Although the clinical presentation and course of schizophrenia is highly variable, it is unclear whether this reflects heterogeneity at an aetiological level. The genetic evidence is reviewed concerning 'traditional' clinical subtypes as more novel categories derived from multivariate statistical methods and Crow's type I-type II classification. Recent data based on a twin series and re-analysis of older published family material suggest that attempts to divide up schizophrenia have resulted in splits between two or more categories of disorder which occupy different positions on the same continuum of liability. Thus, apparent heterogeneity is more likely to be due to quantitative difference in familial-genetic loading rather than qualitative differences. Similarly, the hypothesis that schizophrenia can be broadly divided into two groups, one genetic and the other non-genetic has little to support it. It seems improbable that any further useful and genetically relevant subdivision of schizophrenia can be effected on purely clinical grounds. Further progress awaits developments in the discovery of endophenotypes and the application of molecular genetic marker strategies.