This study explores the relationships among personality, cognitive variables, depressive subtype and outcome in hospital treated patients with unipolar major depressive disorder (MDD). Beck et al. (1983) hypothesized that two personality dimensions, sociotropy and autonomy, may be associated with different types of triggering events for depression, different clinical characteristics, and different treatment responses. Sociotropic individuals are regarded as highly dependent individuals whose self-esteem is governed by the support received from others. Highly autonomous individuals show greater self-investment and act to maximize control over their environment (Beck et al., 1983). Research supports the congruence hypothesis that highly sociotropic subjects may become depressed when they experience events that disrupt gratification of social needs (Clark et al., 1992). Peselow et al. (1992) reported that high levels of sociotropy or autonomy may be associated with nonendogenous or endogenous depression, respectively, and that autonomous individuals respond significantly better to antidepressants than sociotropic subjects. When measured during a depressive episode, high levels of dysfunctional beliefs predict symptom persistence (e.g., Brittlebank et al., 1993 ; Williams et al., 1990). Scott et al. 1 suggest that poor problem-solving ability is associated with a worse outcome in depression. Also, Hagaa et al. (1995) demonstrated that a positive problem orientation is inversely related to severity of depressive symptoms and interpersonal dependency. A subtype of depression is usually defined by interview assessment or neuroendocrine tests. Interview ratings of symptom profile do not always predict outcome (Carney, et al., 1965 ; Peselow et al., 1992). Research into hypothalamic-pituitary-adrenal axis dysregulation demonstrates that 30% to 50% of MDD patients have hypercortisolaemia, loss of diurnal rhythmicity, and dexamethasone nonsuppression (Charlton and Ferrier, 1989). These abnormalities are more strongly associated with endogenous symptom profile but do not consistently predict treatment response (Charlton and Ferrier, 1989).