European Child and Adolescent Psychiatry, Vol.2, Issue 2, April 1993, pp.61-74 Hogrete & Huber Publishers

The Definition and Prevalence of Autism: A Review

Lorna Wing

The Centre for Social and Communication Disorders, Elliot House, 113 Masons Hill, Bromley, Kent, BR2 9HT, UK

______________________________________________________________________________________________ Sixteen studies of the prevalence of autism in childhood, using epidemiological methods in defined populations in Europe, the USA and Japan, in English or with English summaries, were found in the published literature. Age specific rates varied from 3.3 to 16.0 per 10,000. Eight studies gave rates for a sub-group of 'typical' autism varying from 1.2 to 8.4. Reasous for variations were sought by examining geographic and demographic details of the populations screened, methods for initial screening aud final examination of possibly autistic children, demographic and clinical details of children identified as autistic, and criteria used for diagnosis. There was evidence, independent of diagnostic criteria, of a higher prevalence among children of first generation immigrants to Europe from 'exotic' countries. Apart from this, all differences could be due to variations in diagnostic practice and increasing awareness of the manifestations of autistic conditions throughout the range of intelligence, from severely retarded to average and above. However, it remainis possible, though not proven, that rates do vary over place and/or time. The problems of defining a sub-group with 'typical' autism among the wide spectrum of children with the triad of impairments of social interaction, communication aud imagination are discussed and the value of such a sub-grouping questioned. ______________________________________________________________________________________________


Kanner's classic paper (1943) outlined the behaviour pattern, present from early in life, which he named 'early infantile autism'. Prior to this, there were, in the literature, occasional accounts of individual children whose behaviour fitted the picture Kanner later described; the most famous of these was Victor, the wild boy of Aveyron (see translations into English of J. M. G. Itard's reports by Lane, 1977). But, since no one prior to Kanner had published a series of such cases, no infomation exists as to the prevalence of autism before the second half of the twentieth century. Kanner (1943) described only the autistic children referred to his clinic and, later on, those attending a particular special school (Kanner, 1973). He made no estimates of the numbers in the general population, but thought that his syndrome was rare.

Twenty six years ago, Lotter (1966), working in the Medical Research Council Social Psychiatry Unit at the London Institute of Psychiatry, published the first paper to give the results of an epidemiological study of autistic conditions in a complete, large population of children of all levels of intelligence living in a defined geographical area (the former County of Middlesex, England). At about the same time, but independently, Birte Hoeg Brask (1972) was identifying children with 'childhood psychosis' in touch with psychiatric and mental retardation services in Aarhus, Denmark. Since then, there have been a number of other studies of the prevalence of autism, with varying results.

In this paper, a quarter of a century of work in this field is reviewed in order to consider possible reasons for the differences in the reported findings. The studies to be described in detail and tabulated are those meeting the following criteria; they use defined diagnostic criteria; they cover geographically defined populations; the initial selection is made from a wide range of children including those not previously diagnosed as autistic; final identification is based on clinical examinations of selected children by the authors or by the direct care clinical staff who provide information requested by the authors as part of the study. Only papers published in books or journals in English or with detailed English summaries are included, not unpublished reports.

Studies Reviewed

The results of the studies of the prevalence of autism are summarised in Tables I to S. Each is identified by the name of the area in which it was carried out. The authors of the papers, the full names of the areas studied, and the abbreviations of these names used in the tables are listed below, in the order in which they appear in the tables:

- Lotter (1966; 1967a; 1967b) Middlesex, England (Midx)
- Brask (1972) Aarhus, Denmark (Aarh)
- Wing & Gould (1979) Camberwell, London, England (Camb)
- Hoshino et al.(1982) Fukushima-ken, Japan(Fuma)
- Bohman et al. (1983) Västerbotten, Sweden(Väst)
- Cialdella & Mamelle (1989) Rhône, France(Rhôn)
- Matsuishi et al. (1987) Kurume, Japan (Kuru)
- Tanoue et al. (1988) Ibaraki, Japan (Ibar)
- Sugiyania & Abe (1989) Nagoya, Japan (Nago)
- Burd et al. (1987) North Dakota, USA (NDak)
- Ritvo et al. (1989) Utah, USA (Utah)
- Gillberg (1984) Göteborg region, Sweden (Göt1)
- Steffenburg & Gillberg (1986) Göteborg city, Sweden (Göt 2)
- Bryson et al. (1988) Nova Scotia, Canada (NoSc)
- Gillberg et al. (1991) Göteborg city, Sweden (Göt 3)
- Ishii & Takahashi (1983) Toyota, Japan (Toyo)

For the studies in Fukushima, the Rhône, and Utah, the authors gave reasons for considering that the rates found for particular age groups within the total range studied were the most accurate. The findings for these age groups are given in Tables 1 and 2. For the second and third studies in Göteborg the authors also gave rates separately for rural Bohuslän, but only the findings for city Göteborg are quoted in Tables 1 and 2. However, for all these studies, the data in Table 5 refer to all the children finally diagnosed as autistic because, for some of the papers, this was the only information given.


Nine studies were excluded from the tables because they did not meet the criteria listed above.

They are briefly described below.

The prevalence of autism was examined in Kyoto-fu, Gifu-ken, Toyama-ken and Hokkaido in Japan by, respectively, Haga and Miyamoto (1971), Nakai (1971), Tanino (1971) and Yamazaki et al. (1971). Only short English summaries are available for these four studies. Since diagnostic criteria are not given in the summaries the results are not included in the tables. However, from the dates of publication and the references, the diagnoses were most likely to have been based on Kanner's papers. The rates found ranged from 0.9 to 5.02 per 10,000.

One study was excluded because it was based only on inforation from computer print-outs. Treffert (1970) examined the prevalence in Wisconsin, USA among children, aged 3 through 12, seen by psychiatric or mental retardation agencies in the years 1962-67 and diagnosed as having 'Childhood schizophrenia'. He reported a total age-specific prevalence of 3.1 per 10,000 of whom 0.7 were considered to be classically autistic without any known organic cause.

