Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
substitution: evidence
from a study of adults
with a history of
developmental
language disorder
Dorothy V M Bishop* DPhil;
Andrew J O Whitehouse PhD;
Helen J Watt BA;
Elizabeth A Line BSc, Department of Experimental
Psychology, University of Oxford, Oxford, UK.
*Correspondence to first author at Department of
Experimental Psychology, University of Oxford,
South Parks Road, Oxford OX1 3UD, UK.
E-mail: dorothy.bishop@psy.ox.ac.uk
DOI: 10.1111/j.1469-8749.2008.02057.x
Published online 31st March 2008
Rates of diagnosis of autism have risen since 1980, raising the
question of whether some children who previously had other
diagnoses are now being diagnosed with autism. We applied
contemporary diagnostic criteria for autism to adults with a
history of developmental language disorder, to discover
whether diagnostic substitution has taken place. A total of 38
adults (aged 1531y; 31 males, seven females) who had
participated in studies of developmental language disorder
during childhood were given the Autism Diagnostic
Observation Schedule Generic. Their parents completed the
Autism Diagnostic Interview Revised, which relies largely
on symptoms present at age 4 to 5 years to diagnose autism.
Eight individuals met criteria for autism on both instruments,
and a further four met criteria for milder forms of autistic
spectrum disorder. Most individuals with autism had been
identified with pragmatic impairments in childhood. Some
children who would nowadays be diagnosed unambiguously
with autistic disorder had been diagnosed with developmental
language disorder in the past. This finding has implications
for our understanding of the epidemiology of autism.
341
Date of data
collection
Bishop et al.25
Bishop et al.26
Bishop et al.27
McArthur et al.28
Bishop et al.4
Norbury et al.29
Nr of individuals Nr of individuals
who received
who agreed to
study information
participate
19861987
1991
1992
19992001
19992001
20022003
23 SLI, 8 PLI
5 PLI
8 SLI, 7 PLI
16 SLI
8 SLI, 7 PLI
4 SLI, 8 PLI
9 SLI, 2 PLI
2 PLI
1 SLI, 5 PLI
9 SLI
1 SLI, 9 PLI
1 SLI, 3 PLI
Unaffected
ASD
Autism
Unaffected
ADOS-G
ASD
Autism
12 SLI, 1 PLI
0 SLI, 5 PLI
1 SLI, 0 PLI
1 SLI, 1 PLI
1 SLI, 0 PLI
1 SLI, 2 PLI
2 SLI, 2 PLI
0 SLI, 1 PLI
0 SLI, 8 PLI
All three participants with no ADI-R data (2 SLI and 1 PLI) were
unaffected on ADOS-G. The one participant with SLI who refused
ADOS-G was unaffected according to ADI-R. ADI-R, Autism Diagnostic
Interview Revised; ADOS-G, Autism Diagnostic Observation
Schedule Generic; ASD, autistic spectrum disorder; SLI, specific
language impairment; PLI, pragmatic language impairment.
342
Results
Table II shows the number of participants who fell above cutoffs for autistic disorder in relation to their original subtype
of language impairment. The association between language
subgroup and diagnosis from parental report on ADI-R
was statistically significant (2=16.09, degrees of freedom
[df]=2, p<0.001). The association between language subgroup and diagnosis was again statistically significant when
ADOS-G was the basis for diagnosis (2=9.21, df=2,
p=0.01).
If a strict definition of autism was used, requiring scores
above threshold for autism on both ADI-R and ADOS-G, then
eight of 20 cases of PLI and none of the 18 cases of SLI met the
criterion. In relation to criterion standard consensus
autism diagnosis, this criterion has been shown to give sensitivity and specificity of more than 80% in a US sample.14 A
broader definition, requiring a diagnosis of ASD or autism on
both measures, selected 11 of 20 cases of PLI and two cases of
SLI. For the broadest possible definition, where ASD is diagnosed if ASD criteria are reached on either instrument, then
19 of 20 PLI cases and 6 of 18 SLI cases met the criterion.
However, previous research has shown a significant loss of
specificity when this latter criterion is used, with many nonautistic false positives being included.14
Discussion
A high proportion of people who were regarded as languageimpaired rather than autistic when seen in childhood were
deemed to meet contemporary diagnostic criteria for ASD in
adulthood, on the basis of parental report of childhood
symptoms and/or on the basis of current behaviour. This was
particularly true for those who were judged to have pragmatic impairments in childhood. Most of these individuals were
first seen as children when the diagnostic criteria for autism
came from the Diagnostic and Statistical Manual of Mental
Disorders, 3rd edition (DSM-III) or the revised 3rd edition
(DSM-III-R).15 Both of these schemes, especially DSM-III,
adopted more stringent criteria than are currently used, and
milder forms of autism, currently referred to as ASDs, were
not well recognized. The broadening of diagnostic criteria
for autism is apparent in a recent epidemiological study of
9- to 10-year-old children which found a rate of 24.8 per 10
000 for cases where consensus diagnosis of autism was confirmed by both ADI-R and ADOS-G, rising to 38.9 for all cases
of consensus diagnosis of autism, and to 116.1 per 10 000 for
consensus diagnosis of ASD.16
Diagnostic substitution has previously been studied by
comparing long-term trends in prevalence of autism versus
developmental language disorders in epidemiological data,
but findings have been inconsistent. A UK study using the
General Practice Research Database found a decline in frequency of language disorder that mirrored the increase in
autism,17 but a US study of children enrolled in special education found no such pattern.18 However, it is likely that the
latter study included many children with relatively mild
speech or language difficulties, and this may have masked a
decline in diagnosis of rarer receptive language disorders.
