Abstract. Cocoa-related products like chocolate have taken an important place in our food habits and culture. In this work, we aim to examine the relationship between chocolate consumption and cognitive decline in an elderly cognitively healthy population. In the present longitudinal prospective study, a cohort of 531 participants aged 65 and over with normal MiniMental State Examination (MMSE; median 28) was selected. The median follow-up was 48 months. Dietary habits were evaluated at baseline. The MMSE was used to assess global cognitive function at baseline and at follow-up. Cognitive decline was defined by a decrease ≥ 2 points in the MMSE score between evaluations. Relative risk (RR) and 95% confidence interval (95%CI) estimates were adjusted for age, education, smoking, alcohol drinking, body mass index, hypertension, and diabetes. Chocolate intake was associated with a lower risk of cognitive decline (RR = 0.59, 95%CI 0.38-0.92). This protective effect was observed only among subjects with an average daily consumption of caffeine lower than 75 mg (69% of the participants; RR = 0.50, 95%CI 0.31-0.82). To our knowledge, this is the first prospective cohort study to show an inverse association between regular long-term chocolate consumption and cognitive decline in humans.