Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session C: Friday, August 15, 2:00 – 5:00 pm, de Brug & E‑Hall

Working Memory-Supported Reinforcement Learning Related to Mental Health Phenotypes in a Representative Sample

Laura Ana Bustamante¹, Krishn Bera², Guillaume J. Pagnier, Alexis Cruz, Alyssa Zimmerman, Claire Gillan, Deanna M. Barch¹, Megan Boudewyn, Cameron S. Carter, Molly Erickson, James Gold, Steven J. Luck, Daniel J. Ragland, Andrew P. Yonelinas, Angus MacDonald III, Michael Frank²; ¹Washington University, Saint Louis, ²Brown University

Presenter: Laura Ana Bustamante

Working memory-supported reinforcement learning (RL-WM) may differ between participants with and without schizophrenia and mood disorders. It remains unexplored whether such processes are related to distinct dimensional phenotypes. This study tests whether RL-WM parameters are associated with self-reported mental health phenotypes (anxiety, depression, drug use, impulsivity, mania, motivation/pleasure, schizotypy) in a US representative online sample (N=2,300 exploratory). Participants completed a stimulus-response learning task that manipulates WM demands (set size, delay) to disentangle WM and RL contributions to performance (N=1,665 post-exclusion, ages 18-65 years). The RL-WM model estimated participants' reliance on WM (v. RL) reliance, WM capacity, WM decay (interference from intervening trials), RL learning rate, perseveration (negative feedback neglect), and undirected noise (attention lapse) parameters. We tested associations between 8 RL-WM measures and survey items using multivariate sparse partial least squares regression (m-SPLS). m-SPLS Cross-validation identified 1 component with 31 survey items predicting 7 RL-WM measures. Items reflecting schizotypy, mania, and others, were related to reduced RL and WM performance (increased perseveration, WM decay, undirected noise, decreased WM reliance). Results support RL-WM utility for computationally mechanistic community mental health research. Replicability will be tested in an independent sample (N=6,725 confirmatory).

Topic Area: Reward, Value & Social Decision Making

Extended Abstract: Full Text PDF