Autism, primary

pragmatic difficulties,
and specific language
impairment: can we
distinguish them using
psycholinguistic
markers?
Nicola Botting* BSc MA PhD;
Gina Conti-Ramsden BA MSc PhD, Human Communication
and Deafness, School of Education, University of Manchester,
Oxford Road, Manchester M13 9PL, UK.
E-mail: nicola.botting@man.ac.uk
*Correspondence to first author at above address.
Three groups of children with communication disorders were
examined using a series of psycholinguistic markers to explore
whether the tasks could identify children with impairments other
than specific language impairment (SLI), and to examine
whether the different groups within this clinical population could
be distinguished reliably from one another. The groups comprised
children with autistic spectrum disorders (ASD; n=13, all males;
mean age 10 years 10 months, range 10 years 2 months to 12
years 6 months); children with primary pragmatic language
impairment (PLI) but who did not have definite ASD diagnoses
(n=25, 22 males, three females; mean age 11 years 3 months,
range 10 years 2 months to 12 years 5 months); and children with
specific language impairment (SLI) without marked pragmatic
language difficulties (n=29, 25 males, 4 females; mean age 10
years 10 months, range 10 years 2 months to 11 years 9 months).
Clinical markers examined were: the Childrens Non-Word
Repetition (CNRep), the Past Tense Task (PTT), and the Clinical
Evaluation of Language Fundamentals, Recalling Sentences.
First, it was found that the a priori groupings were not sufficiently
defined and that four groups were actually present. The PLI
group was in fact two separate samples: those with PLI pure and
those with some autistic-like behaviours (referred to here as PLI
plus, following Bishop 1998). Second, group comparisons
indicated that CNRep was significantly lower for children with
SLI than all other groups (although this measure was not such a
good discriminator using a specificity analysis). Third, the
markers were able to discriminate between all types of
communication impairment in normal control participants
(n=100; 51 females, 49 males; mean age 11 years, range 10 years
5 months to 11 years 6 months) with sensitivity levels of at least
75% and specificity of 80%. Recalling Sentences was the most
efficient marker for all groups. Finally, analysis showed that
children with PLI plus could be accurately distinguished from all
others, scoring most favourably overall on communication
markers and on performance IQ scores.

The study of markers in communication difficulties has become of increasing interest in the literature. Potentially, the discovery of an efficient marker for a clinical difficulty could have a
great impact on the screening, remediation, and genetic study
of such disorders. It could also inform research about underlying mechanisms and differential diagnosis.
In general, there is a wide population of children with communication impairments. This includes children with general
learning difficulties through to those with highly specific communication needs such as articulation problems. This study,
however, concerns children for whom general communication
problems are central to their difficulties: namely, those with
specific language impairment (SLI), autistic spectrum disorders (ASD), and primary pragmatic language impairment(PLI).
In brief, SLI is usually diagnosed on the basis of abnormal
language ability in the context of normal cognitive skills and in
the absence of any major neurological or physical cause.
However, the definitions of SLI are also under scrutiny. Diagnoses of autism on the other hand are made on the basis of a
triad of impairments: social impairment, communicative impairment, and an impairment in creativity, flexibility of thinking,
and generalization (American Psychiatric Association 1994).
However, there is a debate about how strict criteria need to be
in order to describe usefully an individual as having ASD.
Furthermore, similar disorders such as Asperger syndrome
and children with pervasive developmental disorder not otherwise specified (PDDNOS) who do not quite meet diagnostic
criteria for autism, have made clinicians and researchers aware
that the diagnostic boundaries are not always clear. A subgroup
of children with primary pragmatic language impairment(PLI;
once referred to as having semantic pragmatic disorder) but
who do not have autism have been described a number of
times in the literature (Bishop 1998, Botting and ContiRamsden 1999, Conti-Ramsden and Botting 1999, Bishop and
Norbury 2002). These children tend to be able to produce
complex sentences (although usually not without errors) and
are often verbose. However, they have a poor understanding of
functional communication including turn-taking, poor understanding of roles, limited conversational topics, a lack of sensitivity regarding social cues, and a tendency to give too much or
too little information. It is uncertain whether these children are
also the same as those with PDDNOS previously mentioned
(see Cox et al. 1999, who found similarities between PDDNOS
and language impairment, and also Bishop and Norbury 2002,
who found differences).
It is, therefore, relatively unclear how best to identify accurately different groups of children with these similar types of
communication difficulty. Individuals with SLI, ASD, and PLI
often show similar and overlapping sets of symptoms.
Children who do not match clear-cut examples of individuals
with a disorder are often diagnosed differently by professionals
with different types of training (e.g. psychiatrist versus speechlanguage therapist). At the same time as there being overlap
between groups, there is general clinical agreement that at
least some of these groups have distinct difficulties, requiring
different interventions and having separate prognoses.
The unclear nature of clinical description is exacerbated by
the lack of accurate diagnostic markers. Often children in
these groups are diagnosed by exclusion that is by ruling out
other possibilities rather than by positively identifying particular characteristics. Alongside this search for clarity in the clinical description of SLI is another branch of investigation into

See page 524 for list of abbreviations.

Developmental Medicine & Child Neurology 2003, 45: 515524 515

the heritability of pervasive developmental disorders. In family studies, differing communication impairments are often
found occurring next to one another in genetic relations
(Landa et al. 1990).
Ideally markers should be subclinical in nature in order to
include family members who may genetically contribute to various disorders. That is, they should identify individuals who
have overcome a genetic predisposition to an impairment (and
thus show no overt difficulties) so that the genetic analysis is not
biased. Three potential psycholinguistic tasks currently most
salient in the literature are on language impairment: non-word
repetition, past tense knowledge, and sentence repetition.
Non-word repetition tasks involve repeating nonsense
words with between two and five syllables immediately after
hearing them. A landmark study in the investigation into markers and communication difficulties was that by Bishop et al.
(1996) who found that in families of children with SLI, a nonword repetition task showed high heritability or genetic pattern. Measures of non-word repetition have now been widely
used with a number of different groups including those with
autism (Kjelgaard and Tager-Flusberg 2000), and sensorineural hearing loss (Briscoe et al. 2001), as well as normally developing children (Adams and Gathercole 2000). Non-word
repetition has also been shown to be a good correlate of longterm outcome and reading ability in SLI groups (Stothard et al.
1998, Bishop 2001) and to relate strongly to progress in language learning (Botting and Conti-Ramsden 2002).
Grammar-related markers, such as the ability to formulate
sentences in past tense, have also been examined for some
time, for example by Rice and Wexler (1996) and ContiRamsden and Windfuhr (2002) and have been shown to be
important in describing difficulties. The ability to understand and produce verbs correctly is of particular interest
because it represents a notable difficulty in children with SLI
(Leonard 1998). However, it is not a clear candidate for those
with resolved difficulties and the deficit may result from
other underlying processes.
In contrast, Recalling Sentences (from the Clinical Evaluation
of Language Fundamentals Revised [CELF]; Semel et al.
1987) and other sentence repetition tasks are fairly new to
this type of investigation. Interestingly, a number of studies
and clinical services are currently using a CELF battery
(including the sentence repetition task, Recalling Sentences)
and, therefore, make this task all the more interesting. ContiRamsden, et al. (2001) have recently shown that children
with SLI could be accurately discriminated from normally
developing peers using this task, in which children must
repeat an increasingly complex series of sentences immediately after hearing them. In that study we were able to predict
group membership with 98% accuracy using a 16th centile
(1SD) cut-off point. Furthermore, the majority of children
with a history of SLI but whose language skill was now in the
normal or borderlinenormal range were also identified by
the measure (Conti-Ramsden et al. 2001).
However, several questions remain unanswered. Is this finding specific to SLI, or can this test (or the other markers examined previously) identify other clinical groups within the area
of communication difficulties? Can performance on any of the
marker tests accurately distinguish between groups of children
with different profiles of impairment? This is especially relevant
to those profiles that are notoriously difficult to diagnose, in
particular the borderlands between autism, PLI, and SLI.
516

