Following a 2-week placebo lead-in, schizophrenic patients were randomly assigned to fluoxetine 20 mg/day or placebo added to depot neuroleptic for a 6-week, double blind trial. All patients had received a stable dose of depot neuroleptic for at least 6 months and did not meet criteria for depression. Serum samples were obtained at baseline and at weeks 4 and 6. Scores on the negative symptom subscale of the Brief Psychiatric Rating Scale (BPRS) were significantly lower at week 6, controlling for baseline scores, in patients receiving fluoxetine (n = 20) compared to patients receiving placebo (n = 21). Measures of psychosis, depression, global functioning and extrapyramidal symptoms (EPS) did not differ between groups at week 6. Fluoxetine administration was associated with a mean 65% increase in serum fluphenazine concentrations in 15 patients and a mean 20% increase in serum haloperidol concentrations in three patients. The change in negative symptoms at week 6 did not correlate with serum concentrations of fluoxetine or norfluoxetine, but did inversely correlate with S-norfluoxetine, an active stereoisomer of fluoxetine. For these chronically ill patients, fluoxetine significantly improved negative symptoms and did not worsen EPS, despite causing substantial elevation in serum concentrations of neuroleptics.