Interoceptive psychopathology

The clinical face of interoception: the claim, made as field consensus in Khalsa et al. (2018), that dysfunction of interoception is “an important component of different mental health conditions” — and that it does not respect diagnostic categories. This page holds the nosology (which disorders, which dysfunctions, and the cause-or-consequence question); the mechanism (hierarchical Bayesian accounts, allostatic self-efficacy, precision failures) is on its sibling page computational-psychiatry.

Transdiagnostic, not category-specific

The roadmap’s Table 3 lists interoceptive symptoms and signs across a wide span of conditions — panic disorder (palpitations, dyspnea, dizziness misread; elevated HR, exaggerated startle), depression (appetite and fatigue disturbance), eating disorders (hunger insensitivity, caloric-anticipation abnormalities), somatic symptom disorders (symptom report exceeding medical findings), substance use (craving/withdrawal interoception), PTSD (autonomic hypervigilance, depersonalization), GAD, depersonalization/derealization disorder (physiological hyporeactivity, detachment from the body), and autism (sensory hyper/hyposensitivity). The point of the list is that no single feature is diagnostic; the same broad disturbance surfaces differently across conditions.

Because of that, the paper argues for a dimensional, transdiagnostic approach — the Research Domain Criteria (RDoC) — over DSM categories: several interoceptive processes “might not be readily identified at the symptom report level relied on by clinicians,” so a mechanistic, cross-cutting probe can reveal dysfunction the diagnostic manual never coded. This is interoception offered as an RDoC construct: a knob that is turned differently in different illnesses.

Five candidate dysfunctions

The roadmap proposes that interoceptive investigation in patients might reveal any of five distinguishable failures — a useful checklist because it maps onto the wiki’s taxonomy rather than onto diagnoses:

1. Attentional bias — e.g. hypervigilance to bodily signals.
2. Distorted physiological sensitivity — blunted or heightened magnitude estimation under perturbation.
3. Cognitive bias — catastrophizing an anticipated bodily stimulus (the panic engine; close to anxiety-sensitivity).
4. Abnormal sensibility — a tendency to label one’s experiences in a particular way.
5. Impaired insight — poor confidence–accuracy correspondence on a task.

The wiki already holds worked examples of several. anxiety-sensitivity is #3 in nearly pure form — the belief that arousal is harmful, dissociable from trait anxiety and from accuracy. schema-guided-symptom-perception is a mechanism for #2 (reporting a racing heart that is not racing). The attachment-anxious “high noticing, low not-worrying” profile (oldroyd-2019-attachment-interoception) is #1 and #4 together. The roadmap’s contribution is to file them under one heading and give them a common vocabulary.

A sharper parse: two pathways, not five dysfunctions (Paulus & Khalsa)

The roadmap’s five dysfunctions are a descriptive checklist keyed to the taxonomy. The LIBR group’s own active-inference account (Paulus, Feinstein & Khalsa 2019), restated compactly in the Paulus & Khalsa (2021) editorial, collapses the mechanism into two failure modes of the interoceptive prior:

1. Hyperprecise priors — expectations about the bodily states a situation elicits are held too strongly, swamping sensory evidence. The anxious person who meets a cat in a dark alley stays afraid because “shallow breathing + sudden sound = assault” is weighted with overwhelming precision.
2. Context rigidity — those expectations fail to update when the environment changes. Emerging into a bright open space, the same person keeps feeling threatened because the model is not revised.

Both are faults of the prior, but one is a precision fault (too-strong) and one a plasticity fault (won’t move) — a cleaner mechanistic dissociation than the five-item list, and the frame the editorial uses to read Nord et al.’s mid-insula disruption as “a reduced ability to match the model to how the body is feeling, or vice versa.” The mechanism half lives on computational-psychiatry; it is filed here because it is the parse that connects the five dysfunctions above to the neural locus below.

Cause or consequence — the unanswered question

The roadmap flags, without resolving, the question the wiki keeps hitting: is interoceptive dysfunction a cause of psychopathology, a consequence of it, or a correlated third thing? “Determining whether interoceptive processes are a cause or consequence of developmental psychopathology… will be an important area for future research.” This is the same gap the clinical-training material runs into — the deficit-to-treatment inference (“patients have less X, therefore raise X”) that the anorexia and depression cases show can fail, because a cross-sectional deficit does not say whether it is the illness, a defence against it, or a byproduct. The roadmap recommends developmental and longitudinal samples (younger and older cohorts, premorbid tracking) as the way to break the tie — which is exactly what none of the wiki’s clinical sources have.

