Three methods of assessing problem-solving skills are described in the literature: questionnaires, self-report inventories and verbal assessment of the resolution of problem scenarios. These three approaches were used to assess problem-solving ability in a sample of 20 patients with unipolar major depressive disorder, who were treated with antidepressant medication alone at a hospital out-patient clinic. It was found that baseline deficits in problem-solving skills were significantly associated with prognosis at 3- and 6-month follow-up. Self-ratings of perceived self-control and confidence rather than subjective problem-solving effectiveness were important predictors of final outcome. Differences in perceived and actual problem-solving ability emerged over time. The reasons for these differences in subjective and observer assessments are unclear. The data appear to demonstrate that problem-solving deficits in depression represent both state and trait phenomena.