Causal Inference in the Perception of Verticality.

Area of Science:

• Neuroscience
• Perception Science
• Human Factors

Background:

• Perceptual upright construction involves integrating visual and inertial sensory data with prior knowledge.
• Current models like Forced Fusion (FF) suggest a Bayesian vector sum, weighting signals by reliability.
• However, Cue Capture (CC) and Causal Inference (CI) models propose alternative mechanisms for sensory integration.

Purpose of the Study:

• To investigate the CNS's strategy for constructing perceptual upright.
• To compare the predictive power of FF, CC, and CI models using a novel experimental setup.
• To determine if a vector sum model adequately explains upright perception.

Main Methods:

• Developed an alternative-reality system for independent manipulation of visual and physical tilt.
• Recruited 36 participants to report perceived upright under congruent and incongruent visual-inertial stimuli.
• Compared empirical data against predictions from FF and CI models.

Main Results:

• The Causal Inference (CI) model demonstrated a better fit with participant data than the Forced Fusion (FF) model.
• This preference for CI became significantly clearer with larger sensory discrepancies (±60°).
• Findings indicate limitations of the vector sum approach in explaining upright perception.

Conclusions:

• The perception of upright is not comprehensively explained by a simple vector sum model.
• Causal Inference (CI) provides a more accurate framework for understanding how the CNS integrates sensory information for orientation.
• Future research should explore the nuances of CI in sensory perception.