Objective: The therapeutic alliance is recognized as an important contributor to treatment outcomes. In this study, the session-to-session interplay of the alliance (as perceived by the patient) and alcohol involvement (drinking days and heavy drinking days between successive treatment sessions) was examined. The analyses also tested the extent to which pretreatment changes in drinking altered these interrelationships. Method: Participants (N = 63) seeking treatment for an alcohol use disorder received 12 weeks of CBT for alcohol dependence and completed weekly assessments of the alliance. Results: Higher session alliance scores at a given session significantly predicted lower alcohol involvement (both drinking days and heavy drinking days) in the period until the next treatment session, controlling for previous alcohol involvement. This relationship was further moderated by pretreatment change (changes in drinking before the first treatment session). Among those who demonstrated low pretreatment change, alliances continued to predict alcohol involvement. In contrast, alliances were not associated with alcohol involvement among those who significantly reduced their drinking before the first treatment session (high pretreatment changers). Finally, alcohol involvement during the period preceding a treatment session did not significantly predict alliance ratings. Conclusions: These data demonstrate that more positive patient ratings of the alliance at any given treatment session are associated with less alcohol involvement during the period until the next treatment session, most particularly among patients who have not initiated reductions in their drinking before the first treatment session. For such patients, efforts to maximize therapeutic alliances may be warranted and productive.