Oct. 5, 2005 — Practice makes perfect when people learnbehaviors, from baseball pitching to chess playing to public speaking.Biomedical engineers at Washington University in St. Louis have nowidentified how people use individual experiences to improve performance.
KurtThoroughman, Ph.D., Washington University assistant professor ofbiomedical engineering, and Jordan Taylor, Washington Universitydoctoral student in biomedical engineering, tested a dozen volunteerswho played a video game that involved a robotic arm. Thoroughman andTaylor found that the subjects learned different levels of the game injust 20 minutes of training over different environmental difficulties.Human subjects made reaching movements while holding a robotic arm withperturbing forces that changed directions at the same rate, twice asfast, or four times as fast as the direction of movement, thereforeexposing subjects to environments of increasing complexity acrossmovement space. Subjects learned all three environments and learned thelow and medium complexity environments equally well. They learned thehigh complexity environment, too, though not as well as the other two.
Thoroughmanand Taylor also could detect how individual movements trained people tomake the next movement better. Surprisingly, people could very quicklychange the way errors in one movement induced a learned response in thenext movement. Specifically, subjects lessened theirmovement-by-movement adaptation and narrowed the spatial extent ofgeneralization to match the environmental complexity, showing thatpeople can rapidly reshape the transformation of sense into motorprediction to best learn a new movement task.
"We've demonstratedthat the richness of motor training determines not only what we learnbut how we learn," Thoroughman said. "What we cared about most was notonly what people learned but how they learned from trial to trial,movement to movement.
"The big picture is that in a singlesitting people changed their expectations of the complexity of theworld, in that a single movement's experience could be generalized verybroadly or else generalized very narrowly. We've shown for the firsttime that the learning process itself is flexible. People can figureout that in this particular environment: 'I need to change the way Ilearn, movement to movement.'"
The researchers published their findings in the Sept. 28, 2005 issue of Nature Neuroscience.
Ultimately,these findings can help researchers devise better diagnostic tools forpeople with neurological disorders and could lead to betterneuro-prosthetic devices
Thoroughman and Taylor then modeled thisadaptation using a neural network. They found that, to mimic humanbehavior, the modeled neuronal tuning of movement space needed tonarrow and reduce gain with increased environmental complexity.
Accordingto Thoroughman, prominent theories of neural computation havehypothesized that neuronal tuning of space, which determinesgeneralization, should remained fixed during learning so that acombination of neuronal outputs can underlie adaptation simply andflexibly. Thoroughman showed that this tuning of space is insteadflexible.
"We challenged those well-known theories with evidencethat the neuronal tuning of movement space changed within minutes oftraining our subjects.," Thoroughman said.
"The overall researchgoal for this part of my lab is to characterize as carefully aspossible how people abstract information from single movements to makethe very next movement better. I want to understand more about howhuman motor control goes right before I understand how to help peoplewhen it goes wrong."
Cite This Page: