Interoception Wiki

Last updated: 2026-09-24

A research wiki on interoception: the sensing of the body’s internal state, and its role in emotion, selfhood, health and wellbeing. New here? Start with the overview. Each page opens with an “at a glance” card; the graph (right) colors pages by type.

Help set the research agenda

Vote on 30 open research gaps: the questions this wiki keeps running into, ranked by readers. Sign in with your RADlab account.

Also new: Contradictions register · All studies and all methods as sortable tables · Suggest a paper or a gap

Section	Pages	What’s there
Studies	112	One page per paper: findings, limitations, links
Concepts	128	Theoretical and empirical constructs
Methods	55	Tasks, measures and paradigms
Debates	20	Contested claims, positions in the field
Researchers	61	Key figures and their positions
Applications	6	Clinical and wellbeing translation
Synthesis	1	Cross-study conclusions and reading rules
Gaps	30	Open empirical questions, ranked by reader votes on Research gaps

The full catalog follows.

Studies

• Craig (2002) — redefines interoception as the sense of the whole body’s physiological condition; the lamina I → insula homeostatic afferent anatomy.

• Craig (2009) — the anterior insula as the substrate of human awareness; the ‘global emotional moment’ model.

• Seth (2013) — emotions and the embodied self as active Bayesian inference on interoceptive signals; AIC as comparator.

• Seth & Friston (2016) — formalizes interoceptive inference with the free-energy apparatus; agranular ‘visceromotor areas’ (AIC/ACC/SGC/OFC) as a structural argument for prediction; extends to autism and depression/fatigue.

• Lindquist et al. (2012) — meta-analysis finding no brain region is both consistent and specific for any discrete emotion; proposes a domain-general core affect/conceptualization/language/executive-attention network instead.

• LeDoux (2012) — proposes replacing basic emotions with conserved ‘survival circuits’ (defense, energy, fluid, thermoregulation, reproduction) that are explicitly decoupled from feelings; feelings arise when a cognitive workspace labels the resulting global organismic state.

• Scarantino (2018) — commentary arguing survival circuits and (reformed) basic emotions are the same mechanism under different names; the dispute is over whether ‘emotion’ should denote feelings or motive states.

• Friedman (2010) — review tracing the James–Lange theory to modern psychophysiology; defends autonomic specificity of basic emotions.

• Dror (2017) — archival history showing Schachter–Singer arose from social-influence research, not the James–Lange debate.

• Farb et al. (2015) — integrative theory linking predictive-coding interoception to contemplative practice and clinical health.

• Oldroyd et al. (2019) — two small cross-sectional studies proposing interoception has social antecedents: attachment style tracks MAIA profiles, and maternal rejection of negative emotion predicts lower self-report/physiology congruence in youth.

• Nummenmaa et al. (2014) — the founding paper of the bodily-maps programme and the source of emBODY; five experiments show felt topography survives the change from word cues to vignettes, silent film and faces (rs = 0.71-0.82) — the claim the rest of the literature cites and never retests, here read first-hand.

• Volynets et al. (2020) — 3954 subjects in 101 countries colour where they feel 13 emotions; the maps are emotion-specific and concordant across cultures (rs > 0.82) and sexes, dampen with age, and are — by the authors’ own admission — subjective feeling maps with no physiological warrant.

• Lyons et al. (2021) — first independent replication of the bodily-maps result and first clinical use: depressed patients’ maps stay fully differentiated (unlike schizophrenia), medicated patients’ maps invert to deactivation, and the paper’s between-group percentages are transposed in its own table.

• Ferré et al. (2024) — 651 Spanish speakers rate 1,286 words on the five components of emotion; what makes a word count as an emotion is feeling and interoception, not expression, appraisal or action tendency — the wiki’s first usage data on what ordinary speakers mean by ‘emotion’.

• Bechara & Damasio (2005) — the somatic marker hypothesis stated first-hand, for economists: primary/secondary inducers, body vs as-if loops, a posterior→anterior VM gradient offered as the mechanism behind prospect theory — and, by the authors’ own report, patients whose somatic deficit makes them outperform healthy investors.

• Bechara et al. (2000) — the missing middle text between the two below, and the most patient-focused of the three: it dates the framework’s changes (the convergence-zone parent was gone by 2000, the as-if loop was still the default), eliminates conditioning, memory and working memory as explanations of the VM deficit, and then states in its own words that the mechanism ‘remains unspecified’.

• Damasio (1996) — the same hypothesis nine years earlier, with a discussion transcript in which Damasio, asked whether manipulating the periphery would change behaviour, says it should have relatively little impact; four of the framework’s load-bearing parts do not yet exist and two that do get silently dropped.

• Berntson et al. (2010) — a causal, lesion-based double dissociation of the two core-affect dimensions: insula lesions blunt both valence and arousal for both hedonic signs, amygdala lesions blunt only negative arousal (valence intact) — the wiki’s structural complement to its correlational arousal/valence evidence, from Berntson, Bechara, Tranel and the Iowa lesion group. Small, aetiology-confounded groups; held at the lesion discount.

• Dunn et al. (2006) — the critique the wiki had been carrying secondhand for three ingests, read first-hand: a split verdict in which the neural substrate survives and the psychological mechanism does not, and the news that Everitt’s peripheral-manipulation experiment was run after all — patients with no autonomic feedback outperform controls on the gambling task.

• Dunn et al. (2010) — the critic runs the study he said was missing: interoceptive accuracy predicts nothing on its own (r = .08 with decision quality) but reverses the sign of the body’s effect on choice, so perceiving the body is a gain term rather than a benefit — and the somatic marker result replicates (r = .41) in a task Dunn built to purge the flaws he catalogued in 2006.

• Pollatos et al. (2005) — the consensus Dunn was arguing against, read first-hand: nearly the same experiment five years earlier, reporting the arousal main effect Dunn found absent — and, unanimously with him, nothing at all for valence; the wiki’s first EEG study, and a Jamesian conclusion drawn from a design in which the body was never measured.

• Salomon et al. (2016) — the wiki’s first source where interoceptive timing gates exteroceptive awareness: visual targets flashed synchronously with the viewer’s own heartbeat are suppressed from consciousness (slower to break CFS, less accurate under crowding) and the bilateral anterior insula shows reduced BOLD to them (seen and CFS-invisible), while controls stay flat — a cardiac-frequency (not phase) effect, read as predictive suppression of the heartbeat’s self-generated sensory consequences. Nine experiments, unusually complete controls; small behavioural effects and small fMRI n.

• Pollatos et al. (2007) — the source-reconstruction sister to Pollatos 2005 (same lab, same paradigm, fresh n=32, with stCDR added): reproduces the emotion-specific arousal effect and the valence null, and localizes the high-vs-low-perceiver P300 difference to right insula (unpleasant-specific), ACC, somatosensory and prefrontal cortex but not visual cortex — the source-level anatomy the 2005 paper could only borrow, at the EEG-inverse discount, and still with the body never measured.

• Herbert & Pollatos (2012) — a Herbert & Pollatos framing review, held at one remove: its factual spine (Craig’s insular hierarchy, somatic markers, the heartbeat-perception programme) the wiki holds first-hand from better sources, so kept for two moves — the explicit interoception→embodiment bridge (interoceptive representations as the visceral case of embodied cognition / perceptual symbols) and the “translation deficit” framing of alexithymia and eating disorders as failures to translate bodily signals into felt emotion. Read-out not predictive; all claims by citation; the alexithymia↔IA datum sits on the believed, not the objective-accuracy, side of the measurement split.

• Does et al. (2000) — 709 participants pooled by the four rivals who had spent a decade disagreeing, asking whether the wiki’s foundational task measures anything: under 10% of healthy adults can do it, 95% produce a number anyway, and exercise manufactures accurate perceivers above 100 bpm who dissolve by 95 — the cardiodynamic confound turned from a correlation into a manipulation.

• Ehlers (1993) — the wiki’s first first-hand Ehlers source: her sole-authored review of the panic-interoception programme, seven years before the reanalysis she co-authored against herself. Three characteristics of panic patients (better cardiac perception, attentional bias to threat, fear of bodily sensations) and one rejected hypothesis (they are not more physiologically reactive) — and, as the pro-validity protagonist, she already concedes the heartbeat effect is paradigm-specific and may be a “habit of attending to bodily cues” rather than perceptual ability.

• Barrett (2017) — the constructive half of the critique the wiki carried for eight ingests, in her own words: emotion derived in seven steps from the claim that a brain is for budgeting a body, so interoception is at the core of the internal model and affect is what the budget feels like — and the convergence the wiki had filed as a coincidence between literatures turns out to be the architecture.

• Tong (2015) — the appraisal tradition speaking for itself at last, after being quoted only by its critics: 13 appraisals sort 13 positive emotions well above chance, cross-culturally — vindicating the taxonomy as description while leaving the causal claim untouched, because recall-and-rate cannot show appraisals make the emotions they track.

• Tong & Jia (2017) — the experience-sampling sequel that carries the appraisal taxonomy out of memory and into daily life (61/75 predictions replicate), and explains why positive emotions co-occur: they overlap on their predicted appraisals, and still do after pleasantness is removed — valence is the strongest tie but not the only one, and causation stays untested because EMA is still non-experimental.

• Payne et al. (2015) — the wiki’s first trauma paper: a theory of Somatic Experiencing, which treats trauma as a reversible dysfunctional state of a subcortical network (Core Response Network (CRN)) and heals it by completing thwarted defensive movements through interoceptive attention — coherent, testable in one place (Biological completion), authored by the therapy’s proprietors, and backed by no peer-reviewed evidence, as the authors state outright.

• Kolk (1994) — the paper that named the idea: trauma stored as somatic/implicit memory and re-lived through the biological stress response, not recalled. Splits declarative from nondeclarative memory, gives the amygdala a memory-gating role, and (following LeDoux) calls the subcortical trace indelible — the classic position SE’s reversibility claim argues against (Are traumatic memories indelible?); its interoceptive core is the traumatized “loss of the capacity to use affect states as signals.”

• Farb et al. (2011) — the wiki’s author, first-hand and with data: remitted depressives watch sad films, and 18 months later relapse is predicted by elaborative self-referential (mPFC) reactivity and prevented by sensory (visual) reactivity, while felt sadness predicts nothing — the clinical proof-of-concept of the sensory-vs-elaborative dual mode the interoception theory generalizes, with the honest catch that the protective sensory pole is exteroceptive, not interoceptive.

• Farb et al. (2022) — the RCT sequel: 85 patients scanned before and after MBCT or Well-Being CT, followed two years, turning the 2011 correlation into a two-factor model — a static somatosensory-deactivation vulnerability treatment never moved, plus a dynamic prefrontal reactivity it did (combined C = 0.86); the sensory pole shifts from vision toward somatosensory/insular cortex, and the prophylactic mechanism is the same across both therapies, not mindfulness-specific.

• Farb et al. (2010) — chronologically the earliest Farb data paper, met last: MBSR completers and controls watch sad films at equal felt sadness, and training rebalances the response — less midline self-referential/language elaboration, and the interoceptive right insula kept online where controls deactivate it, with more insula recruitment tracking lower depression. The wiki’s one Farb datum where the protective sensory pole is genuinely interoceptive (not exteroceptive), bought at the price of a cross-sectional design.

• Lutz et al. (2008) — the wiki’s first study of compassion (not mindfulness) and of empathy as a neural process: 15 expert vs 15 novice meditators generate loving-kindness while hearing others’ emotional sounds, and the compassionate state up-regulates the interoceptive insula/ACC to another’s distress in proportion to expertise — an effect that survives covarying out autonomic arousal, so it is empathy, not just arousal. The interoceptive hub turned outward, and trained.

• Khalsa et al. (2018) — the field’s map of itself: 24 authors + the Interoception Summit 2016 agreeing on what interoception is, how to measure it (a third taxonomy), how it goes wrong (transdiagnostically), and how it computes (hierarchical Bayesian). Contradicts nothing; consolidates almost everything — and, to its credit, marks its own unifying framework as “indirect so far.” Fills the wiki’s standing gaps on the panic history and interoceptive-exposure’s real (thin) evidence.

• Wiens et al. (2000) — the pre-Dunn consensus’s first member read first-hand, and better than its reputation as a citation: good heartbeat detectors (the Katkin discrimination task, not Schandry counting) feel emotion more intensely across amusement, anger and fear but no more pleasantly — the intensity-not-valence dissociation, earliest in the wiki (2000). Alone in the consensus it measured bodily arousal and showed the effect survives it, and its ~17% accurate minority matches the counting-task prevalence, so cardiac accuracy is rare on both task families.

• Nentjes et al. (2013) — the wiki’s first psychopathy and first forensic source: 75 offenders given a heartbeat-discrimination task, where reduced accuracy tracks the antisocial dimension of psychopathy (Factor 2, r=−.29) and not its affective core (r=−.01) — so the study’s own emotional-deficit hypothesis failed and the result is a somatic-marker story about behaviour instead. Held loosely: cross-sectional, the sample discriminated at chance on average (mean d’=0.00), and nothing but perception and a trait score was measured.

• Werner et al. (2009) — the study the decision debate named as its highest-value unread source and a prospective conflict with Dunn: 25 good vs 25 poor heartbeat-counters make more advantageous IGT choices (r=±.30), the main effect Dunn (2010) reports as null. Read first-hand it does not settle the dispute — extreme-groups (so Pollatos-style disputable), the marker was never measured (so it cannot test the moderation), and the ‘benefit’ is on deck choice but not net gain — making it the third accuracy↔outcome main effect from a selected sample opposite Dunn’s one unselected null, so the disagreement is about design, not direction.

• Theriault et al. (2021) — the wiki’s first source on social pressure, and the first to run the interoception→affect→Allostasis machinery outward into social cognition: conforming to others’ expectations is derived as an allostatic strategy, because prediction error is metabolically costly and other people are your most unpredictable environment. The felt obligation to conform is the Sense of should — core affect pointed at social prediction error, a “punishment the brain inflicts on itself.” Barrett-lab framework, enormous reach (conformity, guilt, status quo, communication, culture, belief), and the constructed-emotion family’s falsifiability worry.

• Zoellner & Craske (1999) — the wiki’s first first-hand replication of the founding cardiac-perception group difference (Ehlers & Breuer 1992), from an outside lab in an analogue sample: infrequent panickers count their hearts more accurately than controls. Runs the two experiments the validity debate wanted — a caffeine arousal induction (which missed the cardiac channel, so tests nothing) and a within-subject state-anxiety manipulation (which improves the score in a way the wiki’s two camps read as either attention or artefact, indistinguishably).

• MacCormack et al. (2021) — the wiki’s first source on aging, and the purpose-built study behind Age-related interoceptive decline: across two methods (a property-association task and the Day Reconstruction Method), the felt/represented body’s role in emotion weakens with age, specifically for high-arousal interoceptive sensations, while behavior and situation hold — and interoception decouples from situations but not behaviors. Read through Maturational dualism and constructionism; measures conceptual knowledge and self-report, not accuracy or physiology, so the biological mechanism is proposed, not shown.

• Quadt et al. (2018) — the wiki’s first health-neuroscience review, and the Sussex group’s own statement of the Garfinkel/Critchley programme: interoception as a four-level umbrella, the accuracy/sensibility/awareness dimensions first-hand, and the IPP/EPIC anatomy applied to six conditions — sickness behaviours, fatigue, depression, autism, anxiety, eating disorders. Consolidates the wiki’s predictive-coding thread and introduces the immune-to-brain and Alexithymia material the roadmap only gestured at. A review, not data; contradicts nothing.

• Khalsa et al. (2009) — the objective-accuracy counterpart the aging pages had queued by name for three ingests: in 59 adults aged 22–63, heartbeat-discrimination accuracy falls steeply with age (age = 30% of variance, the only demographic; BMI and sex null), reliably across two sessions. So the felt/represented aging decline (MacCormack, Volynets) now sits beside first-hand evidence the ability declines too — but Khalsa’s objective decline is confoundable with a quieter aging heart (no stroke-volume control), so it supplies the accuracy leg without instrumenting the mechanism. An early hands-on Khalsa, Iowa-era, before the roadmap-convener phase.

• Petzschner et al. (2021) — the map of the wiki’s computational quadrant, and the first source to treat predictive coding as one option rather than the theory: everything sorted onto a three-arc loop (infer the state / select the action / forecast the state), Bayesian inference separated from its algorithms, and a genuine rival put on the board — HRL, which is equivalent to active inference computationally yet could regulate the body “in the absence of an interocept.” Names the properties of internal signals (sparse, multi-timescale, oscillatory) that models borrowed from vision cannot handle, and concedes the biomarker programme has almost no empirical test behind it.

• Zeharia et al. (2019) — the “Maps” folder’s other kind of body map, and not an interoception study: fMRI over 20 moved body parts reveals a whole-body somatotopic homunculus in the Precuneus (toes-to-tongue, mirror to the SMA), confirming Penfield & Jasper’s neglected 1954 prediction. Kept for three thematic threads — it forces the distinction between cortical body maps and felt emotion maps; the precuneus is a DMN hub; and the authors link it to their earlier insula homunculus as a putative “body-mind node,” reaching the insula/mindfulness thread from the somatomotor side. Interoceptive relevance is analogical, not measured.

• Bonaz et al. (2021) — the clinical-medicine quadrant of the TICS 2021 special issue, and the wiki’s first source treating gastroenterology, urology, rheumatology and neurology as interoceptive territory: IBS, Chronic pain, pelvic pain, Joint hypermobility, MS, Parkinson’s, Alzheimer’s. Organizes it all under Stephan’s allostatic self-efficacy circuit, where dysfunction can strike at any level of the neuraxis and propagate both ways — and adds what the wiki lacked: the efferent hierarchy, the Microbiota-gut-brain axis, early adversity as cross-cutting determinant, Bioelectronic medicine as intervention, and lesion dissociations of interoceptive dimensions. Broad, shallow, causally unestablished nearly everywhere, and (despite Critchley co-authoring both) drops the heartbeat-task caution Quadt et al. supply.

