Anthropometric methods aimed at assessing muscle size in undernourished subjects assume a constant proportionality between the mass (i.e., size) and composition (specifically protein-energy content) of this tissue. This assumption was examined in three autopsy groups: controls (n = 11, sudden traumatic death), early semistarvation (n = 6), acute preterminal disease), and chronic semistarvation (n = 34, severe weight loss over time). Results of semistarved groups were expressed relative to respective control value. Early semistarvation produced no detectable change in muscle mass, protein, or total energy content (per gram wet weight), although RNA and glycogen were -50 to -70% of control value (p less than 0.05). Chronic semistarvation caused muscle atrophy (-54.2%), but not all measured constituents were reduced to the same degree. The results were H2O--52.9%, collagen--46%, noncollagen proteins--65.3%, total lipids--40%, DNA--54.1%, RNA--81.7%, glycogen--90.3%, and total energy--59.6%. Muscle per unit mass in chronic semistarvation thus reflects relatively more H2O and less protein and energy when compared to normal tissue. About 85 to 95% of muscle protein-energy loss can be detected by anthropometric measurements of muscle size; the remaining 5 to 15% depletion of protein and energy is masked by muscle compositional changes. Proper interpretation of anthropometric data requires an understanding of these unmeasured but important compositional differences in normal and semistarved muscle.