Two further papers were excluded because diagnosis was based on examination of the case notes only. Aussilloux et al. (1989) studied the notes of 6-12 year old children, born 1966-73, known to the relevant services in the Hérault department, France. Using DSM-III criteria, the age-specific rates found were 4.7 per 10,000 for 'Infantile autism' and 1.8 per 10,000 for 'Childhood onset pervasive developmental disorder'. Fombonne and du Mazaubran (1992) carried out a study in four regions in France (Aquitaine, Ile de France, Lorraine and Picardie). The case notes of two annual birth cohorts of children born 1972 and 1976 respectively) known to the relevant services were examined and diagnoses made using the French system of classification (Mises, 1990). The age-specific rate for the combined categories of 'Kanner's autism' and 'Other forms of infantile autism' was 4.9 per 10,000.

The study by Steinhausen et al. (1986) was excluded because the search for children was confined to two sources only, one being the university clinic for child psychiatry and the other a centre providing a programme for autistic children. The authors used Rutter's criteria (see below) to identify children with 'nuclear' autism aged under 15 and living in West Berlin, Germany on June 1, 1992. The age-specific prevalence found was 1.9 per 10,000.

Finally, one study, by McCarthy et al. (1984), was excluded because, although the authors examined the selected children, the initial selection was made by asking child psychiatrists and the staff of relevant institutions for the names of children diagnosed as autistic. Using criteria similar to those in DSM-III (see below), the authors found an age-specific prevalence of 4.3 per 10,000 children aged 8, 9 and 10 years, who were born in the Eastern section of Ireland in 1965-68.

Diagnostic Criteria

In Tables 1 to 5, the studies are arranged in five groups, according to the diagnostic criteria used by the authors. Since variation in the definition of autism is one of the factors that may account for differences in reported prevalence rates, the criteria are summarised as follows.

Kanner's Criteria

Kanner gave vivid descriptions in his first paper (Kanner, 1943), but did not operationalise his diagnostic criteria.

Kanner and Eisenberg (1956) discussed Kanner's original conception of autism and the five features he considered to be diagnostic. These were, a profound lack of affective contact with other people; an anxiously obsessive desire for the preservation of sameness in the child's routines and environment; a fascination for objects, which are handled with skill in fine motor movements; mutism or a kind of language that does not seem intended for inter-personal communication; good cognitive potential shown in feats of memory or skills on performance tests, especially the Séguin form board. Kanner also emphasized onset from birth or before 30 months.

In the same paper, Kanner and Eisenberg modified the diagnostic criteria by selecting two as essential. These were:

1. a profound lack of affective contact
2. repetitive, ritualistic behaviour, which must be of an elaborate kind.

They considered that, if these two features were present, the rest of the typical clinical picture would also be found. Lotter used the two features in his final identification of autistic children and in the division into 'nuclear' and 'non-nuclear groups. He operationalised the criteria and gave examples of the behaviour (Lotter, 1967b). He did not include age ofonset as essential.

Rutter's Criteria

Rutter (1978) suggested four criteria for defining chiildhood autism. These were:

1. impaired social development which has a number of special characteristics out of keeping with the child's intellectual level;
2. delayed and deviant language development that also has certain defined features and is out of keeping with the child's intellectual level;
3. 'insistence on sameness' as shown by stereo-typed play patterns, abuormal preoccupations or resistance to change.
4. Onset before 30 months.

Rutter gave many examples of the above behaviours, based on his own research and clinical experience and on Kanner's descriptions.

DSM-III Criteria

In 1980, the third edition of the Diagnostic and Statistical Manual of Mental Disorders, known as DSM-III, was published by the American Psychiatric Association. In this, the term 'pervasive developmental disorder' was used for the general category of autism and related conditions. A subgroup labelled 'infantile autism' was defined by;

1. lack of responsiveness to others;
2. language absence or abnormalities;
3. resistance to change or attachment to objects;
4. the absence of schizophrenic features;
5. onset before 30 months.

DSM-III also has categories for childhood onset (after 30 months and before 12 years) and for atypical pervasive developmental disorder.

DSM-III-R Criteria

Denckla (1986) published the deliberations of a workshop on the diagnosis of autism and related disorders. The participants agreed that the essential features of autism were social impairment, delayed or deviant language (communication) and repetitive, stereotyped or ritualistic behaviour. These are the familiar features appearing in all systems of definition. The difference was the recognition, in agreement with Wing and Gould (1979), that all these features could occur in widely varying degrees of severity and in many different manifestations. For example, it was no longer felt that autism was diagnosable only if the child was (or had been) aloof and indifferent to other people; passive acceptance of social approaches without any spontaneous initiation of contact, or active but inappropriate, one-sided approaches should also be included as forms of social impairment. This shift of emphasis would permit the inclusion of children with the most subtle as well as those with the most obvious features of autism. Age of onset was considered to be difficult to establish and not to be regarded as an essential criterion. It was also accepted that the diagnosis of autism or related conditions should be made independently of intelligence, aetiology or any associated conditions but that the clinical picture should be evaluated in the light of the developmental age. The definitions in the section on pervasive developmental disorders in the revised version of DSM-III, referred to as DSM-III-R (American Psychiatric Association, 1987) have moved in the direction of the views expressed by the working party.

The diagnosis ofautistic disorder in DSM-III-R requires:

1. impairment in reciprocal social interaction (at least two from a list of 5 items, comprising specified clinical examples);
2. impairment in verbal and non-verbal communication (at least one from a list of 6 items);
3. markedly restricted repertoire of activities and interests (at least one from a list of 5 items);
4. a grand total of at least eight from among the 16 items listed.

Other Criteria

Ishii and Takahashi (1983), in Toyota, used their own version of the diagnostic criteria for autism, as follows:

1. disturbed inter-personal relationships (defined by a list of clinical examples comprising 9 items);
2. absence or deviance in speech and language development (8 items);
3. insistence on the preservation of sameness or resistance to change (6 items)
4. abnormal responses to sensory stimuli or motility disturbance (10 items).

Age of onset, intelligence level and organic aetiology were not mentioned among these diagnostic criteria. The English summary did not specify how many items were needed for a diagnosis of autism, nor whether the autistic subjects were sub-grouped.

Differences in Prevalence Rates

In this discussion, the term 'autism' will include both 'typical' and 'atypical' autism however these are defined in the papers reviewed.

Table 1. Age specific rates per 10,000.


