The ‘stickiness’ of disgust: Testing the hypothesis of disgust intractability in the context of instrumental reversal learning.

Abstract

Discrete emotion theories conceptualise fear and disgust as distinct emotions. This notion is supported by the finding that Pavlovian disgust conditioning is more resistant to extinction than fear conditioning—disgust may be ‘stickier’ than fear. Dimensional emotion theories challenge this, suggesting that categorical differences between emotions reflect variations along shared dimensions of valence and arousal. Reinforcement learning assumes a dimensional framework: assuming that learning from negative events depends on an outcome’s valence and arousal. Here, using a novel online reversal learning task and modelling framework (n = 340), we show that instrumental punishment learning is not ‘outcome agnostic’. Emotion-learning differs from points-based learning, and disgust-learning is uniquely ‘sticky’, altering perseverative error-rate, lose-shift probability, and a model-based ‘stickiness’ parameter, relative to other types of learning. This suggests that whilst some features of learning are shared between emotions, disgust has a distinct behavioural profile, with implications for emotion science and computational psychiatry.