Aim: To explore the effect of acamprosate and naltrexone on craving and alcohol consumption in the treatment of alcohol dependence. Design: A randomized, double-blind, single-dummy, placebo-controlled trial. Setting: Three treatment centres in Sydney, Australia. Participants: A total of 169 alcohol-dependent subjects were given naltrexone (50 mg/day), acamprosate (1998 mg/day) or placebo for 12 weeks, in conjunction with manualized medication compliance therapy. Intervention: During the course of the trial, participants kept a daily diary which included the number of standard drinks they consumed and their peak craving for alcohol that day rated on a 0-10 scale. Measurements: Subjective ratings of daily craving and daily drinking for the first 6 weeks of treatment. Findings: Mixed/hierarchical linear models were employed on an intention-to-treat basis. Analyses revealed that craving was a significant predictor of daily drinking and baseline levels of depression were the best predictor of daily craving. There was no significant improvement in model fit when treatment group was added both in models of daily craving and daily drinking. Daily alcohol consumption was best predicted by a model incorporating baseline dependence and depression scores, and daily craving, entered as a time-varying covariate. However, there was a significant craving × time × treatment interaction (t = -3.365, df = 4413.712, P < 0.001), suggesting that at higher levels of craving drinking was reduced at a significantly greater rate with naltrexone compared to acamprosate. Conclusions: Naltrexone had a greater effect on drinking when craving was high. These results support the role of naltrexone in reducing craving when that craving is highly salient. The role of acamprosate in reducing craving was not supported by these findings.