From Images and Words to Modality-Independent Semantic Representations: When and Where They Emerge

Modality-independent semantic category representations for six categories emerge in MEG decoding from ~70 ms, with cross-modal generalization peaking at ~520–530 ms in anterior temporal and orbitofrontal regions.

• Why it matters: It remains unclear when image- and word-evoked representations converge onto a shared, modality-independent code, and whether such category information appears even when semantics and modality are task-irrelevant.
• What they did: In MEG with 32 participants, classifiers decoded category from six-item sequences of images or written words, then tested cross-modal generalization and localized sources around the generalization peak.
• The result: Generalization peaked at similar latencies (~520–530 ms) with most similar classifier coefficients, and source-space decoding showed shared regions across both transfer directions, including left/right anterior temporal lobe, perisylvian cortex, right visual cortex, and bilateral orbitofrontal cortex.

AI summary of the abstract · may contain errors · not medical advice · read the paper.