Reading the mind's eye: decoding category information during mental imagery.

Area of Science:

• Cognitive Neuroscience
• Neuroimaging

Background:

• Object category information is decodable from fMRI patterns in the ventral-temporal cortex during visual perception.
• The reliability and nature of these neural patterns during mental imagery, in the absence of visual input, remain less understood.

Purpose of the Study:

• To investigate the decodability of object category information from brain activity during visual imagery.
• To compare neural representations evoked by imagined objects versus visually presented objects.
• To determine if category information can be decoded from ventral-temporal cortex during mental imagery.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was used to record brain activity.
• Subjects either viewed or imagined four object categories: food, tools, faces, and buildings.
• Pattern classification techniques were applied to fMRI data to decode object category information.

Main Results:

• Category information was reliably decoded from the ventral-temporal cortex for both viewed and imagined objects.
• Decoding performance was comparable between viewed and imagined conditions, even when classifiers were trained on one and tested on the other.
• Similar decoding was achieved even when excluding information from category-selective brain areas.
• Neural representations during imagery and perception showed significant similarity in ventral-temporal cortex.

Conclusions:

• Neural patterns of representation for object categories are surprisingly similar during visual imagery and actual viewing.
• Cortical back projections can selectively re-activate specific neural activity patterns in the absence of bottom-up visual input.
• Findings suggest that mental imagery relies on similar neural mechanisms as visual perception.