The reproducibility of the widely used Health Habits and History Questionnaire (HHHQ) for estimating "usual past-year" nutrient intake was examined. The HHHQ was self-administered on three occasions during three different seasons; 68 women (avg age 43 yrs) provided usable data for all three questionnaires in the appropriate seasons. Intraclass correlations (ICC) among the three administrations ranged from 0.56 (carotene) to 0.82 (fat as percentage of energy), with a median of 0.72. Thus, reliability was moderate to good, and season of administration/ordinality generally had little impact on ranking of individuals. The point estimates of intake of energy and a number of nutrients were higher in the first administration (winter). Except for dietary fiber and possibly carotene, most differences disappeared when adjusted for energy using a nutrient density approach, as well as using repeated-measures regression models. The higher intake in the first administration may be due more to either learning or fatigue effect rather than an effect of seasonal food availability on perceptions of "usual" intake. These data should be used in conjunction with validity data in the future to help evaluate the gain in precision of group means (and changes in these means) and improved estimates of odds ratios and correlations between nutrients and factors such as serum values, if a questionnaire is administered more than once.