• Quigley et al. (2021) — the functional quadrant of the TICS 2021 issue, asking what interoception is for: energy regulation in and out, memory, affect, and the self. Adds two domains the wiki had nothing on — memory (cut a rat’s upper-GI vagal afferents and hippocampal spatial/episodic memory fails) and excretion (phasic sphincter bursting as a sensory amplifier, an efferent act whose job is to raise afferent precision). Widens Multisensory integration so interoception is a party to perception rather than a channel feeding it. And it is candid where the wiki’s other sources are not: the strongest evidence for interoception’s role in affect is anatomical overlap, the rest is covariation, Schachter-Singer “difficult to replicate,” and studies routinely confound believed with actual bodily change. Proposes Affective touch (C-tactile afferents) as a way out — an interoceptive-anatomy channel that sits in the skin and can actually be manipulated.

• Weng et al. (2021) — the intervention quadrant of the TICS 2021 issue, and the first source here to put contemplative training, a breathing pacer and an implanted nerve stimulator on one diagram as three entry points into one pathway. Uses respiration as its case system — while conceding it is the one interoceptive channel open to immediate voluntary control, so the least representative. New: taVNS and RAVANS, where stimulation gated to exhalation exploits phasic NTS gain — an intervention whose active principle is timing; device-guided slow breathing, the wiki’s first body intervention that needs no awareness at all; MABT with trial results at last (substance use, craving, emotion dysregulation down; RSA and self-reported interoceptive awareness up at 12 months); and brain decoding of interoceptive attention. Carries the wiki’s one result where mindfulness beats the slow breathing it produces. Almost nothing here shows the interventions work through interoception, and every section is written by the author of the work it reviews.

• Berntson & Khalsa (2021) — the anatomical quadrant of the TICS 2021 issue, cited by the other three as the wiring they lean on, and the paper that makes the wiki’s anatomy start one stage earlier: interoception as a continuum from sensors → pathways → networks → circuits → awareness, with awareness the last and rarest stage. New: the receptor inventory (PIEZO baroreceptors, astrocytic central chemosensing, gut TAS2Rs with possibly bacterial ligands); Neuropod cells, gut epithelial sensors synapsing onto vagal neurons that carry a luminal stimulus to the NTS in ~60 ms and decide whether a mouse can tell sucrose from sucralose; and retrograde rabies tracing of the efferent limb, which finds three prefrontal networks reaching the monkey adrenal medulla and the cognitive-control ones absent in rat. Names the field’s structural problem — cortex known in humans, brainstem in rodents, the middle inferred — reports mouse insula anticipating satiety, and states flatly that there are no animal models of interoceptive awareness.

• Allen (2026) — the wiki’s first instrument paper, and its first source where the interoceptive quantity is not a perception but a control performance: an open-source toolbox in which participants breathe to track a sinusoidal target while a visuomotor gain perturbation distorts the visual feedback, yielding baseline control ability plus a normalized perturbation ratio for sensorimotor remapping cost. Answers the standing complaint on Respiratory interoception — a continuous, objective respiratory measure requiring no report, reliability .857 — and names the gap the four taxonomies left: the field measures perception and theorizes control. Held loosely on evidence (N = 1, the developer, one sitting) and with one unresolved problem the paper does not solve: the feedback is visual, so good tracking may be good motor control reading a screen.

• Banellis et al. (2026) — the paper three earlier ingests had queued by name as the most consequential unread source in the backlog, and it delivers: in 241 participants measured on cardiac (HRDT) and respiratory (RRST) axes with one Bayesian pipeline, interoceptive sensitivity, precision and metacognitive efficiency are uncorrelated across organs (moderate Bayesian support for the null, powered to r ≥ 0.179). So “interoceptive accuracy” has no demonstrated person-level referent, and this wiki’s habit of measuring in the heart and concluding about interoception loses its licence — see Is interoception domain-general or organ-specific?, created here. The positive result matters as much: mean confidence correlates strongly across cardiac, respiratory and auditory tasks, so the one thing that generalizes is not interoceptive — which supplies a mechanism for the accuracy/sensibility dissociation recorded on four pages. Brakes: moderate not decisive evidence, healthy young adults at rest, two channels, and the arousal escape untested.

• Chen et al. (2021) — the NIH Blueprint framing paper of the TICS 2021 issue and reference 1 in every other article in it, read last because its survey chapters are superseded by its own companions. Two things are uniquely here. It proposes a seventh definition of Interoception that folds descending body regulation inside — senses, interprets, integrates, regulates — naming regulatory signals, central regulators and interoceptive effectors, while conceding the expansion departs from current usage and blurs interoception into body regulation. And it offers the wiki’s most usable boundary criterion: whether signals represent, rather than originate from, the internal world — which pulls Proprioception and the vestibular system inside and leaves taste mixed. See Exteroception Boundary. Also corrects a stale claim here (VENs are present in macaques, per Evrard 2012), splits the Salience network into a homologous ventral half and a human-elaborated dorsal half, names the non-verbal measurement gap that thins both the developmental and the dementia ends of this wiki, and — most usefully — concedes that for most diseases the links to interoceptive dysregulation are “largely symptomatic and descriptive.” Read as the field’s charter document, and note that all thirteen authors are NIH program staff: it is partly a description and partly an announcement of what will be funded.

• Banellis et al. (2026) — the wiki’s first study of spontaneous thought, and the first to ask how often the drifting mind turns to the body: 536 people probed with an interoceptively-extended MDES battery after a 14-min resting scan. Body-wandering is a stable interindividual dimension (split-half ICCs .88-.94), less prevalent than cognitive content (median 36.3 vs 47), and it comes with negative affect and high autonomic arousal — while correlating with lower ADHD and depression symptoms, where depression instead tracks mental time travel. Its connectivity signature is thalamo-somatomotor, not default-mode (CCA out-of-sample r = .354; right thalamus the hub), aligned with the principal unimodal-transmodal gradient — so embodied mind-wandering engages sensorimotor circuitry rather than decoupling from it. Lands on Is more interoceptive awareness better? as the first row whose verdict splits by outcome class inside one sample, and echoes Farb’s elaborative/sensory dissociation with the sensory pole finally interoceptive. Brakes: rs ~.10-.21, cross-sectional, sensibility not accuracy, and the body sampled is one lying motionless in a scanner.

• Levakov et al. (2023) — the wiki’s first gastric source, and it arrives as a methods audit: 43 subjects with concurrent EGG and resting fMRI, plus an independent single-subject dataset. Two findings held apart. (1) The Gastric network is real but was four-fifths artefact — regressing head motion and non-grey-matter signal shrinks it from 19.89% of grey matter to 1.16%, and the confounds turn out to be themselves phase-locked to the stomach (probably via respiration, which the study did not record). (2) Gastric-brain synchrony has no test-retest reliability (r ≈ −.14 to .08 across runs) while the EGG signal alone retests at r = .74 and the fMRI connectivity is subject-identifiable — both ingredients reliable, their coupling not. Reliability bounds validity, so the interoceptive-biomarker use of brain-body coupling is under a ceiling nobody had measured; see Can brain–body coupling be measured in a person?, created here. Brakes: n = 23 for the reliability null, 24% of runs excluded for signal quality, BMI 18-25 only, and the group-level phenomenon is causally supported (rat vagotomy, human taVNS) and not in dispute.

• Haruki & Ogawa (2023) — the neural counterpart to Banellis et al., and the first study to put two interoceptive channels through one paradigm in one session: 31 people attending, in 10-second blocks, to their heartbeat, their stomach, or a fading word (Interoceptive attention task). Against the visual control the two organs give the same map; against each other they come apart. The right dorsal anterior insula — this wiki’s canonical seat of interoceptive awareness — was the only region preferring cardiac, and the accuracy-to-right-AIC correlation the field rests on fails for breathing and skin conductance, so that region may be about arousal rather than about the body in general (see Insular cortex, Craig (2009)). Gastric attention instead recruited visual cortex overlapping the Gastric network, plus orbitofrontal cortex, hippocampus and motor cortex — a feeding profile reached by a one-word instruction, with no stomach recorded. And MVPA found the left dorsal middle insula classifying which organ was attended (56.3%, d = 0.61) while showing no mean preference for either. Brakes: an attention paradigm with no ground truth and no physiological recording of any kind, a gastric-network overlap asserted from a figure rather than tested, and an imagery account the paper never considers.

• Harrison et al. (2021) — the wiki’s first empirical, model-based test of interoceptive predictive coding in humans, from Stephan’s Zurich TNU with Petzschner on the author list. Sixty adults split by trait anxiety, measured at three levels of one channel at once: questionnaires, the FDT (threshold, bias, metacognition), and a novel breathing learning task at 7T whose Rescorla-Wagner fit supplies per-trial prediction certainty and prediction error as fMRI regressors. What confirms: both quantities have neural correlates with the signs the theory forces — anterior insula, ACC, MFG, dlPFC deactivate with certainty; aIns, ACC, MFG and the PAG activate with error. What does not: the anterior/posterior prediction-vs-error dissociation everyone predicts did not appear, and anxiety modulated the prediction term (a valence x group interaction in bilateral aIns) with nothing in prediction error — a first direct hit on the EPIC-style clinical story ( computational psychosomatics). Also closes two standing gaps: the ASI measured against a non-cardiac task (Anxiety sensitivity), and anxiety tracking worse respiratory sensitivity where the cardiac panic literature has it tracking better — a sign reversal, which is stronger evidence for organ-specificity than any null (Is interoception domain-general or organ-specific?). Its PCA ranks the whole hierarchy by distance from the body, and beliefs win. Brakes: no model won selection (RW used by pre-registered fallback), certainty is a transformation not a parameter, extreme groups, healthy adults, reduced field of view.

• Jones et al. (2010) — the wiki’s evidence about the insula is almost all correlational; this is the dedicated lesion review, written by Critchley’s group to test neuroimaging models of the structure against what happens when it is destroyed, and the models do not survive intact. The anterior/affective vs posterior/interoceptive split holds — but bodily unawareness localizes posteriorly (Anosognosia and disorders of bodily unawareness), not to the anterior seat of awareness; pain awareness survives without the insula entirely (Starr’s two patients rated noxious stimuli as more intense with no insular activation); Disgust is the one emotion-specific deficit and it is left-sided, the wrong side for Craig’s asymmetry; and 42 insular glioma resections left no deficits at three months. Its methodological gift is the MCA confound made concrete: apraxia of speech, localized to left anterior insula in Nature, was reattributed to Broca’s area because the insula “is most commonly damaged because it is most vulnerable to disruption of the middle cerebral artery” (Lesion-symptom mapping). Verdict, from a co-author of the models: they “fall short.”

• Wallman-Jones et al. (2023) — the wiki’s first source on exercise, sedentary behaviour and screen time, and its first attempt to measure interoception repeatedly in ordinary life: thigh-worn accelerometers triggering smartphone prompts, ten a day for a week, in 70 adults. The title finding is the screen-time contrast (every activity category beats screen time except sleeping, including two other sedentary categories — so it is not sitting, it is attention captured by a screen). The finding that matters more is structural: ICC = 0.51, so barely half the variance in self-reported interoception is between people, and the same measure reverses sign between timescales — momentary activity raises it, habitual activity lowers it (β = −.29) in one model on one sample. Baseline heartbeat accuracy both predicted higher daily sensibility (the accuracy↔sensibility association the wiki keeps finding absent) and reversed the activity effect in the top tertile. Brakes: the dependent variable is a mindfulness subscale with one frankly exteroceptive item, no control questionnaire, the sedentary result is cut-off-dependent, intensity was never measured so the paper’s own attention-vs-salience question is undecided, and no outcome was measured at all — ‘more’ is everywhere, ‘better’ nowhere. See State vs trait interoception, Physical activity, sedentary behaviour and interoception.

• Ventura-Bort et al. (2021) — framed as a test of constructed emotion’s ingredients; read here for the psychometric result underneath. Eight interoception and emotion questionnaires in 109 adults do not split into a body factor and an emotion factor — they split into believed accuracy (“Sensibility”) and deployed attention (“Monitoring”), orthogonally, with bodily and emotional instruments interleaved in both. The two halves then point opposite ways: believed accuracy goes with well-being (r = .58) and finer granularity for negative emotions; deployed attention goes with worse granularity, lower felt intensity, and r = .12. So the sensibility entry every taxonomy shares is two constructs, the four definitions do not pick out the same one, and a MAIA composite sums a protective disposition and a costly one. Also: Alexithymia’s TAS-20 administered first-hand at last, loading with the bodily scales (Brewer/Bird at the level of beliefs, and shared method variance explains it equally well). Brakes: no accuracy task anywhere, a two-factor solution forced where four had eigenvalues > 1, 157 recruited → ~88 in the regressions, and the paper’s own laboratory task (Emotion differentiation (ED) task) found nothing at all.

• Wilzok et al. (2023) — an instrument paper whose most consequential result is one it does not foreground. Two new paradigms measure what a person anticipates an upcoming interoceptive stimulus will feel like, what they then experience, and the gap; run in 52 people on graded pinprick pain and graded inspiratory loading with identical cues, scales and timing. Both validate and retest adequately (ICC 0.66-0.92; discrepancy 0.54-0.78 on n=10). Then: the anticipation-experience gap correlates r = 0.57 across the two channels — the wiki’s only positive cross-modal interoceptive correlation, against nulls from Ferentzi, Crucianelli and Banellis. Held at less than face value on three counts (it is a self-report difference score not a psychophysical measure; shared scale use explains it and there is no exteroceptive control; nociception’s status as an interoceptive channel is unsettled) — but both channels were aversively perturbed, which is the condition Is interoception domain-general or organ-specific?’s strongest escape had never been tested under. Also undersold by the authors: people overpredict the strongest stimulus even when it matches the cue, in both channels, after successful familiarization — the pain literature’s overprediction bias, shown to hold for breathing. Brakes: channels not intensity-matched, three cue levels only, retest n=10, MAIA-2 collected and never reported, and one results sentence that calls a p < 0.001 a null.

• Eusebio et al. (2022) — the wiki’s first randomized trial manipulating interoceptive emphasis itself: two isomorphic 10-week yoga curricula, identical in dose, instructors and format, differing only in whether instruction directs attention to bodily sensation. Both self-report hypotheses failed — MAIA rose and symptoms fell equally in both arms — so a curriculum built to cultivate interoceptive awareness did not out-perform one that deliberately avoided mentioning the body. What succeeded was behavioural transfer: SART sustained-attention accuracy improved more under interoceptive instruction (p = .006, corrected) on a task with no bodily content. And the moderation runs in opposite directions across the arms — attentional gain scaled with MAIA under interoceptive instruction (uniquely; six other questionnaires ps > .15), while under movement instruction only the lowest baseline MAIA scorers improved. An aptitude-treatment interaction where dose language has been failing. Brakes: N = 34 analysed against a power analysis calling for 36, total MAIA scores, no accuracy measure, a declared commercial conflict, and a concluding sentence that misstates the study’s own results. See Yoga as interoceptive training, Norman A. S. Farb.

• Farb et al. (2013) — the paper the wiki held secondhand and thought it couldn’t reach, read first-hand now the Farb/ exclusion is lifted: 27 adults alternately attend to the breath or to visual words in the scanner, with respiration recorded. Attention decides which cortex represents the same respiratory signal — posterior insula (primary interoceptive cortex) when the breath is attended (r=0.63), somatosensory cortex when it is not. The insula is a graded posterior→anterior tuning field (interoceptive to exteroceptive), so the anterior insula is “not an area of pure body awareness” — the same deflation Haruki & Ogawa reached a decade later. Interoceptive attention enhances insula↔VMpo thalamus connectivity (the afferent relay), and IA/EA recruit dissociable networks (posterior-limbic/medial-parietal vs. lateral frontoparietal executive). Corrects a conflation on Insular cortex and Respiratory interoception: these are attention effects, not the training effects of the separate SCAN 2013 companion (still pending). Brakes: respiration only (the one voluntarily-drivable channel), a contemplatively-enriched sample, no attentional ground truth, and the anterior-insula EA preference confounded with executive task demand.

• García-Cordero et al. (2017) — the electrophysiological complement to Farb 2013, and the wiki’s first first-hand HEP study. 50 adults do a motor-tracking heartbeat task in three conditions (exteroception / basal interoception / post-feedback interoception); both interoceptive conditions produce more negative frontal HEPs than exteroception at 200-500 ms, dissociating internal from external attention irrespective of accuracy. Feedback produced real interoceptive learning (M=0.47→0.58), carried by connectivity (basal = local short-range; post-feedback = long-range fronto-posterior) not HEP amplitude. And in 2 intracranial patients, posterior insula, amygdala, SSC and IFG rode broadband 35-110 Hz for interoception vs 1-35 Hz for exteroception — an attentional-mode dissociation recorded from inside the human insula. Brakes: a motor component (controlled, not confounding), and Study 2’s n = 2. See Exteroception Boundary, Insular cortex.

• Farb et al. (2007) — the founding paper of the wiki’s sensory-vs-elaborative dual mode, held secondhand by a dozen pages until the Farb/ exclusion lifted: reading trait adjectives in a narrative focus (“what does this say about me”) recruits the mPFC/PCC midline; in an experiential focus it shifts, in mindfulness-trained participants only, to a right viscerosomatic network (insula, SII, IPL). The mechanistic core is a decoupling: the right insula, coupled to the vmPFC by default (R=0.61), is released from the narrative self-model by training (R=0.06) — the neural form of Decentering and the root of Narrative and experiential self-reference. Every later Farb paper is this dissociation with something added (emotion, relapse, an RCT). Brakes: cross-sectional (waitlist vs post-MBSR, so training-per-se is unproven, as the authors state), the “experiential” stimulus is still verbal trait words, small n, and the right-AIC-as-viscerosomatic-seat reading its own 2013 work later narrows.

• Farb et al. (2013) — the training companion to the [[farb-2013-attentional-modulation|Cerebral Cortex attention paper]], and the source of the “mindfulness training raised posterior insula” claim the wiki carried secondhand — read first-hand, it corrects that claim on two points. MBSR graduates vs untrained controls attend to the breath: the group-level training effect is in the dorsal anterior insula (integration with external context), the posterior insula moves with practice dose (r=0.61 with % daily practice, not group membership), and the distinctive finding is DMPFC deactivation during interoceptive attention with negative DMPFC-posterior-insula coupling — an “emergent prefrontal pathway” by which training reorganizes interoceptive attention. The VMpo-thalamus connectivity the secondhand note attached to training is actually the attention paper’s. Cross-sectional and no accuracy task, so it is the cleanest neural exhibit for the “training reorganizes attention/architecture, not accuracy” side of contemplative training increase objective interoceptive accuracy? — and the clearest illustration of why that debate cannot be settled by studies that never measure accuracy.

