Neuropsychologia,Vol. 36, No. 4, pp. 343 351, 1998
~ )
Pergamon
PII: S0028 3932(97)00133 4
1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0028 3932/98 $19.00+0.00
Linguistic dissociations in Williams syndrome:
evaluating receptive syntax in on-line and off-line
tasks
ANNETTE
KARMILOFF-SMITH,*
LORRAINE
K. TYLER,t KATE
KERRY SIMS,* ORLEE UDWIN,+ + PATRICIA HOWLIN§ and MARK
VOICE,t
DAVIES§
*MRC Cognitive Development Unit and University College, London, U.K.; tCentre for Speech and Language, Birkbeck College,
London, U.K.; +West Lambeth Community Care Trust, London, U.K.; §St George's Hospital Medical School, London, U.K.
(Received 11 Not,ember 1996; accepted 16 Juh' 1997)
Abstract--Williams syndrome (WS) is a neurodevelopmental disorder of genetic origin which results in relatively spared language in
the face of serious non-verbal deficits. There is controversy, however, about how intact WS language abilities are. The discussion
has focused on impairments of lexico-semantics and of morphological feature analysis, with the presumption that WS syntax is
intact. We challenged this view and assessed WS receptive syntax by using two tasks testing various syntactic structures: an on-line
word monitoring task and an off-line picture-pointing task. WS performance on the off-line task was generally poor. By contrast,
their performance on the on-line task was far better and allowed us to ascertain precisely which aspects of WS receptive syntax are
preserved and which are impaired. WS participants were sensitive to the violation of auxiliary markers and phrase structure rules
but, unlike both the normal young and elderly controls, they did not show sensitivity to violations of subcategory constraints. The
present study suggests that there exist dissociations within WS language which are not restricted to lexico-semantics or to morphological feature analysis, but which also invade their processing of certain syntactic structures. We conclude by arguing that WS
syntax is not intact and that their language might turn out to be more like second language learning than normal acquisition.
~ 1998 Elsevier Science Ltd. All rights reserved
Key Words: Williams syndrome; genetic disorder; on-line/off-line tasks; preserved syntax; impaired syntax.
acquires a near normal volume relationship to the posterior cortex [5, 28, 74]. Recent work, however, points to
cytoarchitectonic anomalies in the form of an exaggerated horizontal organization of neurons within layers,
increased cell packing density in certain brain regions,
decreased myelination, and abnormally clustered and oriented neurons, particularly in the visual cortex [23]. At
the neurotransmitter level, there appear to be abnormal
levels of calcitonin [17] and of serotonin [2, 45].
Most importantly for the cognitive neuroscientist, WS
is unlike many other developmental disorders, such as
Down's syndrome, in that it results in a very uneven
linguistic-cognitive profile. Some aspects of language
seem relatively spared, whereas many non-linguistic functions, such as spatial cognition, number, planning, and
problem solving, are severely impaired [1,3, 4, 10, 13, 16,
25, 32, 33, 38, 39, 40, 50, 51, 58, 67, 69, 70, 71, 73].
Despite the original claims that WS language is relatively normal, a series of studies using event-related
potentials to assess the timing and organization of neural
Introduction
Williams syndrome (WS) is a rare contiguous gene
disorder found in about one in 25 000 live births [6, 7, 15,
21, 26, 41, 80]. It is caused by a hemizygous submicroscopic deletion at chromosome 7ql 1.23, including
both the elastin gene [19, 20, 42] and the LIM-Kinase 1
gene [22, 61]. At the level of the phenotype, WS patients
have a facial dysmorphology commonly characterized as
"elfin faces", malformations of the connective tissues,
and supravalvular aortic stenosis [30, 44, 45, 53].
In general WS presents no evidence for focal lesions.
Rather, specific patterns of brain growth during embryoand ontogenesis result in different brain volume proportions, with total size being 80per cent of normal
brains. The WS cerebrum is small, but the frontal cortex
*Address for correspondence: MRC, Cognitive Development Unit, 4 Taviton Street. London WC1H 0BT, U.K.
E-mail: annetteca cdu.ucl.ac.uk
343
344
A. Karmiloff-Smith et al./Linguistic dissociations in Williams syndrome
systems during language processing question this conclusion [54, 55]. Participants listened to sentences that
ended either appropriately semantically (e.g., "I take my
coffee with cream and sugar") or to sentences which were
semantically inappropriate (e.g., "I take my coffee with
cream and radiator"). The authors found different temporal patterns in the brain activity of WS participants
compared with normal controls, suggesting that on-line
semantic processing is different in WS compared with the
normally developing brain (see also, discussion in [67]).
However, even if some aspects of semantic processing
are impaired, WS syntactic processing might still be
intact. The purpose of the present experiments was to
determine whether this is so, and if not, which aspects of
syntactic processing are preserved in WS participants and
which are impaired. To examine WS ability to process
syntactic structure when interpreting an utterance we
used two different tasks: one (the word monitoring task)
taps implicit processing, and the other (the sentencepicture matching task) probes explicit syntactic processing.
Implicit and explicit tasks measure different aspects of
language processing. We define implicit tasks as those in
which the subject's response is closely time-locked to
a relevant linguistic variable, and where the subject's
attention is not explicitly drawn to this linguistic variable.
Tasks meeting these criteria reflect more directly the nonconscious, automatic aspects of language processing [63,
66]. Explicit tasks, in contrast, focus more on the controlled aspects of processing and are not directly concerned with the real-time analysis of spoken language
[9, 63, 66]. Whereas we acknowledge that the difference
between implicit and explicit tasks is not always entirely
clear, in general they probe different aspects of the comprehension process. The distinction between them has
been valuable in elucidating both normal development
[8, 31, 36, 37], and the nature of cognitive impairments
in a wide range of adult deficits (e.g., [52, 63, 64, 66, 79]).
