Social communication de¢cits in conduct disorder: a clinical and community survey
J. Gilmour,1 B. Hill,2 M. Place,3 and D.H. Skuse2 1Sub-Department of Clinical Health Psychology, University College London, UK; 2Behavioural and Brain Sciences
Unit, Institute of Child Health, London, UK; 3Hartlepool CAMHS, UK
Background: Increasing numbers of children are referred to Child and Adolescent Mental Health Services because of disruptive behaviour. Recent reviews on the origins of conduct problems indicate that the most severe and persistent forms are found predominantly among males with a range of neurodevelopmental vulnerabilities, which are likely to have biological substrates. In this study, we tested the hypothesis that many children who are identified with conduct disorder actually have a primary deficit in pragmatic language skills, of a quality and degree that is similar to children on the autistic spectrum. We hypothesised that pragmatic difficulties may underlie the antisocial behaviour in a proportion of children who are labelled as conduct disordered. Methods: Using the Children’s Communication Checklist (Bishop, 1998), we surveyed 142 children who had been referred for clinical investigation, with a predominant diagnosis of either an autistic spectrum condition (n ¼ 87) or conduct disorder (n ¼ 55), and 60 typically developing comparison children. Among children with conduct disorders, males predominated 9:1. Results: On the basis of parent and teacher ratings, two-thirds of those with conduct disorders had pragmatic language impairments and other behavioural features similar in nature and degree to those of children with autism, independent of IQ. In a further study, we surveyed 54 children who had been excluded from elementary schools in a socio-economically disadvantaged inner-London borough and found over two-thirds to have comparable deficits. Conclusions: These findings have both theoretical and practical implications. First, they indicate the presence of communicative problems in a sub-group of children in whom conduct rather than language had been the major concern. Second, they indicate that severe deficits in pragmatic abilities and aut- istic-like behaviours can coexist with psychiatric conditions other than autism, especially in boys. Third, they imply that the management of many disruptive children could profitably be addressed to ameliorating their social and communicative skill deficits. Keywords: Social communication, prag- matics assessment, language impairment, autism, conduct disorder. Abbreviations: ADHD: Attention deficit and hyperactivity disorder; CAMHS: Child and Adolescent Mental Health Services; CCC: Children’s Communication Checklist; DfEE: Department for Education and Employment; 3di: The Developmental, Dimensional and Diagnostic Interview; NUT: National Union of Teachers; PIQ: Performance IQ; PDD: pervasive developmental disorder; SCD: social communication disorder; SLI: Specific Language Impairment; VIQ: Verbal IQ; WHO: World Health Organisation.
Permanent exclusions from schools in England in- creased from 3,000 in 1990–1 to 13,500 in 1996–7, including 1,340 from primary schools in 1998–9, a rate of .03 per cent (DfEE, 2000). Each child exclu- ded from school costs public services over £30,000 (Bagley & Pritchard, 1998). Up to 25,000 children are being educated outside the school environment, and they obtain on average only 10% as much edu- cation as their peers do (Parsons, 1996). Those at highest risk are male, have special educational needs, and live in areas of social deprivation. There is great inter-school variability in reasons given for exclusion but, in general, teachers are unsure of the distinction between poor behaviour and behaviour that reflects an underlying problem requiring treat- ment (Social Exclusion Unit Report, 1998). The management of children at risk of exclusion de- mands a framework for assessment and treatment that does not exist at present.
Many children with disruptive behaviour are known to have problems with social understanding (e.g., Milch-Reich, Campbell, Palham, Connelly, &
Geva, 1999). Previous work has focused largely on their misinterpretation of other’s intent, especially attributional biases (Crick & Dodge, 1996; Matthys, Cuperus, & Van Engeland, 1999). These deficiencies and biases are often present from early childhood, and reasons for their development have usually been couched in terms of social learning theory. Whilst this may be true for a proportion, we aimed to test the hypothesis – based on our pilot work and clinical experience – that a sub-sample of these children have a deficit in social communication abilities, including pragmatic skills. Evidence has been accumulating over many years of an association between conduct disorders and deficits in language- based verbal skills, which persist even after con- trolling for potentially confounding variables such as race and socioeconomic status (Hill, 2002). A specific association with deficits in pragmatic skills, al- though suspected, has never previously been invest- igated systematically.
Pragmatics may be defined as the appropriate use and interpretation of language in relation to the
Journal of Child Psychology and Psychiatry 45:5 (2004), pp 967–978
� Association for Child Psychology and Psychiatry, 2004. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA
context in which it occurs (Bishop, 1997). Some children with severe pragmatic skill deficits also possess autistic characteristics. In short, there ap- pears to be blurring of boundaries between deficits in pragmatic skills, broader deficits in social commu- nication, and disorders on the autistic spectrum. We will describe this clinical profile in general terms as a social communication deficit.
Based on our clinical experience, we predicted that pragmatic deficits (and perhaps broader social communication deficits) would commonly be found among children with conduct disorders. We were particularly concerned about the social commun- ication skills of children excluded from mainstream education, and those at risk of exclusion, in the early years at school. Specific and general cognitive defi- cits have been described among children with dis- ruptive behaviour. Moffitt, Caspi, Rutter, and Silva (2001) reviewed findings relevant to sex differences in antisocial behaviour, identified from a population birth-cohort followed up from 3 to 21 years (the Dunedin study). The authors discuss the relevance of neurodevelopmental problems to the origins of persistent and severe antisocial behaviour, which affects males more frequently than females. These risk factors are couched in terms of neuro-cognitive deficits (visuospatial as well as verbal in nature), undercontrolled temperament, a personality trait they term ‘weak constraint’, and hyperactivity. Weak constraint implies that the individual has low self- control, is unconcerned about harm-avoidance, and has little regard for traditional values such as high moral standards and a conservative social environ- ment. These personality descriptors may be more applicable to individuals in late adolescence than to children in elementary school, but their origins probably lie in temperamental variables that could be measured in early life. They do not simply result from acculturation. Family risk factors cannot, as Moffitt et al. (2001) rightly point out, account for the great preponderance of antisocial behaviour among males. They do not discuss the relevance to the onset of antisocial behaviour of autistic-like deficits in so- cial cognitive skills, but they do acknowledge that the male preponderance in pervasive disorders such as autism may share the same neurodevelopmental explanation as a ‘relatively rare’ form of early onset, persistent antisocial behaviour.
Aggression and ‘insolent’ and ‘uncooperative’ behaviour with adults are the most common reasons for exclusion from school (National Union of Teachers (NUT), 1992). We hypothesised that such behavioural problems could be linked in some cases to social communication deficits. Children with such deficits do not tend to use language in a way that takes account of a social hierarchy (peer, teaching assistant, headteacher). Consequently, their failure to treat school staff with due deference makes them appear to be ‘insolent’. They may publicly point out their teacher’s mistakes in class. Children who make
such errors are not necessarily aware of them; typ- ically they are bewildered by the adverse reaction of teachers. The fact that the underlying reason for their solecism is not identified means they are only too likely to repeat it.
