The aim of this rater-blinded randomized study was to evaluate the efficacy and side effects of haloperidol and thioridazine in the treatment of new-onset psychosis in HIV-positive individuals. Participants were 13 men who had no history of psychosis prior to infection with HIV, and whose psychosis was not attributable to delirium or to non-HIV-related organic factors. Participants were evaluated at baseline after at least one month without neuroleptic treatment and then weekly for six weeks of the experimental treatment using several rating scales. The mean daily dose in chlorpromazine equivalents was 124 mg. Both neuroleptics produced modest but significant reduction in overall level of psychosis and in positive symptoms, but not in negative symptoms. All the haloperidol-treated patients developed extrapyramidal side effects and required treatment with anticholinergic medication, whereas three of the five thioridazine-treated patients had noticeable side effects. We make recommendations for the treatment of HIV-associated psychosis with neuroleptics.