Background. There is some uncertainty about the rate and correlates of autism.

Method. Twenty-three epidemiological surveys of autism published in the English language between 1966 and 1998 were reviewed.

Results. Over 4 million subjects were surveyed; 1533 subjects with autism were identified. The methodological characteristics of each study are summarized, including case definition, case-finding procedures, participation rates and precision achieved. Across surveys, the median prevalence estimate was 5·2/10000. Half the surveys had 95% confidence intervals consistent with population estimates of 5·4–5·5/10000. Prevalence rates significantly increased with publication year, reflecting changes in case definition and improved recognition; the median rate was 7·2/10000 for 11 surveys conducted since 1989. The average male/female ratio was 3·8[ratio ]1, varying according to the absence or presence of mental retardation. Intellectual functioning within the normal range was reported in about 20% of subjects. On average, medical conditions of potential causal significance were found in 6% of subjects with autism, with tuberous sclerosis having a consistently strong association with autism. Social class and immigrant status did not appear to be associated with autism. There was no evidence for a secular increase in the incidence of autism. In eight surveys, rates for other forms of pervasive developmental disorders were two to three times higher than the rate for autism.

Conclusion. Based on recent surveys, a minimum estimate of 18·7/10000 for all forms of pervasive developmental disorders was derived, which outlines the needs in special services for a large group of children.