Twenty-five inpatients with acute exacerbations of schizophrenia (by Research Diagnostic Criteria) or schizoaffective disorder underwent a prospective haloperidol dosing procedure and were assigned fixed doses chosen to yield a distribution of haloperidol plasma levels above and below a hypothesized upper therapeutic limit of 18 ng/ml. Changes in Brief Psychiatric Rating Scale scores after 1 week of treatment were negatively correlated with haloperidol plasma levels, and the statistically optimum cutoff point fell near the predicted 18 ng/ml. Plasma level/response relationships over the subsequent 3 weeks were weaker but patients with higher plasma levels had consistently less improvement.