• Rai et al. (2026) — the wiki’s first exercise-intervention trial and its first digital therapeutic: a Farb-group activated RCT (70 undergraduates, smartphone mindfulness app vs a 2048 cognitive-training control) for high-intensity cycling. The manipulation check succeeded (MAIA rose, β=1.83) and endurance rose slightly (β=0.21), but wellbeing did not move and no mediator survived — the raised interoceptive awareness was uncorrelated with the performance gain (r=−0.02) and distress tolerance was the only factor tracking it. The third of this group’s active-comparator trials where the interoceptive mechanism does not carry the result (2022, Eusebio), and the one that measured the mediator and reported its absence. Brakes: no passive control, unequal randomization, 3-week minimal dose, sensibility not accuracy, declared commercial conflict.

• Weng et al. (2020) — the primary source for a method the wiki carried secondhand: the EMBODY framework (EMBODY framework (brain decoding of internal attention)), which trains participant-specific fMRI classifiers on five instructed internal-attention states (breath, mind-wandering, self-referential, feet, sounds) and then decodes an uninstructed 10-min meditation second-by-second into breath/MW/self. Step 1 validates (all five states >twice the 20% chance level, reliable in 14/16, accuracy tracks subjective ratings); Steps 2-3 are feasibility — the meditation decode has no ground truth, but its metrics behave as sustained breath-focus should (breath held in longer events, 10.9 vs 8.1 s, not more frequent ones). The wiki’s first measure of interoceptive attention that neither asks the participant nor counts a heartbeat — and its self-referential state is Farb’s narrative pole read by decoder instead of contrast. Brakes: N = 16, no resting baseline, no subjective check of the decode, classifiers idiosyncratic and mostly cortical, MW class a likely conglomerate.

• Wu et al. (2026) — the fourth Farb relapse paper, and the one that pushes the two-factor model past its prefrontal centre: the same trial cohort as Farb (2022) scanned during a self-referential encoding task instead of a sad film. Behavioural negative self-referential bias predicts concurrent symptoms but not relapse; the frontal DMN/salience ROIs are associated with relapse yet do not independently predict it and their connectivity is inert (a negative test of the dorsal-nexus account); and in the combined survival model the somatosensory markers (right postcentral static + left supramarginal dynamic) are the sole survivors while the subgenual-prefrontal marker drops out. Self-criticism → prefrontal elaboration, self-affirmation → somatosensory embodiment — relapse vulnerability as “insufficient recruitment of somatosensory systems supporting embodied self-experience.” The protective pole is somatosensory, not interoceptive-insular, so the interoceptive reading stays a generalization.

• Farb et al. (2012) — the wiki’s one clinical natural-lesion test of the frontolimbic-integration framework, from Farb’s Baycrest/Rotman dementia-lab period. Resting-state fMRI in frontotemporal dementia (8 bvFTD, 8 SD vs 16 controls) finds “prefrontal isolation”: frontolimbic disconnection (the salience core — right anterior insula, ACC, thalamus, striatum — losing coherence) plus compensatory-looking prefrontal hyperconnectivity. The two halves dissociate onto FTD’s two symptoms — PFC hyperconnectivity tracks apathy (r=.68, surviving atrophy control), limbic disconnection tracks lower disinhibition (the disinhibited symptoms need an intact limbic channel) — and bvFTD adds unchecked DMN elevation (right angular gyrus) tracking stereotypy. The anterior insula’s integration of visceral/emotional information into behaviour, watched in its failure mode. Small n, no autopsy, no DTI, cross-sectional.

• Porges (2023) — the wiki’s first first-hand Porges source, upgrading Polyvagal theory from secondhand scaffolding to primary statement: the “vagal paradox” (RSA-protective vs bradycardia-lethal) resolved by two cardiac-vagal source nuclei, the five principles (autonomic state as intervening variable, three-circuit hierarchy, Jacksonian dissolution, ventral vagal complex / social engagement system, Neuroception), and the measurement apparatus (RSA, vagal brake, vagal efficiency). Its interoceptive hook is Porges’s own claim that neuroception’s felt limb “is functionally equivalent to interoception.” A theory-clarification, not data; concedes human dorsal-vagal bradycardia “is still speculative”; declared Safe-and-Sound-Protocol royalty. Read against its refutation, Grossman.

• Grossman (2023) — the critical source the Polyvagal theory page had long lacked: each of PVT’s five premises argued untenable on ablation/optogenetic evidence (the nucleus ambiguus does essentially all cardiac vagal control; the dorsal DMNX almost none, and never the “massive” bradycardia PVT requires — Veerakumar et al. 2022, Nature) and comparative biology (ventral vagus, myelinated cardiac vagus, RSA-like coupling and sociality all present in non-mammals). The general lesson for this wiki: equating RSA with cardiac (let alone general) vagal tone is a Rylean category mistake. A review, not a decisive test, and entangled in a bitter attribution dispute — but the strongest available statement of the physiological case against PVT. See Is Polyvagal Theory scientifically valid?.

• Kerdabadi et al. (2024) — the clinical-axis counterpart to the Porges/Grossman debate core: a PRISMA meta-analysis (9 studies, 26 effect sizes) of “polyvagal interventions” finding a small pooled benefit (Hedges’s g ≈ 0.26, p < .001) that the paper overstates as “medium”/“large.” Read as dispensability evidence from the pro-PVT side — the “polyvagal” bundle is mostly generic vagal-tone techniques (massage, yoga, slow breathing, HRV biofeedback, positive psychotherapy), only the Safe and Sound Protocol actually Porges-designed, so interventions helping does not touch the theory’s disputed premises. Unaddressed I² ≈ 88% heterogeneity, several dissertations/pilots/single-arm designs pooled as “high quality,” and a bilingual (Persian/English) Iranian-journal source. See Is Polyvagal Theory scientifically valid?, Polyvagal theory.

• Wesarg et al. (2022) — the first of the three queued RSA meta-analyses to land, and the one that asks whether one of the field’s most robust developmental risk factors actually moves vagal tone: two three-level meta-analyses (~27,000 participants) find no overall association between childhood adversity and either baseline vagal activity (r = −.015) or vagal reactivity (r = −.017). Small blunting (r ≈ −.05 to −.11) survives only in clinical samples and for direct/maltreatment adversity, reading as stress-sensitization rather than a population-wide signature. A pre-registered null within the mainstream vmHRV paradigm (Thayer/Neurovisceral-Integration lineage, not a PVT test), it lands on the debate’s evidential-robustness axis and qualifies the “blunted physiology” story on Early adverse life events for the vagal channel — an efferent autonomic readout, not interoceptive accuracy. Its own sharpest caveat: pooling reactivity as group comparisons cannot separate vagal withdrawal from augmentation, so a heterogeneous real effect would average to the null.

• Graziano & Derefinko (2013) — the second queued RSA meta-analysis, and the most pointed for Polyvagal Theory: across 44 studies / ~4996 children, RSA withdrawal predicts fewer externalizing, internalizing and cognitive/academic problems (each r ≈ −.16, small/adaptive) but is unrelated to social functioning (r = −.06) — the “most surprising” result, “counter to what one would expect based on Polyvagal theory and the Social Engagement System,” and one that flips maladaptive in clinical/at-risk samples (M = +.16 vs −.14 in community). A null within the vagal-reactivity paradigm (not a brainstem-anatomy test), it removes an empirical prop from the social-engagement claim and lands on the debate’s evidential-robustness axis beside Wesarg. Sharpest caveats: only linear relations modelled (a quadratic optimum is live), and RSA-W conflates vagal withdrawal with augmentation.

• Zahn et al. (2016) — the third and last queued RSA/HRV meta-analysis, and the most cognitive: across 26 studies / N=2317, resting HRV predicts laboratory self-control at only r = 0.15 — and that small effect drops to a nonsignificant r = 0.04 once publication bias is trim-and-filled, with higher-quality and confounder-adjusted studies showing smaller effects. Speaks more to Thayer’s Neurovisceral Integration Model (HRV as prefrontal inhibitory capacity) than to PVT proper, since it restricted HRV to Task-Force metrics and excluded Porges–Bohrer RSA, conceding results “may only partially be generalizable to Polyvagal Theory.” Completes the trio with Wesarg and Graziano on the debate’s evidential-robustness axis — three different outcomes (adversity, social engagement, self-control), three thin/vanishing associations. Brakes: k=26 underpowers moderators, respiration uncontrolled corpus-wide, resting HRV + lab tasks only.

• Brooks et al. (2005) — the wiki’s first-hand evidence that the dorsal posterior insula (primary interoceptive cortex, the VMpo terminus) is somatotopically body-mapped, with Craig co-authoring a direct fMRI test of his own labelled-line prediction. Painful heat to face/hand/foot in 14 adults produces an orderly dpIns map (face anterior/lateral, foot medial in the circular sulcus) — the only one of five operculo-insular regions to carry a stimulus-site-dependent map, with no SII pain map at all, whose coordinates match human depth-electrode stimulation/recording. Turns the “presumably somatotopic” hedge on the lamina I terminus, and the secondhand Zeharia 2012 “insula homunculus”, into a measured somatotopy of a homeostatic afferent. Sits beside Jones et al. as mapping without necessity (pain finely mapped in dpIns yet not required for the awareness that something hurts). Brakes: n=14 with relaxed thresholds for weak foot activators, mostly fixed stimulus order, a mm-scale effect that only single-subject analysis resolves, left-hemisphere-only map from right-side-only stimulation, non-fibre-selective thermode.

• Folkman & Lazarus (1980) — the wiki’s first first-hand source from the Lazarus causal-appraisal tradition, the camp Barrett attacks and the appraisal debate had held only through her citations. 100 middle-aged adults’ stressful episodes across a year show both problem- and emotion-focused coping used in 98% of encounters, coping more variable than consistent (mean .265), appraisal and context (not who was involved) the potent determinants, and — against stereotype — no gender difference in emotion-focused coping. Not an interoception paper (thematic/structural only); Lazarus first-hand but on coping not emotion, correlational, and in an explicitly reciprocal appraisal-coping loop that complicates Barrett’s serial-ordering charge.

• Schulz et al. (2013) — one acute cold-pressor stressor, two cardiac tasks, 42 people, opposite results: heartbeat-counting accuracy rises after stress while visual heartbeat-discrimination accuracy falls — so “does stress improve cardiac interoception?” has no instrument-free answer. The counting gain tracks a +8.2 bpm / +20.6 mmHg sympathetic response, so it reads as the cardiodynamic confound (a louder heart, not a better perceiver) which the authors concede having not measured stroke volume; the two paradigms did not intercorrelate (r=.22, .08) while the two discrimination modalities did (r=.63). A within-session state manipulation of accuracy for State vs trait interoception, and a manipulation-based data point for Is the heartbeat-counting task a valid measure of cardiac perception?.

• Desmedt et al. (2023) — landmark conceptual review (Desmedt, Luminet, Maurage & Corneille) naming the field’s two foundational discrepancies. (1) Definition: interoception is defined both phenomenally and physiologically (homeostatic pathways) in the same article, but non-homeostatic Aβ somatosensory fibres also process internal states across cardiac/respiratory/gastric systems, so no pathway is necessary-and-sufficient and physiology cannot define it. (2) Measurement: the consensus dimensions are detached from their measures, which converge weakly — HCT vs HDT share 4.4% of variance (Hickman 2020), sensibility scales split into five factors, so cross-measure “replication failure” is statistically expected (Carlson & Herdman). Proposes a phenomenon-based definition (“below the skin,” top-down + bottom-up, exteroceptive senses excluded) and a hierarchical taxonomy by construct specificity. The canonical source for Exteroception Boundary and the two measurement debates.

• Friston (2005) — the primary source for machinery the wiki ran secondhand for fourteen ingests: Friston’s free-energy statement of predictive coding, where inference and learning are one optimization (the E- and M-steps of free-energy minimization), hierarchical/reciprocal/asymmetric cortical anatomy is derived from what empirical-Bayesian inference requires, and an evoked response is prediction error being explained away — yielding one account of repetition suppression, the MMN and P300, with a DCM re-analysis of oddball ERPs confirming the backward+lateral plasticity the theory predicts. Not an interoception paper (vision and audition throughout); its inward life is entirely Seth’s and Barrett’s later work, which is exactly what makes it the machinery read undisplaced. The still-open bet: distinct error/representation unit populations, one algorithm among several (Petzschner) and not yet found in the human interoceptive brain (Harrison). See Karl J. Friston.

• Greenwood & Garfinkel (2025) — Garfinkel’s dimensional programme stated first-hand at last (Annual Review of Psychology), after the wiki reconstructed it from Quadt et al. and Farb’s pushback: the eight-dimension Suksasilp & Garfinkel scheme (signal → representation → preconscious impact → accuracy → beliefs → insight → attention → attribution) worked through emotion, a fifth taxonomy and the only one carrying an attribution dimension. Adds a bottom-up mechanism the wiki lacked (“heartbeat sentinel neurons” — mechanosensitive transduction of vascular pulsation, Jammal Salameh 2024), causal two-factor evidence that afferent manipulation needs context (optogenetic tachycardia → anxiety only when risky, Hsueh 2023), and two clinical conditions new to Interoceptive psychopathology — schizophrenia and EUPD. A consolidation, contradicts nothing; states the person-level/cardiac assumptions Banellis undercut, and (to its credit) concedes them in its own Limitations.

• Engelen et al. (2023) — the wiki’s fullest statement of the rhythm-based view of interoception (Tallon-Baudry’s lab), the counterweight to the heartbeat-detection tradition: the body’s cardiac, respiratory and gastric rhythms are a timing structure woven into how the brain processes the outside world, not a private interoceptive channel. Almost nothing here needs the participant to feel anything. Introduces Cardiac cycle effects to the wiki; treats the HER as the object where emotion, prediction-error and self-readings collide (depersonalization patients lose HER attentional modulation); places the Gastric network and respiratory entrainment under one frame; and lays out three non-exclusive integration mechanisms — oscillatory synchrony (the scaffolding hypothesis), predictive coding, and Multisensory integration — closing that a brain in a body may compute differently from one in silico. A review, not data; directionality mostly unresolved, Fig. 3 collapses species/rhythms/methods, and the three mechanisms are largely untested against each other.

• Desmedt et al. (2022) — the criterion-validity half of the Desmedt/Corneille HCT critique, and the empirical companion to their 2023 review: a pre-registered meta-analysis of 133 studies / 11,524 adults asking whether heartbeat-counting performance predicts the mental-health outcomes the “accuracy” label was adopted to explain. It does not — trait anxiety (r=.03), depression (r=−.04) and alexithymia (r=−.01) are all null, with no publication bias — while the only significant associations are with the cardiac signal and the body producing it: heart rate (r=−.17), BMI (r=−.11), sex (male>female, r=−.14), and age after bias correction. Shared variance ≤2.9% everywhere. A criterion argument independent of every construct argument on Is the heartbeat-counting task a valid measure of cardiac perception?: a valid-but-noisy measure should show attenuated true associations, not flat ones, at this N — so this strengthens the sceptics on prediction the way Van der Does did on prevalence, and lands the objective-task null on Alexithymia that complements the belief-level one. Doubles as a damning practice audit (0.8% pre-registered, 6.8% open data, 0% full-reporting disclosure). Brakes: the ≥10-study inclusion rule biases toward expected-real effects (so the nulls are not underpowered fishing), but heterogeneity is high, covariates unmodelled, and the guessing-reduced modified instructions too rare to test.

• Carter & Ogden (2023) — the wiki’s first purely qualitative interview study: twelve adults narrate how bodily states are felt, attended, made into symptoms and reified, yielding a transcending “series of thresholds” — a state becomes a symptom only by surpassing a movable criterion that expectation, affect, body-focus and meaning-making lower or raise. Its value is register, not mechanism: Schema-guided symptom perception and Interoceptive inference said in the first person, by people who never heard of a prior — a non-predictive-coding tradition arriving at the prior-over-physiology picture, like Van der Does and Ehlers. Also the folk end of emBODY (gut-centred emotion, reported unprompted) and of Alexithymia (insight into emotion↔body varying as a continuum; their “Interoceptive Crossover” construct). Brakes: N=12 self-selected, remote interviews, authors’ health-psychology priors, and thresholds named but never measured — no accuracy task, no physiology. See Jane Ogden.

• Iodice et al. (2019) — the wiki’s first experimentally induced interoceptive illusion, and its first behavioural-causal evidence for predictive interoception: 18 cyclists hear false (faster) heartbeat feedback and the same physical work feels harder (false-feedback pointed at a homeostatic variable, not an emotional one), while their actual heart rate does not move (no entrainment, BF=0.014). The illusion is asymmetric — faster feedback inflates perceived effort, slower does not deflate it — which a symmetric read-out cannot explain but a risk-averse prior (underestimating effort is dangerous) does. The a priori prediction that low MAIA sensibility would strengthen the illusion (from Tsakiris’s RHI work) failed, so its moderators stay open. Senior author Giovanni Pezzulo (Active inference). Brakes: N=18, all-male, one channel, and the effect rides on a laboratory-induced belief that the sound is one’s heart.

• Barr et al. (2022) — not an interoception study, kept as epidemiological backdrop: an All of Us biobank Brief Report (N=329,038, EHR/ICD-10-CM) documenting the comorbidity structure the transdiagnostic framing presupposes but never itself shows — ~51% of diagnoses overlap, cross-disorder tetrachoric correlations r=0.43–0.75, a dominant SUD cluster read as “shared common causes,” and disparities (highest risk in low-SES, women, sexual-minority; schizophrenia the lone racial-risk reversal) that “recapitulate previous epidemiological research.” Warrants the transdiagnostic premise (b: the disorders genuinely cut across each other) while the interoception literature only ever argues premise (a); says nothing about interoception, and a shared-cause reading is compatible with genetics/adversity/ascertainment as easily as with an interoceptive mechanism. Held like Folkman and Friston — context the field borrows. Brakes: treatment-selected biobank with a documented healthy-participant bias (every rate below population), so the comorbidity structure is firmer than the absolute prevalences.

