How do reliability-weighting models explain individual differences in susceptibility to multisensory illusions?

Why would weighting explain who sees an illusion?

A multisensory illusion happens when one sense overrides or reshapes another. In the McGurk effect, for example, seeing a speaker's lips form one syllable while hearing a different syllable can change what listeners report hearing, and listeners vary considerably in how strongly they experience this shift [1]. Reliability-weighting models treat this as a balancing act: the brain estimates how trustworthy each sense is and leans on the more reliable one. If two people differ in how much they trust vision relative to hearing, they should also differ in how much a visual cue distorts what they hear.

The logic extends naturally to touch and body ownership. In the rubber hand illusion, a person watches a rubber hand being stroked while their own hidden hand is stroked at the same time, and some people come to feel the rubber hand is theirs. Susceptibility varies widely between individuals [3]. A weighting account would say this variation reflects differences in how much weight people give to vision versus touch, though the review of personality and trait predictors does not test that mechanism directly.

What actually predicts who is more susceptible?

The clearest evidence for a weighting-like explanation comes from perceptual resolution. In a battery measuring three size illusions, people with lower perceptual resolution, meaning a larger just-noticeable difference in detecting size, showed greater susceptibility to the illusions [4]. In weighting terms, a noisier or less precise signal carries less weight, so the illusory cue gains relative influence. That pattern is consistent with reliability-weighting, although the study measured visual illusions rather than multisensory ones and did not frame its results in those terms.

Brain activity points in a related direction. Among 63 participants, strong and weak McGurk perceivers differed in activation of auditory and visual regions, the left superior temporal sulcus, and multiple motor areas; crucially, engagement of the superior temporal sulcus and motor areas predicted individual susceptibility, while sensory regions did not [1]. The authors suggest that distinct multimodal integration in the superior temporal sulcus, together with coordinated motor processes, may serve as a neural substrate for these differences [1]. That is a proposal about where integration happens, not a direct measurement of reliability weighting.

Do personality and experience shift the balance?

Stable traits appear to matter, but in a split way. The rubber hand illusion review found that hypnotic suggestibility and performance-based personality features were more consistently related to proprioceptive drift, a spatial measure of the illusion, whereas sensory suggestibility, empathy, and self-esteem were more strongly associated with subjective feelings of ownership [3]. Other traits, including schizotypy, dissociation, temperament, and self-other representation, showed less consistent associations [3]. The review also notes that demand characteristics may influence self-reported ownership, so susceptibility appears to reflect both multisensory integration and stable psychological traits [3].

Experience may also shift the balance. A study comparing pilots with controls on sound-induced flash illusions examined susceptibility to fission and fusion illusions, in which a single flash paired with multiple sounds is seen as multiple flashes, or multiple flashes paired with one sound are seen as one [2]. The abstract notes that as the discrepancy between auditory and visual stimuli increases, the likelihood of multisensory integration decreases, and that individual differences have produced inconsistent results across studies of this illusion [2]. This is consistent with a weighting framework in which large mismatches reduce the credibility of combining the senses, but the study does not test weighting models directly.

How strong is the case for reliability weighting?

The honest answer is that the case is indirect. None of these studies directly manipulated or measured the reliability assigned to each sense and then showed that this assignment predicts illusion strength. What they establish is that susceptibility varies reliably between people [1][3][4], that perceptual precision tracks illusion strength in at least one visual domain [4], and that specific brain regions and traits correlate with susceptibility [1][3]. Sensory noise traits have also been reported to track individual differences in illusion susceptibility and in a multisensory mismatch response [5], which fits the idea that noisier signals are weighted differently, though the abstract gives only a brief indication of this link.

A further complication is that susceptibility may not be a single trait. The visual illusion battery found high test-retest reliability for each illusion but little correlation between different illusions [4], meaning a person who is highly susceptible to one illusion is not necessarily susceptible to another. If weighting were a single global setting, you might expect stronger cross-illusion consistency. Instead, the pattern suggests that how much each sense is trusted may depend on the specific pairing of cues and the task at hand, which is exactly the kind of context sensitivity reliability-weighting models can accommodate but which these studies have not yet pinned down.

Sources used in this answer

Related research questions