is acted out, the girls pressure on the car and the cars
movement are directly observable. In the case of a still picture, on
the other hand, the action and movement must be inferred.
Now, suppose we act out an event in front of the child or show
the child a video of the event, and then ask the child to do the same
thing. For example, we might show events involving movement in
different manners, such as rolling/spinning/throwing a stick, or in
different directions, such as pushing/pulling a car, dropping/lifting
a book. We could then give the child a different object from the one
used in the modelled event, and ask him to do the same thing. Will
he act out the event with the same manner or direction even
though the object is different, spinning a ball if the model was
spinning a stick, pushing a box if the model was pushing a car? We
might want to check other aspects of events such as their effects. If
we model events which produce different results, such as opening
and closing a box, or emptying and lling a jug, and give the child
different objects to carry out the same action, such as a bottle or a
drawer to open or close or ll or empty, will he produce the same
result? If he does act out the relevant manner or direction or result,
he must have noticed those aspects of events which differentiate
verbs from each other and so are semantically relevant. But if he
doesnt? Can we infer that the child is not picking out relevant
aspects of events? Not necessarily.
Yet again, if the child does pick out features which distinguish
different verbs, for example manner or direction or effect of move-
ment, we can infer that the child has noticed those features. But if
the child responds differently, this doesnt necessarily mean that
the child is oblivious to those features. It may be that he has based
his response on a different criterion from the one we have focused
on. He may sort pictures according to which things or events look
most similar, or which might occur at the same time or in the same
place. These similarities may override knowledge of other shared
properties which the child nevertheless possesses. The task should
therefore be seen as an exploration of how the child groups repre-
sentations of things or events, rather than a test of semantic
knowledge about those things or events. We would be looking to
nd patterns in the childs responses for example preferences for
certain sorts of features in grouping pictures or actions rather
than scoring them as right or wrong.
As with other tasks, the childs response will depend on his
understanding of the task as well as his ability to process the
distinctions it focuses. We may well use such tasks with a child
whose semantics cannot be tapped through phonology because he
is known to have problems with phonological input. But his
problems with phonological input will make it difcult if not
impossible to explain the task verbally. In this case, we may have
to present the task non-verbally, by giving the child examples of
how pictures might be grouped. We would then be relying on the
child getting the idea of grouping the pictures and attempting to
group them in his own way.
From lexical semantics to lexical phonology:

internal judgement tasks and cued naming
Suppose we know that a childs processing of words in input is
ne. The child has no problem understanding words, so must have
discriminated and recognised their phonology and matched their
phonology to semantics. Yet he has problems producing words.
Where does the difculty arise? Is it in accessing the phonology of
the word for output? Or is the word accessed appropriately, with
difculties arising in its articulatory production?
One of the toughest challenges for the psycholinguist is to tap
the childs accessing of words separately from his saying of words.
This separation is necessary if we are to nd out whether the
phonology he has in his head is different from the phonology he
produces. Imagine a child who, presented with a picture of a cap,
cant name it, or names it as tap. In either case, he may actually
have the form cap in mind. How can we nd out what he has in
mind without asking him to say it?
With some ingenuity, it is possible to probe the sounds in the
childs mind. The relevant tasks are quite challenging for the child:
they require the child to make internal judgements about sound
rather than produce it. By way of illustration, take the following
materials:
(a)
(b)
Look at each set of pictures and decide which two begin with the
same sound. Now decide which two rhyme with each other. Now
consider how you arrived at your decisions. You were presented
with purely visual materials, no words. You had to recognise the
objects in the pictures to select the semantic target. You had to
match that semantic target to a phonological representation
/kp/, /tp/, /k/, /pl'/, /kt/, /kp/, /ht/, /d'/. Then you
had to hold the phonological representations in your head and
compare them to see which shared the relevant phonological
feature the initial sound, or the rhyme. So, in order to make the
judgement, you had to access the word phonology, but you did not
have to speak it.
A phonological judgment task like this may be carried out in a
variety of ways. One of the simplest would be to present just two
pictures, and ask for a yes/no judgement. Do the following pairs
rhyme?
(a)
(b)
Do the following pairs begin with the same sound?

(a)
(b)
Here, the child must hold just two words in mind. Another way of
reducing the load on the child is to present one picture at a time,
and ask the child to make a judgement about just that picture. You
might give the child a heap of pictures illustrating words begin-
ning with /k/ or /t/, and ask the child to post each picture into
either a [k]-box or a [t]-box. In this case, the child has only to hold
the phonology for one picture, and compare it with the two target
sounds. More demanding is a task which presents more pictures,
and includes distractors as in the cap, tap, car, plug set. This
included a phonological distractor picture whose name was
phonologically similar to one of the target pair; and a semantic
distractor picture which was semantically similar to one of the
target pair.
Whatever the particular mode of presentation, this sort of task
involves an internal judgement about word phonology. If the child
is able to make the judgement reliably, we can be sure that he is
accessing the particular aspect of word phonology to be judged
the initial sound or the rhyme of the words. But if the child
responds randomly? As usual, we must think again.
This task is particularly taxing. It requires the accessing of word
phonology, but it requires a good deal more than this. Suppose we
have asked the child to say whether a cap picture rhymes with a
tap picture. First of all, the child must interpret the pictures as we
intend if the child sees the cap as a hat he is already out of the
running. Once he has the appropriate semantic targets, he must
access the correct forms cap and tap. But that is not all. He must
then hold onto cap and tap, and pick out the rhyme part of each in
order to compare them. So even if the child has accessed cap and
tap he may make a wrong judgement for other reasons. He may be
unable to hold onto the two forms to compare them. Or he may not
understand or recognise rhyme, and so be unable to see the
similarity between the words.
To some extent, we can check out these components of the task
independently, and we should do so before using it. After all, there
is no point doing an internal rhyme judgement if the child does not
understand rhyme. But if we have checked that the child recog-
nises rhyme when we say pairs of words such as cap/tap, cap/car,
we know he has the concept of rhyme. We can then rule this out
as the source of difculty with an internal judgement task. How-
ever, other difculties with the task may still account for the childs
responses. It may be that he simply doesnt understand what hes
supposed to be trying to do. After all, making a judgement about
sounds corresponding to pictures is a pretty obscure and demand-
ing activity.
In some cases, there may be no need to resort to such devious
methods in order to probe what is in the childs mind. We may be
able to elicit evidence more directly. Returning to the scenario of
the child who is presented with a picture of a cap and cannot name
it, or struggles to do so, we might see whether we can prompt the
child to produce the word with cued naming. Suppose we give
the child a phonological cue. We may try just the initial conson-
ant of the target: [k-]?. If this makes no difference, we may
extend the cue to the initial consonant + vowel: [k-]?. Where
the target is more than one syllable, we may offer more. With a
word such as candle, the cue may be gradually increased from
initial consonant [k] to initial consonant + vowel [k] to initial
syllable [kn]. The question will be whether, and at what point,
the child succeeds in producing the word. Success implies that the
child does have access to the word form and can produce it. But
gaining that access is difcult. The amount of cuing the child
requires will indicate just how difcult.
If cuing makes no difference to the childs output, this may be
because the child has nothing to cue: he has no further informa-
tion about the word. But it may also be because accessing informa-
tion about the word is not the problem anyway. The barrier to
producing the word lies elsewhere.
From phonological input to phonological output:

repetition tasks
Internal judgement tasks separate out the accessing of word
phonology from the articulation of the word. They provide a
roundabout way of getting at a childs phonological representa-
tion of a word without the child having to articulate it. How can
we do the reverse? How can we tap childrens articulation of words
without them having to access the phonology of the word rst?
Repetition tasks t this bill. Again, puppets can help to intro-
duce the task:
So, weve each got a puppet. My puppet is going to say a word and I
want to see if your puppet can say the same thing. Lets see if your
puppet can say dog.
Here, the child is given the word form, so he does not have to
access it. All he has to do is discriminate the phonology of dog, hold
it, and plan its articulation. If the word is familiar, he will inevi-
tably recognise the phonology and access the semantics. Assum-
ing he recognises the form dog and has it connected to the category
canine creature this information will surface when he is asked to
say dog. But it is not required. If the childs phonological processing
is intact in both input and output, it will be possible to repeat dog
without recognising its form or meaning. The test case for this is
repetition of non-words:
Now my puppet is going to say some funny words. See if your puppet
can say them. deg . . .
In repeating a form such as deg, which is not a word, the child will
have to rely on purely phonological processing. (See Gathercole,
Willis, Baddeley & Emslie 1994 for their Childrens Test of Nonword
Repetition which includes non-words of 25 syllables.)
So, if a child has shown difculties in producing words, repeti-
tion allows us to check whether this difculty is due to problems in
articulatory planning. We can ask the child to repeat words which
he has produced incorrectly himself. Or we can select words or
create non-words to check out articulation of particular phono-
logical targets. For example, if we suspect the child has difculty
with planning articulation of /s/ before a stressed vowel but not
after a stressed vowel, we might ask the child to repeat words and
non-words with /s/ in both positions:
/s/ before stressed vowel /s/ after stressed vowel
sea horse
sock face
seven castle
sandal message
decide listen
cassette basin
Then we could do the same thing with non-words:
/s/ before stressed vowel /s/ after stressed vowel

# #
[ s] sah [ ls] lorse
# #
[ sb] sab [ ves] vace
# #
[ sfl] sefal [ tsm] tahsem
# #
[ smpn] sampen [ nst] nesset
# #
[' sat] gissite [ rsl] rissel
# #
[t sp] tessepp [ 'esl] gayssil
The implications of this task? Where the child repeats the target
appropriately, we can be sure that he has no problem in its
articulatory planning and execution. Where his repetition differs
from the target we need to think further. First, we need to establish
that the child understands the task. With a repetition task, this
should be obvious. If the child is attempting to repeat the words, he
clearly does. In this case, we might need to go back to input
processing and consider whether this is responsible for difculties
in repetition. If the child is falling down in discrimination of
sounds, this will inevitably affect repetition. But if we know that
the childs discrimination is adequate, we have to focus on his
output and dig deeper. What well be digging for is patterns in the
childs repetitions contexts where the child can repeat correctly
versus contexts where he cant.
Suppose the child given the above words with /s/ behaves as
expected, repeating /s/ correctly only in words and non-words
where it follows a stressed vowel. This implies that the child does
have difculty with articulating [s], but only when it occurs before
a stressed vowel.
Or suppose that the child repeats non-words correctly, but not
real words. His correct production of the non-words shows that he
can discriminate and produce the sounds in these words. So the
problem with the real words must be something to do with their
being stored. This suggests that the child has faulty representa-
tions of the words, and is using these faulty representations when
he repeats them. The problem does not arise with non-words
because they are not stored.
So, with an astute search for factors which inuence the childs
repetition, guiding an astute selection of targets to repeat, we can
make inroads into the source of a childs difculties in producing
word phonology.
Reviewing psycholinguistic tasks

Setting out from questions about the difculties children may
have in processing words has opened up many possible paths of
investigation. We have seen how tasks or activities may present
particular aspects of words to which children are asked to respond
in particular ways. They may present word meanings (represented
as pictures or acted out) or word forms. The child may be asked to
make a same/different or right/wrong judgement about the stimu-
li; or to group those which are similar or pick out one which is
different; or to copy one. The nature of the stimuli we present and
the response we elicit whether we present pictures to judge or
phonological forms to repeat or phonological forms to judge will
determine which aspect of input or output processing we are
tapping. The selection of stimuli will determine how nely we tap
the childs processing. If the selection of stimuli to be judged or
grouped includes distractors which are very distinct from the
targets, the skills required by the task will be limited. If, on the
other hand, the stimuli include distractors which are similar to the
targets in some way, the task will require more rened skills. In
working with a child, we will be designing tasks to suit the level
and detail of processing we are aiming to tap. This will depend on
our previous observations of the child, and the psycholinguistic
hunches we derive from our observations.
These psycholinguistic tasks should be regarded as a means of
exploring, rather than testing, the childs processing. With each
task, we have seen that the childs responses will be open to
different interpretations, and will not necessarily lead to denitive
conclusions. In general, if children succeed in a task, we can infer
that they are able to carry out the aspect of processing that task
taps. This conclusion will only be justied, though, if we have
given the child enough examples to be sure that the responses are
reliably correct, and not just correct by chance. In informal explo-
ration with a child, we may not have enough examples to be sure.
In this case, any conclusions must be tentative, and must continue
to be checked out, perhaps in different ways.
If children have difculties with a task, we cannot draw con-
clusions directly. We have to think about what is involved in the
task, and might be responsible for the childs difculties with it.
This might lead us along new paths of inquiry. We might check
out whether the child can carry out the same type of task with a
different type of materials, or with less demanding materials. Or we
might generate a new hunch about the childs difculties and a
fresh line of investigation.
In any case, psycholinguistic tasks should not be viewed in
isolation. Comparison between tasks is crucial. Where we nd
better performance on one task relative to another, we can make
inferences about what is a problem and what is not. Where
different tasks give rise to the same inferences, we can be more
condent that we are on the right tracks.
The tasks presented here are by no means exhaustive. They
simply illustrate psycholinguistic methodology using unsophisti-
cated and readily available materials and activities. There is no
end to the properties of words which may be investigated, and the
methods for investigating them. Different insights into the proper-
ties of words, semantic, phonological and syntactic, open our eyes
to different aspects of words to be investigated. Different technolo-
gies open up all sorts of possibilities for presenting materials and
responding to them. Videos and computers, for example, can be
harnessed for interesting presentation of visual or auditory input.
They also offer different modes of response, such as pressing a
button or moving a curser. The psycholinguistic logic will be the
same: we will still be starting from the processes we are aiming to
tap, devising a task which calls on those processes, and analysing
the childs pattern of response to the task. The possibilities for
pursuing this logic are myriad once the psycholinguistic imagin-
ation is stirred.
All the way along, the psycholinguistic journey will be directed
by the child. The goal is not to carry out every psycholinguistic
task one can think of, or to carry out a task for the sake of it. The
goal is to explore, with the child, what he can and cant do with
words, and what he might yet be able to do.
5 Dant always day dings:
problems with phonology
Listening to some children for the rst time is a bit like trying to see
when youve just moved from bright sunlight into an unlit room.
When they talk, you hear a stream of speech in which its difcult
to nd any familiar, meaningful forms. But gradually, with further
exposure, the ow begins to take shape. Words begin to step out,
and as you differentiate them, you nd you can recognise and
understand what the child is saying.
Stephen is a chatty 5-year-old. Presented with a bunch of
miniature toys, his play is inventive and he keeps up a running
commentary with utterances such as
# # # # #
[a d wn z iz b s n da' s lajn]
I got one of these but its not a tiger its a lion
# # #
[ w iz dld dn]
What are these called again?
# # #
[ d dd desz n]
He cant get the cases on
With a bit of exposure, the listener has no difculty recognising

that
[d] = got
#
[ da'] = tiger
[dld] = called
#
[ dn] = again
[d] = cant
[dd] = get
#
[ desz] = cases
66
Problems with phonology 67
Joseph, like Stephen, has normal 5-year-old conversations as he
plays with miniatures or beats you hands down in a computer
game. His utterances are even more opaque:
# # #
[ a ws az]
I got a rabbit but it died
# # #
[n a lr n rs]
Can I colour it in rst?
Yet even his utterances gradually take shape for the listener,
revealing that
[] = got
#
[ ws] = rabbit
[az] = died
[n] = can
[rs] = rst
How is it that you come to pick out shapes which you could not
pick out before? Somehow, you have adjusted to what is blurred in
the childs speech, and you can get through that blur to recognise
words. This is only possible because the childs blurring of sounds
follows certain patterns. The child is not producing words in
random ways. What we nd is that there is a systematic relation-
ship between the word forms produced by the child and the form of
those words in adult English. Once the listener unconsciously
recognises that relationship, she can adjust to the childs system
and can, in most cases, translate each of the childs word forms
into corresponding adult word forms. Where there is more than
one possible correspondence between child and adult forms, con-
text will usually eliminate all but one of the alternative adult
forms. There will be instances where a word remains unidenti-
able, leaving the listener in the dark. But in general, initially
unintelligible children like Stephen and Joseph become unex-
pectedly coherent.
To illustrate the point, consider rst the small sample of speech
from Stephen. He appears to have difculties with certain sounds
of adult English which he pronounces inappropriately at least
some of the time. /k/ and /'/ are often realised as [d]:
#
cases [ desz]
#
again [ dn]
Once we have been exposed to a number of words in his speech

which show this connection to adult speech, we adjust to the
possibility that Stephens production of [d] may be targeting /d/,
/k/, /'/ and also /t/, and we select the one which yields a word
appropriate to the context. So, when we hear [d] in the above
sample, we are open to the possibilities dot, tot, cot, got, and we go
for the only one of these which is an English word tting the
context, in this case, got.
The listeners adjustment to the childs speech will depend
partly on how closely it corresponds to adult forms. The more
distinctions a child makes in her speech, and the closer those
distinctions are to adult speech, the easier it will be to adjust.
Stephen, for example, is easier to follow than Joseph. This is
because there are many more speech sounds which Joseph fails to
produce. In the sample above, the only consonants which he does
produce are [s], [z], [m], [n], [w], [r], [l], []. Where other conson-
ants are required, he either omits them altogether or he pro-
nounces them as [s] or [z]:
died [az]
#
rabbit [ ws]
#
colour [ lr]
Yet even in Josephs case, the listener can often work out which
consonant would yield the most appropriate word for the context,
and so identify the words which he is targeting. Even though []
could be, for example, pot, tot, dot, cot, got, shot, jot, we are rapidly
able to home in on got in the above sample.
It is only possible for us to do this because the childs output is
perfectly appropriate in other respects. For a start, most aspects of
the childs words are ne. Even phonologically, his words preserve
much of their adult targets. They almost always have the same
number of syllables and the same stress pattern. Their vowels are
virtually always correct. Note, for example, the correct number of
syllables, stress and vowels in Stephens tiger and again, and
Josephs rabbit and colour. Distorted as their speech might at rst
appear, we nd that it is only certain consonants that are system-
atically affected.
The fact that important aspects of word phonology are unim-
paired is partly what enables us to identify the words the child is
targeting. For the rest, we rely on context. We can do this just
because the childs words are syntactically and semantically ap-
propriate. Syntactically, the combinations in which they occur are
normal, with words virtually never omitted or misplaced. Seman-
tically, they convey meanings which are compatible with each
other and make sense. We might also note that the childs combi-
nations of words have appropriate rhythmic patterns. For example
Josephs utterance
# # #
[n a lr n rs]
# # #
Can I colour it in rst?
has stress on the words colour, in and rst, leaving can, I

and it unstressed as we would expect them to be, with the vowel
// of can reduced to [] in its weak form.
Phonological problems
We have narrowed down Stephens and Josephs difculties to
the area of phonology, rather than syntax or semantics. We have
further narrowed it down to the phonology of words. Even more
specically, we have identied consonants in words as the culprits,
these being vulnerable to substitution or omission. Such a problem
has come to be described as a phonological problem. The prob-
lem is so described for two key reasons which are connected to
each other.
The rst is that it is the childs phonology her sound system
which is affected. This means that sounds are not randomly omit-
ted or replaced. Rather, it is certain classes of sound within the
sound system of the language which are affected. They are re-
placed with related sounds or omitted. This results in a smaller
range of sounds and sound contrasts, at least in some places in
words. The sort of substitutions and omissions which occur have
come to be described as phonological processes. These are
processes which affect target consonants in particular ways. For
example, we have seen that Stephen often realises targets /k/ and
/'/ inappropriately. These sounds of adult English share certain
characteristics: they are both produced by forming a closure be-
tween the back of the tongue and the soft palate, or velum, so that
the air ow through the mouth is momentarily blocked. This
means they have the same place of articulation in the mouth.
Stephen, however, replaces /k/ and /'/ with a sound which is
made by forming a closure further forward in his mouth, between
the front of the tongue and a point in the hard palate known as the
alveolar ridge: a sound which we hear as [d]. Thus, he has shifted
the place of articulation of the target. This not unusual substitu-
tion is captured by a phonological process known as fronting,
whereby velar /k/ and /'/ are fronted to [t] or [d].
Producing the sounds [t], [d], [k], ['] involves a complete
closure in the mouth, a characteristic also shared by the sounds
[p] and [b]. This means they have the same manner of articula-
tion. The set of sounds produced in this manner are known as
stops. They contrast with another set of sounds where the lower
lip or the tongue moves towards the upper lip or palate without
forming a complete closure. Here, the airow through the mouth
is partially but not completely obstructed, causing friction. Sounds
produced in this manner are known as fricatives. In English they
include [f], [v], [], [], [s], [z], [], []. These sounds share their
manner of articulation with each other, but vary in their place of
articulation. Some children have difculty with fricatives, and
replace them with stops. This alters the manner of articulation of
the target. Karl exemplies this in the words share, for, some which
he pronounces as dare, bor, dum:
Karl: (playing with a box of keys)
# # #
[a d dwm ]
I gotta share them out
# # # #
[ wm b mi] [ wm b ju]
One for me One for you
# # #
[ am dn ju dm]
Im giving you some
This substitution is described by the process of stopping. Typi-

cally, when Karl targets a fricative, he uses the stop which is
produced at the same or nearest place in the mouth. He replaces
/s/, an alveolar fricative, with [d], an alveolar stop. And he re-
places /f/, a fricative involving the lower lip and upper teeth
(labio-dental), with [b], a stop involving both lips (bilabial). As
well as stopping, Karl shows the fronting process we met in
Stephen, so that /'/ in giving is realised as [d].
The class of fricatives and stops divide into pairs of sounds
which share their place and manner of articulation, but which
differ according to whether the vocal cords vibrate during the
closure to produce what is known as voice. /d/, /'/, /v/, /z/, for
example, are voiced. Their voiceless counterparts are /t/, /k/, /f/,
/s/. Both Stephen and Karl show difculty with this voicing con-
trast. They both exhibit a process known as voicing, whereby
voiceless stop targets are voiced. So, Stephen not only fronts /k/,
which would result in [t]; he also voices it, resulting in [d], so that
the word cases sounds like dases. Karl not only stops /s/, but also
voices the stop, resulting in [d], so that some sounds like dum.
As with Stephen and Karl, the deviations in Josephs speech
affect particular classes of sounds. But the effects are more far-
reaching. They involve a wider range of targets, and more extreme
deformations of those targets. In some cases, he omits stops and
fricatives altogether, showing a process of consonant deletion.
#
This is what happens with the /k/ in [ lr] (colour) and the initial
/d/ in [az] (died). In other cases, he does produce a consonant
where one is expected, but it turns up as [s] or [z]. We see this with
the /b/ in [ws] (rabbit) which becomes [s], and the nal /d/ in
[az] (died) which becomes [z].
While the extent of these deformations varies, the fact that they
are systematic, affecting particular classes of sounds in particular
positions in particular ways, leads them to be classied as phono-
logical. The implication is that the childs problem is not with
individual sounds, but with classes of sounds which are related to
each other and which contrast with other classes of sounds.
A second reason for identifying the problem as phonological
draws on evidence of a different type, but leading to the same
conclusion. This evidence is that the child is physically able to
produce the sound which she is failing to produce. There is no
organic obstacle which prevents her from moving her articulators
as required to produce the target. Indeed, in some cases the child is
able to produce the very same sound in isolation, or in certain
positions. Stephen, for example, does produce [k] and ['] in some
positions in words such as tiger, monkey, take and back. This
proves he is physically capable of moving his tongue in a manner
appropriate to produce velar stops. But if articulation is not re-
sponsible for the problem, what is? Proposed answer: phonology.
What this says is that the problem is somewhere in the childs
mental organisation of speech sounds into classes, rather than in
the physical execution of those sounds. According to this interpre-
tation, the child is capable of producing the gestures for velar
stops, and it is a difculty with sorting out the contrast between
alveolar and velar stops which leads her to realise velars as [d].
This is how children such as Stephen, Karl and Joseph have
come to be labelled as phonologically delayed or phonologi-
cally disordered. If the childs system of sound contrasts appears
to be similar to that of a much younger child, phonological devel-
opment may be described as delayed. On the other hand, the
childs system may show patterns of substitution or omission
which are rarely if ever observed in the majority of children, even
at an earlier stage. Or the childs system may be out of step with
normal development, showing some patterns characteristic of an
earlier stage of normal development and others characteristic of a
later stage. Children exhibiting these atypical patterns may be
described as not just delayed in development, but disordered.
However, the distinction between delay and disorder is not a
clear-cut one. Since there is considerable variation within the
limits of normal development, it is not possible to draw an absolute
distinction between what does and does not occur normally.
But in any case, what do these labels tell us? They tell us that
the childs speech patterns differ from adult patterns, and also from
the patterns observed in most other children, in certain systematic
ways. They do not tell us at what point in the processes of receiv-
ing or producing speech these differences arise. This is a psycholin-
guistic question, and it is to this psycholinguistic question that we
now turn.
Possible sources of phonological problems

Psycholinguistic investigation, as we have seen, sets out from a
hunch. In the case of these children, the initial hunch leaps out.
The deviations observed in these childrens speech are quite speci-
c. We have established that their language is ne in most re-
spects. Once we adjust to their phonological systems, we discover
that they have no problems with the semantics or syntax of words.
We also discover that they use their words in contexts which are
semantically and syntactically appropriate, and in appropriate
conversational interactions with other people. Their problems are
conned to word phonology, with particular aspects of speech
sounds being deformed in particular ways. Our rst hunch, and a
pretty well founded one, will be that the childs difculties reside
specically in phonological processing. This entails the following
components in our model of word processing:
INPUT OUTPUT
Discrimination of relevant Accessing of word

features of speech phonology
Identification of word Articulatory planning

phonology
Storage of word Articulatory execution

phonology
Figure 5.1 Components of phonological processing

So far, the only component of speech processing we have elim-
inated is articulatory execution. Since the children are physically
capable of making the sound distinctions they are failing to make,
articulatory difculties were ruled out as an explanation for their
phonology. This still leaves a number of possibilities. Does the child
have difculties in input which then give rise to the deformations
observed in output? If so, does the problem occur in discrimination
of words, or only in the representations of words which the child
stores? Or is the problem conned to output? If so, does it occur in
the representations which the child accesses for output, or only at
a later stage when representations are programmed for articula-
tion?
These questions can be tackled by looking rst at the speech we
have already observed, to see if this leads us to rule out certain
possibilities and suspect other possibilities. Based on the evidence
we have, our next hunch may be that the problem is conned to
output. For a start, we have not observed difculties in compre-
hension which would suggest difculties with input: all the
children appear to understand and respond appropriately to the
input they receive. If they had difculties in recognising the pho-
nological distinctions they fail to produce, this would surely distort
the input they receive and we might expect this to interfere with
their comprehension. This would be especially true for a child like
Joseph who is producing very few distinct consonants. If he regis-
tered so few consonants in input, he would lose a great deal of
phonological information which might be thought critical to re-
cognising and understanding words.
But how tight is this evidence? Not tight enough. It is just
possible that Joseph does have difculties in input, but that the
phonological information he does receive, together with astute
guesswork based on his interpretation of the non-verbal context,
enable him to understand language in context. In other words, it
may be that Joseph is receiving the same degraded input from
others that we receive from him, and that he is still able to pick out
words and retrieve meaning in others speech just as we do with
his. Evidence of comprehension in context is not precise enough to
rule out input difculties. We need to dig deeper for more reliable
evidence. To do this, we must turn to the sort of input and output
tasks we met in the previous chapter. Such tasks have been carried
out with groups of children and with individual children who have
phonological problems. What do they reveal?
Evidence from tasks tapping different components

of input and output
Investigations of input phonology rely on tasks which ask the
child to make some sort of judgement about phonological forms
or to classify them in some way. The forms used in these tasks are
chosen to test just those phonological distinctions which appear to
be problematic for the child. Take the case of Stephen, who fronts
velar stops so that they sound like alveolar stops. Can he tell the
difference between these two types of stop? To check this, a classi-
cation task was introduced, in the following way:
Look, weve got red cars and green cars and I want you to post them
into this box. But you must listen before you post them. I want you to
post a red car if I say [d] and a green car if I say [']. So see if you
hear [d] or ['], okay . . .
After introducing the task, Stephen was randomly presented with

the two forms [d] and [']. The task was then repeated with
# # # #
the pair [f dt] and [f 't], and the pair [d rkt] and [k rkt].
He selected the appropriate car for all thirty items presented. The
task was then used with three more pairs, [dn] and [kn],
# # # #
[ dn] and [ 'n], [dm pit] and [km pit], but now he was not
told which car went with which member of each pair. This meant
he had to decide whether [dn] was like [d] or like [']. Under
these conditions, Stephens judgement was still perfect. From this,
we know that he could discriminate velars from alveolars, and
assign them to distinct sound categories.
Another way of checking whether the childs input is different
from her output is to have her listen to her own output. If we play
her speech back to her, will she hear it as we hear it? Where two
words sound the same to us, will they also sound the same to the
child? Or will she hear distinctions in it which we cannot hear?
This was investigated in a group of 3- to 5-year-olds who were
making a variety of phonological errors (Dodd, Leahy & Hambly
1989). The children were played a tape of their own production of
words, and had to choose the picture for the word they heard from
a set of four pictures. The set of pictures included one illustrating
the word the child was targeting, and three illustrating words
which were the same as or phonologically similar to the childs
pronunciation. When their own pronunciation sounded like a
different adult word, it was found that the children chose the
picture matching their pronunciation rather than their target.
Suppose, for example, that their production for tree sounded like
tea. Hearing this production on the tape, they would go for the tea
rather than the tree picture. This shows that these children were
hearing their own speech as the adult does, rather than as they
intended it to be.
Such evidence allows us to rule out a problem in phonological
discrimination. It does not rule out a problem in phonological
representation, however. The child may discriminate sounds but
fail to store them distinctly. This possibility was not checked with
Stephen, but it has been investigated in other studies. In the
research study described above (Dodd et al. 1989), the auditory
discrimination task using tape recordings of the childs pronunci-
ation of words was repeated with tape recordings of an adults
production of those words. For example, the child pronouncing
tree as tea would hear the adult pronunciation of tree and have to
select a picture from a set of four which included tree and tea. In
order to select the correct picture, the child must discriminate the
form tree from tea, but must also have that distinct form stored as
the label for tree. The subjects in this study were much better at
identifying the adults pronunciation of words than their own
pronunciation, showing that they have both discriminated and
stored the distinct forms of these words.
Where the childs responses to input tasks are reliably correct,
we can conclude that she has discriminated and stored the phono-
logical distinctions tested in those tasks. We can in turn conclude
that the difculties observed in output arise only in output process-
ing. But at what point in output processing? To investigate this, we
turn to tasks which separate out the accessing of phonological
representations from the programming of their articulation.
These are exemplied by a set of categorisation and production
tasks carried out with a group of 4- to 7-year-olds who were
fronting velar stops, producing target /k/ as [t] or [d], so that cap
sounded like tap or dap (Brett, Chiat & Pilcher 1987). The tasks
were also presented to a group of children of the same age as the
fronting children, but who had no phonological problems.
For the categorisation tasks, children were introduced to two
postboxes and told that one went with the sound [t] and one with
[k]. They then had some practice in posting cards according to
which sound they heard. After the practice, they were given an
auditory judgement task and a pictorial judgement task. In the
auditory task, they heard a mixture of six words beginning with
/k/ and six beginning with /t/, and they had to post a card into the
box matching the sound they heard. In the pictorial task, they
were given pictures of the same twelve words to post. In this case
they had to think of the words themselves, and post the pictures
according to the sound they could hear in their heads. The results
revealed that the fronting children did not differ signicantly from
the normally developing children in these tasks: they were able to
categorise both auditorily presented words and pictures. The
childrens success on these tasks shows that they do not have
difculties in input: they have discriminated and stored the targets
which they are not producing. But their success on the pictorial
task goes further. It shows that they know which words begin with
which of these sounds even when they have to think up the word
forms rather than being given them. That is, the children had in
mind distinct forms for words beginning with /t/ and /k/.
Results in the production tasks, on the other hand, revealed
striking differences between the two groups of children. These
tasks involved naming the pictures, repeating their names, and
repeating similar non-words beginning with /t/ and /k/. Here, the
fronting children made signicantly more errors than the nor-
mally developing children. The difference between the form in the
childs mind and the form he actually produces is neatly illustrated
by one child who picked up the picture of the car, said
[t] [t]
and proceeded to post the picture in the [k] box! This child mispro-
nounces the word, but is not deected by his own mispronunci-
ation, showing that he is hearing a /k/ target internally.
These ndings make it clear that the fronting children had no
particular difculty in discriminating, storing and accessing the
distinction between velar and alveolar stops. Their difculty with
velars must, then, arise at a later point in output processing at
the point where its articulation is planned. This component of
processing is involved in all the production tasks (naming, repeti-
tion of words and repetition of non-words) but not in any of the
categorisation tasks.
With these children, we appear to have come full circle. We set
out with the claim that such children have the articulatory capac-
ity to produce sounds they are not producing, and that their faulty
output cannot be due to articulation. Instead, it was attributed to
limitations in the speech contrasts they make in their organisa-
tion of phonological distinctions. Yet now, having investigated
their processing of those same phonological distinctions, we have
found that the bulk of that processing is unimpaired: they make
the relevant contrasts throughout their processing of words in
input, and even in accessing those words for output. We have
concluded that their difculties arise in getting those words articu-
lated. Is there a contradiction here? On the one hand, we have
concluded that the child has no problem in articulation; on the
other hand, that the childs failure to make certain contrasts
occurs at the point of articulation. These conclusions may appear
contradictory at rst sight, but they are not, once we take into
account the complexity of the articulatory process itself. Between
the accessing of a words phonology and its articulation lie the
intricate processes of speech programming. Theres room for many
a slip in these processes. This emerges very clearly when we
explore patterns of errors in the childrens output, and their impli-
cations for speech processing.
Evidence from patterns in the childs output

We have already observed that childrens difculties with par-
ticular phonological targets are not necessarily consistent. Many
children hit the target in some words but not in others. Stephen,
for example, could produce velars correctly in tiger, monkey, back,
take, but fronted them in got, called, again, forgot, cant, cases.
These childrens ability to produce a sound in one context and
not in another can prove quite puzzling:
Stephen
SC:
Stephen: This puppet is called Mikey. Can you say Mikey?
#
Stephen: [ ma']
SC: (showing a key) Can you say my key?
#
Stephen: [ma di]
Karl
SC: Can you say icing?
#
Karl: [ as]
SC: Can you say I sing?
#
Karl: [a d]
These data reveal a curious phenomenon. Karl can say icing but
not I sing. Stephen can produce a velar in Mikey but not in my key.
When David is presented with aching and a king, he spots a trick:
David
SC:
David: Can you say a king?
#
David: [e d]
SC: Can you say aching?
#
David: [ ek]
SC: Now, a-ching (slowed down)
#
David: [ e k]
SC: Now (showing a picture of a king) a king
# #
David: [ s trk wn] ( = Thats a trick one!)
Davids detection of a trick suggests that he has come face to face

with the paradox which occurs in his speech and the speech of the
other two children. Each child appears to be able and yet unable to
produce a particular target correctly. This apparent paradox leads
to the question: what makes it possible for the child to hit the
target in some words yet not in others? If we can discover the
factors which affect the childs production, we may throw some
light on the nature of the childs difculties and the point at which
they arise. To discover the factors, we need a good range of
examples of target and non-target strikes.
We might start by collecting a sample of spontaneous speech
and separating out all those words which contain incorrect
Table 5.1. Stephen: Velar targets in spontaneous production
Correct take, back, walk, duck, legs, Goldilocks

tiger, nger, monkey, foxes
Fronted got, called, get, cases, gonna, come, Goldilocks
forgot, again
Table 5.2. Karl: Fricative targets in spontaneous production
Correct youve, have, cos, this, his, off, hes, ambulance

diver, even, over, sister, person, scissors, sausages
Stopped for, fit, fish, some, sausages, sister, scissors
realisations of a target and all those words which contain correct

realisations of that same target (see tables 5.1 and 5.2). At rst
glance, the difference between the correct and incorrect cases
looks obvious: targets at the beginning of words are incorrect;
targets in the middle or at the ends of words are correct. So,
Stephen fronts /k/ and /'/ when they are word-initial, but not
word-medial or word-nal. Karl stops word-initial fricatives, but
not word-medial or word-nal fricatives. There is just a hint in
these data, however, that the picture is not quite so simple. In
forgot and again, the velar is word-medial, yet Stephen fronts these
velars. Is this a random error? Or is there some difference between
these word-medial velars and those which he produces correctly?
It turns out that there is. The stress pattern of these words is
different. Forgot and again have stress on the second syllable
# #
/f 't/ and / 'n/. Here, the velar stop precedes the stressed
vowel. The words tiger, nger, monkey, foxes, on the other hand,
have stress on the rst syllable, following the more typical stress
# #
pattern for two-syllable words in English / ta'/, / f'/,
# #
/ mk/, / fksz/. Here, then, the velar stop follows the stressed
vowel. Could stress be the factor which determines whether medial
velars are correct or fronted?
In order to nd out, we obviously need more examples of words
with targets in different positions. There are unlikely to be enough
Table 5.3. Karl: Fricative targets in repetition
Correct Word-nal laugh, cough, safe, dive, give, miss,

dice, rose
Word-medial after offer, toffee, river, answer, easy,
stress pencil
Stopped Word-initial feed, fog, fat, fall, sit, suit, safe, sock
#
Word-medial before before, re fuse, deserve, cassette,
stress decide, beside
examples of such words in the childs spontaneous output. So, we

need to nd a way of eliciting further examples. One possibility is to
ask the child to repeat appropriately selected words. So long as the
child repeats words in the same way as he produces them sponta-
neously, we can use repetition data to probe his difculties.
Repetition tasks were tried out with both Stephen and Karl. The
stimuli were designed to investigate production of the relevant
targets in different positions within the word: word-initial, word-
nal, word-medial before stress, and word-medial after stress. Both
children were very willing to repeat words, making it possible to
sample a good range of targets. Both showed the same pattern of
production. This pattern is exemplied by Karls stopping (see table
5.3)
Word-nal fricatives are correct, whereas word-initial fricatives
are stopped. In word-medial position, production of fricatives de-
pends on stress. They are correct when they follow stress, but
stopped when they precede stress. In the same way, Stephen
produced velar stops correctly if they were word-nal or word-
medial following stress, but fronted them if they were word-initial
or word-medial preceding stress (see table 5.4). The effect of stress
on Stephens production of target /k/ is highlighted by the contrast
# #
between record (correct) and re cord (fronted).
Why should the child succeed in making a distinction in some
positions but not others? Why is it possible for Karl to produce the
fricative /s/ in icing but not in I sing? The observation of such
differences in childrens production of consonant targets offers a
Table 5.4. Stephen: Target velar stops in repetition
Correct Word-nal tick, duck, dog, tag, cake, cook,

kick, gag
Word-medial after making, lucky, wriggle, wagging,
stress tickle, talking, turkey, ticket,
#
baker, angry, whiskers, record,
aching, basket
Fronted Word-initial key, car, go, gay, cake, cook, kick,
gag, cat, kid, get, gate
#
Word-medial before because, agree, begin, re cord,
stress forget, okay, again, become,
secure
mine of information about how speech is processed, and which

speech patterns are easiest for the child to process. One implication
of these differences is that stress pattern, or rhythm, is very import-
ant in speech processing. Children like Stephen and Karl show us
that the alveolar/velar distinction and the stop/fricative distinc-
tion are not produced independently of their place in the rhythm of
a word, since a single one of these distinctions may come out
differently according to its position in relation to the stressed vowel
in the word. The stress pattern of a word appears to act as a frame
on which further details of speech are hung. The stress pattern
affects whether a consonant hangs onto a preceding or a following
vowel, and this in turn affects the way the consonant is produced.
In the examples we have met, the velar closure and the fricative
closure were easier to produce when they followed from a vowel
(word-nal) than when they preceded a vowel (word-initial). In
word-medial position, the relation of the stressed vowel deter-
mined how they behaved. If they preceded the stressed vowel, they
behaved like a word-initial and were produced incorrectly. If they
followed the stressed vowel, they behaved like a word-nal and
were produced correctly.
These rhythmic frames on which the consonant distinctions
hang appear to be word-size. The evidence for this is that the stress
pattern over a word affects the pronunciation of these distinctions,
but the stress pattern over a sequence of words does not. Compare
Stephens production of the single word beacon and the sequence of
words he can:
# #
beacon ; [ bikn]
Here, the target /k/ comes between a stressed vowel and an

unstressed vowel and in the middle of a word. This velar is correct.
# # # #
he can dance ; [ hi tn dns]
Here, the target /k/ still comes between a stressed vowel and an
unstressed vowel, but the unstressed vowel is in a separate word
and the velar is at the beginning of that word. This velar is fronted.
Conversely, when /k/ comes between an unstressed vowel and
a stressed vowel within a word, Stephen fronts it:
# #
be cause ; [b dz]
But if the /k/ occurs at the end of a word in this stress pattern he
preserves the velar:
# #
back out ; [b' t]
(though he changes its voicing).

