How language modulates color perception in a brain-constrained deep neural network

Area of Science:

• Cognitive Neuroscience
• Linguistics
• Computational Neuroscience

Background:

• The linguistic relativity hypothesis posits that language influences perception.
• Color perception studies show differences across languages, e.g., Russian speakers' enhanced discrimination of blue shades ('goluboj'/'sinij') vs. English speakers ('blue').

Purpose of the Study:

• To neuromechanistically explain how language modulates color perception using a brain-constrained neural network.
• To investigate the neural basis of enhanced color discrimination in Russian speakers compared to English speakers.

Main Methods:

• Developed a brain-constrained neural network simulating frontotemporal-occipital cortices.
• Utilized representational similarity analysis to compare neural activity patterns.
• Modeled virtual English and Russian speakers to simulate language-specific effects on color representation.

Main Results:

• The model showed similar neural activity for different blue shades labeled identically in English speakers.
• Distinct neural activations were observed for differently labeled blue shades in virtual Russian speakers.
• Microstructural neural changes, including shifts in neuron encoding, underpinned these language-dependent representational differences.

Conclusions:

• Language learning modulates functionally distinct color representations, impacting perceptual discrimination.
• The findings support the linguistic relativity hypothesis by providing a neurocomputational mechanism for language's influence on perception.
• The model successfully reproduced neurophysiological evidence, validating its approach to bridging linguistic, cognitive, and neuroscientific perspectives.