This paper aims to discuss which one is the best estimator of past diet: a retrospective report or a recent diet recall. The analysis included 13 articles published between 1984-1997 and selected from a MEDLINE search and from other reviews on this subject. The selection criterion was the use of a food frequency questionnaire (FFQ) in a validation study of retrospective report of dietary intake in remote past. Literature review shows that even taking into account misclassification, retrospective report of diet usually yields to a more reliable estimate of past diet pattern than current report. Past diet recall was strongly influenced by current intake and by diet patterns change. Analyzed investigations indicate that agreement between original and retrospective report was higher either for foods eaten rarely or frequently and lower for foods moderately consumed. This review allows considering the FFQ as a valuable instrument when studying the role of diet on the etiology of chronic diseases.