After individual determination of neuroleptic threshold (NT) doses of haloperidol, 106 patients with schizophrenia or schizoaffective disorder (Research Diagnostic Criteria) were treated openly at such doses (mean, 3.7 +/- 2.3 mg/d) for 2 weeks. Ten responding patients were discharged and unavailable for follow-up or refused subsequent randomization, and one non-responding patient refused randomization. The remaining 95 responding or nonresponding patients were then randomly assigned, double-blind, to a dosage of haloperidol two to 10 times higher (mean, 11.6 +/- 4.7 mg/d) or to a continuing NT dosage (mean, 3.4 +/- 2.3 mg/d) for another 2 weeks. Of the 58 patients exposed only to NT dosages of haloperidol, 72% clinically recovered within the 5-week trial. Higher dosages given to 47 patients did not lead to greater improvement in measures of psychosis, but did produce slightly greater declines in measures of hostility. Higher dosages did regularly lead to significant increases in distressing extrapyramidal side effects.