Some recent studies suggest the incidence of schizophrenia to have declined over recent decades. Were such a trend to be ‘real’, it could contain fundamental clues to the nature of schizophrenia. However, these studies are open to potentially serious confounds through reliance on hospital admission data in the face of changes in service provision and diagnostic practice over the same period. The population studied to resolve these critical issues was of all living persons with DSMIIIR schizophrenia (n = 83) from one half of Co. Cavan (population 25,178), as described previously (Youssef et al. (1991) Arch. Gen. Psychiat., 48, 254-258). Morbid risk (MR) for persons born in quinquennia from 1920/24 to 1965/69 was determined using the Strijmgren method. MR was constant for persons born in the years 1920-39 (range 7.29-8.89/1000) but fell markedly for those born in 1940-69 (range 3.96-5.96/1000). Relative risk (R) for schizophrenia in persons born in 1940-69 vs 1920-39 was 0.63, i.e. fall in MR of 37% (p=O.O62); this occurred predominantly in females (R = 0.34, p = 0.041; males: R=0.81, NS), and in cases with age at onset >25 (R=0.34, p=O.Oll; onset 125: R= 1.26, NS). There was an evident parallelism between rates of fall in MR and in infant (but not early neonatal) mortality over this period, but with a temporal incongruity of five years and no sex difference. The incidence of schizophrenia appears to have fallen for persons born after 1940, and such findings are not artefacts of changes in health service provision or diagnostic practice; the fall has been greatest for females and for late onset cases. In the present population, preliminary analysis indicates males, but not females, with a family history of schizophrenia to show a markedly younger age at onset. Thus, this relative sparing of early onset, male cases within the present overall decline may constitute the ongoing ‘purification’ of schizophrenia as an illness of high genetic loading.