Drawing together these observations, it appears that each word
in a sequence of words is produced as a separate rhythmic unit.
This rhythmic unit consists of one or more vowels, which may be
anked by consonants. Whether the consonants attach to a pre-
ceding vowel or to a following vowel depends in part on the stress
of the vowels, that is, on the position they occupy in the rhythm of
the word.
So far, we have only looked at the processes of fronting and
stopping. These involve changes in the place or manner of the
target consonants. What about other aspects of speech which are
affected in children with phonological difculties? One such aspect
is the voicing of consonants. In English, some consonants are
voiceless, with the vocal cords open while the closure is made, as
in [k] or [t] or [s]. Others have vibration of the vocal folds, as in [']
or [d] or [z]. As we have seen, some children have difculties with
this voicing distinction. For example, they may produce voice
when they are targeting voiceless consonants:
Stephen: voiceless /k/ becomes voiced [g]
#
monkey ; [ m']
#
baker ; [ be']
Karl: voiceless /s/ becomes stopped and voiced [d]
sit ; [d]
sea ; [di]
This sort of error again occurs in certain positions only. Again,

it appears to occur at the beginning but not at the end of words, at
least when they stand alone. For example, Stephen does not voice
the word-nal stops in walk and back. But look at what happens
when these words occur in phrases, with a vowel following:
# # #
[ i dn w' m wd]
He can walk on water
# #
[ dm b*' t]
Come back out
Here, the stop at the end of the word is correctly realised as a velar,
but it is incorrectly voiced. Why should the /k/ be voiceless when
back occurs alone, but voiced when it is followed by out? Could it be
that a stop comes out voiced when it is followed by a vowel, even if
that vowel is in a separate word?
This possibility was investigated experimentally (Brett, Chiat &
Pilcher 1987). The subjects in the experiment were ten children
whose speech showed voicing. These children were asked to repeat
words and short phrases containing voiceless stops at the begin-
ning, in the middle, and at the end of the word or phrase. In single
words, the children did exactly as expected: they always voiced
initial stops, but always produced nal stops correctly. What
happened when the words were combined into short phrases?
Now, the word-nal stops which had been correct in isolation
tended to be voiced, and were almost always voiced if the following
# #
vowel was stressed as in get up, back out. This conrms our
suspicion that voicing of consonants is affected by a following
vowel, whether that vowel is in the same word or not. Processes
altering the place or manner of a consonant, on the other hand,
appeared to be affected by its position in relation to the stressed
vowel within the word. This explains the correct placement but
incorrect voicing of /k/ in Stephens production back out. Velars are
correct word-nally, so the velar aspect is correct. But voiceless
stops are liable to be voiced before any vowel even if they are
word-nal, so the voicing aspect is incorrect.
Detailed analysis of the types of phonological errors which
occur and the phonological contexts in which they occur has led
us into the thick of speech processing. We have inferrred that
different consonant distinctions are processed according to their
position in relation to stressed and unstressed vowels. Place and
manner distinctions were seen to be affected by vowels preceding
and following them within the word, while voicing was affected by
any following vowel.
Looking ahead
Drawing on a range of evidence, we appear to have nailed down
the source of childrens phonological difculties. Our conclusion
has been that the problem is in articulatory planning (see model of
phonological processing above), and we have gone on to explore
the nature of that problem. But there is an alternative possibility
which we have overlooked. Throughout our consideration of the
data, we have assumed as most observers do that the child is
failing to make a phonological distinction such as the distinction
between alveolar and velar stops, or between fricatives and stops.
We have then investigated different components of input and
output processing to see where this failure occurs. It is true that
when Stephen produces a word such as car, we perceive a /t/ or
/d/ rather than /k/ at the beginning. When Joseph produces this
word, we perceive it as having no initial consonant at all. But is
our perception reliable? Suppose these children are making a
distinction in their speech, but one which we cannot hear. This is
entirely possible. There is even some evidence that it happens. This
evidence is of a different kind from any we have so far considered.
It lies in measurement of acoustic properties of the childs speech.
An acoustic analysis was, for example, carried out with three
children who omitted stops at the ends of words (Weismer, Din-
nsen & Elbert 1981). The analysis revealed that two of the
children, aged 7, were producing words differently according to
whether the stop was voiced or not, even though the stop was not
itself produced. The difference occurred in the vowel, which was
reliably longer if the word should have ended in a voiced stop than
if it should have ended in a voiceless one. In adult speech, there is
also a difference in the length of the vowel according to the voicing
of the following consonant. Evidence of such contrasts in the
childs speech proves conclusively that the child is making a
phonological distinction, and eliminates any explanation in terms
of a lack of contrasts in the phonological system. This is consistent
with our ndings that some children do make distinctions in input
tasks and even in internal judgement tasks where they identify car
as beginning with /k/ even though they appear to produce it as
[t]. Under this interpretation, these tasks represent no problem
because the child makes the relevant distinction, and the reason
we dont perceive her distinction is that she does not make it in the
relevant way.
The problem shifts from failure to make a distinction, and
becomes failure to make an appropriate distinction. This puts us
back to square one in considering where this problem occurs in
input or output processing. Difculty may arise in picking out the
correct features from the input, or in storing them, or in accessing
them, or in programming their articulation. We may hazard a
guess that the problem still arises in articulatory output. It seems
unlikely, for example, that a child would distinguish /s/ in input
on the basis of features so different from the adults that it ended up
sounding like [t] to the adult ear. To move beyond a good guess,
we would need more sophisticated psycholinguistic tools than we
have so far used. If we are to check not just whether the child
recognises a distinction, but what sort of distinction the child picks
out, we cannot use normal spoken input which contains only
normal adult distinctions. A possible way of checking the childs
sensitivity to particular distinctions is to use synthesised speech.
This allows us to reproduce just the acoustic distinctions we think
the child may be making, to see if she can recognise these distinc-
tions even if we cannot. If we can show that the child does perceive
distinctions which we do not, we would have evidence that she is
focusing on different features in input, which could account for the
output we observe.
The questions which are now emerging and the methods for
pursuing these are at the vanguard of current research. They
show how psycholinguistic exploration leads to the most delicate
dissection of the childs phonological processing.
Note: Stephen, Karl and Joseph were originally presented in Chiat 1983,
Chiat 1989 and Chiat 1994. To avoid confusion with the subject in Chiat
1983, the name of the subject in Chiat 1989 which was also Stephen has
been changed to Karl in this chapter.
6 Stip or step or slip or
what?: problems with
lexical processing
Eamonn is an alert and excitable 6-year-old. He engages intensely

with the world around him, a willing participant and an enthusi-
astic initiator. He is assertive about his own interests and prefer-
ences (not least his passion for Batman) but is at the same time
sensitive to other peoples concerns and points of view (including
their ignorance about Batman). And he loves talking: telling
whats happening at home or school, expressing what touches
him or those around him, nding out whats going on for other
people. His words often tumble out in an eager torrent, as they do
when he starts recounting a class trip into the garden:
E: We [wn] in the garden get some [wi'] get some soil.
#
[b k] we [n] got no soil.
SC: Where havent you got any soil?
# #
E: [bk b k bk] we went in the garden dig and [ kti fan]
#
a snail a snail. [s] not a big snail [s l] snail.
Seven-year-old Michael (Constable, Stackhouse & Wells 1997)
also has his wits about him. His particular passion emerges when
he is asked to name the kinds of work people do and he comes up
with:
making toys, making trainsets, making steam trains, making electric
trains, making diesels . . .
Michael shows considerable ingenuity and insight in his negoti-
ations with words. Unfamiliar with unicorns, he responds to a
picture of a unicorn with the label horse with a party hat on.
Faced with a picture of handcuffs, he does recognise what they are
and strives to get this across:
88
Problems with lexical processing 89
key . . . oh what do you call them . . . oh yeah . . . you put . . . you put
. . . with your . . . with your . . . oh . . . with your . . . when you . . .
when someones stole something . . . and . . . what do you call them
. . . necklace? . . . no . . . I just dont know the word.
First impressions of children like Eamonn and Michael may be

that they are quite intelligible. Many of their words sound just as
they should and can be readily recognised and understood. But
such rst impressions may be deceptive. These children produce
words inconsistently. They may produce some words correctly,
while similar sounding words are produced incorrectly. They may
produce a particular word correctly on some occasions, but not on
others. Incorrect attempts at a word may themselves be inconsist-
ent, with different errors occurring on different occasions. In some
cases, their words may be unrecognisable. When they produce a
whole stream of these, we may have difculty making sense of
what they are saying. Yet even at this extreme, we may perceive
them as being in some sense on the ball. Intuitively, we may feel
that they have coherent meaning intentions but that these have
got obscured or lost in a morass of output.
When these children talk, they present us with a much more
hazy picture than the phonologically impaired children followed
up in the previous chapter. The deviations in their words are at
once less consistent and more diffuse. This suggests that the source
of their difculties is different. To form an initial hunch about the
obstacles in the path of their language processing, we need to take
a closer look at the strengths and weaknesses in their language.
First observations
Eamonns language impairment is evident. At 5, a test of sen-
tence comprehension (Test for the Reception of Grammar, or
TROG) had him performing at the level of a 4-year-old, and a
word-nding test (the Renfrew) at 3. By the age of 7, the gap in
performance is greater. On the TROG and on a word-comprehen-
sion test (British Picture Vocabulary Scales), he is two to three
years behind his age. In contrast, a non-verbal general intelligence
test (the performance section of the Wechsler Intelligence Scales
for Children, or WISC) nds him average/low average for his age.
To get a feel for Eamonns language, consider this more ext-
ended sample of his conversation about the class session in the
garden:
#
E: We [wn] in the garden get some [wi'] get some soil. [b k]
we [n] got no soil.
# # #
E: [bk b k b k] we went in the garden dig and [ kti
#
fan] a snail a snail. [s] not a big snail [s l] snail.
SC: Cassie did?
#
E: Yeah. [ ksis d] and [i] see it [n n] thing. And then Cassie
Cassie throw it away in [] in in the pond. And then [i] the
the snail [km mek] it. No.
SC: The snail?
E: No. Ca We [li] the snail there. The snail in the ground
#
[*s] [b kh ds wi stp] it, [b] it be die.
SC: If you do what to it?
E: [j stp] it.
SC: If you step on it.
E: Yeah. It die.
Looking at this conversation globally, Eamonns contribution is

ne. He takes appropriate turns with the other person, initiating
comments and responding to questions. Each of his comments and
responses relate to the topic of conversation. Eamonn comments
on signicant events, such as the nding of a snail, and conse-
quences of events, such as the fate of the snail if stepped on.
In his responses to questions, we begin to see problems. Al-
though they still relate to the topic of the question, they do not
always answer the specic question that was posed. For example,
to a question about place:
Eamonn appears to respond with a reason, and then rapidly

changes the subject:
# # #
E: [bk b k b k] we went in the garden dig and [ kti fan]
#
a snail a snail. [s] not a big snail [s l] snail.
To a later question about purpose:

SC: What do you need the soil for?
Eamonn responds obscurely and again changes the subject:

#
E: Soil for we [ni pwn] the we [ nid] put some soil in in the
# #
thing in the pot and [n t nun] we [k] take them home.
[wi bi fi fi] good.
As well as producing some utterances which disrupt the thread

of the conversation, Eamonn produces some which are unintelli-
gible and whose relevance to other utterances therefore cannot be
judged:
. . . the snail [km mek] it
Here, where Eamonns words are unclear, we cant decipher

whether or how they relate to the context. When he goes on to
say:
We [li] the snail there
we dont know whether there refers to in the pond (already

mentioned) or in the ground (mentioned subsequently). Clearly
Eamonn is saying something about the location of the snail and its
fate, but the connection between the two remains obscure.
In shifting our view from the more global to the more local
connections between utterances, Eamonns difculties emerge.
When we narrow our focus even further, looking at individual
utterances, his problems are very evident. His combinations of
words do not always match the requirements of English sentences,
sometimes missing elements such as markers of time:
Cassie throw it away = throws? threw?
. . . it be die = will die? will be dead?
One or two combinations are very odd:

Soil for we [ni pwn] the . . .
Finally, there are problems with words themselves. He gropes for a

word:
#
We [ nid] put some soil in
and may come up with an empty substitution:

in in the thing
before nally hitting the target:

in the pot
In some cases, he stops with the empty substitute:

. . . and [i] see it [n n] thing
In others, he goes for the target directly, but the sound of the word
is distorted, often to a point where it is unrecognisable:
#
[ kti] = Cassie
[wi] = if we?
[stp] = step? slip? both?
His attempts at a target are not necessarily consistent, for example

he produces Cassie perfectly on some occasions.
While the sound of word forms is often distorted, Eamonn does
appear to be targeting a single word form. And the meaning of that
form is generally appropriate. At worst, it is over-general (thing).
There are no instances of words being semantically inappropriate
in this sample, though semantic errors were occasionally noted in
other samples of Eamonns output:
(Describing picture of man feeding horse)
#
E: Daddy [ fin] the cat.
SC: Whos hungry?
E: The [k] the horsie.
(Talking about a motor boat)

# # #
E: I got this [ m ba] er [ m bak m bak].
SC: A motor bike?
#
E: A [ m bak] er boat.
From these samples, we can draw a preliminary picture of

Eamonn. He appears to show problems with all aspects of utter-
ance use and utterance structure. His utterances are not always
fully appropriate to their semantic context; their syntactic or-
ganisation is sometimes wanting; and the words they contain are
often hard won, or semantically vague, or phonologically dis-
torted. The picture does not show a clear-cut mismatch between
strengths and weaknesses such as we met in children whose
difculties were conned to phonology. But there is some hint of
mismatches. Pragmatically and semantically, Eamonns output is
relatively strong. His utterances may be imprecise in their content
and relation to the topic, but they are always in the right ball park.
His word forms and combinations of words, on the other hand,
show persistent deviations from those needed to express his in-
tended meanings.
Starting from this apparent mismatch, our rst hunch may be
that Eamonns difculties are primarily with the forms and struc-
tures of language, rather than their meanings and functions. This
predicts that Eamonn will show problems with words and senten-
ces independently of, but possibly responsible for, any problems we
meet in meaning and function.
Through exploration of Eamonns spontaneous output, we
have homed in on his problems with words, and primarily the
form of words. The study of 7-year-old Michael (Constable, Stack-
house and Wells 1997) sets out from a similar point, though this
point is reached mainly via formal assessments rather than spon-
taneous output. Michaels prole resembles Eamonns in a number
of ways. Like Eamonn, he comes out in the low average range on
a test of intelligence (the Kaufman Assessment Battery for Chil-
dren). Also like Eamonn, he has good conversational understand-
ing, but comes down on formal tests of comprehension. His score
for vocabulary comprehension (BPVS) is at the level of children a
year younger, which identies him as low average; his score for
sentence comprehension (TROG) is more than two years below his
age mates. His output, though, shows an even more striking delay,
his score on the Test of Word Finding being way down for his age.
These test results indicate that Michael has difculties. To nd out
whether his difculties lie primarily in semantics or phonology,
some preliminary assessments were carried out.
Two of these were naming tasks. Here, Michael made a few
semantic confusions similar to Eamonns rare semantic errors:
amingo ; octopus
moustache
moustache ; [bj st b st bj bjd st sts bsts]
= beard + moustache?
But he frequently described a picture using a circumlocution:
hammock a net where you sleep on
or a distorted form:
cigarette [ssjt] [si:jt]
In these cases, he is clearly targeting a word which has the

appropriate meaning. This is the strongest evidence that the
source of his problem with words is their form rather their mean-
ing. This evidence is supported by the results of two tests which
focus on meaning. One, known as Pyramids and Palm Trees
(Howard and Patterson 1992), was actually designed for adults. It
presents an array of three pictures, one at the top (e.g. a pyramid)
with two below it (in this case, a palm tree and an oak tree). The
task for Michael was to match one of the bottom pictures with the
top one (palm tree going with pyramid). In order to succeed, the
child must know which pictures are related to each other. Though
billed as a test of semantics, strictly speaking Pyramids and Palm
Trees tests a persons knowledge about the world rather than
knowledge of word meaning. This is shown by one of the few
errors Michael made. Presented with a picture of a windmill, he
selected the daffodil rather than the tulip cos theyre in the
countryside. His wrong choice, then, is due to ignorance of the
Dutch connection rather than ignorance about words. Michaels
impressive score, 22/24, indicates that his world knowledge is
generally good. A second test required Michael to come up with
items in response to a category such as kinds of work people do
(Word Association subtest of the Clinical Evaluation of Language
Fundamentals, or CELF). Here, Michael produced appropriate if
rather limited associates, as we saw with his examples of work:
making toys, making trainsets . . .
which were cited above. The results of these tests at least indicate
that Michael has made normal associations between things, and
they give no hint of semantic difculties.
Turning to phonology, a number of tests point to difculties. In
an auditory discrimination task (Bridgeman and Snowling 1988),
Michael was able to discriminate many pairs of words and non-
words, but had some difculty judging non-word pairs such as
vostvots, which vary the order of a consonant cluster. In a second
task, Michael was asked to nd words which rhymed with another
word. He found this very difcult, managing just three rhymes
compared with the average of 125 produced by 5-year-old chil-
dren. He made imaginative attempts to respond, resorting, for
example, to semantic associates or extensions such as cow ; calf,
hay ; hey dood.
As with Eamonn, initial assessment of Michael has highlighted
a problem with word forms above and beyond any difculties with
semantics. The question is: where do Eamonn and Michaels prob-
lems arise in the process of receiving and producing word forms,
and connecting these to meaning?
Possible sources of precarious words

Our hunch so far is that both children have problems with
words, and that those problems are primarily with the form of
words rather than their meaning. Both children have adequate
discrimination to separate out word forms. Their problems must
arise in the further processing of those segmented word forms. At
this point we have scant evidence of just where, in that processing,
their problems reside.
As far as input is concerned, we know that hearing is within
normal limits for both children (thanks to audiological testing),
which rules out difculties at the very earliest point in input. In
Michaels case, we have a hint of problems in ne auditory dis-
crimination. At the other end of the input process, both children
appear to get to more or less the right meanings behind utterances
in conversation, but we have seen that even in conversation
Eamonns responses are not always fully appropriate, suggesting
he may have difculties when it comes to precise understanding.
Such difculties are conrmed by the results of sentence-compre-
hension tests. Both children, put in a situation where wider cues to
meaning are not available, perform below their age. This could be
due to problems at any point in connecting the sounds they hear to
meaning: in auditory discrimination, in the identication or stor-
features of speech
Identification of word
phonology
Storage of word Storage of word

phonology semantics
Figure 6.1 Michael and Eamonn: possible points of breakdown in lexical input
age of word forms, or in accessing their meaning, as shown in

gure 6.1. All of these possibilities are compatible with our current
evidence of their understanding.
Turning to output, we know that both children produce a wide
range of speech sounds and sound sequences, and do so precisely.
Even more telling, they may produce perfectly on one occasion the
very sound in the very word which they have distorted on another
occasion (as with Eamonns production of Cassie). This is evidence
that articulatory execution is not the problem. Their difculties
must then arise at some point between meaning and articulatory
execution: connecting word meanings to word forms, accessing
precise word forms, or planning their articulation, as shown in
gure 6.2. All these possibilities are compatible with our evidence
so far.
Exploring these different components of input and output
means going beyond the limited evidence of spontaneous output
and existing tests.
Selection of word semantics
Accessing of word phonology
Articulatory planning
Figure 6.2 Michael and Eamonn: possible points of breakdown in lexical output
Evidence from tasks tapping different components

of input and output processing
The point of further investigation is to see whether clearly
identied deviations in the childrens output and roughly identi-
ed constraints on their input can be attributed to difculties in
particular components of input or output processing. We turn to
tasks which probe different processing components.
Each child received a range of tasks using words (or sometimes
non-words) of the sort he had produced incorrectly. Keeping to the
same sort of items ensures that any differences between tasks
using those items reect the processing involved in the tasks,
rather than the items themselves.
The items used in Eamonns tasks varied, but all were either
words which he had attempted spontaneously and were therefore
in his vocabulary, or non-words which were similar to these
words. The words reected his interests. They had cropped up in
talk around favoured subjects such as Batman (hence the item
Riddler naming a character in Batman), and the antics and
health of his cat (hence the item vet).
All Michaels tasks used a core set of ten items: hospital,
elephant, crocodile, microphone, octopus, escalator, bin-
oculars, helicopter, television, caterpillar. These three- and
four-syllable words were known to be in Michaels vocabulary,
and were selected as suitably challenging for him. Some tasks
supplemented these ten items with related items, for example
non-words which were phonologically similar to the words.
In Michaels case, the same tasks were given to two groups of
normal children. One group were selected to match his chrono-
logical age, averaging 7;8. The other group were selected to match
his vocabulary age and were therefore younger, averaging 6;01.
Input processing
Eamonns and Michaels difculties are most evident in their
output. We have, though, witnessed some constraints on input
as well. Could it be that their impaired output is actually a mirror
of what they are taking from the input they receive? To nd out
whether and how their processing of input may be limited, we
need to investigate their ability to perceive and recognise the sorts
of distinctions they are failing to make in output. To this end,
judgement tasks were used with both children.
In Eamonns judgement tasks, he was shown a picture which
the investigator named correctly or incorrectly. Eamonn was in-
vited to give her a tick when the word sounded right, and a cross
when it sounded wrong, so he could see how well she had done at
the end. The rst round of these tasks presented Eamonn with ten
error forms he had produced in his spontaneous output, mixed up
with the ten corresponding target forms. For example, the follow-
# #
ing picture was named once as [ k ru] and once as kangaroo:
Eamonn made no errors in judging these items, indicating that he
could recognise and reject the forms he had himself produced.
Rounds 2 and 3 upped the stakes, focusing on just those distinc-
tions which appeared most vulnerable in Eamonns output. Round
2 tested the voicing of consonants, using the following three pairs
of words and non-words which differ only in voicing:
Word Non-word
boat [pt] poat
sh [v] vish
vet [ft] fet
Eamonn heard each word and non-word three times. His judge-
ment of these was almost always correct. He made only three
errors, and all were in response to the non-word fet which he
incorrectly accepted for the picture of the vet.
Pursuing this curious consistency, round 3 presented Eamonn
with sixteen pictures of words containing [f] or [v]. These were
accompanied by either the correct form, for example re, vest, or a
corresponding incorrect form which reversed the voicing of [f] and
[v], for example [va] (vire), [fst] (fest). This time Eamonns score
was down to 72 per cent, with all but one error involving judge-
ment of a non-word as correct. Of the sixteen non-words, eight
were misjudged.
As the items have become more demanding, requiring sensitiv-
ity to ne distinctions which had appeared problematic in his
output, Eamonns performance has deteriorated. This suggests
that he does have some problem in input. We might then ask
further questions about the source and effects of this problem.
First, does it occur in auditory discrimination or in word
recognition? Either way, can it account for the output problems
we have observed?
The judgement tasks provide some indirect evidence. That evi-
dence, though, looks bafingly contradictory. Eamonn appears to
have problems with certain sound distinctions, particularly the
distinction between voiced and voiceless consonants, and even
more specically, the [f][v] distinction. If a problem occurs with
particular distinctions and cuts across words, the implication is
that the problem is not with the words themselves, but with
auditory discrimination. The paradox is that although certain
auditory distinctions are especially vulnerable, they also appear to
be more vulnerable in some words than in others. Eamonns
responses to words show curious mismatches. He makes consist-
ent errors in judging fet for vet. But he makes no errors in judging
vish for sh. In order to make the sh judgement, he must perceive
[v] as different from [f]. If he can tell [v] from [f], he has the
perceptual capacity to tell fet from vet. His failure to do so, then,
must be due to confusion in his recognition of the word rather
than in discriminating it.
So, it looks as if Eamonns problem with vet is on the one hand
due to general difculties in distinguishing [v] from [f], and on the
other hand due to vagueness in his representation of the specic
word vet. Our ne-grained analysis of Eamonns input appears to
have ended up in a cul de sac. Or has it? Are our two conclusions
necessarily in conict? Or can we reconcile them?
They can be reconciled, if we assume that one component of
processing may interact with another. Suppose Eamonn does have
subtle difculties in auditory discrimination, and may even have
had more extensive difculties at an earlier age. These will have
limited the information available to be stored, and may have
affected the phonological representations of words he has stored to
varying degrees. The point here is that difculties in one compo-
nent of input processing have reached into another component.
If we take this still further, we might hypothesise that Eamonns
input difculties have reached into his output and are enough to
account for it. This is feasible, but hardly compatible with the
evidence of his output which we already have. If we ash back to
Eamonns difculties in nding words, and his phonological dis-
tortion of words, these go well beyond any difculties we might
predict on the basis of his subtle difculties in input. Indeed, we
have seen that he is well able to reject most of his own errors. This
means he perceives and recognises more about words than he is able to
produce. Explanation of his output, and its relation to his input
processing, must wait on further output investigations. Before
embarking on these, we turn to Michael to see how he fares in
input.
Michaels input tasks consisted of lexical decision, presented
with pictures and without pictures. All tasks used his ten test
words. In the easiest round, he was like Eamonn fed back his
own errors, these being the incorrect forms he produced in naming
the ten items, along with the target forms. Here, his judgement
was 100 per cent. The other rounds of lexical decision were
tougher, because they used specially created non-words which
were much closer to the target than his own errors. The ten real
words were mixed up with:
A. Ten non-words created by copying a sound in the real word
e.g. [sklek]
B. Ten non-words created by swapping two sounds in the real
word e.g. [stlek]
C. Ten more real words which were added to keep equal
numbers of words and non-words
Now, Michaels score dropped. He still judged real words perfectly.
But like Eamonn, he came down with non-words to, on average,
70 per cent in judging non-words with pictures and 60 per cent
without pictures. His worst performance was with non-words in
which a sound was copied (type A above) when these were pres-
ented without a picture. Here, he scored 50 per cent, which means
he was performing at chance with these items. In Michaels case,
we have proof that his difculties are abnormal, because the same
tasks were carried out with the age- and vocabulary-matched
groups of children. Even the younger, vocabulary-matched
children performed way better than Michael on the non-words,
their worst average score being 89.17 per cent on judgement of
type A non-words with pictures. It seems that, like Eamonn and
unlike all the normally developing children, Michael was unsure
about non-words despite being sure about very similar real words.
Again we can ask whether Michaels difculties with judge-
ment arise in auditory discrimination or in his word repre-
sentations. We know from earlier testing (the Bridgeman &
Snowling test) that Michael has some auditory discrimination
problems, since he could not always distinguish sequences of
consonants. However, we have reason to think that these prob-
lems are not responsible for his misjudgements. First, his perfect
judgement of the real words suggests that he is making some
discrimination between these and the non-words on which he
falters. Even more telling is the apparent relation between his
judgement and naming of particular words. Most of his incorrect
judgements occurred with words he had also named incorrectly.
For example, he accepted three non-words for binoculars which he
also named incorrectly, whereas he rejected all four non-words for
crocodile which he named correctly. The fact that he makes errors
on the same words in different tasks suggests that he is unsure
about the form of these words.
We are back where we left Eamonn with evidence of some
limitation in auditory discrimination, but further evidence of limi-
tations on specic lexical items. Before we explore this outcome
further, we need to know more about how it tallies with evidence
of the childrens output.
Output processing
Words can only be produced if they have been discriminated
and stored. So, if there are constraints on these input processes, we
would expect constraints on output as well. Having set out from
observations of Eamonns and Michaels output, we already know
they have problems. We have also seen some connections between
their errors in input and output. Both had difculty in making
judgements about words they had misproduced recall Eamonns
problem rejecting non-words for vet and Michaels problem reject-
ing non-words for binoculars, words which the children also had
problems naming. Does this mean their output problems are just a
spin-off from their input? Or do they have further difculties in
output in going from semantics to the phonological representa-
tion, or from the phonological representation to speech?
We already have some reason to think they do. Although their
judgement sometimes matches up with their production of words,
both children make far fewer errors in judgement than in produc-
tion. Their judgement errors involve ne distinctions such as [f] vs
[v], or between sequences of sounds [n][l] vs [l][n]. Their phono-
logical errors in naming and spontaneous speech sometimes in-
volve rather more gross deviations such as:
Eamonn: very ; [vi] [fi]
Michael:
# #
Michael: binoculars ; [ nknz] [ nkmlz]
This accounts for the childrens successful rejection of almost all
their own errors when these are fed back in judgement tasks.
Their output errors, then, appear to go beyond their input
difculties. This could be because their output has been held back
by difculties in input which occurred earlier in their develop-
ment. Perhaps some constraints on input have now been resolved,
but have left their mark on output. Current output representations
are then relics of earlier limitations on input and have not bene-
tted from improved input representations. In this case, the
childrens output is a product of their difculties in input process-
ing. If the deviations in output are relics of past input, they should
occur only when the child produces a word from his store of words.
When he produces a word without accessing a stored representa-
tion, production should be as good as input processing allows. We
have tools for checking this out. Repetition does not require
access to stored representations. In the case of non-word repeti-
tion, there is no representation to access. How do Eamonn and
Michael fare in these tasks?
Eamonn was rst asked to repeat a set of words he had named.
His score on repetition matched his score on naming: both 30 per
cent. All but one of the errors he made in repetition occurred on
the same items as in naming. But the errors were not necessarily
the same:
Target Naming Repetition

# #
guitar [t t] [t t] same
# # # #
kangaroo [ k wu] [ km wu] different
# #
umbrella [ bwl] [m bw] different
The same thing happened in a further test of repetition. This test

comprised a set of words which had proved difcult in production,
and a corresponding set of non-words which altered the vowel and
in some cases also the stress of the word. These are illustrated by
the following examples:
Matched non-words
Words Same stress Different stress

#
after [ ift] eefter
#
garden [ 'idn] geeden
# #
guitar [' t] geter [ 't] gerter
# #
balloon [b ln] bilawn [ bln] bawlin
This repetition test was carried out twice. Eamonns scores were
similar for words and non-words, both 36 per cent rst time
round, 60 per cent for words and 52 per cent for non-words
second time round. He sometimes responded to words and non-
words in the same way
Target Eamonn
#
Word Riddler [ w'l]
# #
Non-word [ rdl] raddler [ w'l]
but did not always do so:
Target Eamonn
#
Word guitar [d t]
#
Non-word [' t] geter correct
Word garden correct
# #
Non-word [ 'idn] geeden [ di'n]
What can we infer from these repetition tasks? First, Eamonn

makes the same sort of errors in repetition as in naming and
spontaneous output. And he makes the same sort of errors
whether he is repeating words or non-words. These ndings point
to problems in production even when he need not or cannot refer
to a stored representation. The implication is that Eamonn has
production difculties which are not due to inadequate representa-
tions resulting from earlier input difculties. As with naming,
Eamonns errors in repetition exceed his errors in judgement. He
makes errors in repetition of the sort he rejects in judgement. This
again indicates problems in output which go beyond his input
problems.
At what point in output might these errors arise? One source of
evidence is a comparison between naming, where a word must be
accessed from a store, and repetition of non-words, where there is
no stored word to access. Eamonn makes similar numbers and
types of errors whether or not a representation is accessed. The
parallels in the errors he makes on these different production tasks
imply that the errors arise in a component of processing common
to word and non-word production. The one component common
to naming, repetition of words and repetition of non-words, is
articulatory planning. In line with difculties in articulatory
planning, Eamonn seems to have problems with certain target
sounds whatever the production task, for example in Riddler,
afternoon, number.
At this point, we strike a paradox akin to the one which
emerged in his input. While certain sounds or sound sequences
appear to be vulnerable, they appear to be more vulnerable in
some words than others. For example, while /f/ is consistently
omitted in afternoon, it is consistently correct in after. While some
words are almost always produced in the same way, for example
vet, very, guitar, others are produced inconsistently. This variation
between words suggests that some words are more stably repre-
sented than others. But this indicates a problem in representations
rather than in articulatory planning. Again, we have come up
with evidence for constraints on articulatory planning of words
and vagueness in representations of words. Putting this together
with the analysis of Eamonns input, we appear to have identied
constraints on all the components of phonological processing we
have considered: on speech discrimination, on articulatory plann-
ing and on the representations to which these connect.
Turning to Michael, output investigations lead to a similar
conclusion. Michael was given his set of twenty words and twenty
non-words to repeat. His performance was weak: only 20 per cent
of the words and 10 per cent of the non-words were repeated
correctly. The scale of his difculty is evident when we compare his
performance with the performance of the other children in the
study:
Percentage of correctly repeated
Real words Non-words
Michael 20.00 10.00

Vocabulary-matched children 99.17 66.25
Age-matched children 100.00 86.00
Like Eamonn, he shows problems with repetition that go beyond

the problems he shows in input. He repeats correctly even fewer
items than he is able to discriminate in lexical decision tasks. Since
he makes errors in repeating non-words as well as words, we
might again infer that there are difculties in a component of
processing common to these: articulatory planning. On the other
hand, as with Eamonn, Michaels production of words varies. He
produces certain words correctly even though these appear no
simpler than words he produces incorrectly. This again suggests
that some words are more precisely represented than other words,
which points to problems in word representations. Michael too
appears to have constraints on auditory and articulatory process-
ing which cut across words and constraints on the representation
of particular words.
Looking ahead
Eamonns and Michaels difculties, it seems, are quite diverse.
They appear to span all aspects of phonological processing beyond
hearing in input and before articulatory execution in output,
including the representation of word phonologies.
Such children tell us that impairment may extend over different
components of word processing. The occurrence of more diverse
difculties with phonology is perhaps unsurprising. The nature of
the connections between those more diverse difculties is harder to
fathom. The picture which has been emerging is one of each
component of phonological processing being constrained in its
own right, but also constraining other components of processing
connected to it. This may lead us to see components of processing
in a fresh light. Rather than being separate and self-contained
INPUT OUTPUT
features of speech
Identification and storage Accessing of word

of word phonology phonology
Figure 6.3 Possible forward cascade in phonological processing
boxes, they may be viewed as collections of information which

cascade forwards and perhaps even backwards. This means that
a constraint on auditory discrimination may weaken phonological
representations which may undermine articulatory planning, as
shown in gure 6.3.
Conversely, a constraint on articulatory planning may weaken
phonological representations which may undermine auditory dis-
crimination though this backward domino effect is likely to be
much weaker than the forward effect, as shown in gure 6.4. Such
cascading of information is taken on board in connectionist
theories of processing.
For Eamonn and Michael, then, the primary impairment may
lie in the ow of information from what they hear to the represen-
tations they store and the articulatory planning of those represen-
tations, and possibly back again. In contrast, the difculties of
children with phonological impairment appeared to lie in very
specic aspects of articulatory planning, with no observable effects
on other components of processing (see previous chapter). This
INPUT OUTPUT
features of speech
Identification and storage Accessing of word

of word phonology phonology
Figure 6.4 Possible backward cascade in phonological processing
suggests that the reverberations from impairment at a very late

point in output processing are negligible or non-existent.
But if this rather hazy notion of cascading information is right,
where and why does the cascade stop? Why does an impairment
which cuts across phonological processing not feed forward, and
even backward, to semantic representations, which are connected
to phonological representations, as shown in gure 6.5?
The answer may be that it does. On present evidence, we know
very little about Eamonns and Michaels semantic processing. We
established at the outset that their semantics was relatively unim-
paired. But semantic errors were noted in both children. It could be
that these errors are simply a spin-off from the problems with
phonological processing already identied. Perhaps when Eamonn
and Michael cant access the phonology they are looking for, they
go for a meaning close to their target or a very general meaning for
which they can nd the phonology. So, when Eamonn cant nd
horse, he lands on another animal, cat; when he cant nd pot, he
falls back on thing until pot comes to him. But there is another
INPUT OUTPUT
Storage of word
phonology
Identification and storage Selection of word

of word semantics semantics
Accessing of word
phonology
Figure 6.5 Possible forward and backward cascade in lexical processing
possibility. Recall that a words phonology acts as a bootstrap to

meaning, and a handle on a particular package of meaning (see
chapter 2). If words phonologies are weak for a child, perhaps this
weakens the childs phonological bootstrapping and his handle on
meaning. While we have reason to think that Eamonn and
Michael are well able to make sense of the world, and do so in
similar ways to other children, this doesnt rule out the possibility
that their packaging of the world into word concepts is affected by
their difculties with word phonology.
We nd that this is more than just a hypothetical possibility
when we turn to childrens difculties with the words from which
sentences are grown.
Note: Eamonn was originally presented in Chiat & Hunt 1993. The
presentation of Michael is based on the data in Constable, Stackhouse &
Wells 1997 and draws heavily on the authors discussion of these data,
for which I am indebted.
MMMM
Part II
Grappling with verb
structure
MMMM
7 Translating events
When children are having difculties with sentences, the most

immediate sign of those difculties will probably be the way they
struggle in everyday conversations. They dont necessarily have
problems taking turns as conversational partners. But as they take
their turns, they falter, producing utterances which are halting, or
limited, or mixed up. Witness Eamonns attempts to explain that
he has been unwell in the night and that his teacher may have to
call his mother:
I not feel well today sick the school why my mum gonna call
Mary gonna call me Anna gonna call m ['d] call my mummy I
sick . . . Im getting sick in [] school Anna gonna t phone my
mum out phone my mummy out.
Here, we see Eamonn start a sentence, stall, have another go, or

change direction and start again. Each stab he makes at express-
ing himself consists of a run of words which comes out uently
even though it may be incomplete. Other children produce se-
quences of words which may be clearer but are more disjointed.
Take 8-year-old Richards attempts to explain what happens to the
lambs on the farm where he lives:
R: Um girls, keep, boys, not keep.
SC: What do you do with the boys?
R: [tsl ts ] on oth people.
SC: Sell them? So why do you keep the girls?
R: Breeding.
A closer look at these childrens utterances shows that some of

113
114 Grappling with verb structure
them are organised in slightly odd ways. As well as leaving out
certain words, they may combine words which dont go together,
or which dont quite get across the meaning they seem to be
aiming for, as with Eamonns
. . . sick the school . . .
. . . phone my mummy out
or Richards
[tsl ts ] on oth people
The order of the words may also go awry, as when Eamonn tries to
explain how his teachers know that he has been unwell:
Because he write my mum because my mum write in my book
Does he mean . . . write my mum or my mum write . . .?
As a consequence of these childrens difculties, the listener is
sometimes unsure what these children are trying to say. Everyday
conversations may suggest that their understanding is precarious
as well. When they respond to other peoples utterances, it may
become obvious that they have missed some detail of what the
speaker was saying, or have got the wrong end of the stick entirely.
Conversations with Eamonn can lead into such cul-de-sacs or
tangles:
SC: Do you like school?
E: Like to go home.
SC: Why do you prefer to go home?
E: Because I dunno.
SC: You like home better than school?
E: No I love home.
Signs of difculties with sentences in everyday conversations
might be conrmed by formal assessments. The child might come
down on something like the Test for the Reception of Grammar
(Bishop 1983), where they must select one out of four pictures to
match a spoken sentence, for example
The girl is sitting on the table
The man is eating the apple
The girl is pushing the horse
The man is chasing the dog
The elephant is pushed by the boy
Translating events 115
If they are below par on this test, it looks as if they have difculties
in understanding sentences. A wide-ranging assessment such as
the Clinical Evaluation of Language Fundamentals Pre-school
(Wiig, Secord & Semel 1992) may reveal problems in the Sentence
Structure section, where the child has to point to the right picture
for sentences such as
The bear is in the wagon

The man opened the door
The girl took some owers to her mother
The woman caught a big sh
The boy was followed by his cat
The girl is being pushed by the boy
This would also point to problems with comprehension. Or prob-

lems might surface in the Recalling Sentences section, where the
child has to repeat back sentences such as
I can carry it
You can wear this old coat
I am putting ketchup and mustard on my hamburger
Here, difculties could be arising in production.

