The prevalence and outcome of persistent tardive dyskinesia (TD) was studied in 131 bipolar patients. There were 34 cases of persistent TD in the subgroup (n = 96) with a history of neuroleptic treatment (prevalence, 35.4%; 95% confidence interval, 25% to 45%); there were no cases of persistent TD in the subgroup (n = 35) without such treatment history. Except in one patient, signs of TD persisted in spite of lithium carbonate treatment in 23 patients (median duration, 16 months; range, five to 24 months), of whom 15 remained off of a neuroleptic regimen during the study period for a median duration of 14 months (range, four to 24 months). Using multiple regression analysis, two variables were found to predict the presence of persistent TD and account for 36% of the variance: longer cumulative duration of maintenance neuroleptic treatment and shorter duration of previous lithium carbonate treatment. There appears to be a significant risk of persistent TD among neuroleptic-treated bipolar patients. High-risk subgroups within this category need to be identified.