The primate wrist is an ideal model system for studying the reference frames in which movements are coded within the central nervous system, as a simple rotation of the forearm allows dissociation between extrinsic and body-referenced coordinates. Important information regarding coordinate frame transformations has been obtained using this system, particularly from studies involving extracellular cortical and spinal recordings from monkeys. Because preferred directions of muscle use were reported to rotate by less than half of the joint rotation, the system was considered to dissociate three reference frames: extrinsic (direction of movement in space), muscle (activity of muscles), and joint (angle of the wrist joint). However, given the relatively minor changes in reported muscle biomechanics with human forearm rotation, the reported distinction between joint space and muscle space is surprisingly large. Here, we reassessed patterns of wrist muscle activity with changes in forearm posture in humans, during an isometric force-aiming task with a device that enabled stringent control of the musculoskeletal configuration. Results show that the preferred directions for wrist muscle activation closely follow forearm orientation (i.e., by 88%). Control experiments confirmed this, whether the hand was clamped passively by a device or grasped a handle. Furthermore, the remaining 12% discrepancy between intended changes in wrist orientation and muscle use also occurred for muscle-pulling directions obtained by intramuscular electrical stimulation. The findings prompt reconsideration of data based on the previously reported dissociation between joint space and muscle space and have critical implications for future investigations of sensorimotor transformations and their adaptation using the wrist.