Area              Total no.                     Rates                          
                  of  autistic         Total   Typical Atypical                       
                  children                      autism  autism          


Kanner's  criteria                                                            
Midx'               35                  4.5      2.0      2.5    
Aarh1               20                  4.3       -        -     
Camb1               17                  4.9      2.0      2.9    
Fuma1              108                  5.0       -        -      

Rutter's criteria                                                             
Väst1               39                  5.6      3.0      2.6    
Rhôn2              112                 10.8      8.0      2.8    

DSM-III criteria                                                              
Kuru2               51                 15.5       -        -      
Ibar2              132                 13.8       -        -      
Nago2               16                 13.0       -        -      
Utah1               66                  4.0a      -        -      
Ndak1               59                  3.3      1.2      2.1    
Göt11               51                  4.0      2.0      1.9    
Göt21               32                  7.5      4.7      2.8    

DSM-III-R criteria                                                            
NoSc2               21                  10.1      -        -
Göt32               47                  11.5     8.4      3.2    
Other criteria                                                                
Toyo2               56                  16.0      -        -      


Note: The terms 'typical' and 'atypical' autism are used to cover the sub-divisions of autism adopted by some of the authors. The definitions and terms used vary.

aThe authors estimated the rate to be higher than that calculated from the numbers found (3.6) to allow for undentified children.

1 = Rate lower than the mean, 2 = Rate higher than the mean

Table 2. Populations screened.

Area     Type             Urban   Total    Years of  Age            
                          rate> children    birth  range 
                          rural in age range 
Kanner's criteria                                                                                

Midx1    Mainly urban       -    77,800      53-55   8-10           
Aarh1    Mainly urban       -    46,500      49-60   2-14           
Camb1    Connurbation       -    34,700      56-70   3-17           
Fuma1    Urban and rural   Yes  217,600      68-74   4-10a          

Rutter's criteria                                                                            
Väst1    Urban and rural    -    69,600      60-79   0-20    
Rhôn2    Mainly Urban       -   103,700      79-83   5-9a                  

DSM-III criteria                                                                            
Kuru2    Urban              -    32,800      71-79   4-12    
Ibar2    Suburban and rural -    95,400      72-78   7              
Nago2    Urban              -    12,300      79-84   3-5     
Utah1    Urban and rural    -   184,800      75-79   8-12a          
Ndak1    Urban and rural    -   181,000        ?     2-18           
Göt11    Urban region       -   128,600      62-76   4-18                   
Göt21    Urban city        Yes   42,900      75-84   0-10           

DSM-III-R criteria                                                                            
NoSc2    Suburban and rural -    20,800      71-79   6-14      
Göt32    Urban city        Yes   40,700      75-84   4-13           

Other criteria                                                                             
Toyo2    Urban              -    35,000      70-76   6-12    

a The authors studied a wider age range prevalence.

1 = Rate lower than the mean, 2 Rate higher than the mean, - = variable not reported

The age-specific rates per 10,000 for autism are shown in Table 1. The mean of all the rates quoted in the table is 8.4 with a standard deviation of 4.4.

The rates include one of 4.5 found among children born in the years 1953 to 1955 in a county in England and another of 16.0 among children born from 1970 to 1976 in a town in Japan. But, before concluding that there are real differences in prevalence related to place or time, it is necessary to consider the effects of the methods used. These include the demographic characteristics of the population studied, the techniques for the initial screening, the examination of the children selected as possibly autistic and the diaguostic criteria applied. Clinical and demographic features of the children diagnosed as autistic may also suggest biases affecting selection.

The Populations Studied

The demographic details of the populations are given in Table 2.

Geographical Area

Among the rates in the Japanese studies tabulated were one that was below the overall mean ( 5.0 in Fukushima) and four that were well above, including the highest of all (16.0 in Toyota).
Rates above the mean were also reported from a study in Canada, one in France and the last one of the series in Göteborg. Other European studies and the two in the USA gave rates below the mean. In the three studies that gave urban and rural rates separately, the urban rates were higher, but this did not account for the differences between the results overall since most of the studies were in urban areas.

Population Size in the Age Ranges Studied

The numbers in the total populations of eligible children varied from 20,800 in Nova Scotia to 769,000 in Utah (the population quoted in Table 2 is only for the age range found to have the highest prevalence). There was no significant relationship between size of population and rates of autism.

Table 3 Initial screening methods.


Area        Methods                              
Kanner's criteria
Midx1       Screening schedule for teachers in all schools.
            Case notes reviewed for children in special schools.
Aarh1       Case notes reviewed in psychiatric, paediatric, neurological
            and mental retardation services and 20 care institutions.
Camb1       Interviews with teachers or carers of children with any physical, psychiatric
            or learning problem.  
Fuma1       Screening schedules to 2,333 agencies - all schools, psychiatric,
            paediatric, child welfare, education and public health centres. 

Rutter 's criteria
Väst1       Detailed information on autism given to almost every organisation working
            with children - then a screening schedule to 716 professionals in services.
Rhôn2       Psychiatrists and psychologists in all psychiatric, and retardation services
            and centres for children with sensory or motor impairments asked
            to identify children diagnosed as 'infantile psychosis'

DSM-III criteria                                                         
Kuru2       Case notes reviewed in all schools, child guidance clinics, welfare
            and public health offices, education board, education institute
            and hospitals.
Ibar2       Child guidance clinic referrals including those from routine
            developmental checks at 3 and 6 years. Counted as annual birth
            cohorts at age 7
Nago2       Found from routine developmental checks at 3m, l8m, 3y and 6y.
Utah1       Information from previous contacts with parents, the Autism Society, clinicians,
            service providers; soliciting families through media publicity;
            referrals from all relevant medical, educational and social agencies;
            search of case records in residential homes and hospitals for
            mentally retarded and mentally ill.
NDak1       Information and screening schedule to medical, psychiatric and developmental
            disabilities agencies.  
Göt11       Information and screening schedule to all paediatricians, child psychiatrists,
            doctors in school health, audiology,
            well-baby clinics and mental retardation institutions.                                                            
Göt21       As for Göt1.                         

DSM-III-R criteria                                                       
NoSc2       Screening schedule to teachers in all schools.                         
Göt32       As for Göt1.                         

Other criteria                                                           
Toyo2       Screening schedule to teachers in all schools.                         