Significant numbers of children with disorders on the autistic spectrum remain undetected in the general population. Characteristically, they have undue difficulty with social aspects of communica- tion, in both verbal and non-verbal modes. They lack social reciprocity, and restricted interests, social isolation and repetitive rituals characterise a pro- portion. There is an increasing recognition of the extent of the problem by child health professionals, both in the UK (General Practice Research Database; Kaye, del Mar Merelo-Montes, & Jick, 2001) and in the USA (California Department of Developmental Services; Fombonne, 2001). Even so, the number coming to clinical attention is significantly lower than the true population prevalence (2–5/1000), as estimated from community surveys that have actively sought out cases (Baird et al., 2000; Fombonne, 1999; Gillberg & Wing, 1999; see Fombonne, 2001). Patterns of comorbidity have not been investigated, largely because standardised case-finding psychiatric interviews are not suitable for this purpose.
We hypothesised that pragmatic difficulties would be associated with antisocial behaviour in a pro- portion of children who are labelled as conduct dis- ordered. There were two phases to our investigation. In Phase One, we aimed to assess the pragmatic competence of children who had been referred to child and adolescent mental health services with conduct disorders. We compared them with children whose primary problem was a disorder on the aut- istic spectrum. In Phase Two, we asked the same question of a sample of inner-city children who had been excluded from school, or who were at high risk of exclusion, but who had not been referred to a clinical service.
Phase One: Participants referred to Child and Adolescent Mental Health Services
All the children in Phase One were clinically referred cases. There were two clinic populations from which children with conduct or pervasive developmental dis- orders were selected. The first sample consisted of all new referrals during the period January 1999 through October 2001 to the Social and Communication Disor- ders Clinic at Great Ormond Street Hospital (GOSH), London. This is a tertiary referral centre, which spe- cialises in the evaluation of children with behavioural problems that are thought to have a neurodevelop- mental origin. Whilst most of these children are sus- pected of having an autistic-like condition, a minority are referred because of persistent behaviour problems of a less specific nature. Referrals are sent from a wide
968 J. Gilmour et al.
geographical area, throughout the southeast of Eng- land, and are heterogeneous with regard to socioeco- nomic background. Those with global learning difficulties, according to clinical evaluation and/or psychometric evaluation (a Full Scale IQ of 70 or below), were excluded, as were children who did not possess sufficient language skills to enable a formal test of their pragmatic abilities to be completed. Cognitive ability was assessed in a proportion of children who attended the clinic at Great Ormond Street Hospital. While IQ data are not available on the whole sample, the sub- group on whom data exist is representative as it was a consecutive series of clinic attendees. Cognitive ability data were not routinely collected in the Sunderland sample. Wherever possible, IQ was assessed using the Wechsler Intelligence Scales for Children (Wechsler, 1992) or Verbal IQ was estimated using the British Picture Vocabulary Scales (Dunn, Dunn, Whetton, & Burley, 1997).
The second clinical population comprised a con- secutive series of children with miscellaneous dis- orders, who had been referred to clinics providing Child and Adolescent Mental Health Services in Sunderland, a city in the north of England with a population of ap- proximately 300,000. According to the Government In- dex of Deprivation 2000, approximately 50% of wards in Sunderland rank in the lowest 10% nationally in terms of multiple indices of income, unemployment, health, education, housing, access to services and child poverty (Department of Transport, Local Government and the Regions, 2000). Exclusion criteria were similar to those in the GOSH sample.
Typically developing comparison children were recruited from paediatric outpatient clinics in Sunder- land. A consecutive clinical sample was chosen which was similar to the psychiatric clinic attendees in terms of age and socioeconomic status. These children were referred for physical complaints such as asthma but were subject to the same structured psychiatric inter- view evaluation as the case children; thus children with a psychiatric condition were excluded from the com- parison group.
Diagnostic procedures. In both the GOSH and the Sunderland sample, diagnoses were made primarily by clinical judgement, by clinicians who were experienced in making diagnoses within the autistic spectrum. The data on which such judgements were based were col- lected by a computerised interview, which has good psychometric properties in terms of both reliability and validity (Skuse et al., 2004). Diagnoses were assigned according to ICD-10 criteria (World Health Organiza- tion, 1996). Identical protocols of assessment and the application of diagnostic criteria were applied to the Great Ormond Street Hospital sample (N ¼ 103) and the Sunderland sample (N ¼ 39).
Eligible cases were subdivided according to their primary diagnosis into the following groups:
i) Conduct disorder (49 boys, 6 girls). The diagnosis of conduct disorder excluded the category of ‘conduct disorder confined to the family context’. Our dia- gnostic criteria required evidence of behavioural disturbance at home and at school. Information concerning the latter context was obtained from
teachers, on the basis of a score ‡4 on the Strengths and Difficulties Questionnaire (Goodman, 1997), followed by interviews where the diagnosis was suspected. We have subsumed both socialised and unsocialised conduct disorders under this heading. We also subsumed ODD within the category ‘con- duct disorder’ for the purposes both of simplifying the analysis and increasing group size.
ii) Autistic Spectrum Disorder (34 boys, 8 girls). These children met diagnostic criteria on two of the three domains of the triad of autistic impairments, according to the Multiaxial Classification of Child and Adolescent Psychiatric Disorders (WHO, 1996).
iii) Autism (40 boys, 5 girls). This group included chil- dren clinically described as both ‘Asperger syn- drome’ and autism, according to ICD-10 criteria. There is little evidence (e.g., Miller & Ozonoff, 2000) that there are any meaningful differences between high-functioning autistic and Asperger’s groups.
iv) Typically developing comparisons (29 girls, 31 boys).
Comorbidity. Some children had more than one dia- gnosis. The primary diagnosis was assigned following team discussions of all clinical material. This was de- fined as the condition that dominated the clinical pic- ture, and it was usually the condition associated with the symptom profile that prompted referral. In some cases one or more additional psychiatric diagnoses were made. Diagnostic evaluations for comorbidity in association with a primary diagnosis of conduct disor- der were based on clinical judgement and ICD-10 guidelines, using as a data source information gathered by means of a standardised psychiatric interview (the 3di; Skuse et al., 2004). Data describing comor- bidity were available for the GOSH clinical samples only. Comorbidity rates in the CD (n ¼ 29) group were as follows: none of the CD group showed evidence of generalised anxiety disorder, phobias, panic disorder, depression, hypomania or bipolar affective disorder. One child (3%) had a separation anxiety disorder and 6 (20%) had hyperkinetic disorder. In the Autism group (n ¼ 34) there was no comorbidity for generalised anxiety disorder, phobia, panic disorder, hypomania, bipolar affective disorder or eating disorders. One child (3%) had separation anxiety disorder, three (9%) had a depressive disorder and 4 (12%) had hyperkinetic dis- order. None of the children in the ASD group (n ¼ 35) showed evidence of generalised anxiety disorder, panic disorder, depression, hypomania, bipolar affective dis- order or eating disorders. Two children (6%) had fea- tures of separation anxiety disorder, two (6%) had a specific phobia, and 6 (17%) had had hyperkinetic disorder.
