The composition of a mixed fast skeletal muscle (rat extensor digitorum longus) was examined to quantify the difference between the relative number of the three major fibre types in a representative muscle and their relative contribution to muscle cross-section, i.e. numerical (NN) and areal (AA) densities, respectively. These two indices clearly differ in their physiological relevance. While the former may be useful in describing hyperplasia, the latter allows for differences in size among fibre types. When estimated as NN, over 20% of fields contained 5-10% SO fibres and less than 5% had 75-80% FG fibres. In contrast, only 2% of fields had an AA of 5-10% for SO fibres while around 30% contained 75-80% FG fibres. The importance of a direct method for estimating AA is emphasised, as an indirect approach may have an error of 20-30% when used for oxidative fibre types. The use of an unbiased sampling regime to minimise error in determining both numerical and areal densities of different fibre types is illustrated.