Developmental Medicine & Child Neurology 2003, 45: 515524

Various methods of assessment of more subtle communication difficulties are currently in use, both clinically and academically. Because of the difficulty in directly observing
pathological behaviours in short time frames, checklists have
been designed with some success. For example, the Checklist
for Autism in Toddlers (Baird et al. 2000), questionnaires by
Klee and colleagues used for identifying language disorders
(Klee et al. 2000), and the Childrens Communication Checklist (CCC) by Bishop (1998) have all achieved useful results.
The Childhood Autism Rating Scale (CARS: Schopler et al.
1980) is also widely used with some successful psychometric
properties being reported despite some anecdotal criticism.
Another form of assessment, that of the parent interview (in
particular the Autism Diagnosis Interview; Lord et al. 1994),
has reported good results regarding the screening of autism
(Lord et al. 1997). Cluster analysis techniques using clinical
observations have also been used in an attempt to identify reliable differences. Prior and coworkers (1998) showed that the
Autistic Spectrum Disorders Checklist identified three clusters
of children who roughly corresponded to autism, Asperger
syndrome, and PDDNOS.
Psycholinguistic markers are the focus of the present study
for two reasons: first, because language is central to all the separate diagnoses within communication disorders and, second,
because quite promising results have been found in the area of
SLI but have not been compared across impairments. Studies
which have examined other important skills (such as theory of
mind, ToM) in different clinical populations have of course
revealed some interesting differences and similarities between
groups with communication disorders. For example, Shields et
al. (1996) found that a group of children with PLI (referred to
as semantic-pragmatic disorder in their article) were not significantly different from children with autism on ToM tasks.
Conversely, Ziatas et al. (1998) found that children with SLI
who were matched to a group with Asperger syndrome were
no more likely to pass a false belief task than those with
Asperger syndrome, and that both these groups were significantly more likely to pass the task than peers with autism.
Children with various other impairments such as learning disabilities* have also been shown to fail ToM tasks (e.g. Charman
and Campbell 1997). Therefore, despite several years of
research, ToM does not present a clear-cut picture and this
measure alone may not be sufficient when describing differences between similar clinical groups. Thus other areas of
impairment, such as language, warrant more comparative
investigation. Tests which tap specifically into underlying linguistic mechanisms (such as non-word repetition and other
process-based tasks) may indeed be more able to provide diagnostic information than standard language tests.
This study builds on previous work showing that four psycholinguistic tasks could accurately identify a group of children
with a history of SLI in relation to typically developing peers
(Conti-Ramsden et al. 2001). Three of these tasks are used
here, namely the Childrens Non-Word Repetition (CNRep;
Gathercole and Baddeley 1990), the Past Tense Task (PTT;
Marchman et al. 1999), and the CELF Recalling Sentences;
Semel et al. 1987). As described earlier, these potential markers
have also been the subject of other research and are widely
used in clinical practice, making them good candidates for use
in comparative analysis and diagnostic screening. In this stage
*US usage: mental retardation.

of the investigation we aimed to examine the following: (1)

were these psycholinguistic markers specific to SLI, or might
they be useful when exploring PLI? (2) which, if any, were best
suited to each clinical group in turn? and (3) could different but
closely related groups of children with communication impairments be accurately distinguished from one another using
these particular markers?
Method
PARTICIPANTS

In total 67 children with communication impairments participated in this study: 29 children with SLI (25 males, 4 females;
mean age 10 years 10 months, range 10 years 2 months to 11
years 9 months); 25 children with PLI (22 males, three females;
mean age 11 years 3 months, range 10 years 2 months to 12
years 5 months); and 13 children with ASD (all males; mean age
10 years 10 months, range 10 years 2 months to 12 years 6
months). These sample sizes were not calculated a priori, but
represented the numbers of children from a wider study who
met clinical criteria (see below).
The groups were defined a priori and are described below.
The first group included all those who met criteria for any
group were taken from a pool of 242 children recruited from
language units at age 7 years and followed up at 11 years. This
age represents an important stage for children in the UK, the
majority of whom are about to move to secondary educational
placements. These children were initially invited to participate
in the Nuffield Foundation Project (see Conti-Ramsden and
Botting 1999) examining school placements after language
unit attendance. Initial recruitment at 7 years was done via telephone contact with schools who in turn agreed to send out
consent forms to families. Every language unit in England was
contacted and in total 108 were visited representing the vast
majority of placements of this kind. The 242 children represented a 50% random sample of consenting children aged 7
attending language units. Where a unit had only one child eligible, they were visited in order to get as wide a spread of units as
possible. By 11 years, these children had been assessed twice
by our team and families knew that the study was ongoing.
They had been asked to give home addresses if they were
happy to be contacted in this way. The second group comprised an additional five children with ASD and six children
with PLI who were recruited from local specialist schools for
children with communication difficulties/autism. This was carried out in a similar way, with initial telephone contact with
teachers in order to identify families to whom consent forms
might be sent via school. In all cases consent forms were sent
with parent-friendly information about the study and assessments and with a stamped addressed envelope for return to
the university. Table I shows group characteristics except for

performance IQ, which is reported within the results section.

