Bayesian comparison of explicit and implicit causal inference strategies in multisensory heading perception

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psychophysics

Background:

• Multisensory perception enhances accuracy when cues share a common cause (e.g., visual and vestibular heading cues).
• The brain must infer causal relationships between sensory cues to optimize integration.
• It is uncertain whether humans use Bayesian inference, heuristics, or causal inference for heading perception.

Purpose of the Study:

• To investigate whether human observers perform causal inference in multisensory heading perception.
• To develop and apply a Bayesian computational framework for comparing causal inference strategies.
• To determine if explicit cause attribution and implicit heading discrimination tasks reveal causal inference.

Main Methods:

• Developed a Bayesian model comparison framework to assess causal inference strategies.
• Utilized an explicit cause attribution task to evaluate judgments of common cause.
• Employed an implicit heading discrimination task to assess cue integration strategies.
• Combined data from both tasks to rigorously test hypotheses.

Main Results:

• In the explicit task, subjects accounted for cue disparity in common cause judgments, but not always in a Bayesian manner.
• The heading discrimination task alone could not exclude a forced-fusion strategy (cue integration regardless of disparity).
• Combining evidence from both tasks allowed ruling out forced-fusion, supporting causal inference.

Conclusions:

• Human observers engage in causal inference during multisensory heading perception.
• A combined approach using explicit and implicit tasks is crucial for understanding complex perceptual strategies.
• The developed computational framework enables rigorous Bayesian analysis of perceptual decision-making under model uncertainty.