Fig. 4. Influence of the cardiac cycle on the absPE-HEP and learning.

a Schematic description of the systole and diastole phases. In red and blue are the systole and diastole periods respectively. Below is a representation of two example trials on which outcome onsets happened at systole or diastole. We then looked at the EEG response locked to the next heartbeat. b Mean amplitude difference between the absPE-HEP of the first heartbeat after all outcomes presented at systole versus diastole (N = 32). A violin plot is used to present all individual participants’ averages, as well as the mean and SEM. c Mean amplitude difference in the absPE-HEP for the low salient outcomes presented at systole versus diastole across subjects (N = 32 participants). d Mean amplitude difference in absPE-HEP for the high salient outcomes presented at systole versus diastole (N = 32 participants) – in b, c data are presented as mean values ± SE e A logistic regression analysis showed that switch/stay (1/0) could be predicted by several predictors, including the Absolute PEs (participants were more likely to switch after a highly surprising outcome) but also residual absPE-HEP (predictors depicted from left to right: Systole/Diastole, Single-trial variability, Absolute PE, Systole/Diastole by Single-trial variability, Systole/Diastole by Absolute PE) f–k Results of the correlation between the regression coefficient for each participant between absPE-HEP and systole/diastole and the mean reward and learning rates in the task. In red, the fit of the robust regression. Any of these results remain true even when including a covariate indexing features of the external outcome type – reward and absolute PE from the model. Particularly, in f learning rates – all task blocks g learning rates – predictive blocks h learning rates – non-predictive blocks i reward – all task blocks j reward – predictive blocks k reward – non-predictive blocks. We did not adjust the p-value for multiple comparison.