Inquiry of the orbitofrontal cortex role in incentive learning: An artificial neural networks simulation

Abstract

This study evaluates, through artificial neural network (NN) simulations, the roles of amygdala (AMG) and orbitofrontal cortex (OFC) in reversal learning as responsible for assigning and updating incentive values, respectively, by taking into account some of its physiological characteristics and its connections with mesolimbic system structures. An architecture was built which contained modules of nodes representing OFC, amygdala (AMG), ventral tegmental area (VTA) and nucleus accumbens (NACC). Simulations were based on experimental responses of rats during a discrimination and reversal learning task in a T-maze. During simulations, intact NN responded similar to control rats. OFC showed retarded encoding in afferent connections as comparing with AMG nodes and an increase in the recruitment of neural nodes during reversal tasks. Activity of neural nodes plausibly emulates reinforcer preference during discriminations and reversals in AMG, COF and NACC. All these findings were disrupted by OFC’s inactivation. NN with OFC nodes inactivated (RN), just like rats with OFC inactivated, achieved discrimination learning although its reversal was impaired. As with rats, perseveration of incorrect responses was not found because subjects stop responding after a few incorrect trials. These findings are in accordance with the hypotheses of AMG as key in encoding novel predicting-outcome cues and OFC as critical for incentive value updating and the crucial role this prefrontal region plays for correct performance in discrimination tasks.