To clarify the plasma level/therapeutic response relationship of haloperidol (HPDL) we used a prospective double-blind design in 95 acutely psychotic patients. After drug washout, patients were randomly assigned to a low, middle or high plasma level range for 2 weeks (phase A), and then 50% of the initial non-responders were randomly reassigned into the putative therapeutic range for an additional 2 weeks (phase B). There were no significant differences in clinical outcome between the three plasma level ranges in phase A. However, in phase B initial non-responders displayed greater improvement in the middle range than in the low or the high ranges. No further benefit was observed when plasma levels were raised to or maintained in the high range.