Unsurprisingly (as they reflect participant group criteria) Test
for Reception of Grammar (TROG: Bishop 1982), Expressive
Vocabulary Test (EVT: Williams 1997), CARS (Schopler et al.
1980), and CCC (Bishop 1998) scores all differed significantly
across groups.
Children with SLI
These children were all required to have had a firm clinical history of SLI (they had in fact all been attending specialist education for language impairments and all had statements stating
language as their primary difficulty although these were not a
priori requirements). However, no current status or history of
pragmatic impairments was allowed as measured by scores
over 132 on the CCC pragmatic scale (Bishop 1998). They were
required to have EVT scores below the 10th centile and TROG
scores below the 50th centile. (Sixteen of these 29 children in
fact had TROG scores below the 10th centile, thus presenting
with both expressive and receptive difficulties, but analysis
showed that there was virtually no difference in results when
using this more impaired group see results. All children were
also required to have short-form Performance IQs of 70 or
above (Wechsler Intelligence Scale for Children, 3rd edn;
Wechsler 1992).
Children with PLI
These children were required to have current or previous
CCC pragmatic scale scores of <132. Four children with
previous low CCC in fact showed borderline 11 year CCC
scores between 132 and 138 at 11 years. CARS scores of <30
(i.e. non-autistic) were also required as were short-form
Performance IQs of 70 or above. No linguistic criteria were
specified.
Children with ASD
These children were required to have CARS scores of 30 or
more and to have definite clinical diagnoses of autism. Shortform Performance IQs of 70 or above were also required. No
other criteria were specified. It should be noted here that the
majority of this group were recruited from language units.
This is likely to mean that these children were higher functioning than might be expected from the general population of those with ASD. It may also be that these childrens
difficulties are qualitatively different from others identified
as ASD. However, no difference was found between the
children with ASD from the language unit and those recruited later. Despite the language unit criteria (which generally
exclude autism), children with definite diagnoses of autism
regularly attend such classes, especially if they are high
functioning.

Table I: Group characteristics medians (interquartile ranges)

Impairment
SLI
PLI
ASD

EVTa

TROGa

CCCa

CARSa

Age, y:m

Males:Females

1 (0 to 2)
11 (0 to 31)
5 (0 to 16)

4 (3 to 30)
31 (4 to 41)
38 (5 to 50)

145 (143 to 151)

125 (119 to 135)
118 (112 to 134)

18 (17 to 20)
21 (20 to27)
34 (33 to 43)

10:10 (10:2 to 11:9)

11:3 (10:2 to 12:5)
10:10 (10:2 to 12:6)

25:4
22:3
13:0

aAll significantly different across groups at p<0.01. EVT, Expressive Vocabulary Test (Williams 1997); TROG, Test for Reception of Grammar (Bishop

1982); CCC, Childrens Communication Checklist (Bishop 1998); CARS, Childhood Autism Rating Scale (Schopler et al. 1980). SLI, specific
language impairment; PLI, primary pragmatic language impairment; ASD, autistic spectrum disorders.

Psycholinguistic Markers and Communication Impairments Botting and Conti-Ramsden 517

MEASURES

MARKER TASKS

Although for some groups criteria were limited to only some

tasks, all children with communication impairments have
data from the following diagnostic assessments:

The following tasks were examined across groups:

Past Tense Task (PTT)
This is a test designed to assess correct grammatical usage of
verbs in past tense form (e.g. he cleaned) consisting of 52
line drawings shown to the child one at a time. With each picture, the assessor reads out a sentence related to the picture,
which the child must complete. The items are balanced and
randomized for frequency of verbs and for regular versus
irregular verb forms.

Expressive Vocabulary Test (EVT)

This is an assessment of expressive vocabulary, which covers
a wide age range (2 years to adult). Children are shown a picture and also told a key word by the assessor. The child must
then elicit another appropriate word that matches both the picture and the key word (e.g. a picture of steps and the word
steps are given. The child must produce stairs or another
similar alternative).

CELF Revised Recalling Sentences subtest

For this task children are given a sentence and asked to repeat
it verbatim. Sentences become increasingly longer and more
complex. Responses are scored in relation to the number of
errors made in each sentence.

Test for Reception of Grammar (TROG)

This is a test of oral comprehension of syntax in which children are shown four pictures while the examiner reads a sentence. The child picks the picture that illustrates the
sentence (e.g. the cat the cow chases is black).

Childrens Non-Word Repetition (CNRep)

This is a test of verbal/phonological short-term memory
consisting of 40 non-words. In this study live presentation
was used rather than the audiotape version. In this paradigm,
the researcher says a non-word (e.g. barrazon) while hiding
their lips behind a screen of paper (to avoid visual cues being
used by the child). The child must repeat the word exactly.
The researcher is not permitted to repeat the word.

Childrens Communication Checklist (CCC; teacher rated)

The CCC is shown by Bishop (1998) to differentiate between
children with PLI and those with more typical SLI. The checklist
consists of nine subscales of communication and interactive
behaviour: speech, syntax, inappropriate initiation, coherence,
stereotyped conversation, context, rapport, social behaviour,
and interests. A composite pragmatic impairment score can be
derived from the middle five scales (inappropriate initiation,
coherence, stereotyped conversation, context, and rapport)
and a score of 132 or below is used as a cut-off point for pragmatic language impairment. Each scale consists of a number of
behavioural items which professionals are asked to rate as does
not apply, applies somewhat, or definitely applies.

PROCEDURE

The study was approved by Manchester University Research

Ethics Committee. Following informed written consent from
schools and parents/carers, children were visited at school and
assessed individually in a quiet room or area. Testing on marker
tasks was completed as part of a wider battery of language and
cognition tests during a single visit, at the childs own pace and
with normal school breaks. Children were, of course, allowed
to stop testing and return to the classroom at any stage. In practice this meant that a few 11-year-olds did not complete the
wider test battery (not included here) if they expressed a wish
to stop but in general children seemed to enjoy the tasks.
Examiners were not blind to placement status or clinical
group, as measures were administered at school, however, the
tasks were considered to be reasonably objective.

Childhood Autism Rating Scale (CARS; teacher rated)

This is a scale of 15 behavioural items each scored from 1 to
4. It covers a range of behaviours typical of autistic spectrum
disorders. Total score ranges from 15 to 60 and a threshold of
30 indicates an autistic disorder. Although a fairly crude measure, this tool was used to provide a guide to the number of
children with marked autistic traits.
Short form Performance IQ (WISC-III)
This consisted of the subtests Block Design and Picture
Completion combined to form an estimated performance or
non-verbal IQ. The performance short form in particular has
been found to correlate well with a full IQ battery and has
been used in other studies of cognitive ability and language
(Sattler 1974, Hohnen and Stevenson 1999).

Results
DIFFERENCES ACROSS GROUPS

Using our three a priori groups, CNRep score, and CELF

Recalling Sentences differed significantly across groups, the
children with SLI scoring most poorly in both cases (Kruskal
Wallis 2, (2) 10.7, p=0.005; KruskalWallis 2, (2) 6.4, p<0.04).