The biomarker aspiration, and its current emptiness

If interoceptive dysfunction is real and mechanistic, objective interoceptive measures could serve as biological indicators of disease. The roadmap wants this and admits the cupboard is bare: “there is currently limited evidence for interoceptive predictors of diagnostic, prognostic, or treatment status.” Its bet is that the sensitive measures will be perturbation probes that engage metacognitive beliefs — an interoceptive “cardiac stress test” that stresses the system under ecologically valid conditions rather than reading it at rest. Note the tension with the wiki’s heartbeat-task material: the roadmap uses the resting cardiac tasks as accuracy measures without caveat, while betting the clinically useful measures are perturbation-based — quietly conceding the resting tasks may not be where the biomarkers are.

The transdiagnostic claim’s most direct neural support (Nord et al. 2021)

Everything above asserts that interoceptive dysfunction cuts across disorders; Nord, Lawson & Dalgleish (2021) is the study that tests whether the neural disruption actually converges, and finds that it does. A preregistered ALE meta-analysis of 33 neuroimaging experiments (626 patients, 610 controls) spanning depression, bipolar disorder, anxiety, remitted anorexia and schizophrenia — and interoceptive tasks from heartbeat detection to hunger, pain, breathing load and affective touch — yields a single convergent cluster of disrupted activation: the left dorsal mid-insula. This is the transdiagnostic, domain-general version of the “reduced dorsal-mid-insula reactivity in depression” this page carries from Quadt et al., and it is the strongest neural content the frame has.

It speaks directly to two of this page’s standing problems:

• The empty biomarker cupboard. Nord et al. explicitly cast the mid-insula as “a neural marker of psychopathology and a putative target for novel interventions,” and their conjunction analyses support the treatment-target half: the locus does not overlap with the regions antidepressants or psychotherapy move (assessed mostly with non-interoceptive tasks). So the biomarker aspiration two sections up gets its first concrete candidate — with the honest rider that a convergent group-difference in task-evoked activation is a marker, not yet a validated diagnostic or prognostic indicator, and that no included study measured interoceptive behaviour in the scanner at all.
• The single-channel worry, partly disarmed and partly not. Because the convergence holds across cardiac, respiratory, nociceptive, hunger and touch tasks, this is the page’s least cardiac-dominated evidence — a shared site of disruption independent of which channel probed it. But it is site-level, not ability-level: Nord et al. could not run domain-specific subanalyses for lack of power and concede the channels “are not necessarily integrated,” so this convergence is compatible with Banellis et al.’s behavioural decorrelation rather than a refutation of it. A common pathology site need not imply a common trait. See is-interoception-domain-general.

What it does not do is resolve cause-or-consequence: a convergent activation difference in patients is as compatible with consequence as with cause, exactly the gap flagged above. And its brakes are the meta-analytic ones — coordinate-based (no effect sizes), a surprising left laterality possibly driven by verbal task probes, and no adjudication of whether the locus is common to all interoceptive-difference conditions or specific to the psychiatric ones it examined (chronic pain, functional GI, functional neurological and connective-tissue disorders were excluded). The mechanistic reading — the dysgranular mid-insula as an interoceptive prediction-error comparator — is on computational-psychiatry and insular-cortex.

The funder’s own verdict on the whole literature (Chen et al. 2021)

Everything above this line is a body of work asserting that interoceptive dysfunction figures in psychiatric and neurological disease. Chen et al. (2021) recite the same list — migraine and chronic pain, alcohol and substance use, anxiety, depression, PTSD, OCD, autism, eating disorders, somatic symptom disorders, stroke and neurodegeneration — and then append the sentence this page should be read with:

A wide range of diseases and disorders are related to interoception, and pathophysiological and causal evidence has been generated for some of these conditions. However, it is important to note that, in most of them, the links to interoceptive dysregulation are largely symptomatic and descriptive.

That is the “cause or consequence” gap two sections above, stated not as an open research question but as a summary judgement on the existing evidence — and stated by the agencies funding it, in the paper the rest of the field cites as its charter. It is the most useful thing in an otherwise thin survey section, because every other source here has an interest in the transdiagnostic claim being strong.

