Contradictions register

This page indexes the places where sources held in this wiki disagree with each other, or where a newer source contradicts something the wiki previously said. Every entry was flagged on the linked pages during ingest; this register only collects them, and nothing here resolves them. “Open” means the wiki’s evidence does not favour either side; “partly reconciled” means a linked page offers a reading under which both claims can stand, though no source has tested it; “resolved” means one claim has been superseded or a wiki summary corrected in place. Larger disputes with field positions and a curator’s view of their own live on the debate pages, linked where relevant. Internal inconsistencies within a single paper (misreported sample sizes and the like) are recorded on the study pages and are not listed here.

Measurement: the heartbeat task and its scores

• Does heartbeat accuracy predict felt arousal? Pollatos et al. (Pollatos et al. (2005)) find good heartbeat counters rate affective pictures as more arousing (r = 0.34), and Wiens et al. (Wiens et al. (2000)) find accurate heartbeat discriminators feel films more intensely; Dunn et al. (Dunn et al. (2010)) find no main effect in a near-identical unselected design, only an interaction with the body’s actual response. Where flagged: Pollatos et al. (2005), Is more interoceptive awareness better?, Heartbeat detection task. Status: open (extreme-groups versus unselected samples and main-effect versus moderation models differ; all three agree there is no valence effect).
• What heartbeat undercounting shows. Ehlers (Ehlers (1993)) treats the counting task as valid, with undercounting as perceived beats being missed; Van der Does et al. (Does et al. (2000)) find under 10% of healthy adults count accurately and that the rest report a regular rhythm slower than their real heart rate, so the score is noise for most people. Where flagged: Is the heartbeat-counting task a valid measure of cardiac perception?. Status: open.
• Is good heartbeat perception a stable trait? Ehlers and Breuer (1996), as recorded on Anke Ehlers, and the wiki’s own definition on Interoceptive sensitivity treat it as a stable characteristic; Van der Does et al. (Does et al. (2000)) find fewer than half of accurate panic patients still accurate after treatment, while inaccuracy is stable. Where flagged: Interoceptive sensitivity, Does et al. (2000). Status: open.
• Continuous and categorical scores reverse each other. Ehlers et al. (1995) report distraction barely changes continuous error scores and Antony et al. (1995) report a general improvement after exercise (both as recorded on Heartbeat detection task); Van der Does et al. (Does et al. (2000)) rescore the same kind of data categorically and find distraction knocks 35% of accurate perceivers out of the category, while exercise only converts a subgroup, briefly. Where flagged: Heartbeat detection task. Status: open.
• Does acute stress raise cardiac accuracy? Schulz et al. (Schulz et al. (2013)) find a cold pressor raised heartbeat counting accuracy but lowered visual heartbeat discrimination accuracy in the same 42 people, so the two tasks give opposite answers to one question. Where flagged: Heartbeat detection task, Schulz et al. (2013). Status: open (the counting gain fits the cardiodynamic confound, since heart rate and blood pressure rose).
• Is counting just guessing from a believed heart rate? Desmedt, Corneille and colleagues (Desmedt et al. (2022), Olivier Desmedt) argue the count is largely belief and time estimation and predicts none of the mental-health outcomes it was adopted for; Murphy et al. (Murphy et al. (2020)) find counting accuracy tracks self-reported accuracy (IAS) but not attention questionnaires or time estimation, which a pure guessing account does not predict. Where flagged: Is the heartbeat-counting task a valid measure of cardiac perception?. Status: open (Murphy’s heart-rate-belief controls showed no effect in that sample).

Measurement: constructs and instruments

• What “sensibility” means. Garfinkel’s framework, as used by Quadt et al. (Quadt et al. (2018)), defines sensibility as belief about one’s own perceptual ability; Khalsa et al. (Khalsa et al. (2018)) define it as the tendency to focus on interoceptive stimuli; Ventura-Bort et al. (Ventura-Bort et al. (2021)) find these load on two orthogonal factors with opposite relations to well-being. Where flagged: Interoceptive taxonomy (Farb et al. Box 1). Status: open.
• What “interoceptive awareness” means. Garfinkel et al. (2015), as recorded on Sarah N. Garfinkel, use it for confidence and accuracy correspondence (“metacognitive awareness”); Farb et al. (Farb et al. (2015)) reject that usage because the term already has a contemplative meaning, and call the correspondence “coherence.” Where flagged: Interoceptive taxonomy (Farb et al. Box 1). Status: open.
• Can MAIA subscales be summed? The Multidimensional Assessment of Interoceptive Awareness (MAIA) page concluded total scores should not be used; Mehling et al. (Mehling et al. (2012)) found a model with one second-order factor over the eight scales fit almost as well as eight separate factors. Where flagged: Mehling et al. (2012). Status: partly reconciled (the case against totals now rests on Ventura-Bort’s finding that subscales predict outcomes in opposite directions, not on the original factor structure).
• Do accuracy and self-reported interoception come apart? The wiki records their dissociation as one of its most robust patterns (Interoceptive taxonomy (Farb et al. Box 1), Banellis et al. (2026)); Wallman-Jones et al. (Wallman-Jones et al. (2023)) find baseline heartbeat accuracy predicted higher daily self-reported interoception. Where flagged: Wallman-Jones et al. (2023), Interoceptive taxonomy (Farb et al. Box 1). Status: partly reconciled (different, state-level instrument aggregated over a week, N = 70 with a wide interval).
• Can brain and body coupling serve as an individual biomarker? The biomarker programme set out in the roadmap (Khalsa et al. (2018)) treats objective physiological indices as the way past report-dependent measures; Levakov et al. (Levakov et al. (2023)) find gastric and brain synchrony has essentially no test-retest reliability within a person. Where flagged: Can brain–body coupling be measured in a person?. Status: open.

