Biased competition in semantic representation during natural visual search

Area of Science:

• Neuroscience
• Cognitive Science
• Visual Perception

Background:

• Humans frequently divide attention among multiple visual targets, increasing search difficulty.
• The biased competition (BC) hypothesis proposes attentional modulation is a weighted linear combination of individual target responses, biased by cortical region selectivity.

Purpose of the Study:

• To investigate if the BC hypothesis explains semantic representation during naturalistic, category-based visual search in dynamic scenes.
• To determine if brain responses during divided attention are a linear combination of single-target attention responses.

Main Methods:

• Participants viewed natural movies and performed single-target (humans or vehicles) or divided-attention (both) tasks.
• Whole-brain BOLD responses were recorded using fMRI.
• A voxelwise linearity index assessed semantic representation linearity during divided attention.

Main Results:

• Semantic representations during divided attention were substantially linear.
• These representations showed bias towards the preferred target in category-selective ventral temporal cortex regions.
• Both target and non-target category representations were analyzed.

Conclusions:

• The biased competition hypothesis provides a compelling framework for understanding attentional modulation of semantic representations.
• Findings support the BC hypothesis in complex, naturalistic visual environments.
• Attentional modulation is linearly combined and biased by intrinsic neural selectivity.