Multidimensional Assessment of Interoceptive Awareness (MAIA)
Developed by Mehling, Price, Daubenmier, Acree, Bartmess & Stewart (2012, PLoS ONE) explicitly to move past the assumption that interoception is one thing you have more or less of. Its design goal was to parse beneficial from maladaptive functions of interoception: to allow “I am very aware of my body” to mean either mindful attunement or anxious body-scanning, and to tell them apart.
Three of its authors (Mehling, Price, Daubenmier) later co-wrote Farb et al. (2015) with Farb, which is why the MAIA sits so neatly inside that paper’s interoceptive-taxonomy — the instrument and the taxonomy come from an overlapping group with a shared motivation.
The eight subscales
| Subscale | What it captures | Items |
|---|---|---|
| Noticing | Awareness of body sensations | 4 |
| Not-distracting | Tendency not to ignore or distract from pain/discomfort | 3 |
| Not-worrying | Tendency not to experience distress or worry with pain/discomfort | 3 |
| Attention regulation | Ability to sustain and control attention to body sensation | 7 |
| Emotional awareness | Awareness of the connection between body sensation and emotion | 5 |
| Self-regulation | Ability to regulate distress by attending to body sensation | 4 |
| Body listening | Tendency to actively listen to the body for insight | 3 |
| Trusting | Experience of one’s body as safe and trustworthy | 3 |
Where it sits in the taxonomy
Under Farb et al.’s Box 1, the MAIA measures sensibility — “an individual’s personal account of their internal-sensation experience, including confidence in their own interoceptive ability and feelings of engagement,” gauged by questionnaire. It is explicitly not a measure of accuracy, which requires an objective criterion and is normally obtained via the heartbeat-detection-task.
This distinction is doing real work, not bookkeeping. The two dissociate: meditators show strong interoceptive attention tendencies without superior heartbeat-detection accuracy (Khalsa et al. 2008; Parkin et al. 2013) — see does-mindfulness-enhance-interoceptive-accuracy. And the accuracy/sensibility mismatch is itself a construct of interest — see interoceptive-trait-prediction-error. Any finding stated as “X is associated with better/worse interoception” on MAIA evidence alone is a claim about sensibility only.
The Not-worrying subscale as a hypervigilance detector
The subscale that carries the most weight in Oldroyd et al. (2019) is Not-worrying, which produced by far their largest effect (attachment anxiety r = -0.43, vs. |r| = 0.18-0.26 for everything else). This is the MAIA’s design paying off: attachment anxiety was associated with higher Noticing and much lower Not-worrying — the profile of someone who attends to their body a great deal and is frightened by what they find. A single-score body-awareness instrument would have scored that person as simply “high,” which is exactly the conflation the MAIA was built to prevent.
Read the reverse-keyed names carefully — high Not-worrying is good (little worry), and the interpretive burden falls on remembering that Noticing means nothing without it.
What it cannot do: measure a moment
The MAIA asks how you generally relate to your body — “always/never” response anchors, trait framing throughout. Wallman-Jones et al. (2023) needed to know how someone related to their body in the last fifteen minutes, ten times a day, and found nothing in the interoceptive literature that would do it. They used the six-item body subscale of the State Mindfulness Scale (Tanay & Bernstein 2013) instead — an instrument built for a different construct, with one item (“sensations caused by my surroundings — heat, coolness, the wind on my face”) that is frankly exteroceptive, and which their own limitations section says straddles self-reported accuracy and self-reported attention (Gabriele et al. 2022; Murphy et al. 2020) so that what it indexes is undetermined.
That is a gap in the instrument landscape rather than a criticism of the MAIA, but it is worth recording here because this is the page a reader arrives at looking for a sensibility measure. There is no validated state measure of interoceptive sensibility, and the reason it matters is that about half the variance in the one state measure anyone has fielded turns out to be within-person (state-vs-trait-interoception). A trait questionnaire administered once is averaging over that half without saying so.
The subscales do not belong to one construct — and the seam runs through the middle of the instrument
The page above treats the MAIA as a single sensibility instrument whose subscales must be read together (Noticing means nothing without Not-worrying). Ventura-Bort, Wendt & Weymar (2021) put five of its subscales into a principal component analysis alongside four emotion questionnaires, and the MAIA came apart:
| MAIA-2 subscale | loaded on |
|---|---|
| Attention regulation, Trusting | Sensibility — believed accuracy about internal states |
| Noticing, Emotional awareness | Monitoring — tendency to deploy attention inward |
| Body listening | both, equally |
The two factors were orthogonal by construction, and they behaved in opposite directions on the outcomes: Sensibility predicted higher granularity for negative emotions and correlated r = .58 with well-being; Monitoring predicted lower granularity, lower emotional intensity, and r = .12 with well-being. On the MAIA’s own Not-worrying subscale the split is signed: Sensibility r = .40***, Monitoring r = −.20*.
Three things this changes on this page.
