In studying the role of diet in chronic disease etioiogy, measurement of past intake may be of more relevance than measurement of that existing either at the time of or just preceding the clinicai onset of disease. in 1983, in metropolitan Adelaide, South Austraiia, estimates of current and previous dietary intake were obtained, by use of a semiquantitative food frequency questionnaire, from 70 individuals (37 male, 33 female), from each of whom an estimate of intake, using the same questionnaire, had been obtained three years eariler. Over this three-year period, median nutrient and energy intake had deciined. In relative, if not absolute terms, individuais were able to make satisfactory retrospective estimates of original nutrient and energy intake; however, their retrospective estimates appeared to have been infiuenced by their current intakes. Nevertheless, the strong correlations observed between retrospective and original estimates of intake suggest that the semiquantitative food frequency method of retrospective dietary assessment allows a reliabie relative ranking of individuals by nutrient intake within a population.