Synthesis at a glance
- Claim: Read a heartbeat counting (Schandry) score as a mix of heart-rate estimation, cardiac signal strength and, for a minority, perception. Treat a result as evidence about interoceptive accuracy only if it survives the checks below; otherwise cite it as evidence about counting.
- Confidence: moderate
Notes
- First synthesis page (2026-09-24). Distils Is the heartbeat-counting task a valid measure of cardiac perception? into a reading rule rather than restating it. Confidence is moderate, not high, because the experimental guessing papers (Desmedt et al. 2018; Zamariola et al. 2018; Ring & Brener 2018) and Schandry (1981) itself are known here only second-hand; they are Tier 1 in docs/find-and-ingest.md, and ingesting them should move this to high or revise it.
Connections
- Supporting studies: Does et al. (2000), Desmedt et al. (2022), Desmedt et al. (2023), Schulz et al. (2013), Wiens et al. (2000), Zoellner & Craske (1999), Ehlers (1993), Banellis et al. (2026), Quadt et al. (2018), Ferentzi et al. (2018), Legrand et al. (2022)
- Challenging studies: Murphy et al. (2020), Wiens et al. (2000), Pollatos et al. (2005)
The debate page Is the heartbeat-counting task a valid measure of cardiac perception? lays out the arguments. This page states where they land, as a rule for reading the rest of the wiki, because a large share of the wiki’s quantitative claims about “interoceptive accuracy” rest on this one task.
The claim
A counting score mixes three things:
- Estimation. Most people cannot perceive their heartbeat at rest (under 10% of healthy adults meet an accuracy criterion; Van der Does et al. 2000), yet nearly everyone produces a count. What they report is a steady rhythm slower than their heart, which is a guess shaped by beliefs about heart rate. That belief is measurable without counting: on the HRDT people judge their heart 7-8.5 BPM slower than it is, and the size of that bias predicts how few beats they count (Legrand et al. 2022).
- Signal strength. Anything that makes the heart louder raises the score with no learning: exercise, a cold-pressor stressor (Schulz et al. 2013), stroke volume. Across 11,524 adults the score tracks heart rate, BMI and sex, and none of anxiety, depression or alexithymia (Desmedt et al. 2022).
- Perception, for a minority. A discrimination task, which counting’s guessing cannot fake, finds the same rare accurate group, and in that group perception does psychological work (Wiens et al. 2000).
So the score is not meaningless. It is a mixture, and in an unselected sample the first two components dominate.
The reading rule
Before citing a counting result as evidence about interoceptive accuracy, check:
- Were beliefs about heart rate or time estimation controlled? If not, the result may be about estimation.
- Was heart rate (ideally stroke volume) controlled, or did the manipulation change it? If a manipulation raised heart rate, a higher score may be a louder heart.
- Is there a second, non-counting measure (discrimination, HRDT) that agrees?
- Is the claim about the heart or about “interoception”? Cardiac scores do not predict other organs (Banellis et al. 2026; Is interoception domain-general or organ-specific?), so even a clean cardiac result is a cardiac result.
A result that passes none of these is evidence about counting. Cite it that way: “heartbeat counting scores were higher in X”, not “X have better interoceptive accuracy”. This is the rule Desmedt and colleagues propose (2023), and Ferentzi et al. (2018) make the same recommendation from the cross-channel side (“heartbeat perception accuracy”, not “interoceptive accuracy”). The wiki adopts it.
What cuts against it
- Selective convergence. Murphy et al. (2020) found counting scores tracked a self-report of believed accuracy and not one of attention, twice. A pure guessing account has to explain that selectivity. The reply on the debate page (a loud heart shapes both the count and the belief) is untested.
- Where perception exists, it matters. The accurate minority is real and behaves differently (Wiens et al.; extreme-group designs such as Pollatos et al. 2005 may be selecting it). The rule does not say counting studies are worthless; it says an unselected continuous score mostly measures something else.
- Second-hand foundations. The experimental guessing studies are held only through reviews; see the note above.
Pages most affected
Counting results used as accuracy evidence on Dunn et al. (2010), Pollatos et al. (2005), Zoellner & Craske (1999), Ribera d’Alcalà et al. (2014) and Meessen et al. (2016), and the “accuracy” rows of Interoceptive taxonomy (Farb et al. Box 1) and Interoceptive sensitivity, should be read through the rule above. The null results (no group difference, no correlation) are less affected than the positive ones: a mixture dilutes true effects, so a null from counting is weak evidence of absence, and a positive result needs the checks.