Abstract
We live in a multisensory world: to estimate properties of the environment, for example, the location of a barking dog, there are usually multiple sensory cues available to base the estimate on. The final estimate of the dog's location depends not only on the multisensory cues, but also on a crossmodal common-cause prior—a prior belief accumulated from past experience about cues in different modalities coming from a common source. We examined whether short periods of exposure (160 trials) to consistently spatiotemporally congruent (association) or incongruent (dissociation) audiovisual stimuli affect observers’ common-cause prior. Crucially, we ensured that the audiovisual stimuli were perceptually rather than physically (mis-)aligned by measuring and adjusting each observer’s relative auditory and visual localization biases. During every trial of the exposure phase, observers localized the stimulus of one modality, cued after stimulus presentation, by adjusting a visual cursor and occasionally made an additional common-source judgment. To measure common-cause priors before and after the exposure phase, observers localized and judged the unity of synchronously presented audiovisual stimuli with a pseudo-random spatial discrepancy. To extract observers’ common-cause prior, we fitted a Bayesian causal-inference model jointly to localization responses and unity judgments made before and after the exposure phase. Significant changes in the common-cause prior emerged after the association and dissociation phases. The majority of observers showed a decrease in the common-cause prior after the dissociation phase; yet, an increase was found, too. Both effects have been reported separately in previous studies. Simulations of the causal-inference model for these opposing results will be discussed. Our findings provide robust evidence for experience-dependent updates of the common-cause prior after short periods of exposure to audiovisual stimuli.