The search for "classes" of movement was based on the idea (hope) that the central nervous system (CNS) might use a relatively small set of "motor programs" in generating the bewildering array of movement of which we are capable. One possibility was that the CNS functions as a "look up table" - the desired movement is determined and the appropriate "program" (in the sense of pattern of muscle activation) is simply read from the table. One of the requirements of such a mode of operation is that there should be a unique relation between muscle activation and the resulting movement. I will describe studies on arm movements in humans which first demonstrate that the "desired" characteristics of movement are not always intuitively obvious and that "identical" movements can be generated using different patterns of muscle activation. The CNS probably determines the "program" of muscle activation from computations based on a constantly updated dynamic model of the limb.