Phase Two: Participants identified from community survey of excluded children
Children were identified with the assistance of the Local Educational Services in the London Borough of Hack- ney. Hackney Borough has a population of nearly 200,000 of which almost a quarter is under 16 years of age. A third of the borough’s population is from ethnic minority groups. Several key contributory factors to antisocial behaviour and crime, such as high unem-
Social communication deficits in conduct disorder 969
ployment, family breakdown, low levels of education and high levels of disaffected young people, are wide- spread in Hackney. According to the Government Index of Deprivation (Department of Transport, Local Gov- ernment and the Regions, 2000), Hackney rates as the second most deprived Borough in the country. All wards are in the worst 10% nationally and nine wards are in the worst 3%. Other indicators of social deprivation in the Borough show that 58% of residents live in social housing; 50% of all children receive free school meals, and long-term illness and disability are the highest in London. Hackney experiences high levels of mental health problems and has amongst the highest admis- sions rates to psychiatric hospitals in the country.
Our sample of antisocial children from this source comprised children between the ages of 5 and 10 years old who had been excluded or were at risk of exclusion from schools in the Borough in June 2001. Of the 57 primary schools in the Borough we received information from 51; 16,000 pupils who were within this age range attended school in the Borough. The total of children in the age range excluded from school on whom we re- ceived information comprised 5 girls and 49 boys at that time. So far as we could ascertain, none was attending CAMHS services at the time of the survey. No diagnostic interviews have been conducted with parents in this sample. We were unable to identify the total number of children in the Borough who were at risk of exclusion at the time of this survey.
Assessment of pragmatic competence
The Children’s Communication Checklist (CCC; Bish- op, 1998) was used for the purpose of assessing prag- matic skills in both the clinical and the community samples. This instrument was developed with the spe- cific goal of distinguishing between those children who have impairments in their social use of language, or pragmatic skills, and those with Specific Language Impairment (SLI), where the principal problems are with language structure. Pragmatic competence is intrinsically dependent on the specific situation in which it is assessed, whereas structural language problems are pervasive. For each item, the rater is presented with a statement and has to check whether this ‘definitely applies’, ‘applies somewhat’, or ‘does not apply’.
The general population scores from Bishop and Baird’s (2001) typically developing group and the maximum score are given as follows for each subscale. Speech (mean 35.1 SD 1.5; maximum score: 38) measures intelligibility and fluency, concerning matters such as whether speech is clearly articulated and flu- ent. Syntax (31.7 SD .7; maximum score: 32) measures the child’s ability to produce developmentally appro- priate length of utterances in a grammatically correct manner. Inappropriate Initiation (27.2 SD 2.1; max- imum score: 30) is a subscale that measures impulsive behaviour, such as interrupting conversations. The Coherence (35.2 SD 1.3; maximum score: 36) subscale measures the child’s ability to tell a story, or to talk about past or future events in an appropriate temporal context, aimed at increasing intelligibility to the lis- tener. Stereotyped Language (28.0 SD 2.1; maximum score: 30) measures a tendency to engage in conversa-
tions that are allied to the interests of the child, con- taining favourite but inappropriate themes, and stereotyped phrases. Use of Context (30.5 SD 1.9; maximum score: 32) assesses different aspects of the use of conversational context, such as the child’s understanding of sarcasm or non-literal remarks. In general, children with poor understanding of contextual cues will interpret another’s remarks over-literally. Rapport (32.8 SD 1.4; maximum score: 34) describes a child’s ability to start a conversation, use gestures to facilitate meaning, interpret non-verbal gestures and facial expressions, and to use eye contact appropriately. Social Relationships (32.7 SD 1.9; maximum score: 34) concerns the child’s ability to make and to maintain friendships. There are questions specifically targeted at identifying children who are loners, or perceived as odd by others, or who are deliberately aggressive. The Interests subscale (31.5 SD 2.1; maximum score: 34) concerns aspects of social behaviour as well as a tend- ency to have overriding specific interests (such as dinosaurs). It also records unusual repertoires of fac- tual knowledge, including obscure words. The Prag- matic Composite (153.6 SD 6.5; maximum score: 162) score is derived from the following subscales: Inappro- priate Initiation, Coherence, Stereotyped Language, Use of Context and Rapport.
Lower scores on the CCC scales indicate greater impairment. A child who obtains a profile indicating average or typical development, with no items describ- ing communication difficulties scored as applying ‘somewhat’ or ‘definitely’, would score 30 on each sub- scale. Scores greater than 30 are possible for those scales that include positive items that describe com- municative strengths. In consultation with Dorothy Bishop, we defined significant clinical impairment to be associated with CCC scores that lay at least 2.0 SD below the population mean, on subscales or the Com- posite score. Population data were determined from data provided by Bishop and Baird (2001). Most items are based on clinical descriptions of a subtype of Spe- cific Language Impairment known as semantic-prag- matic disorder (Bishop & Rosenbloom, 1987). Bishop (1998) predicted that children whose problem was purely a semantic-pragmatic one might be identified by a poor Pragmatic Composite total score, but their Social Relationships and Interests subscale scores would be normal. Children with autistic features in addition to a semantic-pragmatic disorder would obtain abnormal scores on the latter subscales as well.
Specificity of the CCC
In order to investigate the prevalence of pragmatic dis- orders in conditions other than conduct disorder and autism spectrum disorders, we evaluated scores ob- tained from parent ratings of other clinically identified samples, diagnosed according to ICD-10 criteria, from the Sunderland sample of consecutive referrals to CAMHS. We found no evidence that pragmatic skill deficits are a non-specific correlate of significant psy- chiatric disorders in samples of children of approxi- mately the same mean ages and abilities. The mean Pragmatic Composite score for our typically developing group (N ¼ 60) was 151.9 (SD 8.7). Comparable scores were found for children with moderate depression
970 J. Gilmour et al.
(N ¼ 6; mean 151.7; SD 6.8); severe depression (N ¼ 5; mean 156.8; SD 2.6); and generalised anxiety (N ¼ 14; mean 150.6; SD 12.9).
Phase One: CCC ratings and diagnostic categories for clinically referred sample
Table 1 shows parent CCC ratings in relation to diagnostic category for the clinical samples and the comparisons. Table 2 shows corresponding data for teacher ratings from the clinically referred sample. Each table gives the proportion of children in each diagnostic group with scores more than 2 SD below the CCC subscale (and Pragmatic Composite) popu- lation means (data from Bishop & Baird, 2001). Such scores are described as being in the ‘clinical range’, indicating that they are sufficiently low that they would be typical of children requiring clinical inter- vention (Bishop, 1998). Children in the diagnostic categories ‘Conduct Disorder’, ‘Autism Spectrum Disorder’ and ‘Autism’ are compared with a typically developing comparison sample. The main effect of diagnostic group was tested in analyses of variance for differences in means, based on each CCC subscale
and the Pragmatic Composite score. We undertook an analysis of variance, with Bonferroni corrections to avoid the possibility of type 1 errors. The highly sig- nificant main effects of group, according to data on communication skills obtained from parent-rated CCC (see Table 1), were largely attributable to the contrasts between typically developing and clinically referred children (irrespective of whether they were in the conduct disordered, autistic spectrum or autistic groups). We did not have a normal comparison group for teacher data (see Table 2). There were no group differences in pragmatic skills, by diagnostic cat- egory. Age was not significantly related to CCC sub- scale score in the typically developing group (r values ranged between –.16 and .19). There was no signific- ant gender difference in the typically developing group on any of the subscale scores of the CCC (p < .05) (Speech t ¼ 1.76. Syntax t ¼ .76, Inappro- priate Initiation t ¼ 1.1, Coherence t ¼ .18, Stereo- typed Speech t ¼ .71, Context t ¼ .07, Rapport t ¼ .98, Social Relationships t ¼ .11, Interests t ¼ .65 and Pragmatic Composite t ¼ .10) according to parent report. We also re-examined the comparisons de- scribed in Table 1 using an exclusively male sample across all four groups. These results are detailed further in the table.
