Grid representations for efficient generalization
Linda Yu, Matthew Nassar, Brown University, United States
Posters 2 Poster
Pacific Ballroom H-O
Fri, 26 Aug, 19:30 - 21:30 Pacific Time (UTC -8)
Generalization of information from previous contexts to new ones is an important aspect of efficient learning and behavioral flexibility. We propose that grid representations, identified originally in rodent entorhinal cortex, but later across multiple cortical regions in humans, could serve as the substrate for this type of fast structure transfer. We created a predictive inference task that required participants to learn color-location associations, which alternated between two 90- degree rotations. Generalization was tested during a follow-up session in which participants performed the same task with novel rotations. Analysis of fMRI data identified two types of grid representation that are orthogonalized by our task design: a purely spatial one in the precuneus that is consistent across rotations and an abstract cognitive one in the medial prefrontal cortex, which shifted 90 degrees along with the color-location relationships. Behaviorally, participants were able to generalize learning successfully to novel rotations in the transfer task. Our findings provide evidence of a grid system can orient to abstract task space, where its link to behavioral generalization will be further explored.