The volume of the human tibialis anterior (TA) muscle was estimated in vivo by ultrasonography (ULT) and by magnetic resonance imaging (MRI) in six subjects. In both methods, 11 axial scans were taken along the muscle belly, and the cross-sectional area of the muscle in each scan was digitised. Muscle volume was calculated by treating the muscle as a series of truncated cones. To assess the reproducibility of the ULT method, all subjects were scanned twice. A high test-retest reliability was found ( R(2)=0.99), with the two ULT measurements being significantly correlated with each other ( P<0.05). The ULT and MRI methods gave similar results [mean (SD) ULT: 133.2 (20) cm(3); MRI: 131.8 (18) cm(3)]. Nevertheless, a systematic bias of 3.33 cm(3) and a random error of 3.53 cm(3) were found when using the ULT method compared with the MRI method, which results in an error of -0.15% to 5.17%. We conclude that the ULT method is a reproducible and valid method for the estimation of human muscle volume.