Formal procedures for analysing childrens spontaneous ut-
terances might conrm that they are not producing the sort of
structures expected at their age. Examples of such procedures are
the Language Assessment, Remediation and Screening Procedure
LARSP (Crystal, Fletcher & Garman 1976) or the Bloom & Lahey
(1978) or Systematic Analysis of Language Transcripts SALT
(Miller & Chapman 19827). (For further examples and details of
clinical assessments, see Lees & Urwin 1997.)
What is it that stops these children dealing with sentences as
other children do? Are they having difculty with the sorts of
meanings which sentences express? Or do they have the meanings,
but struggle with the words which are needed to create sentences?
Or do they have the words, but struggle when it comes to combin-
ing them? Whatever their difculties, do they arise in output only,
or do they show up in input as well?
To explore these possibilities, we need to know more about
sentences and the sort of words that are needed to grow them.
From scenes to events
Semantically, a sentence encodes a state of affairs, or event.
But the relationship between event and sentence is not a simple
one. Sentences relate to events in a rather special way.
At any moment within the ow of our experience, we tune in to
certain aspects of that experience. Looking at a scene, the sorts of
things we might notice are the presence of an object; something
about the object such as the state it is in; the position of an object
relative to another object; the way it is moving; and so on. The
scene may call to mind some other scene, which we picture in our
minds eye, and again, we attend to certain aspects of the pictured
scene. Or we may be entirely oblivious to the scene in the outside
world, attending only to something in our minds eye. Sentences
enable us to express and communicate what we experience in the
outside world and in our minds eye. But they can only do this
indirectly. We cannot take the whole conglomeration of our ex-
perience at successive moments and encode it all in one sentence.
This is because sentences are highly focused in their meaning.
They pick out certain aspects from the ow of experience. To be
more precise, they pick out a particular event from the ow of
experience. To be even more precise, they pick out a particular
angle on a particular event.
To illustrate this crucial point, take a look at this picture of Fred,
Bob and an old car.
A picture, being a static representation, already limits the informa-
tion we receive about a scene compared with a video or with direct
experience. Yet even the frozen picture includes more than we can
encode in a sentence. Looking at the picture, we might notice what
Fred is doing, what Bob is doing, what is going on between them,
what Fred or Bob or the car is like. If we were to encode these
observations, we might come up with sentences such as:
Fred handed over the money
Fred paid for the car
Bob took the money
Fred bought the car from Bob
Bob sold the car to Fred
The car looked old
and so on. Each of these sentences picks out only certain aspects of
the event in the picture. And that is what sentences necessarily do:
they act like a zoom lens (Fisher, Hall, Rakowitz & Gleitman
1994) focusing in on an event from a particular perspective.
From events to verb structures

Since the semantic heart of the sentence is an event, the
sentence hinges on the word or words which encode that event. In
the examples above, these are words such as buy and sell and
take. It is these words syntactically the verbs which give rise
to the zoom lens effect. If we change the verb, we change the
camera angle on the scene. This may change which participants
in the scene are in view, and are included with the event. It may
also change what it is about the participants that is in the fore-
ground. Pay focuses on the transfer of the money, and the goal of
that transfer. Hand and take both focus on the transfer of the
car, but they take different perspectives on that transfer. Hand
takes the perspective of the starting-point, or source, of the trans-
fer (Fred). Take is from the perspective of the end-point, or goal,
of the transfer (Bob). Look has an entirely different focus. It does
not capture any relationship between participants in the event.
Instead, it focuses on just one of the participants (the car), and says
something about its state.
In some cases, the event is encapsulated in the verb alone. This
is true for buy and sell. Some events, though, are encoded by a
verb along with a further sentence constituent. Take the case of
look old. Here, the verb on its own has a rather general meaning:
it refers to the visual state of an entity, but not to any specic state.
Syntactically, it is the adjectival constituent following the verb
(old) which species the particular state.
This illustrates a pattern which commonly occurs in English.
States and changes of state may be encoded by a general state
verb, such as look, be, seem, feel, get, which combines with
an Adjective Phrase specifying the state:
Verb Adjective Phrase
be sad
look quite ill
seem angry
feel great
get very wet
On the other hand, many changes of state can be expressed by a
single verb which incorporates the particular state: tire, age,
cool, sadden, freeze, melt, break all express a particular
change of state.
Some of these states can be encoded either by a single verb or by
a general state verb with a separate Adjective Phrase:
Single verb Verb + Adjective Phrase
tire get tired
cool get cold
age get old
sadden get sad
A pair of sentences such as:
The soup cooled
and
The soup got cool
express very similar states though you may discern a subtle

difference between them.
Movements, like states, may be expressed by a specic verb or
by a general verb which is further specied. In this case, it is the
direction of movement which is specied separately. Direction is
commonly expressed by a curious category of words known as
particles. These include prepositions such as up, down, in, on,
out, off, away:
Verb Particle
go up
run out
jump in
move off
throw away
However, the verb and particle do not always express a movement
and a direction. Often they combine to express a specic event as a
whole. For example,
give up/sell out/throw up/put up/pick up/go off/put off
can all refer to a very specic event, as in

He gave up his job
The peaches went off
Again, there may be a single verb expressing the same or a similar

event:
He resigned his job
The peaches rotted
Similarly, some movement events can be expressed by either a

single verb or a verb with a particle:
Verb Verb + particle
leave go away
enter come in
insert put in
lower put down
raise hold up
In all these examples, we see that events are expressed by verbs,
either on their own, or together with an Adjective Phrase or
particle. If the semantic heart of a sentence is the event, the syntactic
heart must be the verb. To understand and produce sentences in a
language, we must know verbs. Knowing verbs means knowing
their phonological forms, such as /ba/ (buy) or /sl/ (sell), and
knowing the events which those forms encode, such as exchange
of object for money.
From participants in events to argument

structures
But knowing verbs involves more than the connection between
their form and meaning. This is because events, and the verbs
which encode them, also involve participants the who and
what play a role in the event. The verb, in taking a particular
perspective on an event, decides which participants are in view.
This determines which participants can or must be included with
the verb. For example, a verb which focuses on movement entails
at least that which undergoes the movement. A verb focusing on
possession entails at least that which possesses and that which is
possessed. Returning to our original example, Fred bought the old
car from Bob, the change of possession event involved three
participants: the original possessor Fred, the ultimate possessor
Bob, and the thing changing possession the old car.
The participants entailed by a verb are known technically as its
arguments. Arguments typically refer to things, as with the old
car, or to people, as with Fred and Bob. They play particular
roles in events, known technically as thematic roles. The them-
atic role of an argument identies what the argument is doing in
the event. One such role is the theme. This refers to the argument
which undergoes a state, location or possession. In the following
sentences, the car is a theme undergoing:
State: The car looked quite old
Sue cleaned the car
Location: The car stood in the garage
Fred put the car in the driveway
Possession: That car belongs to Mary
Jane bought the car from Mary
The agent refers to an argument which causes or brings about an
event. In the following sentences, Alice is an agent bringing
about:
State: Alice broke the glass
Location: Alice took the car to the station
Possession: Alice sent the letter to the doctor
The source and the goal refer to the arguments which are the
starting- and endpoints of a change. In the change of location
The bus goes from the station to town
the station is the source and town is the goal. In the changes of
possession
Fred sold the car to Bob

Bob bought the car from Fred
Bob received the message from Fred
Fred is the source and Bob is the goal. Stimulus and experien-
cer are sometimes used to refer to the arguments involved in a
mental or psychological state. In the following sentences, Sam is
the experiencer, with jokes as the stimulus:
Sam enjoys jokes
Jokes amuse Sam
These labels for thematic roles by no means exhaust the different
ways in which arguments participate in events, as we would nd if
we delved further into the vast range of events we can express. But
they enable us to distinguish certain of the roles involved in
certain types of events.
Since sentences express events and events involve arguments,
sentences require not just the verb expressing the event but also
the arguments involved. These arguments take the form of consti-
tuents which combine with the verb. And they combine in ways
which indicate the role each argument is playing in relation to the
verb. In English, word order is the main device for marking the
relation between arguments and verbs. Hence the difference be-
tween two sentences containing the same verb and the same
arguments but in a different order:
The dog tripped up the postman
and
The postman tripped up the dog
Both focus the same type of event the causation of a change of

location and the same characters the dog and the postman. But
the roles played by these two characters change while the dog is
the agent and the postman is the theme in one, their roles are
reversed in the other. In order to understand and produce senten-
ces, then, we must know not only the verbs to express the events
and the arguments they entail. We must also know where the
arguments go in relation to the verb.
The position of an argument depends on the verb, and the
perspective it takes on the event. Human beings show certain
strong biases in the perspectives they take. These are reected in
the way verbs encode events and where they place their argu-
ments. For example, we have a strong tendency to focus on
causation, with vast numbers of verbs encoding the causation of
different states of affairs. Verbs which take this perspective on an
event always involve a causer or agent, and this agent always
occurs in the subject position, which precedes the verb in English:
State: Alice broke the glass
Location: Alice took the car to the station
Possession: Alice sent the letter to the doctor
What happens when an event involves more than one causer?
For example, where there is a change of possession, it may be that
two parties in the exchange play roles which are distinct but
which are both active in bringing it about. When we come to
encode such an event, we must construe it from the perspective of
one or other. Often, the language offers us both possibilities, giving
rise to pairs such as give/take, sell/buy, lend/borrow. Give,
sell and lend take the perspective of the source, placing the
source in the subject position. This focuses on the source as the
agent bringing about the event. Conversely, take, buy and
borrow focus on the goal as the agent of the event. They place the
goal in the subject slot.
It seems that the subject of a verb encodes an agent wherever
the verb entails causation. This is why we typically think of the
subject as being the doer. However, this is not necessarily the
case. Despite our strong inclination to notice causation and caus-
ers, we are quite capable of seeing events from perspectives which
do not involve causation and do not entail an agent. Where a
verbs focus is not causation, the subject slot will be lled by a
different role. We have already met examples where the theme
occupies the subject position:
State: The car looked quite old
Location: The car stood in the garage
Possession: The car belongs to Mary
In these examples, where there is no agent, the subject slot is lled
by the entity which undergoes the event. In contrast, where the
verb took the perspective of the agent and placed this in subject
position, the same theme occurred after the verb, in the position of
the direct object:
State: Sue cleaned the car
Location: Fred put the car in the driveway
Possession: Jane bought the car from Mary
Some verbs can take more than one perspective on an event
and, correspondingly, more than one argument structure. For
example, melt, break, cool, move, drop and many other
verbs can express the causation of a state of affairs. They then
involve an agent in the subject slot and theme in the direct object
slot:
Sue broke the car
Sue moved the car
But they can also refer to the same state of affairs without causa-
tion and without an agent. They then involve just a theme, and
this takes up the subject slot:
The car broke
The car moved
Other verbs take only one of these perspectives. Fall, for
example, refers to the movement of its theme:
The handle fell off
It cannot refer to the causation of that movement by an agent:
*The child fell the handle off
Conversely, pull refers to the agent causing the theme to move:
The child pulled the handle off
but cannot generally refer just to the movement of the theme:
?The handle pulled off
Knowledge of these verbs thus includes knowledge of the perspec-
tives they allow, and the arguments which come into those per-
spectives.
We have now revealed fully the core of the sentence. Semanti-
cally, it is an event which is viewed from a particular perspective
entailing certain participants, or arguments, playing certain roles.
Syntactically, the event maps onto a verb, or verb + Adjective
Phrase, or verb + particle. The arguments map onto constituents
which combine with the verb. Their position in relation to the verb
depends on the perspective the verb takes on the event and its
participants.
Sentences, then, consist of a complex and indirect connection
between events we perceive or imagine, and verbargument struc-
tures. In order to interpret and express events in sentences, we
must conceive events from the sorts of perspectives that verbs take
on events. We must know the connections between particular
perspectives on events and particular verb forms. And we must
know the connections between participant roles in events and
argument positions round the verb.
For children discovering their language, the relation between
events and verbargument structures is a key target. They must
discover the verbs of the language: their forms, and the events on
which they focus. They must at the same time discover the argu-
ments entailed by the verb. And they must discover where those
arguments go in relation to the verb.
From verbargument structure to rhythmic

structure
When the verb and arguments come together to encode an
event, the phonological forms of the verb and arguments join to
form a larger phonological unit. Those phonological forms are not
simply strung together one after the other, each equally promi-
nent. If you try uttering the sentence
you will nd that the sentence has a rhythm in which some words
are more prominent than others. The rhythm or stress pattern of
the sentence is directly related to the way in which the words are
combined. When words combine in a constituent, the last word in
a constituent generally carries the greatest stress in the constitu-
ent. When constituents combine in a sentence, the most stressed
word in the last constituent carries the greatest stress in the
sentence. It is partly this increased stress at the end of constituents
and at the end of the sentence which gives rise to the up and
down rhythm of the sentence.
We can see how this operates in our example:
Fred is a constituent, and since Fred is the only word in the
constituent, it carries the constituent stress. This stress can be
represented by a cross above the word:
x
The car is a constituent, and here it is the nal word car which
carries the constituent stress:
x x
From Bob is a constituent, in which Bob carries the constituent
stress:
x x x
Bob, being the nal word in the nal constituent of the sentence,
also carries the sentence stress, for which we can add another
cross:
x
x x x
The verb bought, which does not end a constituent, is less
stressed than Fred, car and Bob. So, we end up with a stress
pattern with Bob carrying the greatest stress, and Fred and car
carrying greater stress than the other words. This shows the
prominence of certain words in the rhythm of the sentence. You
might check this out by uttering the sentence as naturally as
possible and tapping out its rhythm.
Because nouns frequently occur at the end of a Noun Phrase,
and sentences frequently end with a Noun Phrase, nouns often
carry constituent and sentence stress. Verbs are much less likely to
occur at the end of a constituent or sentence, because they are
typically followed by NP arguments. Verbs therefore tend to be less
prominent than nouns. If you go back to the examples we have
considered and say them out loud, you might notice the promi-
nence of the nouns compared with the verb in the rhythm of the
sentence.
Sentences, then, connect a verbargument structure to a
rhythmic structure in which certain words weigh more than
others. In dealing with sentences, we must be sensitive to the way
words are stressed when they come together in a sentence.
Sentences: connections between events,

verbargument structures and rhythmic
structures
We have now seen that a sentence is a rhythmic unit, made up
of the phonological forms of words which receive varying promi-
nence. The core of that rhythmic unit is the verb and the argu-
ments organised around it. These encode an event from a particu-
lar perspective, and the participants in that event.
The route between scenes as we observe or imagine them and
sentences is not a simple one. Scenes do not map directly onto
sentences. If children are to understand and talk about events,
they must discover the complex connections between events, and
rhythmic structures containing verbs and arguments.
8 Growing verb structures
Children set out with two sources of information about sentences:

the scenes which they observe and participate in, and the sound of
the utterances they hear. We have just seen that the connection
between scene and sound is an indirect one. The sound consists of
word forms organised in a rhythmic structure, and this relates to a
particular event involving particular participants. In discovering
sentences, children must somehow discover which components of
the sound relate to which aspects of the scene. This means ident-
ifying verb forms and the particular events which map onto those
forms. It also means identifying the way arguments are organised
around the verb and the event roles which map onto those argu-
ments.
Most children appear to meet the challenges posed by language
with ease and speed. By 16 months, they will often respond appro-
priately to requests such as
Give Jane the cookie
Show Jane the cookie
suggesting that they understand the verb and the roles of the
arguments (Huttenlocher 1974). This is backed up by evidence of
what they do when they are given peculiar requests which they
are highly unlikely to have heard before, and which therefore call
for active interpretation of the verb and argument roles:
Smell the dolly
Kiss the truck
Presented with bizarre instructions of this sort, children of 1624

127
months carry out a fair proportion correctly (Sachs & Truswell
1978). Where they do, they have clearly understood the verb and
the role of its object. Whether they have also understood the
importance of word order is a moot point. This can only be proven
by asking them to act out contrasting pairs such as
Make the car push the cow
and
Make the cow push the car
Few 18-month-olds will readily comply with such instructions to

show us whether they do or dont understand them. Their likely
response is to ignore the request, be distracted by something else,
or wander off, leaving us none the wiser about their understand-
ing. To get round the hazards of checking out comprehension
directly, we need more subtle probes which catch them unawares.
This is what the preferential looking technique can do
(Hirsh-Pasek & Golinkoff 1996). With this technique, the infant is
seated on a parents lap, between two video screens showing
different scenes involving the same two TV characters. On one,
Cookie Monster is tickling Big Bird. On the other, Big Bird is
tickling Cookie Monster. The child hears a tape playing . . .
Look! Cookie Monster is tickling Big Bird! Where is Cookie Monster
tickling Big Bird?
The question is whether children look longer at the screen which

matches the utterance they hear. If they do, we can infer that they
have understood how word order determines the roles of the
participants in the event. When a group of children around 17-
months-old were tested in this way, the majority did look longer
at the screen matching the utterance. It seems they understood
who was tickling who.
This nding comes as little surprise when we see what infants
do with their own utterances around this age. Between 18 and 24
months, most move beyond single words and start producing their
rst word combinations. The range of combinations they use is at
rst limited, and develops slowly. But by the time they have
Growing verb structures 129
produced around 100 different word combinations typically at
about 22 months their word combinations take off (see Ingram
1989).
Even before they actually produce word combinations, children
give some evidence of communicating events or relationships.
Although the majority of their early words refer to things and
people, not all do. A small proportion of their words pick out states
such as broken, possession such as mine, location or change of
location such as there, down, gone. These words picking out
relations in which things and people participate are the obvious
seeds of verbargument structures. But childrens use of thing
and people words at this stage also anticipates their development
of verbargument structures. Often, these words are used not
simply as labels for the thing or person, but to indicate their
relation to another entity. Take the example of the child who picks
up her mothers slipper and says Mommy. Mommy, and when
asked what it is, says Slipper (Bloom 1970). The child is surely
not using Mommy as a name for the slipper. Much more plaus-
ibly, she is pointing out a relation between an individual,
Mommy, and the slipper. Such a relation is just the sort that
verbargument structures serve to encode relations which later
become specied as possession or location.
These relations become more explicit in the childs utterances
as more relational words are used and as words are combined. The
childs early combinations look like very limited verbargument
structures focusing on participants in events and their relations.
Strikingly, children seem to mark relations between participants
appropriately from a very early stage. The verbargument struc-
tures they use may be incomplete. But they rarely show errors. In
general, their verbs match the perspective the adult language
takes on events, and they order arguments around the verb as the
adult language would order them.
Take the utterances produced by Alison at 2;3. Alison uses a
limited range of word combinations, and the structure of the
combinations she uses is very limited. Often, her utterances ident-
ify just a thing or person, such as Teddy, Shoe, Paul. Often,
they identify just an event or relation:
Drop (after penny dropped)
Fall out (after money falls out of purse)
Hiding (when teddy is in cupboard)
Open (trying to open purse)
Off (taking socks off doll)
Where she does combine events and participants, though, the

event and the participant are ordered as they would be in adult
English:
Agent Event Theme: Ali do it ( = open door)
Michelle do it ( = put doll
in bed)
Agent Event Goal: My daddy go home
Event Theme: Wash teddy
Open it
Push me
Got some
Event Goal: Go in there
Theme State: Teddy wake up
Experiencer
Psychological state Stimulus: I [wan] money
Compare Paul, at exactly the same age. Paul has much more
developed word combinations:
Me play [] puzzles. Me want play [] puzzles. Me want [] look at this
. . . Lets do some drawing . . . My cut it out. Me want to cut it out now.
Add to this sample the following utterances Paul produced ve

months later:
You dont like the skin? . . . I dont like [] skin. I dont like it in in in
in my water.
Within these samples, both Alison and Paul make reference to

psychological states (want and like). These examples of psycho-
logical states provide particularly good evidence that they have
discovered the perspective English takes on certain events and the
way this is reected in verbargument structure. The verbs they
have used take the perspective of the experiencer of the psycho-
logical state, placing this before the verb, with the stimulus follow-
ing the verb. Alison produces one example, and this shows the
correct mapping of the experiencer and stimulus:
I [wan] money
Paul produces more varied and sophisticated examples, again with

the correct mapping:
Me want . . .
You dont like . . .
To appreciate fully the signicance of the mapping the children

have made, notice that it is different from the mapping which must
be made by children acquiring certain other languages. In Italian
and Spanish, the psychological state of liking is expressed from the
perspective of the stimulus rather than the experiencer. The every-
day verb for expressing this state is akin to please rather than
like. I like . . . therefore translates literally as To me pleases
. . .. Children acquiring these languages will therefore acquire a
different verbargument structure from that which Paul has used
correctly with like in English (see Bowerman 1996 and Slobin
1996 for further discussion and evidence).
In Turkish, relations between the verb and its arguments do not
depend on word order, as in English, but on word endings, or
inections, attached to nouns. Children acquiring Turkish use
these inections consistently and productively by the age of 2;0
(Aksu-Koc & Slobin 1985). While their utterances may be short
and simple in the early stages, they still convey verbargument
relations through their virtually error-free use of inections. In
Turkish, word order serves to convey the relative newness or
importance of information in the sentence rather than to mark
verbargument relations. Young Turkish children already order
the verb and its arguments for this purpose.
Evidently, even in their early, very limited word combinations,
children are on the road with verbargument structure. Whatever
language they are acquiring, they already convey events in ac-
cordance with the perspectives that language takes on events, and
with the devices it uses to mark relations between participants in
those events. Children continue apace along this road. Their
limited verbargument combinations become more varied and
more complete. Their range of verbs proliferates, so that they are
able to refer to a wide variety of events from the particular perspec-
tives of their language. At the same time, their verbargument
structures ll out, so that verbs are used with all the arguments
they require.
The point is well illustrated by the following sample of language
from Shel, taken at 2;6, while she was playing with miniatures.
Shels utterances already average four to ve words:
Cant shut that. Them get out and she little babys can go in there.
Cant get she in. Push him. Should I push him? . . . They go [] fall off.
Wheres [] other girls? Look, I found two . . . Can you bend her legs?
. . . Mummy, can I take my shoes off? You wanna wear my shoes? No,
think they too big. Hope theyre not too big for me . . .
Strikingly, these word combinations virtually always include

verbs. And these verbs encode an impressive variety of events:
Causation of change of state: shut
Change of location: get out, go in there,
fall off
Causation of change of location: push, bend, take off
Causation of change of possession: nd
Psychological state: want, think, hope
Shel also includes all the arguments required by these verbs, and
in the appropriate position for their role in the event.
Children vary considerably in their rate of development, and
many will not produce diverse and complete verbargument
structures as early as Shel. But the majority will get there some-
where between 2;6 and 4. The consistency of their verbargument
structures at this point is startling. They do make errors on occa-
sion. But on the whole, their sporadic deviations from adult verb
argument structure still follow the patterns of the adult language.
Far from being evidence of confusion about patterns of verb
argument structure, they provide evidence that the child has
discovered these patterns.
Consider, for example, the following errors produced by vari-
ous children:
2;6 Mummy, can you stay this open ( = make this stay open; keep
this open)
2;8 Daddy go me round ( = make me go round; spin me)
3;8 And the doggie had a head. And somebody fell it off ( = made it
fall off; broke it off)
4;2 He disappeared himself ( = made himself disappear; hid himself)
4;3 It always sweats me ( = makes me sweat; ?bakes me)
(examples from Bowerman 1982)
In each case, the childs verb is in the right semantic ball park. The
child is conveying the intended state (stay open, sweat) or location
(go round, fall off, disappear). The problem is that each of these
verbs is being used with one argument more than it allows. Each of
the verbs refers to a state or location, and entails a theme undergo-
ing that state or location. In adult English, the theme is the subject:
I go round
This stays open
It fell off
In the childrens utterances, this theme occurs after the verb, as

the direct object of the verb:
Daddy go me round
Can you stay this open
Somebody fell it off
So what has the child done with the subject slot? This is lled by
another argument:
Daddy go me round
Can you stay this open
Somebody fell it off
Intuitively, we take this argument to be an agent, bringing about

the event. There are a number of grounds for our assumption. One
is the context, which makes it likely that, for example, daddy is to
bring about the event of going round. Second, we know that
agents are always subjects, so if the child has created a new role for
the subject, it is likely to be an agent. Third, and most telling, we
know many verbs which can be used with an agent, expressing
causation of an event, or without an agent, expressing just the
event.
Agent Theme Theme
Daddy moved me I moved

Somebody broke the head off The head broke off
The childs errors therefore follow a familiar pattern, using a

verb which expresses just a change of state/location to express
causation of that state/location. It is the inclusion of the agent,
placed in the subject slot, which implies causation.
These are the sort of errors which occur in young childrens
language. They hardly warrant being called errors, since they
are not arbitrary. The children are not haphazardly adding an
argument or putting it in the wrong place. On the contrary, they
have discovered a pattern in the complex relation between events
and verbargument structures, and have systematically applied
this to verbs whose meaning and structure do not t that pattern.
These so-called errors are therefore not a sign of difculties in
determining the meaning relations between verb and arguments,
or the organisation of verb and arguments. Rather, they conrm
the childs sophisticated grasp of verbargument structures.
What is more, such errors are vanishingly rare. The utterances
exemplied above are a drop in the ocean of the childs utterances
which generally follow the adult model and use verbs with their
required argument structures (Pinker 1989).
Filtering verbargument structures from scenes

and speech
The speed at which children move from early words through
early word combinations to full and varied verbargument struc-
tures is impressive. But even more impressive, perhaps, is the fact
that they get there at all. Recall that the connections between
scenes and verbargument structures are indirect: scenes can be
interpreted in many different ways corresponding to many differ-
ent verbargument structures. To identify verbargument struc-
tures from the information provided by scenes and sound, the child
must
( identify which event and participants in a scene are being
focused by an utterance. The utterance may focus on a
directly observable event involving things and people in
the scene, for example, Im giving the ball to granny. But
this is not necessarily the case. The utterance may focus
on an event which is anticipated and so is not immediate-
ly observable, such as Give the ball to granny or Kates
coming home soon. Or it may focus on an invisible event
a perceptual or psychological state such as listening, or
wanting, or liking or feeling proud.
( identify the verb or verb structure which maps onto the
focused event: listen or give or pick up or feel proud.
Verbs may occasionally be heard in isolation, in utteran-
ces such as Stop, Watch, Wait. But they will much
more often be buried in a string of words from which they
must be segmented, as in Listen to that, Ill give you a
drink.
( identify the constituents of the sentence which map onto
the arguments in the event, and the positions of those
constituents in relation to the verb:
Youll listen to the story
Youll hear the story
Granny will tell you the story
The story will surprise you
Youll enjoy the story
Any one of these bits of information will help get to the other. If
you can pick out the arguments, this will focus you on the relevant
participants, which will draw your attention to the event in which
they are participating. So if you are the child in the following
scene:
and you hear an utterance containing the familiar words

Granny . . . ball . . . you
this will focus you on the event which currently involves you,
granny and the ball. You can then identify the event focused by
the verb, in this case, a change of possession event. However, this
information will not be enough to distinguish between different
change of possession events. If you also notice the order of the
arguments, this will lead you to a give rather than a take event.
But even this will not narrow your focus down to that of a single
verb for example give as opposed to hand.
Alternatively, if you know exactly what is being focused in the
scene, you will be in a position to make a good guess at the
meaning of the verb. For example, suppose you are in the above
scene, and you know grannys current focus is not just the ex-
change, but her role in bringing about that exchange. This will
enable you to narrow the likely reference of her verb from just any
change of possession event to one in which the source of the
exchange is the agent to a giving rather than taking sort of
verb.
Best of all, if you know the verb as adults do this immediately
directs you to the relevant event in the scene. The moment the
verb is heard and recognised as give or take or hand or listen
or tell or enjoy or whatever, you know exactly which event is
focused.
The problem is that children set out without any of this infor-
mation. To all appearances, they are caught in a vicious circle. To
know what is being focused in even the simplest scene, you need to
know the verb and its argument roles. But the childs task is
precisely to discover the verb and its argument roles. And to
discover the verb and its argument roles, you need to know what is
being focused. Children somehow break this apparently vicious
circle. The evidence is their successful acquisition of verbargu-
ment structures. How do they crack the connection between
scenes and sound to get at the specic meanings and the specic
forms of verbs and their arguments?
Bootstrapping
The answer, according to current theories, is that children are
equipped with bootstraps which give them a leg-up from one type
of information to another. Bootstrapping might start from any of
the information which children receive from scenes or utterances,
and lead them to new information about the way the two are
connected.
Suppose that children are extremely sensitive to what speakers
are focusing on at any moment in time. They notice whether the
speaker is focused on something moving, or moving in a certain
manner, or changing its state, or changing possession, and so on.
If they share the speakers focus on the scene, this will provide a
leg-up to the meaning of the verb used by the speaker. Once they
know what that verb picks out in the scene, they can work out
which participants are involved, and how they are involved. And
once they know the role of different participants, they can work
out where those roles go in relation to the verb. For example, the
child sees an adult pouring water into a glass. If the child shares
the adults focus on the moving water, he will expect the adult to
use a verb referring to movement, and focusing more on the thing
moving than the thing it moves into. He can then deduce that the
thing moving, which is most affected by the event, will be most
directly connected to the verb pour will be followed by the
water rather than the glass. If, on the other hand, the adults
focus is on the state of the glass and the child shares that focus, he
will expect the adult to use a verb referring to the change of state,
and focusing more on what changes state. In this case, he will
expect the glass to be closest to the verb ll will be followed by
the glass.
These are examples of semantic bootstrapping (Pinker
1989). The child is using information about the scene to get to the
argument structure of the verb describing the scene. It appears
that children do this from an early age. The evidence comes from
experiments in which children are shown events they have never
seen before, to nd out if the particular focus of the event inuen-
ces the structure they use to describe the event.
In one experiment (see Pinker 1989 and Gropen, Pinker, Hol-
lander & Goldberg 1991), children were introduced to the follow-
ing two events: a sponge being moved in a zigzag path towards a
cloth, the salient aspect of the event being the particular manner
in which the cloth moves; a sponge being moved towards a cloth
which then changes colour, the salient aspect of the event being
the change of state of the cloth. In each case, the children were
taught a novel verb for the event using a form which does not
require arguments: This is pilking. Children were then asked Can
you tell me what Im doing?. Where children did not provide a
clear direct object, they were given a further prompt Can you tell
me what Im pilking?. The question was whether the different
focus of each event would affect their response. Apparently it did.
Even a group of 3-year-olds were more likely to make the sponge
the object (pilking the sponge) when its movement was salient,
and more likely to make the cloth the object (pilking the cloth)
when its change of state was salient. This shows that they were
sensitive to which of the two objects was most affected in the
event, and knew that affected objects are linked to the slot directly
following the verb. It seems that even 3-year-olds can use seman-
tic information about the verb to get to its syntax.
This points to one possible way out of the vicious circle. If
children can nd the relevant focus in a scene when they hear a
new verb, and if they know where focused argument roles go, they
will be able to work out the argument structure of that verb. But
these are two big ifs. The experimental situation confronting the
child is a long shot from the real-life situation in which children
meet new verbs. In the experiment, children were shown a single
event, and one which involved an eye-catching effect either a
change of colour or a peculiar movement. The childs attention
was therefore easily caught and focused. In real life, events are not
usually isolated. And even when they are, the aspect of the event
focused by the verb is not spotlit. When children hear verbs such
as pour and ll in real scenes, the event is likely to be one in a
sequence in which the movement of the liquid or fullness of the
container will not stand out in any way. If children notice differen-
ces in the speakers focus within such real events, their sensitivity
to the speakers focus must be extremely acute.
Alternatively, they may turn to some other source of informa-
tion about the focus of the event. Perhaps they have other boot-
straps under their belts. Rather than using semantics to get at
syntax, how about using syntax to get at semantics? Enter the
theory of syntactic bootstrapping (Fisher, Hall, Rakowitz &
Gleitman 1994). According to this theory, if children can identify
the verb and its arguments, this will tell them which participants
play key roles in an event, and this will in turn tell them the focus
of the verb. Returning to our pouring/lling scenario, we now
have children observing the event while they hear the utterance
which contains pour or ll. When its pour, they notice that
the verb is followed by the argument the water. This focuses them
on the water rather than the container, they notice what is
happening to the water, and they deduce that pour refers to
movement. When the utterance contains ll, they notice that ll
is followed by the cup, which focuses them on what is happening
to the cup, and they deduce that ll refers to the state of the cup.
Evidence suggests that children can use syntactic bootstrap-
ping. This emerges from another experiment presenting children
with novel verbs (Fisher et al. 1994). Here, children were shown a
video of a familiar event which could be seen from two perspec-
tives. The video might show a rabbit feeding an elephant with a
spoon, an event roughly captured in the following still picture:
This could be seen as a rabbit feeding an elephant, but it could also

be seen as an elephant eating. Children heard the event described
in one of two ways:
The rabbit is ziking the elephant
or
The elephant is ziking
Children as young as 3 to 4 were inuenced by the argument

structure in which they heard the nonsense verbs. If they heard
The elephant is ziking, they were more likely to translate the
nonsense verb with a real verb tting the elephants perspective
eating or licking the spoon. If they heard The rabbit is ziking the
elephant, they were more likely to come up with a verb tting the
rabbits perspective feeding or giving him medicine.
Here, then, is another break in the vicious circle. As long as
they can pick out the verb form, notice the order of arguments
around it, and know what the arguments refer to, they can deduce
something about the verbs meaning. In particular, they can work
out what angle the verb takes on the event.
Again, the child has a lot of processing to do before syntactic
bootstrapping can come into play. Picking out the verb and argu-
ments in the stream of speech is no small feat. This is where
phonological bootstrapping enters the scene. Perhaps the
sound pattern of utterances gives children a leg-up to what they
contain, in particular, to the verb form and its arguments.
We know that verbargument structures have a characteristic
rhythm in which certain word categories are more prominent
than others. In particular, nouns typically receive more stress
than verbs. Suppose children are acutely sensitive to these rhyth-
mic structures. This could help them separate out the verbs and
nouns within the stream of speech. It could also draw their atten-
tion to the order of nouns in the rhythm, which is crucial to
establishing the verbargument structure. The techniques of pref-
erential listening and preferential looking have begun to
reveal just how sensitive infants are to the rhythm and structure of
utterances. Through these techniques, we have seen infants under
a year recognising the way words are grouped into larger units
(chapter 2). We have also seen how early they connect units in a
particular order to participants in events (this chapter). We have
reason to suspect that infants sensitivity is even ner, enabling
them to register the detailed rhythmic characteristics of verbs and
their arguments. Unearthing the evidence to support this suspi-
cion will, however, depend on rening still further our techniques
for penetrating the hidden depths of infants rhythmic processing.
Tying the bootstraps together

Research into bootstrapping has shown that children are sensi-
tive to ne differences in scenes, in sound patterns, and in the
organisation of words within sound patterns. Furthermore, they
know something about how each of these relates to the other. This
means they can use information they have about one to glean
information about the other.
It seems that children are equipped with a number of bootstraps
which are anchored in one type of information and extend this
into new information, and that new information may be reached
by more than one of the bootstraps. The focus in a scene can cue
the child to the verbs focus, but so can the presence and order of
arguments in the utterance describing the scene. The rhythm of
the utterance can cue the child to the presence and order of
arguments, but so can the focus within the scene. Perhaps it is this
convergence of bootstraps which gets children out of the vicious
circle they appeared to be caught in. By putting together
( the speakers perspective on the scene
( the rhythm of the speakers utterance
( the order of words within the rhythm
they can arrive at the meaning of the verb and the structure it
takes. It is the union or coalition of cues in the input (Hirsh-
Pasek & Golinkoff 1996), rather than any one of these cues, that
gets children to this destination.
The implications for the skills children bring to the sentence-
processing task are profound. They must be ready to notice subtle
features of events and of sound patterns. More than this, they must
be ready to notice when particular features of events co-occur with
particular features of sound patterns, and make connections be-
tween these.
Tuning into the childs tune

We have come some way in understanding how children solve
the apparently insoluble problem of mapping events onto verb
argument structures. Our conclusion is that children are attuned
to the rhythm of speech, the highlights of scenes and, most signi-
cantly, to the connection between the two. Only such a nely
tuned mind could steer children to the sort of verbargument
structures they typically use by around 4 years old.
The input the child receives could not on its own account for
this. But it could make a difference. If the input harmonises with
the tuning of the childs mind, it could make the childs task easier.
Given our conclusions about the tuning of the child, we can see
how adults might adapt their utterances to suit it.
First, they might help the child identify the focus of verbs by
using them at just the moment when the relevant event is most
salient to the child. Depending on the way the event happens over
time, this moment may be before, during or after the event. With
events such as ll, open, melt, or break, which have an
observable effect after the event has occurred, it may be best to
hear the verb after the event when its focus is most obvious. Events
such as move, pour, push, spin, on the other hand, may be
quite brief, and leave no conspicuous effect. With such events, it
may be more useful to hear the verb just before the event, alerting
the child to the dynamic change which the verb focuses (see
Tomasello & Kruger 1992, Ambalu, Chiat & Pring 1997). Notice,
though, that such manipulation of adult input can only help the
child to work out the connection between scene and sound. No
matter how effectively adults focus the childs attention on the
event, the child still has to deduce just what the verb is getting at.
Adults might also help the child identify the verbs form and
argument structure by exaggerating the rhythm of the utterance.
This is something adults typically do when they are talking to
children, suggesting they are intuitively aware that it helps. Such
exaggerated rhythm will highlight the nouns acting as partici-
pants in the event. This will help the child zoom in on the relevant
participants, and hence on the event in which they are involved. It
will also highlight the position of those nouns within the rhythm,
which will help the child make the link between argument posi-
tions and argument roles in events.
Such adult cues may be most effective if they work in unison to
push the childs attention in the same direction. By making certain
components of the scene and certain components of the utterance
more salient, adults may help children nd the connections be-
tween scene and utterance and pull out the verb and its argument
structure.
9 Shortfalls with verbs
When children have problems with words, their problems may hit
some types of words harder than others. A number of studies of
language-impaired children have exposed verbs as an area of
particular difculty.
One study involved an extensive comparison of the words and
structures used by a group of language-impaired children and a
group of normally developing children (Fletcher & Peters 1984).
The two groups were of similar age and non-verbal intelligence.
Each childs language was sampled in four different activities: free
play with toys, talking about pictures in a book, playing a board
game, and re-telling a story about a set of pictures. Each sample
consisted of 200 utterances, 50 from each activity. The samples
from the two groups of children were then compared in terms of a
wide range of lexical and grammatical categories. This revealed
that the vocabulary of the language-impaired children was gen-
erally more restricted than the vocabulary of the normally devel-
oping children, and signicantly, the verb range of the language-
impaired children was especially limited.
Similarly, in a study focusing on verb use, it was found that a
group of fourteen language-impaired children had a narrower
range of verbs than normally developing children matched in age
(Watkins, Rice & Moltz 1993). Even more striking, though, was
the comparison between these language-impaired children and a
group of normally developing children who were nearly two years
younger than them, but at the same general language level. The
verb range of the language-impaired group was still more limited
than that of the control group, even though their vocabularies as a
144
Shortfalls with verbs 145
whole did not differ signicantly. This nding was echoed in yet
another study, which compared three language-impaired children
with normally developing children whose mean length of utter-
ance had reached the same level: the language-impaired children
used a smaller number and variety of verbs (Conti-Ramsden &
Jones 1997). When compared with a younger brother or sister at
the same language level as them, the three children were also
found to make less use of the verbs they had available (Jones &
Conti-Ramsden 1997). The implication is clear, and the evidence
solid: verbs are more of an obstacle for language-impaired children
than other aspects of language.
Verbs go hand in hand with arguments the participants
involved in the event. We might therefore expect children who
have problems with verbs to have problems with arguments as
well. And so they do. Rice and Bode (1993) analysed the verb
usage of three language-impaired children. Again they found the
children relying on a limited set of verbs, and sometimes substitu-
ting one verb for another. But they also found that the children
omitted verbs or arguments on some occasions.
How do these difculties with verbs and arguments manifest
themselves? In the most extreme disruptions, children may pro-
duce just a verb or just an argument, as 10-year-old Richard
sometimes does:
My dads boss house. Down lane. Not farm. Walk up.
With no explicit relationship between verbs and arguments, we
are left in the dark about the states of affairs Richard is trying to get
across. In some cases, we might observe a verb accompanied by an
argument but in a rather disjointed structure, as in Richards
comment about the fate of lambs on his farm:
Girls, keep, boys, not keep.
In other cases, verbargument structures may be incomplete, as in
Eamonns repeated stabs at explaining who is going to call who
about what:
. . . my mum gonna call Mary gonna call me Anna gonna call m
['d] call my mummy I sick . . . Im getting sick in [] school Anna
gonna t phone my mum out phone my mummy out.
Or they may be mixed up, as in Eamonns further attempts to
explain the situation:
. . . phone my mummy out . . . Because he write my mum because
my mum write in my book.
Evidence of problems in the production of verbargument

structure may be matched by problems in comprehension. This
was checked out in an experiment where language-impaired
children were asked to act out sentences such as:
The boy pushes the girl The girl is pushed by the boy
The cup is in the box In the box is the cup
Give the boy to the girl Give the boy the girl
(Van der Lely and Harris 1990)
These children had more difculty than a group of younger,

language-matched controls, who themselves had more difculty
than a group of age-matched controls. In a second experiment, the
children had to demonstrate their understanding by selecting one
of four pictures to match the sentence. Again, the language-
impaired childrens scores fell below those of younger but nor-
mally developing controls.
Some experiments have probed childrens comprehension in
more detail. For example, they have investigated whether children
can make connections between syntactic roles of arguments and
the semantic roles these are most likely to play (Van der Lely 1993,
OHara & Johnston 1997). The children are presented with a
syntactic structure containing a nonsense word in place of a verb,
for example:
The woman soogs the bunny
The man keeds the clown to the monkey
(OHara & Johnston 1997)
The children are then asked to make the toys do the little stories,
or to make something up. The question is whether they know
which argument is acting as the agent which brings about the
event, the theme which undergoes the event, or the goal with
which the theme ends up. In the above examples we would expect
the woman and the man to be carrying out an action; the bunny
and the clown to be acted on; and the monkey to be the goal of a
transfer action. Yet again, language-impaired children fared less
well than younger normally developing children. They were less
likely to act out events with each participant playing the role we
would expect from the syntax.
Studies of language-impaired children therefore provide evi-
dence enough that verbs and their argument structures pose
special problems for some children. Where might their problems
arise?
Locating the problem

