Rapid Eye Movement Sleep, Punishment Structure and Decision-Making of the Iowa Gambling Task

Abstract

The Iowa Gambling Task (IGT) is a widely used decision-making task often used to assess decision-making impairment in clinical populations. During learning of the task, individuals receive consistent monetary rewards and inconsistent monetary punishments of different frequencies and magnitudes. Learning relies on in part the ventromedial prefrontal cortex (vmPFC) which integrates reward and loss information to identify the decks with the highest expected value (EV). Currently, little is known about factors that might influence this unique process. This thesis uses experimental techniques to investigate how sleep and punishment structures can influence learning of the IGT. In Chapter 2 we show evidence that a period of post-learning sleep enhances IGT learning, specifically via reduced choice from deck B. Deck B is an initially preferred “bad” deck that requires integrating a large infrequent $1250 punishment to discover its negative EV. In Chapter 3 we link this improvement to increased theta (4-8 Hz) activity in the vmPFC and left prefrontal cortex. We also show evidence that pre-sleep anticipatory emotions correlate to heightened theta activity in the right vmPFC during post acquisition REM. Finally, chapter 4 provides the first evidence that higher-order punishment structures can influence learning of the task. We provide the first evidence that punishment structures with high variance/skew may cause risk-aversion and that IGT decision-making may not be driven by EV alone. Collectively, the work presented provides much needed knowledge on factors that can influence IGT learning. We are the first to show evidence that IGT learning may be enhanced with post learning REM sleep and suggest that similar areas that are involved in online learning (i.e., vmPFC) may be involved. We are also the first to show that performance can be influenced by higher-order punishment structures. Taken together, this body of work reveals that integrating punishments into value estimates may be enhanced by post learning REM sleep and influenced by higher-order punishment structures. These results have widespread implications for how punishment structures in our day-to-day life may influence online and offline decision processes and provides a strong foundation for continued work to understand decision-making in healthy and clinical groups.