Two behavior genetic research strategies have been utilized to understand gene influences in autism. There is overwhelming evidence for gene involvement, although an exact mode of inheritance has not yet been elucidated. Family and twin studies illustrate that the clinical phenotype of autism is not sufficient to characterize the underlying genotype(s) involved. Exactly what should be included in the phenotype remains elusive. Cognitive and social deficits are indicated as milder variants of the autism phenotype, but precisely how to define these deficits requires further research. Furthermore, more complex models of inheritance (e.g., two-locus models--multifactorial and major gene) may be necessary to explain gene influences in autism. Genetic heterogeneity is indicated in autism, with an X-linked disorder, fragile X, and an autosomal dominant disorder, tuberous sclerosis, together accounting for perhaps 8% to 11% or more of cases of autism. Differences in family patterns (i.e., recurrence risks) of neuropsychiatric disorders between autism with and without mental retardation or other clinically defined groups (e.g., males and females) are suggested. Whether these differences represent genetic heterogeneity or multifactorial inheritance with varying thresholds (e.g., of severity or sex differences) cannot be distinguished on the basis of the data available to date. Autosomal recessive inheritance is suggested in a subgroup of families with autism, but the proportion of all autism that may be accounted for by autosomal recessive inheritance is unknown. Evidence exists that stoppage occurs in families with autism, however, and this can affect accurate estimates of segregation ratios when not taken into account. Future family studies need to report (1) exact ascertainment schemes and specification of probands and (2) sex and birth order of affected siblings, including sibship size, so that data may be pooled and such effects can be tested. Investigations of populations with fragile X or tuberous sclerosis as well as those with autism (without known genetic disorders) will identify the etiologic basis of these associations. Such associations may be due to linkage of genes underlying autism and those underlying the known genetic disorders (i.e., linkage disequilibrium) or shared brain pathophysiology or merely shared overt behaviors. Until such mechanisms are elucidated, we can use only empiric risk figures in genetic counseling situations of autism, assuming that no known genetic or environmental cause is identified. Pooling available data from family and twin studies, the following empiric risks are suggested for genetic counseling purposes. An average sibling risk (frequency of affected siblings among all siblings) based on pooled data is 3% (i.e., 57/1698).(ABSTRACT TRUNCATED AT 400 WORDS)