In 45 physically active men (ages 35-67 yr) who underwent hydrostatic weighing to determine body composition, multiple regression equations were developed for the prediction of body density (D), lean body weight (LBW), fat body weight (FBW), and % fat using selected anthropometric measurements. The prediction accuracy for these parameters using several previously generated anthropometric regression equations was also determined. With equations developed from the present data a substantially higher correlation was obtained between measured and predicted LBW (r = 0.95) than between measured and predicted D (r = 0.85), FBW (r = 0.88), or % fat (r = 0.84). When previously developed equations were applied to the present sample, correlations between measured and predicted values were considerably lower (4-42%) than in the original studies; this reduction was least in the case of LBW. Analysis of previous data indicated that in selected populations total body weight can account for a relatively large fraction of the variance in LBW. LBW may be estimated quite accurately (r greater than or equal to 0.90) in physically active men with one of several regression equations which include total body weight as an independent variable.