1 = rate lower than mean, 2 = rate higher than mean

Age Groups

The studies in Fukushima, Västerbotten, Rhöne, Southern Ibaraki, Göteborg, Kurume City, North Dakota, Utah and Toyota all included examination of prevalence rates in different age groups.
There were marked variations; for example in Västerbotten the age specific rate per 10,000 was 12.6 in the 7-9 year olds and 4.6 in those aged 13-15.

The age groups reported to have the highest prevalence rates were in age bands within the 4-12 year range, when the phenomena of autism are likely to be seen in their most easily identifiable form (Rutter, 1970; Wing, 1988). When Gillberg et al. (1991) updated their previous study of children born 1975-84 (Steffenburg & Gillberg,1986), they identified 9 additional children who had been under 4 years in the first study. However, inspection of Table 2 shows that the age ranges chosen do not explain all the differences between the rates.

Years of Birth

The studies that included subjects born in the 1950s and 60s all showed rates below the mean, but the years of birth are closely related to the diagnostic criteria used (see below). Tanoue et al.(1988) in Ibaraki found that the rates of autism in annual birth cohorts from 1973 to 1978 were significantly correlated with the numbers of children admitted to hospital with pneumonia and bronchiolitis in each of the years in which the cohorts were born. No other studies examined the possible significance of the years of birth.

TabIe 4. Final identification of autistic children.

Area    Age of  Methods of                                                
        onset   examination                                                 
Kannier's criteria                                                      
Midx1    No     T child, I teacher & parent, CN                              
Aarh1    No     P child, I teacher & parent, CN                              
Camb1    No     T,N child, I teacher & parent, CN                            
Fuma1    NS     P child, I parent                                            

Rutter's criteria                                                       
Väst1    No     P child, I parent, CN                                        
Rhôn2    Yes    Schedules completed by psychiatrists and psychologists
                who had clinical care of possibly autistic children

DSM-III criteria                                                        
Kuru2    NS     P,N child, CN 
Ibar2    NS     P,N child, I parent                                          
Nago2    Yes    P,N,T child, I parent                                       
Utah1    Yes    P child, I parent                                           
NDak1    Yes    P,N,T child, I parent                                 
    for typical 
Göt11    Yes    P,N,T child, I parent, CN                              
    for typical 
Göt21    Yes    P,N,T child, I parent, CN                              
    for typical 

DSM-III-R criteria                                                      
NoSc2    No     T child, I parent, CN                                        
Göt32    No     P,N,T child, I parent, CN                                    

Other criteria                                                          
Toyo2    NS     P child, I parent                                            
1 = Rate lower than mean, 2 = Rate higher than mean, NS = Not stated if age of onset criterion used, T = Psychological tests, I= Interview, N= Neurological exam, P= Psychiatric exam CN= case notes inspected

Methods of Screening

The nine studies excluded from the tables because the methods used did not meet the criteria for inclusion all found low rates.

The methods of screening the eligible populations in the epidemiological studies are shown in Table 3.

There were wide differences among the studies. Some workers collected information on all children in the eligible population, most of whom had no disabilities, while, at the other end of the scale, some authors used only psychiatric service case records. Two studies in Japan made use of the routine developmental checks that cover the large majority of young Japanese children and their rates were among the highest (13.9 and 16.3).

Various local factors affect the comprehensiveness of the initial screening. For studies that rely on counting cases diagnosed by agencies, the numbers of children with special needs who make contact with those agencies is an important factor. Also important is the level of awareness of the existence and the characteristics of autism among those returning infomation, and their willingness to cooperate. In Västerbotten and Göteborg, the authors sent information on autism to professionals before beginning their studies. Gillberg et al. (1991) discussed the possibility that the higher rates found in the third study in Göteborg were partly due to increased awareness of autism among professionals in the area. However, it was not possible to explain all the variations in rates by these factors, nor by the screening methods used.

Examination of Children Selected as Possibly Autistic

The methods of examination are shown in Table 4. In all but one of the studies tabulated, the children found in the initial screening and, in most cases, their parents or teachers were seen by the authors. The reasons for the differences in rates cannot be explained by the particular method used.

Characteristics of the Children

Demographic and clinical characteristics of the children reported in the studies are given in Table 5. In each case, the figures refer to the total number of children identified, that is, typical plus atypical autism combined if such a distinction is made by the authors concerned.

Occupational Class

A bias towards higher occupational class was found in two studies but not in the other seven in which class was examined. There was no obvious link with the prevalence rates.


In the study in Camberwell (Wing 1979, 1980) the rate for Kanner's autism among children whose fathers were British or, in a very few cases, from Western Europe was estimated to be 4.4 per 10,000 compared with 6.3 for children whose

Table 5. Charactenstics of the autistic children.

Area    Class   High     Male:       IQ            %CNS                                                        
        bias  immigrant  female %<50 %<70 %70+     dysfunc.
               prev'ce   ratio               
Kanner's criteria
Midx1   Yesa     -         2.6   68   16   16       41  
Aarh1   No       -         1.4   -    -    -        80  
Camb1   Nob     Yesd       16.0  35   35   30       59  
Fuma1   Yes      -         9.0   -    -    -        61  

Rutter's criteria
Väst1   -        -         1.6   -    -    -        28  
Rhôn2   No       -         2.0   -    -    -        27 

DSM-III criteria
Kuru2   -        -         4.7   -    -    -        - 
Ibar2   -        Yes       4.1   -    -    -        - 
Nago2   -        -         -     -    -    -        - 
Utah1   No       No        3.7   41   25   34       - 
Ndak1   -        -         2.7   51   31   18       56 
Göt11   Noc      -         1.8   33   45   22       - 
Göt21   No       Yese      3.0   58   31   12       42

DSM-III-R criteria 
NoSc2   -        -         2.5   43   33   24       -
Göt32   No       Yes       2.8   54   28   18       49

Other criteria
Toyo2   -        -         6.0   -    -    -        - 
Note: Evidence of CNS dysfunction is defined in different ways by different authors.

a Lotter (1967), b Wing (1980), c Gillherg & Schaumann (1982), d Wing (1979), e Gillberg et al. (1987) 1 = Rate lower than mean, 2 = Rate higher than mean, - = Variable not reported

fathers were first generation immigrants from third world countries, mainly the Caribbean. The rates for autistic-like conditions, broadly defined (see discussion on defining borderlines below, and Table 6) were, respectively, 9.9 and 37.8. (The rates were related to the origins of the fathers because this was the only relevant sub-grouping of children in the population census current at the time of the study.)

