The therapeutic alliance has a rich history in individual therapy where it has been consistently associated with gains in therapy. Despite this consistent link between alliance and outcome, theorizing and research on the alliance in couple therapy has lagged behind. One reason for this is that the non-independence of data from partners in couple therapy has limited the ability of researchers to address dyadic questions without violating the assumptions of traditional analytic models. This study uses recent analytical advances to examine the therapy alliance using the couple rather than the individual as the unit of analysis. It examines the effect of each partners alliance on his own outcome (actor effect), and on his partners outcome (partner effect). Additionally, this study evaluates whether there are sex differences in the strength of these effects on outcome. Finally, this study examines the differential effects of the between and within systems alliance as well as the goals, tasks and bonds subscales of the alliance on distress. 173 couples receiving treatment for marital distress at 2 university clinics completed the Revised Dyadic Adjustment Scale, a measure of relational distress, and the Outcome Questionnaire-45.2, a measure of individual distress, prior to treatment and again following the 4 session of therapy. The Couple Therapy Alliance Scale was administered following session 4. The actor-partner interdependence model was used to simultaneously regress each partners level of distress at session 4 on 1st session distress and alliance. Results provide support for actor effects on relational distress for both husbands and wives, and on individual psychological distress for wives. Limited support was found for partner effects on distress. The results of this study failed to support the hypothesis that there are sex differences in the relationship between alliance and distress. Models evaluating the relative importance of the between and within systems alliances on distress indicate that the alliance between partners is a stronger predictor of improvement than the alliance between the individual and the therapist. Finally, support was found for the importance of the tasks and goals domains of the alliance in the early sessions of therapy.