The long-term recall of height, weight, and body build was examined as part of a 1988 follow-up of the Boston-based Third Harvard Growth Study (1922-1935). By design, approximately 50 percent of subjects had body mass indices during adolescence of greater than the 75th percentile, and approximately 50 percent of subjects had body mass indices between the 25th and 50th percentiles. Interviewed subjects, aged 71-76 years (n = 181) were asked to recall their high school height and weight and to select outline drawings that best reflected their body size at ages 5, 10, 15, and 20 years. Reports at ages 71-76 years were compared with measurements obtained during childhood and adolescence. High school weight was overestimated by men who were lean as adolescents (mean recalled minus measured difference = 2.5 kg, 95 percent confidence interval 0.78-4.3) and underestimated by women who were obese as adolescents (mean difference = -2.3 kg, 95 percent confidence interval -4.8 to 0.21). Pearson crude correlations between recalled body build and body mass index measured at approximately the same ages were between 0.53 and 0.75 for all ages studied except for a correlation of 0.36 among males at age 5 years. Adjustment for current body mass index only slightly reduced the correlations between adolescent body mass index and recalled build. Adolescent obesity classifications based on indicators derived from recalled weight and build had moderate sensitivities (37-57 percent) and high specificities (98-100 percent). These results indicate that the remote recall of height, weight, and body size can contribute useful information independent of current weight status.