In the two later studies in Göteborg (Gillberg et al. 1987, 1991) it was found that children with autism (typical and atypical) were significantly more likely to have parents from 'exotic' countries ( Asia, South America, or south-east European countries). This factor helped to elevate the rates in the second and third compared with the first Göteborg study.

Tanoue et al. (1988) found a significant difference between the rate for children born in Southern Ibaraki (11.3 per 10,000) compared with that for children born in other parts of Japan whose parents had migrated into the Southern Ibaraki area (17.6 per 10,000).

Table 6. Other sub-groups within the autistic spectrum.

Area     Sub-groups                             No. of           Age-specific 
                                               children          rates/10,000 
Camba    Socially impaired - excluding           57                 16.3  
         autistic - any IQ
Göt4b    Socially impaired - excluding           37                 15.1  
        autistic - IQ<70  
Göt5c    Asperger syndrome - IQ 70+              ?                  10-26 
a Wing & Gould (1979)
b Gillberg et al. (1986)
c Gillberg & Gillberg (1989)

Ritvo et al. (1989) examined prevalence by ethnic origin in Utah. They found no excess of autism in children of parents of any ethnic minority, including Hispanic and Asian, but they did not report whether the parents were first or subsequent generation immigrants. Migration was not discussed in the other studies but it is unlikely that it can account for all the variations.

Male: Female Ratios

The excess of boys noted in Kanner's (1943) original paper was found in all the studies, but with wide variations. It has been obseryed that girls who are diagnosed as autistic tend to be more severely mentally retarded than autistic boys (Lotter, 1966; Brask, 1972; Wing, 1981b; Ritvo et al. 1989; Gillberg et al., 1991). A low male to female ratio could be due to the inclusion of many children with severe mental retardation. Conversely, a high ratio could be the result of the inclusion of a large pro- portion of more able children, some perhaps being from the sub-group with Asperger's syndrome (Wing, 1988). If both these biases were present they would be likely to cancel each other. In the event, no relationsilip with the reported intelligence levels could be seen, but the information on IQ was insufficient to draw any conclusions (see below). The exceptionally high ratio in Camberwell (16:1) was most probably a chance finding due to the small number of autistic children identified (17).

Level of Intelligence

In all the studies where information is available, the majority of the children were mentally retarded with between one third and two thirds being in the severely or profoundly retarded range.

The findings on intelligence in Västerbotten are not given in the table because the authors used different cut-off points for their sub-groups. They also noted that the majority of the children were mentally retarded.

Hoshino et al. (1982), who found a comparatively low prevalence in Fukushima, separated their autistic group from a somewhat larger group with 'autistic mental retardation' (not reported in the tables in the present paper). On the other hand,

Matsuishi et al. (1987) who found a high prevalence in Kurume, noted that their group contained none who were 'high functioning'.

Gillberg (1992) and Gillberg et al. (1991) observed that children diagnosed as autistic who were either severely retarded or of normal or near normal intelligence accounted for the increase in the rates among the children born of native Swedish parents found in the successive Göteborg studies, while there was little change in the rates for the group with mild-moderate retardation. (This trend is not shown in the Table 5 because the figures include children of immigrant parents). They considered that the increases were due to improving awareness of autism in the area as a result of the series of prevalence investigations, not to a real change in prevalence. However, autism associated with immigrant status did increase (see below). No relationship was found between the proportions of the different 19 groups and the prevalence rates shown in the tables. However, in most of the studies in which IQ was reported the measures used were not specified, so it is difficult to compare the results.

Neurological Dysfunction

In the studies listed in the tables, the evidence for the presence of physical conditions affecting the central nervous system comprised identifiable conditions in the history and/or present state, such as maternal rubella, peri-natal trauma, a history of encephalitis, epilepsy, tuberous sclerosis. Different authors used different criteria and, though all found a substantial proportion of children with such conditions, the actual percentages varied from 27 to 80. None included the results of intensive biochemical, genetic or other examinations of all the children. However, Gillberg and colleagues noted that, when the children in the Göteborg studies were given a full, systematic medical work-up, almost all showed some evidence of abnormalities associated with brain dysfunction (Gillberg, 1984; Steffenburg and Gillberg, 1986; Gillberg et al., 1991).

Higher percentages of children with conditions suggesting physical brain dysfunction might have been associated witil the inclusion of more of the severely mentally retarded cihIdren and hence more girls but no relationsilip was found with the prevalence rates or the male: female ratios.

Criteria for Autism

The five diagnostic schemes used in the studies described here all include impairment of social interaction and communication and repetitive, stereotyped behaviour as essential criteria, though there are differences in the details of the definitions of these features.

As shown in Table 1, in seven of the studies, the autistic group was divided into a typical and a less typical sub-group (the exact terms used varied). The classification was based on how closely the clinical picture fitted the clinical criteriaused (Middlesex, Camberwell, Västerbotten, the third Göteborg study) or on age of onset and/or typicality of the clinical picture (North Dakota, the first two Göteborg studies) or absence or presence of probable major neurological abnormalities (Rhône). The results for the studies with the highest rates were not sub-grouped so no conclusions can be drawn from this comparison.

The inclusion in DSM-III of the item 'absence of schizophrenic features' was discussed in the paper on the Rhône study. Cialdella and Mamelle (1989) considered that listening attitudes, formal thought disorders (undefined by the authors) and incongruent laughing and crying were 'schizophrenic features'. When they excluded the children with these types of behaviour, which are common in autism, the rates dropped by half. The rate for the Rhône quoted in Table 3 is that found when this criterion is ignored.

In Nagoya, the authors followed the children in their study from 18 months to 7 years of age; they noted that delusions, hallucinations and thought disorder could automatically be excluded because of the young ages of their subjects. Other authors using DSM-lII criteria did not mention this item.

Age of onset was used as a criterion in some studies and not for others. It can be seen from Table 4 that this did not account for the differences between the rates. The great majority of children with autism have a reported onset before 30 months.

