Flexible decision-making is related to strategy learning, vicarious trial and error, and medial prefrontal rhythms during spatial set-shifting

Abstract

A hallmark of behavioral flexibility is the ability to update behavior in response to changes in context. Most studies tend to rely on error counting around reward contingency or rule switches to measure flexibility, but these measures are difficult to adapt in a way that allows shorter timescale flexibility estimates. Further, choice accuracy does not account for other markers of flexibility, such as the hesitations and decision reversals humans and other animals often exhibit as decisions unfold, a behavior often called vicarious trial and error (VTE). To relate observable information about decision-making to latent aspects like learning and behavioral flexibility, we quantified changes in decision-making strategy using a previously developed, recency-weighted Bayesian inference algorithm. By comparing models of strategy use with decision history to generate strategy likelihood estimates on a trial-by-trial basis, the algorithm enabled us to identify learning points, and served as the basis for the development of a behavioral flexibility score. Aligning flexibility scores to learning points showed that flexibility peaked around estimated learning points and near peaks in VTE rate. However, we occasionally observed VTE during periods of low flexibility, where it often led to incorrect choices, suggesting the likely existence of multiple VTE-types. Additionally, we built on the decades of research suggesting a prominent role for the medial prefrontal cortex in enabling behavioral flexibility by recording field potentials from the medial prefrontal cortex during task performance. We observed changes in different field potential frequency bands that varied with respect to the different behavioral measures we used to characterize learning and decision-making. Overall, we demonstrate the use of multiple measures that jointly assess relationships between learning, behavioral flexibility, and decision-making behaviors. Further, we used these complementary measures to demonstrate that a particular decision-making behavior, VTE, was likely to be a marker of deliberation at some times, and uncertainty at others. Finally, we validate these measures by showing that theta, beta, and gamma rhythms in the medial prefrontal cortex vary with respect to both observable and latent aspects of behavior.