Children are typically successful earlier in development
on tasks that tap their knowledge indirectly and implicitly
than on explicit tasks [8, 27, 31, 36, 59, 62, 72]. Likewise,
language-impaired patients typically show preserved
implicit and impaired explicit processing, with explicit
tasks tending in general to overestimate the severity of a
patient's deficit, strengthening the need for using both
types of task [63, 64]. This is also true in the case of
subjects, such as those with Williams syndrome, who have
serious cognitive impairments [67]. For these reasons,
we investigated the syntactic processing abilities of WS
people by using both implicit and explicit tasks, and by
examining a range of different kinds of syntactic structure.
Experiment 1: Word monitoring study
In this study, we examined WS participants' ability to
use syntactic information in the process of interpreting a
sentence, by measuring their sensitivity to various types
of syntactic violation. We did this by using a word monitoring task in which participants listen to spoken sentences and press a response key when they hear a prespecified target word in the sentence [48, 49, 63, 66]. The
word monitoring task taps into the real-time, automatic
processing of language, because there is a close temporal
relationship between the speech input and the subject's
response. The subject is focused on the task of monitoring
for a target word rather than on the linguistic manipulations introduced by the experimenter.
We compared monitoring latencies to target words
occurring in particular syntactic constructions, and then
violated those constructions to determine whether monitoring latencies increase. In normal controls, monitoring
latencies increase whenever a target word follows a syntactic violation [47]. If WS participants are sensitive to
various types of syntactic structure, their latencies to
target words will also increase following the presence of
a syntactic violation to those structures.
Method
Participants. Eight individuals with Williams syndrome were
tested. There were three male and five female participants. They
had a mean chronological age of 20;7 years (range 14;9 to 34;8),
with a mean verbal IQ on the WISC/WAIS of 71 (range 51-87)
and a mean performance IQ of 58 (range 46-75) [76, 77, 78].
Their mental age on the British Picture Vocabulary Scale
(BPVS) [18] was 11;2 (range 7;5 to 16;4), on the Test for Reception of Grammar (TROG) [11] 7;7 (range 5;6 to 11;0), and on
the Ravens Progressive Coloured Matrices (RPCM) [57] 6;8
(range 5;6 to 8;0). This uneven profile is typical of Williams
syndrome. Eighteen normal participants, nine male and nine
female, formed the control group. They ranged in age from
19 to 29 years. All participants, both WS and controls, were
monolingual and of middle to lower middle socio-economic
class.
Materials. We examined three types of syntactic information
in this study: subcategory constraints on verbs, the constraints
imposed by auxiliaries on the form of the subsequent verb, and
phrase structure rules within local constituents. Examples of
each type of structure are given below. Each sentence containing
the target word was preceded by a short context sentence. The
word in upper case in each of the examples is the target word
for which the subject is monitoring and immediately follows
the correct or violated syntactic structure in italics.
Subcategory constraints: We selected verbs and constructed
sentences into which they could fit, such as (1) and (3) below,
and then violated the subcategory constraints on the main verb,
as in (2) and (4) below:
1. The burglar was terrified. He continued to stru991e with the
DOG but he couldn't break free.
2. *The burglar was terrified. He continued to struygle the
DOG but he couldn't break free.
3. The class was very unpopular. Maria always needed PARTNERS to show her the steps.
4. *The class was very unpopular. Maria always needed jbr
PARTNERS to show her the steps.
The sentences were all of the structure: (NP + verb + object NP)
plus a few additional words. The object NP (e.g., DOG or
PARTNERS in the above examples) was always the monitoring
target word. In the grammatical condition (examples (1) and
A. Karmiloff-Smith et al./Linguistic dissociations in Williams syndrome
(3) above), the object NP was syntactically appropriate in that
it was consistent with subcategory restrictions on the verb. In
the ungrammatical condition (such as (2) and (4) above), the
verb could not take a direct object and thus the presence of the
target noun constituted a grammatical violation. In developing
the stimuli we balanced the set between transitive and intransitive verbs. Therefore, in half the cases the appropriate condition was intransitive and contained a preposition (1), whereas
the inappropriate condition left out the preposition (3). For
the remaining cases, the appropriate condition was transitive
without a preposition (2) and the violation consisted of inserting
a preposition after the verb (4). Normal people are sensitive
to subcategory violations, shown by their faster monitoring
latencies to targets in grammatical compared with ungrammatical sentences [47, 66]. If WS participants are also sensitive
to the syntactic restrictions on the possible arguments which
different verbs can take, their monitoring response times (RTs)
should be slower in conditions (2) and (4) than in (1) and (3).
Auxiliary markers: Sentences in the auxiliary condition were
of the form: ( N P + a u x + v e r b + t a r g e t noun), followed by
additional material, as in (5) and (7) below. In such sentences
the auxiliary is syntactically appropriate for the inflected form
of the following verb. We compared monitoring latencies in
this condition with those to targets occurring in sentences such
as (6) and (8) below, where the auxiliary is changed so that
the combination of (auxiliary+verb) is ungrammatical. The
target word was the direct object of the following verb (e.g.,
SPEECHES and MILK in the examples below). If WS participants are sensitive to constraints on auxiliary choice, then
they should be faster to respond to the target words in sentences such as (5) and (7) than to ungrammatical sentences such
as (6) and (8).
5. It could have been very embarrassing. We didn't realize he
was expecting SPEECHES at the..
6. *It could have been very embarrassing. We didn't realize he
might expecting SPEECHES at the..
7, Fiona's doctor was very worried. He said she shouht have
MILK and protein more often.
8, *Fiona's doctor was very worried. He said she was ha~'e
MILK and protein more often
Phrase structure rules: Our third set of materials consisted of
sentences in which a target word (PILLS and TEST as in examples (9) and (11) below) followed a verb and completed a verb
phrase; the target noun is in the correct configuration with
respect to the verb. In contrast, in sentences such as (10) and
(12) below, the target word follows a sequence which violates
the legal configuration of grammatical categories. In all cases,
the target word occurred immediately after the violation. IfWS
participants are sensitive to the syntactic coherence of the target
word with respect to the prior syntactic context, then their
monitoring RTs should be faster for grammatical than for
ungrammatical sentences.
