Evidence of an optimal error rate for motor skill learning
When acquiring a motor skill, learners must practice the skill at a difficulty that is challenging but still manageable to gradually improve their performance. In other words, during training, the learner must experience success as well as failure. Does there exist an optimal proportion of successes and failures to promote the fastest improvements in skill? Here, we build on a recent theoretical framework for optimal machine learning, extending it to the learning of motor skills. We then designed a custom task whose difficulty dynamically changed along with human subjects performance, constraining the error rate during training. In a large behavioral dataset (N=192), we observed evidence that learning is greatest at around a [~]30% error rate, matching our theoretical predictions.