The authors evaluated the trait/state issues of harm avoidance in depressive-spectrum disorders and its predictive potential for antidepressant response. Subjects with Diagnostic and Statistical Manual of Mental Disorders (4th edn; DSM-IV) major depressive disorder (n = 39), dysthymic disorder (n = 37), depressive personality disorder (n = 39), and healthy control subjects (n = 40) were evaluated with the Temperament and Character Inventory and the 17-item Hamilton Depression Rating Scale (HDRS-17) at baseline and after a 12 week antidepressant treatment period. Higher harm avoidance scores predicted lesser improvement in subjects with dysthymic disorder and major depressive disorder, as determined by lesser decrease in HDRS-17 scores. Mean harm avoidance scores in depressed subjects were consistently greater than those in healthy controls, controlling for age, gender and diagnosis. Mean harm avoidance scores decreased significantly in all depressive-spectrum disorders after treatment, but still remained higher than harm avoidance scores in control subjects. The present study reports that harm avoidance is a reliable predictor of antidepressant treatment in subjects with major depressive disorder and dysthymic disorder and that harm avoidance is both trait- and state-dependent in depressive-spectrum disorders.