The most consistent findings are those using Kanner's criteria, which include the presence of elaborate repetitive routines. Lotter's definitions were used in the studies in Camberwell and Aarhus. The workers in Fukushima stated that they used Kanner's diagnostic criteria but gave no further details. The overall rates from the three studies using these criteria range from 4.3 to 4.9. As already mentioned, Gillberg (1992) pointed out that, in the Göteborg studies, the rate of autism associated with mild or moderate mental retardation, among children born of native Swedish parents, remained almost the same despite changes in other sub-groups. Gillberg argued that the cases Kanner described were mostly mildly or moderately retarded so the rates for typical 'Kanner autism' in this particular population appeared to be stable. However, the prevalence of autism, even when narrowly defined by Kanner's criteria, appears to be affected by parental immigration.

The Problem of Defining Borderlines

Wing and Gould (1979), in the Camberwell study, identified all children who were impaired in their capacity for reciprocal social interaction. These authors found that this kind of abnormality of social interaction was closely associated with impairment of communication and imagination, the latter resulting in a narrow, repetitive pattern of activities.

They referred to this as the 'triad of social impairments'. It is clear that all autistic children, as defined on any of the criteria used by the studies reported here, have the triad of impairments.

Wing and Gould found that there were many more children who also had the triad but who did not precisely fit Kanner's descriptions of his syndrome. (For an account oftue various clinical pic-tures see Wing, 1988). The important point was that there were no clear cut borderlines between typical autism, atypical autism and other manifestations of the triad.

Many children with the triad were severely, even profoundly mentally retarded but others were in the normal range of intelligence. Some of the latter fitted Asperger's description of his syndrome (Asperger, 1944; Wing, 1981; Frith, 1991).

The age specific prevalence, in Camberwell, of those with the triad but not typically autistic was 16.3 per 10,000 makiug the total for all with the triad in any form 21.2 per 10,000 (see Table 6).

Gillberg et al. (1986) found a closely similar prevalence in Göteborg although he included only those with IQs under 70 (see Table 6).

The prevalence of Asperger's syndrome among those with normal intelligence was also investigated by Gillberg and Gillberg (1989). They made estimates based on the results of several studies and concluded that the range lay somewhere between 10-26 per 10,000 (see Table 6).

Thus the possible total prevalence of those of any level of intelligence with the triad of social impairment, referred to as the 'autistic continuum' (or, perhaps more appropriately, the 'autistic spectrum') may be as much as 47 per 10,000.


Sixteen published epidemiological studies, using specified criteria and methods, carried out in different places and at different times, reported age-specific prevalence rates of autism varying from 3.3 to 16.0 per 10,000. Comparison of the methods and results gives no clear evidence for or against the hypothesis that there are real differences in place or time or both. There is evidence that immigration may affect true prevalence and that greater knowledge of autism increases the number of cases identified when screening populations1 but these factors cannot account for all the variation reported.

The evidence does suggest that the definitions used have some effect on the results. The most consistent findings, in the range 4.3 to 4.9 per 10,000, were reported in studies using Kanner's two criteria as operationalised by Lotter. From the present author's experience it is the specifying of social aloofness and indifference, at least in the early years, and elaborateness of the repetitive routines that makes for greater precision and the narrowing of the field. The other systems of diagnosis leave more to the interpretation of the individual worker.

A factor linked to definition is level of intelligence. The presence of elaborate routines means that the child has some capacity for organising the environment, so is likely to have some ability at least on visuo-spatial tasks. But intelligence in the normal or superior range tends to be associated with marked improvement with increasing age, more social approaches and verbal rather than object-oriented routines (Rutter, 1970; Kanner, 1973; Wing, 1988). Thus the manifestations of the triad are less typical of Kanner's autism in children with very low or very high cognitive ability, especially, in the latter group, with increasing age. Lotter's operationalised criteria for the social impairment emphasise aloofness; his criteria for elaborate routines are related to objects or temporal sequences not verbal or abstract repetitive behaviour. The evidence from some of the studies reviewed here gives some but not conclusive support for the view that the higher rates that have been reported may be due to the inclusion of both more severely retarded and more able children.

The real problem for any system of diagnosis is the lack of any clear boundaries between 'classic autism', whichever of the existing definitions is used, and the rest of the autistic spectrum. All those with the triad of social impairments share the features that are regarded by all systems as crucial to autism, the only differences being the ways in which the features are manifested. For example, most workers would classify insistence on doing nothing but lining up toy cars in a precise sequence as an elaborate repetitive routine, but some would not recognise insistence on verbally retailing facts about cars, regardless of the social situation, as another manifestation of the same phenomenon.

The question remains as to whether there is any value in defining a particular 'core autism' as distinct from the rest of the continuum. There is, as yet, no answer to this. Further research is needed to define sub-groups of the spectrum that have some external validity apart from aspects of their observable behaviour. It is possible that Kanner's criteria happen to be found mainly in those who are mildly to moderately retarded and aged between 4 and 12 years of age, not because the criteria define a separate syndrome but because the level of cognitive ability determines the way in which the triad of social impairments is manifested.

All those with the triad share a common need for a structured and organised daily programme to help them cope with the demands of life.


There is no evident reason why the prevalence should be the same in different countries or why it should not change over time. The variations in prevalence rates among the published studies could be due to real differences. The higher rates among cllildren of first generation immigrants to Europe from 'exotic' countries may be one piece of evidence supporting this possibility.

However, there is a strong probability that individual interpretations of the diagnostic criteria used have affected the results of different studies, even if real variations also exist. Considering the wide variety ofpossible behavioural manifestations of the triad of impairments and the changes over time, it is not surprising that there is difficulty in achieving consistency in diagnosis among workers in different areas.

In future prevalence studies, authors should give detailed accounts of their methods, the definitions they use, the characteristics of the population surveyed and of the children selected. They should also be aware of the full spectrum of autistic conditions. Preferably, all those in the spectrum should be identified before they are sub-grouped in ways that are fully defined. It will then be easier to make valid comparisons between studies in different places and at different times.


This paper is dedicated to Victor Lotter (who died in 1988) and to Birte Hoeg Brask, the pioneers of epidemiological studies in this field. It is based on a chapter in the book published on Dr Brask's retirement in recognition of her work (Wing, L. (1988), In A.Trillingsgaard & G.Nemec (Eds.) Tilegnet Birte Hoeg Brask - Et Skrift Om Psykotiske Born. Risskov, Denmark: Psykiatrisk Bornehospital).


