David Rahabi¹, Sam Hutchinson¹, Ammar Marvi², Matthew Glasser³, Nagehan Demirci³, Nancy Kanwisher¹; ¹MIT, ²Harvard, ³WashU

The last few decades of fMRI research have identified many cortical regions with distinctive and highly selective response profiles, including the fusiform face area (FFA), parahippocampal place area (PPA), extrastriate body area (EBA), and visual word form area (VWFA). Because the anatomical locations of these regions vary substantially across individuals, group-based analyses often fail to identify them precisely, motivating the widespread use of individual functional localizers. Although recent work has made these localizers more efficient (e.g., Marvi et al., 2025), they still require costly scan time. Meanwhile, multimodally-defined cortical areas based on the Human Connectome Project (HCP, Glasser et al., 2016) have been widely used to summarize multi-modal MRI data, but it is unknown the extent to which they are capable of providing fine-grained characterizations of local cortical functional selectivity. We therefore quantified the selectivity of the response of the FFA, PPA, EBA, and VWFA when these regions are defined with individual functional localizers versus when they are defined as the HCP cortical areas in the same region. Individual functional localization revealed significant and strong category selectivity in held-out data in all four regions, whereas category selectivity was much weaker (PPA and EBA) or even reversed (FFA and VWFA) in both a) the closest HCP-defined area/s (see also Molloy et al., 2024), and b) the top 10% of vertices within those closest HCP-defined areas with the highest tSNR in resting functional data. Thus, HCP-defined cortical areas do not reproduce the functional specificity obtained from regions defined with individual localizer scans.

Acknowledgements: NIH Project Number: UM1 MH130981-01, ”Functionally guided adult whole 459 brain cell atlas in human and NHP”, PI: Ed Lein