How experience shapes representations of overlapping visual events

University Of Texas At Austin, Austin TX

Investigators

Linked publications & trials

Abstract

PROJECT SUMMARY/ABSTRACT The human memory system is able to maintain both separated memories for overlapping events and combine related events into integrated memories. Neurocomputational models and empirical work suggest that hippocampal subfields dentate gyrus and CA3 (DG/CA3) and lateral prefrontal cortex (PFC) create distinct representations through pattern separation, whereas subfield CA1 and medial PFC integrate overlapping representations. Although the formation of pattern separated and integrated neural codes are both crucial for successful and efficient learning, it is unclear how these two types of representations come to coexist within hippocampal and prefrontal networks and what conditions selectively promote each type of neural code. The overarching goals of the proposed research are: (1) to test the predictions set forth by computational models of the hippocampus by linking separated and integrated neural codes to DG/CA3 and CA1, respectively, and (2) to investigate the role of PFC in memory formation within hippocampus. Here, we will combine high-resolution functional magnetic resonance imaging (hr-fMRI), multivariate decoding, representational similarity analysis, and functional connectivity analysis to test what conditions influence coding biases in hippocampal and PFC subregions. Participants will study initial pairs of images (AB) followed by overlapping pairs (BC). Aim 1 will test the hypothesis that visual similarity of overlapping pairs will affect the degree of memory reactivation during learning and differentially impact engagement and connectivity within integration and separation networks. We predict that neural coding in CA1 will be biased towards integration of overlapping events and have a linear relationship with memory reactivation; DG/CA3 will be biased towards separation of overlapping events and have a non-monotonic relationship with reactivation. Aim 2 will link changes in memory reactivation to representational changes. Across both Aims, we predict that memory representations in hippocampus will be influenced by functional connectivity with PFC. Behaviorally, we expect that memory for directly learned pairs (AB, BC) will be related to separation in DG/CA3. In contrast, the ability to generalize and infer the indirect relationship across overlapping episodes (AC) will be predicted by integration in CA1. Together, this research will provide evidence of coding biases in hippocampal subfields and identify what learning conditions shift coding strategies in these regions during learning. By understanding what conditions bias the human memory system towards separated or integrated coding strategies during learning, the findings from the proposed research can be utilized in interventions for patients with neurological disorders marked by memory disturbances. Specifically, this research will be relevant for patients with Alzheimer's disease and epilepsy, who experience memory interference between overlapping memories during retrieval.