Thigh muscle volume is a useful determinant of functional fitness. However, anthropometric prediction of muscle content is influenced by the variability of adipose tissue accumulation. The aims of this study were to predict thigh muscle and adipose tissue volumes from anthropometry and to assess the validity of the method by examining the various components of the measurements and the assumptions involved. The 19 participants (9 men, 10 women; age 23-49 years) varied in adiposity. They all underwent magnetic resonance imaging (MRI) of the upper leg and the eight men and two women with the lowest adiposity underwent detailed anthropometry involving girths and skinfolds. Using MRI as the reference method, muscle volume was predictedfrom anthropometry using a circular concentric model, and the assumptions inherent in the method were tested further using the MRI data alone. Muscle volume was best predicted by anthropometry in the 10 leanest participants using a five-slice truncated cone model that overestimated the mean MRI value by 30% (R2 = 0.95; standard error of estimate = 288 cm3; P < 0.001). A single skinfold plus girth measurement at the mid-thigh almost matched its predictive ability, but with an increased bias. Measurements of leg circumference by means of the two techniques agreed well. The assumption of a circular cross-section was valid. In contrast, the agreement between skinfold thickness measured by caliper and superficial adipose tissue thickness by MRI was poor, contributing to the scatter of fat and lean area comparisons. An anterior skinfold thickness measurement underestimated the area of superficial adipose tissue at that level, particularly at the most proximal and distal sites. Although these limitations increase the uncertainties of muscle volume determination by anthropometry, they do not prevent its valid prediction in leaner individuals. The prediction of superficial adipose tissue was poorer.