Attention modulates the geometry of auditory representations
Pedersini, C. A.; Muukkonen, I.; Wikman, P.
Show abstract
Selective attention enables the brain to prioritize behaviourally relevant information in complex environments, yet its neural mechanisms remain unclear. Here, we combined fMRI with representational geometry analyses to examine how attention reshapes neural population codes during naturalistic sound scenes. Participants attended to individual sound objects presented either with distractors from different categories (three objects across, 3OA) or the same category (three objects within, 3OW). Attention attracted neural population responses in low-dimensional manifolds toward the representation of each object presented in isolation, consistent with a reorganization of high-dimensional representational geometry. This effect depended on acoustic context: in 3OA, attended representations were significantly closer to object-alone representations than to distractors, whereas in 3OW attentional modulation was minimal. Category-specific effects were observed, with speech engaging left-lateralized language regions beyond the auditory cortex. Finally, stronger behavioural performance predicted greater representational attraction. These findings show that selective attention reshapes neural geometry in a context- and category-dependent manner closely linked to behaviour.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
Similar papers in this journal
- Spatial processing of limbs reveals the center-periphery bias in high level visual cortex follows a nonlinear topography 97%
- Joint population coding and temporal coherence link an attended talker's voice and location features in naturalistic multi-talker scenes. 97%
- Disentangling Semantic Composition and Semantic Association in the Left Temporal Lobe 96%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.