Participants in this study practiced with feedback to anticipate the left-right direction of forehand tennis shots played by stick-figure players. A technique based on principal component analysis was used to remove dynamical differences that are associated with shots to different directions. Different body regions of the stick-figure players were neutralized with this procedure in the pretests and posttests, and in the practice phases. Experiment 1 showed that training is effective if during practice information is consistently present in the whole body of the player, but not if the information is neutralized in the whole body in half of the practice trials. Experiment 2 showed that training is effective if the variance associated with the direction of the shots is consistently present in one body region but neutralized in others, and that transfer occurs from practice with information in one body region to performance in conditions with information preserved only in other regions. Experiment 3 showed that occlusion has a much larger detrimental effect on learning than the applied neutralization technique, and that transfer between body regions occurs also with occlusion. Discussed are theoretical implications for understanding how biological motion is perceived and possible applications in a type of training referred to as reduced usefulness training.