Self-supervised learning yields representational signatures of category-selective cortex.

Overview

1. Neural Dynamics of Visual Cognition Group, Department of Education and Psychology, Freie Universität Berlin, Berlin, Germany

Abstract

The ventral visual stream contains category-selective regions with distinct feature tuning, most prominently the fusiform face area (FFA) and parahippocampal place area (PPA). Why do these brain regions exhibit distinct tuning properties? Recent work suggests that brain-like category-selective features emerge from a general visual learning mechanism without domain-specific biases. Here, we test this proposal by applying a functional localizer approach to both humans and self-supervised neural networks, identifying face- and scene-selective units in the brain and in models. We then compared fMRI and model responses across a broad stimulus set probing classic representational signatures of the FFA and PPA, including preferences relating to curvature, animacy, real-world size, mid-level features, face shapes, and spatial layout information. Category-selective model units largely recapitulate the distinct representational signatures of category-selective brain regions, capturing most of the effects in our test battery. Our findings demonstrate that domain-general learning objectives are sufficient to create humanlike category-selectivity, suggesting that the distinct representational signatures of category-selective cortex may emerge from a unified computational goal akin to self-supervised learning.

Reproduced under the paper's license (CC BY), from the paper cited above.

Code

No file of the authors' code could be read here: it is described below, and read at its source.

Code availability

The stimulus sets and code for all analyses are available at https://osf.io/d9f5e/overview (doi: 10.18112/openneuro.ds007368.v1.0.1)140.

Reproduced under the paper's license (CC BY), from the paper cited above.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

• 2 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
• 0 scripts, each with its path and the digest of its content;
• no match between paragraphs and code yet;
• neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

Datasets cited

• openneuro:ds007368, at OpenNeuro; found in “Data availability”

Data availability

The fMRI dataset collected for this paper is available at https://openneuro.org/datasets/ds007368.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 2 authors, 2 keywords, 11 MeSH terms, 1 funder, 129 references.

Cite

This paper

Janini, D., & Cichy, R. (2026). Self-supervised learning yields representational signatures of category-selective cortex. Communications biology, 9(1), 1153. https://doi.org/10.1038/s42003-026-10843-3

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1523/jneurosci.0038-26.2026 [code]

Multidimensional Feature Tuning in Category Selective Areas of Human Visual Cortex.

Journal: The Journal of neuroscience : the official journal of the Society for Neuroscience

In common: 25 references

[2] doi:10.1073/pnas.2513126123

The cortical scene processing network emerges in infancy, prior to independent navigation experience.

Journal: Proceedings of the National Academy of Sciences of the United States of America

In common: 19 references

[3] doi:10.1016/j.dcn.2026.101765 [code]

Fusiform face area development correlates with development in higher-order social brain regions.

Journal: Developmental cognitive neuroscience

In common: 13 references

[4] doi:10.1162/imag.a.1207 [code]

Investigating the temporal dynamics and modeling of mid-level feature representations in humans.

Journal: Imaging neuroscience (Cambridge, Mass.)

In common: 9 references, author Radoslaw Cichy

[5] doi:10.1038/s41467-026-73938-9 [code]

Higher visual areas act like domain-general filters with strong selectivity and functional specialization.

Journal: Nature communications

In common: 13 references

[6] doi:10.1038/s41467-026-76098-y [code]

A single computational objective can produce specialization of streams in visual cortex.

Journal: Nature communications

In common: 13 references

[7] doi:10.1007/s00429-026-03152-2 [code]

Limb-selective regions in the lateral temporal lobe shrink from childhood to adulthood.

Journal: Brain structure & function

In common: 10 references

[8] doi:10.1038/s42003-026-10169-0 [code]

Shared representations in brains and models reveal a two-route cortical organization during scene perception.

Journal: Communications biology

In common: 10 references

[9] doi:10.1162/imag.a.1246 [code]

A 3.5-minute-long reading-based fMRI localizer for the language network.

Journal: Imaging neuroscience (Cambridge, Mass.)

In common: 8 references

[10] doi:10.1038/s41593-026-02207-1 [code]

Neuronal tuning aligns dynamically with object and texture manifolds across the visual hierarchy.

Journal: Nature neuroscience

In common: 7 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.