Table II: Marker differences using four groups medians (interquartile ranges)

SLI
PLI pure
PLI plus
ASD
KruskalWallis 2 value (df=3)

CNRepa

PTTa

Recalling Sentencesa

Performance IQb

1 (1 to 7)
13 (1 to 18)
15 (13 to 69)
8 (7 to 16)
11.8 p=0.008

4 (1 to 6)
1 (1 to 7)
69 (4 to 90)
5 (1 to 16)
9.0 p=0.029

1 (1 to 4)
2 (1 to 6)
17 (4 to 40)
5 (1 to 9)
10.0 p=0.019

85 (76 to 90)
99 (92 to 119)
110 (92 to 129)
90 (76 to 107)
16.8 p=0.001

aCentiles rounded to whole numbers; bstandard IQ points. CNRep, Childrens Non-word Repetition; PTT, past tense task; SLI, specific language
impairment; PLI pure, primary pragmatic language impairment; PLI plus, PLI with some autistic-like behaviours; ASD, autistic spectrum disorders.

518

Developmental Medicine & Child Neurology 2003, 45: 515524

REGROUPING OF DATA

Further examination of the data suggested that the PLI group

was not as homogenous as the other two groups. This was represented by wide degrees of variance on the measures (Table
II) and bi-modal distributions of scores as seen in Figure 1.
Thus the PLI group was split into two: those with PLI plus
scores (not only <132 on CCC but also <24 block H restricted
interests or <28 block, I social impairment; see Bishop 1998)
and those whom we will refer to as being PLI pure (<132 on
CCC but with no indication of restricted social behaviour or
interests). This resulted in 11 children with PLI plus and 14
with PLI pure. The reason this criterion was used for separating the groups was to avoid a circular analysis taking place in
which defining factors become the same as dependent variables. Instead this more theoretical split was made. This in fact
represented the bi-modal data moderately well, with 11 of 14
children with PLI pure having low Recalling Sentences and
low PTT scores and only two of 11 PLI plus children falling into
this group, where low was defined by the bi-modal split in
data (<20 centiles on Recalling Sentences and <40 centiles on
PTT). The new CCC and CARS median scores for PLI pure were
128 (interquartile range [IQR]=119 to 144) and 24 (IQR=19
to 28) respectively and for the PLI plus group were 124
(IQR=122 to 130) and 23 (IQR=20 to 26) respectively.

did not differ significantly from each other on this measure.

CELF Recalling Sentences
All groups except the children with PLI plus scored below the
10th centile for age on this test, setting them apart in terms of
language impairment. Post hoc MannWhitney U comparisons confirmed that significant differences were only present
between PLI plus and PLI pure and PLI plus and SLI (PLI plus
vs PLI pure p=0.036; SLI p=0.005). The difference between ASD
and PLI plus fell short of significance (p=0.148). The other
groups (PLI pure, SLI, and ASD) did not differ from each other
on this measure.
In summary, group data indicate that CNRep is lower for
children with SLI than all other groups while PTT and Recalling
Sentences are uniformly low for all groups except PLI plus.
COGNITIVE ABILITY AND CLINICAL MARKERS

Although all children showed normal range IQ scores, Performance IQ was significantly different across groups for both
the three-group and four-group analyses, with both PLI groups

16
16
14
14

Number of children

Past tense was not significant (KruskalWallis 2, (2) 1.4,

p=0.49). The more restricted group of children with SLI (those
with expressive and receptive scores below the 10th centile,
n=16) were also used in place of the general SLI group in an
identical KruskalWallis test. Almost identical results were
found. In addition, there was no significant difference found
between children with purely expressive problems (n=13)
and children with both expressive and receptive problems
(n=16) on any marker using MannWhitney U comparisons.
This more limited subgroup was, therefore, not used in future
comparisons and all children with SLI were combined in the
remaining analyses.

12
12
10
10

88
66
44
22
0

ANALYSIS USING FOUR GROUPS

20.0
20

40.0
40

60.0
60

80.0
80

100.0
100

Past Tense
Task centiles
PLI group
past tense
percentiles
forforPLI

The KruskalWallis analyses were performed again using these

four groups. With the separated PLI subgroups, CNRep, and
Recalling Sentences remained significant across groups and
PTT now also showed significant differences by group. Table
II shows the medians and interquartile ranges, which were
analyzed post-hoc using MannWhitney U comparisons, and
are summarized in text for each potential marker below.

12
12

10
10

Number of children

Childrens Non-Word Repetition (CNRep)

Children with SLI performed most poorly on this measure, followed by peers with autism and children with both variants of
PLI. Univariate post-hoc statistics (MannWhitney U) showed
that significant CNRep differences existed between those with
SLI and all other groups (SLI vs PLI plus p=0.006; PLI pure
p=0.048; ASD p=0.011). The other three groups (PLI plus, PLI
pure, and ASD) did not differ significantly from each other on
this measure.

0.0
0

88
66
44
22
0

Past Tense Task (PTT)

For past tense, the SLI, PLI pure, and ASD groups scored much
more poorly with significant differences seen between these
groups and the PLI plus group (PLI plus vs PLI pure p=0.022;
ASD p=0.043; SLI p=0.005). The SLI, PLI pure, and ASD groups

0.0
0

10.0
10

20.0
20

30.0
30

40.0
40

50.0
50

CELF
Recalling sentences
Sentences centiles
for PLI
group
CELF
recalling
centile
for
PLI

Figure 1: Examples of bi-modal distribution in primary

pragmatic language impairment (PLI) group.

Psycholinguistic Markers and Communication Impairments Botting and Conti-Ramsden 519

First, we looked at which test and cut-off point offers the maximum values for identifying each group, compared with a normative threshold. The control group (n=100), used in a wider
study, comprised 51 females and 49 males; mean age was 11
years, age range 10 years 5 months to 11 years 6 months (see
Simkin and Conti-Ramsden 2001). Second, a comparison of
accuracy in relation to other groups with communication
impairments to explore whether any markers could be used to
discriminate between them. Accuracy levels are difficult to
interpret where the sample sizes are very different between
groups (n=100 for controls and <25 for all clinical groups)
and so have not been reported for the first analysis.
In the first approach, the best sensitivity and specificity values per se (using normal control thresholds and regardless of
other groups) were as outlined in Table IV. Values of 75% sensitivity and 85% specificity were used following other studies
with similar values (e.g. Tomblin et al. 1996).

gaining significantly higher scores than peers with SLI or ASD

(see Table II). IQ scores have been represented in typical standard scores (where population mean=100 SD=15) because
this represents the most frequently used presentation, but an
IQ of 100 is approximately equivalent to a 50th centile score,
IQ of 85=16th centile score, and an IQ of 70=2.5th centile.
ANCOVAs to assess the effect of Performance IQ on other markers were not possible due to highly skewed data and relatively
small numbers. However, Spearmans correlations of markers
with Performance IQ were quite different for each group
revealing that only the PLI pure group had significant associations between IQ and markers (Table III). The association
between PLI pure and IQ is discussed later.
SENSITIVITY AND SPECIFICITY

An exploratory examination of sensitivity and specificity was

then undertaken. It should be noted here that with small group
sizes, the results might be imprecise. However, as the aim of
this article was to investigate the clinical use of markers,
some level of individual analysis was deemed interesting.
Two aspects of sensitivity and specificity were examined.

