Plasma amyloid β (Aβ) levels have been associated with an increased risk of Alzheimer's disease (AD). As depression is common before the onset of AD, a few clinical studies tested the cross-sectional association of Aβ levels with depression in elderly and showed incongruous findings. Hence, we tested the longitudinal association between Aβ levels and depressive symptoms in community-dwelling elderly. The study is embedded in a population-based cohort of 980 participants aged 60 years or older from the Rotterdam Study with Aβ levels. Participants were evaluated for depressive symptoms with the Centre for Epidemiological Studies-Depression scale at baseline and repeatedly over the mean follow-up of 11 years. We first performed cross-sectional analyses. Then, we tested the longitudinal association between Aβ levels and depressive symptoms after excluding participants with dementia during follow-up. In cross-sectional analyses, persons with high Aβ(1-40) levels had more clinically relevant depressive symptoms. However, this association was accounted for by persons with clinically relevant depressive symptoms who developed dementia within the next 11 years. In longitudinal analyses, persons with low levels of Aβ(1-40) and Aβ(1-42) without dementia had a higher risk of clinically relevant depressive symptoms during the follow-up. These findings suggest that the cross-sectional association between high plasma Aβ levels and clinically relevant depressive symptoms in the elderly is due to prodromal dementia. In contrast, the longitudinal association between low plasma Aβ levels and depressive symptoms could not be explained by dementia during follow-up suggesting that Aβ peptides may play a distinct role on depression etiology.