Autobiographical memory was examined in participants with seasonal affective disorder (SAD). In Experiment 1, participants with SAD performed an autobiographical memory task (AMT) in the winter, when depressed. The AMT required participants to generate autobiographical memories to positive and negative cue words. Symptom levels were reassessed in the summer, when participants were remitted. The number of overly general memories to positive cues generated when the SAD participants were depressed predicted symptom levels when remitted, over and above initial symptom levels, with greater winter overgenerality being associated with high levels of summer symptoms. However, this was dependent on the exact measure of depressive symptoms used. The degree of overgenerality of memories in SAD participants was further investigated in Experiment 2. Results revealed that SAD participants did not show elevated recall of overgeneral memories relative to controls. The results as a whole indicate that, even when levels of general memories are no greater in a given target group than in controls, the absolute level of general memories to positive cue words is still independently related to symptom outcome.