MARKER ACCURACY BY COMMUNICATION GROUP

For the SLI group all markers were reasonably accurate, thus
identifying one as the most accurate proves difficult. Recalling
Sentences and PTT using 16th or 10th centile thresholds both
showed sensitivity of 85% and above. Non-word repetition
was also able to differentiate 79% of these children from normally developing peers (see Simkin and Conti-Ramsden
2002). For the PLI pure group Recalling Sentences and PTT
using a 16th or 10th centile threshold again provided accurate
group membership. Children with ASD were most accurately
identified from children without language impairment using
Recalling Sentences with a 16th centile cut-off point. For children with PLI plus, none of the potential markers were strong
discriminators.

Table III: Correlations between performance IQ and markers for

each group
Impairment
SLI
PLI pure
PLI Plus
ASD

CNRep
0.30
0.68a
0.05
0.30

Past tense

Recalling Sentences

0.20
0.75a
0.24
0.38

0.23
0.81a
0.14
0.42

a p<0.01. CNRep, Childrens Non-Word Repetition; SLI, specific

MARKER ACCURACY OVER ALL GROUPS

language impairment; PLI, primary pragmatic language

impairment; ASD, autistic spectrum disorders.

If children with all types of communication impairment were

to be discriminated from children without language impairment, Recalling Sentences appears to be the most accurate

Table IV: Results of sensitivity (bold text) and specificity analysis for each group using each marker at three different thresholds

Non-word repetition
SLI
PLI pure
PLI plus
ASD
Normal control group
Past Tense Task
SLI
PLI pure
PLI plus
ASD
Normal control group
Recalling Sentences
SLI
PLI pure
PLI plus
ASD
Normal control group

<16th centile (%)

<10th centile (%)

<2.5th centile (%)

23/29 (79)
9/14 (64)
5/10 (50)
8/13 (62)
13/100 (13)
thus 87/100 above (87 specificity)

23/29 (79)
6/14 (43)
2/10 (20)
7/13 (54)
8/100 (8)
thus 92/100 above (92 specificity)

18/29 (62)
4/14 (29)
2/10 (20)
2/13 (15)
2/100 (2)
thus 98/100 above (98 specificity)

25/28 (89)
12/14 (86)
3/10 (30)
8/11 (73)
11/100 (11)
thus 89/100 above (89 specificity)

25/28 (89)
12/14 (86)
3/10 (30)
8/11 (73)
7/100 (7)
thus 93/100 above (93 specificity)

11/28 (39)
8/14 (57)
1/10 (10)
4/11 (39)
0/100 (0%)
thus 100/100 above (100 specificity)

26/29 (90)
13/14 (93)
5/10 (50)
11/13 (85)
15/100 (15)
thus 85/100 above (85 specificity)

26/29 (90)
11/14 (79)
4/10 (40)
9/13 (69)
8/100 (8)
thus 92/100 above (92 specificity)

21/29 (72)
6/14 (43)
2/10 (20)
5/13 (39)
1/100 (1)
thus 99/100 above (99 specificity)

SLI, specific language impairment; PLI pure, primary pragmatic language impairment; PLI plus, PLI with some autistic like behaviours; ASD, autistic
spectrum disorders.
520

Developmental Medicine & Child Neurology 2003, 45: 515524

tool. This task was able to identify all but the PLI plus group at
>90% sensitivity and 85% specificity. It may also have other
beneficial qualities as a marker, such as being able to identify
individuals with a history of SLI even when language difficulties
have resolved (Conti-Ramsden et al. 2001). The next most useful measure appears to be the PTT followed by CNRep, which
only reached acceptable levels of sensitivity with the SLI group.
MARKER ACCURACY IN DISCRIMINATING BETWEEN DIFFERENT
GROUPS OF IMPAIRMENT

If clinically we were interested in identifying which children

belonged to particular groups within the larger sample of individuals with communication impairments, then alternative
tasks may be optimum. This is the second aspect of sensitivity
and specificity referred to above, and involves examining
directly whether the tasks can discriminate effectively between
different groups of children with communication impairments
(e.g. SLI from ASD). Values are shown in Table V. In this Table,
only the most effective cut-off levels for each marker have
been shown and only those analyses which resulted in 70%
accuracy or above.
Using this approach, it would appear that children with PLI
pure cannot be reliably separated from either the SLI group
or the ASD group using the markers examined here (accuracy
values all below 70%). The PLI plus group on the other hand
can be distinguished from peers with both SLI and PLI pure
with good overall accuracy (around 80%), but not as reliably
from peers with ASD (best accuracy at 71%). Furthermore, SLI
can also be discriminated from ASD at this level (72% best)
despite the fact that both these groups have linguistic difficulties. The results may indicate interesting theoretical groupings (for example that PLI pure appears to fall between SLI
and ASD whilst PLI plus is distinct from SLI), but in general
the levels of accuracy mean that clinical diagnosis using the
markers examined here would be unreliable. One exception
may be the use of such markers to establish a childs difficulty
as PLI plus in nature. These children are often placed in specialist language units and early identification of PLI plus may
mean more accurate educational placement for them.
Again, the Recalling Sentences task has proved to be a useful
discriminator, along with the PTT, both of which were reasonably accurate on four of six comparisons made. Notably, Nonword Repetition was much poorer at differentiating groups in
this individual case analysis than in group comparisons and
was only accurate at distinguishing SLI from PLI plus.
USING MARKERS IN COMBINATION WITH OTHER MEASURES

Data here suggest that the markers chosen because they had
previously tapped into SLI, also tap into non-specific linguistic
impairments, and thus do not clearly separate groups (with
the exception of PLI plus). Therefore, an alternative way of
examining the groups would be to look at markers in relation
to other defining characteristics. Table VI shows the areas of
impairment identified as characteristic of each group and
reveals that each has a different pattern. In many respects this
is unsurprising as two of these characteristics were used as
selective criteria for groups. However, having selected groups
on this basis, the groups with ASD and PLI plus and those with
PLI pure and PLI plus can be distinguished from one another
using the markers examined here. Furthermore, qualitative
analysis of the CARS items show that ASD is significantly different from all other groups on the majority of items. This is

important when considering PLI plus, which shows lower

scores on all but one item of this scale (item 10) compared with
peers with ASD.
Discussion
This report examines the usefulness of various clinical markers, which have already been shown to be effective descriptors
of SLI, among a number of different groups with pragmatic
impairments. It shows that psycholinguistic markers can
indeed be useful in the exploration of this aspect of impairment and the main findings were threefold. First, the results
provide further evidence that the group currently referred to
clinically as having PLI is in fact two quite different subsets of
children. Second, the study reveals that clinical markers can
effectively identify children with all types of linguistic impairment (specific or otherwise) compared with normally developing peers, but not those in the PLI plus group. CELF Recalling
Sentences was shown to be the most effective marker overall.
Interestingly, severe deficits in non-word repetition appear to
be quite specific to the difficulties of children with SLI
(although see later discussion). Last, data suggest that for some
types of impairment, individuals can be assigned to different
groups in the area of communication difficulty with moderate reliability, using markers such as Recalling Sentences. Along
with non-linguistic measures, such tasks might prove a useful
positive system to aid differential diagnosis. Each of these
themes is discussed further below.
THE NATURE OF PRIMARY PRAGMATIC LANGUAGE IMPAIRMENT

