Research gap at a glance
- Question: In a large unselected sample, is the distribution of heartbeat-counting error scores a single continuum, or a mixture of a small minority of genuine perceivers and a majority whose counts are driven by something other than cardiac perception?
- Why it matters: If the scores are a mixture, every continuous correlation computed across an unselected sample dilutes a real effect with noise, and the arbitrary accuracy cut-off could be replaced by an estimated boundary. If they are a continuum, the field’s standard practice of reporting continuous error scores stands.
- Theme: Measurement
- Status: open
What the wiki already knows
- Pooling 709 participants from seven studies, fewer than 10% of healthy adults met an accuracy criterion, and people who counted inaccurately reported feeling a regular rhythm slower than their real heart rate rather than missing beats (Does et al. (2000)).
- The mixture claim rests on histograms inspected by eye; no dip test, mixture model or formal test of unimodality appears in any source the wiki holds (Is the heartbeat-counting task a valid measure of cardiac perception?, Heartbeat detection task).
- Accurate-perceiver status may not be stable: fewer than half of panic patients who were accurate at baseline were still accurate after treatment, yet extreme-groups designs such as Pollatos et al. (2005) select on a single session (Does et al. (2000)).
- Across 133 studies and 11,524 adults, counting accuracy was essentially unrelated to trait anxiety, depression and alexithymia while tracking heart rate, BMI and sex, which fits a guessing account of most scores (Desmedt et al. (2022), Desmedt et al. (2023)).
What would answer it
- Fit finite mixture models and a formal unimodality test to Heartbeat detection task counting errors from a large unselected sample (pooled archives or several thousand new participants) with a retest session, estimate the number of components and each person’s posterior class membership, check that membership is stable across sessions, and then re-estimate a known correlate within each class rather than across the whole sample.
Connections
- Raised on: Is the heartbeat-counting task a valid measure of cardiac perception?, Heartbeat detection task, Does et al. (2000)
- Related concepts: Interoceptive sensitivity, Schema-guided symptom perception, State vs trait interoception
- Related debates: Is the heartbeat-counting task a valid measure of cardiac perception?, Is more interoceptive awareness better?
The wiki’s validity debate about the heartbeat-counting task comes down to two measurement models. On one, every participant perceives their heartbeat to some degree and the error score is a continuous index of that skill. On the other, a small minority perceive and everyone else produces a count from expectation, so the score pools two populations. Is the heartbeat-counting task a valid measure of cardiac perception? records that the two models imply different analyses, different sampling designs and different conclusions from the same data.
The striking thing is how cheap the test is. The categorical claim needs the error distribution to be a mixture, and the evidence offered for it in Van der Does et al. (2000) is three histograms described as a spike near zero error on a broadly normal distribution, which is equally consistent with a continuum. No formal analysis has been reported in anything the wiki has read, which makes this a question about existing data as much as about new data.
The answer also bears on the instability problem. If a mixture model fits, the stability of class membership across sessions can be estimated directly, which speaks to State vs trait interoception: a person who moves between classes from one session to the next is evidence that “accurate perceiver” is a state, not a type. It would also say whether extreme-groups studies such as Pollatos et al. (2005) were selecting a real subgroup or sampling the tails of one distribution. See A heartbeat counting score is not, by itself, evidence of interoceptive accuracy for how the wiki currently reads counting scores in the meantime.