he individual muscle forces in the leg during human walking are unknown, because of a greater number of muscles when compared to degrees of freedom at the joints. The muscle force-joint torque equations can be solved, however, using optimization techniques. A linear programming solution of these equations applied at discrete, time-independent steps in the walking cycle using dynamic joint torque data is presented. The use of this technique, although capable of providing unique solutions, gives questionable muscle force histories when compared to electromyographic data. The reasons for the lack of confidence in the solution are found in the inherent limitations imposed by the linear programming algorithm and in the simplistic treatment of the muscles as tensile force sources rather than complex mechanochemical transducers. The definition of a physiologically rationalized optimal criterion requires both a global optimization approach and more complete modelling of the system.