The Relationship Between Interoception and Metacognition: A Pilot Study

Study at a glance

• Authors: Judith Meessen, Verena Mainz, Siegfried Gauggel, Eftychia Volz-Sidiropoulou, Stefan Sütterlin, Thomas Forkmann
• Year: 2016
• Journal: Journal of Psychophysiology, 30(2), 76-86
• DOI: 10.1027/0269-8803/a000157
• N: 24
• Design: Cross-sectional pilot, healthy students. Heartbeat counting (Schandry; 10 trials, 24-64 s) with ECG and a confidence rating after every trial; a judgment-of-learning memory task (120 word pairs) with confidence after each recall; MAIA Noticing and Attention Regulation; STAI, depression screen. Pearson and robust (bisquare) correlations.

Key findings

• Operationalizes Garfinkel & Critchley’s (2013) three facets in one sample: sensibility = MAIA Noticing and Attention Regulation; accuracy = heartbeat perception score (mean 0.60, Cronbach’s alpha = .95 over 10 trials); awareness = mean squared difference between trial accuracy and trial confidence (lower = better).
• The three facets were mutually uncorrelated (all r between -.04 and .31, none significant): awareness vs accuracy r = .13, accuracy vs MAIA Noticing r = .14, accuracy vs Attention Regulation r = .31. Only the two MAIA subscales correlated with each other (r = .58). Read by the authors as an independent replication of Garfinkel et al. (2015), with the MAIA in place of the BPQ.
• Built the same three facets for memory (self-rated memory, recall performance, confidence-performance mismatch). Interoceptive awareness was unrelated to metamemory awareness (r = .16), and heartbeat accuracy was unrelated to memory performance (r = .03): no sign of a shared metacognitive ability across body and memory.
• Within memory, better recall went with better-calibrated confidence (r = -.44, p = .03; not significant after Bonferroni correction). Within interoception, the corresponding accuracy-awareness link was absent (r = .13). The authors suggest everyday feedback on memory, absent for heartbeats, lets memory confidence calibrate.

Limitations

• n = 24 students, 19 of them women, with BMI restricted to 18.5-24: underpowered for the correlations it reports as absent. The authors call it a pilot.
• The awareness index is an absolute calibration error, not a metacognitive efficiency measure. It mixes confidence bias with confidence-accuracy correspondence and depends on performance level, so it is not the construct the wiki treats as interoceptive insight; see Metacognitive efficiency and metacognitive bias.
• The memory result the paper leans on (better performers better calibrated, r = -.44) is partly built into an uncorrected calibration score and did not survive Bonferroni correction, so the interoception/memory contrast it draws is fragile.
• Metamemory awareness was computed only over trials with a recalled answer, while interoceptive awareness used all trials; the authors acknowledge the two indices are not strictly comparable.
• The MAIA was completed after the heartbeat task, and heartbeat counting carries every objection on Is the heartbeat-counting task a valid measure of cardiac perception?.

Connections

• Methods: Heartbeat detection task, Multidimensional Assessment of Interoceptive Awareness (MAIA)
• Concepts: Interoceptive taxonomy (Farb et al. Box 1), Metacognitive efficiency and metacognitive bias, Interoceptive sensitivity
• Debates: Is interoception domain-general or organ-specific?, Is the heartbeat-counting task a valid measure of cardiac perception?
• Related studies: Banellis et al. (2026), Murphy et al. (2020), Quadt et al. (2018), Harrison et al. (2021)

A small pilot from Aachen that does two things the wiki has otherwise only read about: it measures all three of Garfinkel’s facets of interoception (sensibility, accuracy, awareness) in one sample, and it builds a matching three-facet set for memory, to ask whether knowing how well you perceive your heart has anything to do with knowing how well you remember.

What it found

The three facets did not correlate. Heartbeat counting accuracy, confidence-accuracy correspondence, and two MAIA subscales were mutually unrelated. This is the Garfinkel et al. (2015) dissociation, the backbone of Interoceptive taxonomy (Farb et al. Box 1), reproduced in an independent lab with a different sensibility questionnaire (MAIA rather than BPQ).

Body metacognition and memory metacognition were unrelated too (r = .16). If there were a general “insight” ability, people who knew when their heartbeat counts were right should also know when their recall was right. They did not.

How to read it

Count it as a replication, a weak one. At n = 24 the confidence intervals run from about -.4 to +.6, so “uncorrelated” means “no large correlation”. It agrees with the direction of Banellis et al. (2026) (N = 241), which is the study to cite for the null.

It fits Murphy’s boundary on the sensibility null. Murphy et al. (2020) argue that self-report tracks objective accuracy only when the questionnaire asks about believed accuracy; attention-flavoured scales do not. MAIA Noticing and Attention Regulation are attention-flavoured, and they did not track heartbeat accuracy here (r = .14, .31). So this paper adds to the null Murphy explains, rather than challenging her positive result.

Its “awareness” is not the wiki’s insight. The index is the mean squared gap between trial accuracy and trial confidence. That is a calibration error: it goes up with overconfidence (bias) and with poor tracking (efficiency) alike, and it depends on how well the person performed. Metacognitive efficiency and metacognitive bias separates exactly those parts, and the cross-domain comparison is where it matters. Banellis et al. found that confidence bias correlates across tasks while efficiency does not. A calibration score contains bias, so on that account it should have shown some cross-domain correlation; here it showed little (r = .16). Either the pilot is too small to see it, or scale use differed between a 0-100 heartbeat confidence rating and a recall confidence rating. The paper cannot tell these apart.

The memory contrast is weaker than it looks. The authors explain the memory accuracy-calibration link (r = -.44) by everyday feedback on memory. But an uncorrected calibration score tends to improve with performance anyway, so that link is partly built into the index, and it did not survive Bonferroni correction. The puzzle is the other half: why the same built-in dependence did not appear for heartbeats (r = .13), when the two tasks had similar score distributions (0.60, SD .15 vs 0.62, SD .18). One reading, consistent with Is the heartbeat-counting task a valid measure of cardiac perception?, is that counting scores and counting confidence are both driven partly by estimation rather than perception, which would decouple them. The paper does not test this. See Metacognitive efficiency and metacognitive bias on why cardiac metacognition is hard to estimate at all.

Placement

Adds a replication of the three-facet dissociation to Interoceptive taxonomy (Farb et al. Box 1), a cross-domain (body vs memory) metacognition datum to Metacognitive efficiency and metacognitive bias, and key-study links on Multidimensional Assessment of Interoceptive Awareness (MAIA) and Heartbeat detection task. No contradiction with existing pages: it agrees with them, at low power.