9. Susan seems much happier. I expect the special PILLS she
got from the doctor...
10. *Susan seems much happier. I expect special the PILLS she
got from the doctor...
1I. John's friends were delighted. When he took the new TEST
he passed first time.
12. *John's friends were delighted. When he took new the TEST
he passed first time.
We constructed a large number of sentences, each of which
contained a potential target word for word monitoring.
Ungrammatical versions of these sentences were made by introducing one of the three types of violation detailed above: subcategory, auxiliary and phrase structure. Sentences were
subjected to several pre-tests carried out with normal people.
First an acceptability pre-test was performed in which each
sentence's acceptability was rated on a scale of 1 (very bad) to
345
7 (very good). Only those grammatical sentences which received
a rating of 5 or higher by at least two-thirds of the normal
participants were included in the experiment. The criterion for
ungrammatical sentences was a rating of 3 or less by two-thirds
of the participants. Second, we carried out a cloze pre-test to
ensure that the target words were not predictable in the
sentences. In this test, the grammatical sentences, up to but
not including the target word, were written in a booklet and
participants were asked to provide a one-word continuation.
Any candidate sentences for which participants produced the
target word or a word related in meaning were discarded, thus
ensuring that targets were not predictable.
From the pre-tests we obtained 28 sentences in which the
ungrammatical condition violated subcategory constraints on
the verb, and 24 sentences in which the ungrammatical version
contained an auxiliary which was syntactically inappropriate
for the verb. In a further 22 sentences, the ungrammatical version contained a phrase structure violation based on re-ordering
of the words within the test noun phrase, such that the sequence
violated the phrase structure rules of English. In each set, the
word position of the target varied so that it would not be
predictable.
These 74 test sentences were interleaved with a large number
of fillers consisting of a variety of different syntactic structures.
These served to obscure the regularities of the test sentences. In
total, across the set of test and filler items, there were equal
numbers of grammatical and ungrammatical sentences, Two
versions of the materials were constructed, with items rotated
across the versions so that no item appeared more than once
per version. The materials were recorded and digitized onto
computer hard disk. Timing pulses were placed at the onset of
each target word, triggering a timing device which was stopped
by the action of the participant pressing the response button.
The target words were either monosyllabic or bisyllabic. This
presented no problem with respect to timing onset of latencies
because each word in the ungrammatical condition was
repeated by the same word in the grammatical condition across
the whole set of stimulus items.
Proce&o'e. Before each trial participants were presented with
the target word for which they were to listen in the sentence.
Controls read the word themselves, whereas for the WS participants the words were read aloud by the experimenter to
avoid any reading problems. The card with the target word
written on it stayed in front of the participants throughout the
trial. Participants then heard the sentence over headphones and
pressed a response key when they heard the target word.
Results and discussion
Controls. C o n t r o l s made an average of 1.3per cent
errors (range 0~4 per cent), consisting of time-outs (not
r e s p o n d i n g by 2 s) a n d outliers (RTs exceeding 1 s a n d
less than 100 ms). Two per cent of the control data were
removed from the analyses (0.6 per cent time-outs, 1.4 per
cent outliers); 3.9 per cent of the data exceeded 2 s t a n d a r d
deviations from the m e a n a n d were replaced by the cutoffvalues. The controls' m e a n m o n i t o r i n g R T was 336 ms
(range 268~456 ms). Before r u n n i n g analyses of variance
( A N O V A s ) , we normalized the data by dividing each R T
by the p a r t i c i p a n t ' s m e a n R T and m u l t i p l y i n g by 100,
so that each R T was representative in relation to the
p a r t i c i p a n t ' s m e a n RT. This was to make the control
data c o m p a r a b l e to the W S data for the purposes of
analysis, because the WS data had to be normalized in
346
A. Karmiloff-Smith et al./Linguistic dissociations in Williams syndrome
order to reduce variability introduced by the need for
some participants to perform the task on two different
days because they were unable to complete an entire
version of the experiment in a single session.
The controls showed an overall grammaticality effect,
with monitoring RTs in the grammatical sentences being
faster (307ms) than latencies in the ungrammatical
sentences
(356ms)
[Fl(1,16) = 238.61,
P<0.001;
F2(2,68)=60.52, P<0.001]. The mean latencies for the
three different types of syntactic manipulation are given
in Table 1, where it can be seen that all three types of
ungrammaticality slowed the monitoring RTs for the
controls, The syntax type by grammaticality effect was
not significant [F~(2,32) = 2.32, P = 0.14; F2(2,68) = 1.22,
P=0.29]. Separate analyses carried out on each of the
three syntax types individually revealed a significant
grammaticality effect for each of them: phrase structure
violation: F~(1,16)=76.48, P<0.001; F2(1,20)=22.27,
P<0.001; wrong auxiliary: F~(1,16)=23.5, P<0.001;
F:(1,22)= 15.87, P<0.001;
subcategory violation:
F~(1,16)= 56.98, P<0.001; F2(1,26)= 22.49, P<0.001.
Williams syndrome. The data from the individuals with
WS were cleaned in the same way as for the controls. A
total of 9.6 per cent of the data were removed from the
analyses (time-outs (2.03 per cent) or outliers (7.59per
cent). Of the remaining data, 5.6 per cent exceeded the 2
standard deviations from the mean cut-off values and
were replaced by those values. As with the controls, we
normalized the data before running ANOVAs in order
to reduce the variance attributable to participants being
tested on different days.
