The aim of the present study was to assess reproducibility and relative validity of a self-administered FFQ used in the PREDIMED Study, a clinical trial for primary prevention of CVD by Mediterranean diet in a population at high cardiovascular risk. The FFQ was administered twice (FFQ1 and FFQ2) to explore reproducibility at 1 year. Four 3 d dietary records (DR) were used as reference to explore validity; participants therefore recorded their food intake over 12 d in the course of 1 year. The degree of misclassification in the FFQ was also evaluated by a contingency table of quintiles comparing the information from the FFQ2 and the DR. A total of 158 men and women (aged 55–80 years) were asked not to modify their dietary habits during the study period. Reproducibility for food groups, energy and nutrient intake, explored by the Pearson correlation coefficient (r) ranged 0·50–0·82, and the intraclass correlation coefficient (ICC) ranged from 0·63 to 0·90. The FFQ2 tended to report higher energy and nutrient intake than the DR. The validity indices of the FFQ in relation to the DR for food groups and energy and nutrient intake ranged (r) from 0·24 to 0·72, while the range of the ICC was between 0·40 and 0·84. With regard to food groups, 68–83 % of individuals were in the same or adjacent quintile in both methods, a figure which decreased to 55–75 % for energy and nutrient intake. We concluded that FFQ measurements had good reproducibility and a relative validity similar to those of FFQ used in other prospective studies.