Two qualifications, so this is not overread. “Most of them” is not “all of them” — Chen et al. explicitly allow that pathophysiological and causal evidence exists for some conditions, and the wiki’s own best cases (the lesion dissociations in Bonaz et al., acquired alexithymia from insular damage) are plausibly among them. And “symptomatic and descriptive” is a claim about the current state of evidence, not a prediction that the links are spurious. The remedy they name is the conventional one: pathophysiological studies in clinical populations and in appropriate animal models.

The conditions worked through: Quadt et al. (2018)

Where the roadmap gives a table and a checklist, Quadt, Critchley & Garfinkel (2018) work six conditions through the IPP machinery one at a time — the wiki’s most detailed disease-by-disease treatment, and the source that fills in the conditions the roadmap only listed. The organizing device is Van den Bergh et al.’s (2017) inferential account of symptoms: interoceptive signals usually stay below awareness (low prediction error, within the expected range) and surface only when they generate precise prediction error; they become symptoms when the highest-posterior-probability hypothesis encodes an aberrant, disease-related cause. Each condition is a different failure of that inference:

• Sickness behaviours and fatigue — the immune-to-brain arm the wiki’s cardiac-heavy material had been missing. Inflammation signalled via vagus, humoral cytokines and microglia evokes a stereotyped fatigue/anhedonia/withdrawal syndrome carried experientially by the insula; the shared motivation-and-anhedonia profile is Quadt et al.’s proposed bridge from inflammation to depression. Fatigue gets a top-down reading too (low allostatic self-efficacy) on computational-psychiatry.
• Depression — not a simple accuracy deficit. Healthy accuracy correlates with felt-emotion intensity (raising the prospect of “numbness” as low accuracy), but the empirical picture is “more complex”: the moderately depressed show impaired accuracy while the more severely depressed can match controls (possibly a medication effect). What is firmer: reduced dorsal-mid-insula reactivity, reduced HEP amplitude, and increased amygdala–mid-insula resting connectivity predicting severity. The EPIC “locked-in” account (Barrett, Quigley & Hamilton 2016) supplies the mechanism — over-predicted metabolic demand, downweighted error, a self-maintaining cycle enlisting sickness behaviours to conserve energy. The wiki holds one depression source first-hand, Wiebking et al. (2010), and it sharpens two of the claims above. Its “reduced dorsal-mid-insula reactivity” is shown to be, in part, a rest-baseline effect: the group difference appeared during exteroceptive stimulation but dissolved under correction for the preceding rest period, leaving an abnormally under-deactivating insula at rest rather than a blunted stimulus response — a caution that other studies’ “reduced reactivity” may be baseline effects wherever the design cannot separate them. And it puts the depressive abnormality on the sensibility side, not the accuracy side: on the BPQ, patients scored higher on perceived bodily reactivity but not on the awareness subscale, and their body report was decoupled from insula rest activity (the healthy body-perception↔right-anterior-insula correlation was absent in MDD). See embodied-selfhood for the “material me” framing this comes wrapped in.
• Autism spectrum conditions — divergent accuracy findings (intact in one child sample, impaired in another), possibly driven by comorbid alexithymia rather than autism per se. The signature is a discrepancy: reduced accuracy paired with elevated self-reported sensibility — an enlarged ITPE — read as an imbalance of sensory-evidence precision relative to priors. But the wiki’s dedicated ASD review, DuBois et al. (2016) (earlier and more focused than Quadt et al.), shows the elevated-sensibility half is contested: two of the three sensibility-relevant autism studies find sensibility reduced, and the review reads the literature overall as tending toward hyporeactivity. Consolidated on autism-interoception.
• Anxiety disorders — the mixed accuracy literature (up, down, and null across studies) reconciled by the ITPE: it is the accuracy–sensibility discrepancy, not accuracy alone, that predicts trait anxiety in both autistic and control samples.
• Eating disorders — impaired interoceptive self-report (EDI) predicts ED vulnerability longitudinally (a rare premorbid interoceptive signal), while cardiac-accuracy findings in anorexia and bulimia are inconsistent and confounded by comorbid alexithymia/depression/anxiety; reduced AI and dorsal-mid-insula activation during heartbeat and stomach attention.

The through-line is Quadt et al.’s methodological warning, which the wiki should keep: ED and autism studies “often do not take into account comorbidities” (alexithymia especially), and “objective and subjective dimensions of interoception” get used interchangeably — so a good deal of the disease literature may be measuring the wrong dimension in the wrong-controlled sample. This is the taxonomy problem cashed out as a clinical-evidence problem.