The instrument’s design goal is only half met. The MAIA was built to separate mindful attunement from anxious body-scanning, and the Not-worrying section below shows it doing that well. But the separation runs between subscales rather than being a property of any of them — and the subscale that most obviously names attunement (Noticing) sits on the factor that predicts worse outcomes. Whoever reads Noticing as the good kind of body awareness has it backwards on this evidence.
A composite is worse than this page already says. The caution above is that a total score obscures the Noticing/Not-worrying interaction. The stronger version: a total score sums across two orthogonal dimensions with opposite relations to well-being, so its expected correlation with anything is a weighted average of a positive and a null-to-negative effect. Total MAIA scores should not be used, and their appearance in a source is a reason to discount it.
The subscales’ nearest neighbours are outside the instrument. MAIA Noticing loaded with the Mood Awareness Scale’s Monitoring subscale and the Trait Meta-Mood Scale’s Attention subscale — questionnaires about emotions, not bodies. MAIA Attention regulation loaded with TMMS Clarity and the reversed TAS-20. So the eight subscales are not eight facets of one bodily construct; they are samples from two general dimensions of internal self-knowledge that happen to be worded about the body. See interoceptive-taxonomy, alexithymia.
Two brakes. This is one PCA, forced to two components after four had eigenvalues > 1, in 109 German-speaking students — and only five of the eight subscales were entered (Not-distracting, Not-worrying and Self-regulation were held out as adaptability outcomes rather than predictors, which is itself an interpretive choice this page’s “beneficial vs maladaptive styles” framing should notice). And the study measured no interoceptive accuracy at all, so the split is internal to sensibility and says nothing about the accuracy relationship.
Two things it does that this page had not recorded: fail as an outcome, work as a moderator
Eusebio et al. (2022) randomized 58 adults to two yoga curricula differing only in interoceptive emphasis, and the MAIA behaved in a way worth separating into its two roles.
As an outcome it did not discriminate. MAIA rose over 10 weeks (β = 9.78, p < .001) with no group × time interaction (β = −2.49, p = .443). A curriculum built explicitly to cultivate interoceptive attention tendency and sensibility did not move this instrument more than one in which the instructor deliberately avoided mentioning the body. Two readings, and the study cannot separate them: expectancy raised self-report equally in both arms and masked a real difference, or the instrument is not sensitive to a 20-hour manipulation of exactly the thing it names. Either way, a MAIA change score is weak evidence that an intervention trained interoception — which matters for mabt and much of mindfulness-interoceptive-training, where self-reported MAIA gain is the headline interoceptive outcome.
As a moderator it discriminated sharply, and uniquely. The same MAIA scores predicted who benefited from which training, in a three-way interaction with group and time on SART accuracy (β = −0.004, p = .007) — and the identical analysis run on every other questionnaire in the battery (PSS, Ryff, AAQ-II, Brief-COPE, FFMQ) found nothing (all ps > .15). Baseline MAIA separately gated the movement arm: only the lowest scorers improved there.
That asymmetry is the useful lesson for anyone choosing this instrument. The MAIA appears to be a better individual-difference variable than a change score — better at telling you who someone is than whether they changed. Two caveats hold it loosely: N = 34 analysed, and Eusebio et al. used total MAIA scores throughout, which the section above argues should not be used at all. So the moderation is a moderation by an unknown mixture of the Sensibility and Monitoring factors, and which of the two is doing the work is unidentified.
The change score can discriminate arms and still be behaviourally inert
Eusebio (above) is the case where a MAIA change score failed to discriminate two active arms. Rai et al. (2026) is the complementary and more cautionary case: a MAIA change score that did discriminate — it rose more under a digital mindfulness therapeutic than under a cognitive-training control (β=1.83, p=0.044), serving as the study’s manipulation check — and was then inert with respect to behaviour. Interoceptive awareness went up and exercise endurance went up, but the two were uncorrelated across individuals (MAIA↔time-to-exhaustion r=−0.02, ns; MAIA↔performance factor r=0.05, ns), and MAIA mediated nothing. Its only strong correlate was self-reported attitude (r=0.74) — another questionnaire, inside the self-report cluster.
The lesson sharpens the one above rather than contradicting it. A MAIA gain can be real and group-specific and still tell you nothing about whether the person’s behaviour changed, because the gain lives in the self-report cluster and does not reach out to a behavioural outcome. So a manipulation check on this instrument confirms that an intervention was received as mindfulness training; it does not license the further step, common in this literature, of treating the MAIA rise as the mechanism of any downstream benefit. See does-mindfulness-enhance-interoceptive-accuracy, mindfulness-interoceptive-training.
Caution
Given that the strongest MAIA correlate reported in the wiki’s sources is with attachment anxiety, the concern Farb et al. raise about the Porges Body Perception Questionnaire — that its interoceptive validity is questionable because it may largely index anxiety — is worth keeping live for the MAIA’s affect-laden subscales too. It is a better instrument than that critique’s target, but it is still a questionnaire measuring how people feel about how they feel.