Data on liver enzyme elevations were collected in a retrospective study of 7,263 treatment courses with haloperidol, clozapine, perphenazine, and perazine. Charts of 233 patients hospitalized between 1980 and 1992 at Tübingen University Psychiatric Clinic were selected because clinically relevant increases of liver enzymes had been detected during monotherapy with one of the four examined neuroleptics. At least one hepatic enzyme (mostly alanine aminotransferase [ALAT]) exceeded the established reference range of 3-fold elevations of ALAT, aspartate aminotransferase, gamma-glutamyl transpeptidase, and glutamate dehydrogenase and 2-fold elevations of alkaline phosphatase (AP) during monotherapy with clozapine in 15%, perazine in 7.6%, perphenazine in 4%, and haloperidol in 2.4% of the cases. If all liver enzyme abnormalities with any elevation greater than the conventional upper limits are considered, incidences were as follows: clozapine, 78%; perphenazine, 62%; perazine, 59%; and haloperidol, 50%. Testing for overall differences within the four neuroleptics resulted in significantly different incidences of liver enzyme elevations (chi2 test,p < 0.0001). Threefold increases of AP (>540 U/L) were seen in three patients receiving haloperidol (0.3%) only. Twofold increases of AP (>360 U/L) were distributed as follows: clozapine, 1%; haloperidol, 0.8%; perazine, 0.3%; and perphenazine, 0.1%. Only in the group with 1-fold elevations of AP (>180 U/L) were the differences within the drug regimens significant (clozapine, 40.3%; haloperidol, 33.2%; perphenazine, 23.4%; and perazine, 23.1%; chi2 test, p < 0.0001). In the period under study, no instance of icterus occurred.