16 études de prevalence de l'autisme dans l'enfance faisant appel à des methodes épidémiologiques sur populations définies en Europe, aux U.S.A., Canada et au Japon, en anglais ou avec des résumés anglais ont été relevées dans la littérature publiée. Les taux varient de 3.3 à 16.0 pour 10.000. 8 études donnent des aux pour un sous-groupe d'autisme typique de 1.2 à 8.4. Les raisons de ces variations ont été recherchées en examinant les données geographiques et démographiques de la population concernée, les méthodes de repérage initial et d'examen final des enfants possiblement autistes, les données demographiques et cliniques des enfants identifiés comme autistes, et les cntères utilisés pour le diagnostic. Il y avait à l'évidence, indepéndamment des critères diagnostiques, une prévalence plus élevée parmi les enfants de première génération émigrant en Europe depuis des pays "exotiques". A part cela toutes les différences pouvaient être dues aux variations dans la pratique diagnostique et la prise de conscience croissante des manifestations des conditions autistiques au travers des écarts d'intelligence, depuis les formes sévèrement retardées jusqu' à la moyenne et au-dessus. Cependant, il reste possible, bien que non prouvé, que les taux varient selon les lieux et l'époche. Les problèmes de définition d'un sous-groupe avec autisme typique parmi le spectre large des enfants avec Ia triade de défauts d'interaction sociale, de communication et d'imagination sont discutés ainsi que la valeur d'un tel sous-groupe.


In der englishisprachigen Facliliteratur (mindestens die Zusammenfassung in englischer Sprache) fanden sich 16 Prävalenzstudien zum Autismus im Kindesalter unter Anwendung von epidemiologischen Methoden in definierten Bevölkerungsstichproben aus Europa, USA, Kanada und Japan; die altersspezifischen Raten variierten zwischen 3,3 und 16,0/10.000. Acht Studien gaben Raten für "typischen" Autismus an: 1,2 bis 8,4/10.000. Nach Ursachen für die unterschiedlichen Prävalenzangaben wurden kritisch gesucht unter Prüfung der geographischen und demographischen Details der gescreenten Bevölkerungsstichproben, der Methoden des initialen Screenings und der endgülltigen Untersuchung von möglicherweise autistischen Kindern, der demographischen und klinischen Details von Kindern, die aus autistisch identifiziert wurden, und der diagnostischen Kriterien. Es fanden sichunabhängig von diagnostischen Kriterien - Hinweise für eine höhere Prävalenz unter Kindern von Eltern, die aus "exotischen" Ländern nach Europa eingewandert waren. Hiervon abgesehen, könnten alle Prävalenzunterschiede auf verschiedenen diagnostischen Vorgehensweisen und einem erhöhten BewuBtsein für die Manifestationen autistischer Störungen über die gesamte IQ-Verteilung (von schewer geistig behindert bis durchschnittlich bzw. überdurchschnittlich) beruhen. Es ist jedoch möglich, obgleich bislang nicht bewiesen, daß die Raten von der Zeit bzw. dem Ort abhängen. Die Probleme, die sich aus der Definition der Untergruppe "typischer Autismus" ergeben, werden diskutiert angesichts eines weiten Spektrums von Kindern mit Beeinträchtigungen der sozialen Interaktion, der Kommunikation und der Phantasie. Der Sinn dieser Untergruppenbildung wird in Frage gestellt.


American Psychiatric Association (1980). Diagnostic and Statistical Manual of Mental Disorders (3rd ed.). Washington: APA

American Psychiatric Association (1987). Diagnostic and Statistical Manual of Mental Disorders (3rd ed., revised). Washington: APA.

Asperger, H. (1944). Die autistischen Psychopathen im Kindesalter. Archiv für Psychiatric und Nervenkrankheiten, 17, 76-136.

Aussilloux, Ch., Collery, F. & Roy,J. (1989). Épidémiologié de l'autisme infantile dans le département de l'Hérault. Revue Francaise de Psychiatrie, 7, 24-28.

Bohman, M., Bohman, I., Björck, P. & Sjöholm, E. (1983). Childhood psychosis in a northern Swedish county: some preliminary findings from an epidemiological survey. In M. Schmidt and H. Remschmidt (Eds.), Epidemiological Approaches in Child Psychiatry II (pp.164-173). New York: Thieme-Stratton.

Brask, B. H. (1972). A prevaleuce investigation of childhood psychoses. In Nordic Synaposium on the Comprehensive Care of the Psychotic Children (pp. 145-153).

Oslo: Barnepsykiatrist Forening.

Bryson, S. E., Clark, B. S. & Smith, I. M. (1988). First report of a Canadian epidemiological study of autistic syndromes. JournaI of Child Psychology aud Psychiatry, 29, 433-446.

Burd, L., Fisher, W. & Kerbeshian, J. (1987). A prevalance study of pervasive developmental disorders in North Dakota Journal of the American Academy of Child and Adolescent Psychiatry, 26, 704-710.

Cialdella, P. & Mamelle, N. (1989). An epidenuological study of infantile autism in a Freuch Department (Rhône): A research note. Joimial of Child Psychology aud Psychiatry, 30, 165-176.

Denckla, M. B. (1986). Editorial: New diagnostic criteria for autism aud related behaviour disorders:

Guidehues for research protocols.Journal of the American Academy of Child aud Adolescent Psychiatry, 25,


Frith, U. (Ed.). (1991). Autism and Asperger Syndrome.Cambridge: Cambridge Uiiiversity Press.

Fombonne, E. & du Mazaubran, C. (1992). Prevalence of infantile autism in four French regions. Social Psychiatry and Psychiatric Epidemiology, 27, 203-209.

Gillberg, C. (1984). Infantile autism and other childhood psychoses in a Swedish urban region: Epidemiological aspects. Journal of Child Psychology and Psychiatry, 25, 35-43.

Gillberg, C. (1992). The Emmanuel Miller Memorial Lecture 1991. Autism and autistic-like couditions:

Subclasses among disorders of empathy. Journal of Child Psychology and Psychiatry, 33, 813-842

Gillberg, I. & Gillberg, C. (1989). Asperger syndrome: Some epidemiological considerations., Journal of Child Psychology and Psychiatry 30, 631~38.

