Heartbeat-evoked potential

The HEP is the wiki’s one neural measure of interoception that needs no behaviour from the participant. Every other cardiac measure here — counting, tracking, discrimination — asks the person to report something and then scores the report. The HEP instead averages the EEG relative to the heart itself and reads off the cortical response to each beat. In Quadt, Critchley & Garfinkel’s (2018) scheme it sits at the first, afferent-signal dimension — “the wiki’s cleanest neural index of the raw interoceptive signal, distinct from any perceptual report” — whose amplitude nonetheless correlates with detection ability.

What it is, mechanically

An ERP locked to the cardiac cycle rather than to a stimulus. Cut EEG epochs around the ECG R-wave, average them, and a deflection appears in a late window — most reports use roughly 200-600 ms after the R-wave, maximal at frontocentral sites. That late window is deliberate: it is meant to sit after the electrical wash of the heartbeat itself has passed, though in practice the two overlap and the artefact must be removed rather than waited out.

The cardiac field artefact is the thing that makes it hard

The HEP’s defining problem is that it is locked to the same event that contaminates it. The heart is a large electrical dipole; each R-wave projects a field onto the scalp — the cardiac field artefact (CFA) — that is time-locked to exactly the epoch the HEP occupies. Unremoved, it is a signal that varies with the heart’s electrical activity, so it can produce a between-condition “HEP” difference that is really a difference in cardiac electrophysiology.

This matters most where it is least visible: any comparison whose conditions or groups plausibly differ in heart rate or contractility (most clinical comparisons, and any arousal manipulation) differ in the CFA too. García-Cordero et al. (2017) handle it the standard way — R-wave-locked component removal by ICA plus visual inspection, comparing the HEP before and after removal — but the hazard is structural, not a one-study oversight. It is the electrocortical cousin of the cardiodynamic confound: a louder heart is both easier to detect and a bigger artefact to subtract.

What modulates it, and the interpretive tangle

Three things reliably move the HEP, and their overlap is the measure’s central ambiguity:

• Attention to the heartbeat. Directing attention inward enlarges the (frontal, more-negative) HEP. García-Cordero et al. (2017) is the wiki’s cleanest demonstration: both interoceptive conditions produced more negative frontal HEPs than a matched exteroceptive condition, ~200-500 ms, at scalp and (in the frequency domain) intracranially — a dissociation that appeared irrespective of accuracy, marking the attempt to attend inward rather than success at it.
• Detection ability. HEP amplitude correlates with heartbeat-detection performance across individuals (the correlation the taxonomy leans on), which is what makes it a candidate index of interoceptive-sensitivity and not merely of attention.
• Clinical state. HEP alterations are reported in depression, panic, depersonalization, and borderline personality disorder, and in Alzheimer’s disease the HEP modulation is one of four cardiac-interoceptive deficits (with accuracy, learning, and metacognition) that fall together — see bonaz-2021-diseases-of-interoception and interoceptive-psychopathology.

Because attention, ability, and arousal all shift it, a bare HEP difference is like a bare P300 difference: real, but under-determined. The methodological advance García-Cordero et al. argue for is a matched exteroceptive control condition, so that a clinical HEP abnormality can be read against internal-vs-external attention within the same person rather than against a resting baseline — which most of the patient literature lacks.

What the HEP is for: three frameworks it has been read under (Engelen et al. 2023)

The interpretive tangle above is about what moves the HEP. Engelen, Solcà & Tallon-Baudry (2023) organize the deeper question — what the HEP is for — into the three frameworks their whole review runs on, and the HEP is the object where the three collide most visibly:

• Emotion. HER amplitude depends on the valence of emotional stimuli and on mood and arousal — the somatic-marker intuition that cardiac afference colours feeling, in an evoked-potential register.
• Predictive coding. The HER reconceived not as a sensory response but as a prediction error — the discrepancy between the predicted and the observed cardiac signal — modulated by cardiac-driven prediction of external stimuli and by expected emotional content. This is the reading that makes the HEP a test object for interoceptive-inference rather than merely an index of attention.
• The self. HERs track self vs other in mental imagery, the strength of body ownership during bodily illusions, the stability of preference-based decisions, and — the review’s strongest claim — the subject of conscious experience: HER amplitude predicts conscious detection of near-threshold visual and somatosensory stimuli and indexes residual consciousness in post-comatose patients.

The clinical hinge worth carrying: depersonalization/derealization patients, who feel detached from their own body and emotions, do not show the normal modulation of the HER by interoceptive attention (Schulz et al. 2015; already on the interoceptive-psychopathology radar) — the review’s cleanest thread from this rhythm to the felt bodily self. See embodied-selfhood. Note this raises the interpretive stakes rather than resolving them: a measure that indexes attention, detection ability, arousal, emotion, prediction error and the self at once is powerful and badly underdetermined, which is exactly the caution this page opened with.

Relation to the visual ERP

The HEP and the affective-picture ERPs on the event-related-potentials page share machinery and share the two cautions that govern all scalp electrophysiology: trust the latency, discount the anatomy, and beware between-groups amplitude comparisons. They differ in what they lock to — an internal, self-generated, artefact-carrying event versus an external, experimenter-controlled one — which is why the HEP earns a separate page and why the CFA problem has no counterpart in visual ERP work.