• Murphy (2022) — a two-page Biological Psychology commentary on Suksasilp & Garfinkel’s model, and the wiki’s first Murphy source: names the construct every taxonomy omits — propensity to use interoceptive signals, whether a person deploys a signal they can perceive versus leaning on external cues, split by process and by adaptiveness. Ability is not use; use is a conditional gain term (generalizing Dunn 2010 from accuracy to deployment) whose valence depends on whether the signal is right and insight is good. Its wedge is a sex difference the perceptual taxonomies cannot absorb — women’s poorer lab accuracy with better emotion recognition. The second non-perceptual axis (beside regulation) the five taxonomies leave out, and the disposition the accuracy/sensibility dissociations keep circling. Held for what it is: a proposal with no data, no measure, and an open state status.

• Murphy et al. (2020) — six studies from the Bird/Catmur lab (with Khalsa) developing the IAS and, more consequentially, running the first test of the accuracy/attention distinction against an objective task. Their model is a 2×2 — what is measured (accuracy vs attention) crossed with how (performance vs self-report) — predicting convergence wherever the what matches. It holds on both axes: the two accuracy-belief scales correlate at r = .57–.71 while the attention scale (BPQ) correlates with neither, and heartbeat-counting accuracy tracks the IAS (r = .271, .336) and nothing else in the battery — not the BPQ, not the ICQ, not the TAS-20, not time estimation. That puts a boundary on the wiki’s most-repeated null: the sensibility↔accuracy dissociation recorded across Interoceptive sensitivity and Interoceptive taxonomy (Farb et al. Box 1) was measured with attention instruments every time. Alexithymia likewise lands on the accuracy-belief side (TAS-20 ↔ IAS r = −.43/−.57, ↔ BPQ ≈ 0), surviving self-esteem, depression and anxiety controls, while the objective-task null holds (Desmedt reproduced). A defender’s datum for Is the heartbeat-counting task a valid measure of cardiac perception? — a guessing account must explain the selectivity — held at its real weight: r ≈ .3 at n = 33–52, p ≈ .047, with the paper’s own control-variable findings failing to replicate inside it, and a cardiodynamic account (a loud heart both raises the count and teaches its owner they can feel it) predicting the whole pattern with no perception in it. Created study Murphy et al. (2020) and method Interoceptive Accuracy Scale (IAS); updated Interoceptive taxonomy (Farb et al. Box 1), Interoceptive sensitivity, Alexithymia, State vs trait interoception, Is the heartbeat-counting task a valid measure of cardiac perception?, Heartbeat detection task, Body Perception Questionnaire (BPQ), Jennifer Murphy, Ventura-Bort et al. (2021), Murphy (2022).

• Lazar et al. (2000) — not an interoception study, kept as the wiki’s earliest meditation-fMRI precursor (2000, seven years before Farb): five expert Kundalini practitioners meditate (breath focus + mantra) vs a control task, and the group map reads as attention (frontoparietal) + autonomic control (pregenual ACC, midbrain, hypothalamus, amygdala) — Benson’s Relaxation response given a first brain map. Two things it hands the later literature before that literature had its vocabulary: the insula is absent (susceptibility artifact at 2000-era resolution, not a tested null — so the wiki’s insula-centric account is a fact about later methods), and meditation produced large global signal decreases tracking cardiorespiratory change, so every BOLD meditation study inherits a whole-brain breathing confound ( device-guided breathing (DGB), Respiratory interoception). Activation grew with practice (late>early, scan 2>scan 1), matching meditators’ reports of a slowly evolving state. Held as a baseline, not a result: n=5, all expert, no interoceptive measure, a confounded (animal-naming) control. See body-focused contemplative training, contemplative training increase objective interoceptive accuracy?.

• Moffat & Cross (2024) — not an interoception study, kept as a boundary marker: an fNIRS neuroaesthetics experiment where “embodiment” is the motor-simulation sense (having performed a movement, via the action-observation network), not the interoceptive “material me” this wiki is built on. 43 adults mirror movements then rate watching them: enjoyment rises with synchrony and is greatest when a sequence was both mirrored and recognised as mirrored (awareness of embodiment), engaging sensorimotor cortex (STG, IFG, IPL) — but recognition was near-floor (embodiment/recognition confounded), there was no embodied empathy, and cognitive empathy dampened enjoyment (reversing the authors’ prior result). Its one genuine contact with the wiki is the somatosensory register it shares with Wu/Farb; its use here is conceptual hygiene — the motor “embodiment” the interoceptive literature must not be conflated with. Brakes: small n, fNIRS blind to subcortical reward/AON, stick-figure stimuli, non-preregistered empathy analyses.

• Christoff et al. (2018) — not an interoception paper, kept structurally as the primary statement of a debate the wiki’s Body-wandering construct sits inside: a two-round Trends in Cognitive Sciences Letter exchange on how to define Mind-wandering. Christoff et al.’s dynamic framework says it has an essential feature — thought that arises freely and unconstrained — so rumination and deliberate planning are not mind-wandering; Seli et al.’s family-resemblances reply says it is a cluster concept with no privileged feature and admits them. Both accept graded membership; the whole fight is whether some features are more defining than others. The wiki had no Mind-wandering page until this ingest despite holding Body-wandering and EMBODY, both of which operationalize mind-wandering without defining it. The debate is upstream of the wiki’s body-wandering paradox: free drift toward the body vs constrained fixation on a symptom are classified oppositely, and a single MDES probe cannot tell them apart. No data; the free/constrained cut is the wiki’s experiential and sensory distinction wearing new clothes.

• Parr et al. (2019) — not an interoception paper, kept structurally like Friston 2005: an Active inference simulation (Markov decision process) of two visual phenomena — Troxler fading (a fixated peripheral percept dissolving) as the accumulation of uncertainty about an unsampled stimulus under a volatility prior, and binocular rivalry (and the Necker cube) as the same mechanism driven by covert attentional action rather than an eye movement. Its two gifts to this wiki: it simulates the identity the interoceptive-attention studies show empirically — attention is covert action, which is precision-setting (Farb 2013, Interoceptive precision) — and it makes precision the seat of individual differences (“distinct computational phenotypes”), the exteroceptive twin of the aberrant-precision autism/schizophrenia story behind computational psychosomatics and Harrison 2021. Adds first-hand pages for Thomas Parr and Jakob Hohwy. Brakes: pure simulation (the fade-time∝1/volatility, pupillometry and EEG predictions are advanced, not tested), vision not viscera, discrete-time (so no continuous dominance-time predictions).

• Prescott & Liberles (2022) — the primary review of the transduction stage the wiki carried secondhand: 21 pages from the Liberles lab on vagal sensory neurons across the respiratory, cardiovascular and digestive systems, upgrading Interoceptive sensors and Neuropod cells from Berntson & Khalsa’s one-sentence inventory to receptor, Cre-line and knockout detail. The mechanosensor gets a name and a phenotype (PIEZO2 is the airway stretch sensor behind the Hering-Breuer reflex, lethal on knockout; the baroreceptor needs both PIEZO1 and PIEZO2); many vagal afferents are second-order, reading out from epithelial “sentinel” cells (glomus, taste, tuft, enteroendocrine, NEB, immune) rather than transducing directly; and single-cell RNA-seq finds dozens of vagal sensory types, most functionally uncharacterized — the “dark matter of the vagus nerve.” The gut section is the substrate GLP1-agonist obesity drugs act through. Not an interoception-theory paper — engages no construct debate, proposes no taxonomy, measures nothing — but its documented afferent target-organ heterogeneity is background consistent with Grossman’s efferent case against treating “the vagus” as one unit (Is Polyvagal Theory scientifically valid?), and its closing thesis is a useful counterweight: vagal interoception lags the external senses by decades of mechanistic understanding, and the field theorizes inference over sensors it has mostly not yet identified. Adds first-hand pages for Sara L. Prescott and Stephen D. Liberles. Brakes: a review not data, overwhelmingly mouse molecular genetics, Cre lines that label multiple types, and vagal afferents only (motor control out of scope).

• Nord et al. (2021) — the empirical neuroimaging paper the wiki had been carrying secondhand (“reduced dorsal-mid-insula reactivity in depression,” via Quadt), read first-hand and generalized across disorders: a preregistered ALE meta-analysis of 33 experiments (626 patients, 610 controls) asking whether disrupted interoceptive activation converges anywhere across depression, bipolar, anxiety, remitted anorexia and schizophrenia, and across cardiac/respiratory/pain/hunger/touch tasks. It converges on one cluster — the left dorsal mid-insula (MNI −36, −2, 14) — the same subdivision Haruki’s MVPA singled out for organ-identity, reached from the opposite direction. Conjunction analyses place it lower in the hierarchy than affect (which peaks in left anterior insula, no overlap) and off the map of what antidepressants and psychotherapy move — so it is the transdiagnostic frame’s most direct neural support and a candidate for its empty biomarker cupboard, cast as a “neural marker of psychopathology and a putative target for novel interventions.” Proposes the dysgranular precentral insular gyrus, with hybrid anterior+posterior connectivity, as the interoceptive prediction-error comparator — a different placement from Harrison’s anterior-insula finding, both under-determined. The single convergent locus reads like domain-generality but is site-level not ability-level, so it is compatible with Banellis’s decorrelation (Is interoception domain-general or organ-specific?). Brakes: coordinate-based (no effect sizes), a surprising left laterality possibly a verbal-probe artefact, pools channels it concedes may not be integrated, no behavioural interoception measured in any included study, healthy-only and non-psychiatric-somatic studies excluded, cause/consequence unresolved.

• Paulus & Khalsa (2021) — the invited AJP editorial accompanying Nord et al. in the same issue: no data, held for its framing from the field’s clinical-computational centre (Paulus & Khalsa). Two moves beyond Nord’s own reading — it pushes the mid-insula one synapse downstream, from prediction-error comparator to input for mid-cingulate action selection, making the conjunction nulls the theory’s prediction rather than a surprise; and it restates the LIBR two-pathway model of interoceptive psychopathology (hyperprecise priors vs context rigidity), a cleaner mechanistic parse than the roadmap’s five-dysfunction checklist. Treats the mid-insula-as-error-encoder placement as settled, which the Harrison tension says it is not. Concedes the three gaps between finding and clinic: cause/consequence, modifiability, functional heterogeneity.

• Schwarzlose et al. (2023) — not an interoception study, filed as the exteroceptive mirror of the wiki’s interoceptive transdiagnostic-marker work: sensory over-responsivity (distress at innocuous sounds/textures/smells) in 11,210 ABCD children affects 18% but makes up 57% of children with clinical-range symptoms, associates differentially across disorders (positive for anxiety/depression/OCD/ADHD, negative for conduct disorder) surviving autistic-trait control, prospectively predicts anxiety and prodromal psychosis, and carries a preregistered, twice-replicated resting-FC signature — reduced sensorimotor-network coupling, enhanced sensorimotor↔salience and cingulo-opercular↔Amygdala coupling, altered sensory↔hippocampus coupling read through a sensory-prediction lens. Adjudicates three accounts (dysregulation / subthreshold-autism / sensory-specific) in favour of the sensory-specific. Its value here is comparative and cautionary: the same salience machinery mis-weighting exteroceptive input in one syndrome and interoceptive input in another (Anxiety sensitivity, panic) makes “interoceptive dysfunction across disorders” (Nord) possibly one instance of a domain-general sensory-precision failure — see Exteroception Boundary, Is interoception domain-general or organ-specific?. Held loosely: SOR is a single parent-report item, FC correlational.

• Treves et al. (2025) — the wiki’s first first-hand fMRI study of breath attention: a breath-counting task + dynamic connectivity in 72 high-rumination adolescents. Four recurring brain states (hypoconnected, DMN-anticorrelated, SN-reduced, hyperconnected); the DMN-anticorrelated and hyperconnected states rose during breath counting vs rest (replicating Mooneyham 2017). But the task-vs-rest states were not the on-task-vs-off-task states, and none tracked individual attentiveness — a controlled caution against a single neural biomarker of focused attention, arguing for within-individual imaging.

• Cauda et al. (2011) — the resting-state derivation of the posterior insula dichotomy the rest of the wiki assumes from anatomy: seed-based rsFC plus three clustering methods in 16 adults split the resting insula into a ventral-anterior limbic network (→ rACC, middle/inferior frontal, temporoparietal) and an anticorrelated dorsal-posterior sensorimotor network (→ premotor/SMA, mid-posterior cingulate), with a transitional mid-insula — the gradient recovered from spontaneous BOLD alone, no task, no cytoarchitecture. Partially lateralized: right anterior ROIs preferentially wired to brainstem, pons and right thalamus, early support for the right-insula-as-arousal/attention-hub reading (Craig; Haruki). A founding rsFC parcellation of the insula and the wiki’s example of rsFC used to subdivide a structure. Brakes: correlational, 1.5T, 5mm ROIs exceed cortical thickness, age-heterogeneous sample, anticorrelation dependent on global-signal regression (contested).

• Chang et al. (2013) — the tripartite successor to Cauda, and the wiki’s canonical demonstration of consistency vs specificity: data-driven k-means parcellation of the right insula (validity index → k=3) splits the anterior insula into a dorsoanterior (executive; ACC/DLPFC) and a ventroanterior (affective-chemosensory; amygdala/OFC) node beside the posterior sensorimotor pole, replicated from resting connectivity and meta-analytic coactivation over ~4400 studies. Reverse-inference decoding via Neurosynth gives each subdivision a specific function (ventroanterior=emotion/chemosensation/autonomic, dorsoanterior=inhibition/error/conflict/switching, posterior=pain/sensorimotor/language) — while forward inference shows the dorsoanterior insula is the most consistently activated of the three across nearly all tasks yet is specifically executive, so its ubiquity is high consistency, not domain-generality. Brakes: right insula only, n=17, 2.4-min rest after a decision task, and Neurosynth’s coarse-term/confirmation-bias limits.

• Deen et al. (2011) — the primary source the wiki had cited only secondhand (and mislinked) through Chang: the original functional tripartition of the insula. Voxelwise k-means clustering of resting-state connectivity splits each insula (left and right) into a ventral anterior (→ pregenual ACC, limbic), a dorsal anterior-to-middle (→ dorsal ACC, Dosenbach cognitive-control network), and a posterior (→ somatomotor cortex) subregion — the dorsal-cognitive/ventral-affective anterior seam Chang later grounds by reverse inference and the wiki’s finer studies (Haruki, EPIC) keep re-finding. Adds a cingulo-insular gradient (pregenual ACC↔vAI, dorsal ACC↔dAI, middle cingulate↔PI: three parallel loops, not one insula+ACC unit) and, uniquely, validates the resting-defined clusters on an independent disgust task — both anterior clusters respond to disgusting images, posterior does not, fitting the selective≠specific reframe. Created study Deen et al. (2011); updated Insular cortex, Cauda et al. (2011), Chang et al. (2013), Resting-state functional connectivity, Salience network, Anterior cingulate cortex (ACC). Brakes: correlational, k fixed a priori (no validity index), disgust task shows distinctness not computation.

• Vercelli et al. (2015) — the fourth and highest-resolution insula-parcellation paper, from Cauda’s own Turin group (a self-replication). Its contribution is a method: a “one-step” node-detection procedure where fuzzy c-mean clustering parcellates the insula and detects borders in a single pass (border voxels are nonunivocally attributed between adjacent clusters), dropping 3 mm ROIs at cluster maxima as network nodes. Driven to 12 nodes per insula, it reproduces Cauda’s anterior-salience/posterior-sensorimotor bipartition when nodes are grouped, and — once shared variance is regressed out — finds the finer nodes carrying the “echoes” (Leech 2012) of default-mode, sensorimotor and dorsal-attentional networks nested inside (a hierarchical subparcellation). Validates Touroutoglou 2012’s dorsal (dFP)/ventral (vFP) anterior-insula split, the same dorsal-executive/ventral-affective seam Deen and Chang reach from other methods. Reliable across an 18-subject replication group (17/24 ROIs overlapping). Not an interoception study — no bodily signal, insula as pure connectivity object — so it counts as confirmed anatomy plus a domain-generality data point. Brakes: correlational, 1.5T, a group-template (“fixed-effect”) parcellation, k=12 at the edge of reliability per Kelly 2012.

• Gong et al. (2015) — not an interoception study, filed as the wiki’s clearest evidence that the insula’s anterior/posterior functional split is experience-dependent: taking Cauda’s ten ROIs and A-/P-network scheme as a template, 27 action-video-game champions vs 30 amateurs show higher anterior↔posterior insular functional integration (left-lateralized, r=0.68 with weekly playing time), increased grey matter near the anterior/posterior transitional seam (correlating with integration and playing time), and a rewiring in which the posterior sensorimotor insula couples to a frontal attentional node (MFG) — while the amateurs reproduce Cauda’s bipartite split. Its interoceptive relevance is oblique but real: the insula is treated purely as an attention/sensorimotor integrator with no bodily measure taken, a data point for the domain-generality/“not an area of pure body awareness” side, and a plasticity companion to Farb 2007. Brakes: cross-sectional (cannot show training caused the differences), correlational FC, group confounds.

• DuBois et al. (2016) — the field’s first dedicated review of interoception in autism (2016), and mostly a census: a systematic search returned five original studies, all n<50, male, normal-IQ/verbal, excluding the developmentally-disabled majority of the spectrum. Its payload is a contested direction — the “low accuracy + elevated sensibility” ITPE signature the wiki carries (Garfinkel 2016) is one of two, since two of three sensibility studies find sensibility reduced (Fiene & Brownlow 2015, d≈−1; Elwin 2012) and the review reads the literature as tending toward hyporeactivity. Accuracy splits by sample (intact in children, impaired in adults; Alexithymia confound underexamined); the neural arm is a lesion-free insula hyper/hypoconnectivity story (interoceptive-state hyperconnectivity ∝ ADOS severity); the literature is entirely cardiac where hunger/thirst/bladder cues likely matter more (Is interoception domain-general or organ-specific?). Framed as “autism as a disorder of prediction,” the interoceptive twin of Sensory over-responsivity (SOR). Consolidated on Interoception in autism (ASD).

