Several serotonin uptake inhibitors, including the long-acting fluoxetine, have been found to decrease alcohol intake in moderately dependent alcoholics. While the mechanism of their effect is not fully elucidated, a previous study with citalopram indicated that decreased desire to drink may be an important factor. Therefore, we tested fluoxetine effects on alcohol intake and desire to drink in a placebo-controlled study. Subjects, recruited by advertisement, were mildly/moderately dependent alcoholics (12 male, four female, aged 19-59 years, healthy, non-depressed) who did not believe they had a drinking problem and were not requesting treatment. After a 1 week baseline they received, single-blind, 2 weeks placebo followed by 2 weeks fluoxetine 60 mg/day. As out-patients, subjects recorded daily standard drinks (13.6 g ethanol) and rated interest, desire, craving and liking for alcohol biweekly. Each out-patient period was immediately followed by a double-blind experimental drinking session. Out-patient daily drinks slightly decreased during fluoxetine to 6.6 +/- 0.9 (mean +/- S.E.M.) compared with during placebo (7.16 +/- 0.95, p = 0.07, N.S.) and baseline (7.18 +/- 1.0, p > 0.1, N.S.). Desire, interest and craving for alcohol decreased during fluoxetine vs placebo baseline (p < 0.05), but not vs placebo. Appetite loss and decrease in food intake (p < 0.01, fluoxetine vs placebo) correlated with each other (r = 0.91, p < 0.01) but neither correlated with decrease in alcohol intake (appetite: r = 0.26, N.S.; food intake: r = 0.22, N.S.). Weight loss occurred during fluoxetine (p < 0.05 vs placebo) but did not correlate with decrease in alcohol intake (r = 0.1, N.S.). In the experimental drinking sessions after placebo and fluoxetine treatments subjects rated their desire for each of 18 mini-drinks (each one-third of a standard drink) offered at 5 min intervals. Fluoxetine decreased desire to drink throughout the sessions; both mean and maximum desire ratings were lower after fluoxetine than after placebo (ANOVA, p < 0.05). Therefore, fluoxetine seems to have a robust effect on decreasing desire for alcohol. We propose that in the absence of intention by subjects to reduce drinking, their habitual drinking patterns mitigated against reduced consumption in the out-patient phase. However, fluoxetine could be a useful adjunct for patients in a treatment context who are motivated to reduce their drinking.