Because of the large variability in the proportion of fiber types within a whole muscle, a single biopsy is a poor estimator of the fiber type proportion for a whole muscle. Data on the proportions of type I and II fibers, obtained from cross sections of whole human muscles (vastus lateralis) from young male individuals, have therefore been analyzed statistically in order to determine the sampling errors involved in muscle biopsy techniques. For the purpose of obtaining a good estimate of the fiber type proportion in a whole biopsy, counting all fibers is of great benefit compared with counting only half of the fiber number. The required number of biopsies to obtain a given sampling error of the mean proportion of fiber types in the whole muscle can vary by a factor of six. If less than three biopsies are taken from a muscle, there is a substantial reduction in sampling error taking biopsies with at least 600 fibers. For more than three biopsies there is a small gain in sampling greater than 150 fibers. The precision of the estimate of the mean proportion of fiber types for a group is increased with the number of biopsies per individual and number of individuals. In conclusion, for the muscle in this study, complete counting of three biopsies, each greater than 150 fibers, sampled from different depths of the muscle is recommended