One of the groups initially described here as PLI is characterized by severe pragmatic language difficulties without autistic traits and with a number of linguistic difficulties. The
second PLI group is that described by Bishop (1998) as PLI
plus, who as well as pragmatic language difficulties appear

Table V: Sensitivity and specificity values of comparison groups

CNRep

Past Tense
Task

Recalling
Sentences

ASD vs PLI plus

Sensitivity, score (%)
Specificity, score (%)
Accuracy (%)

8/11 (73)
7/10 (70)
71 (10th c)

11/13 (85)
5/10 (50)
70 (16th c)

PLI pure vs PLI plus

Sensitivity, score (%)
Specificity, score (%)
Accuracy (%)

12/14 (86)
7/10 (70)
79 (10th c)

13/14 (93)
5/10 (50)
75 (16th c)

SLI vs ASD
Sensitivity, score (%)
Specificity, score (%)
Accuracy (%)

25/28 (89)
3/11 (27)
72 (10th c)

26/29 (90)
4/13 (30)
71(10th c)

SLI vs PLI plus

Sensitivity, score (%) 23/29 (79)
Specificity, score (%)
8/10 (80)
Accuracy (%)
80 (10th c)

25/28 (89)
7/10 (70)
84(10th c)

26/29 (90)
6/10 (60)
82(10th c)

PLI pure vs ASD

SLI vs PLI pure

CNRep, Childrens Non-Word Repetition. Only comparisons with

accuracy values of at least 70% are shown. c, centile.

Psycholinguistic Markers and Communication Impairments Botting and Conti-Ramsden 521

to have some autistic-type characteristics (such as narrow

interests, obsessions, and marked social difficulties), but have
none of the linguistic difficulties of the other three groups. The
fact that these two groups are evident in independent data
(i.e. measures not designed to describe PLI or the other
impairments) makes a strong case that two groups exist.
However, in clinical practice, certainly in speech and language services, these two groups are most likely to be treated
as one homogenous group. The existence of two groups
makes more sense of the apparently contradictory literature
about whether this impairment should be described as either
language impairment or an autistic spectrum disorder (e.g.
Brook and Bowler 1992, Shields et al. 1996, Boucher 1998).
Two separate groups would also explain the difficulty experienced by those in research and in the field, in finding universal or key characteristics to define these children (Bishop
and Adams 1989, Botting and Conti-Ramsden 1999).
Thus both schools of thought may be correct: descriptions
of PLI as a language impairment (through lack of obsessive
behaviour, rigidity of thought, or of marked social difficulties
necessary for a diagnosis of autism, and the presence of significant linguistic difficulties not usually seen in those with
Asperger syndrome) could in fact be focussing on those individuals often found in mainstream language units and who are
described here as having PLI pure. On the other hand, the argument for PLI as autism (including evidence of poor ToM skills
[Shields et al. 1996] along with more pervasive difficulties than
expected in SLI) may apply directly to children with PLI plus.
These latter children may find themselves in specialist language schools or educational centres for autism, and indeed
might be regarded as children with Asperger syndrome. Like
individuals with Asperger syndrome, the PLI plus group have
social impairments, communication difficulties, and restricted
interests, and show few severe linguistic difficulties (only three
children had difficulties with the PTT for example). Their IQ
levels were highest of the clinical groups and they could be separated with moderate reliability from the ASD peers who participated in the study on this basis. Furthermore, on closer
analysis of the CARS, the PLI plus and ASD groups scored significantly differently on all but one of the individual items (item
10: Fear and Nervousness). Those with PLI plus also scored
marginally higher than those with PLI pure on CARS items
about social relations and emotional response as well as
regarding adaptation to change. The fact that both the PLI
plus and the ASD children participating here came through

Table VI: Profiles of impairment in different groups

Impairment

CCC
<132

CCC plus
criteria
fulfilled

CARS >30

85% identified
by at least one
LI marker

speechlanguage therapy services, makes it unlikely that

simple explanations, such as different professional bias,
accounts entirely for the fact that the PLI plus children do not
currently have diagnoses of Asperger syndrome. However, the
viability of this description for children with PLI plus warrants
further consideration.
In contrast, the PLI pure group showed a closer similarity
to the SLI group than to ASD peers on marker performance
medians. However, they did not show severe characteristic
deficits in non-word repetition (see next section) and were
not accurately distinguishable from either SLI or ASD groups
using a sensitivity/specificity approach.
NON -WORD REPETITION AND RECALLING SENTENCES : THEIR
RELATION TO SLI

Although CELF Recalling Sentences has been shown to provide the most efficient discrimination for children with SLI or
a history of difficulties, this study reveals that it may also be
useful in identifying wider communication impairments
such as ASD and other pervasive developmental difficulties,
and in discriminating between groups within this area of
impairment.
Our previous speculation that all children with communication difficulties might perform poorly on non-word repetition
was borne out to some extent (Conti-Ramsden et al. 2001).
This is supported by the fact that all group medians fell below
16th centile (1SD) on this task compared with typically developing peers and that only one intergroup comparison was
possible using the CNRep in sensitivity and specificity analysis.
However, the CNRep also revealed that children with SLI were
significantly poorer than all the other groups on this task, and
that using severe thresholds (<2.5th centile; 2SD) children
with SLI were the only group identified in significant numbers
by this task. This is consistent with other literature (e.g. Bishop
et al. 1996, Stothard et al. 1998) which suggests that CNRep
may be a useful marker for SLI. In addition, the fact that nonword repetition does not accurately identify either PLI group
or ASD using any threshold, despite marked communication
difficulties, suggests different underlying mechanisms for
the different disorders in which phonological memory features more strongly in the linguistic impairment seen in SLI
compared with that seen in ASD. Although children with ASD
are identified by markers, performing fairly poorly on nonword repetition as in other studies (Kjelgaard and Tager-Flusberg 2000), the majority do not have the same severe non-word
repetition profile as children with SLI.
The fact that sensitivity and specificity are not as high as they
might be for CNRep is an interesting one given that in group
comparisons, children with SLI perform so poorly on this measure (median=1st centile; <2SD). The apparent contradiction of these findings is probably the result of the wide
heterogeneity found in children with SLI as well as widely varying patterns of intervention, leading to six children in this
group performing above the 16th centile threshold ( Table V).

