Rumination Derails Learning about Potentially Reinforcing Cues

People who brood about their feelings and about what is going wrong in their lives--i.e., ruminate--tend to engage in ineffective behaviors. Ineffective behaviors in turn generate stress and, in the long term, poor mental health. Yet, precisely how rumination disrupts effective behavior is mysterious. Here, we investigate a potential mechanism: impaired reinforcement learning (RL). RL refers to the suite of mechanisms that allow us to make adjustments when something goes different than we expected. RL enables adapting to our environments and altering our actions so that we can act effectively in the future. Thus, if rumination impairs RL, this could explain why it leads to diverse ineffective behaviors. We investigated if rumination interferes with RL in a task with high attentional demands. This is of interest because RL typically depends on attention, as cues in the real world are often embedded in environments that are cluttered, complex, and capacious. Rumination appears to hijack attentional networks, and thus may interfere the ability to home in on potentially reinforcing cues in such environments. We recruited participants (N = 56) with elevated depression symptoms, and had them perform the RL task in the context of experimental rumination and distraction inductions, respectively designed to induce rumination or a neutral state. Manipulation checks confirmed that only the rumination induction elevated state rumination. We compared, within-subject, how an increase in state rumination affected RL. Rumination impaired performance on the task, suggesting that it disrupts RL. This may explain why rumination is associated with diverse ineffective behaviors. When ruminating, participants also responded more quickly in the task. This is noteworthy, as speeded responding is typically associated with learning a RL task--yet, ruminating participants responded faster even as they learned less. We next used computational modeling to investigate if rumination interfered with performance through impairing specific learning, choice, or attention mechanisms. Model parameters representing these mechanisms did not have different values in the two inductions, and simulated data using these parameters could not replicate the impaired learning in rumination. This suggests that rumination did not alter performance by interfering with learning rate, choice stochasticity, or attentional breadth--the mechanisms represented by the model parameters. Next, we applied a model-derived trial-wise marker of choice difficulty that we have established in prior work. We found that, when ruminating, participants did not slow as much on difficult trials. This suggests rumination impairs learning in part by leading to more automatic responding when controlled responding is needed. We discuss the significance of our behavioral findings for theories of how rumination impedes effective behavior, future directions for experimentation and modeling to pinpoint the mechanisms underlying the behavioral and control allocation differences, and the clinical implications of our findings.Ph.D., Psychology -- Drexel University, 201