Table 1 Parental Children’s Communication Checklist ratings in relation to diagnostic category
Diagnostic group ANOVA (Bonferonni-
corrected)+ p < .05
v2 p > .05�
A – Conduct Disorder (n ¼ 55)
B – Autistic Spectrum
Disorder (n ¼ 42) C – Autism (n ¼ 45)
D – Typically developing (n ¼ 60)
Mean age in yrs mths (SD) 10.2 (2.70) 9.7 (2.7) 10.2 (3.7) 10.1 (2.4) ns % male 89% 81% 89% 51% Speech Mean (SD) 32.1 (4.0) 29.9 (5.5) 30.9 (4.9) 34.8 (1.6) D > ABC % in clinical range� 40% 55% 53% 10% ns Syntax Mean (SD) 30.8 (1.8) 30.3 (1.9) 29.6 (3.9) 31.5 (1.3) D > C % in clinical range� 42% 52% 47% 18% ns Inappropriate Initiation Mean (SD) 23.2 (3.4) 22.3 (2.9) 24.5 (3.9) 27.2 (2.7) D > ABC % in clinical range� 54% 63% 40% 12% ns Coherence* Mean (SD) 28.9 (4.2) 26.7 (4.6) 27.1 (4.2) 34.4 (2.6) D > ABC % in clinical range� 78% 89% 89% 21% ns Stereotyped Language* Mean (SD) 22.9 (4.2) 20.8 (4.1) 23.2 (4.3) 26.9 (2.3) D > ABC
C > B % in clinical range� 68% 82% 56% 15% B >> C Use of Context* Mean (SD) 22.8 (3.7) 21.5 (3.5) 22.9 (3.7) 29.9 (2.7) D > ABC % in clinical range� 88% 97% 86% 14% ns Rapport* Mean (SD) 26.2 (4.2) 25.9 (4.1) 24.5 (3.1) 33.2 (1.3) D > ABC % in clinical range� 83% 84% 97% 4% ns Social Relationships� Mean (SD) 25.5 (4.1) 24.1 (4.4) 23.9 (4.6) 32.6 (2.4) D > ABC % in clinical range� 82% 90% 87% 7% ns Interests� Mean (SD) 29.3 (2.8) 28.4 (2.5) 27.6 (2.7) 31.1 (2.4) D > ABC
A > C % in clinical range� 24% 40% 46% 8% C >> A Pragmatic Composite Mean (SD) 127.3 (16.4) 117.29 (2.5) 124.5 (15.1) 151.9 (8.7) D > ABC % in clinical range� 78% 95% 89% 8% B >> A
*Contributes to the Pragmatic Composite score. �Subscales considered by Bishop (1998) to distinguish children with autism from those with ‘pure’ pragmatic disorder. �Proportion of children with scores at least 2 SDs below typically developing mean were considered to have scores in the same range as those seen for clinical evaluation in specialised language units (Bishop, 1998). The TD group was excluded from these analyses. +These analyses were similar using a male only sample in the same sub-groups, with three exceptions: Coherence: A > C. The C > B finding in Stereotyped Language and the A > C in Interests finding no longer hold.
Social communication deficits in conduct disorder 971
We tested the hypothesis that the Pragmatic Com- posite score was related to intellectual ability. In our clinical sample, IQ data from the Wechsler (1992) scales were available for 48 children on whom we had parental CCC information and 41 on whom we had teacher-reported CCC data. Thus it was possible to look separately at relationships between Verbal (VIQ) and Performance IQ (PIQ), and pragmatic competence for each group. In these analyses, age was not entered as a covariate, because neither the parent nor the teacher Pragmatic Composite score correlated significantly with age. Neither VIQ nor PIQ was significantly related to the parent-rated Prag- matic Composite (VIQ: for the Autism group: r ¼ (9) ¼ .35, ns, ASD group r (14) ¼ .26 ns and CD group r (25) ¼ .16, ns. PIQ: for the Autism group: r (9) ¼ ).05, ns, for the ASD group r (14) ¼ .26, ns and CD group r (25) ¼ ).02, ns). The Pragmatic Composite score rated by teachers was not signific- antly correlated with either VIQ or PIQ (VIQ: for the Autism group r (7) ¼ ).09, ns, for the ASD group r ¼ (12) ¼ ).17, ns, and for the CD group r (22) ¼ .21, ns. PIQ: Autism group r (7) ¼ ).62, ns, for the ASD group r ¼ (12) ¼ ).26, ns and for the CD group r (22) ¼ .08, ns. This finding contrasted with the positive association between these variables re- ported by Bishop and Baird (2001).
We had both teacher and parent CCCs for 84 children who were drawn from the clinically identi- fied samples only. Within this group as a whole, the correlation between teacher and parent scores on the Pragmatic Composite was only .18, which failed to
reach a conventional level of significance. In con- trast, teachers and parents did agree reasonably well on specific aspects of speech and language disorder as measured by CCC subscales. These included Speech (r ¼ .56, p < .001) Syntax (r ¼ .57, p < .001), and Coherence (r ¼ .38, p < .001). They were less likely to agree about whether the child used Con- textual Cues appropriately (r ¼ .06, n.s.) or whether a child’s speech was Stereotyped (r ¼ .22, p < .05). Good agreement was found for Interests (r ¼ .41, p < .001) and Inappropriate Initiation (r ¼ .36, p < .001), but there was poor agreement on Social Relationships (r ¼ .16, ns) and Rapport (r ¼ .11, ns).
