Improving performance, whether on the tennis court or in front of the piano, often means reducing the variability of our actions. Yet, no matter how hard we practice, generating identical movements on successive trials is virtually impossible.
But why is it so hard to tame performance variability?
Harvard scientists mainly focused on this question as to the center of a recent study. By using an enormous amount of data from about 3 million rat trials, scientists discovered that rats regulate their motor variability based on the outcomes of the most recent 10 to 15 attempts at a task.
What if previous trials were poor?
In case, if previous trials were poor, the rats found to increase their amount of variability by using a try-anything approach. On the other hand, if the trials were successful, the rats exposed to limit their variability. It suggests that rats to subscribe to the adage “if it ain’t broke, don’t fix it.
Ashesh Dhawale, a postdoctoral fellow in the Department of Organismic and Evolutionary Biology and first author of the study, said, “By using simulations to determine what the optimal variability regulation strategy should be, we found that it was very similar to the one used by rats. We also found that the degree to which individual rats regulated variability could predict how well they learned and performed on the motor task. This means that regulating variability based on performance is important for doing well both in short and the long term.”
For this study, scientists developed a new motor learning task for rats. They also trained rats to press a 2D joystick towards a target angle. At the point when the rats performed well, they got a sip of water. To keep the task from getting excessively simple, the analysts changed the target angle whenever the rat learned its location.
Scientists observed low variability in rats when the rats were regularly getting rewarded. Moreover, the variability increases as the trials went well. If they continued doing poorly, the variability would increase even more.
Maurice Smith, the Gordon McKay Professor of Bioengineering at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) and co-author of the paper, said, “We noticed this was happening on a pretty fast time scale. It was as if the rats were computing their batting average in real-time.”
Dhawale said, “In case of longer-term tasks with less uncertainty, We found that rats stopped regulating variability in response to recent performance, which matches what we found in our simulations. Variability regulation, in this case, had a timescale of several thousand trials, which was much slower than the reward-dependent regulation of variability that we had uncovered earlier.”
“Our results demonstrate that the brain flexibly adapts components of its trial-and-error learning algorithm, such as the regulation of variability, to the statistics of the task at hand. We have shown that the brain uses a sophisticated algorithm to regulate motor variability in a way that improves task performance.”