COMPUTATIONAL MODELING OF PREFRONTAL-HIPPOCAMPAL INTERACTIONS IN VERBAL MNEMONIC DISCRIMINATION

Abstract

Background: Mnemonic Discrimination, the ability to distinguish between highly similar memories is impaired in people with various neuropsychiatric disorders. While hippocampus and prefrontal cortex are attributed to perform this strategic process, there has been no study to understand the computational processes of executive control systems trained with reinforcement learning such as prefrontal cortex in this paradigm. In our study, we present a novel framework to study how prefrontal cortex lesions affect verbal mnemonic discrimination. Methods: We employ a computational model to simulate a yes/no recognition task which is built upon Becker and Lim's work on the verbal free recall model. We extend the model by representing words as continuous word embeddings to facilitate the calculation of mnemonic discrimination performance in verbal learning paradigms. We model hippocampus as a continuous Hopfield network and prefrontal cortex as a reinforcement learning agent for memory retrieval based on task requirements. We first validate our model by ensuring the free recall results are consistent with prior results. Using our novel implementation of a verbal mnemonic discrimination task, we then examine the role of prefrontal cortex lesions on mnemonic discrimination performance. Results: Our results indicate that the lesions in prefrontal cortex have a significant negative impact on the mnemonic discrimination performance (ß=-0.07, p=0.004) and overall recognition performance (ß=-0.02, p=0.014). Conclusion: Our study is the first one to build a computational model of interactions between the prefrontal cortex and hippocampus to study verbal mnemonic discrimination. Our study highlights the role of intact reinforcement based executive control system for mnemonic discrimination and recognition performance.