Causal inference shapes crossmodal postdiction in multisensory integration

In our environment, stimuli from different sensory modalities are initially processed within a temporal window of multisensory integration spanning several hundred milliseconds. During this window, stimulus processing is influenced not only by preceding and current information, but also by input that follows the stimulus. The computational mechanisms underlying crossmodal backward processing, which we refer to as crossmodal postdiction, are not well understood. We examined crossmodal postdiction in the Illusory Audiovisual (AV) Rabbit and Invisible AV Rabbit Illusions, in which postdiction occurs when flash-beep pairs are presented shortly before and shortly after a single flash or a single beep. We collected behavioral data from 32 participants and fitted four competing models: Bayesian Causal Inference (BCI), forced-fusion, forced-segregation, and non-postdictive BCI. The BCI model fit the data well and outperformed all other models. Building on previous findings that demonstrate causal inference during non-postdictive multisensory integration, our results show that the BCI framework can also explain crossmodal postdiction phenomena. Our findings suggest that the brain performs causal inference not only across concurrent sensory inputs but also across temporal windows, integrating information from past, present, and subsequent events across modalities to construct a unified percept.