Compressibility of high-dimensional conceptual information across the human cortex
Presented During:
Thursday, June 26, 2025: 11:30 AM - 12:45 PM
Brisbane Convention & Exhibition Centre
Room:
M4 (Mezzanine Level)
Poster No:
828
Submission Type:
Abstract Submission
Authors:
KAIXIANG ZHUANG1, Xinyu Liang1, Yun Wang1, Joern Alexander Quent1, Deniz Vatansever1
Institutions:
1Fudan University, Shanghai, Shanghai
First Author:
Co-Author(s):
Introduction:
Humans excel at organizing high-dimensional conceptual information into low-dimensional semantic spaces [1], facilitating complex mental operations such as analogical reasoning, inferences and generalization [2]. Despite extensive behavioral evidence for this compression process, its neural basis and cortical topography remain largely unexplored. Using ultra-high resolution 7T fMRI and a comprehensive set of 1,854 object concepts, here we investigated how cortical responses represent and compress semantic information into low-dimensional manifolds. Our results highlight the default mode network (DMN) as a critical hub for semantic compression, providing novel insights into how humans efficiently organize and utilize conceptual knowledge [3].
Methods:
We presented 22,256 high-quality object-centric images, representing 1,854 concepts from the THINGS database [4], to 20 participants (24.56 ± 2.42 years, 14F/6M) during 7T fMRI across five consecutive sessions (60 runs in total) (Fig. 1a). A vertex-by-concept embedding matrix was constructed by averaging GLMsingle-derived single-trial beta maps across images and participants (Fig. 1b). Dimensionality reduction using Principal Component Analysis (PCA) uncovered cortical manifolds of conceptual information, which we interpreted via 66 behaviorally-driven semantic dimensions [5] and 53 semantic categories [6]. Cortical semantic compressibility was assessed by mapping vertex-wise reconstruction errors, with lower errors indicating greater alignment to low-dimensional representations (Fig. 1c-d).
Results:
The posterior DMN, including the PCC, TPJ, and IPL, demonstrated maximal semantic compressibility, reflecting stronger alignment with low-dimensional cortical semantic manifolds (Fig. 1e). In contrast, regions such as the OFC, temporal area TF, and primary visual cortex showed minimal compressibility, indicating more specialized representations. At the macro-scale, transmodal networks associated with higher-order cognition (e.g., DMN, frontoparietal networks) exhibited greater compressibility than unimodal sensorimotor networks (e.g., auditory and language networks) (Fig. 1f). Compressed cortical semantic manifolds aligned closely with behaviorally-driven dimensions (Fig. 2a), enabling flexible delineation of semantic categories (Fig. 2b) and encoding multiple pathways of conceptual access in the posterior DMN (Fig. 2c).
Conclusions:
Our findings elucidate the neural mechanisms of semantic compression, identifying the posterior DMN as a central hub for transforming high-dimensional conceptual information into low-dimensional semantic manifolds. These manifolds not only align with human-interpretable dimensions but also enable flexible categorical access via diverse representational pathways, advancing our understanding of the cortical organization of conceptual knowledge for healthy and adaptive cognition.
Higher Cognitive Functions:
Higher Cognitive Functions Other 2
Language:
Language Comprehension and Semantics
Learning and Memory:
Long-Term Memory (Episodic and Semantic)
Learning and Memory Other 1
Modeling and Analysis Methods:
Multivariate Approaches
Keywords:
Cognition
Cortex
Data analysis
Design and Analysis
Experimental Design
FUNCTIONAL MRI
Memory
Multivariate
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Provide references using APA citation style.
[2] Hebart, M. N., Zheng, C. Y., Pereira, F., & Baker, C. I. (2020). Revealing the multidimensional mental representations of natural objects underlying human similarity judgements. Nature human behaviour, 4(11), 1173-1185.
[3] Bottini, R., & Doeller, C. F. (2020). Knowledge across reference frames: Cognitive maps and image spaces. Trends in Cognitive Sciences, 24(8), 606-619.
[4] Hebart, M. N., Dickter, A. H., Kidder, A., Kwok, W. Y., Corriveau, A., Van Wicklin, C., & Baker, C. I. (2019). THINGS: A database of 1,854 object concepts and more than 26,000 naturalistic object images. PloS one, 14(10), e0223792.
[5] Contier, O., Baker, C. I., & Hebart, M. N. (2024). Distributed representations of behaviour-derived object dimensions in the human visual system. Nature Human Behaviour, 8(11), 2179-2193.
[6] Stoinski, L. M., Perkuhn, J., & Hebart, M. N. (2024). THINGSplus: New norms and metadata for the THINGS database of 1854 object concepts and 26,107 natural object images. Behavior Research Methods, 56(3), 1583-1603.