SLI

PLI pure

PLI plus

DIFFERENTIAL DIAGNOSIS OF CHILDREN WITH PRAGMATIC

ASD

IMPAIRMENTS

CCC, Childrens Communication Checklist (Bishop 1998); CARS,

Childhood Autism Rating Scale (Schopler et al. 1980). LI, language
impairment; SLI, specific language impairment; PLI, primary
pragmatic language impairment; ASD, autistic spectrum disorders.

522

Developmental Medicine & Child Neurology 2003, 45: 515524

A role for psycholinguistic markers?

Communication disorders are often difficult to define in positive rather than exclusionary ways. For example, although universal features of SLI such as a difficulty with verb morphology
have long been discussed (Leonard 1998), these have been

difficult to pin down and are by definition not subclinical in

nature. Furthermore, as children develop, many familiar formal language tasks can be learned. This is evident here in the
relatively high TROG scores shown by the ASD group, as this is
a task they will have encountered a number of times in their
educational progression. This study confirms previous reports
(e.g. Gathercole and Baddeley 1990, Bishop et al.1996, ContiRamsden et al. 2001) that certain psycholinguistic measures
may be useful in the accurate identification and screening of
language difficulties. Our finding that those with severe
receptive SLI scored no differently from those with a more
expressive impairment may also indicate that these markers
can tap into underlying difficulties that are not always evident from standard language tests.
Continuums and categories
Data here appear to show that different groups of children with
communication impairments experience subtly different patterns of difficulty in psycholinguistic terms as well as in terms of
autistic features. We feel that, despite the difficulty in separating
out these differences, it is also important to be able to distinguish children with one type of impairment from another (that
is to categorize difficulties), in order to access relevant treatment, support, and educational systems. Furthermore, psycholinguistic markers in combination with other forms of
assessment might provide key information on the course of
action in diagnosis and remediation, such as whether to use
other techniques and measures like EEG results or motor skills.
A categorization process may also enhance work that is seeking
to establish theoretical and causal groupings. Unlike Bishop
and Edmundsons article (1987), data here do not point to a
mountain profile of severity whereby a continuum from wordlevel difficulties through syntax impairment is evident. Qualitative examination of the data does not reveal any obvious
patterns in impairment profile across the groups. However, to
examine this issue comprehensively, the wider cognitive and
social characteristics much investigated in autism need to be
applied to populations with other communication disorders in
order to tease out meaningful differences and similarities.
We need also to consider what these psycholinguistic tools
are measuring. There is increasing evidence, for example, that
general working memory difficulties are central to language
development problems (Adams and Gathercole 2000). One
could speculate, therefore, that a combination of this type of
general impairment (subtle enough not to cause general learning difficulties) and a deficit in linguistic processing leads to the
poor performance on psycholinguistic tests highlighted here.
However, further investigation is needed into differing combinations of this memory element alongside linguistic components in order to provide optimum screening tools and to
examine further the issue of continuums and categories.

given that verbal short-term memory has been shown to be

impoverished in these populations (e.g. Jarrold et al. 2000).
It is also the case that while for the majority of children with
communication impairments the markers are not significantly related to cognitive ability, for the PLI group this association was much stronger. It is unclear why this should be the
case. It is possible that for these children non-verbal IQ is a
powerful protective factor which enables many of their linguistic difficulties to be masked at least in every day conversation. In the same way, these individuals may rely much more
strongly on alternative strategies (such as visual short-term
memory) to complete marker tasks.
There are also difficulties with group data when applied to
a clinical setting. The authors have tried to show that individual discrimination in the form of sensitivity/specificity analysis
is at moderate levels for clinical use. However, these values
remain arbitrary in that it depends on the cost of over-inclusive
and under-inclusive screening measures. In the area of SLI versus typical development we would argue that it is probably
advantageous for the measure to be over-sensitive and under
specific that is, to risk falsely identifying a typically developing
individual rather than miss a child with SLI. However, in the
case of identifying a child as having autism compared to SLI, the
risks of misdiagnosis are much less clear and clearly, a number
of other types of measures need to be employed.
Conclusion
The main finding of this study was the presence of two subgroups of children with PLI (who do not fulfil criteria for
autism) and this needs to be investigated further in order to
clarify data from future studies in the area. The study suggests
that psycholinguistic markers can be useful not only in discriminating SLI from typically developing peers but also when
exploring the differences between different groups of children
with pragmatic impairments.
In this respect, CELF Recalling Sentences proved to be the
most effective universal marker for communication impairments. In contrast, CNRep appeared to be particularly associated with SLI rather than other groups.
Some discrimination between clinical groups was possible
using these markers and in combination with differences seen
in other measures (e.g. CARS), although levels of accuracy
were not optimum.
Further research is needed, particularly a better controlled
(such as experimenters blind to clinical group) and more comprehensive examination of key linguistic and cognitive tasks
(such as those tapping short-term and working memory).
Studies enrolling large numbers of children are also needed to
help identify genuine diagnostic patterns. These may help lead
to the non-exclusionary and clinically applicable definitions of
communication disorders that are still much needed in order
to aid clinical support for children and their families.

SENSITIVITY AND SPECIFICITY AND NON -VERBAL COGNITIVE

DOI: 10.1017/S0012162203000963

ABILITIES

There are many reasons as to why the markers investigated

here may not be ideal. They are not, in fact, entirely subclinical and they may be influenced by other factors not tested as
part of the present study. For example, all of the children in
this study had performance IQ subtests above 70. It is highly
plausible that groups with generally low cognitive abilities
(including those with autism) would also be identified as
extremely poor performers on these measures, especially

Accepted for publication 1st April 2003.

Acknowledgements
We gratefully acknowledge Grant R000223262 from the Economic and
Social Research Council (ESRC). We would like to thank Zo Simkin
and Emma Knox for their help with data collection and, most importantly, the schools and families who helped us with this research.

Psycholinguistic Markers and Communication Impairments Botting and Conti-Ramsden 523

