It has previously been suggested that providing variable experience during motor practice is crucial for learning, retention, and transfer of motor skills. However, it is not clear what should be varied to optimize learning. In this study, we investigate whether practicing to modulate or regulate performance parameters, during practice of a virtual shuffleboard task, leads to better transfer of learning between tasks. Our results indicate that during practice, the parameter that was held constant improved the most during initial learning, and transferred best to a task requiring modulation of that parameter. Translation of this improvement into task performance depended upon the sensitivity of each task to changes in each variable. We conclude that during variable practice, learning to regulate, not modulate task variables leads to optimal learning and transfer. Further, optimal transfer occurs when variable practice elicits synergistic covariation between performance parameters.