Verbs have particular characteristics which could make every
component of word processing more challenging:
( Verbs rarely occur in isolation, and when they occur in a
stream of words, they are typically less stressed than the
nouns around them. So, if children have problems picking
out the sound patterns of words, this may particularly
affect their identication and storage of verb phonol-
ogy.
( Verbs refer to events from certain perspectives which
focus on certain participants. They are typically dynamic,
and may be very brief. So, if children have problems
picking out relevant aspects of scenes, these may be par-
ticularly acute when they have to pick out an event from
a certain perspective, and may markedly affect their
identication and storage of verb semantics.
Where children have difculties with verbs, these could arise at
any point in the processing of verbs:
Figure 9.1 Processes in acquiring verbs
However, this view of verb processing is still only part of the

story. The special phonological and semantic properties of verbs
ow from their inherently relational nature: they express relations
between arguments. The acquisition of verbs therefore goes be-
yond the acquisition of single words. It also entails the acquisition
of argument structures. To acquire argument structures, the child
must pick out the participants involved in the event, pick out the
arguments which accompany the verb, and relate the two. Prob-
lems could arise in any of these processes:
Identification of form and

Identification of focused
position of arguments in
roles in scene
stream of words
Figure 9.2 Processes in acquiring argument structure
The acquisition of verbs and the acquisition of argument struc-

tures are not independent: the argument structure of a verb pro-
vides information about the perspective the verb takes on the
participants. This in turn provides information about the verbs
meaning. These connections lie behind syntactic bootstrapping
the possibility of using argument structure to work out a verbs
focus. If children have difculties picking out the arguments, they
Figure 9.3 Processes in acquiring verbargument structure
will be less able to use arguments as a cue to the meaning of the

verb. Conversely, the meaning of the verb provides information
about which participants are focused and how they will be or-
ganised around the verb. These connections lie behind semantic
bootstrapping, where knowledge of a verbs meaning is used to
predict the structures it will or will not take. If children have
limited knowledge of a verbs meaning, this may limit the predic-
tions they can make about the verbs structure.
Putting together the processes involved in connecting senten-
ces and events, we see where problems may arise and how a
problem in one process may impact on another (see gure 9.3).
Even if children make good headway with these processes in
input, they may falter in output. When they come to talk about
scenes, they may have difculty homing in on a particular event
which would lead them to a particular verb. Or they may know
exactly what event they are going for, but have trouble accessing
the phonological form of the verb, or accessing information about
where the participants go in relation to it. Or they may successfully
select a verb and arguments but struggle to hold these in mind in
order to articulate them.
How can we begin to track down the source of a childs difcul-
ties within the complex web of verb processing?
Exploring the problem

Psycholinguistic tasks used to investigate single word process-
ing (see chapter 4) can be adapted to investigate verbs in particu-
lar. For example, we may want to check whether the child recog-
nises verb forms regardless of whether she knows their meaning.
This would require a lexical decision task, but one in which the
stimuli were verbs together with similar-sounding non-words:
Real verbs Matched non-words

run [ln] lun
sing [] shing
break [plek] plake
spin [stn] stin
think [vk] vink
More likely, though, we will want to home in on connections
between verb forms and verb meanings. This could be done using
a picture-pointing task such as the one shown opposite:
Place each of the following pictures in front of the child and ask:
Which one is spinning?
Which one is cutting?
Which one is giving?

To pick the right picture in the pairs, the child must appreciate
particular aspects of the verbs meaning, such as direction of
movement, or manner of movement, or change of state, or direc-
tion of change of possession. Alternatively, an acting-out task
can be used. Here the child is given a set of appropriate props and
asked Can you show me spinning?, Can you show me giving?,
and so on. We can then see if the child includes crucial features
such as manner or direction of movement in the acting out.
If the child appears to have verb forms, but interprets them in
odd ways, or produces them in odd contexts, we may want to
explore the childs perception of events independently of verbs
themselves. For this, we might use an event copying task which
involves no use of language at all. Here, the child watches a
particular sort of event such as spinning, or giving, or breaking,
and is asked to do the same thing. The child might also be invited
to demonstrate events for the adult to copy. Through the childs
actions, we will gain some insight into what the child notices in
events, and most importantly, whether she notices aspects of
events which are relevant to particular verbs of the language.
These tasks can throw light on the childs processing of verb
semantics and phonology. To investigate the processing of argu-
ments in relation to the verb, we must move beyond single word
tasks. The tasks we use may involve similar procedures, but the
input or output required will be sentences rather than words.
From events to verbargument structures: event

description tasks
The surest entry point to the childs verbargument processing
is what she has to say about events. As a rst step, it would be
helpful if we could pin down more precisely what the child does
and doesnt do with verbs. Samples of everyday talk may give us
some clues, but they are unlikely to provide a comprehensive
picture. To sample a wide range of events, we need to devise
scenarios which will communicate different sorts of events to the
child, and nd out how the child translates these into words.
Events are dynamic, so they are more clearly captured using
dynamic representations such as videos or real-life actions than
static pictures. To elicit responses, we need to maintain the childs
interest in the events, and provide some motivation to describe
them. The use of novel events can do the trick. Events can be made
novel by casting things or people in roles which they do not
normally play a boy kissing a key, a book hitting a bus, a girl
tickling a ower, an animal pushing a car, a tree pulling a bike, a
man biting a dog, a baby pushing a pram . . . The more novel the
events, the more likely they are to hold the childs attention and
elicit a response, and the more sure we can be that the response
reects the childs ability to create verbargument structures she
has not met before. To make the task more meaningful to the child,
we can give it a communicative purpose:
(Set-up: child sits behind a screen with Sam, a friend, sitting on the
other side)
Im going to show you something funny. Sam cant see whats
happening. So watch carefully, and see if you can tell Sam what
happened.
This technique can be extended by giving the same set of props to

the child and to the party on the other side of the screen, and
saying:
Lets see if Sam can make the same thing happen. You tell him what
happened, and then well take the screen down and see if he does it
right.
Sampling a range of events in this way may reveal particular

areas of difculty. We will need to scan the childs responses for
descriptions which match the event and are complete, and descrip-
tions which are incompatible with the event or contain errors or
omissions. We can then compare these to see whether difculties
occur more with verbs conveying certain sorts of events, or requir-
ing certain sorts of argument structure. In searching out the roots
of those difculties, we will want to check whether parallel difcul-
ties crop up at other points in processing.
From verbargument structures to events: acting
out tasks
The most robust method for tapping the childs understanding of
verbargument structures is to ask the child to act them out. We
might, for example, offer the child a number of miniature gures
and say
Can you show me The boy gives the shoe to the granny
A correct response would involve putting the shoe with the boy
and then making the boy move it such that it ended up with the
granny. In contrast,
The boy takes the shoe from the granny
would have the shoe starting with the granny and ending up with
the boy. What is crucial about such sentences is that understand-
ing of the words and knowledge of the world are not enough to
interpret them correctly. This is because grandmothers are as
likely to take shoes from boys as boys are to take shoes from
grandmothers. Correct interpretation can only be achieved if the
child recognises that the boy is in the subject position, and that
the subject is the source with give but the goal with take. Such
sentences are said to be reversible because both characters can
plausibly occupy both roles. Contrast a sentence such as
The boy gives the carrot to the rabbit
Since boys are much more likely to give carrots to rabbits than
rabbits are to give carrots to boys, the child could get by if she
knew what boy, carrot and rabbit mean, and had experience of
feeding rabbits. The advantage of using reversible sentences is that
they truly test the childs mapping between syntax and semantics
and push it to the limit. However, interpreting sentences without
support from context or world knowledge is much more demand-
ing than interpreting sentences in everyday contexts in real life.
Recognising the argument structure of verbs:

judgement tasks
Some children clearly have connections between verbargu-
ment structures and events. They show understanding of these in
everyday life, and their understanding is conrmed by, say, an
acting-out task. They also use verbs with arguments to talk about
events. Yet their production is unreliable. They sometimes
struggle to recount events, or leave out verbs, or leave out argu-
ments, or put arguments in the wrong place. If they have connec-
tions between structures and meanings, why do they not use
structures consistently to express meanings? It looks as if the
structures themselves are the problem. That is, they have not
established patterns of sentences, which are to some extent inde-
pendent of the meaning we want to convey. They have not estab-
lished that sentences require at least a verb and the arguments
entailed by that verb even if these elements are not all crucial to
their message.
Trying to tap a childs knowledge of sentence structure is
akin to trying to tap knowledge of word phonology. In both
cases, our target is linguistic knowledge which is independent of
meaning. If we are to probe this knowledge, we must devise tasks
which require processing of word and sentence form, but do not
require connections to meaning. Judgement tasks allow us to do
this. Lexical decision tasks, for example, asked the child to judge
whether something sounded like a word or not, a judgement
which requires knowledge of word phonology but makes no de-
mands on word meaning. At the sentence level, sentence judge-
ment tasks offer a similar opportunity. If we are checking knowl-
edge of verbargument structure, we need to create sentences
which are correct, and sentences which violate verbargument
structure, and ask the child to judge these. Presentation might run
along the following lines:
Remember this puppet? Shes the puppet who sometimes says things
right and sometimes says funny things. When she says something
right, you can give her a tick. And when she says something wrong,
you can give her a cross. Then at the end we can see how many she got
right. Lets try this one: Broke the vase the boy. How does that sound?
Are you going to give her a tick or a cross? And now how about this
one: The boy cleaned the plate. How does that sound?
In this example, the non-sentence violated the order of arguments in

English. The expected order would be The boy broke the vase.
Other examples might leave out an argument which the verb de-
mands:
The man found
where nd requires a direct object. Or they might include an

argument which the verb does not take:
The woman disappeared the hat
where disappear cannot take a direct object. If the child is able to

distinguish the good and bad sentences, we might conclude that
she does know which arguments occur where in relation to the
verbs used. In this case, any errors she makes in talking about
similar events could not be attributed to ignorance of the verb
argument structures.
What about a child who makes no response at all, or responds
yes to every example, or responds at random? In such cases, no
immediate conclusions can be drawn. First of all, she may have
problems with the task. Making judgements about words is a hard
enough task; making judgements about sentences is even harder.
But the sentence-judgement task is problematic in other ways.
Although its aim is to elicit syntactic judgements about sentences,
the child may go about judging the sentences on some other
grounds. The sentences and non-sentences used contain real
words which have meaning for the child. It is highly likely that
some children will judge the examples according to whether they
can make sense from the words, ignoring how they sound. For
example, on hearing Broke the vase the boy, the child may
conjure up an image of a boy breaking a vase, decide this is a
plausible scenario, ignore the wording she actually heard, and
accept the sentence. The task has then tapped the childs word
semantics and world knowledge, and her responses do not reect
her syntactic knowledge. One way round this might be to make all
the sentences nonsensical:
Syntactically okay: The boy broke the milk
Syntactically out: Broke the boy the milk
and see if the child can make any distinction between these
semantically odd examples. Very obviously, this makes the task
even more abstract and obscure. We would only contemplate
attempting such a high-level task in an older language-impaired
child who had shown sufcient ability and motivation to under-
take the linguistic challenge.
Sentence judgement tasks, then, are not at all straightforward.
They must be handled with caution. As with any psycholinguistic
task, they can be practical and informative provided they are used
discriminatingly and as a tool for exploration, rather than mech-
anically. We can try out the task just to see how the child reacts,
look out for any evidence that the child notices differences between
the examples, and invite the child to say how she might like to
change the sentences. The results of such exploration may not be
denitive, but may still yield some evidence that the child notices
when something is wrong.
From verbargument input to verbargument

output: repetition tasks
Sentence judgement relies on conscious reection about senten-
ces to tap the childs unconscious awareness of how sentences
should sound. A rather different way of coming at this is via a
repetition task asking the child to reproduce verbargument
structures such as
The boy gives the shoe to the granny
To do this, the child must at least register the phonology of the

sentence its rhythm, and the sound patterns of the individual
words within that rhythm. She must hold onto these rhythmically
organised word forms. And she must plan their production. She
may well take in their meaning as well, but repetition does not
require this. Nor does it require conscious reection, concentration
or effort. Children are therefore unlikely to struggle with the
general demands of the task, so that their responses will provide a
clear window onto processing skills. If the child has the capacity to
recognise, hold and produce the sentences in a task, we can be
pretty sure this will show up in correct repetition.
Conversely, if repetition falls short of the target, we can be
pretty sure the child does have problems with phonology. Since
repetition involves input, storage, and output of phonology, the
problems could be arising in any or all of these processes. In order
to distinguish these possibilities, we must make comparisons with
other tasks:
( event description: involves semantics and phonological
output but not phonological input
( acting out events: involves phonological input and se-
mantics but not phonological output
As well as revealing problems at one or more stages of phono-
logical processing, repetition offers a key to the nature of those
problems. The key will be the pattern of errors. Does the child omit
words, or does she distort them? Which words are most likely to be
affected? Are nouns preserved better than verbs? Is word order
preserved? The answers to these questions will indicate which
aspects of sentence phonology are strongest for the child, and
which are vulnerable.
Making the mapping: semantic and syntactic

bootstrapping tasks
The tasks we have surveyed so far all draw on verbs the child
has already met. The childs response to these is a product of all the
components of verb processing which she has at her disposal.
Components of verb processing are, as we have seen, highly
interdependent. It may therefore be difcult to sort out which
components of verb processing are chickens and which are eggs
when we look at verbs about which the child has already received
information. We can get round this by providing the eggs our-
selves, and seeing what chickens they hatch. If we introduce new
verbs, we can control the information the child receives about
those verbs, and see what the child can do with that information.
Experiments on bootstrapping in normally developing children
offer us the tools we need (see chapter 8).
In these experiments, children are introduced to nonsense verbs
in contexts which give them limited information about those
verbs. They may be given just semantic information, using a
semantic bootstrapping task (Gropen et al. 1991). Here, the
adult introduces a novel verb in a context which provides no
syntactic information about it:
let me show you what keating is . . .
The adult then demonstrates the event, in this case, moving a

packet of marbles onto a square of cloth causing the cloth to sag.
The childs attention is drawn to what is important in the event:
when I do this [moving a packet directly towards an unsupported
square] . . . and it ends up like that [placing the packet onto the square,
causing it to sag] . . . its called keating.
This is reinforced by showing an event in which the marbles move

onto the cloth without causing the cloth to sag:
now let me show you something thats not keating . . . when I do this
[moving a packet towards a supported square] . . . and it ends up like
that [placing the packet onto the square, without it changing its shape]
. . . its not called keating.
(Gropen et al. 1991: 168)
After demonstrations of keating by the adult and the child, the
action is performed again and the child is asked a question design-
ed to elicit a structure with the novel verb:
Can you tell me what Im doing?
If this does not elicit a structure with the verb, further prompts
may be used:
Can you tell me what Im keating?
or
Keating what?
(based on Pinker 1989: 346)
The question is whether the child will select the cloth or the
marbles for the direct object slot. If she uses semantic bootstraps,
she will note that the verb focuses on the sagging state of the cloth
and will make this the direct object of the verb, in which case she
will come out with Keating the cloth or The cloth. We would
expect a different response if the focus of the verb were different.
This is tested with a second type of event. In this case, the marbles
are moved onto the cloth in a zig-zagging manner. This event is
demonstrated in a similar way to the previous one, with the adult
saying:
when I do this . . . and it ends up over there . . . its called pilking.

(Gropen et al. 1991: 169)
To exemplify an event which is not pilking, the marbles are moved

in a bouncing manner. The focus of the verb is now the zig-zagging
movement of the marbles. Semantic bootstraps will therefore lead
the child to select the marbles for the direct object slot, and
produce Pilking the marbles or The marbles in response to the
prompt questions.
Investigation of syntactic bootstrapping requires input
which gives syntactic information about the verb, but limited
semantic information. One example of this starts with the follow-
ing introduction:
Mac [a puppet] doesnt speak English very well, so sometimes he uses

puppet words. Can you help us gure out what the puppet words
mean?
(Fisher, Hall, Rakowitz & Gleitman 1994: 344)
The child then hears a sentence such as The bunny is nading the
ball from the elephant, and watches a video of a scene, captured
here in a still picture:
In this case, the child is watching an elephant handing a ball to a
bunny who takes the ball. After seeing the video and hearing the
sentence, she is asked:
What does nading mean?
She may be probed further with the prompts:

Whats happening? Whats going on?
Any change of possession verb such as give, or hand, or

take, or get would be compatible with the scene. But only verbs
such as take or get are compatible with the argument structure
accompanying the nonsense verb and the focus implied by this.
Will the child come up with a verb such as taking or getting? If
so, she has translated the nonsense verb with a real verb sharing
the same argument structure as the nonsense verb and therefore
the same focus. This would suggest that she has recognised the
position of the arguments and has established the links between
positions in argument structure and the way roles are focused in
events. If she has, she should make a different response when she
meets the arguments in a different order, as in The elephant is
nading the ball to the bunny. Here, the argument structure and
the implied focus should elicit real-verb translations such as
give or hand.
We might think about using these experimental techniques to
tap bootstrapping processes in language-impaired children, but
with some important reservations and adaptations. In the case of
language-impaired children, the use of nonsense verbs might be
considered unethical and inappropriate. But it would not be dif-
cult to nd real verbs which, like nonsense verbs, would be new for
the child. These could be verbs which the child had shown dif-
culty understanding or producing. Or they could be verbs which a
child of the same age would not be expected to know anyway.
Unlike nonsense verbs, these verbs would be potentially useful to
the child. This would overcome ethical objections to teaching the
child nonsense words.
However, we may have concerns about other aspects of these
experimental techniques, such as the high demands they impose.
A semantic bootstrapping task would seem quite feasible with
language-impaired children who are at least at the stage of
producing two-word utterances. The syntactic bootstrapping task
described above, however, requires a more sophisticated type of
response. In order to succeed, the child must have the notion of
using one word to explain the meaning of another. Only children
who have shown that they have this notion will be able to under-
take syntactic bootstrapping tasks of this nature. If they do succeed
in these tasks, we know for sure that they have links between
argument positions and roles focused in events.
Any pursuit of verbargument processing in a child will require

careful selection of tasks which are not only relevant to the childs
problems, but are appropriate to the childs general level of devel-
opment and interests. The following case study of 6-year-old
Travis illustrates the use of various tasks to investigate a childs
processing of verbargument structures. By the end of this explo-
ration, we arrive at some understanding of the ways in which a
childs bootstraps might fall short, and so limit his ability to talk
about events.
10 Thing out. Tip in there:
problems with verb
processing
Travis has much in common with the children we followed up in

chapter 6. He too comes across as a bright, responsive child who
interacts well with other people. He takes turns in conversations,
and his contributions bet the context. Pragmatically, he is on the
ball. But when we home in on the specic content of his utteran-
ces, we start to notice problems. These problems also show up in
formal tests of his language. At 6, he performs at the level of a 4- to
412-year-old in tests of word and sentence comprehension:
British Picture Vocabulary Scales (BPVS): scored at age
level 4;1
Test for the Reception of Grammar (TROG): scored at age
level 4;6
In contrast, his score on a test of non-verbal intelligence is average
for his age. Like the children presented in chapter 6, Travis shows
particular difculties when it comes to dealing with words and
combinations of words. In presenting Traviss case, however, the
spotlight will move from words in general to the particular words
needed to express events, namely verbs and their arguments. Our
purpose will be to nd out where, in the connections between
events and verbargument structures, Traviss difculties arise.
Our starting-point is an event-description task which estab-
lished that Travis did have particular difculties in talking about
events. In this task, Travis watched his therapist act out 127
different events. These events represented thirty verbs, which were
enacted in different contexts and with different participants. The
verbs spanned a variety of event categories, including:
163
( something moving in a particular manner or direction,
e.g. a ball rolling, a penny sliding, books or cards falling
down
( something changing state, e.g. a spider dying, a door
opening
( the therapist causing something to move in a particular
manner or direction, e.g. rolling a ball, rocking a doll,
pushing a pram or a box
( the therapist causing something to change state, e.g.
opening a card or box or mouth, cutting bread or paper
After he had watched each event, Travis was asked What is
happening? or What am I doing?.
In order to see whether the description of these events posed
unusual problems for Travis, they were also presented to three
normally developing children. These children were matched with
Travis on vocabulary age, which put them well below his chrono-
logical age they were two to two and a half years his junior. We
have already observed that normal 4-year-olds have very good
control of verbargument structures in the output they produce
spontaneously (see chapter 8). But how would they fare in an
elicitation task calling for them to produce 127 verbargument
structures on command? Just ne, as it turns out. They co-oper-
ated in the activity quite happily, and were virtually spot-on in
their descriptions. For at least 90 per cent of the 127 events, these
children used a verb which was consistent with the event, and
combined it with any arguments it required to depict that event.
Between them they produced 363 such descriptions. Typical
examples are their descriptions of pouring sugar into a bowl:
Hazel: Pouring something into the bowl
Rosie: Pouring it into there
Ross: Tipping it all into there
or dropping apples:
Hazel: They all comed out
Rosie: You dropped them
Ross: They dropped out
or a top spinning:
Problems with verb processing 165
Hazel: Spinning
Rosie: Its twirling round
Ross: The top is spinning
These results conrm that normally developing 4-year-olds are
extremely adept at seeing an event in an appropriate way and
representing this with an appropriate verb or verbargument
combination.
How about Travis? He too provided appropriate event descrip-
tions. His description of dropping straws as
Travis: Straw [] fall on carpet
and breaking a pencil as
Travis: Pencil broke
contain adequate verbargument structures to characterise what
he had observed. But such descriptions account for just over 50
per cent of his responses, showing that he is far less consistent than
the control children. The majority of his inappropriate descriptions
fall short in the following two ways:
Verb omission
Twenty-two of Traviss descriptions contained no verb, as in his
versions of:
Pouring sugar into the bowl: Sugar in the pot
Dropping apples: Fruit on oor
Shaking the st: Strong
A top spinning: Round
In contrast, the one and only example of verb omission from the
control children was Rosss description of shaking the head as
Nos.
Use of a bare verb

Thirty-three of Traviss descriptions consisted of a bare verb with
no arguments and no verb ending, as in his versions of:
Breaking an egg: Break.
Tractor pushing mud: Dig, dig, dig
Shaking a bottle: Shake
The controls produced only ve examples of bare verbs between
them.
Evidently Travis is not able to describe events as reliably as
these much younger children. This suggests that the event-de-
scription task posed special problems for him.
What is the source of those problems? Our initial hunch might
be that events themselves are not the obstacle. If they were if
Travis understood events differently from other people we would
surely expect his behaviour to be odd in ways which it is not. His
generally appropriate interactions with the world, both non-ver-
bal and verbal, argue for a normal orientation towards things and
people and the events in which they participate. If the problem is
not with events, it must be somewhere in the connection between
events and verbargument structures.
Questions arising from Traviss event description

Our investigation so far has revealed limitations in Traviss
output. Those limitations could, though, have their origins in
input. To be able to produce his own verbargument structures,
Travis must rst process verbargument structures which he
hears. At rst sight, Travis seems able to do this. His comprehen-
sion in context is good, which suggests that he is interpreting
events appropriately. But such evidence is loose. Comprehension
in context can be aided by all sorts of non-linguistic cues from the
situation, the direction of the speakers gaze, the speakers ges-
tures. It may be that Travis is able to use these sort of cues very
effectively and so mask difculties with precise processing of verb
argument structures. One hint that he does have such input
difculties comes from his performance on formal tests. Recall that
his score on a test of sentence comprehension, the TROG, had
found him nearly two years behind his peers. This does not prove
that Travis is behind in interpreting events, since the TROG is too
wide-ranging to identify such specic decits, but it does indicate
that all is not well in input.
It is therefore possible that the problems observed in Traviss
output do have their origins in input. Perhaps he omits verbs in his
output because he has failed to identify verb forms in the stream of
speech, or to connect these verb forms to corresponding events, or
to recognise that they are obligatory. Perhaps he uses bare verbs
because he has failed to identify the way arguments are organised
around verbs, or to connect arguments to participant roles, or to
recognise that they are sometimes obligatory.
A closer look at Traviss description of events throws up some
relevant evidence. If omissions were due to ignorance of verb
forms and structures, we would expect these omissions to be
consistent. But they are not. On some occasions Travis succeeds in
using the very forms which he could have used on other occasions where
he used no verb at all. Of his twenty-two verb omissions, sixteen
were in response to an event which could have been described
with a verb he did use for other events. For example, he used the
verb tip for:
Emptying a jar: Tip in there
which would have served him perfectly well for describing all the
other emptying events where he omitted the verb:
Emptying bottle: Drink in there
Emptying glass: Water and drink in there
Emptying pocket: Thing out
Emptying box: Thing out
Similarly, compare his descriptions of different rocking events with

the verb rock:
Rocking side to side: Rocking
and without it:

Rocking arms: Wobbly jelly
Rocking forwards and backwards: I dont know
In the same way, Traviss use of bare verbs cannot have been
due to ignorance of argument roles and positions, because on some
occasions he uses the very same verbs with arguments. For example,
he used just shake for:
Shaking a bottle: Shake
yet combined it with the direct object argument for

Shaking the head: Shake [] head
Shaking hands: Shake [] hand
There was only one case of a verb being used without arguments
in all instances.
These inconsistencies provide evidence that Travis has recog-
nised and stored verb forms and argument structures which he
does not always use. When he does use them, they are generally
appropriate. On only a handful of occasions does he use a verb
unsuited to the context, as was the case in:
Pushing a pram: Drive [] baby chair
where drive is wrong for the movement of a pram. But he was not
alone in doing this. Ross, one of the control children, also made a
handful of verb selections which were slightly odd for the event:
Wetting the brush: Twisting it in there
This shows that Travis, like the control children, has verb forms
and argument structures associated with appropriate events and
event roles.
The inconsistencies we have observed in Traviss event descrip-
tion rule out a total blockage in his recognition of verb form or
meaning. What they do not rule out is some sort of limitation in
the connections he has made between form and meaning. It may
be that he has acquired certain verb forms and their argument
structures, has established some of the events they can refer to, but
has not established their full semantic range. For example, he may
have the verb tip associated with movement from a jar, but not
from a bottle, glass, pocket or box. Or he may have adequate
semantic information about many verbs including tip, but fail to
realise that verbs are essential in sentence structure leading him to
treat tip or any other verb as optional in the sentence.
Assuming Traviss output is constrained by some such limita-
tions in verb information, these might stem from limitations in
input processing which have repercussions for output, or they
might arise only in output.
These possibilities were followed up by designing a variety of
input and output tasks. All the tasks used the same set of verbs in
the same structures, so that any differences in Traviss responses
could be attributed to the task rather than the linguistic items. The
verbs used were:
Change of possession: give, take, offer, accept, buy, sell
Change of location: put, throw, drop, chase, follow
Psychological state: like, hate
Possession: have
The tasks were again carried out with three normally developing
children matched to Traviss vocabulary age.
Input investigations
Two tasks tested Traviss comprehension of verbargument
structures.
In an acting out task, the therapist presented Travis with an
appropriate set of toys to act out sentences such as the following:
The girl gives the apple to the boy
The pig buys the orange from the sheep
The boy puts the box on the ladder
The balloon has a picture on it
She introduced the task by saying:

I want you to make the toys do what I say
and then gave him each sentence to act out. To show correct
comprehension, he had to use all the participants mentioned in
the sentence, and put them in their appropriate role in the event.
For The boy puts the box on the ladder, for example, he would
have to make the boy the agent causing movement; the box the
theme, undergoing movement; and the ladder the goal, where the
theme ends up. Responses which overlooked a participant or put it
in the wrong role were scored as incorrect.
Traviss performance on this task was not signicantly different
from the control childrens, though it did fall at the bottom end of
the group. Travis also made the same sort of errors as the controls.
Over half of these involved the reversal of arguments in acting out
the sentence. For example, he acted out
The pig buys the orange from the sheep
with the pig, who should end up with the orange, as the starting-
point (source) and the sheep, who should start out with the
orange, as the end-point (goal). Travis made at least twice as many
reversal errors as any of the controls. However, the verbs which
were most troublesome for Travis were also those which gave most
trouble to the controls. These verbs were take, accept, buy and
sell.
In a picture-pointing task, the therapist presented Travis
with two pictures, one representing a sentence such as
The boy takes the apple from the girl
and the other representing the same event but with the roles
reversed as in
The girl takes the apple from the boy
The therapist then said:

I want you to point to the picture that goes with what I say.
Again, Travis made some errors, but he performed above chance

level, and just as well as the control children. Again, he had most
difculty with possession verbs such as give/take, buy/sell,
offer/accept.
What do these two tasks tell us about Traviss comprehension
of argument roles? First, since his performance was not signicant-
ly worse than the vocabulary-matched controls, the tasks do not
expose any problems in identifying argument roles beyond those
which occur in normal development. But nor do they expose just
how much Travis or the controls understand from verbargument
structures. While Travis shows some success with these tasks, he
could be succeeding with many of the items on the basis of a rather
broad understanding of the verbs and the argument structures.
Suppose he knew roughly what sort of event the verb expressed,
for example whether it referred to state or location or possession.
Suppose he also knew the most typical mapping relationship be-
tween syntactic and thematic roles around the verb, that is
subject NP = agent
direct object NP = theme
PP = goal
Such broad knowledge would get Travis a long way. He would
only come unstuck where the sentences contained verbs which go
against these generalisations. Take, buy and accept are just
such verbs: all map their goal onto the subject NP, contrary to the
expectation that the goal will map onto the PP. And it was with
just these verbs that Travis, and some of the controls, came most
unstuck. Thus, Traviss relatively good performance on this task
does not rule out limitations on his understanding of verbs or
verbargument structure.
On the other hand, it is also possible that his understanding is
better than it appears in these tasks. It may be that the demands of
remembering the sentence, picking out the characters and their
role, and then having to perform the actions or select one of two
almost identical pictures to match his interpretation are just too
much, and he loses track of who was doing what. In this case,
difculties with the task would be masking a perfectly good under-
standing of the sentence. This is an unavoidable hazard of input
tasks, which can only tap language processing indirectly.
The upshot is that these tasks have left us little the wiser about
the precise nature of Traviss input processing. We can be sure that
Travis is able to discriminate the order of arguments in the sen-
tence, and knows something about the relationship between the
position of an argument and its role in the event. But this much
was already evident from his output: he might omit arguments,
but when he does include them, he does not mis-order them. Input
tasks have not revealed just how much he knows about the event
roles which different verbs impose on their arguments.
To make more headway, we might turn to output tasks, which
tap the childs language much more directly, to see whether these
allow us to make more precise and reliable inferences about input.
Output investigations
Three output tasks were carried out. These were description of a
picture, repetition of a sentence, and repetition of a sentence with
the support of a picture. These tasks give us further evidence of
Traviss output. But they could also yield evidence of his input,
since the tasks use different sorts of input a picture or a heard
sentence or both in order to elicit the same output.
In the picture-description task, the children were shown a
picture and given a verb to see if they could put that verb into a
sentence matching the picture. The pictures were the same as
those used in the picture-pointing task. The lead-in went like this:
This picture shows pushing. Tell me all about this picture.
As in the original event-description task, Traviss performance was

way below the control childrens: the lowest of their scores was
almost double Traviss. His picture descriptions were sometimes
reminiscent of the event descriptions we have already seen, for
example those containing no verb:
(Picture shows boy putting ladder on box; the verb put is supplied)
Travis: Train track on the box

But on this task, where he was given a verb to use, his errors more
often involved substitution of a different verb and sometimes
multiple verbs for the target:
(Picture shows cat accepting cake from dog; the verb accept is supplied)
Travis: Cat dog get the cake [] cat.
(Picture shows panda buying pear from monkey; the verb buy is
supplied)
Travis: Monkey panda want a pear monkey want money.

