[Background] Psychiatric morbidity studies in developing countries have used diagnostic procedures of low reliability, without a clinica definitions of caseness, producing descriptive data with limited application for mental health planning. [Method] A two-stage cross-sectional design (with a sample size of 6476) was conducted to estimate the prevalence of DSM - III psychiatric diagnoses in three metropolitan areas of Brazil (Brasilia, São Paulo and Porto Alegre). All subjects were screened for the presence of psychopathology with a 44-item instrument (the QMPA) and a subsample was selected for a psychiatric interview. [Results] Age-adjusted prevalence of cases potentially in need of care ragend from 19% (São Paulo) to 34% (Brasília and Porto Alegre). Anxiety disorders comprised the higest prevalences (up to 18%). Alcoholism yielded the most consistent prevalence levels, around 8% in all sites. Depression showed grat variation between areas: from less than 3% (São Paulo and Brasília) to 10% (Porto Alegre). [Conclusions] Overall prevalences were hih in comparison with previous studies conducted in Brazil. A female excess of non-psychotic disorders (anxiety, phobias, somatisation and depression) and a male excess for alcoholims were consistently found.