Past psychiatry entirely: Bonaz et al. (2021)

The roadmap gives a table, Quadt et al. work six psychiatric conditions through one mechanism, and Bonaz et al. (2021) then take the same frame out of mental health altogether. This is the wiki’s first source arguing that gastroenterology, urology, rheumatology and neurology contain interoceptive disorders in the same sense psychiatry does.

What it adds to this page, beyond breadth:

• A neuraxis claim in place of a nosological one. Dysfunction can originate at any level — receptor, brainstem, CAN, visceromotor cortex, metacognitive belief — and propagate in both directions. So the organizing question is not which disorder but at which level, and the same interoceptive failure can surface as a gut symptom, an autonomic symptom, or a mood symptom depending on where it enters. This is Stephan’s allostatic-self-efficacy circuit (allostasis, computational-psychiatry) promoted from a theory of fatigue and depression to a general theory of psychosomatic medicine.
• Three proposed shared clinical features across these disorders: anticipatory symptom-related anxiety, central sensory amplification, and aberrant autonomic activation. If the transdiagnostic claim has empirical content beyond the family resemblance, this is where it lives.
• Neurological dissociations by lesion, which are cleaner evidence than anything psychometric on this page — viscerosensory and somatosensory heartbeat experience coming apart after insula/ACC damage; the four-way interoceptive deficit in Alzheimer’s mapping onto accuracy, HEP, learning and metacognitive awareness; interoceptive deficits predicting seizure frequency in functional seizures. See insular-cortex.
• Somatic conditions with the same discrepancy signature. IBS patients show high sensibility with impaired processing; functional motor disorder patients show abnormal bodily attention allocation; chronic-pain shows accuracy predicting severity while sensibility and awareness findings stay inconclusive. The accuracy-low/sensibility-high profile the wiki learned from Garfinkel’s autism and anxiety work turns up in literatures with no contact with it.
• Early adversity as the shared upstream determinant, with a rodent maternal-deprivation model producing visceral hypersensitivity, gut dysfunction, raised HPA response and anxiety in the same animal — the review’s best demonstration that one early perturbation expresses across organ systems.
• joint-hypermobility, the awkward case: a heritable connective-tissue variant with heightened interoceptive sensitivity, insula reactivity, anxiety and dysautonomia. Interoceptive dysfunction originating in tissue rather than in representation, which nothing else on this page accommodates.

The caution to carry forward is that this breadth is bought at a cost in rigour. Bonaz et al.’s unifying claim — similar network alterations across unrelated brain-body disorders — rests on thin citation, causal direction is unestablished nearly everywhere, and the review leans on the resting cardiac tasks throughout without the validity caveat Quadt et al. supply from the same research group. See is-the-heartbeat-counting-task-valid.

The transdiagnostic frame has a measurement precondition it has not met (Banellis et al. 2026)

Everything above — the five dysfunctions, the Table 3 nosology, the biomarker aspiration — presupposes that “interoceptive dysfunction” is a property a patient has, expressible across their body and detectable in whichever channel is convenient to test. That presupposition has now been tested and failed.

Banellis et al. (2026) find cardiac and respiratory interoceptive sensitivity, precision and metacognitive efficiency uncorrelated in 241 healthy adults. Their own clinical inference is stated plainly: this “challeng[es] the assumption that processing deficits in one organ system reflect a more general interoceptive dysfunction,” and they note that many studies generalise from a single modality, “typically cardiac.”

Three consequences for this page, in ascending order of awkwardness.

The evidence base is mostly one organ. Nearly every deficit catalogued above — in anxiety, depression, eating disorders, autism, substance use — was measured cardiacally. If dysfunction is organ-specific, the transdiagnostic claim is currently a claim about cardiac perception across many disorders, which is a narrower and less interesting thing than it reads as.

Panic is the sharpest case, and it is the wrong way round. cognitive-model-of-panic and anxiety-sensitivity are largely about misread respiratory sensation — hyperventilation, air hunger, suffocation alarm — and the wiki’s evidence for them is overwhelmingly heartbeat tasks (ehlers-1993-interoception-panic, zoellner-1999-interoceptive-accuracy-panic, van-der-does-2000-heartbeat-perception-reanalysis). The decorrelation says those scores do not speak for the respiratory channel the theory is about.

