Why do trained musicians seem to separate what they see from what they hear more sharply than most people, while experienced wine tasters appear to lean more heavily on a wine’s color when judging its aroma and flavor? A new theoretical paper in Frontiers in Human Neuroscience argues that this apparent contradiction in the multisensory expertise literature is not a contradiction at all, but the result of conflating two formally distinct quantities that have rarely been kept apart.
The authors, Alessandro Bortolotti, Riccardo Palumbo, Fabio Lokwani-Di Matteo and Pier Luigi Sacco, propose a generative model in which two parameters are free to vary independently: the prior probability that two concurrent signals arise from a common source, and the coupling between the values of signals that do share one. The model also includes channel precision and a decision criterion, and it contains existing causal-inference models and coupling-prior models as special cases.
According to the analysis, the correlation between the values of two signals across an environment reflects only the product of the two priors. This means a learner that simply tracks correlations acquires the coupling whenever the signals always share a source, but separating the two priors requires representing causal structure itself. In other words, an expert can become more tightly coupled to a co-occurring signal without becoming any more likely to believe the signals come from the same object.
The framework also exposes methodological pitfalls. Binary common-source reports, the authors show, confound the common-cause prior with the decision criterion, so a person who says two signals belong together more often may simply be using a laxer criterion rather than holding a stronger prior. Moreover, the apparent strength of coupling trades off against the measurement noise of the influencing channel unless that channel’s precision is measured independently. The two priors do, however, leave distinct signatures in how cross-modal influence varies with disparity between signals, which offers a way to tell them apart empirically.
Applied to wine expertise, the model yields a striking result. When the source is fixed, as it is for the color and odor of a single wine, a tighter coupling by itself produces both greater displacement of odor judgments by color and better discrimination between wines, the conjunction reported for experienced tasters. No change in odor precision can produce that combination, the authors argue, which helps explain why experts can be simultaneously more accurate and more susceptible to color.
To test whether the framework is usable in practice, the authors ran simulations of a test battery combining continuous estimates, common-source reports and discrimination thresholds. The simulations showed that the parameters can be recovered, that a within-subject dissociation between the two priors can be detected with samples of conventional size under illustrative effect sizes, and, importantly, in which cases recovery fails.
The account extends beyond wine and music. The authors note that the same ambiguity attends claims about altered multisensory processing in autism and aging, where studies may be measuring changes in the base rate of common causation, changes in coupling strength, or changes in decision criteria without being able to say which. They propose a rule that classifies environments by their base rate of common causation and their coupling strength, offering a common vocabulary for comparing findings across domains.
The paper also delivers a pointed critique of the existing literature: the most cited demonstration of expert susceptibility to color, the authors state, contains no expertise manipulation at all. They suggest that apparent expert biases may instead reflect calibration to real-world statistics, in which experts have simply learned the true correlations between a wine’s color and its sensory properties in their environment.
As the authors note in their abstract, the article is a theoretical contribution with simulations rather than new experiments, and its conclusions about specific expert populations rest on reinterpreting previously reported findings. Its practical value will depend on whether future studies using the proposed test battery can indeed recover the parameters and reveal the predicted dissociations in human participants. If they do, the long-standing puzzle of why some experts seem sharper and others seem more biased may dissolve into a clearer question: which kind of statistical structure, common cause or coupling, each expert has actually learned.
Subject of Research: A computational model distinguishing common-cause priors from coupling priors in multisensory perception and perceptual expertise.
Article Title: Resolving the apparent paradox of multisensory expertise: coupling priors, common-cause priors, and how to distinguish among them
Article References: Bortolotti, A., Palumbo, R., Lokwani-Di Matteo, F., & Sacco, P. L. (2026). Resolving the apparent paradox of multisensory expertise: coupling priors, common-cause priors, and how to distinguish among them. Frontiers in Human Neuroscience, 20, Article 1960361. https://doi.org/10.3389/fnhum.2026.1960361
Image Credits: AI Generated
DOI: 10.3389/fnhum.2026.1960361
Keywords: multisensory perception, causal inference, coupling prior, common-cause prior, perceptual expertise, wine tasters, musicians, temporal binding window, cross-modal correspondences, parameter recovery, autism, aging
News Source: Ophelia Keating. (October 29, 2026). Why Experts Sometimes Trust Their Eyes Too Much: A New Model Separates Two Kinds of Multisensory Learning. Scienmag.