Gillberg, C., Persson, E., Grufman, M. & Themner, U. (1986). Psychiatric disorders in mildly and severely mentally retarded urban children and adolescents: Epidemiological aspects. British Journal of Psychiatry, 149, 68-74

Gillberg, C. & Schaumann, H. (1982). Social class and infantile autism. Journal of Autism and Developmental Disorders, 12, 223-228.

Gillberg, C., Steffenburg, S., Borjesson, B. & Andersson, L. (1987). lnfantile autism in children of immigrant parents: A population-based study from Göteborg, Sweden. British Journal of Psychiatry, 150, 856-857.

Gillberg, C., Steffenburg, S. & Schaumann, H. (1991). Is autism more common now than ten years ago?

British Journal of Psychiatry, 158, 403~09.

Haga, H. & Miyamoyo, Y. (1971). A survey on the actual state of so-called autistic children in Kyoto prefecture.

Japanese Journal of Child Psychiatry, 12, 160-167.

Hoshino, Y., Yashima, Y., Ishige, K., Tachibana,R., Watanabe, M., Kancko, M., Kumashiro, H., Ueno, B., Takahashi, E. & Furukawa, H. (1982). The epidemiological study of autism in FukushimaKen. Folia Psychiatrica et Neurologica japonica 36, 115-124.

lshii,T. & Takahashi, O. (1983). The epidemiology of autistic children in Toyota, Japan: Prevalence. Japanese Journal of Child and Adolescent Psychiatry, 24, 311-321.

Kanner, L. (1943). Autistic disturbauces of affective contact. Nervous Child, 2, 217-250.

Kanner, L. (1973). Childhood Psychosis: Initial Studies atid New Insights. New York: Winston/Wiley.

Kanner, L. & Eisenberg, L. (1956). Early infantile autism 1943-1955. American Journal of Orthopsychiatry, 26, 55-65.

Lane, H. (1977). The Wild Boy of Ave yron. London: Allen and Unwin.

Lotter, V. (1966). Epidemiology of autistic conditions in young children: I. Prevalence. Social Psychiatry, 1, 124-137.

Lotter, V. (1967a). Epidemiology of autistic conditions in young children: II. Some characteristics of the parents and children. Social Psychiatry, 1, 163-173.

Lotter, V. (1967b). The prevalence of the autistic syndromes in children. Unpublished PhD thesis, University of London.

McCarthy, P., Fitzgerald, M., and Smith, M. (1984). Prevalence of childhood autism in Ireland. Irish Medical Journal, 77, 129-130.

Matsuishi, T., Shiotsuki, Y., Yoshimura, K., Shoji, H., Imuta, F. & Yamashita, F. (1987). High prevalence of infantile autism in Kurume City, Japan. Journal of Child Neurology, 2, 268-271.

Mises, R. (1990). Classification Francaise des troubles mentaux de l'enfant et de l'adolescent. Neuropsychiatrie de l'Enfance, 38, 523-539.

Nakai, M. (1971). Epidemiology of autistic children in Gifu-ken. Japanese Journal of Child Psycitiatry, 12, 262-266.

Ritvo, E.R., Freeman, B.J., Pingree,C., Mason-Brothers, A., Jorde, L., Jenson, W. R., McMahon, W.M., Peterson, P. B., Mo, A. & Ritvo, A. (1989). The UCLA-University of Utah epidemiological study of autism: Prevalence. American Journal of Psychiatry, 146, 194-245.

Rutter, M. (1970). Autistic children: Infancy to adulthood. Seminars in Psychiatry, 2, 435-450.

Rutter, M. (1978) Diagnosis and definition. In M. Rutter & E. Schopler (Eds.), Autism:A Reappraisal of Concepts and Treatment (pp.1-25). New York: Plenum Press.

Steffenburg, S. & Gillberg, C. (1986). Autism and autistic like conditions in Swedish rural and urban areas: A population study. British Journal of Psychiatry, 149, 81-87.

Steinhausen, H., Göbel, D., Breinlinger, M. & Wohllebhen, B. (1986). A community survey of infantile autism. Journal of the American Academy of Child Psychiatry, 25, 189-189.

Sugiyama,T. & Abe, T. (1989) The prevalence of autism in Nagoya, Japan: A total population study. Journal of Autism and Developmental Disorders, 19, 87-96.

Tanino, Y. (1971). An investigation into infantile autism and autistic children in Toyama prefecture. Japanese Journal of Child Psychiatry, 12, 150-159.

Tanoue, Y., Oda, S., Asano, F. & Kawashima, K. (1988). Epidemiology of infantile autism in the Southern Ibaraki, Japan. Journal of Autism and Developmental Disorders, 18, 155-167.

Treffert, D. A. (1970). Epidemiology of infantile autism. Archives of General Psychiatry, 22, 431-438.

Wing, L. (1979). Mentally retarded children in Camberwell (London). In H. Hafner (Ed.), Estimating Needs for Mental Health Care (pp.107-112). Berlin:SpringerVerlag.

Wing, L. (1980). Childhood autism and social class: A question of selection. British Journal of Psychiatry, 137, 410~417.

Wing, L. (1981a). Asperger's syndrome: A clinical account. Psychological Medicine, 11, 115-130

Wing, L. (1981b). Sex ratios in early childhood autism and related conditions. Psychiatry Research, 5, 129-137.

Wiug, L. (1988). The continuum of autistic disorders. In E. Schopler & G.M. Mesihov (Eds.), Diagnosis and Assessment In Autism (pp.91-110). New York: Plenum.

Wing, L. & Gould, J. (1979). Severe impairments of social interaction aud associated abnormalities in children Epidemiology and classification. Jounial of Atism and Childhood Schizophrenia, 9, 11-29.

Yamazaki, K., Yamashita, I., Suwa, N., Kuroda, T., Iwahuchi, J., Imamura, S., Miyamoto, M., Fujino, T., Ito, N. & Sugaya,K. (1971). Survey on the morbidity rate of "autistic children" in the Hokkaido district. Japanese Journal of Child Psychiatry, 12, 141-149.

© Hogrefe & Huber. Used by permission.

Back to NAS Surrey Branch Welcome Page