But the null is from healthy adults at rest, which is exactly where it is weakest as a clinical argument. A general factor could exist in clinical populations, or emerge only under the physiological perturbation that clinical interoception is about — sahib-khalsa’s standing position, and the escape the authors themselves flag. So this section is a scope warning on the existing evidence, not a refutation of the transdiagnostic hypothesis. The discriminating study — the same multi-channel battery in a patient sample, and under challenge — has not been run. See is-interoception-domain-general.

The direction that runs the other way (Banellis et al. 2026, PNAS)

Every disorder above is characterized by an interoceptive deficit or distortion, and the implied remedy is to correct it. Banellis et al. (2026) supply the page’s first evidence that a form of spontaneous bodily attention is associated with lower symptom burden — in 536 healthy adults, thoughts about the heart, bladder and skin during a resting scan correlated negatively with both ADHD (rs −0.104 to −0.211) and depression (rs −0.105 to −0.149).

Three things this adds, and one it does not.

It gives ADHD an interoceptive entry on this page. ADHD appears nowhere in the roadmap’s Table 3, in Quadt et al.’s six conditions, or in Bonaz et al.’s neuraxis survey. Here it is the strongest symptom correlate, and in a coherent direction: ADHD is characterized by attentional lapses and physiological hypoarousal, and a partial correlation analysis found it specifically linked to fewer embodied heart-related thoughts alongside more repetitive self-related content.

It separates depression’s interoceptive profile from its cognitive one. Depression here is not primarily a body story — its positive correlate is mental time travel (past and future, rs 0.123 to 0.164), which is the rumination literature’s prediction, and its body correlate is a reduction in bladder-related content. On this dataset, what marks depressive symptomatology is temporal decoupling from the present, and what marks its absence is presence in the body. See cognitive-reactivity and body-wandering.

It sharpens the cause-or-consequence problem rather than solving it. All of the above is cross-sectional with rs ≈ 0.10–0.21 in a sample where only 2% exceeded the MDI cutoff for high depression. It is compatible with “attending to the body protects,” with “being unsymptomatic frees attention for the body,” and with a common third cause. The roadmap’s prescription — developmental and longitudinal designs — applies here exactly as it applies to every other row on this page.

What it does not do is rescue the biomarker aspiration. A correlation of −0.15 between a thought probe and a symptom score is not a diagnostic indicator, and the measure is sensibility — self-reported attention with no ground truth — which is the dimension this page has repeatedly warned gets used interchangeably with objective ones. Read alongside the section above it: the decorrelation paper says the transdiagnostic frame lacks a person-level referent for performance, and this paper shows that what does behave person-level is a relationship to the body. Both are from the same group and the same cohort, and together they point the clinical programme away from “which organ does this patient read badly” and toward “how does this patient inhabit their body” — a shift in construct, not a supply of biomarkers.

Two conditions the wiki lacked, read emotion-first (Greenwood & Garfinkel 2025)

Quadt et al. above work six conditions through inference; Greenwood & Garfinkel (2025) work them through emotion, and their table adds schizophrenia and emotionally unstable personality disorder (EUPD, borderline) — neither of which appears in the roadmap’s Table 3, Quadt et al.’s six, or Bonaz et al.’s neuraxis survey.

• Schizophrenia. Reduced HRV and elevated heart rate (a Swedish cohort of >1M men found resting HR > 82 bpm at conscription predicting a 21% higher later risk of the diagnosis), greater cardiovascular-disease risk, and altered cardiac structure/function tied to polygenic risk. Reduced cardiac accuracy tracks both negative and positive symptoms. The distinctive interoceptive angle is self–other and body-ownership disturbance: diminished insula deactivation when observing another being touched (impaired suppression of self-referential affective arousal), read as an early sign of the self-disturbance. Framed as an interoceptive-predictive-coding failure in which both hypoprecise and hyperprecise priors relative to sensory data generate large prediction errors — positive symptoms from misattributed sensory input under low prior precision, negative symptoms from a hindered construction of emotion. This is computational-psychiatry’s precision story extended to a condition the wiki had no interoceptive account of.
• EUPD (borderline). Emotional lability, altered sense of self, heightened negative affect; at rest, higher HR/BP and lower HRV and baroreflex sensitivity, with blunted HR acceleration during stress anticipation. The interoceptive marker is HEP amplitude: resting HEP is positively associated with emotion-regulation ability and is reduced in EUPD, with remitted patients falling between patients and controls — a candidate state marker that tracks symptom improvement. (Elevated HEPs have also been reported, plausibly reflecting distinct endophenotypes — alexithymia, PTSD comorbidity.) Mechanistically: hyperreactive salience regions (amygdala, insula) with under-recruited frontal control regions.

