PAGE 1 OF 1
A Noisy Brain Is a
by Lindsey Pujanauski
Background activity in the brain may be more prevalent—and more important—than previously thought.
Neural activity in the brain changes gradually, even when nothing new is being learned.
New research from H. Sebastian Seung's laboratory at the Massachusetts Institute of Technology (MIT) suggests that scientists still have quite a bit to learn about learning.
The researchers were following up on intriguing observations made by MIT collaborator Emilio Bizzi while studying behavioral tasks of macaques in 2003. Bizzi's group saw slow changes in neural activity even while the macaques were performing familiar tasks, during which no learning was going on. Previously, scientists had assumed that slow changes in neural firing corresponded to learning of motor activities, but these changes in neuron firing rates produced no corresponding changes in motor behavior.
It was Uri Rokni, a postdoc in Seung's lab, who realized that these slow changes in the macaque brain, dubbed background noise, were distributed randomly. The changes were seen during the learning tasks as well and could represent the existence of an unstable neural network in the motor cortex. For decades, scientists have thought that the process of learning can be detected as changes in a stable neural network: when there is no learning, there are no changes in firing rates. According to Seung, the background noise was surprising. "Then the question became how to interpret [it]," he says.
One idea is that there are two components in the brain—a teacher and a tinkerer. The tinkerer is constantly adjusting things, which produces the background noise; the teacher goes back and fixes or optimizes the changes. According to Seung, "if you get rid of the noise, which is made by the tinkerer, you get rid of any ability to learn."
Rokni developed a simple mathematical model to represent how changes in neural activity during the familiar tasks could be irrelevant to behavioral performance. His hypothesis is rooted in the notion that the motor cortex is a redundant network, meaning that it uses more neurons than it needs. As a result, changes that produce background noise can affect the wiring of the brain without affecting motor behavior.
Photo: Mehau Kulyk / Photo Researchers, Inc.