Recent studies have linked compassion with higher vagally mediated heart rate variability (vmHRV), a measure of parasympathetic activity, and meta-analytic evidence confirmed significant and positive associations. Compassion, however, is not to be confused with soothing positive emotions: in order to engage in actions aimed to alleviate (self or others) suffering, the pain should resonate, and empathic sensitivity should be experienced first. The present study examined the association between vmHRV and the empathic sensitivity and action components of trait and state compassion. To do so, several dispositional questionnaires were administered and two videos inducing empathic sensitivity (video 1) and compassionate actions (video 2) were shown, while the ECG was continuously recorded, and momentary affect was assessed. Results showed that (i) scores on subscales assessing the empathic component of trait compassion were inversely related to resting vmHRV; (ii) vmHRV decreased after video 1 but significantly increased after video 2. As to momentary affect, video 1 was accompanied with an increase in sadness and a decrease in positive affect, whereas video 2 was characterized by an increase in anger, a parallel decrease in sadness, and an increase (although non-significant) in positive affect. Overall, present findings support the notion that it is simplistic to link compassion with higher vmHRV. Compassion encompasses increased sensitivity to emotional pain, which is naturally associated with lower vmHRV, and action to alleviate others’ suffering, which is ultimately associated with increased vmHRV. The importance of adopting a nuanced perspective on the complex physiological regulation that underlies compassionate responding to suffering is discussed.