Project HeartBeat! is a four year mixed-longitudinal study of the development of cardiovascular risk factors in White and African American children who at baseline comprised three age cohorts 8, 11, and 14 years. This paper focuses on the anthropometric variables which were chosen to reflect body fat and fat-free mass. Selected anthropometric dimensions are compared with those of samples from the combined National Health and Nutrition Examination Surveys I and II to explore the similarities of the samples in terms of central tendencies and variances. The measurements were then explored in terms of their ability to estimate the two compartment model of body composition: fat-free mass (FFM) and body fat (BF) from bioelectrical impedance (BIA). Project HeartBeat! children are slightly larger than NHANES children and have variances that are generally comparable to the national surveys. Over seven percent (7.7%) of children were overweight (BMI) and 25% had 'mild obesity' by %BF. Three different factor analytic methods (incomplete principal components, alpha and maximum likelihood) produced two latent variables from 17 anthropometric dimensions which together accounted for 76-83% of the variation: (1) A body mass factor (F1) which was weighted highly on six circumferences, weight and six skinfolds, and (2) a linear growth factor (F2) which was strongly associated with height, arm length, and sitting height. Triceps, subscapular and midaxillary skinfolds were consistently highly loaded on the body mass factor and their sum was highly correlated to %BF and fat mass (0.90-0.99). This suggests that this sum could be used to estimate fatness in children in studies where the BIA or other body composition techniques are unavailable. FFM and %BF were predicted from the anthropometric factors. Both factors contributed to the estimate of FFM (R(2) = 0.81-0.93), although F2 contributed proportionately more. The 'body mass' factor (F1) was the main predictor of %BF (R(2) = 0.86-0.93), though at some ages the linear factor (F2) was significantly and negatively related to %BF. This set of anthropometric dimensions, taken for the purpose of estimating body composition and summarized as two latent vectors by factor analysis, strongly reflects body fat and FFM in children and adolescents. Am. J. Hum. Biol. 11:69-78, 1999. Copyright 1999 Wiley-Liss, Inc.