Imagine playing a game of catch. To catch the ball, you must make an estimate of its trajectory. Only then you can decide how to move your body in order to catch it. How do we estimate the ball’s trajectory? One idea is that our brain constructs what’s termed an “internal model” of the ball and the environment. That is, even though we aren’t consciously aware of it, our brains know from past experience the physics of how a ball moves through space. Taking advantage of the fact that the laws of physics remain unchanged, we use this internal model so that we don’t have to re-learn each time how to catch a ball. Recent work has suggested that an area of the brain termed the anterior cingulate cortex is a likely region for representing such models of the environment. However, it has been extremely difficult to relate patterns of neural activity to distinct features of the model, such as how we estimate the ball’s position or velocity. Working in rats, we plan to use behavioral experiments, recordings of the activity of many neurons at once, sophisticated genetic tools, and cutting-edge mathematical analysis techniques to understand how the anterior cingulate cortex constructs internal models. We will investigate how rats learn to make inferences about their environment by presenting the rat with choices, but only particular patterns of choices will lead to a reward. The rat will have to learn which pattern leads to the reward—i.e., the rat will have to form hypotheses about the rules of the task, and test those hypotheses to construct an internal model of the task. Using advanced recording technology, we can then ask which patterns of neural activity in the anterior cingulate cortex relate to the discovery of rules about the environment, paving the way to understand how internal models are constructed more generally.
Alla Karpova, Howard Hughes Medical Institute, Janelia Farm Research Campus
Shaul Druckmann, Howard Hughes Medical Institute, Janelia Farm Research Campus
Joshua Tenenbaum, Massachusetts Institute of Technology