(Picture shows man offering biscuit to lady; the verb offer is supplied)
Travis: The man make pizza give a man lady

(Picture shows sheep selling pear to pig; the verb sell is supplied)
Travis: Pig want want pear in basket sheep give it pig pig want
pear um basket and sheep and he give it and lot lot of 5p
Such examples provide evidence that Travis is trying to convey all
aspects of the event entailed by the target verb. Though he never
uses buy or sell he covers their meaning pretty exhaustively with
want, give, money and lot of 5p. But he has separated these elements
out rather than integrating them into a single verb. The controls
made no such responses.
In the repetition task, Traviss performance sank to its lowest
and moved furthest from that of the control children. Here, the
children were required to repeat the sentences used in the acting-
out and picture-pointing tasks. The controls did this effortlessly
and almost without error. In contrast, three-quarters of Traviss
responses showed omission of a verb or argument, or substitution
with weak, unintelligible forms consisting largely of []:
Target: The pig chases the sheep
Travis: [] pig [h ] sheep
Target: The cow gives the banana to the horse
Travis: [] cow [h h] horse
Again, Travis was inconsistent, producing the same target verb
correctly in some instances:
Target: The boy chases the girl
Travis: [] boy chase [] girl
Target: The boy gives the apple to the girl
Travis: [] apple [] give [] girl
Following Traviss very poor performance on pure repetition,

the task was repeated with the support of pictures. In the repeti-
tion with pictures task, the therapist showed Travis a picture,
produced the corresponding sentence, and asked him to say that
sentence. Travis now made fewer omissions, but more often sub-
stituted a different verb for the original one:
Target: The monkey chases the panda
#
Travis: [] monkey [ fl ] panda
Target: The boy gives the apple to the girl
Travis: Boy [h] apple [] girl
His responses on this task were therefore more like his picture
descriptions, suggesting he was using the picture input to generate
a sentence rather than simply repeating what he heard.
It seems that repetition is the greatest stumbling block for
Travis and even harder than picture description (hence the im-
provement when repetition was supported by a picture). This
reverses the response patterns of the control children. For them,
repetition is no problem, yielding even higher scores than picture
description, where they already surpassed Travis by a long shot.
The difference between these tasks lies in input one starts from
pictures, the other from heard sentences. Traviss response to them
tells us something about his processing of these different inputs.
His overriding difculty with pure repetition points a nger at
phonology. Suppose Travis has particular difculty registering the
phonological forms in the input utterance. This would certainly
stymie his attempts at repetition. It would explain the occurrence
of those weak unintelligible forms such as [] and [h] and partial
forms such as [h] where verb forms should occur. It looks as if
Travis registers the overall rhythm of the utterance. He also regis-
ters the phonological form of two or three key words. But he does
not register the phonology of all the words, and in the absence of
any other cues, he can only ll the rhythmic slots with some sort of
phonological padding. When repetition is supported by pictures,
on the other hand, he can supplement the information from the
heard utterance with the information from the picture. Having
missed the form chase in the input utterance, he can observe the
chasing/following in the picture and go from semantics to a form
he has stored.
If phonology is the culprit for Travis, the gulf between his
pattern of response and the control childrens falls into place. For
the control children, input phonology is no challenge at all. Hear-
ing each sentence, they can register the word forms and their
order and can produce the verb and its arguments even more
reliably than when they generate these from meaning. It could be
that they are repeating the sentences without accessing their
meaning fully or at all. What lies in the gulf between Travis and
these children, it seems, is their rock-solid capacity for processing
input phonology.
If a childs phonological processing is shaky, we might expect
problems with verbs and verb structure as fall-out. The verb form
will be vulnerable because of its typical position within the sen-
tence. Recall that verbs, because they often take arguments, rarely
occur in the nal position which carries the greatest stress. Verbs
are typically surrounded by words which are more stressed than
they are. So Travis is likely to have more difculty with verb forms
than, say, nouns. This could account for his less reliable produc-
tion of verbs, both in repeating sentences and in describing events.
Argument structure will also be vulnerable. Though the nouns
in the arguments will be more stressed, more salient, and more
strongly represented, this will not protect their relationship to the
verb. In order to represent the verbargument structure stably, the
child must hold the verb unit together with the arguments. A
constraint on phonological input may prevent Travis from regis-
tering the verb unit and all accompanying argument units simul-
taneously. Unstable representation of elements in the rhythmic
structure will undermine the phonological bootstraps by which
children reach for verb structure and will prevent them establish-
ing a secure relationship between verbs and arguments. This could
explain why, even when Travis knows and can use verbs with
their arguments, he does not do so reliably.
The insecurity of verb structure will in turn affect syntactic
bootstrapping. Only by noticing that pour is followed by a NP
referring to a moving substance while ll is followed by a NP
referring to a container can the child use their arguments to work
out the different events focused by each verb. If Traviss informa-
tion about what follows the verb is precarious, his access to the
verbs focus will be precarious too. He will have to rely much more
on information from the scene he is observing to work out what
the focused event is likely to be.
Such information should allow him to use semantic boot-
strapping, which involves working out the syntactic position of
participants in the sentence from their role in the scene. However,
even semantic bootstrapping may be undermined. Semantic boot-
strapping relies on the child having secure knowledge of the way
participant roles link to argument positions around the verb. If the
childs access to verb structures is haphazard, it may take longer to
identify and consolidate these links. He may have specic
examples of verbs which focus on different thematic roles, such as
ll and pour, but these may be too few or too insecure for him to
establish the ways in which focused participants connect to syn-
tactic positions. We saw above (chapter 8) that the different boot-
straps semantic, syntactic and phonological are mutually
dependent. It seems likely that the potential of semantic bootstraps
may be curtailed where it it not supported by other bootstraps.
On our interpretation of Traviss difculties with verb structure,
we would not expect either semantic or syntactic bootstraps to be
totally lacking, but we would expect both to be affected to some
degree. An in-depth investigation of bootstrapping in a child like
Travis would throw more light on what information he can and
cannot use in learning verbs. It would also indicate what kind of
information can best help the child get on top of verbs and their
structure.
The journey we have taken through Traviss processing has

landed us in a rather unexpected place. We started off from an
observed problem with verbs and verbargument structure, which
would seem to suggest difculties with syntax and semantics. Yet
we have concluded that Traviss difculties lie in his processing of
phonological input, which is crucial for acquiring and using verb
argument structures to convey events. The implication is that any
children whose phonological processing is similarly limited as
evidenced by their particular difculties in repeating sentences
will struggle to establish and produce verbs and verb structures.
Another implication is that these children will have even more
problem with words which are even more phonologically chal-
lenging than verbs. It is to such words that we now turn.
Note: This chapter is based on research carried out by Mary Evelyn,

Speech and Language Therapist, Mid Anglia Community Health NHS
Trust, at City University, London (Evelyn 1996).
Part III
Missing function
morphemes
MMMM
11 Filling out sentences
When we hear Ruth say:

Stay under water long time, that hurt my eyes
and Richard say:
Girls, keep, boys, not keep
we automatically ll in certain missing bits. These bits will
depend on our interpretation of the utterances and knowledge of
their context. In the light of this information, we might expand
Ruths and Richards utterances into:
When I stay under water for a long time, that hurts my eyes
The girls we keep, the boys we dont keep
The missing bits are a well-known stumbling block for children
with language impairment, a classic sign of their difculties. Col-
lectively known as function morphemes, these bits are distin-
guished by their minor status relative to the content words
such as nouns and verbs on which they depend.
Phonologically, they are weak. Their phonological bittiness
shows up in different ways. In English, they are typically un-
stressed words or inections.
Unstressed words take up the weakest beats in the rhythm of a
sentence. Examples are auxiliary verbs such as can, has/have,
was/were. When these occur in context, they typically reduce,
with the vowel of their full form shortening:
he can dance he [kn] dance
they have danced they [v] danced
we were dancing we [w] dancing
183
184 Missing function morphemes
or they lose their vowel altogether, contracting onto a previous
word:
theyve (they have)
hell (he will)
Inections are part of a word. Examples are the sound segments
-s, -ed and the syllable -ing, which attach to verbs:
walks,walked, walking
These unstressed words and inections contrast with content
words which always consist of at least a syllable with an irreduc-
ible vowel. So, we cannot do to the content words pan and fur what
we can do to the function words can and were: we cannot reduce
them to [pn] and [f].
Semantically, function morphemes typically do not refer to
things or events in the outside world. They depend on the refer-
ences which are made by the content words to which they attach.
Some specify a dimension of that content word, such as the
number of a noun (books) or the timing of a verb (walks, walked).
Function morphemes in different languages pick out different di-
mensions, and cut them up to different degrees. For example, the
marking of number on nouns may be obligatory, or conned to
certain classes of nouns, or optional, or absent. It may divide into
just the categories singular and plural, or it may include a further
category, dual, which distinguishes two items from more than
two. Time, too, is divided in different ways. Some languages make
a two-way distinction between past and non-past, or future and
non-future; some make a three-way distinction between past,
present and future. Some mark other dimensions of time, such as
the time course or completeness of an event.
Other function morphemes contribute nothing to the meaning
of the content word. They are determined by a syntactically re-
lated content word, and act as a marker of agreement between
these content words, indicating their syntactic relation to each
other. Many languages, for example, mark verbs according to
features of their subject or object, such as its number or gender. In
English, this accounts for the difference between
you walk
Filling out sentences 185
and
he walks
where the inection -s is determined by the 3rd-person-singular
subject. Many languages mark adjectives and determiners accord-
ing to particular dimensions of the noun they modify, for example
its number, gender or animateness.
Syntactically, the hallmark of function morphemes is that
they are obligatory. Languages vary widely in the dimensions and
relations of content words which they choose to mark, but within
any one language those distinctions are sacrosanct. That is pre-
cisely why we can ll in the bits which Ruth and Richard omit.
Given the context and the content words they have used, English
speakers know which aspects of content words have to be spelled
out.
Somehow, children must discover these phonologically
meagre, semantically abstract items on which their language
insists. This does not happen overnight. Function morphemes
unfold gradually, even in the language of normally developing
children.
Output: the normal road

The clearest evidence comes from their own utterances. As we
have seen, childrens early words are typically content words
identifying things, people, and events or relationships in which
these participate. When children start producing word combina-
tions, these are also largely conned to content words. The pre-
dominance of content words was apparent in the sample of ut-
terances produced by Alison (chapter 8):
Utterances produced by Alison, aged 2;3
Drop (after penny dropped) Wash teddy
Fall out (after money falls out of purse) Open it
Hiding (when teddy is in cupboard) Push me
Open (trying to open purse) Got some
Off (taking socks off doll) Go in there
Ali do it ( = open door) Teddy wake up
My daddy go home I [wan] money
Phonologically, almost all her words are those which carry stress
and cannot reduce. Semantically, they are words which convey
events and participants in those events. The only exceptions are
the pronouns it, I and me, the possessive determiner my, the
preposition in, and the verb inection -ing. These are usually
amongst the earliest function morphemes to emerge.
These rst signs of function morphemes are often unreliable.
Some children will use a particular item in just one or two phrases
to start with. For example, it may appear only in Do it and Get
it. This suggests that the child is using each of these phrases as a
whole, frozen chunk, and does not yet have it as a form in its own
right. For other children, the earliest signs of function morphemes
are more prolic. They turn up in a wide range of contexts, but
take a very rudimentary form. They consist of weak, unintelligible
syllables which occur in positions where function morphemes
would be expected like Claires [n#n], [s], [], [am], [m]:
Claire, aged 2;2
SC: Shall we look at these?
C: [n#n] in there? (looking at bag of toys)
SC: Ill show you.
C: [u] got?
SC: Huh?
C: [s] in []? [] slide . . . Get out (tries to get swing out of bag) . . .
Cant. Cant . . . Done it! . . . Go . . . Swing . . . [am] head. [m]
bang [] head.
In the wake of these rst glimmers, it is not long before function

morphemes take off. Their forms become distinct and recognisable,
and they are used more extensively and consistently. These devel-
opments are gradual and to a certain extent predictable within
any one language (Berko 1958, Brown 1973, de Villiers & de
Villiers 1973, Slobin 1985, and see Owens 1996 for a very useful
overview).
As Alisons utterances illustrated, the earliest examples in Eng-
lish are:
( certain pronouns (I/me, you, it)
( the odd possessive determiner and pronoun (my/mine)
( the verb inection -ing
These emerge between 12 and 24 months. Later, around 27 to 33
months, nouns will be elaborated with
( the plural marker -s
and a little later still, with
( determiners a/the
The emergence of verb markers is a drawn-out affair. The inection
-ing, which is well in the lead, is followed by sporadic uses of
( certain irregular past forms such as came, went, broke.
These are followed by the use of
( the regular past form -ed
( 3rd-person-present form -s
( auxiliaries have, be, do, can, will
The auxiliary verbs rst appear in contexts where they must take
their full form, for example, in elliptical responses such as I am, He
is, and in negatives cant, dont. This precedes their use in contexts
where they can contract, as in Im, hes. From about 30 months,
the range of auxiliaries grows, with the use of
( other modal verbs such as shall and could
(See Brown 1973, Owens 1996, Fletcher & Garman 1979, 1986
for further details.)
By this stage, the child is producing most obligatory function
morphemes in most contexts which require them, and we perceive
the child as having pretty well reached adult language. Some
children get there as early as 212 to 3 years; others are closer to 4.
So we see that Shel, just four months older than Claire, deploys a
far wider range of function morphemes in her utterances which
sound almost complete and adult:
Shel, aged 2;6
S: Cant shut that (miniature box, in which she has tried to put
playpeople). Oh yes I can . . . Them get out and she little babys
can go in there. Cant get she in.
SC: Huh?
S: Cant get her in.
SC: There. Where do they go?
S: On swing. Push him. Should I push him? Oh.
SC: What happened?
S: Oh, poor little girl! Don let them fall . . . Wheres [] other girls
. . . Look, I found two. Can you bend her legs?
Shels output is still not entirely in line with the English of those
around her. She makes odd errors which are characteristic of
emerging function morphemes. These errors do not usually affect
meaning. Take Shels incorrect use of pronouns:
Them get out . . . Cant get she in.
She puts these pronouns in the right position, so their role in the
sentence is perfectly clear. Its just that the form of the pronouns is
at odds with their position and role. When Shel eventually sorts
out which pronoun goes where and uses each correctly, this will
have no effect whatsoever on the meaning she communicates.
The same applies to overgeneralisations where children use the
regular form of an inection on a word which is exceptional. This
gives rise to child-typical plurals mans, womans, sheeps, tooths or
teeths; and pasts maked, breaked or broked, falled or felled, comed or
camed; and possessives hims, shes, thems.
When children do use a form which is semantically not quite
right, it is likely to be just off target. A good example is the use of
preposition by in place of with:
Can I pick it up by my hands?
Cover me up by my silk
(Tomasello 1987)
Such are the ne-grained errors which dot the slow but predictable
path to function morphemes which most children follow.
Output when the road is blocked

For children identied as language-impaired, the path is a rocky
one and may stop short of its destination. Time and again, these
childrens use of function morphemes has been found to be limited
compared with normally developing children, even when the com-
parison is with younger children matched on some other measure
of language such as their mean length of utterance (MLU). For
example, a large-scale study of 287 language-impaired children
ranging in age from 3;0 to 16;2 (Johnston & Schery 1976) found
that their acquisition of function morphemes was not only way
behind normally developing children of the same age. It was also
behind younger normally developing children of the same language
level. The language-impaired children were not only older but
tended to be at a higher MLU level when they achieved reliable use
of function morphemes. Yet the stage at which they rst used each
morpheme, and the order in which different morphemes emerged,
were much the same as those observed in normally developing
children.
Such disproportionate delays have been observed in studies of
individuals and groups, and of different function morphemes.
These studies have highlighted persistant omission of articles a/
the, plural -s, possessive -s, auxiliary verbs be and have, past tense
-ed, 3rd-person-singular -s (Ingram 1972, Leonard, Bortolini,
Caselli, McGregor & Sabbadini 1992, Leonard, Eyer, Bedore &
Grela 1997, Oetting & Horohov 1997).
Difculties may be life-long. This is evident from a study of three
generations of a family, about half of whom were language-
impaired (Gopnik & Crago 1991). The study focused on six of the
older language-impaired members of the family who could be
assumed to have completed their language development. They
ranged from a grandchild of 16 to a grandmother of 74. Together
with six unimpaired relatives, aged 817, they participated in a
variety of language tasks. On many, the language-impaired mem-
bers of the family did not differ from their unimpaired relatives.
The tasks on which they did differ involved function morphemes.
Presented with the following sort of input:
This is a zoop (pointing to an imaginary animal)
These are ? (pointing to two or three of these animals)
the language-impaired members were less able to create a plural

form for the word they had not met before. Similarly, they were
less able to generate auxiliary verbs and inections in response to
lead-ins such as
Table 11.1. Difculties with function morphemes reported for
languages other than English
Noun-related Verb-related
Germana Omission of articles Errors in marking of verbs

Errors in case marking of to agree with subject
nouns (through use of verbs in
Errors in marking of an uninected, innitive,
articles and adjectives to or incorrectly inected
agree with nouns form)
Omission of auxiliaries
Italianb Omission of articles Errors in marking of verbs
to agree with
3rd-person-plural subject
Hebrewc Omission of marker of
direct object
aClahsen1989
bLeonardet al. 1992, Bortolini, Caselli & Leonard 1997
cRom & Leonard 1990, Dromi, Leonard & Shteiman 1993
Every day he walks eight miles

Yesterday he ?
The boy always cries

Right now he ?
Yet function morphemes are not all equally problematic. Com-

parison between languages reveals that while function morphemes
are commonly affected, the marker for a particular function may
be more affected in one language than another. Table 11.1 draws
together some of the cross-linguistic evidence.
It turns out that English-speaking children have greater dif-
culty with plural and 3rd-person-singular markers than their
Italian- or Hebrew-speaking counterparts (Leonard 1992).
Italian-speaking children struggle more with articles than He-
brew-speaking children. Hebrew-speaking children, it seems, have
an easier time all round with function morphemes. When com-
pared with MLU matches, they show no special difculty with
present and past tense inections on verbs, number inections on
nouns, or gender inections on adjectives; the only function mor-
pheme which gives them more trouble is the marker which He-
brew requires before a denite direct object (Dromi, Leonard &
Shteiman 1993).
Comparison within languages also reveals variable levels of
difculty. In English, the 3rd-person-singular marker -s on verbs is
more affected than the plural marker -s on nouns. For example, a
group of 4- to 6-year-old language-impaired children showed
some delay in their use of plural -s compared with 2- to 4-year-old
normally developing children matched on MLU, but the discrep-
ancy was greater for the 3rd-person -s (Rice & Oetting 1993). In
Italian, a language-impaired group produced plurals as reliably as
their language matches, but not determiners (Leonard et al.
1992). In German, children were observed to inect verbs correctly
for past participles, but not for agreement with the person and
number of the subject (Clahsen 1989).
Whether they are more or less acutely affected, it is striking that
function morphemes are not entirely absent. What is notable is the
unreliability of their use. Children may, for example, use plural -s or
past tense -ed in some or even many of the contexts requiring
them, yet still not attain the reliable use of the normal 3 to
4-year-old. They may even be inconsistent within a single lexical
context, producing the same word both with and without a re-
quired function morpheme. An analysis of spontaneous speech
samples produced by nine language-impaired children revealed
that 50 per cent of the words they produced in two or more
contexts requiring an inection were produced at least once with
that inection and at least once without it (Leonard et al. 1997).
Every one of these children exhibited inconsistency with at least
some words.
This nding is important. It rules out the possibility that the
children have acquired individual inected items as wholes, for
example balloons for a group of balloons or walked for a past
occurrence of walking, without separating out the inections and
their plural or past function. If they had done this, they would use
those specic inected forms consistently, but would not use other
similarly inected forms.
Even stronger evidence that they have separated out the inec-
tion rather than acquiring it as a xed part of a word is that they
produce inected forms which they have not met. They sometimes
use overregularised forms which they cannot have heard. They
produce past tense forms such as throwed and catched, and plural
forms such as peoples and mans (Rice & Oetting 1993, Oetting &
Horohov 1997, Leonard et al. 1997). They also show novel use of
function morphemes when they add them to nonsense words.
Such inected forms are elicited by introducing children to a
nonsense word and then asking them to complete a sentence
containing that word:
Here is a boy and here is a nep. See, a boy and a nep. This is the boys
food and here is the . . .
(Leonard et al. 1997)
Nine language-impaired children did produce the required func-

tion morpheme with a nonsense word on at least some occasions.
These novel uses are evidence that the childs acquisition of func-
tion morphemes is not piecemeal or frozen. It is just slow and
unreliable, trailing even further behind their acquisition of content
words.
Input
Considering the spotlight that has fallen on function mor-
phemes in output, surprisingly little has been said about what
children do with function morphemes in input. This is probably
because it is much harder to track childrens recognition and
understanding of function morphemes than their production. The
problem is that most function morphemes do not make reference,
as noted above, so they are largely redundant to communication
in context. This means that children could make good headway in
understanding everyday utterances even if they didnt understand
the function morphemes. We may try to get round this problem by
testing comprehension of function morphemes outside of context
only to encounter other problems. How do we design a compre-
hension test for items which cannot be represented by pictures or
actions?
A minority of function morphemes do convey referential mean-
ing. Prepositions such as in, on, under, behind, for instance,
may be used to make reference to spatial relations which are
observable in the outside world. In the case of these items, experi-
ments can easily be designed to test childrens understanding. We
can ask children to put the doll in/on/under/behind the box and
see if the child does this. Judging from this sort of evidence,
normally developing childrens comprehension of spatial preposi-
tions extending from in and on through to above, below, in
front of, behind unfolds between around 18 months and 4
years. (See Owens 1996 for an overview.)
The majority of function morphemes are less easy to test. The
time relations expressed by -s, -ed and -ing cannot be depicted by an
action or picture, nor can the specicity of reference expressed by
the. To nd out whether children recognise and understand these
forms, we have to nd ways of probing childrens reactions to
function morphemes indirectly. Techniques which do this have
produced gems of evidence indicating that normally developing
and language-impaired children may be more sensitive to function
morphemes in input than their output would have us believe.
These techniques do not elicit conscious judgement or full
interpretation of the function morpheme. Instead, they check
childrens responses to sentences with and without correct func-
tion morphemes to see if they react to these differently. If they do,
they must be registering the function morphemes. In one study,
children were presented with sentences which included obligatory
function morphemes:
Throw me the ball!
and sentences which, like their own, were telegraphic:

Throw ball!
or
Ball!
(Shipley, Smith & Gleitman 1969)
Children whose own utterances were mostly a single word obeyed

the shorter child-like commands more frequently than the longer,
well-formed commands. However, children whose own utterances
were just a bit more advanced mostly two words or less obeyed
the longer, well-formed commands more often than the shorter,
telegraphic commands. This shows that children who are not yet
using certain function words may nevertheless distinguish and
prefer sentences which include them.
More rened probing reveals childrens sensitivity not only to
whether a function word occurs, but to which function word
occurs where. Children are presented with sentences containing
an appropriate function word and sentences containing a different
and inappropriate one:
Find the dog for me (appropriate: determiner the before noun)
Find was dog for me (inappropriate: auxiliary was before noun)
(Gerken 1996)
Children of 2128 months did show a different response to these

different inputs. Some of the children those whose own utteran-
ces had a mean length of over 1.5 morphemes were better able to
point to the correct picture for the noun following a determiner
(appropriate) than following an auxiliary (inappropriate).
What about language-impaired children whose own produc-
tion of function morphemes is unreliable? Will they respond differ-
ently to correct and incorrect use of those function morphemes? To
follow up this question, a group of language-impaired children
were presented with instructions containing appropriate or inap-
propriate use of function morphemes:
Find the bird for me (appropriate)
Find bird for me (omitted)
Find was bird for me (inappropriate)
Find gub bird for me (nonsense)
(McNamara, Carter, McIntosh & Gerken 1998)
The children were signicantly better at selecting the correct

picture when they heard correct the than incorrect was. They were
less affected by the omission of the function morpheme and its
substitution with gub. These ndings show that the children were
at least sensitive to the form the and its context. If they were not,
they would respond to the appropriate and inappropriate instruc-
tions in the same way.
Where children react differently to the determiner and the
auxiliary, they clearly know the form of the function word which
goes with the noun. They dont necessarily know the meaning. This
brings us back to the difculty of tapping childrens knowledge of
abstract meanings. We can overcome this only by devising situ-
ations where that abstract meaning has repercussions for the
concrete interpretation of a sentence. This can happen with deter-
miners, whose function is to identify which thing a noun refers to
in a particular context. Names such as Susan or Fred or Fido
inherently refer to a particular individual rather than a class of
individuals, so do not normally occur with a determiner:
*The Susan
*A Fido
The presence or absence of a determiner can therefore indicate

whether an unknown word is a name or a common noun:
This is dax (taken to be a name)
This is a dax (taken to be a common noun)
Do children interpret these differently?

This has been checked out experimentally (Katz, Baker & Mac-
namara 1974). Two-year-old normally developing children were
introduced to toys they had not met before either small plastic
dolls or large plastic blocks. Some children received the lead-in:
This is dax
and were subsequently asked to point to dax. Others received the

lead-in
This is a dax
and were asked to point to a dax. Girls as young as 1724

months distinguished the name from the noun when it referred to
a doll. Those who met dax were more likely to point to the same
doll they had been introduced to, rather than a different doll.
Those who met a dax were as likely to point to a different doll.
This shows they already know that dolls can be labelled as individ-
uals or as a class. More than this, they must know that the
determiner signals the class of dolls rather than the name of the
doll. They did not show this distinction with blocks, presumably
because they do not expect blocks to be identied as unique
individuals. (Two-year-old boys showed no difference with dolls or
blocks. Either they were insensitive to the possibility of dolls being
individualised like people reecting a quite plausible gender bias!
Or they were less linguistically sophisticated than girls and did not
appreciate the contribution of the determiner.)
Such experiments have not yet been tried with language-im-
paired children. Evidence of their comprehension of function mor-
phemes remains sparse. Anecdotal reports of their misunderstand-
ings often involve confusion over matters such as time which can
crucially depend on function morphemes. When they fall down on
clinical tests of comprehension, it may well be that items which
rely on function morphemes are amongst those which trip them
up. Limited data from the Italian group of language-impaired
children (Leonard et al. 1992) suggest that their difculties in
output may indeed be parallelled in input. When these childrens
comprehension of function morphemes was tested, they performed
poorly on the same two morphemes which they produced less
reliably than their language-matches. Poor comprehension,
though, does not imply no comprehension. On the contrary, the
Italian language-impaired children scored above chance on all but
one of the function morphemes tested. This suggests that the
children could recognise their form and function.
It is possible that a speaker may recognise and understand
function morphemes, but not realise that they are obligatory. This
aspect of function morphemes is picked up in the study of three
generations mentioned above (Gopnik & Crago 1991). The lan-
guage-impaired members of the family understood plural -s in
instructions such as:
Please touch the balloons
Put the crayon on the balloons
Here, their scores were on a par with their unimpaired relatives.

But they were much less able when it came to making judgements
about sentences such as
The boy eats three cookie
The little girl is play with her doll
where the judgement hinges on the plural marker or a verb

inection. They were also less able to correct the errors in such
sentences. They often left the error unaltered and changed some-
thing that was already correct, for example turning
The boy eats three cookie
into
The boy eats four cookie
While the language-impaired individuals clearly recognised and

understood the plural form, they did not recognise that its marking
is obligatory, in contrast to their intact relatives who did.
Stretched and overstretched

Unlike most children who progress to full and reliable use of
function morphemes, some children do not make it all the way.
The processes by which children normally acquire function mor-
phemes are somehow falling short. To nd out more, we need to
consider what is involved in those processes.
Consider their starting-point. Children are not handed function
morphemes on a plate. On the contrary, function morphemes
come welded to content words, in whole utterances relating to
whole scenes. Children must discover these morphemes using
their own resources for unpicking the complex information they
receive. The patterns of emergence of function morphemes within
a single language, and across widely different languages, are
beginning to throw light on the nuts and bolts of this process.
Cues from phonology

Young children have acute antennae for the complex rhythms
of the speech they hear, as we have seen. Early on, they pick out
the rhythmic peaks which highlight the key content words. They
rarely make errors of word segmentation. They rarely use a con-
tent word together with a function morpheme but with the mean-
ing of just the content word, an error which would suggest they
had not separated out the function morpheme. Function mor-
phemes are more likely missing from their early speech, or reduced
to some neutral and unrecognisable form. When they have con-
solidated a good number of content words, they move on to the
phonological forms which embellish those now familiar content
word forms. What prompts this move?
The ball probably starts rolling with children beginning to
notice the phonological snippets whose detail they have previously
ignored or skimmed over. Their increasing sensitivity to the pho-
nological forms may trigger their search for the function which
these subsidiary forms fulll. This possibility is in keeping with
evidence of the ways in which phonology inuences childrens
acquisition of function morphemes.
Studies of language development in English have indicated that
children tune into unstressed forms which follow the stress in a
rhythmic unit more than those which precede the stress. As we
saw in chapter 2, this is true within words, where children are
much more likely to leave out an unstressed syllable which preced-
es stress than one which follows it. Hence, the typical
# # # #
raffe, mato, tato, jamas
for
# # # #
gi raffe, to mato, po tato, py jamas
As further evidence, when 2-year-olds are introduced to novel
four-syllable forms such as
zam pa ka sis
they reproduce these forms differently according to where the
stress occurs (Gerken 1996). Where a weak syllable precedes the
stress, for example pa in
pa ZAM ka SIS
children omit that syllable. Where it follows the stress as in
ZAM pa ka SIS
it is preserved, at the cost of the subsequent weak syllable.
Likewise, when children produce word combinations, they are
more likely to omit or collapse an unstressed word which precedes a
stressed word than one which follows. They are very likely to miss
or blur a preposition such as on in:
Draw paper
or
Draw [] paper
yet produce it clearly in:

Shoe on
Light on
The repercussions of this phonological bias on the acquisition of

function morphemes are evident. We nd that function mor-
phemes which have the same function appear at different stages
depending on their position. This emerges from comparisons be-
tween children acquiring different languages. Take the case of
determiners which precede the noun in English:
Determiner Noun, as in a table, the table
but follow it in Bulgarian:
Noun Determiner
Children acquiring Bulgarian use determiners earlier than their

English-learning counterparts. An even more striking contrast has
been cited from the speech of bilingual children acquiring Hungar-
ian and Serbo-Croatian. Serbo-Croatian marks spatial relations
like in and on with prepositions which precede the noun (as
does English):
Spatial marker Noun
In Hungarian, on the other hand, such spatial relations are

marked by sufxes, word endings which follow the noun:
Noun Spatial marker
Though these children are acquiring Serbo-Croatian and Hungar-

ian simultaneously, they use the spatial sufxes of Hungarian in
advance of the spatial prepositions of Serbo-Croatian. They have
clearly grasped the notion of spatial relations, since they mark
these in Hungarian. The hold-up in Serbo-Croatian must therefore
reect the childrens greater difculty in getting hold of the form. It
seems that the form following the content word is easier for them
to get hold of. (Examples taken from Slobin 1973.)
The greater salience of phonological forms which follow the
stress can to some extent account for the road to function mor-
phemes which we have observed in children acquiring English.
The plural marker, a sufx following the noun, emerges before
determiners, which precede it. Verb inections -ing and -ed, which
follow the verb, are acquired earlier than the accompanying aux-
iliaries be and have which precede the verb.
But such phonological biases cannot be the whole story. We
can, after all, nd examples of unstressed words which precede the
stress, and so should be harder for the child, emerging earlier than
others which follow the stress and should be easier. In and on,
which are prepositions, emerge relatively early, and before some of
the inections which follow the verb.
The use of cues from phonology, commonly known as phono-
logical bootstrapping, may launch the child on the road to func-
tion morphemes. But other forces will inuence the direction that
road takes.
Cues from meaning

Children are initially focused on things and people, and their
actions, states, location and possession in the here-and-now, as we
have seen. We might expect their sensitivity to relations in the
world to be a launching pad for discovering what function mor-
phemes contribute to the meaning of sentences. Is this what
happens?
Consider the morphemes which take the lead in English: -ing,
in/on, I/me, my, mine. Each of these relates to one of the
salient aspects of the here-and-now scene. -ing identies events
as on-going, so when it is used in the present, it applies to events
which are directly available to the child in the here-and-now.
Early uses of in/on represent the rare case of function morphemes
with referential meaning: they refer to salient spatial relations
between things. In contrast, prepositions which apply to less sali-
ent space relations, time relations and relations of an even more
abstract nature are not yet used. I/me/my/mine identify the
speaker, the most salient role in the communication event, who of
course coincides with the child herself when she is speaking.
My/mine also identify possession which is in itself a salient rela-
tion, but especially so when the possessor or would-be possessor is
the child!
Turning to later acquired function morphemes, these identify
less salient aspects of scenes, or relations which have no direct
reection in the scene. Determiners a/the limit the reference of a
noun by indicating whether or not it is a specic entity already
focused in the context. Tenses -ed/-s and auxiliary verbs such as
will/can/could/would mark an event as outside the here-and-
now past, habitual, future, hypothetical, and more.
The slowest-to-be-acquired function morphemes not only mark
notions which are less accessible. They conate two or more such
notions, as in the case of auxiliary verbs which convey complex
information about timing of the event and mark agreement of the
verb with the subject:
He is eating present tense + 3rd person singular
They are eating present tense + 3rd person plural
He was eating past tense + 3rd person singular
On top of all this, these auxiliary verbs play a role in marking
questions and negation:
Is he eating?
He isnt eating
Phonological cues rst

Function morphemes, it seems, are more or less accessible to
the child depending
( how closely they t the childs phonological biases
( how closely they t the childs semantic biases
The most easily accessible and earliest acquired are those which
have consistent phonological forms, occur in phonologically sali-
ent positions, and pick out a single salient aspect of here-and-now
scenes. The effect of both phonological and semantic factors indi-
cates that the child draws on both semantic and phonological
bootstraps. This is very much in line with our conclusion about
childrens acquisition of verb structure: that it must result from a
coalition of bootstraps. However, when it comes to function mor-
phemes, the partners in this coalition appear far from equal.
This is not surprising. After all, function morphemes of different
languages pick on different aspects of entities and events. Children
cannot know in advance which aspects will be obligatorily marked
in their particular language. They cannot know, for example, that
English insists on distinguishing past events according to whether
their occurrence at a specic time in the past (fell) or their effect at
a later time (has fallen) is being pinpointed. The only evidence
available to children at the outset is the different linguistic forms
they encounter. The English-acquiring child encounters verbs
with have and verbs without, both referring to events in the past.
This can alert the child to look for a difference between the two
forms for past reference, and notice that has fallen occurs in scenes
where the focus is on the effect following from the event. At this
point, semantic bootstrapping comes into play: childrens sensitiv-
ity to different aspects of the scene will inuence their discovery of
the feature which goes hand in hand with the form.
Where function morphemes have no direct relation to the
context, the argument for phonological bootstrapping is even
stronger. Take the case of agreement markers. These have no
signicance in themselves, so they cannot be reached by semantic
bootstrapping. The only cue to agreement children receive is the
phonological variation which occurs in the word which agrees.
For example, where verbs are marked for number, the distinction
between one and more than one does not arise from the event to
which the verb refers. It is the subject of the verb which is one or
more than one, and the marker on the verb does no more than
echo this property of its subject. Only the phonological variation in
the verb form can cue the child to this subjectverb agreement.
Logically, phonology must drive the search for the function of
these forms. Actual evidence of the way they emerge supports the
case. Children easily pick up phonological patterns which have no
semantic correlate. For example, where gender carries no mean-
ing, as is the case with nouns which are inanimate and do not
come in male or female guise, children readily negotiate the pho-
nological patterns associated with it. In languages as diverse as
Hebrew, French and Polish, they acquire gender between the ages
of 2 and 3, appropriately matching the phonology of determiners
or noun inections to the noun. In some languages, such as
Russian, gender marking is slower to emerge. The difference is that
gender marking is less distinctive and less consistent in Russian,
making it harder for children to sort out the phonological patterns
(Slobin 1985).
The strength of childrens phonological bootstraps is such that
phonological cues to gender can take priority even when semantic
cues are available. This is illustrated by an experiment carried out
with French-speaking children which pitted phonological and
semantic cues against each other (Karmiloff-Smith 1978).
Children were presented with pictures of martian-like persons of
female or male appearance. These characters were given novel
names with typically masculine or typically feminine endings.
How would children react when the gender of the character in the
picture and the gender marking of the word were in conict: when
males were given feminine-sounding names, and vice versa? They
might use the semantic information from the picture, and assign
masculine gender to males. Or they might use the phonological
information from the word ending, and take the name for the
males to be feminine. The nding was that phonology won out
over semantics. Children as young as 3 to 4 determined gender
from the name. The mismatch between their gender marking and
the gender in the picture apparently did not trouble them.
If phonology is the rst cue to function morphemes, we might
expect these to be a problem for children to whom phonology does
not come so naturally. A special difculty with items of relatively
short duration has been put forward as a possible explanation for
the vulnerability of function morphemes in language-impaired
children (see, for example, Leonard & Eyer 1996). According to
this explanation, function morphemes which take the form of
syllables which are unstressed or precede the stress, or of single
consonants, should be most affected. This is consistent with the
differences between language-impaired children acquiring differ-
ent languages which were noted above. We saw, for example, that
language-impaired children acquiring Italian showed much less
problem with verb and noun inections than their English
counterparts. The inections which the Italian-speaking children
produce as reliably as children matched on MLU are precisely
those which are phonologically more robust. They are phonologi-
cally stronger than the inections which so challenge English-
speaking children.
The interaction between phonological and

semantic/syntactic bootstrapping
Phonological bootstrapping is clearly a necessary and powerful
force in the acquisition of function morphemes, driving the process
of discovering the semantic and syntactic contexts which deter-
mine their occurrence. But noticing the phonological left-overs of
content words is not enough to see this process through. Children
must also have some idea of the sorts of functions these forms
fulll. If they were not pre-set to notice the co-occurrence of these
phonological forms with certain sorts of semantic or syntactic
context, it is hard to see how they would ever hit on their specic
function and know when they are required. Phonological boot-
straps may kick-start the process, but semantic and syntactic
bootstraps must kick in. The ease with which children acquire a
function morpheme depends on the salience of its semantic/syn-
tactic function and the salience of its phonology, but also on the
connection between these.
Phonological forms or patterns which are distinct and consistent
for a particular function are picked up more easily than those
which are less distinct or vary. This accounts for the observation
that children acquiring Russian take much longer to establish
gender than those acquiring Polish who, as we saw in the previous
section, achieve this at an early age. In Russian, the phonological
marking of gender is notably less distinct and consistent than in
Polish.
The leap from phonological form to function is easier where
phonological links mirror semantic links than where they are
mismatched. Function morphemes which are attached to the
relevant content word are established earlier than those which are
syntactically remote from the relevant content word. Take forms
which mark deniteness. Deniteness is the notion conveyed by
the English determiner the (as opposed to a). Its function is to
identify the reference of the noun. The determiner marking de-
niteness is separate from but closely connected to the noun. In
Hungarian, verbs are marked according to the deniteness of their
object. Here, the marker of deniteness is attached to a verb, but is
not semantically relevant to the verb. This marker of deniteness
on the verb poses a much greater problem for children than
markers for deniteness which are connected to nouns, including
the determiners of English, German, Romance languages, the
noun prex of Hebrew and the noun sufx of Turkish. (See Slobin
1985 for extensive examples and discussion.)
The connection between the phonology and semantics of func-
tion morphemes, as well as their phonological and semantic sali-
ence, determines how far they stretch the childs bootstraps.
Working through the complex processes which go into the acqui-

sition of function morphemes, it becomes clear that limitations on
the use of function morphemes which we observe in some children
could have many different sources. To illustrate how we may
explore such limitations, we turn to the case of Ruth.
12 That one not working, see:
problems with auxiliary
verb processing
Ruth is a spirited 10-year-old who loves horse riding and wants to

work in a stable when she grows up. She has lively exchanges with
her class mates and teacher, chats enthusiastically about school
and home activities, jokes, fantasises, gets embarrassed, has fun
putting down friends or family, throws side-long glances at adults
when she knows they are discussing her, is both keen and shy to
take the stage. She is observant and sensitive and initiates conver-
sation in just the ways you would expect of a girl approaching
adolescence:
(Ruth has just come into the room)
R: I know, your hair cut.
SC: Im going to have one? Or I have had one?
R: You have one.
SC: Yeah. I did have one. Its too short, isnt it? But it will grow.
R: Like me.
Ruths open and interested demeanour only clouds over and closes
up when her difculties with language come into focus. At this
point, she is likely to re angry and conversation-closing words:
SC: Yeah, your hairs getting long. Who cuts your hair?
R: Tressers (rather distorted).
SC: Tracys?
R: Nothing!! [n] your business.
SC: (catching on, but too late) Oh, hairdressers! Like me.
Once her difculties are touched on, and especially if they become
apparent in public, she is likely either to withdraw or to lash out
206
Problems with auxiliary verb processing 207
with expressions of rejection. These negative feelings also surface
in role-play games which are in her control. She often acts out
aggressive fantasies through the roles she creates:
(Ruth has SC working for her, and after ordering her about, nds ways
of ensuring that her employee cannot get away)
#
R: There is [ bz] bar.
SC: Huh?
#
R: [ mz] bar there.
SC: Bar?
R: Yeah me-tal bar.
SC: Oh, a metal bar.
R: Yeah [sdn] there. And there too. [du] cant get out . . . [zu]
scape, [] f we nd you. You come back here, no food.
SC: Oh no! Will I get food now?
R: No.
SC: If I dont escape, can I have food?
R: Think about that.
SC: What will happen if I have no food?
R: Dead. No sweat . . .
Such negative or aggressive reactions are hardly surprising. The

gulf between what Ruth wants to express and what she is able to
express intelligibly would be enough to drive a person to distrac-
tion.
Ruths language is severely impaired. She produces a ow of
speech which, at a distance, might sound like normal English.
Closer up, the impression changes dramatically. Many utterances
are partially or entirely unintelligible. Many are fragmented. In
contrast to her very disrupted output, Ruth seems to understand
language well, at least in context. Up to the age of 5, her perform-
ance on a comprehension test (the Reynell Developmental Lan-
guage Scales) was normal. At 5, her comprehension dipped rela-
tive to her peers, and appeared to plateau at this point. At 10 her
comprehension, as assessed on the Test for the Reception of Gram-
mar, is at the level of a 6-year-old. It looks as if Ruths input skills
are in some way limited, but they see her through everyday life
quite well and outstrip her output skills by a long shot.
Focusing on Ruths output, the more detailed impression is of
easily recognisable words interspersed with phonological chunks
which are unrecognisable but sound like English. Where Ruth
produces enough recognisable words, we observe that she orders
them appropriately for her intended meaning. She also assigns
them their appropriate stress, producing a rhythmically appropri-
ate utterance:
I kafter you, you kafter me ( = look after)
Told you I got lots
Pretend you don know me
Stay under water long time, that hurt my eyes
To this extent, Ruths utterances appear to be syntactically and
phonologically on track.
Because Ruth produces so much unintelligible speech, calculat-
ing her mean length of utterance (MLU) (following Brown
1973) is not at all straightforward. Do we include in our count
every form we judge to be an attempt at a word? Or do we count
only those which are recognisable? Alternatively, do we avoid
such decisions by using only utterances in which every word is
recognisable? For the purposes of rough comparison with nor-
mally developing children, Ruths MLU was calculated counting
only the recognisable words in the usual sample of 100 consecu-
tive utterances. Using this rather ungenerous measure, Ruths
MLU came out at 3.7, with utterances ranging from one to nine
morphemes. This MLU is normally characteristic of a 3- to 312-
year-old. It is obvious that Ruths output is very different from a
typical child of either her age or her MLU. She has a profound
problem with words.
Sometimes, the intended target of her unintelligible forms can
be worked out from the context:
#
No inerbrup [n n brp]
uttered when Ruth is in the full ow of a story and the listener
cuts in to seek clarication is surely an attempt at
Dont interrupt
Sometimes Ruths unintelligible form sounds very like a word
which has just been used and which we can assume to be her
target:
SC: Nobody can see it. Its invisible (referring to an imaginary aero-
plane).
#
R: Yeah, [ v'zl klen]
SC: Huh?
#
R: [ vvbl] plane.
Sometimes, with repeated attempts, Ruth arrives at a recognisable
target. This happened in the conversation cited above: from
# #
[ bz] bar she progressed to [ mz] bar and then hit the target
with metal bar.
Since we can sometimes identify specic targets for Ruths
unrecognisable forms, we can be sure that at least some of the time
Ruth is aiming for particular English words even when what we
are hearing is a stream of unintelligible phonology. In so doing, we
discover that words of all sorts are a problem for Ruth. Her distorted
forms include nouns, verbs and adjectives (plane, interrupt, invis-
ible). And when we compare these with their known targets, we
discover that Ruth distorts content words in particular ways:
1. While they usually preserve stress and number of syllables,
unstressed syllables are sometimes omitted or reduced, especially
where they precede stress:
#
disgusting ; [ 'stn]
# # # #
invisible ; [ v'zl], [ vvbl], [ vidbl], [ vizbl]
# # # #
injection ; [pr n], [ dns], [ sdn], [ d'n]
2. Consonant clusters are often reduced:
#
problem ; [ pblm]
next ; [n]
3. Within syllables, individual vowels or consonants are often
omitted, added or replaced with a closely related sound:
stay ; [s]
#
luggage ; [ l'wz]
#
festival ; [ fsbl]
In all these cases, Ruth has a specic word meaning in mind. Only
the phonology of the word is obscured often out of all recognition.
If content words are so affected, we would certainly expect
function morphemes to suffer. First impressions of Ruths speech
indicate that they do. Function morphemes are obligatory in
certain contexts and where the context is clear enough from the
scene or from the words Ruth has used, omissions and distortions
of function morphemes are apparent:
Ruth Likely intention
Me borrow mum camera us go Ill borrow mums camera when
out we go out
I ring you last time I rang you last time
We walk up We walked up
You and me getting married You and me are getting married
[wnw] nished, we play the When weve nished, well play
game the game
[tz] got my telephone Have you got my telephone
number? number?
These examples illustrate problems with pronouns, possessive -s,

tense and auxiliaries, most of which would be well established in a
normally developing child with Ruths MLU.
Where do Ruths problems with function morphemes lie? Does
she have problems recognising or producing the form of function
morphemes? Or is their function the problem?
We will set about exploring these questions through the par-
ticular case of tense and auxiliaries. We will start by looking at
Ruths spontaneous output.
Ruths production of auxiliary verbs and inections is inconsist-
ent, but some patterns emerge within the inconsistency. On some
occasions, Ruth produces recognisable auxiliary verbs and inec-
tions, including those listed in table 12.1.
Where Ruth does not produce obligatory auxiliary elements,
these are either entirely absent, or some unintelligible form ap-
pears where the auxiliary would be expected (see table 12.2).
In questions and negatives, whose structure depends on aux-
iliaries, Ruth very rarely produces auxiliaries correctly. In ques-
tions, an auxiliary should precede the subject Noun Phrase. This is
sometimes omitted:
You go [] church?
More commonly, an unintelligible weak form turns up just where

the auxiliary and a subject pronoun are called for:
Table 12.1. Ruths correct use of auxiliaries: some examples
be + present have + past

participle participle can/will Past tense
Pretend youre Ive been on She can eat [m] hated

tickling me television sweets food on []
before plane
You can see me
These are Ive grown up You said no!
hurting me
Hell help with
me
I am She will like it
dreaming
Table 12.2. Ruths omission of auxiliaries: some examples
be gonna
or modal
such as
be have can/will Past tense
You and me Oh no you Her come in I ring you last

getting married gotten some time time
This is my We walk up
Ian what
borrowed
Us going on Your baby
Friday come out soon
Here I do it for
you
[d] go home now? ( = shall we)

