ABSTRACT
Twin and family studies provide strong evidence that autism has a largely genetic aetiology. The pattern of familial aggregation suggests that in individual families, a small number of genes act together to cause the phenotype. However, it is unlikely that the same genes act in all families. Thus, the total number of genes involved could be large. One key to finding genes for disorders with considerable locus heterogeneity is to detect genetically more homogeneous subsamples. There exist several traits in families who have a child with autism--biochemical, physical, or behavioural--that are likely to reflect underlying genetic heterogeneity and can thus be used to divide families into more homogeneous subsets. These traits (1) show variation in autism samples; (2) are found in non-autistic family members more often than controls; (3) aggregate in particular autism families; and (4) result in increased signals when used in linkage analysis to define 'affected'.
