Effects of Time Manipulation on Reinforcement Learning

Eslam Hassan

Advisor: James C Thompson, PhD, Department of Psychology

Committee Members: Allison Jack, Martin Wiender

Online Location, https://gmu.zoom.us/j/98452889797
August 13, 2026, 09:00 AM to 11:00 AM

Abstract:

This dissertation examines how time manipulation influences reinforcement learning (RL) and working-memory (WM). RL and WM are two cognitive processes that overlap in underlying mechanisms and support each other in learning and decision-making. Disruptions to RL and WM are a common trait of psychological disorders and previous research has shown that manipulating temporal factors using massed and spaced learning schedules can provide information on these discrepancies. To examine these effects, three experiments were conducted on dual-system frameworks central to our cognition: Experiment 1 assessed Pavlovian-instrumental transfer (PIT), Experiment 2 observed goal-directed and habit-based learning/devaluation, and Experiment 3 evaluated model-based and model-free RL. Methodological consistency was ensured using the same time manipulation and WM paradigms across all studies, including massed and spaced conditions, the Operation Span task (OSPAN), and immediate and delayed (24-hour) retention tests. Results revealed that massed and spaced learning did not influence PIT, goal-directed and habit-based learning/devaluation, or model-based and model-free RL. WM capacity also did not predict or relate to any RL system. However, outcome-specific and general PIT, goal-directed learning, and devaluation-sensitivity were observed in Experiments 1 and 2 by showing evident learning performance. These findings emphasize that RL is persistent and insensitive to time manipulation as spaced learning produced no measurable effects. Instead, RL presented to be affected by PIT effects and outcome value, highlighting the importance of associative and motivational processes in learning and decision-making. Lastly, learning mechanisms such as consolidation, reinforcement contingencies, and repetition appear as more suitable constructs to use in explaining differences in learning performance than individual differences in WM capacity.