The monitoring RTs of the WS participants were very
similar to those of the normal controls (WS mean
RT = 386 ms, range 275-487 ms; controls mean RT = 336,
range 268-456 ms). A combined analysis on all eight WS
participants was carried out. This showed a significant
grammaticality
effect
[F~(1,6) = 44.64,
P < 0.001;
F2(2,68) = 7.5, P < 0.001] with monitoring latencies being
faster to targets in grammatical (371 ms) compared with
ungrammatical sentences (401 ms). This is again similar
Table I. Syntactic violations: mean monitoring RTs (ms)
Phrase structure:
Grammatical
Ungrammatical
Mean difference
Subcategory:
Grammatical
Ungrammatical
Mean difference
Wrong auxiliary:
Grammatical
Ungrammatical
Mean difference
Controls
WS
286
347
61 *
345
395
49*
316
361
45*
407
404
- 3
318
360
42*
362
405
43*
*Difference is significant at P<0.05 or above.
to the control data (grammatical 307, ungrammatical
356). However, the WS participants reacted differentially
to the three types of syntax. In keeping with the controls,
they showed a grammaticality effect for both phrase
structure and auxiliary violations, but unlike the controls
they were insensitive to subcategory violations. This can
clearly be seen in Table 1, where we also present the WS
monitoring latencies for each of the three types of syntax.
This differential effect showed up as a marginally significant interaction in the items analysis [F2(2,68)= 2.38,
P = 0.09], but it was not significant on the subjects analysis [F~(2,12) = 2.34, P = 0.17], presumably because of the
relatively small number of WS participants tested. To
explore this further, we carried out separate analyses on
each of the three types of syntax, finding a significant
grammaticality effect for phrase structure [F~(1,6) = 8.18,
P = 0.02, F2(1,20)= 12.82, P = 0.001] and wrong auxiliary
[F~(1,6)=13.33, P=0.01; F2(1,22)=5.57, P=0.02], but
no effect for the subcategory violations [F' and F2< 1].
In sum, WS and controls show very similar overall
mean RTs, with similar ranges. Both populations also
show similar overall monitoring latency differences to the
set of grammatical vs. the set of ungrammatical sentences.
In both groups, too, significant effects were found in
separate analyses of phrase structure and auxiliary structures. Where the two populations differed, however, was
with respect to the violation of subcategory structures.
One interpretation of these data is that the WS people
are insensitive to syntactic constraints which are lexically
specified, such as subcategory information. However,
another interpretation may turn out to be more plausible:
namely, that WS are sensitive to subcategory constraints,
but they are slow to integrate this type of information into
the developing sentential representation. The evidence
for this stems from the RTs of WS participants in the
grammatical subcategory condition. Their RTs averaged
407 ms in this condition, compared with 345 ms for grammatical phrase structures and 362 ms for grammatical
auxiliary markers. This difference shows up in a marginally significant main effect of type of syntax
[F~(2,12) = 5.11, P = 0.06; F2(2,68) = 2.65, P = 0.07]. The
normal controls do not show this pattern; their RTs in
the three grammatical conditions for the three types of
syntax only vary by 30 ms (from 286 to 316 ms; see Table
1). We will consider this further in the Discussion section.
Experiment 2: Sentence-picture matching task
In the sentence-picture matching task, the subject
hears a spoken sentence and chooses, from an array of
pictures, the one that corresponds to the sentence. We
chose this task for our explicit tasks for a number of
reasons. First, in some circumstances it can be more
informative than tasks such as grammaticalityjudgement
in which the subject is just making a yes/no response. In
the sentence-picture matching task, we can look at the
type of errors that subjects make, i.e., whether they
A. Karmiloff-Smith et al./Linguistic dissociations in Williams syndrome
choose, for example, a lexical distractor or a syntactic
distractor. Our second reason for using this task was that,
unlike the word monitoring task, it allowed us to look at
a wide variety of different types of syntactic structure.
The on-line monitoring task necessarily restricts the types
of structure we can examine because of the need to have
the target word occur immediately after a linguistic violation. Because our aim was to obtain data on a wide
variety of sentence structures, in this second experiment
we did not limit ourselves only to those types of violation
we had used in the monitoring study.
Method
Participants. We tested the same eight WS and 18 normal
control participants in this study as had been tested in the word
monitoring task in Experiment 1.
Materials. The materials were from the Birkbeck Reversible
Sentences Test [12]. On each trial, participants are presented
with an array of three pictures: one corresponding to the
sentence, one a reverse role distractor and the third a lexical
distractor. For example, for the sentence: "The clown photographs the policeman", the three pictures are: (1) correct: a
clown photographing a policeman; (2) reverse role distractor: a
policeman photographing a clown; (3) lexical distractor: a
clown lifting the helmet off a policeman. The test sentences were
presented orally and consisted of 70 sentences, 10 for each of
the following seven types of construction: active sentences with
agentive and non-agentive verbs, passive sentences with agentive and non-agentive verbs, sentences containing adjectives.
deverbal adjectives and locatives.
Results and discussion
The performance of the controls was virtually errorfree on this task. In contrast, the WS participants performed very poorly, making an average of 24 per cent
errors (range 14-37per cent). Most of their errors consisted of the syntactic error of choosing the reverse role
distractor, which they did 81 per cent of the time. Only
19 per cent of their total errors involved choosing the
lexical distractor.
Because the normal controls were at ceiling on all the
structures, we also compared the WS results with those
from a group of elderly participants between the ages of
60 and 75 years, all of whom had left school by the age
of 14years [68]. They made many more errors than the
young controls on this task (mean errors for elderly subjects 17 per cent, range 4-30 per cent). Our aim was to see
whether the pattern of errors on the different sentence
structures made by the WS individuals was similar to that
of normal but elderly controls. We found that the order
of difficulty of the various linguistic structures was identical for the elderly controls and the WS group: most
errors were made on sentences (both actives and passives)
with non-agentive verbs (WS 33 per cent errors, controls
31 per cent errors), followed by sentences with deverbal
adjectives (WS 26 per cent, controls 23 per cent), those
with adjectives (WS 24per cent, controls 21 per cent),
347
passive sentences with agentive verbs (WS 17per cent,
controls 10 per cent), active sentences with agentive verbs
(WS 17 per cent, controls 5 per cent), both groups finding
the sentences with locatives the easiest (WS 14 per cent,
controls 3 per cent).
