An image-computable model of how endogenous and exogenous attention differentially alter visual perception

Area of Science:

• Cognitive Neuroscience
• Computational Vision

Background:

• Endogenous (voluntary) and exogenous (involuntary) attention typically enhance visual task performance.
• Existing models do not differentiate the effects of these two attention types, failing to explain observed behavioral differences.

Purpose of the Study:

• To develop a computational model distinguishing endogenous and exogenous attention effects on visual perception.
• To explain differential impacts of attention on texture segmentation across the visual field.

Main Methods:

• Developed a model emulating sensory encoding for figure-background segmentation.
• Incorporated distinct operating regimes for endogenous and exogenous attention based on spatial frequency.
• Validated the model against behavioral data from texture segmentation experiments.

Main Results:

• The model successfully emulates sensory encoding and predicts behavioral performance.
• It replicates the parafoveal advantage in texture segmentation.
• It explains uniform improvements by endogenous attention and peripheral benefits with foveal costs by exogenous attention.

Conclusions:

• A computational dissociation exists between endogenous and exogenous attention.
• The model provides a generalizable framework for predicting attentional effects on visual perception.
• Distinct attentional mechanisms operate across different spatial frequencies and visual field locations.