Search Papers | Poster Sessions | All Posters
Poster C24 in Poster Session C - Friday, August 9, 2024, 11:15 am – 1:15 pm, Johnson Ice Rink
Representational geometry, not topography, best characterizes human neural activity in transmodal brain areas
Bogdan Petre1 (), Martin A Lindquist2, Tor D Wager1; 1Dartmouth College, 2Johns Hopkins University
Conserved topographies (e.g. retinotopy, tonotopy, somatotopy) and idiosyncratic population codes provide complementary accounts of neural representation at different scales. We use fMRI to examine which best explains human cortical function at mesoscale (mm to cm) where these accounts intersect and remain unsettled. We compare interindividual functional correspondence up to a low dimensional topography preserving warp (diffeomorphisms) or a high dimensional projection that preserves population code feature spaces (hyperalignment). Unlike previous studies our comparison is matched on training data and spans the full brain. When alignments computed from spontaneous activity were applied to a battery of task evoked responses, diffeomorphic warps predicted responses better in unimodal cortical areas while hyperalignment predicted responses better in transmodal areas. This is consistent with classic topographic representations of peripheral sensation and control, but demonstrates conserved feature spaces rather than topographies provide a better account of higher order computations in biological neural networks.
Keywords: cortical maps population codes representation interindividual alignment