Is interoception one capacity?

• A unitary interoceptive sense, or separate channels? The coherence tradition around Herbert and Olga Pollatos links cardiac accuracy to gastric sensitivity and treats interoception as one sense; Ferentzi et al. (Ferentzi et al. (2018)) find no general factor across six channels, and Banellis et al. (Banellis et al. (2026)) find no cardiac and respiratory correlation in sensitivity, precision or metacognitive efficiency. Where flagged: Is interoception domain-general or organ-specific?. Status: open.
• Cross-modal metacognitive insight. Garfinkel et al. (2016), as recorded on Sarah N. Garfinkel, report a cardiac and respiratory correlation in metacognitive insight of about r = 0.36; Banellis et al. (Banellis et al. (2026)), powered to detect r of 0.179, fail to replicate it for insight while replicating it for confidence. Where flagged: Banellis et al. (2026), Is interoception domain-general or organ-specific?. Status: open.
• Does the insula integrate channels or separate them? Wilzok et al. (Wilzok et al. (2023)) infer, without imaging, that discrepancy awareness is computed in the insula across modalities; Haruki and Ogawa (Haruki & Ogawa (2023)) find the insula is where cardiac and gastric attention diverge. Where flagged: Wilzok et al. (2023). Status: open.
• Is respiration a poor or a privileged channel? The Respiratory interoception page argues voluntary control makes breathing the least representative channel, so respiratory evidence may not generalise; Allen (Allen (2026)) treats the same property as what makes interoceptive control measurable at all. Where flagged: Respiratory interoception. Status: partly reconciled (both are held together, implying interoceptive control may be a fact about breathing rather than a general construct).

Insula and anatomy

• Where awareness of the body sits. Craig (Craig (2009)) makes awareness the business of the anterior insula; Jones et al. (Jones et al. (2010)) report that anosognosia for hemiplegia and somatoparaphrenia follow posterior insula damage. Where flagged: Jones et al. (2010), Insular cortex, Anosognosia and disorders of bodily unawareness. Status: open.
• Is the insula needed for pain? Craig (Craig (2002)) classes pain as an interoceptive modality represented in the insula, and posterior lesions have been reported to raise pain thresholds or abolish its aversiveness; Jones et al. (Jones et al. (2010)) cite patients with insular strokes who rated noxious stimuli as more intense, with no insula activation. Where flagged: Insular cortex, Jones et al. (2010), Chronic pain. Status: open.
• Craig’s forebrain asymmetry. Craig (Craig (2009)) assigns withdrawal emotions to the right anterior insula and affiliative ones to the left; Lindquist et al. (Lindquist et al. (2012)) find no region specific to any emotion category, and Jones et al. (Jones et al. (2010)) note disgust deficits follow left, not right, insula damage. Where flagged: Lindquist et al. (2012), Insular cortex, Jones et al. (2010). Status: open.
• Where feelings are constituted. Craig (Craig (2009)) places the integrated momentary feeling state in the anterior insula; LeDoux (LeDoux (2012)) places a parallel construct in a prefrontal and parietal workspace and gives the insula no role. Where flagged: LeDoux (2012), parietal workspace?. Status: open.
• Shared activation across emotion categories. Barrett and Lindquist (Barrett (2017), Lindquist et al. (2012)) read it as degeneracy, a real many-to-one mapping; LeDoux (LeDoux (2012)) reads it as an artefact of imaging resolution concealing distinct cell populations. Where flagged: Degeneracy, Do discrete emotions map onto distinct brain regions or networks?. Status: open.
• Von Economo neurons in macaques. Craig (Craig (2009)) reports them absent in macaques; Chen et al. (Chen et al. (2021)) report them present, citing Evrard et al. (2012). Where flagged: Von Economo neurons (VENs). Status: resolved (the 2012 finding supersedes the 2009 claim).
• Is the right anterior insula the seat of body awareness? Farb et al. (Farb et al. (2007)) treat it as such; Farb et al. (Farb et al. (2013)) find a graded posterior to anterior field in which the anterior insula is “not an area of pure body awareness,” and Haruki and Ogawa (Haruki & Ogawa (2023)) suggest it may track arousal. Where flagged: Farb et al. (2007). Status: partly reconciled (the 2007 finding stands; its regional interpretation is the coarser one).
• Can agranular cortex receive prediction error? Seth and Friston (Seth & Friston (2016)) argue agranular visceromotor cortex is unfit to receive ascending prediction error; Harrison et al. (Harrison et al. (2021)) find prediction error tracked in the agranular anterior insula as strongly as anything measured. Where flagged: inferential?. Status: open (functional imaging cannot separate cortical layers, and the fitted model was a fallback).

