Aims To test two central assumptions of the Transtheoretical Model (TTM) regarding recovery from alcohol problems: (i) individuals making a forward transition from pre-action to action stages will show greater drinking improvements than those remaining in pre-action stages; and (ii) individuals remaining in pre-action stages will not demonstrate improvements in drinking outcomes. Design and setting Large, multi-centre, randomized controlled trial of treatment for alcohol problems [United Kingdom Alcohol Treatment Trial (UKATT)]. Measurements Stage of change, drinks per drinking day and percentage days abstinent at baseline, 3- and 12-month follow-ups. Findings In support of TTM assumption 1, improvements in drinking outcomes were consistently greater among clients who showed a forward stage transition (Cohen's d = 0.68) than among those who did not (d = 0.10). Two tests of assumption 2 showed a significant improvement in drinking outcomes in non-transition groups, inconsistent with the TTM; one test showed a significant deterioration and the other showed equivalent drinking outcomes across time. An explanation is offered as to why, under the relevant assumption of the TTM, clients in non-transition groups showed small changes in drinking outcomes. Conclusions In contrast to a previous study by Callaghan and colleagues, our findings largely support the TTM account of recovery from alcohol problems in treatment. The discrepancy can be explained by the use in our study of a more reliable and valid method for assigning stage of change.