We had expected agreement in Pragmatic com- posite scores rated by teachers and parents to differ by diagnosis. For example, distinct impairments in pragmatic skills should be obvious in cases of aut- ism, whether the rater was a parent or a teacher. In fact, the value of the correlation coefficient was very similar within each of the four diagnostic categories. In general, teachers tended to rate children with disorders on the autistic spectrum as less language- disordered than did their parents. This is clear from a comparison of the proportions of children who obtain scores that are more than 2 SDs below the population mean. Further analyses indicates there was 60% agreement between parent and teacher in their allocation of individual children to the clinical range status on the Pragmatic Composite scale, 58% agreement for Speech, 73% agreement for Syntax, 77% agreement for Coherence, 78% agreement for Use of Context, 55% agreement for Stereotyped
Table 2 Teacher CCC ratings in relation to diagnostic category
Diagnostic group ANOVA (Bonferonni-
corrected) p < .05
v 2 p < .05
A – Conduct Disorder (n ¼ 29)
B – Autistic Spectrum
Disorder (n ¼ 31) C – Autism (n ¼ 24)
Mean age Mths (SD) 120 (33) 109 (32) 125 (50) ns % male 92% 84% 87% Speech Mean (SD) 32.0 (3.3) 32 (5.5) 30.3 (4.7) ns % in clinical range� 46% 41% 58% ns Syntax Mean (SD) 30.7 (2.3) 30.8 (1.08) 29.8 (2.3) ns % in clinical range� 31% 29% 54% ns Inappropriate Initiation Mean (SD) 25.5 (3.4) 24.7 (3.4) 26.6 (3.3) ns % in clinical range� 31% 39% 17% ns Coherence* Mean (SD) 29.3 (4.9) 29.1 (4.9) 28.1 (4.6) ns % in clinical range� 76% 74% 83% ns Stereotyped Language* Mean (SD) 25.2 (3.9) 23.9 (4.1) 24.8 (3.6) ns % in clinical range� 31% 58% 46% ns Use of Context* Mean (SD) 25.0 (3.5) 24.5 (3.8) 24.1 (6.0) ns % in clinical range� 79% 80% 67% ns Rapport* Mean (SD) 26.8 (4.4) 27.6 (4.2) 26.1 (3.5) ns % in clinical range� 79% 68% 82% ns Social Relationships Mean (SD) 26.3 (4.1) 25.7 (4.3) 25.1 (5.1) ns % in clinical range� 69% 80% 74% ns Interests Mean (SD) 28.5 (3.3) 29.4 (2.6) 27.3 (2.8) ns % in clinical range� 31% 23% 48% ns Pragmatic Composite Mean (SD) 131.7 (15.1) 129.9 (15.1) 130.1 (13.1) ns % in clinical range� 69% 68% 77% ns
*Contributes to the Pragmatic Composite score. �Proportion of children with scores at least 2 SDs below typically developing mean.
972 J. Gilmour et al.
Speech, 60% agreement for Interests, 64% agree- ment for Inappropriate Initiation, 70% agreement on Social Relationships and 71% agreement for Rapport.
Table 3 provides data for both the parent- and teacher-rated Pragmatic Composite, in specified ranges of scores, by diagnostic group. Scores are divided into approximately 1.0 SD categories below the population mean (Scores in band 133–139 are >1.0 SD below mean, 123–132 are >2.0 SD, and those less than 122 are >3.0 SD. The values we have taken are conservative, and derived from our own sample of typically developing children and com- parable data collected by Bishop (Bishop, personal communication). With very low Pragmatic Composite scores on the CCC (below 122) there is a trend for parents and teachers to disagree, insofar as parents are more likely to give lower ratings. Nevertheless, the proportion of children with conduct disorders in this lowest banding is very similar to the proportion of children with autism scoring in that band, whether the rating is made by teachers or by parents. Parent ratings indicate about 50% of clinically identified conduct disordered children are in this lowest category, compared with 51% of those with childhood autism. Teachers rate 27% of autistic children in the lowest category, compared with 34% of those with clinically defined conduct disorder.
Comorbidity. We specifically examined the extent to which evidence existed of comorbid conduct disorder in children assigned diagnoses of autism or an aut- ism spectrum disorder. We looked separately at co- morbidity between these conditions in the two clinical samples referred to the Social Communica- tion Disorders Clinic and the CAMHS service in Sunderland. The proportions of conduct disordered children who were clinically significantly impaired on two of the three aspects of the autism triad (i.e., had an autism spectrum disorder) were 38% (13 of 34) in the Social Communication Disorders Clinic and 19% (4 of 21) in the CAMHS services, a non- significant difference (v2 ¼ 2.2, ns, effect size ¼ .20). Table 4 compares the CCC subscale scores for the
CD group comorbid for an ASD and without. As expected, in general, parents and teachers reported a trend for the ASD affected group to have lower scores (poorer functioning). The same table also shows the proportion of children who fall into the clinical range according to parents and teachers.
Phase Two: Children excluded from school for disruptive behaviours
In Phase Two we selected the conduct-disordered sample of children from elementary schools in the London Borough of Hackney, on the basis that they were so disruptive they were at risk of permanent exclusion from mainstream education. In Table 5, Teacher CCC ratings for the 54 excluded children are given for individual subscales, showing the propor- tions of those children rated as having a degree of pragmatic disorder that falls more than 2.0 SDS below the population mean (see p. 8) (Bishop & Baird, 2001). In the Hackney excluded children sample, the proportion with a Pragmatic Composite score which is >2.0 SDs below the mean is 69%. In the clinically ascertained sample of conduct disor- dered children, the equivalent proportion as rated by teachers is 48% (scores less than 132 are equivalent to )2.0 SDS), and the proportion with low scores rated by parents (total N ¼ 55) is 65%. The equival- ent proportions in the autism spectrum disorder sample are 55% and 88% respectively. These pro- portions do not differ significantly from one another.
