In treating an acute episode of winter depression, cognitive-behavioral therapy for seasonal affective disorder (CBT-SAD) and light therapy are comparably efficacious, with improvement in depression symptoms during CBT-SAD mediated by reduced seasonal beliefs (i.e., maladaptive thoughts about the seasons, light availability, and weather). Here, we tested whether the enduring benefit of CBT-SAD over light therapy following treatment is associated with offsetting seasonal beliefs during CBT-SAD. Currently depressed adults with Major Depression, Recurrent with Seasonal Pattern (N = 177) were randomized to 6 weeks of light therapy or group CBT-SAD and followedup one and two winters after treatment. Outcomes measured during treatment and at each follow-up included depression symptoms on the Structured Clinical Interview for the Hamilton Rating Scale for Depression-SAD Version and Beck Depression Inventory-Second Edition. Candidate mediators measured at pre-, mid-, and posttreatment were SAD-specific negative cognitions (Seasonal Beliefs Questionnaire; SBQ); general depressogenic cognitions (Dysfunctional Attitudes Scale; DAS); brooding rumination (Ruminative Response Scale-Brooding subscale; RRS-B); and chronotype (Morningness-Eveningness Questionnaire; MEQ). Latent growth curve mediation models found a significant positive path from treatment group to the slope of SBQ during treatment, with CBT-SAD showing larger improvements in seasonal beliefs with overall change in seasonal beliefs in the medium-effect range, and significant positive paths from SBQ slope to depression scores at the first and second winter follow-ups, indicating greater change towards more flexible seasonal beliefs during active treatment was associated with less severe depression symptoms following treatment. Estimated indirect effects (treatment group → SBQ change * SBQ change → outcome) were also significant at each follow-up for each outcome with β