AbstractAccuracy of methods to assess past diet in retrospective studies is of concern. Adjustment of current intake for subjects’perceptions of change is one alternative to estimate past intake. The authors studied the potential utility of perceived change in diet in residents of a rural Japanese community who participated in three sequential surveys with self‐administered questionnaires. Frequencies of consumption for 27 foods were investigated in the first and second questionnaires administered in 1988 and 1993 (n = 451), and perceived changes in their frequencies during this period were asked in the second and third questionnaires surveyed within two weeks (n = 214). Perceptions of change in diet were highly reproducible. At a group‐level analysis, they were consistent with the changes in food frequencies in 21 of the 27 items. Nevertheless, whereas current diet adjusted for perceived change correlated better with past intake than did current intake alone in 17 foods, improvement in Pearson's correlations was only marginal (median = 0.02). Perceptions of change in diet would therefore be of limited utility in improving estimate of past intake among individuals in retrospective studies.