What does crossmodal postdiction actually look like?
The simplest example is intermodal apparent motion: a light flashes in one place, then a sound occurs in another, and you perceive the light as having moved toward the sound—even though the light itself never moved [3]. The later sound reaches backward, so to speak, and alters your experience of the earlier flash. The Illusory Audiovisual Rabbit illusion works similarly: a rapid sequence of flash-beep pairs followed by a single flash or beep makes you perceive extra flashes or beeps that were never presented [1]. In both cases, a stimulus that arrives after the critical event changes what you consciously see or hear.
These effects are not confined to one modality. A visual stimulus can postdictively influence an auditory percept, and an auditory stimulus can postdictively influence a visual percept [3]. The key feature is that the later stimulus in one sense modality revises the earlier percept in another—hence 'crossmodal' postdiction.
How is predictive integration different from postdiction?
Predictive multisensory integration is about anticipation, not revision. Your brain constantly generates expectations—based on prior experience and top-down models—about what sensory signals should go together, and it combines incoming inputs from different senses according to those expectations [2]. For example, when you see a person's lips move, your brain predicts the accompanying speech sounds and integrates the visual and auditory streams into a coherent percept. This happens as the signals arrive, not after the fact.
In a predictive framework, mismatches between what you expect and what you receive generate prediction-error signals, which the brain tries to resolve by adjusting its models or reweighting sensory inputs [2]. The direction of influence is forward: prior expectations shape how you process current input. In postdiction, the direction is backward: later input reshapes how you experienced earlier input [1][3].
Does the brain use the same machinery for both?
There is evidence that a common computational process—causal inference—may underlie both phenomena. In the Illusory Audiovisual Rabbit paradigms, a Bayesian Causal Inference model fit participants' behavioral data better than models that forced fusion, forced segregation, or ignored postdiction [1]. That model assumes the brain weighs the probability that sensory signals come from a common cause versus separate causes, and uses that weighting to combine them. The same framework had previously been shown to explain non-postdictive multisensory integration, and the new results suggest it can also account for postdiction across temporal windows [1].
This does not mean postdiction and prediction are the same thing. It means the brain may apply similar inferential rules in both directions—forward in time for predictive integration, and backward in time for postdiction. The finding is consistent with the idea that the brain integrates information from past, present, and subsequent events to construct a unified percept [1], but it does not establish that the mechanisms are identical.
Why does the difference matter for understanding perception?
The distinction matters because it changes what we think consciousness is doing. Predictive integration fits comfortably with the idea that perception is a proactive construction based on expectations. Postdiction is stranger: it implies that conscious experience can be edited after the fact, so that what you 'saw' or 'heard' is partly determined by what happened next [3]. This has been described as conscious perception behaving like revisionist history—the final percept is a story the brain tells after weighing all the evidence, including evidence that arrived late [3].
Crossmodal postdiction also opens a window into how the senses talk to each other. The review literature notes that postdictive effects can reveal crossmodal crosstalk that might otherwise be invisible, and that models originally developed for unimodal postdiction can be adapted to multisensory cases [3]. That suggests postdiction is not just a curiosity but a tool for mapping the brain's integrative machinery.
Sources used in this answer
Causal inference shapes crossmodal postdiction in multisensory integration
A Bayesian Causal Inference model explained crossmodal postdiction in the Illusory Audiovisual Rabbit illusions better than forced-fusion, forced-segregation, or non-postdictive models, suggesting the brain performs causal inference across temporal windows.
Predictive coding, multisensory integration, and attentional control: A multicomponent framework for lucid dreaming
A multicomponent framework proposes that predictive coding and multisensory integration, guided by attentional control, help explain lucid dreaming phenomena such as vivid imagery and kinesthetic sensations.
Crossmodal Postdiction: Conscious Perception as Revisionist History
Crossmodal postdiction occurs when later stimuli in one modality influence earlier stimuli in another, and three neuropsychological models of unimodal postdiction can be adapted to multisensory cases.
Crossmodal Postdiction: Conscious Perception as Revisionist History †
This review describes crossmodal postdiction phenomena, including intermodal apparent motion and the Illusory Audiovisual Rabbit illusion, and proposes adapting unimodal postdiction models to multisensory processing.
Sensory Noise Traits Predict Illusion Susceptibility and Multisensory Mismatch Negativity
Sensory noise traits are examined in relation to illusion susceptibility and multisensory mismatch negativity, with reference to both crossmodal postdiction and predictive aspects of multisensory integration.