Traditional models of the genetic transmission of human diseases have often assumed that the phenotype is a simple dichotomous trait, which is unrealistic for many psychiatric conditions, and may result in loss of valuable information. We describe a new model for complex phenotypes, implemented in the program COMDS, which subclassifies normal and affected individuals into polychotomies correlated with the underlying genetic liability to the disorder. The model is applied to 18 Scottish pedigrees ascertained for schizophrenia, in which auditory P300 latency had been measured as a possible correlate of the genetic predisposition to schizophrenia. The results suggest that there may be a major locus for schizophrenia, but that there are also other familial determinants, possibly a second modifier locus. In addition, the results indicate that auditory P300 latency may be a useful measure of the genetic predisposition to schizophrenia among asymptomatic relatives, although the relationship between P300 latency and the degree of genetic predisposition in clinical cases was not significant, presumably because other factors are operating on P300 latency. Because of the possible selection biases in this sample, there is a need to replicate these findings in systematically ascertained pedigrees.