Previous research suggests poor social problem-solving may function as a vulnerability factor for depressive symptoms. However, the ecological validity of previous findings is questionable, with recent research using real-life performance based approaches to ascertain how solutions are implemented in the ‘real world’. The current study employed a longitudinal design to examine the role of real-life problem-solving as a predictor of future depressive symptoms. Participants completed a diary of the interpersonal problems they encountered, and their attempts to solve them. They also completed traditional measures of social problem-solving (SPSI-R and MEPS task). Real-life problem-solving performance predicted depressive symptoms 3 months after the initial testing session, beyond the variance accounted for by traditional measures of social problem-solving. This suggests that the ability to hypothetically problem-solve is distinct from the ability to generate and implement problem-solving strategies in real-life, and that it is the latter which is most important in predicting depressive symptoms.