Causal Inference Modulates Auditory Spatial Recalibration in Accordance with Bayesian Principles

Poster Session A - Saturday, March 29, 2025, 3:00 – 5:00 pm EDT, Back Bay Ballroom/Republic Ballroom

Wenshu lou¹ (wenshulou795@gmail.com), Uta Noppeney¹; ¹Donders Institute for Brain, Cognition and Behaviour

Our sensory systems continuously adapt to changing sensory statistics in our environment. To maintain internal consistency and external accuracy with respect to the outside world, an observer must rapidly recalibrate auditory and visual senses. Crucially, evidence suggests that auditory spatial recalibration may inherently rely on causal inference—that is, determining whether audiovisual signals originate from a common source. To investigate the role of causal inference in recalibration, we presented observers with synchronous audiovisual (AV) signals at varying spatial disparities, followed by unisensory A or V signals. Observers reported: (1) whether the AV signals originated from a common source, (2) their confidence in this causal judgment, and (3) the perceived location of the subsequent unisensory signals. Psychophysics results show that recalibration of the less reliable A cue depends non-linearly on spatial disparity and is enhanced when observers perceive a common cause. Moreover, the uncertainty of causal judgment (i.e., confidence) also influences recalibration—observers recalibrated less when reporting higher confidence. To further examine the underlying mechanisms, we developed a set of observer models incorporating both Bayesian and heuristic strategies for recalibration and causal inference. Group-level Bayesian model comparison results indicate that Bayesian strategies outperformed their non-Bayesian counterparts in terms of BIC. Importantly, behavioral patterns were captured and well explained by Bayesian causal inference models but not by heuristic models. Our findings demonstrate that causal inference and its uncertainty modulate auditory spatial recalibration in response to varying spatial discrepancies, with underlying computations following Bayesian principles.

Topic Area: PERCEPTION & ACTION: Multisensory