Three observations challenge Kraepelin's binary view of the functional psychoses: a bimodal distribution of the clinical features of manic-depressive illness and schizophrenia has not been demonstrated; affective illness appears to predispose to schizophrenia in later generations; and 'schizoaffective' illnesses cannot be separated in family studies from either of the prototypical psychoses. The alternative concept is that psychosis is a continuum extending from unipolar, through bipolar affective illness and schizoaffective psychosis, to typical schizophrenia, with increasing degrees of defect. According to this concept the genes predisposing to psychosis have a degree of stability that ensures that the form of the psychosis tends to remain the same within families, but there is also the possibility of change, implying that the genetic mechanisms themselves are variable. It is proposed that quantal changes in the 'virogene' are due to replications within the genome (e.g. the generation of tandem repeats of the element or a component of it); that such replications occur at a critical stage (e.g. gametogenesis, fertilisation, very early embryogenesis) in the course of reproduction; and that the 'season of birth effect' reflects the operation of the mechanism responsible for these replications.