In a 5-year longitudinal study in a cohort of 169 schizophrenic outpatients treated with neuroleptics, we found a twofold increase (from 22% to 44%) in prevalence of tardive dyskinesia (TD) meeting the Schooler and Kane research diagnostic criteria. If we include cases of TD that were considered definite but did not meet the research criteria, the prevalence increased from 31% to 58%. In the cohort of 131 patients who did not present with the disorder in 1975, we found parkinsonism and increase in parkinsonism to be the best predictors of subsequent development of the disorder. Poor schizophrenic prognosis and long treatment duration also appeared to be risk factors. Another finding was the importance of changes in neuroleptic and antiparkinsonian dosage in both covering and uncovering TD.