Poorly defined cohorts and weak study designs have hampered cross-cultural comparisons of course and outcome in schizophrenia. To describe long-term outcome in 18 diverse treated incidence and prevalence cohorts. To compare mortality, 15- and 25-year illness trajectory and the predictive strength of selected baseline and short-term course variables. Historic prospective study. Standardised assessments of course and outcome. About 75% traced. About 50% of surviving cases had favourable outcomes, but there was marked heterogeneity across geographic centres. In regression models, early (2-year) course patterns were the strongest predictor of 15-year outcome, but recovery varied by location; 16% of early unremitting cases achieved late-phase recovery. A significant proportion of treated incident cases of schizophrenia achieve favourable long-term outcome. Sociocultural conditions appear to modify long-term course. Early intervention programmes focused on social as well as pharmacological treatments may realise longer-term gains.