This study examined an alternative approach to the acquisition of a complex motor skill, using constraints. To determine the efficacy of two alternative methods of learning tennis strokes, forehand and backhand strokes were practiced over five successive days using two contrasting training methods. The novel, multiple movement method required alternate forehand and backhand strokes, whereas the traditional, simple movement method required repeated forehand or backhand strokes as separate series of events. Five subjects were randomly assigned to each of the two conditions. Their strokes were recorded and analyzed kinematically before and after the training sessions. The simple movement method resulted in no changes in the range of trunk rotation perse, but the center of the range shifted unfavorably, in the direction opposite to that in which the ball was struck, for both forehand and backhand strokes. By contrast, the multiple movement method increased the range of trunk rotation, and the center of the range shifted favorably in the direction toward which the ball was struck, for both strokes. These differences were confirmed by the trajectories observed in hyper-cylindrical phase space as dynamical systems. In the multiple movement group, the forehand and backhand clusters converged after training, whereas in the simple movement group the two clusters diverged after training. From a dynamical systems perspective, we argue that the multiple movement method achieves its superiority by exploiting the inertia of the trunk rotation movement as a constraint that is produced by the preceding striking action, whether it be forehand or backhand.