Through an experimental study of the stability properties of the human neuromuscular system while it performs simple point-to-point arm movements, this paper evaluates the concepts of equilibrium and virtual trajectories as a means of executing movement of the arm. Human subjects grasped the instrumented handle of a two-link robot manipulandum and performed specified point-to-point planar arm trajectories. Computer-controlled brakes were used to subtly change the movements by constraining the trajectory along an arc of radius equal to the length of one link of the manipulandum. Target points were arranged to lie along the arc so that the subject could complete the movement even when constrained. These situations were tested: (1) unconstrained throughout the movement, (2) constrained through the entire movement, and (3) initially constrained and then released during movement. Experimental results showed that the constraint evoked significant forces strongly oriented so as to restore the hand to the unconstrained hand path. In addition, when released from the constraint, these forces caused a strong tendency to return the hand to the unconstrained path before the end of the movement was reached. Such strong positional stability properties of the arm reinforce the notion that a moving attractor point dominates the dynamics of the arm during movement. Additionally, bounds on the shape of the virtual trajectory were found which indicate that the equilibrium point remains close to the actual movement produced. These results, showing that a controlled equilibrium point may be used for planning and coordinating multijoint movements, are consistent with an equilibrium point hypothesis.