Research is reviewed on the association between alcohol outcome expectancies and consumption which has led many to argue that manipulating expectancies might be a route to manipulating consumption for problem prevention and treatment. Studies indirectly and directly evaluating this latter position are reviewed. Expectancies predicting treatment outcome: two studies have shown that the more positive expectancies held at treatment, the poorer is treatment outcome, but five other studies have failed to find this. Three related studies have shown that the more negative expectancies held at treatment, the better the treatment outcome. This evaluation provides evidence inconsistent with the main position for positive expectancy and limited support for negative. Expectancy manipulations and ad libitum consumption: three studies in the laboratory have shown that increasing positive expectancies through word priming increases subsequent consumption and two studies have shown that increasing negative expectancies decreases it. A single study in the field showed a similar relationship. This evaluation provides evidence consistent with the main position but is limited by measuring consumption changes over only 1-2 hours. Prevention programmes with expectancy components: seven projects are reviewed in which positive expectancies were targeted, but only two report an expectancy change analysis and in both cases the expectancy change did not relate to subsequent consumption. This evaluation provides evidence inconsistent with the main position. Expectancy challenge: two related studies are reviewed in which positive expectancy challenges reduce subsequent consumption but changes in expectancy were not evaluated as predictors of consumption change. Two studies are reviewed which found a reduction in positive expectancy following expectancy challenge but no reduction in consumption. One study is reviewed in which when negative expectancy was increased in treatment there was a better treatment outcome at 3 months follow-up than when it was not. This evaluation provides evidence inconsistent with the main position for positive expectancy and limited consistent evidence for negative. It is concluded that the research has still to be done that securely links expectancy manipulations with subsequent changes in consumption, and fulfils the early promise from association studies.