[d] type that for me please ( = will you)
[] got a husband ( = have you)
In negatives, an auxiliary should precede the negative particle nt

or not. Again, the auxiliary is often lost:
That one not working, see ( = isnt/s not)
She no coming back in ( = isnt/s not)
Strangely, the negative particle is sometimes lost as well:

Andrew got any telephone number yet ( = hasnt)
Nanny got any teeth up [] top got any teeth down [] bottom
( = hasnt)
Unintelligible substitutions which squash a subject pronoun,

auxiliary and negative particle are also common:
[am] ask her yet ( = I havent)
[an] telling you ( = Im not)
[amn] steal things, no ( = I dont)
Ruths only reliable use of auxiliaries is in elliptical re-

sponses, where the auxiliary stands for a previous Verb Phrase
and occurs at the end of the utterance:
Oh yes you are
She tickling me, I not go on [] oor. She can!
But here, where an auxiliary is always included, it is not always

the correct one for the context. The auxiliary in the elliptical
response should match the utterance to which it is a response, and
sometimes does not:
(Shall we go out now?) ; Course we are
(I do like you really) ; No you wont
(Is she at home?) ; Yeah, in the mornings shes does
In summary, the pattern which emerges is:

1. inconsistent use of be, have, a limited range of modals, and
tense/participle inections
2. unintelligible substitutions, especially in questions and
negatives
3. auxiliary substitutions in elliptical responses
What is stopping Ruth from achieving consistent use of the full
range of auxiliaries?
One possibility is that she struggles with the sort of concepts
that auxiliaries convey the situating of events in relation to time
and reality. Ruth gives us strong reason to think that she does have
these concepts and attempts to convey them.
First, when she does use auxiliary markers, they are almost
always appropriate for the context, suggesting she does under-
stand them.
Second, she sometimes conveys time relations of just the sort
that tense and auxiliaries convey, but using time adverbials (see
Moore & Johnston 1993 for experimental evidence that adverbials
may be less taxing than tense for some language-impaired
children). Ruth uses last time referring to a past event:
I ring you last time
and yet referring to a past event relevant to the present:
[am] ask her yet
Third, the context and content of certain utterances provide
evidence that she has understood just the sort of timing and
contingency relations between events which auxiliaries capture.
Take her impressive attempt at time reference in the following
interchange:
(Will you still know me when youre 15?) ; Forgotten you by now
The verb with which Ruth responds to the question, forgotten,
clearly signals an event after the present time, when Ruth does
know the questioner. The past participle form forgotten clearly
signals an event prior to another event, that of being 15. This adds
up to a reference to a future event which is past relative to a later
future:

Time of utterance Time of forgetting Time of Ruth being 15
It is true that Ruth uses by now rather than by then. But the
implications of her response indicate that neither this confusion,
nor the omission of the auxiliaries Ill have . . . are due to con-
fusion about the time relations between the events.
We nd similar evidence that Ruth talks knowingly about
hypothetical and possible events:
SC: If we leave our bags here, what might happen? If we leave our
money?
R: The people come up [t] [t] ladder [] come in there that
window down that window down to come and pinch your
money. [kt] take it with us. [] case. [t] money with us. []
case.
Here, Ruth conveys two events which are clearly related in her
and our experience of the world someone breaking in, and not
leaving money around. Her linking of these events and her use of
[] case ( = in case) imply the contingency of one event on
another: one event is a possible one and its possibility is the
motivation for the other event. But she does not use any of the
modal verbs, such as might break in and should take our money,
which would normally make explicit the status of these events and
their interrelations. If she has a grasp of these notions and is trying
to convey them, why does she not do so with the modal verbs
which make them explicit? The implication is that she has prob-
lems not with their semantic function, but with their form or with
the connection between their form and function.
We already have solid evidence that Ruth struggles with the
phonological form of words: even when she is targeting a particu-
lar content word semantically, she is liable to distort its phonology.
It may be that she has a similar problem with the phonology of
function morphemes such as auxiliaries. The phonology of auxili-
aries is particularly challenging. They are typically unstressed,
and have weak forms which reduce and contract, though only in
certain contexts (see chapter 11). We would expect auxiliaries to
be especially affected by difculties with phonological forms.
The pattern which has emerged from Ruths spontaneous out-
put points to phonology as a prime suspect. The fact that we
frequently nd unintelligible phonological forms where auxiliaries
should occur often alongside other weak forms suggests Ruth
knows something is required at these points in the rhythm. But she
doesnt know quite what.
A second reason for suspecting phonology is the way that
Ruths use of auxiliaries varies. Her distorted forms occur only in
positions where auxiliaries are unstressed, liable to reduction, and
even merge with other weak forms. For example, the auxiliary and
subject pronoun which are normally subject to reduction or
merger as in
shall we ; [w]
have you ; [vju]
are liable to be inappropriately squeezed in Ruths speech:

[] go home now? ( = shall we)
[] got a husband? ( = have you)
In contrast, Ruths clearest use of auxiliaries is in elliptical re-

sponses, where the auxiliary either carries the sentence stress or
follows it. Both are phonologically salient positions. In these posi-
tions, the auxiliary cannot take its reduced form.
Curiously, in just these contexts where Ruth produces a fully
intelligible auxiliary, we nd her sometimes selecting the wrong
one:
(I do like you really) ; No you wont
In this case, the phonological form is correct and cannot be the

problem. Ruths confusion between auxiliaries must stem from
some other difculty. This looks like evidence against our suspi-
cion that phonology is the primary culprit. But it may not be.
Consider the function of auxiliaries in elliptical responses: to afrm
or deny a state of affairs referred to previously. Semantically, the
auxiliary carries no new information. It simply echoes the auxili-
ary in the utterance to which it responds. This matching of auxili-
aries appears to be the sticking point for Ruth. At rst sight, this
looks like a syntactic problem: Ruth doesnt realise that the auxili-
ary in the response must agree with the original. The identication
of this syntactic connection, however, relies on phonological boot-
strapping. If the child is to extract the pattern of auxiliary match-
ing, she must register the phonological form of the two separate
auxiliaries which match. Any difculty with phonological forms
will surely be an obstacle to discovering connections between
phonological forms which are separated. Such a problem could
therefore account for Ruths errors in elliptical responses.
So, the patterns of auxiliary use in Ruths spontaneous output
are in keeping with our hunch that their phonological forms may
be at the root of her problems. If they are, we should nd that she
has the same problems producing auxiliaries when she does not
have to deal with their function. We can check this by seeing what
she does with auxiliaries when she is repeating rather than
generating utterances.
Repetition
Ruth was asked to make her puppet say sentences, with no
pressure to achieve correct repetition, and no correction of what
she said. Under these conditions, Ruth responds to repetition tasks
readily, and her responses are interesting.
Her repetition is certainly impaired. She was given sentences of
two, three, or four words. With two-word sentences, she has little
problem. No words are omitted, and 90 per cent are reproduced
correctly. As sentence length increases, her repetition deteriorates.
The percentage of correctly repeated words reduces to 80 per cent
in three-word sentences, and 60 per cent in four-word sentences.
These correct words are interspersed with word omissions or with
Ruths by now familiar unintelligible forms. The net effect is that
the utterances Ruth produces in repetition sound very like her
own spontaneously produced utterances. With spontaneous out-
put, we cannot be sure what the target words are. With repetition,
we know exactly what the targets are. Her responses to repetition
therefore offer a clear window onto what she is doing with words.
Looking through that window, we nd that Ruths repetition is
selective. Her response to an utterance which is too long is not
simply to reduce the number of words by, say, repeating only the
rst two or the last two or just any two. This becomes apparent
when we focus on different types of words and see what happens to
them in different positions in the sentence and in sentences of
different lengths. What happens to auxiliary verbs compared with
content words like nouns and verbs?
The repetition stimuli included nouns, verbs and auxiliaries in
varying positions. The position of auxiliaries was varied by placing
them in questions, statements and elliptical structures (see table
12.3).
Ruths responses to these stimuli show that she is not simply
parroting the stimulus sentence, without recourse to her knowl-
Table 12.3. Examples of repetition stimuli
Initial Medial Final
Two-word Is John? John is

Can she? She can
Wont Mary? Sue cant
Doesnt he? He hasnt
Three-word Is John driving? Mary will cry Soon John will
Can she go? They have left Old men do
Four-word Can boys read John will be John knows he
books? coming is
Will Jane leave They must She says she
soon? have left can
Mary will not
cry
He was not
driving
Table 12.4. Percentage of words repeated correctly
Category of target
Position in sentence Noun Verb Auxiliary
Final 99 94 92
Medial 87 53 30
Initial 84 73 44
edge of words. The nal word in a sentence is the one most reliably
repeated, whatever its category. In other positions, nouns are
better preserved than verbs, which are in turn preserved better
than auxiliaries. A summary of all the repetition data shown in
table 12.4 illustrates this.
Focusing on auxiliaries, it is apparent that Ruth can recognise
and reproduce these reliably in two positions. One is in two-word
sentences, where her only problem is with negative auxiliary
targets:
Target Ruth
Wont Mary Not Mary
Mary wont Mary not
Doesnt she [dn] she
The other is the nal position of all stimuli, whatever their length.
In this position, auxiliaries even outstrip preceding nouns and
verbs: see table 12.5. But as soon as they lose this privileged
Table 12.5. Repetition of auxiliaries in sentence-nal position
Target Ruth
Three-word Old men do The old man do

Soon John will John John can
There Jane is Jane is
Four-word John knows he is Knows he is
Paul thinks he will [bl z w]
Anne knows she mustnt [s] mustnt
position, they are extremely vulnerable: see table 12.6. Here, we

nd Ruth replacing them with the same sort of form she would use
in her spontaneous output, for example
auxiliary + pronoun ; [d]
This suggests that she has separated out and recognised the auxili-
ary or the chunk in which it occurs. She occasionally substitutes a
quite distinct form with a related function:
wont ; not
will/was ; is
This suggests that she has processed the word further still, access-
ing at least some aspect of its semantic/syntactic function. These
ndings reinforce the evidence that the difculty with auxiliaries
stems from difculties with their phonology rather than their
semantic or syntactic function.
Is the problem in recognising their phonology or producing it?
Table 12.6. Repetition of auxiliaries in sentence-medial and
sentence-initial position
Position of
auxiliary Target Ruth
Medial Mary will cry Mary [z] cry

She will cry She is cry
They have left [h] left
John is feeling sick John feel sick
She can go now She go now
Mary has been crying [] Marys is crying
They must have left [mz v] left
Initial Will Mary cry Mary [n] cry
Will she cry She [z] cry
Was Bill driving [d] Bill driving
Have they left [dd] left
Is John feeling sick John is feel sick
Can Mary go now [] Mary go now
Has he eaten late [d] eaten late
Will she leave soon [z] leave [] soon
Judgement
In order to check out Ruths ability to recognise auxiliaries, she
was asked to judge sentences. These sentences consisted of either
NounVerbNoun, or NounAuxiliaryVerb. For each correct
sentence, three incorrect sentences were created by replacing each
word in turn with a nonsense word: see table 12.7.
These sentences were presented to Ruth in blocks, each contain-
ing the correct sentences mixed up with one set of the nonsense
sentences.
Ruths responses were not random. She said yes in response to
115 out of 120 correct sentences. In response to the nonsense
sentences, she said no as often as she said yes 62 yes to 58
no. Nor were these yeses and nos randomly spread. When a
nonsense word occurred in the middle or at the end of the sen-
tence, she was quite likely to say no (56 out of 80). When the
Table 12.7. Examples of judgement stimuli
Correct Nonsense substitutes
Cats drink milk [ri] drink milk Cats [r'] milk Cats drink [v]
Birds can y [v] can y Birds [fn] y Birds can [ri]
Bill is eating [k'] is eating Bill [k] eating Bill is [fn]
nonsense word occurred at the beginning of the sentence, on the

other hand, she almost never said no (2 out of 40 times), even
though these nonsense words occupied a noun position. In fact she
made more errors judging a nonsense word occupying a noun slot
at the beginning of the sentence:
[k'] is eating
than a nonsense word occupying an auxiliary slot in the middle of

the sentence:
Bill [k] eating
These responses show clearly that Ruth can distinguish a correct

auxiliary form from a nonsense form, even in the medial position.
They also show that she has greatest difculty noticing unfamiliar
material at the beginning of an utterance.
It looks as if Ruths recognition of auxiliary forms, reected in
her judgement, is better than her production of those forms,
reected in her repetition and spontaneous output. However, our
conclusions must remain tentative, as we do not know whether
Ruth recognises which auxiliary is the correct one, since she was
not asked to judge sentences where an incorrect auxiliary as
opposed to a nonsense word was inserted.
Putting the pieces together

Our nal source of evidence, the judgement tasks, reveals two
things. One, Ruth can recognise auxiliary forms in input. Two, her
capacity for input is limited. She has difculty noticing the begin-
ning of an utterance when it exceeds two words. Both ndings are
in keeping with her output.
The previous source of evidence, the repetition tasks, reveals
three things. One, Ruths capacity for repetition is also limited.
Beyond two words, she comes into problems. Two, Ruth can
recognise and can produce auxiliary forms, since she does so when
they occur at the end of an utterance. Three, Ruth has great
difculty holding onto or reproducing auxiliaries when these are
outside her focus on the end of a sentence. In contrast, she
preserves nouns quite well even when they are at the beginning of
a three- or four-word utterance. This difference indicates that
Ruth is gleaning at least some semantic and syntactic information
from the words in the utterance she is to repeat something she
cannot do with the nonsense words in the judgement tasks. It
must be this further information which strengthens her repetition
of nouns even when they occur in a position on which she is weak
in the judgement task.
Putting the evidence from repetition together with the evidence
from judgement, the picture begins to take shape. The results of
the judgement tasks show that her difculties are not conned to
output. She also has difculties registering forms in input. These
difculties will affect her repetition, since repetition involves input
as well as output. The results of repetition further show that she
registers forms selectively, depending on their position and their
function. These factors inuence which words she produces best,
and which she skips or blurs in her production.
How do these pieces t together with Ruths everyday under-
standing and use of language? Ruths difculties with registering
forms, particularly when they fall outside her focus on the last
word or two of an utterance, have massive implications for her
acquisition of words. The more words occur in positions which
pass her by, the harder it will be for her to notice their phonology,
and the harder it will be for her to work out their meaning and
their syntactic relation to the words with which they combine. We
would predict particular difculties with words such as auxiliaries
which typically occupy weak phonological positions and whose
function is dependent on other words. It should be hard for Ruth to
establish their phonology and to establish the contexts in which
they occur. In contrast, she should have less difculty with the
phonology of words which frequently occur in prominent posi-
tions. She should also have less difculty with words whose mean-
ing is not so dependent on the words with which they combine.
Nouns are most favoured on both these scores. Verbs are less so, as
we saw in our analysis of Traviss difculties with verbs (chapter
10), and the consequences for Ruth would be apparent if we
turned the spotlight onto her verbs.
The disruption in registering phonology and the consequent
disruption in connecting phonology to semantic/syntactic func-
tion will have some impact on Ruths comprehension. However,
she should get quite far with the words which are more accessible
to her, these being the information-carrying ones. And she does.
Her comprehension in context is generally good. Even so, every-
day conversation must often stretch Ruth, and it is not surprising
that misunderstandings and frustration do occur. But only the
high-level demands of comprehension tests reveal clearly that
something is amiss. We dont know exactly where comprehension
falls down, but we might expect it to happen when the interpreta-
tion of an utterance depends on those forms or syntactic relations
we know to be vulnerable.
The impact on production will be much more damaging. To
produce an utterance, Ruth must map her meaning onto the word
forms and structures she has picked up from the input. These
include the markers of meanings or syntactic relations which are
obligatory in English, for example auxiliaries. The difculties Ruth
has had in getting to the functions of these forms will affect the
planning of her own utterances. She will be less predisposed to pick
out features of the context which are obligatorily marked, such as
the timing of an event. This may lead her to mark a time distinc-
tion sporadically but unreliably. Her unstable phonological repre-
sentations of these forms will compound the problem. Even when
she has established a function and plans to mark it, she is likely to
have difculty accessing the correct form. Take the form of her
questions. We can tell Ruth is asking a question thanks to her
questioning intonation and the context of her utterance. She also
appears to know that something at the beginning of a question
differentiates it from a statement. But her phonological representa-
tion of that something is rough, and she comes out with one of her
llers:
[tz] got my telephone number? ( = have you)
In contrast, she has been able to establish solid phonological

representations of the required forms when they occur in strong
positions, in elliptical structures. Here she produces the identical
forms clearly:
You have now
Her difculties with phonology will be highly disruptive for

longer-distance syntactic relations, where the form of one word
depends on the form of another. If it is difcult for Ruth to discover
these interdependencies, it will be even more difcult for her to
reproduce them. This may account for her odd errors in elliptical
responses, where her auxiliary fails to match the auxiliary on
which it depends:
(Ive got to start?) ; Oh yes you are
The picture may be incomplete, but the pieces appear to t

together. Even with severe limitations on her processing of phono-
logical forms, Ruth can make headway with words particularly
words which are phonologically strong. This takes her a long way
in understanding language in context which does not depend on
recognition and comprehension of every function morpheme. The
repercussions are much more evident in Ruths production of
language. Just to get by in production, the speaker must preserve
at least enough phonological information for words to be intelli-
gible to the listener. To sound normal, much more is required: the
speaker must have precise information about meanings and about
syntactic relations which are marked in the language. And she
must have precise information about the phonological forms
which mark these. If the speaker has not secured this information,
it will either be missing or it will be insecurely represented in her
output. It may be that different speakers deal with insecure infor-
mation differently, some tending to skip it altogether, others to ll
in partial information. Ruth appears to do both. Her general
uncertainty about the phonological form of words shows up in her
variable and sometimes unintelligible pronunciation of words. Her
particular difculty with phonologically weak forms shows up in
her omission and distortion of unstressed syllables and unstressed
words, especially those which precede stress, such as auxiliary
verbs. Her difculties in discovering which meanings or relations
must be marked in English show up in her unreliable use of forms
such as auxiliaries.
It seems that Ruths difculties with phonological forms cas-
cade through her language processing. The relatively minor rufes
in her comprehension become tidal waves in her production of
utterances, which can end up as a stream of unintelligible speech.
Hence the yawning and for Ruth painful gap between the rich
meanings she has in mind and the blocks she faces in getting them
out.
Part IV
Hidden meanings, bafing
meanings
MMMM
13 The roots of meaning
Some children use plenty of words. Yet their utterances are in

some way bizarre. Take this conversation with 9-year-old Sally:
SC: What languages do you speak?
S: I speak with Oxford Junior.
SC: You speak with Oxford Junior?
S: Oxford Junior English.
SC: Do you speak Chinese?
S: Yeah I speak Chinese. But English but I cant speak Chinese.
SC: You cant speak Chinese? Or you can?
S: I can speak Chinese but my mum cant speak English. But he can
speak English because but he cant speak Chinese. And she can
speak Chinese too.
Every one of Sallys words is clear enough. The way she combines
these words into sentences is, in certain respects, ne. But taken
together within this dialogue, they leave the listener reeling. What
does Sally mean? Information about her linguistic background
that her parents speak Cantonese while her six brothers and sisters
all speak English does not answer that question.
Meaning: the packaging of experience

Any exploration of a childs problems with meaning in lan-
guage must set out with some handle on what language can
mean. In previous chapters, we have grappled with the meanings
conveyed by different types of words and their integration into
sentences. We have seen that sentences relate to scenes which we
227
228 Hidden meanings, bafing meanings
properties properties
of entities of events
location entities States and events in location

of entities which entities participate of events
possession timing means/goals conditions/causes/effects

of entities of events of events of events
Figure 13.1 Core components of scenes
observe in the world or evoke in our minds. Words focus particular

components of those scenes, from particular perspectives. They fall
into different categories which reect the way we structure scenes
in terms of certain core components, including those shown in
gure 13.1.
But what is the content of these abstract components of scenes?
In what do word meanings consist?
Word meanings are cut from the vast and intricately woven
material of human experience. That material comes in different
qualities which we might conceive of as different tiers of experi-
ence:
( perceptual experience: experience of the senses (seeing,
hearing, feeling, smelling, tasting)
( affective experience: experience of emotions/moods
( mental experience: experience of having something in
mind
These tiers of experience are distinct: no one tier of experience is
the same as, nor can it be reduced to, another. We may be
predisposed to associate experience on one tier with experience on
another. For example, something we experience perceptually:
The roots of meaning 229
may be associated with something we experience affectively: hap-
piness. The perceptual experience may therefore trigger the affec-
tive experience. But registering the perceptual features of the face
above is not the same as registering the affective state of happiness.
Experience on one tier may also provide the content for experience
on another. We can, for example, have an image in our mind, in
which case we have a mental experience of a perceptual experi-
ence. Here, the perceptual experience is embedded in, but still
distinct from, the mental experience.
The points are well illustrated by our everyday experience of an
utterance scene, depicted below:
Imagine you are Y. What do you experience? Perceptually, you see

the conguration of Xs face, including the moving mouth. You
also hear the sounds emanating from Xs mouth. Mentally, you
take X to have something in mind and to be conveying that
something to your mind. Affectively, you take X to be wanting to
convey something to you and also to be feeling anxious. Based on
these core experiences, you are likely to make further connections.
You may, for example, take Xs anxiety state to be the trigger for
Xs intention in speaking (this would of course be inuenced by
what X is saying, which we are ignoring). But however automatic
and intricate the interconnections we make, they stem from tiers
of experience which are ultimately distinct.
Childrens packaging of experience

If words are a packaging of experience, the words children use
can provide a window onto the experiences they have packaged
and the way they have packaged them. Looking through the
window of childrens earliest words, we have seen that from the
earliest stage children take a range of perspectives on the world
around them (chapter 2). It is true that the majority of their
earliest words identify entities, and do so at what is known as the
basic level. They typically come out with nouns such as dog,
rather than poodle or animal; chair rather than armchair or
furniture. But they are not conned to labelling at the basic level,
nor to using only one label for one entity (Clark 1997), nor to
labelling only entities:
( They refer to entities at different levels and from different
perspectives. They may refer to one woman with a specic
term mummy and another woman with the basic-level
term lady. They may refer to a ower with the basic-level
term ower or may use the term that which simply
points it out.
( They sometimes refer to a single entity at different levels
and from different perspectives. They may refer to the
family pet as Fido and doggie, or to the plate in front of
them as dinner or food.
( They sometimes pretend that one thing is another thing
and refer to it by its pretend label as well as its usual
label. They may use a banana as a telephone and call it a
telephone.
( They focus on an entity but refer to its state, location or
possession. They may say hot in relation to food, bro-
ken or mine in relation to a toy.
( They refer to actions, events and relations in which enti-
ties are involved. They may express these with preposi-
tions such as in, on, off , or verbs such as look,
come, push.
The semantic variety already evident in childrens early words is a
seedbed for the breadth and complexity of meaning in language
which they will soon attain. The number of items at this stage may
be small. What is signicant is their range. They span different
components of experience, such as entities, locations, states and
events. They also span different tiers of experience, with some
drawing more on perceptual experience (hot, broken) and some
drawing more on affective interpersonal experience (mummy,
mine, yours).
In order to use such words in appropriate contexts, children
must have picked out the particular experience on which they
focus and must have registered that other speakers were focusing
on that experience when they used that word. Childrens appropri-
ate use of words proves not only their own differentiation of
experiences, but also their sensitivity to others experiences. Their
use of mine illustrates the point particularly clearly. Take the
scenario of a child handing out cups, then pointing to the cup in
front of her and saying Mine. In order to arrive at her appropriate
use of mine, the child must have recognised the experience of
attachment between people and entities about which they say
mine, and this must have resonated with her own experience of
attachment. If she had not identied with this experience of attach-
ment of the speaker to the entity, she would come up with the
wrong meaning for mine. She might incorrectly take mine to
refer to other peoples connection to an entity, but not her own. In
this case, she would use mine to mark possession by her ad-
dressee, but not her own possession. She might say mine when
pointing to something belonging to a child she is talking to.
Alternatively, she may take mine to mark possession by a par-
ticular individual she has heard using it, say her mother. In this
case, she might focus on her mothers nose or coat and say mine.
She may be even further off the mark, and take mine to be a label
for the entity itself rather than a persons relationship to it. She
could then use mine to indicate an entity or all entities which she
has heard other people claim as mine. Childrens normally cor-
rect use of mine, as well as yours, indicates how easily they
recognise that the speaker is focusing on the attachment of
speaker/listener to an entity.
It is true that certain sorts of experience are not represented
among childrens earliest words. While some words may have an
affective dimension (dirty, mine, mummy), children are unlikely to
use words which focus on affective or mental experience. Their use
of words, though, is proof of their sensitivity to affective and
mental experience in themselves and others. They do not utter
words aimlessly and to an empty space (though they may play
games with words when they are on their own). They use utteran-
ces intentionally and direct them towards an addressee. They
request and reject:
Biscuit! Down! No!
and react negatively if they dont get the response they are looking
for. They point out:
Bus!
and check to see that their addressee takes notice of what they are
pointing out. They question:
Dirty? [wst]? [huzt]?
and expect a response. You can only seek to draw a persons

attention to something if you experience that person as having
attention to draw. You can only seek to question a person if you
experience the person as having in mind something which you do
not have in mind. Likewise, if you comprehend these intentions in
what other people say, you must be assuming that they experience
you as having attention and a mind that they can put things into
or take things out of. (See Tomasello 1995 for more extensive
discussion.)
This appreciation of affective and mental perspectives is implicit
in childrens appropriate use of utterances and responses to other
peoples utterances. It is not long before it surfaces explicitly in
their words. By 212 to 3, children start using mental state verbs
such as know, think, remember, pretend, guess (Shatz,
Wellman & Silber 1983). They use them in some cases to indicate
discrepancies between a persons mental state and reality:
Before I thought this was a crocodile; now I know its an alligator
or between one persons mental state and anothers:

I was teasing you; I was pretending cept you didnt know that
(Shatz et al. 1983: 309)
Such uses prove conclusively that the child distinguishes between

direct experience of the world and mental experience.
Over this period, around ages 24, childrens meanings also
become wonderfully diverse. They have words which make ne
differentiations in entities and events, from different types of ani-
mals through different manners of speaking (whispering, shout-
ing) and different types of mental/affective experience (hoping,
dreaming, imagining, guessing) and different affective/mental
goals of speaking (telling, persuading, promising, lying). At the
same time, words extend to components of experience which were
previously unrepresented times, causes, effects and conditions.
Their range of time modiers and auxiliaries grows, and they use
connectors such as when, while, to, because, so, if to
indicate the ways one event may depend on another.
By 4, most children have in some sense acquired all the qualities
of adult meaning. They have the breadth of adult meanings, in
that all the core tiers of experience are represented in their lan-
guage. They have the depth of adult meanings, in that they embed
experiences within experiences. Obviously they do not control the
same range and complexity of meaning as older children and adults.
But later developments involve the elaboration of these core struc-
tures, affecting the variety and subtlety of their packages of mean-
ing, and the scale of embedding of experiences within one another.
What are children doing as they make their way to the vocabu-
lary of the average 4-year-old? We must assume that they can
only package into words what they themselves can experience.
They can only acquire the meaning of look if they understand the
experience of vision, and that human beings each have a visual
perspective. They can only acquire the meaning of happy if they
can separate out the experience of positive affect both for them-
selves and other people. The fact that their earliest words span
different tiers of experience and different components of experience
is evidence that they have registered different sorts of experience
and organised these in ways which are suited to the meanings of
words. That organisation of experience is the starting-point for the
earliest semantic bootstrapping (see chapter 2): it is the impetus
for children to nd words phonological forms for the experien-
ces to which they are sensitive.
At this point, phonological bootstrapping necessarily comes
into play. Only by registering phonological forms which consist-
ently occur with certain experiences can children discover the
particular ways experience is packaged into words in their lan-
guage. Phonological bootstrapping, then, shapes childrens initial
organisations of experience. It will play a more minor role with
words whose meanings most closely match their initial organisa-
tions of experience. Its role becomes increasingly important as
they tackle words which go beyond the sorts of experience on
which they rst focus. Only by noticing the wealth of phonological
forms which occur with similar experiences will the child be
alerted to look for the subtle differences in those experiences which
serve to differentiate the meanings of the words. When it comes to
the packaging of non-perceptual experiences such as time, whose
packaging varies dramatically between languages, phonological
bootstrapping will be crucial (as we saw when we considered the
acquisition of tense and auxiliaries in English).
Once phonological bootstrapping comes into its own, it is a
powerful force in driving the childs extraction of increasingly
differentiated and abstract aspects of experience. In the extreme,
phonological bootstrapping can be used to create meanings when
children have no access to the experience on which their meaning
relies.
Children with experience disabilities

Take the case of blind children. They lack the perceptual experi-
ence which is the focus of words such as the verbs see, look and
show. Yet they come to understand the meanings of such words.
How do they get there? They cannot start from the normally
salient experience of vision and use semantic bootstrapping to
zoom in on words which refer to categories of visual experience.
Instead, when they meet the phonological forms see, look, show,
these must provide the impetus to nd a meaning for visual terms.
The child can then discover what in her experience is focused
whenever she hears these words. The common experience for her
will be in the domain of touch rather than vision, and so the
meaning of these words will be a packaging of experience in this
domain. For her, look will mean explore by hand, even if she
comes to understand that for others it means explore with the
eyes and relates to an experience which her eyes cannot give her.
(See Landau & Gleitman 1985 for much more extensive evidence
and discussion.)
Blindness blocks visual experience. It leaves the childs potential
for other types of experience unaffected. These readily ll in the
gaps in her experience in ways which allow her to deduce the
meanings of vision-focused words.
Suppose it is not the perceptual tiers but the affective and
mental tiers of experience which are disrupted. This will have
dramatic effects on the sense the child makes of any words whose
reference is to things experienced affectively or mentally. If the
child can only approach these words through other tiers of experi-
ence for example, perceptual experience the child will come up
with a partial and oddly orientated meaning for them, or will nd
them meaningless.
But the effects of decits in these core tiers of human experience
may be expected to spread beyond just the meanings of words
which depend on these tiers. Such decits are liable to affect word
meaning more generally. Recall that all words take a perspective on
experience, whatever the tier of that experience. A child can
therefore only discover the meaning of a word if its meaning
coincides with the childs own perspective on experience, referring
to what is most salient to the child at the moment of hearing the
word and/or the child is able to identify the speakers perspective
on experience at the moment of using the word.
This second possibility using the speakers perspective as a cue
to meaning is only open to children if they appreciate that people
have a perspective (see also Tomasello 1995). This is precisely the
issue for children whose difculties lie in the tier of affective/
mental experience. The more limited their ability to recognise
other peoples perspectives, the less they will be able to use other
peoples perspectives as a cue to the meaning of their words, and
the more they will equate words with whatever is salient for them
when they hear a word. They may then come up with partial or
oddly orientated meanings even for words which are not focused
on affective or mental experience.
The effects of decits in these tiers of experience may run even
deeper, striking at the very essence of language. As we saw in our
analysis of the utterance scene above, the use of language is
itself an event drawing on affective and mental experience. It
involves a speaker, who has something in mind and an intention to
convey that something to the mind of a listener. If a childs
capacity for affective and mental tiers of experience is disrupted,
this will strike at language as a tool for communication for
putting something into someones mind or taking something out
of it.
The difculties of children diagnosed as autistic are widely
attributed to decits in affective and mental experience. We nd
autistic children described as profoundly limited in their capacity
for and experience of personal relatedness . . . Their sense of
psychological connectedness to and affective engagement with
other people is seriously impoverished (Hobson 1993: 194). It is
as if their capacity to experience desire for another . . . is either
absent or distorted, and their difculties set limits on the building
of an inner fantasy life (Mayes, Cohen & Klin 1993: 4601). Their
difculties have also been construed as the consequence of mind-
blindness: an unawareness that people have minds and an inabil-
ity to read minds (Baron-Cohen 1995).
Consistent with decits in the experience of emotion and mind
are the oddities which autistic children display in aspects of lan-
guage and uses of language which rest on affective/mental experi-
ence. Commonly recorded observations are:
( Reduced use of terms referring to mental states (Tager-
Flusberg 1993)
( Difculty picking out which words refer to emotion and to
what the mind can do (Hobson 1993, Baron-Cohen,
Ring, Moriarty & Schmitz 1994)
( Confusion of 1st- and 2nd-person pronouns (I/me vs you):
You want candy (for I)
I write (for you)
(Tager-Flusberg 1993)
( Use of a question for a statement/request:
Did we say bye?
Want a cracker?
(Tager-Flusberg 1993)
( Use of language primarily to satisfy a need or desire, with
little use of language to request information or comment
(Hobson 1993)
All of these oddities point to problems in distinguishing the
affective and mental orientation of different individuals. Children
may use terms which normally refer to these experiences, and
their use may even be appropriate to the context. But it may stem
from partial understanding, based on tiers of experience which are
intact. For example, in the absence of affective experience, words
such as happy and sad could be understood by their association
with different visual congurations of the face; enjoy and hate
could be understood by their association with positive or negative
perceptual experiences such as tasting chocolate as opposed to
medicine. Similarly, I and you could be understood as referring
to the one physically speaking, and the one at whom the speech is
directed, without appreciating the sense of own self and other
self.
Some individuals are able to become aware of their difculties in
understanding certain kinds of experience. For example, a 20-
year-old man who cannot grasp what a friend is keeps trying to
nd out, and repeatedly asks Are you a friend? Is he a friend?
(Hobson 1993). A 16-year-old describes a character in a story as
distraught and promptly asks What does distraught mean? (L.
Rosen, personal communication). A young autistic man describes
his experience of people:
I really didnt know there were people until I was seven years old. I
then suddenly realised that there were people. But not like you do. I
still have to remind myself that there are people . . . I never could have
a friend.
(Cohen 1980, cited in Hobson 1993)
Difculties with meaning and use of language are not conned
to children diagnosed as autistic. Some children convey bizarre
meanings and use utterances in odd ways, but do not t the
general prole of autism. Clinically, these children have been
grouped under the umbrella term semanticpragmatic dis-
order (Rapin & Allen 1987, Bishop & Rosenbloom 1987, McTear
1991, Mogford-Bevan & Sadler 1991, Boucher 1998), though
more recently the term pragmatic language impairment has
been proposed as more appropriate (Bishop in press). Typical
features of their language are:
( odd associations and reasoning
( tangential or inappropriate remarks
( undue attention to literal rather than underlying mean-
ing
( problems in understanding normal conversation, for
example, descriptions of sequences of everyday events
which are related temporally or causally
( lack of awareness of what knowledge is shared between
speaker and hearer
(as specied by McTear 1991: 22, with reference to Bishop &
Rosenbloom 1987)
A prole of this sort may emerge from informal observation of
conversations, or from more systematic assessment using a tool
such as the Childrens Communication Checklist (Bishop in press).
Samples of conversation taken from a group of 8 to 12-year-
olds who tted this clinical picture illustrate the inappropriacy of
their language (Bishop & Adams 1989). To generate conversation,
they were shown pictures of a doctor examining a sick boy; a girl
having a birthday party; and a man and girl with a broken down
car. Their responses to these pictures were characterised by cer-
tain oddities which did not occur in the samples of normal
children, even those who were considerably younger. These oddi-
ties were classied as:
Too much information:
Adult: Is that a good place to break down?
Child: The answer whether its a good place to break down is no,
because if see if anybody broke down, cos theres no telephone
to telephone, theres no telephone for the breakdown.
Too little information:
Adult: So what did you do when you were sick?
Child: I cant remember. I did though when I was run over by a car.
[What does did refer to?]

Unusual or socially inappropriate content or style:
Adult: Whats going on there?
Child: Its someones birthday. Something could be dangerous you
know like a re from the candles.
(Bishop & Adams 1989: 2504)
These children, it seems, have perspectives on the content of the

conversation and on their conversational partners which are dif-
ferent from those which we would expect and which other children
take. Their experience of the situations which are the focus in the
conversation and of the conversation situation itself must in some
way be different. But while it is obvious to conversational partners
that these children have an odd perspective, it is not at all obvious
just how their perspectives differ from those we would expect. Nor
is it obvious just what differences in their way of experiencing the
world are responsible for their odd perspectives. To approach these
questions, we need to nd ways of tapping the sorts of experiences
to which the children do and do not have access.
Exploring a childs capacity for experience through

language
How can we establish the quality of another persons experi-
ence which we cannot observe directly? One way is for the person
to tell us about her experience. But this depends on her ability to
reect on her experience and express it in words, and is ruled out
when reection on experience is the experience under suspicion
and in question. In this case, we can only nd out about the
persons experience by making inferences based on what she does
and says. Her language may offer the tightest source of evidence,
provided our probes are sensitive enough to pick it up.
We know that words convey highly specic and quite precise
aspects of experience. They can only be acquired and used appro-
priately to the extent that the child is able to identify the experi-
ence relevant to them. Suppose the child can identify some but not
all tiers of experience relevant to a word. This will enable her to use
the word appropriately in contexts which fall within her experi-
ence. But where a context involves tiers of experience which are
unavailable to her, her understanding and use of the word will be
inappropriate. A careful comparison of a childs appropriate and
inappropriate use of words can therefore cast light on the tiers of
experience which are or are not available to her. Does the child
deal with words differently depending which experiential tier they
draw on? Does the child deal with a word differently depending on
the context in which it occurs, and the experiential aspects of the
word which are pertinent to that context?
The obvious way to investigate this is to test the childs under-
standing and use of different sorts of words in different sorts of
contexts. But experimental investigation is in some ways unsuited
to the purpose. The experimental method lends itself best to testing
words whose meaning can be demonstrated. Demonstration de-
pends on the perceptual dimensions of words. Even when the mean-
ing of words is not essentially perceptual, we are likely to test them
through the perceptible aspects of their meaning. We test under-
standing by asking the child to point to a picture matching the
word or to act out the word; we test production by providing a
picture or demonstration of the word and asking the child to
describe it. Pictures and actions: the perceivable faces of words. We
focus on these not because the perceptual dimension of words is
the most signicant aspect of word meaning in general, and not
even because it is the most signicant aspect of word meaning for
children, but because the perceptual dimension is the one which,
by its very nature, can be demonstrated in a way that the experi-
menter can see. It is much harder to elicit demonstrations of affec-
tive and mental aspects of meaning.
The point is well illustrated by a set of experiments carried out
with a group of autistic youngsters (Lee, Hobson & Chiat 1994).
The purpose of the experiments was to investigate their under-
standing and use of pronouns. The experiments presented them
with photographs of people, including themselves and the experi-
menter. In order to elicit pronouns, they were asked questions
about the photographs:
Who is this a picture of?

