Few studies have prospectively examined dietary patterns and adult weight change, and results to date are inconsistent. This study examines whether a Mediterranean diet (MD) pattern is associated with reduced 3-y incidence of obesity using data from the Spanish cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC-Spain). The sample included 17,238 women and 10,589 men not obese and aged 29-65 y at baseline (1992-96). Height and weight were measured at baseline; weight was self-reported in a follow-up survey a mean of 3.3 y later. Detailed dietary history data, collected using a validated method, were used to construct a MD score. Logistic regression models were used to estimate odds of becoming overweight or obese. Among initially overweight subjects, 7.9% of women and 6.9% of men became obese, whereas 13.8% of normal weight men and 23.0% women became overweight. High MD adherence was associated with significantly lower likelihood of becoming obese among overweight subjects, with stronger associations after adjusting for underreporting of dietary data. Associations (odds ratios with 95% CI) were similar in women (0.69, 0.54-0.89) and men (0.68, 0.53-0.89). Adjusting for the plausibility of reported dietary intakes increased the magnitude of these associations, which were approximately 0.8 without this adjustment. MD adherence was not associated with incidence of overweight in initially normal-weight subjects. Nonetheless, results suggest that promoting eating habits consistent with MD patterns may be a useful part of efforts to combat obesity.