We tested the hypothesis that there would be similar mean scores in respect of each of the CCC subscales and the pragmatic composite for the clin- ical and the community samples, in respect of con- duct disordered children. Teacher ratings on the CCC for the 54 excluded children are reported in Table 5, and should be compared with equivalent data for the clinically referred sample, shown in Table 2. We undertook a further set of analyses, comparing the teacher CCC scores for the excluded children with those in the diagnostic categories shown in Table 2. The mean values for the CCC subscales and the Pragmatic Composite score are
Table 3 Numbers (percentage) of children in relation to range of score on the Pragmatic Composite Score of Children’s Com- munication Checklist. Parent and teacher ratings
Parent rating A – Conduct Disorder
(N ¼ 55) n (%) B – Autistic Spectrum Disorder (N ¼ 42) n (%)
C – Autism (N ¼ 45) n (%)
D – Typically Developing (N ¼ 60) n (%)
140 or more 12 (22%) 2 (5%) 5 (11%) 56 (93%) 133 to 139 7 (13%) 3 (7%) 7 (16%) 2 (3%) 123 to 132 8 (14%) 6 (14%) 10 (22%) 1 (2%) Below 122 28 (51%) 31 (74%) 23 (51%) 1 (2%)
Teacher rating A – Conduct Disorder
(N ¼ 29) n (%) B – Autistic Spectrum Disorder (N ¼ 31) n (%)
C – Autism (N ¼ 24) n (%)
140 or more 9 (31%) 10 (32%) 6 (27%) 133 to 139 6 (21%) 4 (13%) 3 (14%) 123 to 132 4 (14%) 5 (16%) 7 (32%) Below 122 10 (34%) 12 (39%) 6 (27%)
Social communication deficits in conduct disorder 973
Table 4 Parent- and teacher-rated Children’s Communication Checklist (CCC) scores for the Conduct Disordered (CD) group comorbid with Autism Spectrum Disorders (ASD), compared to remainder
Parent rating Teacher rating
CD minus ASD (CD)) (N ¼ 38) Mean (SD)
CD plus ASD(CD+) (N ¼ 17) Mean (SD)
ANOVA (Bonferonni- corrected)+ P < .05
v 2 p > .05
CD minus ASD (CD)) (N ¼ 18) Mean (SD)
CD plus ASD (CD+) (N ¼ 11) Mean (SD)
ANOVA (Bonferonni- corrected) P < .05
v2 p > .05
Speech 32 (3.9) 30 (3.9) TD > CD), CD+ 32 (3.5) 31 (2.8) ns % in clinical range� 32% 59% CD+ > CD) 38% 60% ns Syntax 31 (1.5) 30 (2.2) ns 31 (2.1) 30 (2.5) ns % in clinical range� 48% 35% ns 28% 36% ns Inappropriate Initiation* 23 (3.3) 24 (3.5) TD > CD), CD+ 26 (3.1) 25 (3.9) ns % in clinical range� 60% 42% ns 28% 36% ns Coherence* 30 (3.7) 28 (4.8) TD > CD), CD+ 30 (5.1) 28 (4.2) ns % in clinical range� 77% 79% ns 67% 90% ns Stereotyped Language* 23 (4.5) 23 (3.5) TD > CD), CD+ 26 (2.6) 23 (5.1) ns % in clinical range� 65% 72% ns 17% 54% CD+ > CD) Use of Context* 23 (2.6) 22 (3.7) TD > CD), CD+ 26 (3.7) 23 (2.5) ns % in clinical range� 84% 93% ns 73% 91% ns Rapport* 26 (4.2) 23 (3.7) TD > CD), CD+ 28 (4.2) 25 (4.1) ns % in clinical range� 77% 93% ns 67% 100% CD+ > CD) Social Relationships� 26 (4.1) 23 (3.3) TD > CD), CD+ 27 (4.2) 24 (3.0) CD) > CD+ % in clinical range� 76% 94% ns 50% 100% CD+ > CD) Interests� 29 (2.7) 28 (2.7) TD > CD), CD+ 30 (1.8) 27 (4.1) CD) > CD+ % in clinical range� 21% 29% ns 11% 63% CD+ > CD Pragmatic Composite 130 (16.8) 121 (13.9) TD > CD) > CD+ 135 (14.7) 124 (13.6) ns % in clinical range� 44% 65% CD+ > CD) 61% 82% ns
*Contributes to the Pragmatic Composite score. �Subscales considered by Bishop (1998) to distinguish children with autism from those with ‘pure’ pragmatic disorder. +ANOVA comparison with TD group. �Proportion of children with scores at least 2 SDS below typically developing mean.
9 7 4
J . G ilm
o u r e t a l.
very close indeed in the two samples. There were just two significant differences between the excluded children and others, on the basis of CCC scores. On both the Speech and Coherence subscales ‘excluded’ children obtained higher (indicating better func- tioning) scores than children with autism, but they did not differ significantly from clinically referred conduct disordered children, nor from autistic spectrum disorders.
Social communication deficits in children with conduct disorders
In this investigation we have shown that a substan- tial proportion of children with conduct disorders, in both clinically referred and community-ascertained samples, have deficits in their pragmatic skills that are as severe as those of children with clinical dia- gnoses on the autistic spectrum. Our data show that a sub-set of children presenting as CD actually have an unidentified ASD. There are still further children with CD who do not reach a formal ASD diagnosis but who nonetheless have pragmatic problems. It may be that pragmatic problems we have described in the CD group have the same origins as ASD but this remains untested. The primary neuro-cognitive impairments in children with pragmatic problems remain poorly understood, though executive
dysfunction (e.g., McDonald & Pearce, 1996) is one likely candidate. For these high-functioning chil- dren, their pragmatic difficulties are not simply a function of intelligence; as there is no significant correlation between these abilities and either Verbal IQ or Non-Verbal IQ in our study. While we acknowledge we have measured cognitive ability on only a sub-sample of children, we are confident it is representative of the whole clinic population as it was a consecutive series of children from the GOSH clinic.
We considered the possibility that a systematic bias may have been introduced in the clinical sam- ples, insofar as parents who brought their children to the CAMHS service in Sunderland, or to the service at Great Ormond Street Hospital (which specialises in treating children with social commu- nication problems), may have exaggerated their child’s pragmatic deficits. This seems unlikely. First, the profile of CCC subscale and pragmatic composite scores for both conduct-disordered and other dia- gnostic groups was similar in both parent- and teacher-rated data. Second, teacher ratings of the conduct-disordered excluded sample of children from inner-city schools (who had not attended CAMHS services, so far as we could ascertain) yiel- ded almost identical mean scores for CCC subscales and the Pragmatic Composite to the clinically re- ferred samples. Third, although Great Ormond Street Hospital (GOSH) is a Tier 4 service, there were
Table 5 Teacher CCC ratings for school excluded sample
F – Children excluded from Hackney schools* (n ¼ 54)
Significant difference from groups A-C on Table 2 (Bonferroni-corrected p < .05)
Age yrs.mths Mean (SD) 9.4 (2.0) ns Sex ratio (M:F) 49:5 Speech Mean (SD) 32.2 (3.7) F > C % in clinical range� 44 Syntax Mean (SD) 30.4 (4.0) ns % in clinical range� 33 Inappropriate initiation Mean (SD) 24.5 (3.5) ns % in clinical range� 35 Coherence Mean (SD) 30.5 (4.4) F > C % in clinical range� 72 Stereotyped Language Mean (SD) 26.6 (5.0) ns % in clinical range� 28 Use of Context Mean (SD) 24.7 (4.0) ns % in clinical range� 78 Rapport Mean (SD) 27.7 (4.2) ns % in clinical range� 67 Social Relationships � Mean (SD) 23.8 (3.6) ns % in clinical range� 94 Interests� Mean (SD) 29.5 (3.0) ns % in clinical range� 20 Pragmatic Composite Mean (SD) 131.5 (14.2) ns % in clinical range� 69
*Two children from Hackney school exclusion children sample were omitted from analysis because their quality of language was insufficient to compute a Pragmatic Composite score. �Subscales considered by Bishop (1998) to distinguish children with autism from those with ‘pure’ pragmatic disorder. �Proportion of children with scores at least 2 SDs below typically developing mean were considered to have scores in the same range as those seen for clinical evaluation in specialised language units (Bishop, 1998).
Social communication deficits in conduct disorder 975
no substantial differences in the proportions of conduct disordered children with severe pragmatic deficits seen at GOSH, compared with the sample from the Royal Sunderland Hospital Tier 3 CAMHS service. Because of small numbers, statistical com- parisons between the two sub-samples were inap- propriate.