This second experiment revealed that WS participants
have far more problems with syntax when tested by
means of an explicit task, although the types of structure
on which they have most difficulty are those which the
elderly controls who left school at 14years of age also
find difficult. By contrast, it is noteworthy that, unlike
the WS participants, when tested with the on-line technique the elderly group were not selectively insensitive to
subcategory violations but showed the same pattern as
the normal controls [68]. The low performance levels
of the WS individuals on the sentence-picture matching
study are similar to the levels of syntactic performance
that they achieve on the T R O G (see Participants section),
which covers a range of other structures that they find
difficult, such as left-branching relative clauses [39] for
an item-by-item analysis of WS T R O G results).
General discussion
Despite the differences in performance between Experiments 1 and 2, it is clear from both, and in comparison
with the results of both the younger and elderly control
groups, that WS syntax is not intact. The sentence-picture matching task revealed an across-the-board impairment. WS participants' ability to select the correct picture
from an array in response to a spoken sentence was
considerably worse than normal on the wide range of
syntactic structures we tested. However, they did not
show a selective deficit for any specific syntactic constructions; their performance on the various sentence
types was similar to the normal pattern, although overall
accuracy was much worse. The word monitoring study,
on the other hand, showed that some syntactic operations
remained intact whereas others were selectively impaired.
The poorer performance on the off-line task is not so
surprising given that most explicit tasks do not tap into
language processing directly, but also involve a number
of additional cognitive processes shown to influence the
performance of both children [31, 36] and adults [63, 64,
66]. In the sentence-picture matching task, for example,
the participant has to listen to and decode the sentence,
maintain it in their memory (when it is spoken), process
the pictures and compare the sentence to each one of the
pictures, eventually selecting the one that best matches.
Because WS individuals have clear cognitive impairments, tasks which rely heavily upon general cognitive
abilities may overestimate the extent of the linguistic
impairment. This is not to say that people with WS find all
picture-matching tasks uniformly difficult. As mentioned
earlier, WS individuals perform rather well on picturematching tasks involving single words, such as the BPVS
[18]. However, given that this task involves single words
348
A. Karmiloff-Smith et al./Linguistic dissociations in Williams syndrome
and simpler pictures, the cognitive demands it places on
the individual are undoubtedly less than those involved
when whole sentences need to be matched to a picture.
The contrast between the explicit and implicit data
has important implications for studying the language
comprehension abilities of people with WS as well as
other impaired populations. Explicit tasks can, in general,
only provide an overall measure of performance.
Although we can determine whether participants have
problems with particular aspects of language, we cannot
determine which aspects of the comprehension process
are actually impaired. If, for example, we had only carried
out an explicit task on the materials we used in the monitoring study, we might have obtained the same outcome - - namely, that WS participants had difficulty
comprehending subcategory information but not phrase
structure and auxiliary markers. However, we would not
have been able to propose an account of their problems
in terms of deficits in the basic processes involved in
language comprehension - - processes of activation and
integration [46]. The advantage of on-line tasks is that
they can provide a more detailed assessment of the nature
of the impairment, as has been shown with brain-damaged adults [66].
The on-line study revealed a more varied pattern than
the off-line task. It showed that whereas WS participants
had no difficulty assembling the relevant elements together to form coherent noun phrases or in appreciating
the relationship between an auxiliary and a main verb,
they had clear problems with subcategory structures. For
subcategory violations, monitoring latencies did not significantly increase as they did for violations of both
phrase structure and auxiliary markers. Although it
might be tempting to interpret this pattern as indicating
that WS cannot process subcategory information, we
provisionally favour a more conservative interpretation - - namely, that WS are able to access subcategory
information associated with a verb, but they are slow to
integrate this information with the upcoming input. This
account is more consistent both with the data and with
some current theoretical approaches to syntax. Those
aspects of the data which support this view are the fact
that, coupled with no significant increase in latencies in
the subcategory ungrammatical condition, RTs in the
grammatical condition are in and of themselves also slow
compared with the other categories.
The above interpretation is also more consistent with
a lexicalist approach to syntactic processing [14, 43, 65,
73, 75], in which much of what was traditionally thought
to be due to the operations of a syntactic parser is seen
to be part of lexical representations. On this kind of
account, subcategory information is lexically represented
and activated when a word is heard, providing potential
argument structures into which the upcoming words are
integrated. Similarly, there are lexically specified constraints on the permissible configurations for nouns,
adjectives, determiners, etc., which determine agreement
and word order. Therefore, on a lexicalist account, all
three types of syntactic information that we manipulated
in the word monitoring experiment are lexically
represented. This suggests that they should stand or fall
together in cases of language impairment. If this obtains,
then it is more likely that the differences in the word
monitoring study for WS are not due to differential representational impairments but to differences in the ease
with which various types of syntactic or lexical information can be integrated on-line. This is a picture which
we have observed many times in studies of normal adult
brain-damaged patients whose language is impaired
either as a result of focal lesions accompanying stroke or
more generalized damage from progressive degenerative
disease [68].
However, although our suggestion that the three categories should stand or fall together holds for adults
who have acquired language normally and then suffered
neurological damage, it may not hold for developmentally impaired children whose brains develop
differently from the outset [34, 35]. Several studies have
now suggested that WS language is acquired via a somewhat different route to normal acquisition and involves
different brain processing [39, 40, 50, 54, 55, 60, 67].
Furthermore, in normal development subcategorization
structures are acquired later than word order and auxiliary categories [24]. They also seem to be more difficult
for second language learners [29]. Of course, these two
data sources come from very different methods and age
groups, the first being the infant preferential looking task
with 12-27month olds, the second written grammaticality judgements with non-native but fluent adult
speakers of English. Nonetheless, the congruence of order
of difficulty of these two sources with our WS on-line
data suggests that subcategorization, whether it turns out
to be lexically or syntactically specified, is more difficult
to acquire. Therefore, unlike brain-damaged adults who
had acquired language normally and show integrational
problems rather than representational impairment, it
remains an open question whether our WS subjects have
only integrational or also representational impairments
for certain linguistic categories.
