Psychosis is a low prevalence disorder with high cost to those affected, their families, and society in general. Enormous effort to determine the causes and pathophysiology of schizophrenia has been relatively unrewarded, and no robust biological markers have been identified. It is argued that reliance on the Kraepelinian dichotomy model of psychosis, as demonstrated in ICD-10 and DSM- IV, impedes research, especially in psychiatric genetics. Modelling the psychoses from first principles demands a population based a theoretical approach. By considering the whole spectrum of psychosis in a general population the natural boundaries of the underlying disorder(s) may be best understood. This thesis describes the use of both dimensional and categorical approaches in the same data, providing complementary approaches to delineating the psychosis phenotype. Classes and dimensions thus identified are validated by their pattern of associations with many variables previously known to be important in schizophrenia. The findings are anchored in the literature by making comparisons with traditional diagnostic categories and first rank symptoms in addition to comparison with other studies.