As advanced as the brain is, it makes its share of mistakes in how it directs our muscles to work. We drop things. We miss the waste basket. We drive the car over the curb.
One cause is brain and neural circuitry, with a delay occurring from the point of sensory perception to the muscular response, much like the pause between turning on the television and the picture appearing.
“Sensory-process delays cause everything to be out of date by one-tenth of a second,” said Steve Chase, an assistant professor in the department of biomedical engineering and the Center for the Neural Basis of Cognition at Carnegie Mellon University.
That’s enough delay to cause us to make incorrect movements.
How the brain contends with these delays is the topic of a CMU study published online in eLife. Rather than one’s perceptions leading directly to a motor response, the brain requires an intermediate step.
It filters perceptions through a predictive model likely involving the motor cortex and the cerebellum, to determine how best to react. So if you want to throw a ball into a hoop, the brain uses its predictive model to determine what muscle or motor action would do the trick. If that model is incorrect, as often occurs, you miss. It can explain why golfers duff shots, baseball players make errors and musicians hit sour notes.
The research led jointly by Mr. Chase, who holds a Ph.D., and Byron Yu, associate professor in CMU’s department of electrical and computer engineering, and its department of biomedical engineering, explains how incorrect predictive models can lead to erroneous movements.
“To successfully guide limb movements, the brain takes in sensory information about the limb, internally tracks the state of the limb, and produces appropriate motor commands,” the study says. “It is widely believed that this process uses an internal model, which describes our prior beliefs about how the limb responds to motor commands.”
Understanding how mistakes occur required development of a new mathematical program known as an algorithm that could extract the predictive model from the brain’s activity.
“I think this is revolutionary, exciting and creative work and a novel idea that shows practical benefits,” said Jonathan Pillow, an assistant professor in neuroscience and psychology at Princeton University. ”What it tells us is, if you take a basketball shot and it doesn’t hit, those errors involve the brain making a wrong prediction about how the arm should respond. It is trying to do it but it is using the wrong model. The brain calibrates its own model of how the muscles work.
“This group has done cutting-edge work on how the brain works, and they are using these insights on engineering applications,” said Mr. Pillow, who holds a Ph.D.
The CMU study involved two monkeys using a “brain-machine interface” or BMI to direct a cursor on a computer screen to hit a target, all while the activity of multiple neurons was measured. Analysis of results identified what predictive model the monkeys were using, thereby explaining 65 percent of “the behavioral errors.”
But it also showed how learning occurs, with repetitive tasks leading to better predictive models and results. “We’re updating internal models every day,” said Matthew Golub, primary author on the study and a postdoctoral fellow in CMU’s department of electrical and computer engineering.
The team studied brain signals and worked backward to characterize the predictive model. The approach shows how the brain sets up a model to allow the limbs to complete a task successfully, and how improvements occur through learning.
“The brain appears to build a predictive model of the world,” said Mr. Yu, who also has a doctoral degree. “Exactly how and where in the brain this predictive model is formed is an active part of our research.”
David Templeton: dtempleton@post-gazette.com or 412-263-1578.
First Published: February 23, 2016, 5:00 a.m.
