The aim of our study was to investigate whether depression assessed by different markers predicts the risk of incident dementia in elderly individuals. Data was derived from the 3C cohort study conducted in community-dwelling individuals aged 65 years and over, randomly recruited from electoral rolls of three French areas and followed up for four years (1999-2001 through 2004) with assessments every 2 years. The study sample comprised 7989 dementia-free individuals (mean age, 74.0 years; 61% women) assessed at baseline for current or past Major Depressive Episodes (MDE) based on MDE module of the Mini International Neuropsychiatric Interview, self-reported lifetime treated depression, and level of depressive symptoms using the Center for Epidemiologic Studies-Depression Scale. Fully adjusted Cox proportional hazards models were applied to examine the risk of incident dementia associated with these markers of depression. MDE and self-reported lifetime treated depression did not increase incident dementia risk. Conversely, high level of depressive symptoms at baseline was associated with a 50% increased risk of dementia (adjusted Hazard Ratio [HR], 1.5; 95% Confidence interval [CI], 1.2-2.2). This result was driven by a five-fold increased risk of vascular dementia (HR, 4.8; 95% CI, 2.2-10.7; p < 0.0001), whereas there was no increased risk of Alzheimer's disease (1.0; 0.7-1.6). In elderly individuals, high level of depressive symptoms is predictive of vascular dementia within a few years. This close temporal association suggests that depression is less a risk factor for than a prodromal symptom of vascular dementia.