Anthropometry is frequently used for nutritional assessment. Little is known in infants about the validity of anthropometric measurements in relation to whole-body fat (TBF) and fat-free mass (FFM) composition. We compared TBF and FFM estimations by total-body electrical conductivity (TOBEC) with anthropometry in 435 healthy infants ages 21-365 d. TBF was best correlated with weight-for-length and calf circumference (r2 = 0.84, r2 = 0.83). FFM was best correlated with body weight (r2 = 0.93). Upper-arm anthropometry, skinfold thickness, and Quetelet's and Ponderal indexes were poorly correlated with TBF and FFM (r2 < 0.65). New anthropometry-based prediction equations were calculated (r2 = 0.90 for TBF and r2 = 0.95 for FFM). New simple indexes (analogous to Quetelet's index) were calculated for TBF (weight x calf circumference/length; r2 = 0.87) and for FFM (square root of weight x length; r2 = 0.95). Prediction equations and indexes were cross-validated in a second population by a second observer. Interobserver variation was largest for equations with skinfold thicknesses included. We conclude that anthropometry can be used for rough estimations of infant body composition, although indexes different than those used in children and adults are preferred.