Muscle fiber numbers were estimated in vivo in biceps brachii in 5 elite male bodybuilders, 7 intermediate caliber bodybuilders, and 13 age-matched controls. Mean fiber area and collagen volume density were calculated from needle biopsies and muscle cross-sectional area by computerized tomographic scanning. Contralateral measurements in a subsample of seven subjects indicated the method for estimation of fiber numbers to have adequate reliability. There was a wide interindividual range for fiber numbers in biceps (172,085-418,884), but despite large differences in muscle size both bodybuilder groups possessed the same number of muscle fibers as the group of untrained controls. Although there was a high correlation between average cross-sectional fiber area and total muscle cross-sectional area within each group, many of the subjects with the largest muscles also tended to have a large number of fibers. Since there were equally well-trained subjects with fewer than normal fiber numbers, we interpret this finding to be due to genetic endowment rather than to training-induced hyperplasia. The proportion of muscle comprised of connective and other noncontractile tissue was the same for all subjects (approximately 13%), thus indicating greater absolute amounts of connective tissue in the trained subjects. We conclude that in humans, heavy resistance training directed toward achieving maximum size in skeletal muscle does not result in an increase in fiber numbers.