Enhanced cardiac perception is associated with benefits in decision-making
Werner, Jung, Duschek & Schandry (2009), Psychophysiology 46(6):1123–1129, doi:10.1111/j.1469-8986.2009.00855.x (printed on the paper — no derivation needed). Received 13 Nov 2008, accepted 9 Jan 2009.
This is the study the somatic-feedback debate had named as the highest-value unread source on the page — the missing bridge between the Iowa programme and interoception — and framed as a prospective conflict with Dunn et al. (2010): Werner’s title asserts the main effect Dunn reports as null. Read first-hand, both the conflict and its resolution are sharper than the queue note anticipated. It is also the paper Nentjes et al. (2013) leaned on for “poor detection → disadvantageous IGT play,” so a psychopathy result the wiki already holds rests on this one. Senior author Schandry built the counting task the whole cluster runs on; this is the wiki’s second first-hand paper of his (after Pollatos et al. 2005).
What it did
Split 50 people into good and poor cardiac perceivers by the Schandry count (screened twice, ≥1 week apart, to stabilize the classification), match the groups on age/BMI/sex, and run each through a standard 100-trial Iowa Gambling Task. Record ECG throughout; administer a battery of personality controls. Ask one question: do people who perceive their hearts better choose better?
The answer is yes for which decks they pick and no for how much money they end with, and the gap between those two answers is the first thing to hold about the paper.
| index | good vs poor | statistic |
|---|---|---|
| disadvantageous (A/B) selections | good pick fewer | F(1,48)=5.32, p<.03, η²=.10 |
| advantageous (C/D) selections | good pick more | (same univariate test) |
| perception score × disadvantageous | negative | r=−.30, p=.02 |
| perception score × advantageous | positive | r=+.30, p=.02 |
| net gain (€) | 417 vs 143, ns | F(1,48)=1.51, p>.05, η²=.03 |
| perception score × net gain | — | r=.18, ns |
Why it matters here: the same design that makes Pollatos disputable
The debate page filed this as a conflict with Dunn’s null main effect and guessed the reconciliation would be “Werner’s task supplied bodies that were mostly right — Dunn’s moderation restated.” That guess is available but the paper cannot support it, because Werner never measured the body’s per-choice signal. ECG yielded heart rate, group means of which did not differ; no anticipatory-marker differentiation term was constructed. So Werner is not the IGT version of Dunn’s interaction. It is a main-effect, extreme-groups study — structurally the decision-quality twin of Pollatos’s arousal main effect, run on the same lab’s instrument (Schandry counting), with the same design and the same disputability.
That reframes the “conflict” cleanly. It is not Werner-positive vs Dunn-null as two estimates of one number. It is the wiki’s recurring pattern:
- extreme-groups / clinical designs find an accuracy↔outcome main effect — Pollatos (felt arousal), Nentjes (antisocial psychopathy), and now Werner (IGT choice);
- the one unselected, continuous, body-measured design finds no main effect — Dunn (r=.08), only a moderation.
Read through Van der Does et al., the design difference is not merely inflation. If the counting task is valid only for the accurate minority, Werner’s selection of a good group scoring .92–.95 is purification — isolating the subpopulation the task works on — while Dunn’s unselected correlation is diluted by the ~80% for whom the score is noise, attenuating toward zero. Werner’s poor group (.61) is, on that reading, largely people the task does not measure. Both the inflation and the purification stories predict exactly what Werner and Dunn jointly show, and this paper does not distinguish them. See is-the-heartbeat-counting-task-valid.
What the paper genuinely adds, and what it doesn’t
Adds: the two obvious third variables are measured and dead. No HR difference (so not arousal — a control Pollatos skipped and Wiens ran on SCR), and no personality difference across anxiety, impulsivity, sensation-seeking, and the Big Five. And it is a genuine IGT × interoception result, the pairing Dunn et al. (2006) named as the study to run — five years before Dunn ran a version of it on a different task.
Doesn’t: it does not measure the marker, so the somatic-marker reading is inference, not demonstration (same limit as Nentjes). It does not escape the cardiodynamic confound — a louder heart is both easier to count and a bigger anticipatory signal, so “perception” and “signal strength” are not separated. It does not run the moderation that would connect it to Dunn’s surface. And its “benefit” is on deck selection only: net gain did not differ, so the title over-reaches on the paper’s own numbers.
The no-interaction result is a quiet further complication. If markers are what let good perceivers learn to avoid the bad decks, the group gap should widen across blocks; instead the group main effect rides on top of an identical learning curve (F for the interaction = 0.31). A constant offset present from early blocks is not what a marker-acquisition account predicts — it is more consistent with good perceivers bringing a standing advantage (or a louder standing signal) to the task.
Bottom line
A real, oft-cited datum — good cardiac perceivers make more advantageous IGT choices — that is weaker on inspection than its title and its citation record suggest: extreme-groups (hence inflated and Pollatos-style disputable), marker-free (hence inferential about mechanism), cardiodynamically confounded, and significant on choices but not on winnings. Its best use in the wiki is not as proof that perceiving the body benefits decisions, but as the third accuracy↔outcome main effect from a selected sample sitting opposite Dunn’s single unselected null — which makes the design, not the direction of the effect, the thing the field has never resolved. See does-somatic-feedback-guide-decisions, is-more-interoceptive-awareness-better.