Retrospective analysis of the stringency of diagnosis and therapeutic response was carried out in 80 patients with major recurrent winter depression (SAD) who had participated in controlled light therapy trials in Switzerland from 1984 to 1990. Two groups were formed with respect to anamnestic information: patients whose previous episodes of seasonal depression had been reconstructed graphically, and those who could only globally recollect prior depressive phases. These data were taken to test conformity to DSM-III-R criteria for seasonal pattern, as well as its prognostic usefulness for response to light therapy. The more liberal ‘Rosenthal criteria’ for SAD of at least two consecutive, seasonally recurring major depressive episodes were sufficient to predict improvement with light: none of the other DSM-III-R criteria differentiated further, and few patients could remember previous depressive episodes in precise detail. Our data support suggestions to revise the restrictive SAD diagnostic criteria for DSM