Muscle mass is a determining factor in skeletal muscle function and is affected by inactivity, immobilization, disease, and aging. The aim of this study was to develop an objective and time-efficient method to quantify the volume and cross-sectional area of human calf muscles using three-dimensional magnetic resonance images. We have estimated the errors incurred in muscle volume measurements arising from artifacts known to occur in magnetic resonance imaging (MRI). The largest source of error was due to partial volume effects, which resulted in overestimation of phantom volumes ranging from 145 to 900 cc by 6% to 13%. The magnitude of this effect has been shown to increase with decreasing object size and decreasing spatial resolution. We have presented a straightforward correction for this effect, which has reduced the volume measurement error to less than 4% for all cases. Through the use of computer simulations, the correction algorithm has been shown to be independent of object shape and orientation. To reduce user subjectivity, a semiautomated computer program has been developed to segment MRI data for particular muscle groups. Images from seven human subjects were analyzed by the program, yielding muscle volumes of 154.2 +/- 23.2, 281.2 +/- 35.8, and 432.2 +/- 83.7 for the lateral gastrocnemius, medial gastrocnemius, and soleus, respectively.