Rarely categorical, highly separable representations along the cortical hierarchy
9 affiliations
- 1ICM Paris Brain Institute, Sorbonne Université, CNRS, INSERM, Paris, France. [email protected]
- 2Zuckerman Institute, Columbia University, New York, NY, USA. [email protected]
- 3School of Computer and Communication Sciences, EPFL, Lausanne, Switzerland
- 4School of Life Sciences, EPFL, Lausanne, Switzerland
- 5Zuckerman Institute, Columbia University, New York, NY, USA
- 6Department of Statistics, Columbia University, New York, NY, USA
- 7Kavli Institute for Brain Science, Columbia University, New York, NY, USA
- 8Kavli Institute for Brain Science, Columbia University, New York, NY, USA. [email protected]
- 9Zuckerman Institute, Columbia University, New York, NY, USA. [email protected]
Abstract
A long-standing debate in neuroscience concerns whether individual neurons are organized into functionally distinct populations that encode information differently (categorical representations1–3) and the implications for neural computation. Here we systematically analysed how cortical neurons encode cognitive, sensory and movement variables across 43 cortical regions during a complex task (14,000+ units from the International Brain Laboratory public Brainwide Map dataset4) and studied how these properties change across the sensory–cognitive cortical hierarchy5. We found that the structure of the neural code was scale dependent. At the whole-cortex scale, neural selectivity was categorical and organized across regions in a way that reflected their anatomical connectivity. However, within individual regions, categorical representations were rare and limited to primary sensory areas, and neuronal responses were instead very diverse. With theoretical arguments and empirical evidence, we demonstrate that the diversity of neural responses enables high-dimensional representations and therefore high separability, allowing linear readouts to separate experimental conditions in many arbitrary ways. Indeed, when accounting for information that is actually encoded in each area, all cortical regions exhibit maximal separability. Our results indicate that cortical circuits prioritize diversity over categorical structure, supporting a computational regime geared towards high-dimensional, highly separable neural representations. Cortical circuits prioritize diversity over categorical structure, supporting a computational regime geared towards high-dimensional, highly separable neural representations.
Related
- Individualized morphometric-similarity deviations in autism linked to cortical hierarchy and microarchitecture
- Cortical Hierarchy Dynamically Organizes Large-Scale Neural Propagation
- Structure-function tethering of human locus coeruleus connectivity reveals a noradrenergic cortical hierarchy
- An adaptive fractional state links circuit mechanisms to cortical dynamics across the visual hierarchy
- Generalizable sex differences in functional brain organization follow the cortical hierarchy in youth
BibTeX
@article{posani2026rarely,
title = {{Rarely categorical, highly separable representations along the cortical hierarchy}},
author = {Lorenzo Posani and Shuqi Wang and Samuel P. Muscinelli and Liam Paninski and Stefano Fusi},
journal = {Nature},
year = {2026},
month = jul,
doi = {10.1038/s41586-026-10668-4},
url = {https://doi.org/10.1038/s41586-026-10668-4},
note = {CC BY},
}
RIS
TY - JOUR
TI - Rarely categorical, highly separable representations along the cortical hierarchy
AU - Lorenzo Posani
AU - Shuqi Wang
AU - Samuel P. Muscinelli
AU - Liam Paninski
AU - Stefano Fusi
PY - 2026
DA - 2026/07/15
JO - Nature
DO - 10.1038/s41586-026-10668-4
UR - https://doi.org/10.1038/s41586-026-10668-4
AB - A long-standing debate in neuroscience concerns whether individual neurons are organized into functionally distinct populations that encode information differently (categorical representations1–3) and the implications for neural computation. Here we systematically analysed how cortical neurons encode cognitive, sensory and movement variables across 43 cortical regions during a complex task (14,000+ units from the International Brain Laboratory public Brainwide Map dataset4) and studied how these properties change across the sensory–cognitive cortical hierarchy5. We found that the structure of the neural code was scale dependent. At the whole-cortex scale, neural selectivity was categorical and organized across regions in a way that reflected their anatomical connectivity. However, within individual regions, categorical representations were rare and limited to primary sensory areas, and neuronal responses were instead very diverse. With theoretical arguments and empirical evidence, we demonstrate that the diversity of neural responses enables high-dimensional representations and therefore high separability, allowing linear readouts to separate experimental conditions in many arbitrary ways. Indeed, when accounting for information that is actually encoded in each area, all cortical regions exhibit maximal separability. Our results indicate that cortical circuits prioritize diversity over categorical structure, supporting a computational regime geared towards high-dimensional, highly separable neural representations. Cortical circuits prioritize diversity over categorical structure, supporting a computational regime geared towards high-dimensional, highly separable neural representations.
SN - 0028-0836
SN - 1476-4687
ER -