Participants in the 1987-1989 recall of the Charleston Heart Study were asked to report their current weight and to recall their weight in 1984 and 1960. Reported weights were compared with weights measured in the respective time periods. Subjects included male and female blacks and whites between ages 62 and 100 years. Correlations between reported and measured weights over all subjects were 0.979 for current, 0.935 for 4-year, and 0.822 for 28-year recall. Subjects in the lowest body mass index quartile overestimated their weight, while subjects in the highest quartile underestimated their weight. This tendency increased as the elapsed time increased. Deviations between measured and reported weights increased as performance on cognitive tests declined.