The rise in serum prolactin concentration in patients treated with neuroleptic drugs is well documented, but attempts to relate this rise to clinical response have yielded conflicting results. These conflicting results could be explained by design flaws in those studies attempting to relate prolactin to clinical response. Seventy-three newly (re)admitted drug-free schizophrenic men were randomly assigned to receive haloperidol either 5, 10, or 20 mg daily for 4 weeks. Prolactin levels post-treatment were significantly (p less than 0.02) related to global outcome by logistic regression. A serum prolactin level may be a useful guide to the lowest effective dose of haloperidol in newly treated schizophrenic men. Above a plasma prolactin level of approximately 30 ng/ml there was very little increase in response. Patients on the 5, 10, and 20 mg daily haloperidol doses had mean prolactin levels of 16, 32, and 34 ng/ml, respectively.