Who is wearing the hat/scarf?
Tell me who they are
The probes for testing comprehension were:
Point to the picture of me/you
What am I wearing? What are you wearing?
According to their teachers, the majority of these youngsters had

difculties with pronouns. By way of example one teacher reported
that, on her return from sick leave, she was met with the comment
Im better now. Yet in the photograph experiments these same
youngsters did not make a single pronoun reversal of this sort.
Bafingly, one 19-year-old performed perfectly on the tasks and
then left the room saying Thank you for seeing you Tony, revers-
ing the pronoun in an idiomatic construction where we might
least expect it. How can we make sense of these apparently contra-
dictory responses? It looks as if some people may actually perform
better in a highly structured experimental situation which focuses
on pictures and action, than in everyday interaction where ident-
ifying the focus is precisely the problem. This is the reverse of the
more common nding that childrens performance is poorer in
experiments, which remove the contextual support offered by
everyday situations.
A similar observation was made about a group of 810-year-
old children who tted the clinical picture of semanticpragmatic
disorder (Bishop & Adams 1991). These children fell well outside
normal limits on measures of conversational inappropriacy, ob-
taining high scores for the categories too much information and,
in some cases, too little information in the conversational analy-
sis procedure described above (Adams & Bishop 1989). Yet these
children were well within normal limits in a structured communi-
cation task. This task presented them with an array of eight
pictures which differed from each other in one of a set of three
features, for example, colour, size or shape of an item. The children
were required to describe one of these pictures to an adult who had
to nd the card within the array. Under these conditions, the
information the children provided was no less appropriate than
age-matched controls. Again, we nd that communication of
visual information in a structured setting does not bring out prob-
lems which are evident in open-ended conversation.
It is striking that non-verbal tests of IQ typically rely on
perceptual material which is presented in a highly structured con-
text. This may explain why some children who go off the rails in
quite basic everyday communication can come out as average or
above average in an intelligence test.
If we are to probe these childrens access to different aspects of
word meaning, we may do better to probe this in more everyday
contexts which vary in respects which cannot be picked up
through purely perceptual cues. It is not easy to do this within a
tight experimental format. Within a more informal setting,
though, it is possible to manipulate contexts systematically and
explore how changes in context inuence a childs understanding
or use of words. If we are sufciently ingenious in deciding which
sorts of experience may be critically involved, and in devising
scenarios which probe for these, we should gain valuable insights
into the invisible tiers of the childs experience.
For a avour of how research in this vein might be pursued and
what it may reveal, we turn to selective observations of children
which point to important differences in what they can experience.
We start with Eamonn and Ruth, whose difculties were followed
up in chapters 6 and 12. When we explore their sensitivity to
different types of experience, we nd strengths which are in strik-
ing contrast to their difculties. We then move onto children
whose difculties appear to lie in their ability to experience the
world as other people do.
14[] you dont tell nobody
this?: strengths in
pragmatic processing
Eamonn, at 6, and Ruth, at 10, are sometimes hard to understand,

as we saw when we met them in previous chapters (see chapters 6
and 12). They often produce words which the listener doesnt
recognise. In full ow, they can leave the listener completely at sea
with their meaning. On other occasions, their utterances are quite
clear, but fragmented and lacking some of the forms that sentences
demand. Where enough words can be recognised, the listener gets
the gist of what they are saying, but may still be in the dark about
certain details which we expect in language. The listener might
gather who the children are talking about, what happened to
them, but be confused about whether it has happened, may hap-
pen, or will happen. The listener may know they are asking
something about an event, but not be sure whether they are
asking who was involved, where it occurred, or when. Conse-
quently, conversations with Eamonn and Ruth are not as smooth
as with other children of their age: their conversational partners
often nd themselves confused or at cross purposes, and the con-
versation has to backtrack to get back on course. And yet these
children are perceived as bright, on the ball, responsive and par-
ticipating appropriately in conversations. Despite the considerable
obstacles to communication, the gut feeling of the listener is that
the meanings they have in mind are normal. What is the basis for
that feeling? What is the evidence that they have access to the
normal range and complexity of experience on which meaning is
built, even if these are not reliably encoded in their language?
The evidence emerges when we scrutinise the meanings they
convey through their use of words in context. We necessarily rely
243
on utterances which contain recognisable words. These allow us
to identify the word meanings which the child has selected. By
considering these word meanings in relation to the context of their
utterance, we can make inferences about the meaning the child is
trying to convey. This in turn reveals the quality of experience
which underpins that meaning.
Of special interest are the more abstract, non-perceptual tiers of
experience. These are not directly observable through childrens
visible responses to objects, actions or pictures. But they can be
inferred from the contributions children make in scenarios which
are dened by non-perceptual states, events and relations.
Children may themselves conjure up such scenarios. They may
initiate references to a state of affairs which is not perceivable. Or
they may string together references to states of affairs which are
themselves perceivable but whose connection to each other is not.
Their initiation of such scenarios can provide powerful evidence of
experience which we cannot observe. This can also be tapped by
presenting children with similar sorts of scenarios, and exploring
their responses to these.
In the context of such scenarios, both Eamonn and Ruth prove
their sensitivity to invisible events, and to invisible relations be-
tween events.
Sensitivity to mental and affective events

Eamonn and Ruth both refer to invisible events which are in the
mind of one person, and not in the mind of another.
Forbidden fruits kept secret Eamonn initiates a scenario in which

an event he desires must be kept secret, that is, outside the minds
of others:
SC: So what do you want for next week?
E: Ribena ( = a blackcurrant drink).
SC: Are you allowed to have Ribena at school?
E: No. You can put it You can put it [n#] bag and the eibody know
you
SC: Huh?
E: You put it in [] bag and edibody know [jun] taking Ribena.
Strengths in pragmatic processing 245
SC: You mean nobody would know I was taking it?
E: You put the Ribena the big bag and then edibody know you
taking Ribena.
SC: And then nobody would know?
E: No.
SC: It would be a secret . . . You would tell somebody!
E: I not!
SC: You would say Anna, S. brought me some Ribena.
E: I not. I not tell him.
Some of Eamonns contributions to this interchange are quite hard
to decipher. In one of his more intelligible utterances, he appears to
omit a crucial element of meaning: the negative element needed to
indicate that if the Ribena is in the bag nobody will know about it.
This could be a sign that Eamonn is confused as to whether other
people will or wont know about the Ribena. However, the context
strongly suggests that Eamonn is seeking to prevent other people
knowing about the Ribena, and that he does intend the negative. It
is he who comes up with the suggestion of putting the Ribena in
the bag which, in the context of Ribena not being allowed, carries
the implication that he wants to stop other people knowing about
it. It is also he who initiates reference to people knowing about the
Ribena. Given the coherent relation between the events he selects
to talk about and the state of affairs posed in the question, we can
infer that the meaning he intends is nobody knowing, and that
his confusion is between the words everybody and nobody. Despite
this confusion, he is quite clear in his mind about the discrepancies
between the contents of different minds.
A touchy matter in the family Eamonn has started recounting a

family saga when he realises hes got himself in emotionally deep
water. Well into his tale, he suddenly says:
#
E: Oh, I can tell you this. Oh! Can you keep this a [ swikw]?
SC: Can I keep it what?
#
E: Keep it a [ swikw] becau my mum said dont tell this t . . .
and Eamonn launches back into the story. After he reports:
E: Ms little bit upset
he returns to the secrecy:
E: [] you dont tell nobody this?

SC: Of course not.
E: Because my mum said dont tell nobody . . .
In this lengthy exchange, Eamonn conveys a complex layering of

affective and mental experiences. Some of these are expressed
explicitly, and some are implicit in the sequence of utterances. He
has understood Ms affective state (explicit in upset). He has under-
stood that Ms state and the events leading up to it are a sensitive
matter: this sensitivity is implicit in his understanding that the
matter should be kept from other peoples minds (explicit in cant
tell, secret, dont tell). Apart from the affective and mental experien-
ces which he conveys within his account, he reveals his own
conicting feelings about giving the account. He is bursting to go
on with it, but this desire battles with the guilt of divulging matters
which he knows he should not. His way out of the conict is to
commit the account to secrecy.
Eamonns understanding of mental experience goes a step fur-
ther when he initiates a scenario of pretence about mental experience.
After receiving a robot for his birthday, he conjures up an impres-
sive fantasy:
Mental telepathy with a robot

E: You put the two wires on them and then you push this button and
[] wire goes to my the wire come to the robot [unintelligible
word] and [] robot come to [] my brain so I talk like him. And he
talk like me. So thats clever.
Although it is not absolutely clear what Eamonn has in mind, this

much is clear: he is playing with his knowledge that one persons
brain is not normally available to another; his pretence is that a
wire connected to a robot can overcome this reality.
Ruths engagement in fantasy shows a similar understanding of
mental states.
Monkey business In this case, I have initiated a fantasy, the

fantasy being that we each have a secret pet monkey. Ruths
response shows a full understanding of the content of the fantasy,
and that it is a fantasy. Ruth has named her secret monkey
Aeroplane, while mine is named Coconut:
SC: Did you bring Aeroplane with you?
R: Yeah, [m] head. [d] got Coconut with you?
SC: Yeah. Where?
R: In your head! . . . Monkeys in ours head. Nobody know us got
monkeys in our head.
When Ruth situates the monkeys in ours head, she shows she
has them rmly established as characters in a fantasy and not in
the real world. When she says nobody knows, she proves her
understanding that the contents of one mind are not available to
other minds.
Ruths understanding of secrets comes in handy as a way of
deecting the discomfort she feels when her difculties with lan-
guage are exposed:
The right to silence

R: I know my phone number is.
SC: What?
R: Im not telling you. Secret.
Here again Ruth refers to information which she has in her mind
(or claims to), and the withholding of that information from
another mind.
Sensitivity to conditions and consequences of

events
Eamonn and Ruth not only refer to the abstract events of
knowing, hiding of knowledge and communicating of knowledge.
They also refer to abstract relations between events. This is evident
in their linking of two events within a single utterance. The events
which they link may themselves be abstract.
The precise nature of the relation between events is not always
made explicit. Where it is not, it can usually be inferred from the
content of the combined events, because these hold a particularly
plausible relation to each other.
Eamonn volunteers many examples of one event embedded
within another:
Utterance Gloss of utterance identifying

events
I tell off I get Ribena so I not Ill get told off if I get Ribena so
tell anybody I wont tell anybody
[] you dont tell nobody this Can you not tell anybody this.
. . . Because my mum said Because my mum said dont
dont tell nobody tell anybody
Where he uses because and so, the specic link between the
events is clear. In some cases, the link is less clear. He appears to
use the form [] to introduce a hypothetical or anticipated event
as a condition for another event, where the link should be if:
Utterance Gloss of utterance

[wi stp] it, it be die If we step on it, it will die
. . . we can take them home We can take them home if we
[] we be [fi fi] good are very very good
Eamonn does not stop at one embedding. He produces utterances

which embed already linked events within a further event:
E: We gonna be good because [] we good we llowed to take our
owers home.
Here, one event, being good, is the condition for another event,
being allowed to take owers home; these linked events, being
allowed to take the owers home on condition of being good, are
the cause of another event, being good.
Ruths linking of events is less explicitly marked than
Eamonns. The links she intends can be inferred, though, because
the events she has chosen to link hold a clear relation to each
other. In some cases, one event is appropriate as a condition for the
other:
Utterance Gloss of utterance

Talking about communication
between sessions
You can ring up [] want to You can ring up if you want to
Playing the role of boss talking
to secretary
[td] someone call, [td] If someone calls, you put a
put a note down, right note down
Playing the role of prison guard,
talking to prisoner
[zu] scape, we nd you If you escape, well nd you
In some, one event is appropriate as the consequence of another:
I do I want to, syou shut up I do I want to, so you shut up
right
On occasion, Ruth layers events. This happens when she explains
why, after unlocking the door, she has kept one key separate from
the rest of the bunch:
SC: Why do you need to know which key?
R: Because [] wrong key, [t] cant do it.
SC: Yeah, but weve already opened the door. We dont need it now.
R: Yes, lock it!
SC: When?
R: After us.
Here, Ruths reason for one event (keeping the key separate) is the
event which would follow from the hypothetical situation of hav-
ing the wrong key (cant do it).
Use and content of utterances as evidence of

invisible experience
When we comb through conversations with Eamonn and Ruth,
we nd strong evidence of the breadth and depth of the experien-
ces to which they are sensitive. The fact that they initiate and
respond to communication appropriately already implies an
understanding that language can convey what is in one persons
mind to another persons mind. When they communicate, they
provide further evidence of this understanding, by referring to
affective and mental experiences, and responding to other peoples
references to these. They also refer and respond to dependencies
between events, and they can do this with events which tap
different tiers of experience. Their language may be severely
limited and limiting (see chapters 6 and 12), but it shows that their
ability to experience is not.
Not so with some children. We turn now to three children
whose language may be quite intelligible, but whose meanings are
elusive or bizarre.
15 I can speak Chinese. But I
cant speak Chinese:
problems in pragmatic
processing
The children presented in this chapter come out with language

which is odd. But the oddness of these childrens behaviours is not
conned to language. They are all reported to have shown certain
obsessive behaviours. All are observed to react to situations in
unusual ways. Their play is very limited: when offered toy minia-
tures, they are likely only to reproduce scenarios they have seen,
rather than constructing novel scenarios with the toys. When
they talk, they commonly echo another persons utterance.
All of these behaviours suggest that their experience of the
world and of other people is unusual. Since shared perspectives are
fundamental to establishing meaning in language, we would ex-
pect them to attach unusual meanings to the language they hear,
and to express odd meanings when they talk. It is not surprising,
then, that they have all been found to perform well below their age
level on tests of comprehension. They have also been very slow to
produce language. When we home in on the language they pro-
duce, it proves quite odd.
Tony, who was observed between the ages of 3 and 7 in a study
by Conti-Ramsden and Gunn (1986), showed all of these charac-
teristics. Alongside these, he showed certain strengths. He was
very good at tasks using visual materials. He successfully sorted,
matched and sequenced items by size; recognised and matched
numbers and symbols; copied patterns; and was able to do puzzles.
Throughout the period of observation, his score on a test of non-
verbal intelligence was consistently above average. Yet his use of
language remained odd.
He reversed pronouns such as I/you, and he was still having
251
difculties with these at the age of 6. This confusion points to
problems with perspective. His language revealed other evidence
that he had difculty taking and maintaining perspectives. At 5,
when he played the game What am I?, Tony took it upon himself
both to ask and to answer the question, thereby defeating the
purpose of the game:
T: I have long ears and a short tail, what am I? A rabbit.
(Conti-Ramsden & Gunn 1986: 345)
At 6, his responses to questions about various scenarios showed
that he had picked up some aspects of each scenario. But he did not
maintain perspectives on individuals, times of events, and rela-
tions between events in the scenario:
Teacher: What might happen if you didnt look and listen before you
crossed the road?
T: Shes falling in the road. She went to the hospital.
(Ibid.: 346)
The references to road and hospital are potentially relevant to
the scenario in the question. But Tony has not maintained the
perspective on person (you shifts to she) or time (would shifts
to is falling, went). Nor has he maintained the perspective on
events: he leaps from not looking when you cross the road to the
unconnected event of falling.
Teacher: Why do you have to be quiet when theres a baby in the
room?
T: Cos shes crying. (Ibid.: 347)
This response illustrates clearly Tonys failure to appreciate the
perspective on events which is presented to him. His response
refers to crying, which is certainly not a reason for the event in
the question, nor is it a likely effect, condition, or goal of that event.
It looks as if Tony has simply found an event which is loosely
connected to various references in the question, crying being
associated with quiet, and baby.
Similar shifts in perspective occur in his response to a play scene
in which Carl, a doll, has fallen:
Teacher: Is Carl frightened?
Problems in pragmatic processing 253
T: Yes.
Teacher: Why is he frightened?
T: He is falling down, he cried, he is sore mouth. (. . .)
Teacher: Why has he got a sore mouth?
T: Because you are falling (to doll). Why are you getting on the
oor? (Ibid.: 347)
Tonys utterances again refer to events which are plausibly asso-
ciated with each other and with the question (frightened ; fall-
ing down, cried, sore) and this time, the sequence of events is
more coherent. But relations between the events are still by no
means clear: hes frightened because he fell down and he cried
because he got a sore mouth? he fell down and got a sore mouth
and so he cried? In addition, Tony has confused time reference (is
falling, cried). He also appears to have shifted person reference
(he to you), though if he has turned away from the teacher to
address the doll this shift of person reference is appropriate.
Tony has clear forms of reference to people, times and events,
yet his use of these shows considerable confusion. A 10-year-old
observed by McTear (1985) shows similar peculiarities. Unlike
Tony, this childs non-verbal as well as his verbal abilities are
depressed: he comes out as low average/borderline on a test of
non-verbal intelligence. However, by 10 he makes substantial
contributions to conversations. And he produces intelligible ut-
terances of some complexity, albeit with odd voice quality and
intonation.
This child appears to sustain references better than Tony. But
he still shows evidence of difculties with perspective. For example,
he makes assertions which assume a perspective at odds with the
context in which they are made:
Adult: Now do you want to see if you can play some games with me?
Child: Yes
Adult: Theyre very easy games um
Child: They are indeed (McTear 1985: 133)
As he did not yet know what the games were, he could not have
known that they were easy. The perspectives he takes on individ-
uals and events can also be at odds with each other:
Adult: Which race would you like to be in?
Child: I like to be in X ( = place) at the Sports Day
Adult: In X?
Child: Yes
Adult: What do you mean?
Child: I mean something
Adult: Is there a Sports Day in X?
Child: There is not, there is a Sports Day in Y ( = childs school)
Adult: Then whats X got to do with it?
Child: Nothing
Adult: Then why did you mention it?
Child: Indeed I did mention it (Ibid.: 1356)
Here, the child makes clear enough references to events and

places, but he shifts his perspective on these to the point of contra-
dicting his own assumptions: he refers to a Sports Day in X and
then states that there is no Sports Day in X. Sometimes, he atly
contradicts himself:
(referring to a boy in a picture)
Adult: Does he have to pay for his dinner?
Child: Yes
Adult: Why?
Child: Because he wants to eat it
Adult: Do you have to pay for your dinner?
Child: Yes
Adult: Do you?
Child: No I dont pay for my dinner. (Ibid.: 138)
Similar contradictions crop up in conversations with Sally. At

9, Sally appears to relate well to her conversational partner. She
makes good eye contact. She initiates and responds to talk, and her
talk is generally uent and delivered with normal-sounding voice
quality and intonation. Yet it can be very difcult to follow who or
what she is talking about, or what she is saying about them. This is
illustrated by the conversation with her which was quoted at the
outset of chapter 13:
SC: What languages do you speak?
S: I speak with Oxford Junior.
SC: You speak with Oxford Junior?
S: Oxford Junior English.
SC: Do you speak Chinese?
S: Yeah I speak Chinese. But English but I cant speak Chinese.
SC: You cant speak Chinese? Or you can?
S: I can speak Chinese but my mum cant speak English. But he can
speak English because but he cant speak Chinese. And she can
speak Chinese too.
In this conversation, Sally uses words which clearly hold some

connection to each other and to the questions she is answering:
English, Chinese, Oxford Junior English, speak all relate to
language speaking, which was the focus of the opening question.
The problem is the reference of the words. Her shifts from my
mum to he to she are confusing, because we cant tell whether
these refer to the same person or not. If we take her to be maintain-
ing reference to one person, she is directly contradicting what she
says about that person without turning a hair. If we take her to be
referring to different people, she is shifting reference haphazardly,
making it impossible to identify who she is talking about. Alterna-
tively, it may be that she loses track of which language she is
focusing on. Whichever way we play it, we nd ourselves in the
appropriate ballpark language-speaking in Sallys family but
none the wiser as to who speaks what.
The impression is that Sally cannot hold a perspective on a
person or a state of affairs. The same thing appears to be happen-
ing in the following exchange:
(Sally has spontaneously made reference to a teacher who has had a
baby; seeing the baby in her tummy; and the teacher having eaten too
much)
SC: You havent got any babies in your family have you? (S is the
youngest)
S: I have a aunty.
SC: Has she got a baby?
S: Yeah.
SC: Whos the baby in your family?
S: Um. Ellen. My sister. And a baby name called Sophie. Sophie called
mummy. And she called daddy. I mean and she called mummy
and she called papa.
SC: Whos called mummy?
S: Sophie.
SC: And whos papa?
S: Sophie. Sophie called papa. I heard that.
SC: Oh, is Sophie the baby?
S: Yeah.
SC: And she can say words? (no response). Is she a little baby?
S: No, er, then she was a zero years old. And when she grown up and
she was one years old and when she grown up again she was two
years old.
Sally again uses a number of related words: aunty, sister, mummy,
daddy; Ellen, Sophie, name, called; baby, grown up, years old. But
again, her references appear to oat, unmoored, in an uncharted
sea. She refers to people in her family, but we have no idea who is
who. She refers to ages, but we are still in the dark about the age of
the baby in question. Her reference to events is similarly unan-
chored. It looks as if she is sliding between two meanings of
called, so that her talk about what the baby is named (called)
suddenly ips into talk about what the baby said (called). A similar
slipperiness in her event reference may account for her initial
response I have a aunty. With some inside information about her
family that she has an older sister who has a baby it seems
likely that the reference in the question to having a baby in the
family has slipped into the related event of being an aunty to a
baby in her response.
Sallys confusion of perspectives is exposed most dramatically
when she plays a game called Guess who?. In this game, each
player picks out a picture card from a set representing different
characters. These characters have names, and are distinguished
from each other by sex, age and a variety of features such as hair
colour, having a beard or moustache, wearing glasses or a hat (see
(GUESS WHO? 1999 Hasbro International Inc.).

illustration). Players can look at their own card, but not at anyone
elses, the aim of the game being to guess the identity of other
players characters. They do this by asking questions such as
Has this person got white hair?
Has this person got a beard?
which enable them to identify the character by a process of elimin-

ation. Look at Sallys explanation of the game:
(We have each taken our picture card)
SC: So, what do we have to do?
S: Dont look the card. Thats what Jane said.
SC: Dont look at the card?
S: Yeah. Guess what it is. If Jane and could is it or is it a Peter or is
it a or is it a Joe.
SC: Can I look now?
S: No. Dont look at it and trying to guess (S points to my card as she
speaks).
SC: Right . . . You looked at yours.
S: U-huh.
SC: Is that all right?
S: Dont look for it otherwise its just cheating.
SC: But you did look at yours. Can I look at mine?
S: Er no. Jane say you cant and then you have to guess.
SC: Which must I guess? Must I guess this one (pointing to my card)?
Or that one (pointing to her card)?
S: This one (S points to my card).
SC: So I mustnt look at it.
S: No. Um. Put over here. Put over here (S points to the slot for my
card, which is facing me).
SC: But then Ill see it. Can I look at this one (my own, which she is
telling me to put in the slot)?
S: No. You Can you tell Jane?
SC: Huh?
#
S: Will you tell Jane. I dunno what [z badz].
SC: Youre not sure what I think you look at yours and I look at
mine.
S: Yeah.
Sally knows that not looking at is important in this game. She has
clearly heard this many times, and heard that it is cheating. But
she seems unable to sort out who is allowed to look at which card.
When she is directly confronted with the contradictions in the
set-up as she presents it, she cannot resolve these contradictions,
and she appeals to the higher authority of her therapist (Jane).
Interestingly, she uses the verb tell when her object is in fact to ask
Jane.
Once her confusion is set aside, Sally proceeds to play the game
appropriately, with no confusion about which picture each of us
looks at. She asks appropriate questions which enable her to
narrow down and ultimately identify my picture. In the course of
the game, she makes just one slip in response to one of my
questions:
SC: Has the person got white hair?
S: (mistakenly looking at the set of pictures in front of her rather
than her own picture) One has, two has . . . What did you say?
SC: Has the person got white hair?
S: No. No. Sorry I made a mistake.
Sallys responses to the scenario of this game are bafingly

inconsistent. They are worth working through in some detail to
see whether we can bring any coherence to their seeming incoher-
ence. On the one hand, she can play the game. She can select
features appropriately as the focus for questions such as Has this
person got a big nose?. She can also use responses appropriately. If
the response is yes, she throws out all those characters who have
not got a big nose; if no, she throws out those who have. Here,
her logic is impeccable. Yet in explaining the game, she cannot
hold onto its simple logic: that you must look at the picture you are
answering questions about. How can we make sense of this appar-
ent paradox?
The paradox may be resolved in terms of the different tiers of
experience tapped by the different activities. Suppose the perceptual
tier of Sallys experience is normal. She observes and holds onto
what things and people look like, and what they are seen to be
doing. This would enable her to recognise the way players in the
game are visually oriented towards their picture cards. Equally,
she would be able to recognise the visual features which distin-
guish the characters on the cards. When the material is visual, her
logic is intact: she recognises that having a beard is mutually
exclusive with not having a beard. This would account for her
ability to work through the process of elimination systematically.
Talking about the same scenario is a different matter. It is not
enough to recognise the way the game plays itself out visually: the
visual position of players and their materials, and the visual fea-
tures of her picture. In order to explain the game, she must select
and stay with the perspective of one player. She must recognise
how that players perspective is distinguished from the perspective
of other players. That is, she must appreciate that each of the
players has a distinct visual perspective, which makes them privy
to different information from other players. The idea that each
player sees and knows what other players do not see and know is
fundamental to the aim of the game: guessing who. It is precisely
this which puts Sally in a spin. She fails to distinguish my perspec-
tive from hers, and allow me to see and know what she must not.
Sallys difculties with perspective are brought into relief when
we compare the explanation of the game offered by 10-year-old
Richard, whose words are far more limited and distorted than
Sallys:
SC: What do I have to do?
R: [pn] card out. [p] your card out. And we guess who you got the
person. Someone [juaju] guess someone and [] winner.
SC: So what do we have to do rst?
R: [pn] card out. No tell me, no. Put in there (pointing to my slot).
#
[s kntnz] (probably = ask questions).
[] you got bald head or not. Bald head or hair. [b] him or
her. Something else. Got hat big eyes and big nose.
SC: And how many questions can I ask?
R: [] [] [u] guess mine. [] turns. You, me, you . . .
For all the limitations of his language and his sometimes unclear
references, Richard highlights just those differences in perspective
that Sally obscures. He clearly takes my perspective in the game
and he presents this consistently in relation to his perspective: he
tells me to pick out a card, shows me where to put it, and instructs
me not to tell him about it. His explicit references to guessing and
turn-taking conrm his grip on who knows what and who does
what. A further interesting difference between the two children
emerges from the examples of questions they offer. All of Sallys
examples pick on features of the characters which are purely
visual. In contrast, Richard includes a question about the gender
of the character (him/her) which is not a simple visual feature.
The indications are that Sally, unlike Richard, experiences the
world very differently from other children, and that the oddness of
her utterances stems from the oddness of the experience on which her
language is built. The suggestion is that her access to non-percep-
tual tiers of experience to mental and affective tiers is disturbed.
We have seen that these tiers of experience are central to the
meanings of many words; that they are central to discovering the
meanings of words in general; and that they are central to appreci-
ating the functions of words in utterances. A disturbance in these
tiers can account for the mix of oddities observed in Sallys lan-
guage: her tendency to echo other peoples utterances; her limited
comprehension and production of language; her unanchored ref-
erences to people, events and times which make her spontaneous
utterances so disorientating for the listener; and her confusion of
perspectives.
Homing in on the mental/affective tiers of experience is really
only the start of another story. The tiers of affective and mental
experience are themselves complex (see Hobson 1993 and Mayes
et al. 1993, for example), and may be disrupted to different degrees
and in different ways. We have looked at data from three children
who all appear to have difculties with perspective, but the way
these difculties manifest themselves is by no means the same.
Having identied perspective as the domain of difculty, the wide
differences between children who show difculties in this domain
invite us to look more closely at the complexities of affective and
mental experience and the different ways in which such experi-
ence may be disrupted. Where children have some ability to
understand and use language, their language is a potential source
of evidence of the experience which lies behind it. In-depth investi-
gation of carefully selected words in carefully devised scenarios
could reveal more about the sorts of experience to which children
are or are not sensitive. It could also reveal what they do with
words which relate to experiences which elude them.
This type of psycholinguistic investigation has not yet been
pursued with children like Sally whose difculties with words lie in
their meaning and use. From the data analysis in this chapter, the
indications are that their words may provide a unique window
onto the unusual quality of their experience and the unusual sense
of the world to which this gives rise.
Endpoint and springboard
Some children can hear; they can speak; yet their language is not
developing normally. The source of their difculties is not immedi-
ately obvious. It lies beyond the auditory and motor ends of
speech, in processes which are hidden from consciousness and
from direct observation. This book has been a search into the mind
of the child to uncover those hidden processes and to nd out
where they stop short.
The search sets out from the childs spontaneous language. In
some cases, this provides leading clues. It reveals clear-cut
strengths which eliminate certain components of processing from
suspicion, and point a nger at others. Take the child whose
output is unintelligible or near-unintelligible at rst brush, but
once deciphered, is found to contain the sorts of words and word
combinations we would expect in that child. It can be condently
concluded that this child has no difculties with the meaning of
words or with their syntactic organisation into sentences, in either
input or output. Difculties with the phonological (rhythmic) or-
ganisation of sentences may also be eliminated. The problem is
then conned to the phonological form of words. The question is:
at what point (or points) in perceiving, representing or producing
the sound patterns of words does the problem arise? We can probe
the source of the problem further through input and output tasks
which tap different sounds in different sound contexts.
Conversely, a childs output may be phonologically impeccable.
Words may sound just as they should. They may also be appro-
priately organised into sentences, with appropriate rhythmic pat-
terns. The childs output is nevertheless odd. The choice of words
262
and structures somehow jars with the social and semantic context.
This points to problems with meaning. The child is not making
sense of the world in the same way as other children, and so is
attaching unusual meanings to the forms of language, these being
no trouble in themselves. We would expect the peculiarities which
show up in language to show up outside of language as well, in the
childs response to scenarios and interactions which she experien-
ces differently from other children. These initial observations drive
further exploration. Systematic investigation into the childs lan-
guage may reveal that she understands or uses words which tap
certain types of experience but not others, say perceptual but not
mental experience. Or it may reveal that the child understands or
uses words in contexts which draw on some aspects of their
meaning but not others, say perceptual but not affective aspects.
By delving into subtle differences in the childs responses to words
and contexts, we can throw more light on the oddness of her
language and the experience of experience which underpins it.
While some childrens output shows clearly polarised strengths
and weaknesses, many children present a mixed picture:
( their words are limited, and the words they do produce

may be off-beam phonologically or semantically
( their organisation of words into sentences is limited, and
the sentences they produce may be characterised by
omissions or errors in syntactic, phonological (rhythmic)
and semantic structure
The incidence of such a scattered prole is not surprising. If a

spanner falls into the middle of the works, the disruption is likely to
be widespread. If a child has difculties at crucial intersections in
language processing, at or close to the point where form and
meaning join, we might expect these to reverberate through the
childs language.
We have seen that the union between the processing of phonol-
ogy and the processing of experience is highly potent. In phonologi-
sing experience packaging experience into phonological chunks
language directs and controls our interpretation of that experi-
ence. Paradoxically, in tying down our experience, language
264 Endpoint and springboard
Picks out
meanings and
meaning relations
Picks out Picks out

focus in scene chunks in speech
Picks out
word forms and
word order
Figure 15.1 Interactions in mapping between form and meaning
affords us the freedom to manipulate and communicate experience

in untold ways.
The phonologising of experience relies on the childs facility for
identifying phonological chunks within speech, and for identifying
speaker focus within scenes. We have seen how each of these
facilities acts as a bootstrap to the other. The chunks which the
child nds in speech push the search for meanings which corre-
spond to those chunks. The perspectives on scenes which the child
picks out push the search for forms which correspond to those
perspectives. The result may be seen as a push-me-pull-you rela-
tionship between form and meaning (see gure 15.1).
It is through these intertwining bootstraps that children con-
struct the extraordinarily elaborate mapping between meaning
and form which is language. Where there is a weakness at one or
other end of the bootstraps, bootstrapping between form and
meaning will be disrupted, and the childs mapping between these
will be undermined.
Limitations in the childs ability to process rhythmic patterns, to
pick out phonological chunks within these, or to hold onto the
details of these chunks will limit the childs phonological grip on
words. This limitation will roll along the bootstraps: it will affect
the childs phonological bootstrapping to word semantics and to
syntactic relations between words. We would expect to see limita-
tions or distortions in word phonologies, but also in word seman-
tics and in syntax. The extent of the gaps and errors will depend on
( the extent of the childs access to phonology
( the childs response to shortfalls in her processing, for
example whether she includes information which is only
partially specied or whether she leaves it out
( the extent to which the child is able to draw on strengths
in processing to compensate for her limitations
The childs freedom to manipulate experience mentally, which
comes with ready access to words and sentences, may to some
degree be curtailed.
Limitations in the childs ability to identify speaker focus within
scenes will again ow through the childs language processing,
though with different effects. It will limit the childs semantic
bootstrapping to the meaning which attaches to word phonologies
and to the meaning relations which map onto syntactic structures.
If the childs phonological and syntactic processing are intact, she
may exploit these to the full and come out with words in sentences
which sound positively sophisticated. But their meanings would
extend only as far as her limited or odd abilities to experience
allowed. This results in the polarised prole of strong word and
sentence phonology and weak semantics which was described
above. On the other hand, the childs limited semantic bootstrap-
ping may cramp the connections she makes to phonology, leading
her to ignore phonological and syntactic forms even if she has the
capacity to process these.
Where the deviations in a childs output are scattered across
phonology, syntax and semantics, our exploration of the childs
processing will focus on each of these and their interconnections.
This will mean selecting particular words or structures which are
known to be problematic, in order to investigate what about them
is more or less problematic. Can the child process the phonology
266 Endpoint and springboard
which is needed for the form of the word or structure in question?
Can the child process the types of experience which underlie the
meaning behind that word or structure? Can the child process the
connections between these? Through tasks judiciously chosen to
tease out different aspects of a word or structure, we can expose
the strengths and weaknesses in the childs bootstraps.
The children in this book have personied some of these pro-
les. They have provided the impetus for investigation into the
processing behind the prole. They have illustrated the logic of
psycholinguistic exploration, and the sort of techniques which can
be used to uncover what the child can and cant do with words.
This does not mean we will always recognise and understand the
childs words, or fully appreciate the nature of experience to which
those words give expression. Nor will we be able to restore
processing skills which the child does not naturally have. But our
detection of strengths in the childs processing may point to ways
round the weaknesses. We may highlight for the child what she
can do with words, and encourage optimal use of her psycho-
linguistic resources. More than this, we may play into the childs
strengths by adjusting our input to highlight those aspects of a
target phonological, semantic, syntactic which the child can
process. In this way, we may facilitate the childs access to aspects
of a target which are not directly accessible.
As we rene our insights into the childs processing of lan-
guage, we also come closer to the childs experience of language
and language disability. Ruth, now 23, articulates the importance
of such understanding:
Some people dont understand it means. They dont understand what
we going through like. And doctors dont. They know about it, but
they dont know what caused it from . . . They should put something
on TV. [z] get people understand why we have it.
This book isnt TV, but it turns a spotlight on what Ruth has in
mind.
Further reading
Language impairment in children

See Bishop (1997) and Leonard (1998) for comprehensive, in-depth
and clear reviews of research into specic language impairment (SLI) in
children, spanning:
the nature and source of their difculties with language and

language processing
other aspects of their cognitive processing
genetic and neurobiological characteristics of children with SLI
Research studies documented in these texts also furnish further examples

of techniques which may be used to tap different aspects of language and
wider cognitive processing.
See Lees & Urwin (1997) for a clinical perspective on language-impaired
children dealing with clinical assessment, treatment and management,
and illustrating these with individual cases.
Speech impairment in children

See Stackhouse & Wells (1997) for extensive presentation and illustra-
tion of psycholinguistic investigation into speech processing in children
with speech impairments.
Normal language development

See Owens (1996) for a comprehensive and clear review of research on
the stages and processes of normal language development.
See Fletcher & MacWhinney (1995), Hirsch-Pasek & Golinkoff
(1996), MacWhinney & Bates (1989), Morgan & Demuth (1996),
Tomasello & Merriman (1995) and Slobin (1985, 1997) for further
reading on aspects of language development focused on in this book.
267
268 Further reading
Relationships between semantics, syntax and phonology
See Saeed (1997) for more on semantics in general, and Pinker (1989)
for detailed analysis of the relationship between verb semantics and
syntax.
See Selkirk (1984) and Cooper & Paccia-Cooper (1980) for extensive
analysis and investigation into the relationship between the syntax and
phonology of sentences.
Meaning in and outside language

