1. The question we addressed was the following: what rules does the CNS employ, given the initial and final positions of the arm for a pointing movement, to decide which shoulder and elbow muscles ("agonists") to activate for initiating movement? 2. Widely varying initial and final positions were used, so that the movements studied encompassed much of the reachable work space within the horizontal plane. For each movement, the initial electromyographic (EMG) activity at each joint was classified qualitatively in terms of the "sign," i.e., flexor or extensor muscle activity, and quantitatively in terms of the integral of the rectified EMG. 3. The sign of initial muscle activity at each joint was found to be related to the angular excursions at both joints during movement to the final position. 4. Two different hypothesized rules, derived from previously proposed strategies for control of multijoint limb movements, were tested for their ability to predict correctly the sign of initial muscle activity at each joint. Although four variables are needed to describe the initial and final positions of a two-segment arm, only two combinations of these four were relevant for testing the rules. The two positional variables were the spatial direction of the final tip position with respect to the initial forearm orientation (psi) and the initial elbow angle (theta Einit). 5. According to one of the rules tested, which was based on statics, the initial muscle activity at each joint should be such that the distal tip of the limb exerts an initial force in the direction of the final tip position. Our data concerning the sign of initial shoulder muscle activity clearly contradicted this rule in two distinct regions of the (psi, theta Einit) plane. 6. According to the other rule tested, which was based on dynamics, the initial muscle activity at each joint should be such that the initial acceleration of the distal tip is in the direction of the final tip position. The data contradicted the predicted sign of initial shoulder muscle activity for a certain range of psi. This shows that the activation of muscles is not always appropriate even qualitatively for a straight-line path. Furthermore, the effects of added inertial loads predicted by this rule were not observed for trials in which a 1.8-kg mass was attached to the distal portion of the limb.(ABSTRACT TRUNCATED AT 400 WORDS)