• Wiebking et al. (2010) — the wiki’s first study of the depressed “material me” (Northoff’s programme): 17 medicated MDD vs 17 controls counting heartbeats/tones/resting, plus Porges’s BPQ. Two dissociations survive the interoceptive-imaging discounts — (1) the group difference is at rest, not in the stimulus response (the exteroceptive difference dissolved under baseline correction, leaving an abnormally under-deactivating insula at rest, a caution that “reduced insula reactivity in depression” may be a rest-baseline artefact), and (2) patients score higher on BPQ perceived reactivity but not awareness (sensibility up, awareness flat) while the healthy body-perception↔right-anterior-insula rest correlation is absent in MDD — body report decoupled from its interoceptive cortex, with reduced left-AI rest deactivation instead tracking severity (BDI r=0.57). Brakes: all medicated, “rest” only 4–10 s fixation, “material me” a frame not a measured self, n=17/arm. See Embodied selfhood, Interoceptive psychopathology.

• Ribera d’Alcalà et al. (2014) — the wiki’s first primary data on Chronic pain: 22 patients vs 37 controls, no difference in heartbeat counting or body awareness, and no link between the two. Heavily confounded (22-year age gap, no severity analysis), so a tension with the review-level accuracy claim, not a contradiction.

• Meessen et al. (2016) — all three Garfinkel facets (MAIA sensibility, heartbeat accuracy, confidence-accuracy awareness) measured in 24 students and mutually uncorrelated, plus a matched memory set: body and memory metacognition unrelated. An independent, underpowered replication; its awareness index is a calibration error, not efficiency.

• Ferentzi et al. (2018) — six channels in 118 students (heart, stomach, pain, taste, proprioception, balance): measures correlate within a channel and never between, no general factor, and the cardiac–gastric link comes out slightly negative. The primary source for the organ-specific side of Is interoception domain-general or organ-specific?, now held first-hand.

• Legrand et al. (2022) — the methods paper for the HRDT (N = 223, two sessions): people judge their heart 7-8.5 BPM slower than it is, and that bias predicts undercounting on the counting task; cardiac judgments less precise and less metacognitively efficient than tone judgments; threshold reliable (r = .51), slope not (r = .10).

• Mehling et al. (2012) — the development paper for the MAIA: 32 items, 8 scales, built and validated only in experienced mind-body practitioners and only against other questionnaires. One factor fits poorly, a second-order factor nearly as well as eight; in practitioners Noticing goes with lower anxiety, a finding the authors warned may not generalize.

• Paulus & Stein (2010) — the framework behind Interoceptive psychopathology: anxiety and depression as “altered interoceptive states” from noisy afferents, alliesthetic amplification and self-referential beliefs acting on anticipated body states. A review with no new data; predicts reduced accuracy with heightened reactivity.

Concepts

• Interoception — the sense of the internal physiological condition of the body; common substrate of the cluster.
• Sensory over-responsivity (SOR) — exteroceptive construct held for its parallel to interoceptive transdiagnostic markers: distress at innocuous external stimuli (sounds/textures/smells), a cross-diagnostic childhood risk feature with a salience/sensorimotor connectivity signature, read as over-precision on exteroceptive prediction error — the mirror of the interoceptive precision failures on Interoceptive psychopathology.
• Interoception in autism (ASD) — what the wiki knows about interoception in ASD, consolidated: between-group differences repeatedly found but heterogeneous, accuracy intact in children/impaired in adults, sensibility running both directions (so the ITPE’s elevated-sensibility premise is one of two), insula hyper/hypoconnectivity, and the Alexithymia confound — framed as “autism as a disorder of prediction,” the interoceptive twin of Sensory over-responsivity (SOR).
• Symptom threshold — the point at which a bodily state stops being a state and becomes a symptom; the claim (Carter & Ogden) that its height is a movable individual difference — lowered by expectation, negative affect and body-focus, raised by normalising — and that it comes in a staged series, so a state can be noticed without being symptomatised. The folk/phenomenological register of Schema-guided symptom perception and Interoceptive inference; names a criterion where those supply the mechanism.
• State vs trait interoception — roughly half the variance in self-reported interoception is within people, moving with what they did in the last fifteen minutes; every taxonomy on the wiki names traits, and single-session measurement inherits an error the size of the signal.
• Physical activity, sedentary behaviour and interoception — movement as a naturally occurring Interoceptive exposure and screens as its opposite; two mechanisms (signal salience vs directed attention) that nobody has yet separated, and an athlete finding cited in both directions.
• Distress tolerance — the capacity to withstand aversive bodily states and persist, as opposed to how intensely they are felt or how accurately perceived; a regulation-side construct, and the one factor a digital mindfulness intervention moved (and the one that tracked exercise performance) while interoceptive awareness moved in parallel and did nothing.
• Anosognosia and disorders of bodily unawareness — acquired unawareness of one’s own body: denial of paralysis, denial of lost sensation, and disowning a limb (somatoparaphrenia); localizes to posterior insula, which is the wrong end of the gradient for an anterior seat of awareness.
• Disgust — the one discrete emotion with a selective insular lesion deficit (left anteroventral insula, recognition and experience); the wiki’s test case for functional selectivity without specificity, and a problem for Craig’s forebrain emotional asymmetry.
• Periaqueductal gray (PAG) — the midbrain structure the wiki had pointed at from four directions (defensive output, bottom of the visceromotor hierarchy, respiratory threat anticipation, placebo analgesia) without a page; now with a first-hand human result — it tracks interoceptive prediction error and not prediction certainty, which is the vertical division of labour the hierarchical account requires.
• Gastric network — cortical and cerebellar regions phase-locked to the stomach’s ~0.05 Hz pacemaker rhythm; the first resting-state network individuated by an organ rather than by its own connectivity. Unimodal sensory/motor rather than transmodal, causally tied to vagal afferents — and, as an individual measure, currently unmeasurable.
• Cardiac cycle effects — the dependence of perception, cognition, action and emotion on where in the heartbeat an event falls (systole vs diastole). The pulsed-inhibition hypothesis (baroreceptor firing at systole → transient cortical inhibition) is robust for pain, mixed elsewhere. A rhythm-based interoceptive phenomenon needing no perceptual report, and one of the most artifact-prone measurements in the field. The wiki’s entry point to interoception-as-carrier-rhythm.
• Body-wandering — the component of spontaneous thought that is about the body (breath, heart, stomach, bladder, skin, movement); a stable dispositional dimension that trades off against socially-oriented thought, runs negative in tone and high in arousal, and — paradoxically — goes with fewer trait symptoms. The wiki’s entry point to mind-wandering, met in its embodied form first.
• Mind-wandering — self-generated thought unhooked from the current task; a construct with no agreed definition (dynamic-framework “unconstrained dynamics” vs family-resemblances “cluster concept”, see How should mind-wandering be defined?). The wiki meets it at an odd angle — mostly as Body-wandering, the drift toward the body — and holds the canonical DMN/executive-control literature only by citation. Connective stub, created with the Christoff et al. (2018) ingest.
• Exteroception Boundary — what makes a signal interoceptive rather than exteroceptive. The anatomical answer (the skin) fails on vestibular, taste, proprioception and affective touch; three functional criteria were proposed in one journal issue in 2021 and nobody has adjudicated between them. Determines what the measures measure and how far the clinical claims reach.
• Proprioception — the sense of body configuration and movement. Interoceptive under Chen et al.’s functional criterion, partly so under Craig’s anatomical one, exteroceptive-adjacent by tradition — and load-bearing for Somatic Experiencing (SE), which names it alongside interoception as its therapeutic lever without saying why the pairing is natural.
• Interoceptive sensors — the transduction stage the wiki had been skipping: chemo-, gluco-, osmo-, thermo-, humoral and PIEZO mechanoreceptors, free nerve endings, and the astrocytes and Schwann cells that transduce alongside them. Much of it characterized after 2015, which is a reason not to treat the afferent anatomy as settled.
• Neuropod cells — enteroendocrine cells that are wired sensory transducers, synapsing onto vagal nodose neurons; a gut-to-brain line as fast and transmitter-specific as taste, and one that sets an innate sugar preference below both tongue and reward circuitry.
• Multisensory integration — sense data combined into a product not decomposable into its parts; widened here to include interoceptive×interoceptive and interoceptive×exteroceptive pairs, which makes interoception a party to perception rather than a separate channel.
• Affective touch (C-tactile afferents) — C-tactile afferents in hairy skin: interoceptive fibre morphology, insular projection, pleasant by report — and, being in the skin, the field’s leading candidate for an interoceptive channel that can actually be manipulated in healthy humans.
• Gut-hippocampal memory — interoceptive energy signals priming hippocampal encoding of where and what you ate; the wiki’s first interoceptive function that is not affective, clinical, or about the self, and its best causal evidence (selective vagal afferent ablation).
• Urinary interoception (lower urinary tract) — excretion as an interoceptive function: urgency penetrated by environmental cues (the front-door effect), and an efferent act (phasic EUS bursting) whose function is to defeat afferent accommodation and keep the sensory channel alive.
• Respiratory interoception — the breath as an interoceptive channel, and the argument that it is an exceptional one: uniquely open to immediate voluntary control, and a rhythm that phasically gates afferent traffic one synapse in. Most of the field’s intervention evidence is respiratory evidence, which is a problem for generalizing it.
• Interoceptive control — the efferent half: driving an interoceptive channel to a prescribed state, and adapting when the mapping changes. The field’s models are closed-loop and its measurements are afferent-only; regulation sat in the taxonomies for a decade with no task attached, because most channels cannot be voluntarily driven at all. Measurable so far on exactly one channel, which may make it a fact about breathing rather than a general construct.
• Interoceptive propensity — whether a person actually uses the interoceptive signals they can perceive, versus relying on external cues — an individual difference (Murphy 2022) absent from every taxonomy, which partition perception but not deployment. Use is downstream of ability but not reducible to it, and it is a conditional gain term with no fixed valence, not an intrinsic good. The second non-perceptual axis beside regulation; theorised heavily, measured not at all, trait/state open.
• Central autonomic network — the efferent hierarchy (ACC, insula, amygdala, hypothalamus, brainstem, spinal centres) that acts on the body interoception senses; homeostatic reflexes “enslaved” by descending allostatic policy.
• Relaxation response — Benson’s whole-body hypometabolic state (lowered O₂ consumption, heart rate, respiration, sympathetic tone) elicited by simple meditative techniques and framed as the physiological opposite of fight-or-flight; a behavioural-medicine construct agnostic about the felt body, so it sits beside Homeostasis (and feelings as homeostatic sensations)/Allostasis rather than among the perceptual constructs. The target the wiki’s contemplative interventions were later argued to work through, given a first brain map by Lazar et al. (2000).
• Functional disorders — conditions impairing bodily/cognitive function with no structural pathology (functional seizures, FMD, IBS, fibromyalgia); the disorder class the inferential account of symptoms was invented to explain.
• Disorders of brain-gut interaction — the Rome successor to “functional GI disorders”; IBS as a regulatory-loop failure, and the accuracy-low/sensibility-high profile reached independently of the cardiac tradition.
• Microbiota-gut-brain axis — an interoceptive channel whose signals do not originate in the organism: bacterial metabolites and lipopolysaccharide acting on enteroendocrine cells and vagal afferent TLR4. Real pathways, weak causal evidence.
• Chronic pain — pain past 3 months or past healing; Craig’s reclassification of pain as interoceptive, cashed out clinically — accuracy predicts severity, sensibility and metacognition inconclusive.
• Joint hypermobility — Ehlers-Danlos-spectrum connective tissue with heightened interoceptive sensitivity, insula reactivity, anxiety and PoTS; the wiki’s only heritable peripheral tissue variant with an interoceptive phenotype, and its clearest bottom-up case.
• Early adverse life events — childhood adversity as cross-cutting determinant of symptom expression; blunted physiology with amplified report, and the wiki’s third independent route to caregiver responsiveness as the origin of interoceptive competence.
• Craving — addiction’s interoceptive feeling state, varying by drug state; carries the insular-lesion finding (damage the insula, lose the craving) that is the wiki’s most direct evidence a felt bodily state constitutes a drive.
• Lamina I spinothalamocortical pathway — the homeostatic afferent pathway (small-diameter afferents → lamina I → VMpo → insula).
• Insular cortex — the cortical home of interoception; posterior→mid→anterior gradient from primary mapping to subjective feeling.
• Anterior cingulate cortex (ACC) — ‘limbic motor cortex’ (motivation/agency); the ACC half of Craig’s emotion dyad.
• Homeostasis (and feelings as homeostatic sensations) — feelings as homeostatic sensations; interoception as the afferent side of homeostasis (Craig).
• Von Economo neurons (VENs) — spindle neurons in AIC/ACC linked to self-awareness; degenerate in frontotemporal dementia.
• Global emotional moment (Craig’s model of awareness) — Craig’s model of awareness as a timebase-indexed sequence of unified self-representations.
• Salience network — AIC + ACC + amygdala + hypothalamus; detects/integrates behaviourally salient signals.
• Interoceptive inference — emotions as top-down predictive models of interoceptive causes (Seth).
• Predictive coding — hierarchical generative models; predictions down, prediction errors up; precision-weighting.
• Active inference — suppressing prediction error by acting (here: autonomic reflexes as fulfilled predictions).
• Perceptual inference (interoceptive) — suppressing prediction error by updating priors to match sensation; the contemplative counterpart to active inference.
• Simulation map — layered ‘as-if’ body representation; the construct closest to interoceptive awareness (Farb et al.).
• Presence and agency — felt connection to the moment and felt control over outcomes, as markers of successful prediction-error minimization.
• Interoceptive taxonomy (Farb et al. Box 1) — seven-construct breakdown (awareness, coherence, attention, sensitivity, accuracy, sensibility, regulation) replacing loose ‘interoceptive sensitivity’.
• Allostasis — physiological stability through adaptive change, not a static set-point.
• Alliesthesia — Cabanac’s term for pleasantness depending on internal state; used by Paulus & Stein (2010) as the valence-amplification mechanism in anxiety and depression.
• Subtle body — contemplative-tradition analog of the simulation map (chi/prāṇa channels).
• Autonomic specificity of emotion — hypothesis that basic emotions have distinct ANS patterns (James–Lange corollary).
• Experience of body ownership (EBO) — felt ownership of the body; probed by the (cardiac) rubber hand illusion.
• Interoceptive illusion — an experimentally induced, transient misperception of an internal state (effort, thermal pain, body ownership) produced by a resolvable multisensory conflict — the interoceptive counterpart of the rubber-hand illusion, and both a phenomenon and a method (hold the body fixed, move the percept, so any change in report is inference not signal). Collects the thermal grill, false-feedback effects and Iodice’s illusion of effort under one logic; its asymmetry is the wiki’s cleanest behavioural signature of a cost-shaped prior.
• Interoceptive sensitivity — individual sensitivity to interoceptive signals; measured by heartbeat detection. Now shown not to travel between organs, so it is cardiac sensitivity, not a person’s.
• Interoceptive precision — the wiki’s most theoretically load-bearing quantity and its least measured: the Bayesian weight on an interoceptive prediction error, and separately the slope of a psychometric function. The two senses are related and not the same, and the first cross-modal data decorrelate the second.
• Metacognitive efficiency and metacognitive bias — confidence splits into bias (how sure you feel) and efficiency (how well confidence tracks correctness, performance controlled). Bias generalizes across the body and out to audition; efficiency generalizes nowhere — which reframes the taxonomies’ “awareness”/“coherence”/“insight” construct as a bundle of two different things.
• Cognitive appraisal (of emotion) — emotion as interpreted/labeled arousal (two-factor theory) and its generalizations.
• Transactional (cognitive-phenomenological) model of stress — Lazarus’s cognitive-phenomenological account: stress as a person-environment relationship mediated by appraisal (primary: harm/threat/challenge; secondary: coping options) and coping, with the two in a continuous reciprocal loop rather than a fixed sequence. The other appraisal tradition — of stress, not emotion — and the pre-predictive-coding ancestor of Allostasis.
