Little is known about how psychiatric patients' source of referral relates to treatment outcome. This study examines the effect of referral source on clinical outcome with computer-aided cognitive-behavior therapy (CCBT) for anxiety and depressive disorders. Three hundred fifty-five referrals to a clinic that offered CCBT with brief backup from a clinician were classified into general practitioner (GP) referrals (34%), mental health (MH) professional referrals (42%), and self-referrals (SR, 24%), and compared on sociodemographic and clinical features and treatment outcome. At intake, referrals from all 3 sources had similar sociodemographic features and problem duration, but GP referrals had less comorbidity, whereas MH professional referrals were being treated for their problem more often and were less motivated to change than were SR. Among treatment completers, SRs had the least and MH professional referrals had the most impaired work/social adjustment. Each referral group improved on generic and syndrome-specific measures; however, GP referrals improved the most and MH professional referrals the least. The 3 groups received similar therapist support and were equally satisfied after treatment. We conclude that GP referrals had the best outcome with CCBT for anxiety/depressive disorders. Referral source can be important in psychotherapy research because it may affect the type of patient seen and may predict treatment outcome.