Internal States Emerge Early During Learning of a Perceptual Decision-Making Task

Abstract

Recent work has shown that during perceptual decision-making tasks, animals frequently alternatebetween different internal states or strategies. However, the question of how or when these emergeduring learning remains an important open problem. Does an animal alternate between multiplestrategies from the very start of training, or only after extensive exposure to a task? Here we addressthis question by developing a dynamic latent state model, which we applied to training data frommice learning to perform a visual decision-making task. Remarkably, we found that mice exhibiteddistinct “engaged” and “biased” states even during early training, with multiple states apparent fromthe second training session onward. Moreover, our model revealed that the gradual improvement intask performance over the course of training arose from a combination of two factors: (1) increasedsensitivity to stimuli across all states; and (2) increased proportion of time spent in a higher-accuracy“engaged” state relative to biased or disengaged states. These findings highlight the power of ourapproach for characterizing the temporal evolution of multiple strategies across learning.

Keywords: Internal states, Learning, Perceptual Decision-Making, Behavioral Strategies, Hidden Markov Model