• Coping (problem-focused vs emotion-focused) — the two functions of coping: managing the situation vs regulating the emotion; both used in ~98% of stressful episodes, so coping reduces to neither problem-solving nor emotional defence. A distant, loose ancestor of the wiki’s sensory-vs-elaborative and approach-vs-avoidance splits.
• Basic emotions and natural kinds — discrete universal emotions vs the natural-kinds critique (Barrett).
• Embodied selfhood — selfhood grounded in bodily/physiological representation (‘material me’).
• Somatic marker hypothesis — Damasio’s Neo-Jamesian account of bodily feedback guiding decisions.
• Visceromotor areas (VMAs) — agranular AIC/ACC/SGC/OFC positioned atop the interoceptive hierarchy as a source of descending predictions (Seth & Friston).
• Core affect — Barrett’s psychologically primitive, non-emotion-specific bodily-valence construct; realized by a domain-general network.
• Interoceptive trait prediction error (ITPE) — accuracy/sensibility mismatch proposed as a trait-level interoceptive prediction error; elevated in autism, correlated with anxiety.
• Social origins of interoception — the proposal that interoception is built developmentally out of caregiver responses to the infant’s bodily states.
• Attachment style — secure/anxious/avoidant patterns of relating, proposed to produce distinct interoceptive profiles (hypervigilance vs. distrust).
• HPA axis — the stress axis as proposed mechanism linking early social experience to interoceptive signal strength and processing.
• Survival circuits — LeDoux’s conserved function-specific circuits (defense, energy, fluid, thermoregulation, reproduction), proposed as the replacement unit of analysis for basic emotions.
• Global organismic state — the whole-organism state produced by survival-circuit activation (circuit activity + motivation + generalized arousal); raw material for feelings, not the feeling itself.
• Amygdala — threat/appetitive nuclei read three incompatible ways: fear centre, domain-general salience detector, or nucleus-specific survival circuit components.
• Affect programs — Ekman’s latent-variable mechanism for a basic emotion; the input-open/output-closed specification that made bodily signatures a prediction, and Scarantino’s input-output open revision.
• Homeostatic property cluster kinds — Boyd’s essence-free natural kinds; the move by which New BET makes emotion variability a prediction rather than a refutation.
• Higher-order theories of consciousness — a state becomes conscious when represented by a higher-order state; the two-stage account shared by LeDoux and Scarantino and denied by Craig.
• Bodily sensation maps (bodily fingerprints of emotion) — the felt topography of activation/deactivation per emotion (‘bodily fingerprints’); reproducible, universal, and of contested referent (felt body or shared concept?).
• Cortical somatotopy (the homunculus) — the orderly cortical mapping of the body surface (the Penfield ‘homunculus’); the exteroceptive body map, kept distinct here from felt emotion maps and from the visceral interoceptive map.
• Precuneus — the medial parietal wall: a DMN/self-processing hub that is also sensorimotor and carries a whole-body somatotopic map — a place where the model of the self and the map of the body share cortex.
• Age-related interoceptive decline — felt bodily emotion dampens from 18 to 90 without changing shape; proposed to explain why older adults regulate emotion better.
• Antidepressant emotional blunting — the numbing patients report on antidepressants; a topographical correlate exists, and it is the case where less felt body is plainly unwanted.
• Emotion prototypicality — how strongly a word refers to an emotion; graded, and constituted mostly by feeling and interoception — so the folk concept of emotion is a bodily concept.
• Primary and secondary inducers — what triggers a somatic state: present stimuli (obligatory, amygdala) vs thoughts and memories (deliberate, VM cortex); the split that predicts the Iowa double dissociation.
• Body loop and as-if body loop — enacted in the body and read back, or simulated intra-cerebrally; where the somatic marker hypothesis decides how much body it actually needs.
• Background somatic states — the somatic marker framework’s model of mood: pre-existing states filter subsequent ones by congruence, so negative states breed pessimism.
• Ventromedial prefrontal cortex (VM) — the Phineas Gage lesion site; trigger structure for secondary inducers, and — under another name — one of Seth & Friston’s visceromotor areas.
• Convergence zones and dispositional knowledge — Damasio’s discarded foundation: content-free ensembles holding only the potential to reactivate patterns elsewhere; a generative model pointed backwards in time.
• Anxiety sensitivity — the belief that arousal sensations are harmful; the only thing distinguishing accurate heartbeat perceivers from inaccurate ones, which makes the wiki’s accuracy measure partly a sensibility measure.
• Schema-guided symptom perception — Pennebaker’s symptom schemata: perception of the body driven by past experience rather than present physiology. Interoceptive inference in clinical-psychology vocabulary, thirteen years before Seth.
• Cognitive model of panic — catastrophic misinterpretation of benign bodily sensations; the clinical theory that made cardiac interoception a research question, and which turns out not to have needed it.
• Attentional bias to threat — the anxious tendency to preferentially process threat cues and the body itself; the second of Ehlers’s three characteristics of panic patients, and the seam along which “attending to the body” and “perceiving the body accurately” come apart.
• Theory of constructed emotion — Barrett’s positive account: emotion as a categorization performed by the same machinery as all perception, arriving seventh in a derivation that starts from the body budget and never needs an emotion mechanism.
• Degeneracy — dissimilar structures, same function; the one-step argument that turns a century of failed signature-hunting into a prediction — and makes the theory very hard to disconfirm.
• Default mode network — the wiki’s third intrinsic network, proposed to host the internal model; the whole prediction cascade, not the summary, is the concept.
• Situated conceptualization (concepts as predictions) — predictions are concepts, assembled ad hoc and grouped by goal rather than by feature; so ‘happiness’ means something specific and something different each time.
• The mental inference fallacy — inferring the feeling from the freezing; the error Barrett names and LeDoux built a programme on, which is one of the few things they agree about.
• Population thinking — Darwin’s move via Mayr: categories are populations, not types — and the anti-essentialist argument Barrett and Scarantino make identically before reaching opposite conclusions.
• Appraisal overlap — why positive emotions co-occur: they travel together to the degree they share their predicted appraisals, beyond shared pleasantness (Tong & Jia); a correlational construct that describes a regularity without naming a mechanism.
• Core Response Network (CRN) — SE’s unit of trauma: four subcortical systems (autonomic, limbic, emotional-motor, reticular-arousal) as one complex dynamical system that can get stuck in a dysfunctional, reversible attractor; a third carving of the emotional subcortex alongside Survival circuits and visceromotor regions.
• Biological completion — SE’s proposed mechanism of change and its one falsifiable claim: completing a defensive movement frozen mid-execution supplies the proprioceptive feedback that fires the parasympathetic reset — movement-specific, and untested.
• Polyvagal theory — Porges’s three-division ANS (sympathetic / dorsal-vagal shutdown / ventral-vagal social engagement); borrowed by SE, now read first-hand (Porges (2023)) and under strong physiological challenge (Grossman (2023), Is Polyvagal Theory scientifically valid?).
• Respiratory sinus arrhythmia (RSA) — the respiration-linked fluctuation of heart rate, quantified as HF-HRV; the near-universal non-invasive index of cardiac vagal tone and the single measure the whole vagal-psychophysiology literature (all of PVT included) leans on. The wiki keeps the index (RSA) distinct from the phenomenon (cardiac vagal tone) and from “general vagal tone,” because collapsing them is exactly the category mistake the Porges/Grossman debate turns on.
• Neuroception — Porges’s reflexive, pre-conscious detection of safety/threat that sets autonomic state without awareness or appraisal; its bottom-up limb is, on Porges’s own account, “functionally equivalent to interoception” — the bridge from PVT into this wiki, and one that overlaps heavily with Interoceptive inference and Amygdala-based threat detection under other names.
• Tonic immobility — the freeze/collapse pole of trauma, physiologically the co-activation of sympathetic and parasympathetic under inescapable threat — a state the reciprocal-tone model cannot produce, on old animal evidence.
• Traumatic memory — trauma as a memory-system disorder: overwhelming arousal impairs declarative encoding, so the event is stored implicitly as somatic sensations and images, re-lived rather than recalled, and keyed to arousal state; the interoceptive core is the “loss of affect states as signals.”
• Stress-induced analgesia — the endogenous-opioid mechanism of trauma numbing: a trauma cue triggers opioid release that blunts pain and affect together (naloxone-reversible, ~8 mg morphine-equivalent, decades later); the numbing pole opposite HPA axis hyperarousal.
• Cognitive reactivity (to mood) — why a recovered depressive who feels fine can still be at high relapse risk: mild dysphoria reactivates the ruminative self-referential processing of the past episode, and the degree of reactivation (not the mood) predicts relapse; Farb (2011) locates it in self-referential mPFC.
• Decentering — the protective counter-pole: experiencing thoughts and feelings as transient mental events rather than truths or cues to elaborate; the trainable core of MBCT and the psychological name for Perceptual inference (interoceptive), with a relapse-protective neural correlate (trait acceptance, sensory reactivity).
• Narrative and experiential self-reference — the dual modes of self-awareness the whole Farb corpus rests on: the self extended across time (mPFC-centred narrative elaboration, James’s “me”) vs the self in the present moment (right viscerosomatic representation, James’s “I”). Habitually integrated via right-insula-vmPFC coupling, decoupled by mindfulness training; the conceptual root of the sensory-vs-elaborative dyad running from 2007 through the relapse work.
• Two-factor model of relapse vulnerability — Farb & Segal’s account of depression relapse as two dissociable factors: a static sensory-deactivation trait that treatment does not move, and a dynamic prefrontal cognitive reactivity it does; vulnerability as over-reliance on elaboration to the detriment of sensory integration (not the Schachter-Singer two-factor theory). Wu et al. (2026) re-weight it: the somatosensory pole carries the predictive load.
• Negative self-referential bias (NSB) — depression’s self-schema made measurable: the tendency to endorse negative over positive trait words as self-descriptive (SRET). Its behavioural form tracks concurrent symptoms but not relapse; the relapse signal is the neural activity underneath it, on the prefrontal-elaboration-vs-somatosensory-embodiment axis. A cousin of Cognitive reactivity (to mood), dissociating from it in what it predicts.
• Empathy — feeling with another, modelled as re-representing their affective/bodily state on one’s own interoceptive machinery (insula/ACC); distinguished from mentalizing (TPJ/pSTS/mPFC), and shown trainable by compassion meditation — the wiki’s first construct about other people’s feelings.
• Interoceptive psychopathology — interoceptive dysfunction as a transdiagnostic dimension (panic, depression, eating, addiction, PTSD, somatic-symptom…), better probed by RDoC than DSM; five candidate dysfunctions, an empty biomarker cupboard, and the unresolved cause-or-consequence question. The clinical face of interoception.
• computational psychosomatics — the mechanistic half of the same frame: interoceptive disorders as failures of hierarchical Bayesian inference/control — mis-set priors, aberrant error, maladaptive precision; Stephan’s allostatic self-efficacy routes chronic dyshomeostasis to fatigue and depression. Elegant, unifying, and (the authors concede) evidentially indirect.
• Floatation-REST — the mirror image of interoceptive exposure: attenuate the exteroceptive world in a dark silent float tank so the breath and heartbeat get loud; preliminary anxiolytic effect reported in high-Anxiety sensitivity individuals — a rare “more contact, better outcome” candidate.
• Psychopathy — the wiki’s first forensic/personality-disorder construct: a two-factor disorder (Factor 1 interpersonal/affective, Factor 2 antisocial) whose behavioural pole, not its emotional core, is the one that tracks reduced heartbeat-discrimination accuracy — read via somatic markers as weak bodily guidance of behaviour.
• Acquired sociopathy — the clinical face of the ventromedial lesion, and the bridge the somatic marker framework walks across to reach psychiatry: normal premorbid conduct, blunted emotional engagement, ruinous real-life choices, intellect intact. Kept separate from Psychopathy because the two literatures currently point at different halves of that construct — the acquired form at the affective pole, the interoceptive evidence at the antisocial one.
• Sense of should — the felt obligation to conform to others’ expectations, held to be a motive distinct from reputation-seeking and moral obligation; experienced as an anticipatory aversion to violating expectations, and functioning to keep one’s social world predictable and metabolically cheap. The wiki’s first construct about social pressure.
• Metabolic cost of prediction error — the biological engine of the sense-of-should framework: encoding prediction error costs energy, so unpredictability has a price and predictability is a resource organisms act to secure. The bridge from free-energy predictive coding to a motivational claim; also the constructing-vs-coasting (explore/exploit) trade-off.
• Interactive inference — Theriault et al.’s constructive answer to Barrett’s The mental inference fallacy: since minds cannot be read off behaviour, you infer them by forming a hypothesis from priors, acting, and reading the resulting prediction error (and its arousal) as evidence — an embodied, trial-and-error route to inferring other minds.
• Maturational dualism — Mendes’s proposal that mind–body links weaken across adulthood, so emotion involves fewer, less intense bodily sensations in older age; a body-based complement to motivational/expertise accounts of emotional aging, and the theory behind Age-related interoceptive decline. Predicts both easier regulation and worse gut decisions from one fading signal.
• Sickness behaviors — the immune-to-brain arm of interoception: inflammation signalled via vagus/cytokines/microglia evokes a stereotyped fatigue/anhedonia/withdrawal syndrome carried experientially by the insula; adaptive acutely, and the wiki’s proposed bridge from inflammation to depression when chronic. The humoral channel the cardiac-heavy material had been missing.
• The sensory-control loop — the closed circuit (states → sensations → actions → states) used to locate every computational model by which arc it models; the wiki’s one page that can hold rival frameworks side by side without adjudicating.
• Homeostatic reinforcement learning (HRL) — reward redefined as drive reduction, so ordinary RL machinery delivers body regulation; formally equivalent to active inference, and the wiki’s first account in which regulating the body may not require representing it.
• Forecasting (internal-state prediction) — predicting how internal states will evolve, under one’s actions and their own dynamics; assumed by every model of body regulation, formalized by almost none — a gap page, and the one the clinical “mis-set anticipation” claims are resting on.
• Emotional granularity — how finely a person’s emotion labels cut the same affective material, measured as a reverse-scored ICC across emotion ratings rather than asked about; the constructionist tradition’s favourite individual difference, and — on the wiki’s one source — two constructs, since positive and negative granularity are uncorrelated and only the negative half predicts anything.
• Alexithymia — a trait deficit in reading one’s own emotions, highly comorbid with autism and eating disorders, and repeatedly the variable that — once controlled — explains away interoceptive deficits credited to those conditions; on Brewer & Bird’s strong reading, “a general deficit of interoception.” The constructionist interoception→emotion link stated as a personality trait.
• Discrepancy awareness — noticing that the body felt different from what you expected, and the signed size of that gap as the person reports it. Kept deliberately apart from prediction error (a model parameter) and from failure to anticipate interoceptive states (a clinical disposition) — three things one vocabulary serves, of which only this one has ever been measured directly in a human here. Reliable and cross-modal, which is the profile the wiki has otherwise only seen in confidence — and the control that decided the confidence case (an exteroceptive comparison task) has never been run on this one.
• Nociception — pain has been an example on a dozen pages and the subject of none. Whether it counts as interoceptive is genuinely unsettled (Craig: all of it; Ceunen et al.: visceral only), and the choice is not cosmetic — it flips whether Wilzok’s cross-modal correlation is a within-interoception finding, and whether Ferentzi’s and Crucianelli’s nociceptive nulls count as evidence in Is interoception domain-general or organ-specific? at all. Also the channel that brings a thirty-year literature on biased expectation the interoception field does not cite.

