The epidemiology of first-episode psychosis is poorly understood because of the paucity of systematic studies, yet it constitutes the fundamental basis for understanding the disorder and the foundations on which clinical, biological, therapeutic, and long-term outcome studies are built. A particular need is to clarify the diagnostic breadth of first-episode psychosis and, on this basis, to undertake systematic comparisons across representative populations of the psychoses, to include comparisons with first-episode mania. Considered here is the new generation of prospective studies that may be able to inform in some way on these issues. Attainment of the above goals requires prolonged accrual of "all" cases of nonaffective, affective, and any other psychotic illness, including first-episode mania, to derive the required representative populations. To illustrate some of the challenges, the structure of the Cavan-Monaghan prospective first episode study is described and its interim findings are outlined, as rural Ireland provides psychiatric care based on strict catchment areas and is characterized by substantive ethnic and socioeconomic homogeneity and stability. It is argued that there are 3 primary diagnostic nodes (schizophrenia spectrum psychosis, bipolar disorder, and major depressive disorder with psychotic features) around which there exist numerous additional, overlapping, and well-populated diagnostic categories that are distinct only in terms of their operational definition. Only through systematic, epidemiologically based studies that access this intrinsic diversity are we likely to understand fully the origins and pathobiology of first-episode psychosis.