Method at a glance
- Modality: multimodal self-report (21 interoceptive domains)
- Measures: Trait believed interoceptive accuracy — how accurately a person thinks they perceive their own bodily signals, deliberately excluding how much attention they pay to them. Sits in the self-report/accuracy quadrant of the Murphy–Catmur–Bird 2×2, alongside post-task confidence ratings and opposite the attention-flavoured BPQ and MAIA Noticing.
- Procedure: 21 statements of the form ‘I can always accurately perceive when I am hungry / breathing fast / need to urinate / am going to sneeze’, each rated on a 5-point agree–disagree scale (21–105 total, higher = greater believed accuracy). Instructions ask respondents to rate perception without external cues (taking one’s pulse does not count) and give a worked example of inaccuracy (feeling the need to urinate and then finding you do not). Domains span cardiac, respiratory, alimentary, thermoregulatory, nociceptive, sexual, vestibular-adjacent and affective-touch signals.
- Also called: IAS, Murphy Interoceptive Accuracy Scale
Strengths
- Fills the empty quadrant. Existing self-report instruments measure interoceptive attention (BPQ) or bundle attention with belief (MAIA, Body Awareness Questionnaire); the IAS asks about believed accuracy only, which is what makes the accuracy/attention distinction testable rather than stipulated.
- It is the one self-report instrument here that tracks an objective interoceptive task. Heartbeat-counting accuracy correlated with the IAS (r = .271 and .336 in two samples) and with no other questionnaire administered alongside it (Murphy et al. 2020).
- Psychometrically sound on the basics: α = .88–.90, 30-day test–retest r = .754 — statistically indistinguishable from the ICQ (.814) and the BPQ (.684).
- Domain-general by design. Where nearly every objective task is cardiac, the IAS samples 21 channels, so it is usable in populations and questions where the heart is the wrong signal (hunger, bladder, thermoregulation).
- Trait-level, so it complements moment-by-moment confidence ratings rather than duplicating them — and the two behave alike, which is evidence both are indexing believed accuracy at different timescales.
Limitations
- Factor structure unresolved. PCA gives two components (19% and 18.6% of variance) with no clean interpretation — plausibly signals perceivable from interoception alone versus signals with external correlates or social sanction (itch, bruise, wind, blood sugar). Confirmatory analysis rated the two-factor solution acceptable-but-imperfect (RMSEA = .072, CFI = .863, TLI = .847) and only modestly better than one factor. All published results use the total score.
- Validated against one channel. The scale asks about 21 domains; the criterion evidence is cardiac heartbeat counting. Given that interoceptive ability does not travel between organs, a total score summed across channels is validated at exactly one of them.
- The criterion evidence is small (r ≈ .27–.34 at n = 52 and n = 33, p ≈ .047) and its objective anchor is the heartbeat counting task, so it inherits that instrument’s validity problem wholesale.
- It is a belief measure with no ground truth of its own. A high score is a claim the respondent makes; nothing in the instrument distinguishes an accurate self-assessment from a confident one — the scale explicitly is not a measure of interoceptive insight (confidence–accuracy discrepancy was unrelated to it, r = .090 and .000).
- Strongly related to Alexithymia (r = −.43, −.57 with the TAS-20), which is either convergent validity or construct overlap depending on one’s view of what the TAS-20 measures.
- Discriminant validity against a general disposition to endorse positive self-descriptions rests on a single self-esteem control (r = .448 with the Rosenberg scale; the alexithymia relationship survived partialling it).
Connections
Developed by Murphy, Brewer, Plans, Khalsa, Catmur & Bird in Murphy et al. (2020) to occupy a quadrant of their 2×2 model of interoceptive measurement that had almost nothing in it: self-reported accuracy.
Why the quadrant was empty, and why it mattered
The Murphy–Catmur–Bird model crosses what is measured (accuracy of perception vs the body being the object of attention) with how (objective performance vs self-reported belief). Three quadrants were well populated — objective accuracy by the heartbeat tasks, self-reported attention by the BPQ and MAIA Noticing, objective attention (thinly) by Ecological Momentary Assessment (EMA). Self-reported accuracy had one instrument, the Interoceptive Confusion Questionnaire (Brewer, Cook & Bird 2016), with imperfect psychometrics, plus post-task confidence ratings — which are state-dependent and confined to whatever channel the task uses.
The gap was not cosmetic. Without a trait accuracy-belief measure you cannot test whether the accuracy/attention distinction is real, because every available self-report is on the attention side, so every self-report↔task comparison in the literature is confounded with the what axis. That confound is the wiki’s most-repeated null.
The items, and what they are trying to avoid
Every item takes the same frame — “I can always accurately perceive when I…” — across heartbeat, hunger, breathing, thirst, urination, defecation, taste, vomiting, sneezing, coughing, temperature, sexual arousal, passing wind, burping, muscle soreness, bruising, pain, low blood sugar, affective touch, tickle and itch.
Two design decisions are worth recording, because they are what distinguish it from the MAIA:
- No attention items. Nothing asks how often the respondent notices, monitors or listens to the body. The deployed-attention split is built into the instrument rather than discovered in it afterwards.
- No worked examples inside the items. Unlike the ICQ (“I frequently forget to eat”), IAS items stay abstract. The stated rationale is that an individual’s difficulty perceiving hunger may not manifest as the specific behaviour an example names — it may show up as overeating rather than as forgetting to eat.
The instruction to rate perception without external cues is doing real work: it is the only thing separating the IAS from a measure of how well a person infers their internal states by any available route, which is a different construct entirely.
Its one genuine claim on the wiki
The IAS is the only self-report instrument here that has been shown to correlate with objective interoceptive performance. That result is small (r ≈ .27–.34), doubly underpowered, and cardiac — but it is selective in a way that matters more than its size: in the same participants and the same sessions, the BPQ correlated with the same task at r = .091 and below.
Read across Interoceptive taxonomy (Farb et al. Box 1), that is the taxonomies’ central distinction doing empirical work for once. Read against Interoceptive sensitivity, it says the wiki’s accumulated “sensibility does not predict accuracy” nulls were mostly measured with attention scales, and the claim needs restating with the what axis held fixed.
Adoption
Used by Ventura-Bort, Wendt & Weymar (2021), where it loaded at +.79 on their “Sensibility” (believed-accuracy) component alongside reversed TAS-20 and MAIA Trusting — independent confirmation of where the instrument sits, from a lab with no stake in the 2×2.