Background: In younger adults, depression has been associated with hypercortisolemia. In older depressed patients, however, both low and high cortisol levels have been reported. We examined the possibility of a U-shaped association between depression and cortisol in older people, suggesting both hypo- and hyperactivity of the hypothalmic-pituitary-adrenal axis. We also examined whether this might represent different depression subtypes. Methods: This population-based study included 1185 subjects aged 65 and older. Depression was measured at both diagnostic (major depression) and symptomatic (subthreshold depression) levels of caseness. Plasma concentrations of cortisol (CORT) and corticosteroid binding globuline (CBG) were determined. From these (CORT/CBG), a free cortisol index (FCI) was computed. Results: The association between cortisol and major depression was U-shaped (B CORT = -9.50 [SE 3.85] p = .014; B CORT2 = .008 [SE .003] p = .021). Hypocortisolemic depression (lower cortisol tertile) was associated with female sex, joint diseases, and smoking. Hypercortisolemic (upper cortisol tertile) depression was associated with older age, male sex, cardiovascular diseases, nonsteroidal antiinflammatory use, and (borderline significant) cognitive impairment. Conclusions: In older people, the association between cortisol and major depression is U-shaped. Hypo- and hypercortisolemic depression may represent different depression subtypes, requiring different clinical management.