This is not the first study to indicate that the division between pervasive and specific develop- mental disorders is not a sharp one (e.g., Norbury & Bishop, 2002), although it is the first to find such a substantial comorbidity for autistic features among children with conduct disorder. Indeed, increasing evidence is emerging to indicate that there is con- siderable continuity between disorders that have traditionally been regarded as quite distinct from one another. An overlap between conduct disorders and autistic traits was suspected by Moffitt et al. (2001). Other evidence more directly implies that there are links between the relatively poor verbal abilities of children with conduct problems and emotion pro- cessing skills. For example, Speltz, DeKlyen, Cald- eron, Greenberg, and Fisher (1999) reported that preschool boys with oppositional defiant disorder had poorer vocabularies for describing affective states than comparison boys. The result held after general vocabulary knowledge and test behaviour were controlled.
To our knowledge, this is the first study to investigate pragmatic communication deficits in an unreferred sample of antisocial children who were excluded, or about to be excluded, from school. Our findings indicate that such deficits are characteristic of a significant proportion, perhaps as many as two- thirds, of children excluded or at risk of exclusion from school during their first few years of education. We aim to establish if there is a causal relationship between social communication problems and exclusion in forthcoming investigations, as we sus- pect there is. However, it is important to acknow- ledge that social, cognitive, psychological and neurobiological factors may play a part in engen- dering disruptive behaviour at school. These factors almost certainly interact with one another and con- tribute in a complex manner to that outcome. Fu- ture studies might aim to include full psychosocial investigations in order to explore the relationship between psychosocial issues and the type of social communicative difficulty we describe in the current study.
Autism and autistic spectrum disorders are highly heritable conditions, with a strong genetic predis- position (Bailey et al., 1995). We do not at this stage know to what extent the pragmatic communication deficits we have described have a similar aetiology, and to what extent circumstances of upbringing contribute to the measured behaviours and skills. To date, genetic research on conduct disorder has raised more questions than it has answered, and basic issues such as the heritability of childhood
antisocial behavior have not yet been clarified (Simonoff, 2001).
Parent-professional agreement on pragmatic deficits
In the age range under investigation the CCC is not merely acting as an alternative way of identifying verbal ability: rather it picks up communicative dif- ficulties that may not be detected by conventional psychometric assessments. As in the Bishop and Baird (2001) investigation, our study showed that parental ratings of children’s pragmatic competence have discriminant validity. Previous studies have also found that parent–teacher agreement on many aspects of child behaviour is not necessarily good. This issue has been most extensively studied in relation to the parent-rated Child Behavior Checklist and the Teacher Report Form (Achenbach 1991a & b). Parent-teacher correlations for ratings of beha- vior/emotional problems are typically in the range of .2 to .4 (e.g., Verhulst & Akkerhuis, 1989). There are various reasons for the lack of agreement in our own investigation of the CCC. It is possible that the verbal and non-verbal communicative behaviours in ques- tion are insufficiently well defined to allow for ob- jective assessment. However, it is notable that the parent–teacher agreement on CCC subscales was good for Interests and for Inappropriate Initiation. Perhaps the subscales measuring Social Relation- ships and Rapport are relatively less well specified. For subscales specifically concerned with speech and language competence, there was good agree- ment about Speech, Syntax and Coherence, which may be rather more obvious than the more subtle behaviours that contribute to the Pragmatic Com- posite score. Nevertheless, in both the teacher and parent ratings we did find a relationship between CCC scores and clinical diagnoses. Disagreements between raters may in part reflect the fact that communicative abilities are context dependent. This latter interpretation, which is discussed by Bishop and Baird (2001), is suggested by the fact that higher parent–teacher correlations were found for scales assessing language structure (such as Speech and Syntax), which tends not to vary much with context. We concluded that parental ratings may be more valid than teacher ratings, insofar as they correlate rather better with the child’s diagnostic status in the clinical samples studied, although the instrument was originally devised to be rated by teachers and speech-language therapists (Bishop, 1998).
There is a striking sex ratio, with a preponderance of males in our clinically referred conduct-disordered population (just 11% were female). This ratio is very similar to that found among ASD and autistic groups. The number of females with conduct prob-
976 J. Gilmour et al.
lems in both the clinic and the excluded children samples was too small to warrant separate analysis. The very small number of girls engaging in antisocial behaviour at elementary school is in keeping with previous reports, such as the Dunedin study (Moffitt et al., 2001) which found that antisocial activities involving girls usually do not commence until early adolescence. However, the 9:1 male to female ratio we report is very high, greater than was found in that investigation at any stage. It is likely to reflect the dif- fering characteristics of a clinically referred sample, compared with an epidemiologically ascertained one.
We are aware that these surprising findings require replication. The results indicate that a significant minority of children with disruptive behaviour in the community have significant, previously unidentified, social communication difficulties. Our findings could have particular significance for chil- dren identified as having conduct disorder early in their school careers. The findings from our inner- London schools survey are of potential importance for service planning. We caution that our prelimin- ary finding will have to be followed by a detailed evaluation of the children concerned. This should include not only a formal evaluation of psychiatric morbidity, but also a neurological examination, and detailed cognitive testing. Whether these excluded children would meet diagnostic criteria for an aut- istic spectrum disorder is at present an open question, but one that is clearly accessible to empirical verification. We do not intend to imply that psychosocial and family factors play an insig- nificant role in engendering antisocial behaviour (see Hill, 2002).
With accurate identification comes the possibility of a new approach to the management of many excluded children and an opportunity to ameliorate their social communication skill deficits. Tradition- ally, interventions for children with antisocial beha- viour draw from social learning theory models (e.g., Miller & Prinz, 1990). The long-term efficacy of such models of intervention has been questioned (Kazdin, 1997). Because we suspect such social commun- ication problems could be causally related to the onset of behaviours that put children at risk of ex- clusion, a window of opportunity may exist among those in their early years at primary school, for pre- vention. There are a number of effective intervention strategies developed specifically for children with social communication deficits of this nature, particularly children with average range general in- telligence, such as those we are targeting. Special- ised social skills groups, peer tutoring (Kamps, Leonard, Vernon, Dugan, & Delquadri, 1992; Ozo- noff & Miller, 1995) and pivotal response training (Koegel, Koegel, Hurley, & Frea, 1992; Pierce &
Schreibman, 1997) increase socially appropriate behaviour, which, in turn has positive secondary gains in competence and adjustment.
The authors would like to thank Dorothy Bishop, Margaret Dimmock, Jessica Hulsmeier, Susan Woollacott, Tony Charman, Nichola Baboneau and the families and teachers who took time to complete our questionnaires and interviews.
Jane Gilmour, Sub-Department of Clinical Health Psychology, University College London, Gower Street, London WCIE 6BT, UK; Email: Jane.Gilmour@ ucl.ac.uk
References Achenbach, T.M. (1991a). Manual for the Child Beha- viour Checklist and 1991 profile. Burlington, VT: University of Vermont Department of Psychiatry.
Achenbach, T.M. (1991b). Manual for the Teacher’s Report Form and 1991 Profile. Burlington, VT: Uni- versity of Vermont Department of Psychiatry.
Bailey, A., Le Couteur, A., Gottesman, I., Bolton, P., Simonoff, E., Yuzda, E., & Rutter, M. (1995). Autism as a strongly genetic disorder: Evidence from a British twin study. Psychological Medicine, 25, 63–77.
