Genes are now accepted as being important in the aetiology of schizophrenia (Gottesman & Shields, 1982; McGuffinet al,1987), and over the past decade the emphasis in genetic research has shifted away from genetic epidemiology to searching the chromosomal DNA for the genes themselves. Despite this increasing technical sophistication, the application of linkage analysis to families multiply affected by schizophrenia has been accompanied by the familiar controversy over the exact borders of the adult clinical phenotype (Sherringtonet al,1988; St Clairet al,1989). Indeed, the preoccupation of researchers with the vagaries of the clinical definition has resulted in repeated attempts to use genetic studies to determine the relative validity of different operational definitions of schizophrenia (McGuffinet al,1984; Farmeret al,1987). To us, such studies beg the question of how precisely genes are involved in the aetiology of schizophrenia; after all, genes code for proteins, not for auditory hallucinations in the third person.