We plan further experiments to tease apart the contribution of representational vs. integrational impairments in WS syntax. Indeed, the use of on-line techniques
for addressing such issues is crucial in revealing the subtleties of abilities and impairments that off-line tasks
often fail to capture. The present study has shown that the
on-line techniques used with adult neuropsychological
patients [66] can very profitably be extended to individuals whose impairments are of genetic origin. Whatever the full extent of the WS impairments (see also [39,
73]), the two experiments reported here provide further
evidence that, contrary to a popular view in the literature
(e.g., [56]), people with Williams syndrome do not have
entirely normal syntax.
Acknowledgements--We should like to express our appreciation
A. Karmiloff-Smithet al./Linguistic dissociationsin Williams syndrome
to the WilliamsSyndrome Foundation, U.K., for their generous
help in putting us in touch with families whom we warmly
thank for their participation in this research. The WS participants were tested at the MRC Cognitive Development Unit
and the normal controls at the Centre for Speech and Language,
Birkbeck College, London. This research was partly supported
by an MRC programme grant to Lorraine K. Tyler and W.
D. Marslen-Wilson and by an MRC Senior Research Leave
Fellowship to L.K.T.
14.
15.
16.
References
1. Arnold, R., Yule, W. and Martin, N., The psychological characteristics of infantile hypercalcaemia: a
preliminary investigation. Developmental Medicine
and Child Neurology, 1985, 27, 49-59.
2. August, G. J. and Realmuto, G. M., Williams syndrome: serotonin's association with developmental
disabilities. Journal of A utism and Developmental Disorders, 1989, 19, 137-141.
3. Bellugi, U., Marks, S., Bihrle, A.M. and Sabo, H.,
Dissociation between language and cognitive functions in Williams Syndrome. In Language Development in Exceptional Circumstances, eds D. Bishop
and K. Mogford. Churchill Livingstone, London,
1988.
4. Bellugi, U., Bihrle, A., Neville, H., Jernigan, T. and
Doherty, S., Language, cognition and brain organization in a neurodevelopmental disorder. In Developmental Behavioral Neuroscience, eds M. Gunnar
and C. Nelson. Lawrence Erlbaum Associates,
Hillsdale, N J, 1992.
5. Bellugi, U., Wang, P., and Jernigan, T. L., Williams
syndrome: an unusual neuropsychological profile. In
Ao'pical Cognitive Deficits in Developmental Disorders: Implications Jor Brain Function, eds S. Broman and J. Graffman. Lawrence Erlbaum
Associates, Hillsdale, N J, 1994, pp. 23-56.
6. Beuren, A. J. Supravalvular aortic stenosis: a complex syndrome with and without mental retardation.
Birth Defects, 1972, 8, 45-46.
7. Beuren, A. J., Schulze, C., Eberle, P., Harmjanz, D.
and Apitz, J., The syndrome of supravalvular aortic
stenosis, peripheral pulmonary stenosis, mental
retardation and similar facial appearance. American
Journal of Cardiology, 1964, 13, 471-482.
8. Bialystok, E. Factors in the growth of linguistic
awareness. Child Development, 1986, 86, 498-510.
9. Bialystok, E. and Ryan, E. B., Toward a definition
of metalinguistic skill. MerillPalmer QuarterO,,
1985, 31,229-251.
10. Bihrle, A. M., Bellugi, U., Delis, D. and Marks, S.,
Seeing either the forest or the trees: dissociation in
visuospatial processing. Brain and Cognition, 1989,
11, 37-49.
11. Bishop, D. V. M., Test for Reception q/ Grammar.
Medical Research Council, U.K., 1983.
12. Black, M., Nickels, L. and Byng, S., Patterns of sentence processing deficit: processing simple sentences
can be a complex matter. Journal of Neurolinguistics,
1991,6,79 101.
13. Capirci, O., Sabbadini, L. and Volterra, V., Lan-
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
349
guage development in Williams syndrome: a case
study. Cognitive Neuropsychology, 1996, 7, 1017-1039.
Carlson, G. and Tanenhaus, M., Thematic roles and
Language comprehension. In Thematic' Relations, ed.
W. Wilkins. Academic Press, London and San Diego,
1987.
Creery, R. D. G. Idiopathic hypercalcaemia in
infants. Lancet, 1953, 2, 17-19.
Crisco, J. J., Dobbs, J. M. and Mulhern, R. K.,
Cognitive processing of children with Williams syndrome. Developmental Medicine and Child Neurology, 1988, 30, 650-656.
Culler, F., Jones, K. and Deftos, L., Impaired calcitonin secretion in patients with Williams syndrome.
Journal qf Paediatrics, 1985, 107, 720-723,
Dunn, L. M., Dunn, L. M. and Whetton, C., British
Picture VocabulaO' Scale. NFER-Nelson, 1982.
[!wart, A. K., Morris, C. A., Atkinson. D., Jin, W.,
Sternes, K., Spallone, P., Dean Stock, A., Leppert,
M. and Keating, M. T., Hemizygosity at the elastin
locus in a developmental disorder, Williams
syndrome. Nature Genetics, 1993, 5, 11-16.
Ewart, A., Morris, C., Ensing, G., Loker, J., Moore,
C., Leppert, M. and Keating, M., A human vascular
disorder, supravalvular aortic stenosis, maps to chromosome 7. Proceedings o/ the National Academy o['
Science, 1993, 90, 3226-3230.
Fanconi, G., Gritted, P., Schlesinger, B., Butler, N.
and Black, J. A., Chronische Hypercalcaemie, kombiniert mit Osteosklerose, Hyperazotaemie, Minderwuchs
und
kongenitalen
Missbildungen.
Heh'etica Paediatrica Acta, 1952, 7, 314-334.
Frangiskakis, J. M., Ewart, A. K., Morris, C. A.,
Mervis. C. B., Bertrand, J., Robinson, B. F., Klein,
B. P., Ensing, G. J., Everett, L. A., Green, E. D.,
Proschel, C., Gutowski. N. J., Noble, M., Atkinson,
D. L., Odelberg, S. J. and Keating, M. T., LIMKinase 1 hemizygosity implicated in impaired visuospatial constructive cognition. Cell, 1996, 86, 59-69.