See Gumperz & Levinson (1996), Jackendoff (1997) and Slobin
(1997) for extensive analysis and evidence of the relation between lan-
guage and other cognitive structures in line with the thinking in this
book.
Theories and models of language processing and wider cognitive

processing
See Aitchison (1994), Harley (1995) and Levelt (1989) for overviews
of word and sentence processing.
See Ellis & Young (1988) and Temple (1997) for introductions to
cognitive neuropsychology and extensive illustration of its application to
cognitive impairments in adults and children respectively.
References
Aitchison, J. (1994). Words in the mind: an introduction to the mental

lexicon. Oxford: Blackwell.
Aksu-Koc, A. A. & D. I. Slobin (1985). The acquisition of Turkish. In The
crosslinguistic study of language acquisition. Volume 1: The data, edited
by D. I. Slobin. Hillsdale, NJ: Lawrence Erlbaum Associates.
Ambalu, D., S. Chiat & T. Pring (1997). When is it best to hear a verb? The
effects of the timing and focus of verb models on childrens learning of
verbs. Journal of Child Language 24: 2534.
Anthony, A., D. Bogle, D. Ingram & M. W. McIsaac (1971). The Edinburgh
Articulation Test. Edinburgh: Churchill Livingstone.
Baldwin, D. A. (1993). Infants ability to consult the speaker for clues to
word reference. Journal of Child Language 20: 395418.
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of
mind. Cambridge, MA: MIT Press.
Baron-Cohen, S., H. Ring, J. Moriarty & B. Schmitz (1994). Recognition of
mental state terms: clinical ndings in children with autism, and a
functional neuroimaging study of normal adults. British Journal of
Psychiatry 165: 6409.
Baron-Cohen, S., H. Tager-Flusberg & D. J. Cohen (eds.) (1993). Under-
standing other minds: perspectives from autism. Oxford: Oxford Univer-
sity Press.
Berko, J. (1958). The childs learning of English morphology. Word 14:
15077.
Bishop, D. V. M. (1983). Test for Reception of Grammar. Available from Age
and Cognitive Performance Research Centre, Department of Psychol-
ogy, University of Manchester.
Bishop, D. V. M. (1997). Uncommon understanding: development and dis-
orders of language comprehension in children. Hove: Psychology Press
Limited.
269
270 References
Bishop, D. V. M. (In press). Development of the Childrens Communication
Checklist (CCC): a method for assessing qualitative aspects of com-
municative impairment in children. Journal of Child Psychology and
Psychiatry.
Bishop, D. V. M. & C. Adams (1989). Conversational characteristics of
children with semantic-pragmatic disorder. II: What features lead to
a judgement of inappropriacy? British Journal of Disorders of Com-
munication 24: 24163.
Bishop, D. V. M. & C. Adams (1991). What do referential communication
tasks measure? A study of children with specic language impair-
ment. Applied Psycholinguistics 12: 199215.
Bishop, D. V. M. & L. Rosenbloom (1987). Classication of childhood
language disorders. In Language development and disorders. Clinics in
Developmental Medicine No. 101/102, edited by W. Yule & M. Rutter.
London: MacKeith Press.
Bloom, L. (1970). Language development: form and function in emerging
grammars. Cambridge, MA: MIT Press.
Bloom, L. & M. Lahey (1978). Language development and language disorders.
New York: John Wiley.
Bortolini, U., M. C. Caselli & L. B. Leonard (1997). Grammatical decits in
Italian-speaking children with specic language impairment. Journal
of Speech and Hearing Research 40: 80920.
Boucher, J. (1998). SPD as a distinct diagnostic entity: logical consider-
ations and directions for future research. International Journal of
Language and Communication Disorders 33: 7181.
Bowerman, M. (1982). Evaluating competing linguistic models with lan-
guage acquisition data: implications of developmental errors with
causative verbs. Quaderni di Semantica 3: 566.
Bowerman, M. (1996). The origins of childrens spatial semantic catego-
ries: cognitive versus linguistic determinants. In Gumperz & Levin-
son.
Brett, L., S. Chiat & C. Pilcher (1987). Stages and units in output process-
ing: some evidence from voicing and fronting processes in children.
Language and Cognitive Processes 2: 16577.
Bridgeman, E. & M. Snowling (1988). The perception of phoneme se-
quence: a comparison of dyspraxic and normal children. British
Journal of Disorders of Communication 23: 24552.
Brown, R. (1958). How shall a thing be called? Psychological Review 65:
1421.
Brown, R. (1973). A rst language: the early stages. Cambridge, MA:
Harvard University Press.
References 271
Carey, S. (1978). The child as word learner. In Linguistic theory and
psychological reality, edited by M. Halle, J. Bresnan & G. Miller. Cam-
bridge, MA: MIT Press.
Chiat, S. (1979). The role of the word in phonological development.
Linguistics 17: 591610.
Chiat, S. (1983). Why Mikeys right and my keys wrong: the signicance
of stress and word boundaries in a childs output system. Cognition
14: 275300.
Chiat, S. (1989). The relation between prosodic structure, syllabication
and segmental realization: evidence from a child with fricative stop-
ping. Clinical Linguistics and Phonetics 3: 22342.
Chiat, S. (1994). From lexical access to lexical output: what is the problem
for children with impaired phonology? In First and second language
phonology, edited by M. Yavas. San Diego: Singular Publishing
Group.
Chiat, S. & J. Hunt (1993). Connections between phonology and seman-
tics: an exploration of lexical processing in a language-impaired
child. Child Language Teaching and Therapy 9: 20013.
Chiat, S., J. Law & J. Marshall (eds.) (1997). Language disorders in children
and adults: psycholinguistic approaches to therapy. London: Whurr
Publishers.
Clahsen, H. (1989). The grammatical characterization of developmental
dysphasia. Linguistics 27: 897920.
Clark, E. V. (1993). The lexicon in acquisition. Cambridge: Cambridge
University Press.
Clark, E. V. (1997). Conceptual perspective and lexical choice in acquisi-
tion. Cognition 64: 137.
Clark, R. (1974). Performing without competence. Journal of Child Lan-
guage 1: 110.
Cohen, D. J. (1980). The pathology of the self in primary childhood autism
and Gilles de la Tourette syndrome. Psychiatric Clinics of North Amer-
ica 3: 383402.
Constable, A., J. Stackhouse & B. Wells (1997). Developmental word-
nding difculties and phonological processing: the case of the miss-
ing handcuffs. Applied Psycholinguistics 18; 50736.
Conti-Ramsden, G. & M. Gunn (1986). The development of conversa-
tional disability: a case study. British Journal of Disorders of Communi-
cation 21: 33952.
Conti-Ramsden, G. & M. Jones (1997). Verb use in specic language
impairment. Journal of Speech, Language, and Hearing Research 40:
1298313.
272 References
Cooper, W. E. & J. Paccia-Cooper (1980). Syntax and speech. Cambridge,
MA: Harvard University Press.
Crystal, D., P. Fletcher & M. Garman (1976). The grammatical analysis of
language disability: a procedure for assessment and remediation. London:
Edward Arnold.
Cutler, A. (1996). Prosody and the word boundary problem. In Morgan &
Demuth.
de Villiers, J. G. & P. A. de Villiers (1973). A cross sectional study of the
development of grammatical morphemes in child speech. Journal of
Psycholinguistic Research 3: 26778.
Dodd, B., J. Leahy & G. Hambly (1989). Phonological disorders in
children: underlying cognitive decits. British Journal of Developmen-
tal Psychology 7: 5571.
Dromi, E., L. B. Leonard & M. Shteiman (1993). The grammatical mor-
phology of Hebrew-speaking children with specic language impair-
ment: some competing hypotheses. Journal of Speech and Hearing
Research 36: 76071.
Dunn, L. M. & L. M. Dunn (1981). Peabody Picture Vocabulary Test
Revised. Circle Pines, MN: American Guidance Service.
Dunn, L. M., L. M. Dunn, C. Whetton & D. Pintilie (1982). British Picture
Vocabulary Scale. Windsor: NFER-Nelson.
Ellis, A. W. & A. W. Young (1988). Human cognitive neuropsychology.
Hove: Lawrence Erlbaum.
Evelyn, M. (1996). An investigation of verb processing in a child with a
specic language impairment. Unpublished MPhil thesis, City Uni-
versity, London.
Fisher, C., D. G. Hall, S. Rakowitz & L. Gleitman (1994). When it is better
to receive than to give: syntactic and conceptual constraints on
vocabulary growth. Lingua 92: 33375.
Fletcher, P. & M. Garman (1979, 1986). Language acquisition. Cambridge:
Cambridge University Press.
Fletcher, P. & B. MacWhinney (1995). The Handbook of Child Language.
Oxford: Blackwell.
Fletcher, P. & J. Peters (1984). Characterizing language impairment in
children: an exploratory study. Language Testing 1: 3349.
Gathercole, S. E., C. S. Willis, A. D. Baddeley & H. Emslie (1994). The
Childrens Test of Nonword Repetition: a test of phonological work-
ing memory. Memory 2: 10327.
Gerken, L. (1996). Phonological and distributional information in syntax
acquisition. In Morgan & Demuth.
German, D. J. (1986). National College of Education, Test of Word Finding
References 273
(TWF). Allen, TX: DLM Teaching Resources.
German, D. J. (1987). Spontaneous language proles of children with
word-nding problems. Language, Speech, and Hearing Services in
Schools 18: 21730.
German, D. J. & E. Simon (1991). Analysis of childrens word-nding skills
in discourse. Journal of Speech and Hearing Research 34: 30916.
Gleitman, L. R., E. L. Newport & H. Gleitman (1984). The current status of
the motherese hypothesis. Journal of Child Language 11: 4379.
Gopnik, M. & M. B. Crago (1991). Familial aggregation of a developmental
language disorder. Cognition 39: 150.
Greeneld, P. & J. H. Smith (1976). The structure of communication in early
language development. New York: Academic Press.
Gropen, J., S. Pinker, M. Hollander & R. Goldberg (1991). Affectedness
and direct objects: the role of lexical semantics in the acquisition of
verb argument structure. Cognition 41: 15395.
Gumperz, J. & S. C. Levinson (eds.) (1996). Rethinking linguistic relativity.
Cambridge: Cambridge University Press.
Harley, T. A. (1995). The psychology of language: from data to theory. Hove:
Erlbaum (UK) Taylor & Francis.
Hayman, K. (1996). Phonological awareness therapy for a child with
developmental word nding difculties. Unpublished MSc thesis, City
University, London.
Hirsh-Pasek, K. & R. M. Golinkoff (1996). The origins of grammar: evidence
from early language comprehension. Cambridge, MA: MIT Press.
Hobson, R. P. (1993). Autism and the development of mind. Hove: Lawrence
Erlbaum Associates.
Howard, D. & K. E. Patterson (1992). Pyramids and Palm Trees. Bury St.
Edmunds: Thames Valley Test Company.
Huttenlocher, J. (1974). The origins of language comprehension. In
Theories in cognitive psychology, edited by R. Solso. New York: Erl-
baum.
Ingram, D. (1972). The acquisition of the English verbal auxiliary and
copula in normal and linguistically deviant children. Papers and
Reports on Child Language Development 4: 7991.
Ingram, D. (1989). First language acquisition: method, description and expla-
nation. Cambridge: Cambridge University Press.
Jackendoff, R. (1997). The architecture of the language faculty. Cambridge,
MA: MIT Press.
Johnston, J. R. & T. K. Schery (1976). The use of grammatical morphemes
by children with communication disorders. In Normal and decient
child language, edited by D. M. Morehead & A. E. Morehead. Baltimore:
274 References
University Park Press.
Jones, M. & G. Conti-Ramsden (1997). Verb use in children with SLI. First
Language 17: 16593.
Jusczyk, P. W. & D. G. Kemler Nelson (1996). Syntactic units, prosody,
and psychological reality during infancy. In Morgan & Demuth.
Karmiloff-Smith, A. (1978). The interplay between syntax, semantics,
and phonology in language acquisition processes. In Recent advances
in the psychology of language, edited by R. Campbell & P. Smith. New
York: Plenum.
Katz, N., E. Baker & J. Macnamara (1974). Whats in a name? A study of
how children learn common and proper names. Child Development
45: 46973.
Kyle, J. G. & B. Woll (1985). Sign language: the study of deaf people and their
language. Cambridge: Cambridge University Press.
Landau, B. & L. R. Gleitman (1985). Language and experience: evidence from
the blind child. Cambridge, MA: Harvard University Press.
Lee, A., R. P. Hobson & S. Chiat (1994). I, you, me, and autism: an
experimental study. Journal of Autism and Developmental Disorders 24:
15576.
Lees, J. & S. Urwin (1997). Children with language disorders. Second edition.
London: Whurr Publishers.
Leonard, L. B. (1992). Specic language impairment in three languages:
some cross-linguistic evidence. In Specic speech and language disorders
in children, edited by P. Fletcher & D. Hall. London: Whurr Publishers.
Leonard, L. B. (1998). Children with specic language impairment. Cam-
bridge, MA: MIT Press.
Leonard, L. B., U. Bortolini, M. C. Caselli, K. K. McGregor & L. Sabbadini
(1992). Morphological decits in children with specic language
impairment: the status of features in the underlying grammar. Lan-
guage Acquisition 2: 15179.
Leonard, L. B. & J. A. Eyer (1996). Decits of grammatical morphology in
children with specic language impairment and their implications
for notions of bootstrapping. In Morgan & Demuth.
Leonard, L. B., J. A. Eyer, L. M. Bedore & B. G. Grela (1997). Three
accounts of the grammatical morpheme difculties of English-speak-
ing children with specic language impairment. Journal of Speech and
Hearing Research 40: 74153.
Leonard, L. B., L. Sabbadini, J. S. Leonard & V. Volterra (1987). Specic
language impairment in children: a cross-linguistic study. Brain and
Language 32: 23352.
Levelt, W. J. M. (1989). Speaking: from intention to articulation. Cambridge,
References 275
MA: MIT Press.
MacWhinney, B. & E. Bates (eds.) (1989). The crosslinguistic study of
sentence processing. Cambridge: Cambridge University Press.
Markman, E. M. (1989). Categorization and naming in children: problems of
induction. Cambridge, MA: MIT Press.
Marshall, J., M. Black, S. Byng, S. Chiat & T. Pring (1999). The Sentence
Processing Resource Pack. Bicester: Winslow Press.
Marshall, J., T. Pring & S. Chiat (1993). Sentence processing therapy:
working at the level of the event. Aphasiology 7: 17799.
Mayes, L., D. Cohen & A. Klin (1993). Desire and fantasy: a psycho-
analytic perspective on theory of mind and autism. In Baron-Cohen,
Tager-Flusberg & Cohen.
McNamara, M., A. Carter, B. McIntosh & L. Gerken (1998). Sensitivity to
grammatical morphemes in children with specic language impair-
ment. Journal of Speech, Language and Hearing Research 41: 114757.
McTear, M. F. (1985). Pragmatic disorders: a case study of conversational
disability. British Journal of Disorders of Communication 20: 12942.
McTear, M. (1991). Is there such a thing as conversational disability? In
Child language disability, Volume II: Semantic and pragmatic difculties,
edited by K. Mogford-Bevan & J. Sadler. Clevedon: Multilingual Mat-
ters.
Miller, J. & R. Chapman, (19827). SALT: Systematic Analysis of Language
Transcripts. Madison: University of Wisconsin, Language Analysis
Laboratory, Waisman Center.
Mogford-Bevan, K. & J. Sadler (eds.) (1991). Semantic and pragmatic
difculties or semantic-pragmatic syndrome? Some explanations. In
Child language disability, Volume II: Semantic and pragmatic difculties,
edited by K. Mogford-Bevan & J. Sadler. Clevedon: Multilingual Mat-
ters.
Moore, M. E. & J. R. Johnston (1993). Expressions of past time by norma-
land language-impaired children. Applied Psycholinguistics 14: 515
34.
Morgan, J. L. & K. Demuth (eds.) (1996). Signal to syntax: bootstrapping
from speech to grammar in early acquisition. Mahwah, NJ: Lawrence
Erlbaum Associates.
Nelson, K. (1973). Structure and strategy in learning to talk. Monographs
of the Society of Research in Child Development 38, No. 149.
Newport, E. L., H. Gleitman & L. R. Gleitman (1977). Mother, Id rather do
it myself: some effects and non-effects of maternal speech style. In
Talking to children: language input and acquisition, edited by C. E. Snow
& C. A. Ferguson. Cambridge: Cambridge University Press.
276 References
Oetting, J. B. & J. E. Horohov (1997). Past-tense marking by children with
and without specic language impairment. Journal of Speech and
Hearing Research 40: 6274.
OHara, M. & J. Johnston (1997). Syntactic bootstrapping in children with
specic language impairment. European Journal of Disorders of Com-
munication 32: 189205.
Owens, R. E. (1996). Language development: an introduction. Fourth edition.
Needham Heights, MA: Allyn & Bacon.
Peters, A. M. (1983). The units of language acquisition. Cambridge: Cam-
bridge University Press.
Pinker, S. (1989). Learnability and cognition: the acquisition of argument
structure. Cambridge, MA: MIT Press.
Rapin, I. & D. A. Allen (1987). Developmental dysphasia and autism in
preschool children: characteristics and subtypes. Proceedings of the
First International Symposium on Specic Speech and Language Disorders
in Children. London: AFASIC.
Renfrew, C. (1980). Word-nding vocabulary scale. Oxford: C. E. Renfrew.
Rescorla, L. A. (1980). Overextension in early language development.
Journal of Child Language 7: 32135.
Rice, M. L. & J. V. Bode (1993). GAPS in the verb lexicons of children with
specic language impairment. First Language 13: 11331.
Rice, M. L., J. C. Buhr & M. Nemeth (1990). Fast mapping word learning
abilities of language-delayed preschoolers. Journal of Speech and Hear-
ing Disorders 55: 3342.
Rice, M. L. & J. B. Oetting (1993). Morphological decits of children with
SLI: evaluation of number marking and agreement. Journal of Speech
and Hearing Research 36: 124957.
Rom, A. & L. B. Leonard (1990). Interpreting decits in grammatical
morphology in specically language-impaired children: preliminary
evidence from Hebrew. Clinical Linguistics & Phonetics 4: 93105.
Sachs, J. & L. Truswell (1978). Comprehension of two-word instructions
by children in the one-word stage. Journal of Child Language 5: 1724.
Saeed, J. I. (1997). Semantics. Oxford: Blackwell.
Selkirk, E. O. (1984). Phonology and syntax: the relation between sound and
structure. Cambridge, MA: MIT Press.
Shatz, M., H. M. Wellman & S. Silber (1983). The acquisition of mental
verbs: a systematic investigation of the rst reference to mental state.
Cognition 14: 30121.
Shipley, E. F., C. S. Smith & L. R. Gleitman (1969). A study in the
acquisition of language: free responses to commands. Language 45:
32242.
References 277
Slobin, D. I. (1973). Cognitive prerequisites for the acquisition of gram-
mar. In Studies of child language development, edited by C. A. Ferguson
& D. I. Slobin. New York: Holt, Rinehart & Winston.
Slobin, D. I. (ed.) (1985). The crosslinguistic study of language acquisition,
Volumes 12. Hillsdale, NJ: Erlbaum.
Slobin, D. I. (1996). From thought and language to thinking for speak-
ing. In Gumperz & Levinson.
Slobin, D. I. (ed.) (1997). The crosslinguistic study of language acquisition,
Volume 5. Hillsdale, NJ: Erlbaum.
Snow, C. E. (1977). The development of conversation between mothers
and babies. Journal of Child Language 4: 122.
Snow, C. E. (1995). Issues in the study of input: netuning, universality,
individual and developmental differences, and necessary causes. In
The handbook of child language, edited by P. Fletcher & B. MacWhin-
ney. Oxford: Blackwell.
Stackhouse, J. & B. Wells (1997). Childrens speech and literacy difculties: a
psycholinguistic framework. London: Whurr Publishers.
Tager-Flusberg, H. (1993). What language reveals about the understand-
ing of minds in children with autism. In Baron-Cohen, Tager-Flus-
berg & Cohen.
Temple, C. (1997). Developmental cognitive neuropsychology. Hove: Psy-
chology Press.
Tomasello, M. (1987). Learning to use prepositions: a case study. Journal
of Child Language 14: 7998.
Tomasello, M. (1995). Joint attention as social cognition. In Joint atten-
tion: its origins and role in development, edited by C. Moore & P. J.
Dunham. Hillsdale, NJ: Lawrence Erlbaum Associates.
Tomasello, M. & M. Barton (1994). Learning words in nonostensive
contexts. Developmental Psychology 30: 63950.
Tomasello, M. & J. Farrar (1986). Joint attention and early language.
Child Development 57: 145463.
Tomasello, M. & A.C. Kruger (1992). Joint attention on actions: acquiring
verbs in ostensive and non-ostensive contexts. Journal of Child Lan-
guage 19: 31133.
Tomasello, M. & W. E. Merriman (1995) Beyond names for things: young
childrens acquisition of verbs. Hillsdale, NJ: Lawrence Erlbaum Associ-
ates.
Tomasello, M., R. Strosberg & N. Akhtar (1996). Eighteen-month-old
children learn words in non-ostensive contexts. Journal of Child Lan-
guage 23: 15776.
Tomasello, M. & J. Todd (1983). Joint attention and lexical acquisition
278 References
style. First Language 4: 197212.
Van der Lely, H. K. J. (1993). Specic language impairment in children:
research ndings and their therapeutic implications. European Jour-
nal of Disorders of Communication 28: 24761.
Van der Lely, H. K. J. & M. Harris (1990). Comprehension of reversible
sentences in specically language impaired children. Journal of Speech
and Hearing Disorders 55: 10117.
Watkins, R. V., M. L. Rice & C. C. Moltz (1993). Verb use by language-
impaired and normally developing children. First Language 13: 133
43.
Weismer, G., D. Dinnsen & M. Elbert (1981). A study of the voicing
distinction associated with omitted, word-nal stops. Journal of Speech
and Hearing Disorders 46: 3207.
Wepman, J. M. & W. M. Reynolds (1987). Wepmans Auditory Discrimina-
tion Test. Second edition. Los Angeles: Western Psychological Servi-
ces.
Wiig, E. H., W. Secord & E. Semel (1992). Clinical Evaluation of Language
Fundamentals Pre-school. Sidcup, Kent: The Psychological Corpor-
ation.
Index
Note: References to gures, tables and notes are given as follows: e.g. 80f, 219t,
87n.
abstract words, 19 auxiliary verbs, 1834, 187, 189,

acoustic analysis, 856 1945, 201, 211t
acting-out tasks, 56, 152, 154, 158, concepts, 21214
16970, 1712 contraction, 184, 187
Adams, C., 2389, 241 elliptical responses, 187, 212, 215,
adjectives, 16, 185 223
phrases, 118, 119 input, 220, 221
semantic discrimination, 53 negatives, 21112, 21718
adultchild interaction, 224, 2930, normal development, 187, 233
312, 1423 omission, 211t
adverbials, 213 phonology, 21416, 218
affective experience, 2289 position, 21719, 218t, 219t
decits in, 2359, 260 processing problems, 21024
sensitivity to, 2313, 2447 questions, 210, 211, 222
agent, 121, 122, 1234, 146 recognition of, 21920, 220t
agreement markers, 1845, 201, 202 reduction, 183
Akhtar, N., 24 repetition, 21619, 217t, 2201
Aksu-Koc, A. A., 131
Allen, D. A., 237 Baddeley, A. D., 62
Ambalu, D., 23, 143 Baker, E., 195
Anthony, A., 33 Baldwin, D. A., 24
argument structures, 1204 Baron-Cohen, S., 236
comprehension, 1278 Barton, M., 25
processing, 1478, 148f basic-level terms, 21, 230
see also verbargument structures Bedore, L. M., 189, 191, 192
arguments, 120, 148 Berko, J., 186
problems with, 145 bilingualism, 199200
articulation, 33, 614, 70, 78, 105 bilabial stops, 71
auditory discrimination, 756, 77, Bishop, D. V. M., 114, 237, 2389,
99100, 101 241
Auditory Discrimination Tests, 33 blindness, 2345
autism, 2367, 2401 Bloom, L., 115, 129
279
280 Index
Bode, J. V., 145 connectors, 233
Bogle, D., 33 consonants
bootstrapping, 301, 13742 deletion, 71, 209
interaction, 2045, 2646 discrimination, 434
phonological, 140, 178, 200, 202, fricatives, 701, 80f, 81f
203, 215, 2334 phonological processes, 6971
semantic, 1389, 1489, 15860, position, 43, 44f
1612, 179, 202, 233 reduced clusters, 209
syntactic, 13940, 148, 1601, sequencing, 44
162, 1789 stops, 701, 845
Bortolini, U., 189, 190, 191, 196 voicing, 835, 99
Boucher, J., 238 Constable, A., 34, 88, 93, 109n
Bowerman, M., 131, 133 content words, 183, 184, 1856
BPVS see British Picture Vocabulary distortion, 209
Scales normal development, 1856
Brett, L., 77, 84 see also adjectives; nouns; verbs
Bridgeman, E., 94 Conti-Ramsden, G., 145, 251, 2523
Bridgeman & Snowling test, 945, 101 conversation, 2389, 2412
British Picture Vocabulary Scales Crago, M. B., 189, 196
(BPVS), 33, 89, 93, 163 Crystal, D., 115
Brown, R., 21, 26, 27, 186, 187 cued naming, 61
Buhr, J. C., 34 Cutler, A., 29
Carey, S., 17 de Villiers, J. G., 186

Carter, A., 194 de Villiers, P.A., 186
Caselli, M. C., 189, 190, 191, 196 deafness, 68
causation, 1223, 1334 deniteness, 205
CELF see Clinical Evaluation of determiners, 185
Language Fundamentals in language-impaired children, 189,
Chapman, R., 115 191, 1945
Chiat, S., 23, 27, 39, 56, 77, 84, 87n, normal development, 186, 187,
109n, 143, 240 1956, 199, 205
child-directed language, 22, 1423 Dinnsen, D., 85
Childrens Communication Checklist, direct objects, 123
238 discrimination tasks, 426, 48, 63
circumlocution, 34, 94 distractor, 501, 60
Clahsen, H., 190, 191 Dodd, B., 75, 76
Clark, E. V., 17, 20, 31, 230 Dromi, E., 190, 191
Clark, R., 26 Dunn, L. M., 33
classication tasks, 758 dysphasia, 12
Clinical Evaluation of Language
Fundamentals (CELF), 94, 115 Edinburgh Articulation Test, 33
Cohen, D., 236, 260 Elbert, M., 85
Cohen, D. J., 237 Emslie, H., 62
communication Evelyn, M., 180n
aids, 46 events, 537, 11626
tasks, 2412 acting out, 56, 152, 154, 158,
see also conversation 16970, 1712
comprehension, 334, 11415, conditions and consequences of,
1278, 146, 154, 16972, 207 2479
Index 281
copying tasks, 152 Gathercole, S. E., 62
description tasks, 1523, 158, gender, 203
1639, 1736 Gerken, L., 194
invisible, 135, 2447 German, D. J., 33, 34
participants, 1204, 129 Gleitman, H., 19
sensitivity to, 2449 Gleitman, L. R., 19, 117, 139, 160,
experience 193, 235
childrens packaging of, 22934 goal, 117, 121, 122, 1467
disabilities, 2349, 260 Goldberg, R., 138, 159, 160
exploring through language, Golinkoff, R. M., 29, 128, 142
23942 Gopnik, M., 189, 196
phonologising of, 2634 Greeneld, P., 22
sensitivity to, 2312, 2447, Grela, B. G., 189, 191, 192
24950, 260 Gropen, J., 138, 159, 160
tiers of, 2289, 2301, 233, Guess who? game, 25660
23940 Gunn, M., 251, 2523
experiencer, 121
experiments, the structured situation Hall, D.G., 117, 139, 160
of, 2412 Hambly, G., 75, 76
Eyer, J. A., 189, 191, 192, 203 Harris, M., 146
Hayman, K., 13
failure at tasks, 45, 48, 56, 57, 60, hearing impairment, 68
61, 635 Hirsh-Pasek, K., 29, 128, 142
Farrar, J., 23 Hirson, A., 26
fast mapping, 34 Hobson, R. P., 236, 237, 240, 260
ngerspelling, 6 Hollander, M., 138, 159, 160
Fisher, C., 117, 139, 160 homophones, 15
Fletcher, P., 115, 144, 187 Horohov, J. E., 189, 192
fricatives, 701, 80f, 81f Howard, D., 94
fronting, 70, 71, 778, 802 Hunt, J., 109n
function morphemes, 1834 Huttenlocher, J., 127
in language-impaired children,
18892, 1945, 1967, 203, inections, 131, 183, 184, 1867,
21024 189, 1912
meaning cues, 2001, 203 Ingram, D., 17, 33, 129, 189
normal development, 1858, input processing, 3, 36, 37f, 74, 149
1934, 1956, 198 acting-out tasks, 16970, 1712
novel uses, 192 classication tasks, 758
obligatory aspect, 1967 lexical processing problems, 956,
other languages, 1901, 190t, 196, 96f, 98102
199200, 2034 picture-pointing tasks, 1702
phonology and, 198200, 2015 repetition tasks, 1778
position, 199 internal judgement tasks, 5761
semantic factors, 184, 2023, 204, IQ tests, 2412
205
stress, 2034, 21415 Johnston, J. R., 146, 189, 213
syntax, 185, 204 joint attention episodes, 223, 25
unreliability of use, 191, 21415 Jones, M., 145
judgement tasks, 469, 1001, 1557
Garman, M., 115, 187 Jusczyk, P. W., 29
282 Index
Karmiloff-Smith, A., 203 repercussions for lexical
Katz, N., 195 development, 389
Kaufman Assessment Battery for Liberator, 45
Children, 93 light pointers, 4
Kemler Nelson, D. G., 29 lip-reading, 6
Klin, A., 236, 260 location, 1201, 122, 123, 129
Kruger, A. C., 23, 143
Kyle, J. G., 6 Macnamara, J., 195
mapping, 1719, 34, 127, 131, 142,
labio-dental fricatives, 71 264f
Lahey. M., 115 Markman, E. M., 21
Landau, B., 235 Marshall, J., 39, 56
Language Assessment, Remediation & Mayes, L., 236, 260
Screening Procedure (LARSP), McGregor, K. K., 189, 190, 191, 196
115 McIntosh, B., 194
language development McIsaac, M. W., 33
function morphemes, 1858, McNamara, M., 194
1924, 1956, 197205 McTear, M. F., 2378, 2534
verbs and verb structures, 12743 mean length of utterance (MLU), 189,
words, 1732, 22934 208
language impairment meaning, 3, 2279
function morphemes, 18892, childrens development of, 22934
1945, 1967, 2034, 21024 experience decits, 23442, 251
mean length of utterance (MLU), form and, 264f
189, 208 tiers of experience, 2289, 2301,
non-words and, 161 233, 23942
verbs, 14480 see also pragmatic processing
language processing, 34, 6, 2626 mental experience, 2289, 2313
connectionist theories, 107 decits, 2359, 260
decits, 359, 14750 sensitivity to, 2313, 2447
LARSP (Language Assessment, mental lexicon, 15, 389
Remediation & Screening Miller, J., 115
Procedure), 115 mind-blindness, 236
Law, J., 39 mine and yours, 231
Leahy, J., 75, 76 MLU (mean length of utterance), 189,
Lee, A., 240 208
Lees, J., 33, 115 modal verbs, 187, 214
Leonard, L. B., 189, 190, 191, 192, Mogford-Bevan, K., 238
196, 203 Moltz, C. C., 144
lexical decision tasks, 469, 1001, Moore, M. E., 213
150 Moriarty, J., 236
lexical processing problems, 3335, movement events, 119, 1234
88109
evidence from tasks, 4065, negatives, 21112, 21718
97106 Nelson, K., 22
input processing, 36, 37f, 956, Nemeth, M., 34
96f, 98102 Newport, E. L., 19
output processing, 368, 96, 97f, non-words
1026 discrimination and, 43
possible sources, 358, 956 judgement, 469, 989, 101
Index 283
language-impaired children and, acoustic analysis, 856
161 evidence from input/output tasks,
repetition, 62, 1035, 1056 758
nouns, 16 evidence from output patterns,
count, 16 7885, 1778
mass, 16 possible sources, 735
normal development, 223, 24, 25, word position, 221
1856 phonology, 14, 15
number, 184 auxiliary verbs, 21416, 218
phrases, 126 distractors, 501, 60
plural markers, 187, 191 function morphemes, 198200,
semantic discrimination, 512 2015
processes, 6971, 7587
obsessive behaviour, 251 processing, 734, 73f, 867, 1059,
Oetting, J. B., 189, 191, 192 107f, 108f, 17780
OHara, M., 146 representation, 369, 469, 74,
on-line processing, 389 767, 956, 98102, 1059
ostensive conditions, 23 sentences, 1256, 141, 143, 183
output processing, 23, 36, 37, 38f, verbs, 147
149 see also bootstrapping; phonological
lexical processing problems, 96, 97f, discrimination tasks;
1026 phonological problems
picture-description tasks, 1736 picture-description tasks, 1523,
repetition tasks, 1767 1736
repetition with pictures tasks, picture-pointing tasks, 1502, 1702
1778 picture-sorting tasks, 516, 57
overgeneralisation, 21, 188 Pilcher, C., 77, 84
Owens, R. E., 186, 187, 193 Pinker, S., 134, 138, 159, 160
Pintilie, D., 33
participants, 147 pitch, 2930
see also arguments possession, 1201, 122, 123, 129,
particles, 119, 212 1701, 201
Patterson, K. E., 94 pragmatic language impairment,
Peabody Picture Vocabulary Test, 33 2389, 241
perceptual experience, 2289, 2345 pragmatic processing
perspective, 117, 1224, 231, 25160 problems, 25161
Peters, A. M., 25 strengths, 24350
Peters, J., 144 preferential listening technique, 289,
phonological bootstrapping see 141
bootstrapping preferential looking technique, 128,
phonological delay, 72 141
phonological discrimination tasks, prepositions, 28, 119, 193, 199200,
426 2001
consonants, 434 Pring, T., 23, 56, 143
failure, 45 pronouns, 186, 188, 2401, 2512
syllables, 44 psycholinguistic processes, 23
vowels, 44 psycholinguistic tasks, 4265, 15062
phonological disorder, 72 Pyramids and Palm Trees, 94
phonological problems, 6975, 107,
108, 21424 questions, 210, 211, 222
284 Index
Rakowitz, S., 117, 139, 160 tests of, 94
Rapin, I., 237 see also bootstrapping; semantic
Renfrew, C., 33 problems
Renfrew word-nding test, 89 Semel, E., 115
repetition tasks sentences
auxiliary verbs, 21619, 2201 comprehension, 11415, 1278
lexical processing, 614, 1035, focus, 11617, 1348
1056 judgement tasks, 1557
with pictures, 1778 production, 11314, 115
verbargument structures, 1578, reversible, 154
1767, 178 rhythm, 1256, 141, 143, 183
Rescorla, L. A., 21 semantics, 11724
Reynell Developmental Language structure see phonology, sentences;
Scales, 207 syntax
Reynolds, W. M., 33 Shatz, M., 232
rhyme, 589, 601, 95 Shipley, C. S., 193
rhythm Shipley, E. F., 193
sentences, 1256, 141, 143, 183 Shteiman, M., 190, 191
speech processing, 823, 141, 143 Sign language, 6, 7
word acquisition, 289 Silber, S., 232
Rice, M. L., 34, 144, 145, 191, 192 Simon, E., 34
Ring, H., 236 Slobin, D. I., 131, 186, 200, 203, 205
Rom, A., 190 Smith, C. S., 193
Rosen, L., 237 Smith, J. H., 22
Rosenbloom, L., 237, 238 Snow, C. E., 30
Snowling, M., 94
Sabbadini, L., 189, 190, 191, 196 source, 117, 121, 122
Sachs, J., 128 speech
Sadler, J., 238 ltering, 2531, 1347
SALT (Systematic Analysis of impairment, 36, 8, 669
Language Transcripts), 115 processing, 803, 141, 143
scenes synthesis, 4, 86
core components, 2278, 228f Stackhouse, J., 33, 34, 88, 93, 109n
encoding, 11617 state, 118, 1201, 122, 123, 129
ltering, 204, 301, 1347 psychological, 1301
Schery, T. K., 189 stimulus, 121
Schmitz, B., 236 stopping, 701, 802, 83
Secord, W., 115 stops, 701, 80t, 82t
segmentation, 1720, 26, 29, 30 voicing, 845
semantic bootstrapping see stress
bootstrapping function morphemes, 2034,
semantic problems, 401, 4957 21415
events, 537 sentences, 1256
objects, 512 speech processing, 803
properties, 53 word acquisition, 268, 29, 1989
semanticpragmatic disorder, 2379, see also rhythm
241 Strosberg, R., 24
semantics syllables
distractors, 501, 60 discrimination, 44
information, 14, 1516 unstressed, 209
Index 285
symbol boards, 4 problems, 1456, 1789
syntactic bootstrapping see see also verbs, problems
bootstrapping processing, 1479, 149f, 162
syntax, 14, 15, 16, 11824, 185, repetition tasks, 1578, 1768
204 repetition with pictures tasks,
see also bootstrapping 1778
Systematic Analysis of Language verb omission, 165, 1667
Transcripts (SALT), 115 verbs, 11726
acting-out tasks, 152
Tager-Flusberg, H., 236 agreement, 201
tense, 184, 201, 213 auxiliary see auxiliary verbs
Test for the Reception of Grammar comprehension, 1278
(TROG), 89, 93, 114, 163, 207 inections, 184, 186, 187, 191
Test of Word Finding, 33, 93 lexical decision tasks, 150
thematic roles, 1201 mental state, 232
see also verbargument structures modal see modal verbs, 187, 214
theme, 120, 1234, 133, 1467 movement, 119, 1234
thought, 56 normal development, 234, 245,
time, 184, 201, 213, 234, 2523 28, 1324, 232
Todd, J., 23 overregularisation, 192
Tomasello, M., 23, 24, 25, 143, 188, phonology, 147
232, 235 picture-pointing tasks, 1502
TROG see Test for the Reception of possession, 1201, 122, 123,
Grammar 1701
Truswell, L., 128 problems, 14480
processing, 14750, 148f, 15862
undergeneralisation, 21 semantic discrimination, 537, 147,
Urwin, S., 33, 115 1501
utterance scenes, 229, 236 state, 118, 1201, 122, 123
structures, 11720
Van der Lely, H. K. J., 146 timing, 184, 201, 213
velar stops, 70, 80t, 82t visual impairment, 2345
verbargument structures, 1206, vocabulary, 335
1401 mental lexicon, 15, 389
acting-out tasks, 16970 problems, 144
bare verbs, 1656, 1678 see also language development,
bootstrapping tasks, 15862 words; lexical processing
comprehension, 1278, 154, problems
16972 voice, 71
development of, 12934, 1389 voicing, 71, 835, 99
errors, 1324 vowels
event-description tasks, 1523, discrimination, 44
1639 reduction, 183
ltering, 1347
input investigations, 16972 Watkins, R. V., 144
judgement tasks, 1557 Wechsler Intelligence Scales for
output investigations, 1739 Children (WISC), 8990
picture-description tasks, 1523, Weismer, G., 85
1736 Wellman, H. M., 232
picture-pointing tasks, 1701 Wells, B., 33, 34, 88, 93, 109n
286 Index
Wepman, J. M., 33 see also phonology, representation
Whetton, C., 33 word-to-picture matching tasks,
Wiig, E. H., 115 4951
Willis, C. S., 62 words, 1316
Woll, B., 6 acquisition see language
word combinations, 12831, 199 development, words
see also verbargument structures content, 183, 184
word-nding, 33, 34, 93 as packaging of experience, 22934
see also lexical processing problems as perspective on experience,
Word-nding Vocabulary Scale, 33 23042
word order, 1212, 128, 130, 131 unstressed, 1834
word representation, 14, 101, 103, see also vocabulary; lexical
104, 105, 106 processing problems

Shula Chiat

Uploaded by

Copyright:

Available Formats

Shula Chiat

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Shula Chiat

Uploaded by

Copyright:

Available Formats

MMMM

This page intentionally left blank

Dr Shula Chiat is Senior Lecturer in Linguistics at the Depart-

Also in the series

Cambridge University Press

Shula Chiat 2000

This book is in copyright. Subject to statutory exception and to the provision of

First published in print format 2000

ISBN-13 978-0-511-06788-4 eBook (EBL)

ISBN-13 978-0-521-57386-3 hardback

ISBN-13 978-0-521-57474-7 paperback

Cambridge University Press has no responsibility for the persistence or accuracy of

Part I Problems with words

Part II Grappling with verb structure

Part III Missing function morphemes

Endpoint and springboard 262

input Forms in bold highlight technical terms which are accom-

Ruth: My mum and dad dan me. And my brothers.

But what does it leave him feeling? Not anger:

Anger is from the your mind

he says, pointing to his head,

For Damian, it is a relief to talk his own language, which he feels

Some children, as Ian says, get trouble by talking. They may

In general, we experience the connections as instantaneous. We

What would happen if Anthony did not have communication

Anthony Robertson with Liberator

When a child cannot produce recognisable speech and does not

A: I wasnt thought to be bright till when people sat down and

So, Anthonys apparent silence does not reect a silence in his

We continue in speech, supported by some use of Sign:

Sarahs experience at her work-place provides an example of what

Note: To clarify my use of gender-marked personal pronouns when refer-

Seven-year-old Steven struggles to name a microphone:

Six-year-old Eamonn requests a favoured toy:

Eamonns search is for the deep-sea diver.

Semantically, hair is most similar to words which refer to parts

Det Adjective Noun

the famous skin/nail/wire/soil/bread

In acquiring language, children acquire thousands of words.

Who tunes into who?

Conversely, an unstressed syllable which is part of a word may be

Connecting speech and scenes

Ooh, feel the ice!

What of children who are not developing language normally? For

In attempting to produce words, children may make actual errors.

All of these children are clearly attempting to produce words and

Where the blocks may lie

Discrimination of relevant Discrimination of relevant

Identification of word Identification of word

Storage of word Storage of word

Figure 3.1 Components of input processing

As with input, problems could conceivably arise in one or more

Selection of word Accessing of word

Figure 3.2 Components of output processing

Repercussions for lexical development

Thinking about where problems could arise begs the question of

The rst appears to be a semantic confusion, with Eamonn select-

From hearing speech to phonology: phonological

Discrimination Discrimination Discrimination Discrimination

keytea sitsick eatbeat beebeat