The only other follow-up study of which we are aware that
used algorithms from autism diagnostic instruments with
children diagnosed with developmental language disorder
was performed by Conti-Ramsden et al.19 They studied a
group of children originally recruited from language units
(special classes) at 7 years of age. These children were given
the ADOS-G and their parents were given the ADI-R when the
children were 14 years old. The proportion of participants
who met diagnostic criteria for autism on both ADI-R and
ADOS-G was relatively low (3/76), but in total 11 children
met criteria for autism on ADI-R, and 19 met criteria for ASD
or autism on ADOS-G. These are lower rates than found in
our study, but this could be explained by the fact that most of
the sample were recruited in the early to mid 1990s, after the
publication of DSM-III-R, which used broader diagnostic criteria than DSM-III. In addition, our sample was selected to
include a high proportion of individuals with PLI. ContiRamsden et al. noted that there was no difference on language tests between children who did and did not meet
criteria for an ASD. However, they did not present any data
on childrens pragmatic abilities.
These authors argued against the idea of diagnostic substitution, instead proposing that some children had developed autistic symptoms as they grew older. They noted that
this sample were all definite cases of SLI as late as 7 years of
age, when the stereotypical behaviours and atypical social
skills characteristic of autism would have been visible if they
had been present (p 626). In a similar vein, another study
found autistic-like symptomatology in adulthood in a sample
of individuals originally recruited because of developmental
receptive language disorder in childhood, but the authors
argued that this had developed with age, rather than being
part of the original presentation.20,21 In our sample, this
explanation is most plausible for those participants who did
not show signs of autism according to parental report on the
ADI-R but did score in the ASD or autism range on the ADOSG (three SLI cases and three PLI cases in Table II).
However, for most individuals with autistic symptomatology in the current study, and for just over half of those in the
study by Conti-Ramsden et al., autistic symptoms were evident on parental report on the ADI-R, which uses an algorithm based on behaviour observed at 4 to 5 years of age.
Clearly, one must be cautious about interpreting retrospective reports of early childhood behaviours that are made
some 20 or so years after the event. Furthermore, reports
could be contaminated by parents having read about autism
and thereby developing biased memories of their childs
early development. Nevertheless, we found that parents
often gave vivid and highly specific examples of behaviours
that would lead to a clear coding of abnormality on ADI-R,
despite the fact that nobody had discussed a possible autism
diagnosis with them. Some illustrative vignettes are given in
Appendix I. This provides clear evidence of autistic symptomatology at the time when the children had been diagnosed
with language disorder.
343
were focusing on individuals who had serious communication problems, often associated with poor literacy skills.
Nevertheless, we were able to show that the responders were
representative of the larger pool of potential participants in
terms of language subtype, nonverbal ability, and receptive
language level. The value of this sample is that, although
small in size, it is unique in being well documented in terms
of language characteristics in early childhood, and including
many individuals first seen before the advent of DSM-IV. At
follow-up, we were able to assess ASD symptoms in the
young people themselves and to obtain a retrospective
report of symptoms in childhood from their parents, and
thus demonstrate in individual participants how a given
symptom profile related to changes in diagnostic practices
over time.
Even though rates of autistic behaviours were high in our
sample, especially in those with PLI, our study agrees with
others4,19 in emphasizing the lack of a clear dividing line
between language disorder and autism.
Conclusion
This study provides direct evidence of diagnostic substitution, indicating that many children who were diagnosed with
severe language disorders in the 1980s and 1990s displayed
behaviours that would be regarded as meriting a diagnosis of
ASD according to contemporary criteria. This appears to be a
direct result of changing diagnostic criteria from DSM-III
through to DSM-IIIR and DSM-IV. As noted by Rutter,2 it
would be rash to conclude that an increasing prevalence of
autism is entirely explicable in terms of broadening diagnostic criteria, but the data reported here illustrate how secular
changes in diagnostic concepts and clinical awareness have
led to diagnostic reassignment from language disorder to
autistic disorder. It is likely that similar factors have operated
to lead to a diagnosis of autism in other children who would
hitherto have been regarded as cases of learning disability*
or attention-deficithyperactivity disorder. Our study also
has implications for our evaluation of the research literature
on developmental language disorders. Many studies of children with receptive language disorder that were published
in the last century need to be re-evaluated on the grounds
that they will have included children who would nowadays
be regarded as having ASD.
Accepted for publication 19th December 2007.
Acknowledgements
The authors would like to thank all of the participants who
generously gave up their time to take part in this study. Additional
thanks are due to Courtenay Frazier Norbury for help in tracing
participants. This research was supported by a Programme Grant
from the Wellcome Trust.
References
1. Williams JG, Higgins JPT, Brayne CEG. Systematic review of
prevalence studies of autism spectrum disorders. Arch Dis Child
2006; 91: 815.
2. Rutter M. Incidence of autism spectrum disorders: changes over
time and their meaning. Acta Paediatr 2005; 94: 215.
3. Bishop DVM. Autism, Aspergers syndrome and semanticpragmatic disorder: where are the boundaries? Br J Disord
Commun 1989; 24: 10721.
*North American usage: mental retardation.
344
List of abbreviations
ADI-R
ADOS-G
ASD
PLI
SLI
Aspergers syndrome. His ADI-R scores were 14, 14, and 4 for the
domains of social interaction (autism cut-off 10 or more)
communication (autism cutoff 9 or more) and repetitive behaviour
(autism cutoff 3 or more), respectively. In adulthood, his ADOS-G
total score was 12 for communication and social interaction, above
the autism cutoff of 10.
PARTICIPANT PLI07
345