Galaburda, A. M., Wang, P. P., Bellugi, U. and
Rossen, M., Cytoarchitectonic anomalies in a genetically based disorder: Williams syndrome. Cognitive
Neuroscience
and
Neuropsycholoyy,
NeuroReport, 1994, 5, 753 757.
Gleitman, L. R. The structural sources of verb meanings. Language Acquisition, 1990, I, 3 55.
Grant, J., Karmilofl-Smith, A., Gathercole, S., Paterson, S., Howlin, P., Davies, M. and Udwin, O.,
Verbal short-term memory and its relation to language acquisition in Williams syndrome. Cognitive
NeuropsychiatJ3,, 1997, 2, 81-89.
Greenberg, F. Introduction to special issue on Williams syndrome. American Journal o1" Medical Genetics Supplement, 1990, 6, 85-88.
Hakes, D. T., The Development of Metalinguistic
Abilities in Children. Springer-Verlag, Berlin, 1980.
Jernigan, T. L., Bellugi, U., Sowell, E., Doherty, S.
and Hesselink, J. R., Cerebral morphological distinctions between Williams and Down's syndromes.
Archives Of Neurology, 1993, 50, 186-191.
Johnson, J. S. and Newport, E. Critical period effects
350
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
A. Karmiloff-Smith et al./Linguistic dissociations in Williams syndrome
in second language learning: the influence of maturational state on the acquisition of English as a
second language. Cognitiz'e Psychology, 1989, 21, 6099.
Jones, K. L. and Smith, D. W., The Williams elfin
faces syndrome: a new perspective. Journal of Paediatrics, 1975, 86, 718-723.
Karmiloff-Smith, A. From metaprocesses to conscious access: evidence from children's metalinguistic
and repair data. Cognition, 1986, 23, 95-147.
Karmiloff-Smith, A., Abnormal phenotypes and the
challenges they pose to connectionist models of
development. Technical Report Series on Parallel
Distributed Processing and Cognitive Neuroscience,
No. PDP/CNS.92.7, Carnegie Mellon University,
Pittsburgh, 1992a.
Karmiloff-Smith, A., Beyond ModulariO,: A Developmental Perspective on Cognitive Science. MIT Press/Bradford Books, Cambridge, MA, 1992b.
Karmiloff-Smith, A., Self-organisation and cognitive
change. In Brain Development and Cognition: A
Reader, ed. M. H. Johnson. Blackwells, Oxford,
1992c.
Karmiloff-Smith, A., Is adult neuropsychology an
adequate model for genetic disorders? Developmental
Neuropsychology in press.
Karmiloff-Smith, A., Grant, J., Jones, M.-C., Sims,
K. and Cuckle, P., Rethinking metalinguistic awareness: representing and accessing what counts as a
word. Cognition, 1996, 58, 197 219.
Karmiloff-Smith, A., Johnson, H., Grant, J., Jones,
M. C., Karmiloff, Y. N., Bartrip, J. and Cuckle, C.,
From sentential to discourse functions: detection and
explanation of speech repairs in children and adults.
Discourse Processes, 1993, 16, 565-589.
Karmiloff-Smith, A., Klima, E., Bellugi, U., Grant,
J. and Baron-Cohen, S., Is there a social module?
Language, face processing and theory-of-mind in
subjects with Williams syndrome. Journal of Cognitive Neuroscience, 1995, 7, 196-208.
Karmiloff-Smith, A., Grant, J., Berthoud, I., Davies,
M., Howlin, P. and Udwin, O., Language and Williams syndrome: how intact is "intact"? Child Development, 1997, 68, 246-262.
Kataria, S., Goldstein, D. J. and Kushnick, T.,
Developmental delays in Williams ("elfin faces") syndrome. Applied Research in Mental Retardation,
1984, 5, 419-423.
Lightwood, R. Idiopathic hypercalcaemia with failure to thrive. Proceedings of the Royal Society of
Medicine, 1952, 45, 401-410.
Lowery, M. C., Morris, C. A., Ewart, A., Brothman,
L. J., Zhu, X. L., Leonard, C. O., Carey, J. C.,
Keating, M. and Brothman, A. R., Strong correlation of elastin deletions, detected by FISH, with
Williams syndrome: evaluation of 235 patients.
American Journal o[" Human Genetics, 1995, 57, 4953.
Macdonald, M., Perlmutter, N. and Seidenberg, M.,
Lexical nature of syntactic ambiguity resolution.
Psychological Review, 1994, 101, 676-703.
McKusick, V., Mendelian Inheritance in Man: Catalogs ot" Autosomal Dominant, Autosomal Recessive
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
and X-linked Phenotypes. Johns Hopkins Press, Baltimore, 1988.
Marriage, J., Central hyperacusis in Williams
syndrome. Paper presented at the conference on
Building Bridges Across Disciplines: Cognition to
Gene. Williams Syndrome Association Sixth International Professional Conference, San Diego, 1994.
Marslen-Wilson, W. D., Functional parallelism in
spoken word recognition. In Spoken Word Recogniton, eds U. Frauenfelder and L. K. Tyler. MIT
Press, Cambridge, MA, 1987.
Marslen-Wilson, W. D., Brown, C. and Tyler, L. K.,
Lexical representations in language comprehension.
Language and Cognitive Processes, 1988, 3, 1-17.
Marslen-Wilson, W. D. and Tyler, L. K., Processing
structure of sentence perception. Nature, 1975, 257,
784-786.
Marslen-Wilson, W. D. and Tyler, L. K., The temporal structure of spoken language processing. Cognition, 1980, 8, 1-71.
Mervis, C. B. and Bertrand, J., Williams syndrome:
findings from an integrated program of research. In
Developmental Relations between Cognition and
Research on Communication and Language Disorders:
Contributions to Theories of Language Development,
ed. H. Tager-Flusberg. Brookes, New York, in press.
