Sensory reliability shapes perceptual inference via two mechanisms

Area of Science:

• Neuroscience
• Cognitive Science
• Psychophysics

Background:

• The brain integrates common sensory signals and segregates independent ones for coherent environmental perception.
• Human audiovisual integration aligns with Bayesian Causal Inference (CI), but decision strategies and reliability effects are unclear.

Purpose of the Study:

• To investigate decision strategies in implicit and explicit Bayesian Causal Inference (CI).
• To evaluate Bayesian CI across a range of sensory reliabilities.
• To elucidate how sensory reliability influences multisensory integration.

Main Methods:

• A psychophysics study presenting participants with congruent and discrepant audiovisual signals.
• Four levels of visual reliability were used.
• Participants performed auditory localization (implicit CI) and source judgment (explicit CI).

Main Results:

• Humans use model averaging for implicit CI, weighting spatial estimates by posterior probabilities.
• Explicit CI involves inferring a common source when posterior probability exceeds 0.5.
• Sensory reliability sharpens the integration window and influences signal weighting in common-source scenarios.

Conclusions:

• Bayesian Causal Inference (CI) is fundamental for integrating sensory signals with variable reliability.
• Decision strategies for implicit and explicit CI were clarified.
• Sensory reliability critically shapes multisensory integration through distinct mechanisms.