Hirao, T.; Terada, K.; Miyamae, M.; Yamada, M.

The heartbeat-evoked potential (HEP) reflects the cortical processing of cardiac afferent signals. However, it remains unclear whether trial-level interoceptive prediction errors can be quantified directly from spontaneous resting cardiac fluctuations and whether these model-derived errors are associated with HEP amplitude. Here, we applied a Kalman filter, implemented as a sequential Bayesian estimation procedure, to resting-state EEG and ECG recordings from 21 healthy adults to estimate trial-by-trial signed prediction errors in RR-intervals. Positive prediction errors reflected unexpected cardiac deceleration, whereas negative prediction errors reflected unexpected cardiac acceleration. Cluster-based permutation tests showed that unexpected cardiac acceleration was associated with greater fronto-centro-parietal HEP amplitude than unexpected deceleration in an early post-R-peak window, spanning FC1, CP1, Pz, CP2, Cz, C4 and FC2 from 215 to 250 ms. A Bayesian linear mixed-effects model further indicated a credible negative association between signed prediction error and HEP amplitude after controlling for respiratory phase and preceding RR interval. In a secondary connectivity analysis, unexpected acceleration was associated with stronger Cz-to-frontal beta-band phase synchrony during a later post-R-peak window from 250 to 500 ms. Exploratory individual-difference analyses suggested that neuroticism was negatively correlated with late frontal HEP amplitude during unexpected acceleration, but not during unexpected deceleration or when trials were pooled across conditions. These findings demonstrate that spontaneous cardiac fluctuations can be used to derive trial-level computational estimates of interoceptive prediction error and that these estimates are reflected in early HEP amplitude. They further suggest that the cortical processing of unexpected cardiac acceleration may be related to individual differences in affective personality traits.