Abstract

People with misophonia experience intense distress at everyday sounds, especially the orofacial sounds of others (chewing, sniffing). A striking, under-explained feature is the urge to mimic the trigger i.e. to reproduce the offending action oneself, which is prevalent, scales with severity, and, when enacted overtly, brings relief. We ask whether active inference, in which agents act to make their predictions come true, can explain this urge. We develop a graded series of formal models. A discrete partially observable Markov decision process (POMDP) casts the urge as the expected-free-energy– minimising policy once an orofacial trigger is inferred: hearing another’s orofacial sound engages the observer’s own motor model (a mirror-like misattribution, “I am doing this”), which — because the observer is not the author — generates a high-precision prediction error that cannot be resolved; mimicry resolves it by restoring authorship and recruiting the sensory attenuation that self-generated acts entails. In this work, we show that relief from mimicry, choice among coping routes (mimicry, withdrawal, dissociation), and the attenuation mechanism emerge from precision assigned to the sensory motor system. We then test the account against a survey (N = 676). The model’s mechanistic claims are supported and severity-independent: relief is tightly coupled to a restored sense of control (ρ = 0.67), and mimicry becoming spontaneous is experienced as involuntary (ρ = −0.59); exploratory factor analysis recovers, unbidden, the model’s core distinctions — sensitisation vs habit, agency vs habit, and orofacial vs non-orofacial triggers as separate factors. Its quantitative between-person predictions are weaker in cross-sectional data. We argue misophonic mimicry is Bayes-optimal action to reclaim agency over an aversive percept, and that its hardening into compulsion reflects habit and sensitisation as two separable learning channels

Here is my poster about this work

Here are the slides from my talk about this work