This prospective study tested the self-complexity buffering hypothesis that greater self-complexity moderates the adverse impact of stress on depression and illness. This hypothesis follows from a model that assumes self-knowledge is represented in terms of multiple self-aspects. As defined in this model, greater self-complexity involves representing the self in terms of a greater number of cognitive self-aspects and maintaining greater distinctions among self-aspects. Subjects completed measures of stressful events, self-complexity, depression, and illness in two sessions separated by 2 weeks. A multiple regression analysis used depression and illness at Time 2 as outcomes, stressful life events and self-complexity at Time 1 as predictors, and depression and illness at Time 1 as control variables. The Stress X Self-Complexity interaction provided strong support for the buffering hypothesis. Subjects higher in self-complexity were less prone to depression, perceived stress, physical symptoms, and occurrence of the flu and other illnesses following high levels of stressful events. These results suggest that vulnerability to stress-related depression and illness is due, in part, to differences in cognitive representations of the self.