Discrete Subfields and Continuous Gradients Coexist: A Multi-Scale View of Hippocampal Organization

Abstract

The human hippocampus is studied via two competing frameworks: one dividing it into discrete anatomical subfields with distinct computational processes, and another describing it as a continuous, functional gradients along the anterior-posterior and medial-lateral axes. How these organizational principles relate to one another, particularly regarding intrinsic neural timescales of the hippocampus, remains unknown. Here, we used high-resolution resting-state fMRI to investigate how single-voxel autocorrelation, a measure of intrinsic neural timescale, maps onto hippocampal subfields. We found evidence for a hybrid organization. First, consistent with our predictions, we observed significantly higher autocorrelation (longer timescales) in the subiculum compared to the other subfields. Contrary to our hypotheses, we found that CA1, which is implicated in integration, had low autocorrelation whereas CA2/3 and CA4DG, which are linked to pattern separation, had intermediate autocorrelation. Second, we discovered that the overarching anterior-posterior and medial-lateral gradients of autocorrelation were recapitulated within each individual subfield. Finally, data-driven clusters of autocorrelation values aligned more strongly with the continuous gradients than with the discrete anatomical boundaries, particularly in the right hemisphere. These results suggest that the discrete and continuous views of hippocampal organization are not mutually exclusive but coexist across different spatial scales. We therefore propose a new comprehensive model of hippocampal function that integrates both its modular, subfield-specific properties and its graded, large-scale organization.