Proceedings of the Annual Meeting of the Cognitive Science Society

Debates about the neural organization of semantic knowledge have long contrasted modular accounts, which posit anatomically specialized regions, with distributed accounts, which emphasize patterns of activity spanning many neural populations. Although multivariate neuroimaging has provided strong evidence for distributed representations, both perspectives often presume that the representational solutions supporting cognition are largely shared across individuals. Here, we challenge that premise by asking whether semantic representations are not only distributed, but also stable within individuals and meaningfully idiosyncratic across people. Using individualized multivariate fMRI decoding across two scanning sessions, we show that representations distinguishing faces, places, and objects are widely distributed across cortex, stable within individuals over time, and heterogeneous across participants. We then use these individualized representational profiles to guide transcranial magnetic stimulation (TMS), showing that stimulating non-canonical, decoding-defined sites selectively disrupts face similarity judgments in a manner comparable to stimulation of a canonical face-selective region. Together, these findings provide causal evidence that semantic representations arise from distributed neural systems whose organization is both stable and person-specific, calling for theories of cognition that make room not only for shared structure, but for the idiosyncratic ways meaning is represented across brains.