To summarize quantitatively the literature comparing hypothalamic-pituitary-adrenal (HPA) axis function between depressed and nondepressed individuals and to describe the important sources of variability in this literature. These sources include methodological differences between studies, as well as demographic or clinical differences between depressed samples. The current study used meta-analytic techniques to compare 671 effect sizes (cortisol, adrenocorticotropic hormone, or corticotropin-releasing hormone) across 361 studies, including 18,454 individuals. Although depressed individuals tended to display increased cortisol (d = 0.60; 95% confidence interval [CI], 0.54-0.66) and adrenocorticotropic hormone levels (d = 0.28; 95% CI, 0.16-0.41), they did not display elevations in corticotropin-releasing hormone (d = 0.02; 95% CI, -0.47-0.51). The magnitude of the cortisol effect was reduced by almost half (d = 0.33; 95% CI, 0.21-0.45) when analyses were limited to studies that met minimal methodological standards. Gender did not significantly modify any HPA outcome. Studies that included older hospitalized individuals reported significantly greater cortisol differences between depressed and nondepressed groups compared with studies with younger outpatient samples. Important cortisol differences also emerged for atypical, endogenous, melancholic, and psychotic forms of depression. The current study suggests that the degree of HPA hyperactivity can vary considerably across patient groups. Results are consistent with HPA hyperactivity as a link between depression and increased risk for conditions, such as diabetes, dementia, coronary heart disease, and osteoporosis. Such a link is strongest among older inpatients who display melancholic or psychotic features of depression.