Hierarchical Bayesian modeling of interoceptive psychophysics.

Abstract

Interoception, the capacity to sense, perceive, and metacognitively appraise viscerosensory and homeostatic signals, is a growing focus in psychology and psychiatry. Adaptive psychophysical tasks now allow quantification of perceptual sensitivity, bias, and precision in cardiac and respiratory domains. However, accurately estimating these parameters often requires large numbers of trials or participants, posing practical challenges, especially in clinical research where participant availability and tolerance are limited. One approach to reduce participant burden while maintaining statistical rigor is to optimize data analysis. Here, we present hierarchical Bayesian models tailored for cardiac and respiratory interoceptive psychophysics that efficiently estimate sensitivity, bias, and precision at both individual and group levels. Using simulations and empirical data, we validate these models and demonstrate that they allow enhanced inference relative to conventional approaches. To support adoption, we provide openly accessible resources, including a tutorial on implementing and fitting these models in R (written with researchers without modeling expertise in mind) and an app for sample-size justification. These tools facilitate robust, efficient, and generalizable modeling of interoceptive performance, enabling rigorous studies even with limited trials or participant availability.