To test mechanisms underpinning action prediction, we directly controlled experience in a dart-throwing training study. A motor-visual group physically practiced throwing darts and a perceptual training group learned to associate dart throw actions (occluded video clips) with landing outcomes. A final control group did not practice. Accuracy was assessed on related prediction tests before and after practice (involving temporally occluded video clips). These tests were performed while additionally performing simple, action-incongruent secondary motor tasks with either the right (observed throwing arm) or left effector, in addition to an attention control task. Motor proficiency tests were also performed. Although both trained groups improved their prediction accuracy after training, only the motor-visual group showed interference associated with the right-arm secondary motor task after practice. No interference was shown for the left-arm motor task. These effects were evidenced regardless of whether predictions were made in response to video stimuli or static clips. Moreover, improvements on the motor proficiency test were only shown for the motor-visual group. These results show evidence in support of motor simulation processes during action prediction among observers with motor experience. Prediction accuracy can be achieved via nonmotor processes (for the perceptual group), but there was no evidence that physically experienced performers could effectively switch processes to maintain prediction accuracy.