Psychophysiological outcome responses in human Pavlovian fear conditioning: a prediction error analysis

Prediction errors (PE) are thought to drive associative learning. While neural signals consistent with PE encoding have been identified, the expression of PE in psychophysiological indices remains debated. Here, we sought to fill this gap by investigating responses to unconditioned stimulus (US) occurrence and probability in skin conductance responses (SCR), pupil size responses (PSR), heart period responses (HPR), and respiration amplitude responses (RAR). Data set 1 consisted of 8 published studies (N1 = 264) using differential fear conditioning with partial reinforcement (50%), and novel data set 2 (N2 = 29) parametrically varied US probability (20%/50%/80%). Across both data sets, all data modalities showed differential response to the US compared to US omission. In data set 1, there was evidence for responses to unexpected US omission in all modalities, but no responses were consistent with signed PE encoding. Similarly, data set 2 provided no evidence that US or US omission responses scaled with outcome probability, which is incompatible with PE encoding. In conclusion, all recorded psychophysiological signals responded strongly to US and less strongly to unexpected US omission, with no evidence of a signed prediction error encoding.