Bagley, C., & Pritchard, C. (1998). The billion dollar costs of troubled youth: Prospects for cost-effective prevention and treatment. International Journal of Adolescence and Youth, 7, 211–225.
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. Journal of the Amer- ican Academy of Child and Adolescent Psychiatry, 39, 694–702.
Bishop, D.V., & Baird, G. (2001). Parent and teacher report of pragmatic aspects of communication: Use of the children’s communication checklist in a clinical setting. Developmental Medicine and Child Neurology, 43, 809–818.
Bishop, D.V.M. (1997). Uncommon understanding: Development and disorders of language comprehen- sion in children. Hove, UK: Psychology Press.
Bishop, D.V.M. (1998). Development of the Children’s Communication Checklist (CCC): A method for asses- sing qualitative aspects of communication impair- ment in children. Journal of Child Psychology and Psychiatry, 39, 879–891.
Bishop, D.V.M., & Rosenbloom, L. (1987). Classifica- tion of childhood language disorders. In W. Yule & M. Rutter (Eds.), Language development and disorders. Clinics in Developmental Medicine (double issue). London: Mac Keith Press.
Crick, N.R., & Dodge, K.A. (1996). Social information- processing mechanisms in reactive and proactive aggression. Child Development, 67, 993–1002.
Social communication deficits in conduct disorder 977
Department for Education and Employment (2000). Statistics of education: Schools in England 2000. London: The Stationery Office. ISBN 0-11-271095-6.
Department for Transport, Local Government and the Regions. (2000). Indices of deprivation. http://www. regeneration.dtlr.gov.uk/research/id2000/
Dunn, L.M., Dunn, L.M., Whetton, C., & Burley, J. (1997). The British Picture Vocabulary Scale, Second Edition, NFER-Nelson, Windsor.
Fombonne, E. (2001). Is there an epidemic of autism? Pediatrics, 107, 411–412.
Fombonne, E. (1999). The epidemiology of autism: A review. Psychological Medicine, 29, 769–786.
Gallagher, T.M. (1993). Language skill and the devel- opment of social competence in school-age children. Language, Speech and Hearing Services in the Schools, 24, 199–205.
Gillberg, C., & Wing, L. (1999). Autism: Not an extremely rare disorder. Acta Psychiatrica Scandina- vica, 99, 399–406.
Goodman, R. (1997). The Strengths and Difficulties Questionnaire: A research note. Journal of Child Psychology and Psychiatry, 38, 581–586.
Hill, J. (2002). Biological, psychological and social processes in the conduct disorders. Journal of Child Psychology and Psychiatry, 43, 133–164.
Kamps, D.M., Leonard, B.R., Vernon, S., Dugan, E.P., & Delquadri, J.C. (1992). Teaching social skills to students with autism to increase peer interactions in an integrated first-grade classroom. Journal of Applied Behavior Analysis, 25, 281–288.
Kaye, J.A., del Mar Merelo-Montes, M., & Jick, H. (2001). Mumps, measles, and rubella vaccine and the incidence of autism recorded by general practitioners: A time trend analysis. British Medical Journal, 322, 460–463.
Kazdin, A.E. (1997). Practitioner review: Psychosocial treatments for conduct disorder in children. Journal of Child Psychology and Psychiatry, 38, 161–178.
Koegel, L.K., Koegel, R.L., Hurley, C., & Frea, W.D. (1992). Improving social skills and disruptive beha- vior in children with autism through self-manage- ment. Journal of Applied Behavior Analysis, 25, 341– 353.
McDonald, S., & Pearce, S. (1996). Clinical insights into pragmatic language theory: Frontal lobe deficits and sarcasm. Brain and Language, 72, 81–104.
Matthys, W., Cuperus, J.M., & Van Engeland, H. (1999). Deficient social problem-solving in boys with ODD/CD, with ADHD and with both disorders. Journal of American Academy of Child and Adolescent Psychiatry, 38, 311–321.
Milch-Reich, S., Campbell, S.B, Pelham, W.E., Connel- ly, L.M., & Geva, D. (1999). Developmental and individual differences in children’s on-line represen- tations of dynamic social events. Child Development, 70, 413–431.
Miller, J.N., & Ozonoff, S. (2000). The external validity of Asperger disorder: Lack of evidence from the
domain of neuropsychology. Journal of Abnormal Psychology, 109, 227–238.
Miller, G.E., & Prinz, R.J. (1990). Enhancement of social learning family interventions for childhood conduct disorder. Psychological Medicine, 108, 291–307.
Moffitt, T.E., Caspi, A., Rutter, M., & Silva, P.A. (2001). Sex differences in antisocial behaviour (pp. 237–239). Cambridge, UK: Cambridge University Press.
National Union of Teachers. (1992). Survey on pupils’ exclusions. London: Author.
Norbury, C.F., & Bishop, D.V. (2002). Inferential processing and story recall in children with commun- ication problems: A comparison of specific language impairment, pragmatic language impairment and high-functioning autism. International Journal of Language Disorders, 37, 227–251.
Ozonoff, S., & Miller, J.N. (1995). Teaching theory of mind: A new approach to social skills training for individuals with autism. Journal of Autism and Developmental Disorders, 25, 415–433.
Parsons, C. (1996). Exclusion from school: The public cost. London: Commission for Racial Equality.
Pierce, K., & Schreibman, L. (1997). Multiple peer use of pivotal response training to increase social behaviors of classmates with autism: Results from trained and untrained peers. Journal of Applied Behavior Analy- sis, 30, 157–160.
Simonoff, E. (2001). Gene–environment interplay in oppositional defiant and conduct disorder. Child and Adolescent Psychiatric Clinics of North America, 10, 351–374.
Skuse, D., Dimmock, M., Warrington, R., Lau, J., Nicholls, D., & Place, M.. (2004). The Developmental Dimensional and Diagnostic Interview for Children (3di): A novel computerized assessment for Autistic Spectrum Disorders. Journal of the American Acad- emy of Child and Adolescent Psychiatry, 45, 548–558.
Social Exclusion Unit. (1998). Truancy and school exclusion; Report by the Social Exclusion Unit. ISBN 0-10-139572-8 (updated 27 March 2000).
Speltz, M.L., DeKlyen, M., Calderon, R., Greenberg, M.T., & Fisher, P.A. (1999). Neuropsychological characteristics and test behaviors of boys with early onset conduct problems. Journal of Abnormal Psy- chology, 108, 315–325.
Verhulst, F.C., & Akkerhuis, G.W. (1989). Agreement between parents’ and teachers’ ratings of beha- vioural/emotional problems of children aged 4–12. Journal of Child Psychology and Psychiatry, 30, 123– 136.
Wechsler, D. (1992). Wechsler Intelligence Scale for Children –Third UK edition. London: Psychological Corporation.
World Health Organization. (1996). The ICD-10 classifi- cation of mental and behavioural disorders: Multiaxial classification of child and adolescent psychiatric dis- orders. Geneva: World Health Organization.
Manuscript accepted 29 September 2003
978 J. Gilmour et al.