References
Adams AM, Gathercole SE. (2000) Limitations in working memory:
implications for language development. Int J Lang Commun
Disord 35: 95116.
American Psychiatric Association. (1994) Diagnostic and Statistical
Manual of Mental Disorders. 4th edn. Washington DC: American
Psychiatric Association.
Baird G, Charman T, Baron-Cohen S, Cox A, Swettenham J,
Wheelwright S, Drew A. (2000) A screening instrument for autism
at 18 months of age: a six-year follow-up study. J Am Acad Child
Adolesc Psychiatry 39: 694702.
Bishop DVM. (1982) Test for Reception of Grammar. Published by
author at Manchester University, Manchester, UK.
Bishop DVM. (1998) Development of the Childrens
Communication Checklist (CCC): a method for assessing
qualitative aspects of communicative impairment in children.
J Child Psychol Psychiatry 39: 879893.
Bishop DVM. (2001) Genetic influences on language impairment
and literacy problems in children: same or different? J Child
Psychol Psychiatry 42: 189198.
Bishop DVM, Edmundson A. (1987) Language-impaired four-yearolds: distinguishing transient from persistent impairment.
J Speech Hear Disord 52: 156173.
Bishop DVM, Adams C. (1989) Conversational characteristics of
children with semantic-pragmatic disorder. II: What features
lead to a judgement of inappropriacy? Br J Disord Commun
24: 241263.
Bishop DVM, North T, Donlan C. (1996) Non-word repetition as a
behavioural marker for inherited language impairment: evidence
from a twin study. J Child Psychol Psychiatry 37: 391403.
Bishop DVM, Norbury CF. (2002) Exploring the borderlands of
autistic disorder and Specific Language Impairment: a study
using standardised diagnostic instruments. J Child Psychol
Psychiatry 43: 917930.
Botting N, Conti-Ramsden G. (2002) Non-word repetition and
language development in children with language impairments.
Int J Lang Commun Disord 36: 421432.
Botting N, Conti-Ramsden G. (1999) Pragmatic language impairment
without autism: the children in question. Autism 3: 371396.
Boucher J. (1998) SPD as a distinct diagnostic entity: logical
considerations and directions for future research. Int J Lang
Commun Disord 33: 7181.
Briscoe J, Bishop DVM, Norbury CF. (2001) Phonological processing,
language and literacy: a comparison of children with mildmoderate sensorineural hearing loss and those with specific
language impairment. J Child Psychol Psychiatry 42: 329340.
Brook S, Bowler D. (1992) Autism by another name? Semantic and
pragmatic impairments in children. J Autism Dev Disord 22: 6181.
Charman T, Campbell A. (1997) Reliability of theory of mind task
performance by individuals with a learning disability: a research
note. J Child Psychol Psychiatry 38: 725730.
Conti-Ramsden G, Botting N. (1999) Classification of children with
specific language impairment. J Speech Lang Hear Res
42: 11951204.
Conti-Ramsden G, Botting N, Faragher B. (2001) Psycholinguistic
markers for SLI. J Child Psychol Psychiatry 42: 741748.
Conti-Ramsden G, Windfuhr K. (2002) Productivity with word order
and morphology: a comparative look at children with SLI and
children with normal language abilities. Int J Lang Commun
Disord 37: 1730.
Cox A, Klein T, Charman T, Baird G, Baron-Cohen S, Swettenham J,
Drew A, Wheelwright S. (1999) Autism spectrum disorders at 20
and 42 months of age: stability of clinical and ADI-R diagnosis.
J Child Psychol Psychiatry 40: 719732.
Gathercole SE, Baddeley AD. (1990) Phonological memory deficits
in language disordered children: is there a causal connection?
J Memory Lang 29: 336360.
Hohnen B, Stevenson J. (1999) The structure of genetic influences
on cognitive, language, phonological and reading abilities. Dev
Psychol 35: 590603.
Jarrold C, Baddeley AD, Hewes AK. (2000) Verbal short-term memory
deficits in Downs Syndrome: a consequence of problems in
rehearsal? J Child Psychol Psychiatry 41: 233244.
Kjelgaard MM, Tager-Flusberg H. (2000) An investigation of language
impairment in autism: implications for genetic subgroups. Lang
Cognitive Processes 16: 287308.

524

Developmental Medicine & Child Neurology 2003, 45: 515524

Klee T, Pearce K, Carson DK. (2000) Improving the positive predictive

value of screening for developmental language disorder. J Speech
Lang Hear Res 43: 821833.
Landa R, Piven J, Wzorek MM, Gayle JO, Chase GA, Folstein S. (1990)
Social language use in parents of autistic children. Psychol Med
22: 245254.
Leonard LB. (1998) Children with Specific Language Impairment.
Cambridge, MA: MIT Press.
Lord C, Pickles A, McLennan J, Rutter M, Bregman J, Folstein S,
Fombonne E, Leboyer M, Minshew N. (1997) Diagnosing autism:
analyses of data from the Autism Diagnostic Interview. J Autism Dev
Disord 27: 501517.
Lord C, Rutter M, LeCouter A. (1994) Autism Diagnostic Interview
Revised: a revised version of a diagnostic interview for caregivers of
individuals with possible pervasive developmental disorders.
J Autism Dev Disord 24: 659685.
Marchman VA, Wulfeck B, Ellis Weismer SE. (1999) Morphological
productivity in children with normal language and SLI: a study of
the English past tense. J Speech Lang Hear Res 42: 206219.
Prior M, Eisenmajer R, Leekam S, Wing L, Gould J, Ong B, Dowe D.
(1998) Are there subgroups within the autistic spectrum? A cluster
analysis of a group of children with autistic spectrum disorders. J
Child Psychol Psychiatry 39: 893902.
Rice ML, Wexler K. (1996) Toward Tense as a clinical marker of specific
language impairment in English-speaking children.
J Speech Hear Res 39: 12391257.
Sattler JM. (1974) Assessment of Childrens Intelligence. Philadelphia:
WB Saunders.
Schopler E, Reichler RF, Devellis RF, Daly K. (1980) Toward objective
classification of childhood autism: Childhood Autism Rating Scales
(CARS). J Autism Dev Disord 10: 91103.
Semel E, Wiig E, Secord W. (1987) Clinical Evaluation of Language
Fundamentals Revised. San Antonio, TX: Psychological
Corporation.
Shields J, Varley R, Broks P, Simpson A. (1996) Social cognition in
developmental language disorders and high level autism. Dev Med
Child Neurol 38: 487495.
Simkin Z, Conti-Ramsden G. (2001) Non-word repetition and
grammatical morphology: normative data for children in their final
year of primary school Int J Lang Commun Disord 36: 395404.
Stothard SE, Snowling MJ, Bishop DVM, Chipchase BB, Kaplan CA.
(1998) Language impaired preschoolers: a follow up into
adolescence. J Speech Lang Hear Res 41: 407418.
Tomblin JB, Records N, Zhang X. (1996) A system for the diagnosis of
specific language impairment in kindergarten children.
J Speech Hear Res 39: 12841294.
Wechsler D. (1992) Wechsler Intelligence Scale for Children. 3rd edn.
San Antonio, TX: Psychological Corporation.
Williams K. (1997) Expressive Vocabulary Test. Circle Pines,
MN:American Guidance Service.
Ziatas K, Durkin K, Pratt C. (1998) Belief term development in
children with autism, Asperger Syndrome, specific language
impairment and normal development: links to theory of mind
development. J Child Psychol Psychiatry 39: 755764.

List of abbreviations
ASD
CARS
CCC
CELF
CNRep
EVT
PDDNOS
PLI
PTT
SLI
ToM
TROG

Autistic spectrum disorders

Childhood Autism Rating Scale
Childrens Communication Checklist
Clinical Evaluation of Language Fundamentals
Childrens Non-Word Repetition
Expressive Vocabulary Test
Pervasive developmental disorder not
otherwise specified
Primary pragmatic language impairment
Past Tense Task
Specific language impairment
Theory of mind
Test for Reception of Grammar