Some skills are hard to teach with a rulebook. In a virtual reality study, people learned to sort certain visual patterns faster when they responded with whole-arm movements instead of pulling a controller trigger.
The study, published in Science Advances, found this boost only when the task required people to combine several clues to make a judgment. When they could use an easy-to-explain rule, like sorting by color or shape, whole-arm movements made little difference.
Researchers at Brown University and Florida Atlantic University used virtual reality to test whether the way people move affects how they learn to sort things into groups. In daily life, this kind of learning helps us decide whether a fruit is ripe enough to pick. Yet many theories treat movement as the last step, something that happens after the brain has made a decision.
In the first experiment, 80 volunteers wore headsets and learned to sort gray striped discs into two groups. The discs varied in the thickness and angle of their stripes. Participants were recruited near Brown University and averaged about 21 years old. Half learned a sorting task with a rule they could put into words. The other half had to weigh both features together, using a pattern that was hard to describe in words. In each task, some volunteers pulled a controller trigger, while the others swung an arm to hit the virtual disc. Everyone completed 600 practice trials with feedback.
A second experiment with another 80 volunteers, again averaging about 21, tested other explanations. Swinging might have helped because the virtual tool touched the disc or because slower responses gave people more time to look at the pattern or think. This time, volunteers making the larger movement punched a virtual box without touching the disc. Each disc stayed on screen for the same amount of time, regardless of response speed. After training, volunteers switched to the other response method for 100 final trials.
In both experiments, whole-arm movements helped volunteers learn patterns that required combining clues. In the last 100 training trials of the second experiment, the volunteers who punched the box averaged about 91 percent correct, compared with about 81 percent for those who pulled the trigger. Taking longer to respond did not explain the advantage. Slower responses were actually linked to lower accuracy. Making the rule-based task harder in the second experiment still produced no clear movement benefit, suggesting that the earlier task's simplicity was unlikely to explain the lack of a benefit.
The switch test offered another clue. Among volunteers learning to combine features, those who had trained by punching still scored nearly 90 percent after switching to the trigger. Those who had trained with the trigger and switched to punching scored about 83 percent. If the bigger movement had helped only by jogging their memory at the moment of answering, the advantage should have disappeared when they switched to the trigger. Instead, the results suggest that the benefit developed during practice.
Why movement helped only on one kind of task remains unclear. Learning to combine clues is thought to rely on brain circuits linked to movement, while learning rules draws on areas near the front of the brain that help us test ideas and hold information in mind. The authors suggest that larger movements might engage the circuits linked to movement more strongly or help people pay closer attention to the features that matter. The study did not measure brain activity, so both ideas remain untested.
The study had only 20 participants in each of the four groups in each experiment. The volunteers were mostly young adults from the area around one university, so the results may differ for other age groups. Researchers also did not record which hand each participant favored or balance the groups on that basis. They tested only striped patterns and did not examine other ways of sorting things, such as learning rules with exceptions. Whether the same benefit carries over to training outside the lab remains an open question.