Mervis, C. G., Morris, C. A., Bertrand, J. and Robinson, B. F., Williams syndrome: findings from an integrated program of research. In Neurodevelopmental
Disorders: Contributions to a New Framework from
the Cognitiee Neurosciences, ed. H. Tager-Flusberg.
MIT Press, Cambridge, MA, in press.
Milberg, W. and Blumstein, S., Lexical decision and
aphasia: evidence for semantic processing. Brain and
Language, 1981, 14, 371-385.
Morris, C., Demsey, S., Leonard, C., Dilts, C. and
Blackburn, B., Natural history of Williams syndrome: physical characteristics. Journal of Paediatrics, 1988, 113, 318-326.
Neville, H. J., Holcomb, P. J. and Mills, D. M.,
Auditory, sensory and language processing in Williams syndrome: an ERP study. Journal of Clinical
and Experimental Neuropsvchology, 1989, 11, 52.
Neville, H. J., Mills, D. L. and Bellugi, U., Effects
of altered auditory sensitivity and age of language
acquisition on the development of language-relevant
neural systems: preliminary studies of Williams syndrome. In Cognitive Deficits in Developmental Disorders: Implications jor Brain Function, eds S.
Broman and J. Grafman. Erlbaum, Hillsdale, N J,
1993, pp. 67-83.
Pinker, S., The Language Instinct: The New Science of
Language and Mind. Alan Lane, The Penguin Press,
Harmondsworth, Middlesex, 1994.
Raven, J. C., Ravens Progressive Matrices and
Vocabulary Scales. Lewis, London, 1986.
Rossen, M., Bihrle, A., Klima, E. S., Bellugi, U.
and Jones, W., Interaction between language and
cognition: evidence from Williams syndrome. In
Language Learning and Behaviour, eds J. H. Beitchman, N. Cohen, M. Konstantareas and R.
Tannock. Cambridge University Press, New York,
in press.
A. Karmiloff-Smith et al./Linguistic dissociations in Williams syndrome
59. Smith, C. L. and Tager-Flusberg, H., Metalinguistic
awareness and language development. Journal q/'
Experimental Child Psychology, 1982, 34, 449~,68.
60. Stevens, T. and Karmiloff-Smith, A., Word learning
in a special population: do individuals with Williams
syndrome obey lexical constraints?, Journal q[+Child
Lanquage, in press.
61. Tassabehji, M. K., Metcalfe, K., Fergusson, W. D.,
Carette, M. J. A., Dore, J. F., Donnai, D., Read, A.
P., Proschel, C., Gutowski, N. J., Mao, X. and Sheer,
D., LIM-Kinase 1 detected in Williams syndrome.
Nature Genetics, 1996, 13, 272-273.
62. Tunmer, W. E. and Grieve, R., The development
of the child's awareness of grammatical structure.
Educational Research and Perspectives. 1980, 7, 47
56.
63. Tyler, L. K. Real-time comprehension processes in
agrammatism: a case study. Brain and Language,
1985, 26, 259 275.
64. Tyler, L. K. Spoken language comprehension in a
fluent aphasic patient. Cognitive Neuropsycholo.qy,
1988, 5, 375~400.
65. Tyler, L. K., The role of lexical representations in
language comprehension. In Lexical Representation
and Process, ed. W. D. Marslen-Wilson. MIT Press,
1989.
66. Tyler, L. K., Spoken Language Comprehension: An
Experimental Approach 1o the Study qic Normal and
Disordered Processin#. M IT Press, Cambridge, MA,
1992.
67. Tyler, L. K., Karmiloff-Smith, A., Voice, K., Stevens,
T., Grant, J., Davies, M., Howlin, P. and Udwin, O.,
Do people with WS have bizarre semantics? A
primed monitoring study. Cortex, 1997, 33, 515 527.
68. Tyler, L. K., Moss, H. E., Patterson, K. and Hodges,
J., The gradual deterioration of syntax and semantics
in a patient with progressive aphasia. Brain and Lan9ua.qe, 1997, 86, 426-477.
351
69. Udwin, O. and Yule, W., Expressive language of
children with Williams syndrome. American Journal
of Medical Genetics Supplement, 1990, 6, 108-114.
70. Udwin, O. and Yule, W., A cognitive and behavioural phenotype in Williams syndrome. Journal of
Clinical and Experimental Neuropsychology, 1991,
13, 232 244.
71. Udwin, O., Yule, W. and Martin, N., Cognitive abilities and behavioural characteristics of children with
ideopathic infantile hypercalcaemia. Journal of Child
Psycholoqy and Psychiat W, 1987, 28, 297-309.
72. Van Kleeck, A. The emergence of linguistic awareness: a cognitive framework. Merrill Palmer Quarterly, 1982, 28, 237-265.
73. Volterra, V., Capirci, O., Pezzini, G., Sabbadini, L.
and Vicari, S., Linguistic abilities in Italian children
with Williams syndrome. Cortex, 1996, 32, 663 677.
74. Wang, P. P., Hesselink, J. R., Jernigan, T. L.,
Doherty, S. and Bellugi, U., The specific neurobehavioral profile of WS is associated with neocerebellar hemispheric preservation. Neurology,
1992.42, 421-521.
75. Wasow, T., Postscript. In Lectures on Contemporar),
Syntactic Theories, ed. P. Sells. CSLI, Stanford, 1985.
76. Wechsler, D., Wechsler Intelli,qence Scale ./'or Children - - Revised UK. The Psychological Corporation,
London. 1974.
77. Wechsler, D., Wechsler Adult Intelligence Scale - Retqsed UK. The Psychological Corporation,
London, 1981.
78. Wechsler, D., Wechsler hltelligence Scale ./br Children --- 3rd Edn, UK. The Psychological Corporation, London, 1992.
79. Weiskrantz. L., Blindsight: A Case Study and Implications. Clarendon Press, Oxford, 1986.
80. Williams, J. C. P., Barratt-Boyes, B. G. and Lowe,
J. B., Supravalvular aortic stenosis. Circulation,
1961, 24, 1311 --1318.