OBJECTIVES:: To investigate whether specific domains of cognitive functioning predict the natural course of depressive symptoms in older people. DESIGN AND Participants: Using the nationally representative, population-based cohort of the Longitudinal Aging Study Amsterdam, 281 participants with clinically relevant depressive symptoms (Center for Epidemiological Studies Depression Scale 16) aged 55 years and older were followed longitudinally during a period of 6 years. MEASUREMENTS:: Using a maximum of 14 successive Center for Epidemiological Studies Depression Scale observations, three clinical course types of depressive symptoms were defined. At baseline, general cognitive functioning was assessed using the Mini-Mental State Exam, memory performance (immediate recall and retention) by means of the auditory verbal learning test, and processing speed by means of a timed coding task. Results: Remission, fluctuating course, and chronic course were seen in 22%, 50%, and 28%, respectively. In univariate analyses, a slowed processing speed was associated with a chronic course of depressive symptoms, as compared with remission (mean: 21.5, SD: 6.6, versus mean: 24.6, SD: 6.8, t = 2.78, df = 139, p < 0.001). Using multivariate regression techniques, this association remained significant after correcting for potential confounders and a number of risk factors for vascular brain damage (odds ratio: 1.08, 95% confidence interval: 1.01-1.14). Neither global cognitive functioning nor memory performance was associated with any course type of depressive symptoms. Conclusion: We found an independent association of a slowed processing speed with a poor natural course of depressive symptoms in older people. In clinical practice, when dealing with an older depressed person with comorbid cognitive decline, processing speed might be a more useful tool than the Mini-Mental State Exam in predicting the prognosis. © 2011 American Association for Geriatric Psychiatry.