Muscle activity was related to the direction of a static force at the human wrist. For each muscle the force direction of maximal activity and the directional tuning characteristics were determined. Electromyographic (EMG) activity was recorded from nine superficial elbow and/or shoulder muscles while subjects held the right arm stationary in one of six postures. The direction of the force at the wrist was varied in two orthogonal planes. In each experiment a cable was attached to the subject's wrist, and a constant force magnitude was applied in various directions with the use of a pulley system. The relationship between the averaged EMG level and the force direction was described for each muscle, in each posture, and in each plane. The EMG data were fit with a nonlinear, multiple cosine function, which allowed the identification of one, two, or sometimes three separate cosine peaks. Two-cosine functions often provided the best fit to the EMG data. All nine muscles were best fit with a two-cosine function in at least two of the six postures. Four of the muscles had a second peak of activity in more than one-half of the experimental situations. The second peak was often in a direction that was nearly opposite the direction of the first peak and represented a negative contribution to the total force produced at the wrist ('coactivation'). We suggest that multimodal directional tuning results from the convergence of multiple sources of descending signals onto motoneurons. The mechanical actions of nine elbow and/or shoulder muscles were estimated with the use of published data from a cadaver study by Wood and co-workers. Postural changes in the mechanical actions of muscles were substantial. A 45° rotation of the shoulder, for example, might cause a 30-50° change in the direction of force at the wrist that could be produced by the contraction of a given muscle. The magnitude of these postural changes suggests that arm position is an important determinate of EMG patterns. Postural changes in the direction of maximal EMG activity usually paralleled the postural changes in mechanical pulling direction. Postural changes in EMG amplitudes usually covaried with postural changes in mechanical advantage. The posterior deltoid (PD) was an exception to the general rule of covariation of mechanical actions and EMG activities. Instead of reflecting the muscle's mechanical action, the EMG activity of the PD closely resembled the EMG activity of the medial deltoid (MD). An analysis of the correlation between the activities of PD and MD suggested that the activities are tightly coupled by the CNS, in spite of the muscles' dissimilar mechanical actions. In the Back/Forward (or sagittal) plane, shoulder muscles can produce a force at the wrist in only one of two opposite directions. Thus, for forces in this plane, one might expect EMG activity to be a single cosine function of force direction. In the Lateral/Medial (or frontal) plane, however, the mechanical action of each muscle has two orthogonal components: medial/lateral and up/down. When a pure medial or lateral force at the wrist is desired, unwanted upward or downward forces also occur. The necessity to work within this mechanical framework may explain the deviations from single cosine directional tuning.