Both fit this page’s organizing claim — no single dysfunction is diagnostic, and each condition loads a different dimension (schizophrenia: representation + accuracy + self-model; EUPD: neural representation via HEP). And both inherit the page’s standing caveat, which the review states plainly in its own Limitations: the evidence is cross-sectional and correlational, cause-or-consequence is unresolved, and much of the accuracy data is cardiac — the single-channel problem two sections up, now spanning eight conditions instead of six.

The comorbidity premise the transdiagnostic move assumes (Barr et al. 2022)

The argument at the top of this page has two premises: that interoception is disturbed across many disorders, and that those disorders genuinely cut across each other rather than being clean, separable categories. The interoception sources only ever argue the first. Barr, Bigdeli & Meyers (2022) — a psychiatric-epidemiology Brief Report, not an interoception study — supply population-scale evidence for the second.

In 329,038 All of Us biobank participants, EHR-recorded psychiatric diagnoses were heavily comorbid: ~51% of diagnoses involved overlap, most participants with any disorder carried ≥2, cross-disorder tetrachoric correlations ran r = 0.43–0.75, and the commonest comorbid cluster was mood + anxiety + substance use (64% of the multi-diagnosis group). The authors read this as evidence the disorders “may in part share common causes.” That is exactly the co-occurrence structure a dimensional, mechanism-first nosology (RDoC, the p-factor tradition) is built to explain — and the structure the transdiagnostic-interoception claim quietly presupposes when it proposes interoception as one such cross-cutting knob.

Two limits keep this in its place. It says nothing about interoception: a shared-common-causes reading of comorbidity is compatible with a transdiagnostic interoceptive mechanism, but equally with shared genetics (the authors’ own follow-up direction), shared early adversity, or ascertainment overlap — so it warrants the frame, not the specific claim, and lands squarely back on the cause-or-consequence gap two sections up. And it is a treatment-selected biobank with a documented healthy-participant bias (every disorder rarer than in nationally representative samples), so the comorbidity structure is firmer than the absolute rates. Filed as the epidemiological backdrop the transdiagnostic premise leans on, in the same spirit the wiki keeps Folkman & Lazarus and Friston — context the interoception literature borrows without supplying.

The same shape, from the outside of the body (Schwarzlose et al. 2023)

Everything on this page is about interoceptive dysfunction across disorders. Schwarzlose et al. (2023) — not an interoception study — is filed here as the exteroceptive parallel that the transdiagnostic claim has to reckon with. Sensory over-responsivity (distress at innocuous exteroceptive stimuli) in 11,210 ABCD children is transdiagnostically associated with depression, anxiety, OCD and ADHD, survives control for autistic traits, prospectively predicts later anxiety and prodromal psychosis, and carries a salience/sensorimotor connectivity signature — the exact evidentiary shape as Nord et al.’s interoceptive mid-insula marker, in a far larger sample.

That parallel is a double-edged gift to this page. It shows the transdiagnostic-sensory-marker logic is not unique to interoception — which lends the whole enterprise external credibility. But it also poses a specificity problem the interoceptive literature does not raise on itself: if an exteroceptive sensory feature predicts the same broad psychopathology through the same salience machinery, then “interoceptive dysfunction across disorders” may be one instance of a general sensory-precision dysfunction rather than an interoception-specific mechanism. The discriminating study — interoceptive and exteroceptive over-responsivity measured in the same sample, dissociated against outcomes — has not been run. See interoception-exteroception-boundary, is-interoception-domain-general.

The applied debate this sharpens

Everything on this page feeds is-more-interoceptive-awareness-better. The transdiagnostic frame makes the debate unavoidable: if interoceptive dysfunction takes opposite forms in different conditions (hypervigilance in panic, hyporeactivity in depersonalization, avoidance in anorexia), then no single intervention direction — “raise awareness” or “lower it” — can be right across the board. Profile, not diagnosis, and certainly not a single composite score, has to guide treatment. See interoceptive-training-clinical, where that argument is worked through case by case.