Methods

• Lesion-symptom mapping — the only design in this wiki that speaks to necessity: case report, case series, lesion overlay, VLSM. Produces dissociations imaging cannot — and is confounded by vasculature, since lesions follow arteries rather than functional boundaries.
• Activation likelihood estimation (ALE) — the wiki’s first coordinate-based meta-analysis method: pool published peak coordinates from many neuroimaging studies, model each as a sample-size-scaled 3D Gaussian, and test whether they converge on the same location more than chance allows. Answers “where do these studies agree?” and nothing about how much (no effect sizes, no meta-regression). The design behind Nord et al.’s transdiagnostic mid-insula locus, and the tradition Lindquist used for the brain-basis-of-emotion question; the convergent-correlational counterpart to lesion mapping’s necessity test.
• Resting-state functional connectivity — task-free whole-brain fMRI; the method behind the wiki’s intrinsic-network constructs (DMN, salience, executive). Correlation structure of spontaneous low-frequency BOLD, read via ICA / seed / fALFF / REHO / PLS. Usable where task compliance fails (dementia), which is why the clinical FTD work depends on it; correlational and cross-sectional by nature.
• Meta-analytic reverse inference (Neurosynth decoding) — decoding what a brain region does from a large study database (Neurosynth, ~4400+ studies) by separating two quantities single studies conflate: forward inference = consistency (P(activation|state)) and reverse inference = specificity (P(state|activation)). A region can be consistently active for many states yet specific to one; only whole-literature base rates yield reverse inference. The method behind Chang’s insula decoding and Lindquist’s disgust-module reframe. Coarse-term only, and it inherits the source literature’s biases.
• Dynamic functional connectivity (DFC) — how the BOLD correlation structure changes over time within a scan: sliding windows clustered into recurring “brain states” (hypoconnected, DMN-anticorrelated, hyperconnected), quantified by dwell time and switching. More sensitive to within-scan / on-task variation than static FC, but analytically sensitive and — per Treves 2025 — its states need not track the behaviour of interest.
• Breath-counting task (BCT) — respiratory/attentional paradigm (Levinson 2014) that scores focused-attention meditation: count breaths in fixed cycles, press keys for counts, cycle-ends and self-caught lapses, so miscounts/resets index mind-wandering. With a breath belt, separates on-task from off-task periods for neural analysis. The behavioural bridge between the wiki’s contemplative and attention material.
• Heart rate variability (HRV) — beat-to-beat variation in heart timing; the wiki’s main efferent (vagal) cardiac measure, contested as an index of ‘vagal tone’
• Heartbeat detection task — cardiac; operationalizes interoceptive sensitivity.
• Heart Rate Discrimination Task (HRDT) — cardiac psychophysics; faster/slower judgments against your own live heart rate, separating sensitivity from precision and carrying a matched auditory control. Fixes counting’s response problems, not the cardiodynamic one.
• Respiratory Resistance Sensitivity Task (RRST) — respiratory psychophysics; which of two inhalations was harder, with a motor-driven resistive load on an adaptive staircase. The perceptual counterpart to the tracking task, and the instrument that made the cross-modal null credible.
• Filter Detection Task (FDT) — the RRST’s predecessor and the instrument the wiki named for three ingests without a page: add clinical breathing filters or a sham, ask yes/no plus confidence, staircase to ~60-85% accuracy. Four dissociable quantities from one session (threshold, decision bias, metacognitive bias, M-Ratio) and no slope; its staircase pins first-order accuracy in the band where M-Ratio behaves, which no cardiac task can do.
• Breathing Learning Task (BLT) — the wiki’s first interoceptive task built to be modelled trial by trial with concurrent imaging: predict from a cue whether the next breath will be loaded, 80 trials with four hidden reversals, fitted with Rescorla-Wagner or a Hierarchical Gaussian Filter to yield per-trial prediction certainty and prediction error as fMRI regressors. Not Forecasting (internal-state prediction) — what is learned is a cue-outcome contingency, not the body’s own dynamics.
• Multidimensional Assessment of Interoceptive Awareness (MAIA) — multimodal self-report; eight subscales of interoceptive sensibility, separating adaptive attention from anxious monitoring.
• Interoceptive Accuracy Scale (IAS) — 21-item self-report of believed accuracy across interoceptive domains (“I can always accurately perceive when I am hungry”), built to fill the empty self-report/accuracy cell of the Murphy-Catmur-Bird 2x2. The only questionnaire here shown to correlate with an objective interoceptive task — selectively, where the attention scales do not. Factor structure unresolved; validated in one channel.
• physiology congruence — electrodermal x self-report; within-person correspondence as a proxy for interoceptive coherence, usable in children.
• Ways of Coping checklist — episode-anchored self-report of coping thoughts and actions, sorted into problem- and emotion-focused functions (Folkman & Lazarus 1980); one of stress research’s most-used instruments. Measures coping used, not coping that worked, and its P/E item split turns out to be partly situational rather than fixed.
• Rubber hand illusion (RHI) — multimodal; manipulates body ownership, incl. cardio-visual variants.
• Continuous flash suppression — visual; high-contrast masks to one eye render a target to the other invisible for seconds, and the time to ‘break through’ into awareness (or a full-suppression fMRI version) probes unconscious visual processing and conscious access. Breaking-time is a contested proxy, best paired with a full-suppression or accuracy measure.
• Visual crowding — visual; a peripheral target flanked by distractors becomes hard to identify though still visible, scored by discrimination accuracy — no reaction-time or interocular-competition confound, so it serves as a converging test of effects first found with CFS.
• Adrenalin (epinephrine) injection paradigm — cardiac/autonomic; the two-factor theory’s core method.
• Cold pressor test — acute-stress induction (not an interoceptive measure); a hand in near-freezing water raises heart rate, blood pressure and (socially-evaluated variant) cortisol. Treacherous to pair with a cardiac task because the same sympathetic drive makes the heartbeat a louder signal — the cardiodynamic confound built into the stressor.
• False (bogus) physiological feedback paradigm — cardiac; tests effect of believed (bogus) bodily change on emotion.
• body-focused contemplative training — interoceptive attentional training; seven proposed mechanisms of action.
• loving-kindness meditation — loving-kindness / compassion practice; cultivating prosocial affect turned outward, and (unlike mindfulness) trained toward another’s distress via insula/ACC empathy circuitry.
• Interoceptive exposure — deliberate, non-avoidant contact with feared interoceptive sensation.
• device-guided breathing (DGB) — respiratory; paced to ~5 breaths/min, usually by device. The wiki’s only body-directed intervention that requires the person to notice nothing — pure reflex physiology (stretch receptors, baroreflex), which makes its status as an interoceptive intervention an interpretive choice rather than a property of the method.
• Thermal-grill illusion — thermosensory; ‘cold inhibits pain’ demonstration of central homeostatic integration.
• Pavlovian defense conditioning — multimodal; the best-characterized survival circuit paradigm (‘fear conditioning’, renamed on principle).
• emBODY (topographical bodily sensation mapping) — (Nummenmaa’s emBODY, not to be confused with EMBODY framework (brain decoding of internal attention)) whole-body; subjects colour where they feel activity rising and falling, at ~50,000-pixel resolution; measures felt topography with no physiological ground truth.
• EMBODY framework (brain decoding of internal attention) — (Weng’s EMBODY, “Evaluating Multivariate Maps of BODY Awareness” — unrelated to Nummenmaa’s emBODY (topographical bodily sensation mapping)) whole-brain fMRI decoding; per-person MVPA classifiers trained on five instructed internal-attention states, then applied across-task to decode an uninstructed meditation into breath / mind-wandering / self-referential focus. The wiki’s only method that reads what a person’s internal attention is doing without a probe — and the only one whose validated training step feeds an unvalidated decode.
• Emotion-word rating norms (multidimensional semantic norming) — semantic self-report; ~20 raters per word across a large lexicon, yielding reusable per-word norms; measures what a concept contains, not what a body does.
• Iowa Gambling Task — behavioural; decision-making under genuine ambiguity (probabilities never learnable), the paradigm the somatic marker hypothesis was tested on.
• Skin conductance response (SCR) — electrodermal; one sympathetic channel, used as the operational definition of ‘somatic state activation’ — it can show a state occurred, not which one.
• Delay tasks (delayed response and delayed non-matching to sample) — behavioural; the primate working-memory delay paradigms with a verbal distracter added, imported into the somatic marker programme to show the VM patient’s decision deficit is not a memory deficit — and delivering an asymmetric dissociation plus a basal-forebrain split that makes ‘the VM patient’ at least two patients.
• Intuitive reasoning task (IRT) — behavioural; the Iowa task’s purpose-built successor, repairing the reversal and magnitude/variance confounds its own designers catalogued — and, so far, a one-laboratory instrument.
• Event-related potentials — electrocortical; millisecond timing bought at the cost of knowing where the signal came from — trust the latency, discount the anatomy.
• Heartbeat-evoked potential — cardiac electrocortical; the ERP locked to your own R-wave, not to a stimulus — a neural index of interoception needing no behavioural report, haunted by the cardiac field artefact it shares its latency with.
• Affective picture viewing — visual emotion induction; normed IAPS images plus SAM valence/arousal ratings, the shared paradigm behind the wiki’s sharpest live disagreement.
• Ecological Momentary Assessment (EMA) — naturalistic self-report; signal people through daily life to rate the present moment — memory-free and ecologically valid, but non-experimental, so it observes correspondence and cannot test causation.
• Multidimensional Experience Sampling (MDES) — laboratory thought probes; a fixed multi-item battery rating the content and form of ongoing self-generated thought, factor-analysed into latent dimensions rather than scored. Smallwood’s framework, extended by Banellis et al. with eight interoceptive/somatomotor items. The wiki’s only instrument that measures spontaneous bodily attention — what the mind drifts to unbidden — and firmly a sensibility measure: nothing is detected, nothing scored against ground truth.
• Acoustic startle response — reflex to sudden loud stimuli whose habituation rate is the measure; nonhabituation is a cardinal PTSD signature (93% vs 22%) and possibly a pre-trauma vulnerability marker, since it persists after recovery.
• Microneurography — percutaneous recording from peripheral nerves in awake humans; the wiki’s only method that measures interoceptive signalling at the nerve — and, because it reaches only near-surface nerves, unusable for most interoceptive afferents. Held as a capability (CT afferents, and now the human vagus), not an evidence base.
• Respiratory tracking task with visuomotor perturbation (respyra) — respiratory; breathe to follow a sinusoidal target with real-time visual biofeedback from a chest belt, with a visuomotor gain perturbation distorting the display while leaving the goal intact. The wiki’s only method that scores what a participant does to their body rather than what they perceive of it — objective, continuous, report-free, reliable (.857) — and the only one whose interoceptive credentials rest on the substrate rather than on the information, since the feedback loop is visual.
• Electrogastrography (EGG) — gastric; cutaneous abdominal electrodes recording the stomach’s ~0.05 Hz pacemaker rhythm, MRI-compatible so it can run inside the scanner. The wiki’s first method that asks the participant for a body and no judgment at all — and, at r = .74 across months and settings, one of its most reliable measurements. Costs: ~24% of runs discarded for signal quality, BMI-restricted by construction, and confounded with respiration.
• Phase-locking value (PLV) — multimodal; the consistency of the phase lag between two oscillations, amplitude-independent, compared against a time-shifted surrogate null. The statistic every Gastric network finding rests on — and the one that stops working at the individual level: it inflates on any confound that happens to be synchronized with the body signal, and its run-to-run reliability is ~0 while both its inputs retest well.
• Interoceptive attention task — multimodal fMRI; cue an organ by name (“HEART”, “STOMACH”), hold attention on its sensations for ~10 s, subtract a matched exteroceptive-attention block. The dominant paradigm in interoceptive neuroimaging and the wiki’s most-cited unpaged method until now. Its virtue is that it perturbs nothing, so it is the only way two organs can be compared in one session; its cost is that it has no correct answer — nothing verifies that the participant attended to the organ named, and papers using it are routinely titled about awareness while measuring attention.
• Day Reconstruction Method (DRM) — next-day episodic self-report; reconstruct yesterday as bounded episodes (time, activity, place, company) and rate emotions on each. Kahneman’s compromise between the life-satisfaction survey and Ecological Momentary Assessment (EMA) — enough within-person observations for granularity at a fraction of the burden, at the price of being recall rather than sampling.
• Emotion differentiation (ED) task — IAPS pictures rated twice, once for valence/arousal and once against a fixed set of discrete emotion labels; the laboratory counterpart to computing granularity from daily life. The wiki’s one datum on it is a comprehensive null, which its authors read as too few labels and stimuli that move Core affect without evoking specific emotions.
• Property association task — semantic self-report; rate how strongly a property (interoceptive/behavioral/situational) comes to mind for an emotion category, isolating the content of the concept. Measures interoceptive knowledge, not accuracy or physiology — the instrument behind MacCormack’s aging Study 1.
• Accuracy of Interoceptive Anticipation and Interoceptive Discrepancy paradigms — multimodal; three grey-shade cues mapped to three stimulus intensities, then either rate anticipation retrospectively on unstimulated trials (AIA) or rate it before every stimulus and violate it by ±1 level (ID). The wiki’s only tasks whose dependent variable is a mismatch between a forecast and an outcome, and the only ones running the identical protocol in two interoceptive channels. Not psychophysics: no threshold, no slope, no correct answer — a signed gap between two VAS marks, with all that implies for difference-score noise and shared scale use.
• Sustained Attention to Response Task (SART) — non-interoceptive visual go/no-go; respond to every digit except a rare target. Held here because it isn’t interoceptive: an improvement on it after contemplative training is a transfer effect, not practice on trained material, and it was the only measure in an eight-instrument battery that distinguished two yoga arms where every questionnaire failed to.
• Ex-Gaussian decomposition of reaction times — analytic; split an RT distribution into a Gaussian body (mu, sigma) and an exponential tail (tau), the field’s standard implicit index of lapses and mind-wandering. Its cautionary use here: tau can stay flat while what it means inverts (lapse-slowing vs deliberate control-slowing), so a null on tau is uninformative rather than negative.
• Self-Referential Encoding Task (SRET) — cognitive/affective (not interoceptive): judge whether positive/negative trait words describe you (self), another (other), or their letter-case (a non-evaluative baseline). Yields the behavioural negative self-referential bias and, in the scanner, the neural substrate of evaluative self-processing. The wiki’s first dedicated self-referential paradigm; enters via Farb’s dual-mode reading of self-processing, and delivers the lesson that its behavioural and neural readouts predict different clinical outcomes.
• Inspiratory Breathing Load (IBL) paradigm — respiratory; cue an upcoming aversive breathing restriction, deliver it on some trials, and derive “interoceptive prediction accuracy” from experienced-minus-anticipated. Paulus’s 2012 paradigm, the operational form of his clinical thesis, and the empirical footing of the wiki’s predictive-psychopathology material — one respiratory task at one load level, with the anticipation rating taken after the moment of non-arrival and therefore relief-contaminated. Held at one remove; no primary IBL paper is in raw/.
• Body Perception Questionnaire (BPQ) — Porges’s BPQ, a self-report sensibility instrument (belief/report about body and autonomic reactivity, no ground truth) long cited across the wiki without a page. Its subscales dissociate — awareness vs perceived autonomic reactivity — which is why it earns its keep: in Wiebking et al. depressed patients’ reactivity scores rose while the awareness subscale stayed flat. Sibling of the MAIA; reactivity subscales confounded with anxiety/distress.

Debates

• Do basic emotions have distinct autonomic signatures? — do basic emotions have distinct autonomic signatures? (open)
• inferential? — is interoception bottom-up read-out or top-down inference? (open)
• What were the real origins of the Schachter–Singer two-factor theory? — what were the real origins of Schachter–Singer? (open)
• contemplative training increase objective interoceptive accuracy? — does contemplative training raise objective interoceptive accuracy, or just shift attention/regulation? (open)
• Is perceptual inference itself a genuine regulatory mode, or does the act of ‘choosing to let go’ collapse it back into active inference? — is perceptual inference genuine regulation, or does ‘choosing to let go’ collapse it into active inference? (open, author’s take recorded)
• Do discrete emotions map onto distinct brain regions or networks? — do discrete emotions map onto distinct brain regions/networks, or emerge from domain-general ones? (open)
• Are the origins of interoception social or biological? — are core feelings like hunger and warmth a phylogenetic endowment, or constructed socially in infancy? (open)
• parietal workspace? — is subjective feeling constituted in the anterior insula (Craig) or a prefrontal/parietal cognitive workspace (LeDoux)? (open)
• Can we know whether other animals have conscious feelings? — does behavioural and circuit similarity license inferences about animal subjective experience? (open)
• Should emotion terms refer to feelings or to motive states? — should ‘emotion’ name feelings (LeDoux) or motive states (Scarantino), when both agree on the mechanism? (open)
• Are emotions culturally universal? — are emotions the same everywhere, and is the felt body more universal than the face because it cannot be observed and copied? (open)
• Is more interoceptive awareness better? — the applied pages assume more bodily contact helps; several sources report less contact with better outcomes, and more with worse. (open)
• Does somatic feedback guide decisions? — do bodily signals bias choice toward advantage, must they be felt, and does perceiving them help? (open; both sides now read first-hand, and the critics’ verdict is split — the anatomy survives, the mechanism does not)
• Is the heartbeat-counting task a valid measure of cardiac perception? — is the wiki’s foundational number a continuous measure of a real ability (Ehlers), or a valid score for under 10% of people pooled with a majority measuring something else (Van der Does)? (open; the field went with Ehlers’s practice without answering the argument)
• Is interoception domain-general or organ-specific? — is there one interoceptive sense expressed in every organ, or only channel-specific abilities sharing a name? The best-powered test finds cardiac and respiratory ability uncorrelated while confidence generalizes even to tones. (open; no general ability demonstrated, but the arousal defence is untested rather than refuted)
• Can brain–body coupling be measured in a person? — brain-body coupling measures are offered as the objective escape from report-dependent interoceptive tasks; the first direct test finds gastric-brain synchrony has no test-retest reliability in individuals, while both the gastric and the brain signal retest well. Real at the group level, causally supported, and currently unmeasurable in a person. (open; the discriminating experiment — reliability against recording length — is straightforward and unrun)
• Are appraisals causes of emotion, or descriptions of it? — do appraisals produce emotion, or describe an episode produced by something else? (open; the wiki has heard the case against causal appraisal four times and the case for it never)
• Are traumatic memories indelible? — is the subcortical trauma trace permanent-but-inhibitable (van der Kolk, following LeDoux) or a reversible dynamical state that can be resolved (Levine/SE)? both cite LeDoux to opposite ends. (open)
• Is Polyvagal Theory scientifically valid? — does the cardiac vagus have two functionally opposite brainstem sources (Porges), or does the nucleus ambiguus do essentially all cardiac vagal control while the dorsal motor nucleus does almost none (Grossman/Taylor, on ablation/optogenetic evidence)? The physiological premises are under strong expert challenge; Porges calls the critiques a strawman; clinical utility is a separate axis — where a meta-analysis finds a small benefit for a bundle that is mostly generic vagal-tone techniques — and evidential robustness a third, where a ~27,000-participant meta-analysis finds childhood adversity barely moves vagal tone at all. (open)
• How should mind-wandering be defined? — does Mind-wandering have an essential defining feature — relatively unconstrained dynamics (Christoff et al.’s dynamic framework), so rumination and deliberate planning fall outside it — or is it a cluster concept with no privileged feature and graded membership (Seli et al., family-resemblances)? Both accept graded membership; the disagreement is whether any feature is more defining. Upstream of the wiki’s Body-wandering paradox, because the two views split free drift toward the body from constrained fixation on a symptom. (open; conceptual, unadjudicated by any first-hand wiki source)

