One hundred and thirty men and women attending psychiatric hospitals with depressive disorders were interviewed at the time of their initial contact. After a mean four month interval, 119 were reassessed in order to test the hypothesis that initial levels of social support predict clinical improvement even when other potential risk factors such as age, sex, diagnosis and severity of depression are controlled. Severity and duration of the episode emerged as the only significant background predictors of recovery. The explained variance in recovery from depression due to social support was equal in men and women, and was not diminished by the background clinical predictors. According to subset analyses however, the aspects of personal relationships and perceived support that predict recovery in men and in women appear to be different. The available multiple regression models of outcome favoured a main effect of social support and provided persuasive if inconclusive evidence for a statistical interaction effect with sex. The implications for further research and for theory are discussed.