The objective of our study was to elucidate distinct paths to depression in a model that incorporates age, measures of medical comorbidity, neuroanatomical compromise, and cognitive status in a sample of patients with late-life major depressive disorder (MDD) and nondepressed controls. Our study was cross-sectional in nature and utilized magnetic resonance imaging (MRI) estimates of brain and highintensity lesion volumes together with clinical indices ofClinically significant mood disorders are among the most common psychiatric illness in the elderly (Blazer et al. 1987;Parmalee et al. 1992;Henderson et al. 1997). They increase psychosocial and economic morbidity and contribute to the relatively high rate of suicide in the elderly Wells et al. 1989;Unutzer et al. 1997;Penninx et al. 1999). Depression in the elderly is consistently associated with medical illness, and this relationship has been observed in virtually all clinical settings including acute, long-term care, and ambulatory programs (Coulehan et al. 1990;Lustman et al. 1992;Caine et al. 1993;Gierz and Jeste 1993;Katz et al. 1994;Lacro and Jeste 1994;Frasure-Smith et al. 1995;Katz 1996;Borson et al. 1998). Depression is associated with a broad spectrum of medical disorders including cardiovascular, cerebrovascular, musculoskeletal, metabolic, and pulmonary illnesses and malignancies (Coulehan et al. 1990;Lustman et al. 1992;Caine et al. 1993;Gierz and Jeste 1993;Katz et al. 1994;Lacro and Jeste 1994;Frasure-Smith et al. 1995;Katz 1996;Zubenko et al. 1997;Borson et al. 1998). Acknowledging this robust connection between medical disorders and depression, the National Institutes of Health consensus conference on depression in late life concluded, "The hallmark of depression in the elderly is its association with medical comorbidity"(National Insti- tutes of Health 1992). Despite these observations, the precise nature of the relationship between mood and medical disorders and the pathways that lead from medical illness to clinical depression remain elusive.The application of magnetic resonance imaging (MRI) to the study of late-life depression has provided information on the neuroanatomical substrates of depression (Rabins et al. 1991;Coffey et al. 1993;Krishnan 1993;Sheline et al. 1996;Kumar et al. 1997aKumar et al. , 1997bKumar et al. , 1998Lai et al. 2000). Late-life depression is associated with smaller focal brain volumes together with increased MRI-determined high-intensity lesion in the brain (areas that appear bright on T2 and proton density weighted images) when compared with nondepressed controls. The decreases in brain volume appear to be relatively focal and have been described in the prefrontal lobe, hippocampus, and the caudate nucleus (Krishnan 1993;Sheline et al. 1996;Kumar et al. 1998). The high-intensity lesions are also distributed in the neocortical and subcortical areas. The increased prevalence of high-intensity lesions in the brain together with their putative relationship to vascular disease in some samples has led to a resurgence of...