Poster Presentation

Poster C in Poster Session C: Wednesday, August 5, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Neural mechanisms underlying valence bias in reinforcement learning and episodic memory across development

Sahithyan Sivakumaran¹, nikita sossounov¹, Wangjing Yu¹, Catherine Hartley¹; ¹New York University

Presenter: Sahithyan Sivakumaran

Individuals learn differently from positive and negative outcomes. Such valence asymmetries shift across development and have been shown to yield corresponding biases in subsequent memory. The neural mechanisms underlying this link between biases in learning and memory, however, remain unclear. To investigate this, in the current fMRI study, we recruited participants aged 9-25 to complete a reinforcement learning (RL) task in which each choice outcome was associated with a trial-unique image. We then probed participants' memory for these images in a surprise recognition test. Using a risk-sensitive RL model, we computed each participant's asymmetry index, reflecting their valence bias in updating values during learning. Consistent with prior research, we showed that the information individuals retained in memory reflected their corresponding RL valence asymmetries. Moreover, this correspondence between the valence asymmetries in RL and episodic memory emerged with age and was stronger in older than younger participants. We focused on post-learning consolidation processes as potential mechanisms underpinning the developmental emergence of valence-asymmetric memory prioritization. We found that increased resting-state connectivity between hippocampus and ventral temporal cortex after learning predicted better subsequent memory for information congruent with individual learning biases in older, but not younger, participants. Together, our findings show that, over development, valence biases in RL increasingly influence what is maintained in memory and suggest offline consolidation as a potential mechanism underlying this developmental shift.

Topic Area: Memory, Learning & Knowledge Structures