This study examines the predictive validity of several clinical variables—including marital distress, individual symptomatology, and family‐of‐origin experiences—on the formation of the alliance in couple therapy. Eighty people who were treated with a naturalistic course of integrative conjoint psychotherapy at a large midwestern outpatient clinic were assessed on the clinical variables before session 1. They also completed ratings of the therapeutic alliance after sessions 1 and 8. Individual symptomatology did not predict alliance formation at either treatment stage. Higher levels of marital distress predicted poorer alliances to treatment between partners at session 1. Marital distress also predicted therapeutic alliance quality for men and women at session 8. Family‐of‐origin distress predicted alliance quality for men at session 1, and for women at session 8. Family‐of‐origin distress for men and women predicted split alliances early in treatment, and marital distress predicted split alliances for women at session 8. Clinical implications for the assessment and treatment of couples are discussed.