Search Papers | Poster Sessions | All Posters
Poster B145 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
Latent state space of the nucleus accumbens BOLD activity indicates internal feedback mechanisms for learning temporal regularity
Wei Tang1 (), Pradyumna Lanka2, Zhenghan Qi2; 1Indiana University Bloomington, 2Northeastern University
Statistical learning (SL) is an unconscious cognitive process in which the brain extracts regularities from the environment through repeated exposure. Because of its implicit nature and potentially high individual variability, SL posts a major challenge for studying its neural mechanisms. In this work, we investigate the latent brain states that drive multivoxel patterns of functional neuroimaging data during the learning of temporal regularity embedded in sequential visual inputs. This approach allows the latent states to be individual-specific while preserving meaningful group-level consistency. We found that, consistently across individuals, a state in the nucleus accumbens was associated with the perceptual facilitation effect as the subjects were learning the temporal pattern implicitly. This state occurred more frequently during random sequences than structured sequences, suggesting a potential error-driven feedback signal for training the internal prediction. These findings open the door to further elucidating network dynamics using the found latent states as guidance.
Keywords: Statistical learning Striatum Hidden Markov model fMRI