In order to evaluate the repeatability of nutrient values estimated from a semiquantitative food frequency questionnaire being used in a longitudinal study of the relationships between diet, hemostatic factors, and stroke risk in the elderly in Western Sydney, a subsample of 62 participants (24 men, 38 women) completed a repeat questionnaire approximately 1 month after baseline data were collected. The mean age was 78 years (range, 65 to 88; median, 78). Nutrient values calculated from the repeat questionnaire were not significantly different from the baseline results by paired t test. Intraclass correlation coefficients ranged from 0.63 for beta carotene to 0.82 for saturated fat. Quadratic weighted kappa values were calculated for quintile categories and these ranged from 0.50 for fiber to 0.86 for ethanol. These values are comparable to previously published results in elderly subjects and confirm that repeatability of nutrient intakes estimated using semiquantitative food frequency questionnaires is very high in the elderly. Older subjects may be more established in their dietary habits than younger subjects, so any tendency for repeatability to decrease due to impaired memory associated with advanced age is offset by a lower intraindividual variability in dietary habit.