Researchers

• Amie Wallman-Jones — the wiki’s only source on exercise and interoception, arriving from sport science rather than psychiatry: physical activity as naturally occurring interoceptive exposure, and interoception as a state that everyday behaviour moves.
• William James — originator of the peripheral/embodied (James–Lange) theory of emotion.
• Walter B. Cannon — Cannon–Bard centralist theory; undifferentiated-arousal critique.
• Stanley Schachter — two-factor (cognitive-arousal) theory of emotion.
• Anil K. Seth — interoceptive inference; predictive processing of the embodied self.
• Karl J. Friston — the free-energy principle / predictive coding read at source; cited by name across this wiki’s interoception literature (Seth, Barrett, Stephan, Petzschner, Harrison) and held secondhand until now. Not an interoception researcher; the upstream theorist whose machinery the field imports inward.
• Giovanni Pezzulo — active-inference theorist (with Friston) of homeostatic regulation and motivated control — the efferent/allostatic half of the predictive picture the wiki usually meets through Seth (perception) or Stephan (psychopathology). Enters through the experimental door: senior author of Iodice et al. (2019), the false-heartbeat illusion of effort. Thin page (one study in raw/).
• A. D. (Bud) Craig — interoception and the insular hierarchy; ‘How do you feel?‘.
• Lisa Feldman Barrett — constructionist critique of basic emotions / natural kinds; and, since the 2017 ingest, the wiki’s most ambitious positive theory — a predictive-coding theorist who derives emotion from the body budget.
• Otniel E. Dror — historian of science; origins of the two-factor theory.
• Norman A. S. Farb — simulation-map model; active/perceptual inference; contemplative neuroscience.
• Joseph LeDoux — amygdala defense circuitry; survival circuits; skepticism about animal feelings.
• Andrea Scarantino — philosopher of emotion; New BET (emotions as motive states); emotions as homeostatic property cluster kinds.
• Lauri Nummenmaa — bodily maps of emotions; emBODY; the wiki’s principal empirical universalist and Barrett’s unengaged opposite number.
• Antonio R. Damasio — somatic markers; the emotion/feeling distinction the field runs on; the body loop and the as-if loop.
• Antoine Bechara — the Iowa Gambling Task and the pre-hunch result; the experimentalist half of the somatic marker programme; neuroeconomics.
• Barnaby D. Dunn — the wiki’s principal critic of the somatic marker hypothesis, who named interoceptive ability as the framework’s untested moderator and then went and measured it.
• Olga Pollatos — the pre-Dunn consensus in person: good heartbeat perceivers feel emotion more intensely, shown electrocortically — on a design with no bodily measurement in it.
• Beate M. Herbert — the Tübingen half of the Herbert & Pollatos partnership; the embodied-cognition reading of the Munich/Tübingen heartbeat tradition, and the “cardiac awareness indexes translation of body into feeling” framing behind alexithymia-as-disturbance-of-embodiment. One first-hand document (the 2012 review); her empirical programme known only by citation.
• Rainer Schandry — author of the heartbeat-counting task nearly every interoception claim here rests on, and of the stroke-volume finding that undermines reading it as perceptual skill.
• Edward Katkin — Schandry’s counterpart: author of the heartbeat discrimination task (the signal-detection alternative to counting), and of the earliest demonstration that good detectors feel emotion more intensely but not more pleasantly; the Katkin pole of the counting-vs-discrimination debate.
• Anke Ehlers — the clinician who gave the heartbeat task its stakes (good perceivers do worse in treatment) and its chief defence, then co-authored the reanalysis against her own conclusions.
• Eddie M. W. Tong — appraisal theory as a living programme rather than a target; specialist in differentiating positive emotions, and the wiki’s first first-hand voice from the Lazarus/Smith & Ellsworth/Scherer lineage.
• Richard S. Lazarus — founder of the appraisal-and-coping tradition in stress research and the head of the lineage Tong works in; the transactional model, primary/secondary appraisal, and (with Folkman) the problem-/emotion-focused coping distinction. Held first-hand at last through Folkman & Lazarus (1980) — but on coping, not the emotion-causation claim Barrett targets; and read first-hand he holds a reciprocal loop, not the serial ordering the critique attributes to him.
• Peter A. Levine — originator of Somatic Experiencing; a complex-dynamical-systems model of trauma built over 45 years, synthesized from Gellhorn, Porges, Damasio and LeDoux — and the therapy’s trademark-holder, so read as advocacy.
• Bessel van der Kolk — the clinician-synthesist who named “the body keeps the score”; trauma as somatic/implicit memory and biological stress response, welded from Janet-era clinical observation and 1990s neurobiology — and the wiki’s indelibility pole against Levine’s reversibility.
• Zindel V. Segal — co-developer of MBCT and theorist of cognitive reactivity/differential activation; senior author on Farb (2011) and the clinical anchor of the sensory-vs-elaborative distinction the wiki’s interoception theory rests on.
• Richard J. Davidson — founder of affective neuroscience; ‘affective style’ and well-being as trainable neural skills; senior author of the Wisconsin adept studies, whose compassion work routes prosocial emotion through the interoceptive insula.
• Antoine Lutz — neurophenomenologist and methodological lead of the Wisconsin contemplative program; first author on the compassion-meditation study; insists on distinguishing families of practice by their neural signatures.
• Sahib S. Khalsa — the wiki’s first convener rather than combatant: organizer of the Interoception Summit 2016 and first author of its roadmap; advocate of perturbation-based interoceptive assessment (isoproterenol, an interoceptive “cardiac stress test”) over resting self-report — and, in an earlier Iowa-era guise, the hands-on empiricist of Khalsa et al. (2009), the wiki’s evidence that objective heartbeat-detection accuracy declines with age.
• Hugo D. Critchley — foundational interoception imager the wiki had cited only secondhand until now: ‘Neural systems supporting interoceptive awareness’ (2004), the visceral-influences synthesis, and (with Garfinkel) the accuracy/sensibility/awareness distinction the taxonomy pages turn on.
• Martin P. Paulus — senior author of the roadmap and source of its clinical hinge: Paulus & Stein (2010), mood/anxiety disorders as failures to anticipate interoceptive states — the “noisy afferent inputs” framing behind Interoceptive psychopathology.
• Jordan E. Theriault — Barrett/Young-lineage social neuroscientist who runs the constructionist body-budget account into social cognition and morality: lead author of the sense-of-should framework and originator of Interactive inference.
• Michelle Craske — anxiety-disorder researcher (Barlow lineage) and originator of Interoceptive exposure; enters the wiki through one first-hand study (Zoellner & Craske (1999)), a name doing load-bearing work on the clinical pages while almost none of that work is yet in raw/.
• Jennifer K. MacCormack — lead of the wiki’s emotional-aging thread; applies psychological constructionism to the body’s changing role in emotion across the lifespan, and separates “interoceptive knowledge” from accuracy and sensibility.
• Kristen A. Lindquist — Barrett’s closest empirical collaborator; first author of the 2012 meta-analysis and senior author of MacCormack (2021), running constructionism from the brain-basis debate down the lifespan; theorist of language and development in emotion construction.
• Klaas Enno Stephan — founder of the Zurich TNU and the most-cited name in this wiki without a page until now: source of allostatic self-efficacy, the hypothesis that unresolved interoceptive surprise becomes a metacognitive verdict that one cannot regulate one’s body, and thence fatigue and depression. Held at one remove on eight pages; now senior author on the wiki’s first empirical test of interoceptive predictive coding, which reports a null against his own side of the framework.
• Olivia K. Harrison (née Faull) — (née Faull) the wiki’s respiratory instrument-builder: the FDT, the BLT, co-author on the RRST, and — under her former name — the PAG respiratory-threat work the wiki had been citing as an unrelated author. Organized by an organ rather than a construct, on the argument that breathing is the one channel where an aversive visceral stimulus can be applied and withdrawn safely and repeatedly.
• Frederike H. Petzschner — the Zurich TNU modeller behind computational psychosomatics, cited secondhand here for two ingests and read first-hand at last; keeps Marr’s levels apart, which is how she can be the source of the wiki’s two most deflationary computational claims while working inside the framework they deflate.
• Sarah N. Garfinkel — author of the accuracy/sensibility/awareness dissociation and the ITPE the wiki’s taxonomy and clinical pages turn on; the Sussex interoception programme’s methodologist, read first-hand at last through Quadt et al. (2018) after being cited by name across a dozen pages.
• Bruno Bonaz — Grenoble hepato-gastroenterologist and vagus researcher; the wiki’s first clinical voice from internal medicine rather than psychiatry, and the source of its VNS/cholinergic-anti-inflammatory and microbiota material.
• Karen S. Quigley — Northeastern psychophysiologist in the Barrett orbit; a methodologist in a theorist’s paper, who names the anatomical cause of the field’s correlational evidence base (unmyelinated diffuse afferents) and proposes Affective touch (C-tactile afferents) as the way around it.
• Manos Tsakiris — the bodily-self researcher the wiki has been running on since its early ingests without a page: “just a heartbeat away from one’s body” (low interoceptive sensitivity → stronger RHI), “mentalizing homeostasis” (the source of Social origins of interoception), and the free-energy self. Came to interoception from the outside in, via illusions of body ownership.
• Helen Y. Weng — UCSF contemplative neuroscientist; architect of the four-domain intervention framework that refuses to privilege meditation, and author of the wiki’s first measure of interoceptive attention that neither asks the participant nor counts heartbeats (brain decoding of meditative states).
• Cynthia J. Price — developer of MABT, present in the wiki since its first ingest as an unsourced vignette and read first-hand at last; holds that interoceptive awareness is an explicitly teachable skill rather than a byproduct of sitting practice — and directs the nonprofit for the therapy whose trials are the evidence.
• Emeran A. Mayer — UCLA neurogastroenterologist behind the reframing of “functional GI disorders” as disorders of brain-gut interaction — naming a class by a mechanism rather than by an absence.
• Micah Allen — Aarhus/Cambridge computational psychophysicist and head of the Embodied Computation Group; enters the wiki as a builder rather than a theorist, and both his instruments are respiratory by programme, not by accident. Holds that the field measures perception while theorizing control, and that interoceptive ability may not generalize across channels at all — he is senior author on the study reporting cardiac and respiratory ability uncorrelated, and also the author of the tools that would let each channel be measured on its own terms.
• Gary G. Berntson — Ohio State autonomic psychophysiologist, the older measurement tradition the wiki had been drawing on without a representative; holds that afferent interoception cannot be described apart from the efferent systems that change what is sensed, so the unit is a circuit, not a channel. Declines to take a side on the predictive debate and proposes a dynamic systems approach instead — the first alternative here to both the read-out and predictive-coding framings.
• Stephen W. Porges — author of Polyvagal theory, the most widely adopted and most contested ANS-emotion framework; autonomic state as an intervening variable, the ventral/dorsal cardiac-vagus split, the vagal brake, vagal efficiency and Neuroception. Read first-hand at last; enormous clinical reach, foundational premises under sustained physiological challenge.
• Paul Grossman — psychosomatic-medicine physiologist and long-time RSA methodologist; the wiki’s voice for the physiological case against PVT (with comparative physiologist E.W. Taylor). Holds that RSA is only an approximate index of cardiac vagal tone and that PVT’s dorsal-vagal mechanism is falsified by modern ablation/optogenetic evidence.
• Olivier Desmedt — the field’s most systematic measurement critic: argues interoception’s definition is internally contradictory (physiological vs phenomenon-based) and its dimensions detached from their weakly-converging measures. Both the conceptual argument and the data forcing it — his experimental work showing the heartbeat counting task is largely guessing from a believed heart rate. Proposes a phenomenon-based definition and a specificity-ordered hierarchical taxonomy.
• Jane Ogden — Surrey health psychologist who treats symptoms as psychologically constructed; the wiki’s outside-arrival voice on symptom perception, source of the threshold model and (with Laura Carter) the “Interoceptive Crossover” construct — how far a person perceives emotions and physical symptoms as similar. Held at one remove: only the 2023 qualitative study is in raw/.
• Jennifer Murphy — Royal Holloway psychologist in the Bird/Catmur alexithymia lineage; the “multidomain, multidimensional failure of interoception” phrasing (Murphy, Catmur & Bird 2018) and, in raw/, the propensity proposal — the distinction between interoceptive ability and the use of interoceptive signals. Two sources in raw/: the propensity commentary and the 2020 IAS validation, the wiki’s clearest case of a taxonomic distinction making a wrong-able prediction and surviving.
• Thomas Parr — the formalist of Friston’s Active inference programme; worked out the discrete-state (Markov decision process) belief-updating and its neuroanatomy, and applied it to attention/salience as precision-setting. Not an interoception researcher; enters through Parr et al. (2019), a visual paper that isolates attention as covert action and precision as the seat of computational phenotypes. Thin page: one paper in raw/.
• Jakob Hohwy — Monash philosopher of the predictive mind; named on Predictive coding as a founder (with Clark and Friston) from the first ingest, read first-hand at last. Author of The Predictive Mind (2013) and of self-evidencing (2016) — the motivational floor under Active inference. Reaches the wiki because interoceptive inference is a special case of the picture he articulates. Thin page: one paper in raw/.
• Stephen D. Liberles — Harvard/HHMI molecular sensory biologist; the wiki’s representative of interoception studied one level below the pathway — which vagal sensory cell expresses which receptor, projects where, and controls which reflex, dissected by single-cell genomics and Cre-line optogenetics. Senior author of Prescott & Liberles (2022); his lab produced much of the baroreceptor-PIEZO, airway and gut-satiety primary work. Holds that the field should identify sensors before modelling inference over them. Thin page: one review in raw/, overwhelmingly mouse.
• Sara L. Prescott — MIT (formerly Liberles lab) biologist of the airway as a sensory surface; lead author of Prescott & Liberles (2022) and of the 2020 vagal airway-sensory atlas that defined the P2RY1 water-sensing protective circuit. The receptor-level anatomy under Respiratory interoception’s exceptionality claim. Thin page: one review in raw/.
• Camilla L. Nord — Cambridge computational psychiatrist whose organizing move is transdiagnostic: find the mechanism-level disruption several disorders share and treat that, not the diagnosis, as the target. First author of the mid-insula meta-analysis and the wiki’s first primary methodologist of the transdiagnostic frame (where Khalsa and Quadt assert it, she runs the convergence test). Reads the dysgranular mid-insula as an interoceptive prediction-error encoder; also works transdiagnostic reward/punishment learning. Thin page: one paper in raw/.
• Tim Dalgleish — Cambridge clinical psychologist and affective neuroscientist at the MRC CBSU, and a leading advocate of transdiagnostic approaches to psychopathology; senior author of Nord, Lawson & Dalgleish (2021), whose cross-diagnostic logic is his programme applied to interoception. Thin page: one paper in raw/.
• Rebecca P. Lawson — Cambridge cognitive neuroscientist in the predictive-processing tradition, best known for the finding that adults with autism overestimate sensory volatility (aberrant precision) — the idiom behind the mid-insula-as-prediction-error reading of Nord et al. (2021), on which she is middle author. The interoceptive twin of the visual aberrant-precision phenotype in Parr et al.. Thin page: one paper in raw/.
• Georg Northoff — neurophilosopher of the self, whose signature move splits it into a mental aspect (cortical midline) and a physical “material me” (insula), and treats each as separately disorderable in psychiatric illness; also the rest-stimulus interaction thesis (an abnormal rest baseline can masquerade as an abnormal stimulus response). Senior author of Wiebking et al. (2010), the depressed material me. The bridge from Craig’s anatomy to the clinic. Thin page: one paper in raw/.

Applications

• contemplative interoceptive training for well-being — dimension: wellbeing; contemplative interoceptive training for stress resilience and well-being.
• Interoceptive training in clinical practice — dimension: clinical; trauma, chronic pain, substance use, and anorexia (with explicit cautions).
• Somatic Experiencing (SE) — dimension: clinical; Levine’s body-first trauma therapy (resourcing, titration, pendulation, completion); the best-specified and least-evidenced clinical application in the wiki.
• Mindful Awareness in Body-oriented Therapy (MABT) — dimension: clinical; Mindful Awareness in Body-oriented Therapy — the wiki’s most interoception-specific psychological intervention (interoceptive awareness taught explicitly and in stages, not as a component of something broader), and the only one with RCT evidence and a physiological readout (RSA at 12 months). Still does not show it works through interoception.
• Yoga as interoceptive training — dimension: wellbeing; the most widely practised candidate interoceptive training and among the least characterized, because yoga is simultaneously exercise and contemplative practice and almost no study specifies the mixture. Holds the exercise-vs-contemplative hypothesis pair and the wiki’s one trial that separates them.
• Bioelectronic medicine — dimension: clinical; VNS, spinal cord stimulation, TMS, tDCS, DBS as “electroceutical” intervention on brain-body communication. The wiki’s first application targeting the interoceptive pathway rather than interoceptive attention — a different bet about where these disorders live, held entirely at one remove.

Synthesis

• A heartbeat counting score is not, by itself, evidence of interoceptive accuracy — confidence: moderate; a heartbeat counting score mixes heart-rate estimation, cardiac signal strength and (for a minority) perception, with a four-check reading rule for citing counting results as accuracy evidence.

Gaps

• Does interoception decline with age because the signal fades or because perception does? — Does interoception decline with age because the signal fades or because perception does?
• Is alexithymia a deficit in perceiving the body or in beliefs about it? — Is alexithymia a deficit in perceiving the body or in beliefs about it?
• Does anxiety sharpen heartbeat detection but blunt breath detection? — Does anxiety sharpen heartbeat detection but blunt breath detection?
• Do bodily emotion maps record bodily states or shared emotion concepts? — Do bodily emotion maps record bodily states or shared emotion concepts?
• How do interoceptive signals travel from the human brainstem to cortex? — How do interoceptive signals travel from the human brainstem to cortex?
• Do people with panic disorder control their breathing differently? — Do people with panic disorder control their breathing differently?
• Do cross-channel interoceptive self-reports reflect the body or response style? — Do cross-channel interoceptive self-reports reflect the body or response style?
• Does a general interoceptive ability appear when the body is perturbed? — Does a general interoceptive ability appear when the body is perturbed?
• Does early caregiving shape how children come to feel their bodies? — Does early caregiving shape how children come to feel their bodies?
• Can people’s forecasts of how their bodily state will evolve be measured? — Can people’s forecasts of how their bodily state will evolve be measured?
• Is gastric-brain coupling unreliable because recordings are short, or because it is a state? — Is gastric-brain coupling unreliable because recordings are short, or because it is a state?
• Does gastric-brain coupling relate to anything a person feels? — Does gastric-brain coupling relate to anything a person feels?
• Can gastric interoceptive accuracy be measured with modern psychophysics? — Can gastric interoceptive accuracy be measured with modern psychophysics?
• Is heartbeat-counting performance a continuum or a mixture of two groups? — Is heartbeat-counting performance a continuum or a mixture of two groups?
• Do posterior insula lesions make body-ownership illusions weaker or stronger? — Do posterior insula lesions make body-ownership illusions weaker or stronger?
• Is interoceptive dysfunction in mental illness specific to interoception? — Is interoceptive dysfunction in mental illness specific to interoception?
• Does better heartbeat perception intensify felt emotion, or only amplify the body’s signal? — Does better heartbeat perception intensify felt emotion, or only amplify the body’s signal?
• What shape is the relationship between interoceptive contact and wellbeing? — What shape is the relationship between interoceptive contact and wellbeing?
• Is interoceptive dysfunction a cause or a consequence of mental illness? — Is interoceptive dysfunction a cause or a consequence of mental illness?
• Does interoceptively focused training raise objective accuracy under randomisation? — Does interoceptively focused training raise objective accuracy under randomisation?
• Does the interoceptive trait prediction error hold when sensibility is low? — Does the interoceptive trait prediction error hold when sensibility is low?
• Can report-physiology mismatch identify who benefits from interoceptive training? — Can report-physiology mismatch identify who benefits from interoceptive training?
• How can interoception be measured in people who cannot report it verbally? — How can interoception be measured in people who cannot report it verbally?
• Does physical activity change objective interoceptive accuracy, not just self-report? — Does physical activity change objective interoceptive accuracy, not just self-report?
• Does acutely removing bodily feedback impair decisions under ambiguity? — Does acutely removing bodily feedback impair decisions under ambiguity?
• Does interoceptive accuracy really moderate how bodily signals guide decisions? — Does interoceptive accuracy really moderate how bodily signals guide decisions?
• Can body-focused treatment shift the stable sensory marker of depressive relapse? — Can body-focused treatment shift the stable sensory marker of depressive relapse?
• How much of interoceptive accuracy is a passing state rather than a stable trait? — How much of interoceptive accuracy is a passing state rather than a stable trait?
• Does cardiac signal strength explain what heartbeat tasks seem to measure? — Does cardiac signal strength explain what heartbeat tasks seem to measure?
• Does completing an interrupted defensive movement resolve trauma better than other movement? — Does completing an interrupted defensive movement resolve trauma better than other movement?