Category representations in the brain are both discretely localized and widely distributed

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Neuroimaging

Background:

• The brain's representation of visual categories is debated: discrete localization vs. widespread distribution.
• Early functional MRI studies suggested localized representations.
• Recent machine learning fMRI studies indicate distributed representations, but often overlook shared information.

Purpose of the Study:

• To investigate whether visual category information is localized or distributed in the brain.
• To differentiate between unique and shared information across brain regions.
• To reconcile conflicting findings on category representation in the brain.

Main Methods:

• Utilized functional MRI (fMRI) data.
• Applied multivariate analyses to parse fMRI signals into unique and shared variance.
• Examined information representation across different brain regions.

Main Results:

• Distributed representations are evident when brain regions are analyzed independently.
• Localized representations emerge when unique and shared information are separated.
• Shared information exhibits a distributed pattern, whereas unique information is localized.
• Brain connectivity likely facilitates information spreading, while distinct regions handle category-specific computations.

Conclusions:

• Visual category information is represented through both localized and distributed mechanisms.
• Information is initially localized and subsequently shared across regions, leading to observed distributed patterns.
• Findings support a hybrid model reconciling localizationist and distributed theories of brain function.