Neural noise helps you learn 30.03.2015

Twenty years ago, Stanford neuroscientists discovered strange noise activity in some brain neurons: they responded to stimuli that seemed to have nothing to do with them. And such activity arose precisely when the brain made a decision. The experiment itself was as follows: the experimental animals had to determine how dots on the screen move, from right to left or from left to right; in case of a correct answer, a reward was given. With the help of such a model, it is possible to study what processes in the brain accompany the formation of categories. The categorization of objects and phenomena is one of the most general features of the psyche that underlies learning, and it would really be interesting to know what is happening in the brain at this moment. In this case, as is easy to understand, it was necessary to distinguish two classes of objects: those that move in one direction, and those that move in the opposite direction.

As a result, it was possible to find a group of neurons that respond to movement, and among them were those that became especially active at the very moment of making a decision. However, their activity looked as if some cells shouted “from right to left!” in response to the point, while others shouted “from left to right!” in response to the same point, regardless of where the point actually moves. The noise level was reduced with the help of a reward for the correct answer - it tuned the neurons, making them more picky and less noisy, so that they began to respond mostly only to points of one, "their" category. And what was especially strange was that the neural noise did not occur at all in those areas of the cortex that are usually associated with decision making.

Why neurons are noisy in the "non-core" part of the brain, we managed to partially find out only now, with the help of a computer model developed by Tatiana Engel and her colleagues; the results of their work are published in Nature Communications. The model imitated the work of neural circuits connecting sensory areas of the brain with categorizing ones. Virtual neurons "observed" points that moved in different directions and which had to be divided into the same two classes, "right" and "left" - as in the original experiment with animals.

A simulated neural circuit, unlike a real one, can be deprived of the ability to make noise, which the researchers did. But it turned out that without the neural noise that accompanies the choice, the formation of categories is impossible. In other words, in order for a class of points moving from right to left to form in the mind, the brain must make a choice in "noisy" conditions, when some of the neurons will simultaneously "agitate" for the wrong answer. If we ignore the points and pick up a more realistic example, then imagine that every morning you choose between a cup of coffee and a cup of tea. You make a choice every day for a week, two weeks, a month, six months, and in the end you come to the conclusion that a morning cup of coffee is exactly what you need. But if it suddenly happens that your brain makes a choice without any noise, then you simply will not form a connection between the morning hours and coffee, the very concept of morning coffee will be absent.

Of course, there is a great temptation here to interpret neural noise as "doubt", or as "the need to consider all possible solutions." However, such formulations belong rather to the field of philosophy, which we can hardly correlate with specific neurophysiological phenomena so far. However, it may well be that new data in the future will allow the creation of some hardware methods that improve cognitive abilities - through the management of neural noise. But for now, it remains to be seen where it actually comes from: whether it is the sensory departments that generate it, or whether it is produced by other areas of the brain that are directly related to decision making, or whether both sensory and cognitive departments are involved here.