Clinical

• Which way does anxiety move interoceptive accuracy? Ehlers (Ehlers (1993)) and Zoellner and Craske (Zoellner & Craske (1999)) find panic goes with better heartbeat perception; Harrison et al. (Harrison et al. (2021)) find anxiety goes with worse respiratory sensitivity; Desmedt et al. (Desmedt et al. (2022)) find counting accuracy flat against trait anxiety across 40 studies; Paulus and Stein (Paulus & Stein (2010)) predict reduced accuracy. Where flagged: Harrison et al. (2021), Paulus & Stein (2010), Is interoception domain-general or organ-specific?. Status: open (channel-specific effects or a counting-task artefact remain the two readings).
• Is the panic advantage specific to panic? Ehlers and Breuer (1992), as reported in Ehlers (1993), find panic patients count heartbeats more accurately than specific phobics and controls; Van der Does et al. (Does et al. (2000)) find no difference between panic and social phobia, generalised anxiety or specific phobia. Where flagged: Anke Ehlers, Cognitive model of panic. Status: partly reconciled (the advantage survives only as a prevalence difference that other anxiety disorders share).
• Is anxiety a disorder of prediction error? Paulus and Stein (Paulus & Stein (2010)) and Barrett and Simmons’s account (as recorded on computational psychosomatics) locate anxiety or illness in elevated interoceptive prediction error; Harrison et al. (Harrison et al. (2021)) find no anxiety effect anywhere in prediction error, only in prediction certainty. Where flagged: Harrison et al. (2021), Klaas Enno Stephan. Status: open (one study, n = 58, a respiratory load rather than a disorder).
• Alexithymia and heartbeat accuracy. Herbert and Pollatos (Herbert & Pollatos (2012)) cite high alexithymia going with low heartbeat perception; Desmedt et al. (Desmedt et al. (2022)) find the two share 0.01% of variance across 23 studies. Where flagged: Alexithymia. Status: partly reconciled (the link holds for self-reported interoception and vanishes for task performance).
• Interoceptive sensibility in autism. Garfinkel et al. (2016), as recorded on Interoceptive trait prediction error (ITPE), find autistic adults over-report their interoceptive ability; most sensibility studies in DuBois et al.’s review (DuBois et al. (2016)) find decreased awareness. Where flagged: Interoception in autism (ASD). Status: partly reconciled (the accuracy and sensibility gap, rather than the level, may carry the signal; untested when sensibility is low).
• Does stress raise or lower interoceptive accuracy? Oldroyd et al. (Oldroyd et al. (2019)) propose two stress routes (stroke volume, cortisol) that both predict higher accuracy; Bonaz et al. (Bonaz et al. (2021)) report childhood trauma correlating with lower interoceptive ability. Where flagged: HPA axis, Early adverse life events. Status: open (an acute versus chronic reading is available on HPA axis, but neither source makes it).
• Does early adversity blunt vagal physiology? Bonaz et al. (Bonaz et al. (2021)) report blunted physiological responses after childhood maltreatment; Wesarg et al. (Wesarg et al. (2022)), pooling about 27,000 participants, find no overall association with vagal activity or reactivity. Where flagged: Early adverse life events. Status: partly reconciled (a small blunting appears only in clinical samples and for direct maltreatment).
• Does trauma dampen or sensitise pain? Van der Kolk (Kolk (1994)) describes trauma producing opioid-mediated analgesia; Bonaz et al. (Bonaz et al. (2021)) have early adversity raising chronic pain risk. Where flagged: Chronic pain. Status: open (acute analgesia and chronic sensitisation could both hold; neither source addresses the other).
• Interoceptive accuracy in chronic pain. Bonaz et al. (Bonaz et al. (2021)), relaying Di Lernia et al. (2016), report accuracy deficits that predict symptom severity; Ribera d’Alcalà et al. (Ribera d’Alcalà et al. (2014)) find no accuracy or body-awareness difference between patients and controls. Where flagged: Chronic pain, Ribera d’Alcalà et al. (2014). Status: partly reconciled (the review concerns severity within patients, the study only a confounded group difference).
• The heartbeat-task caveat does not travel. Quadt et al. (Quadt et al. (2018)) say the tracking task’s validity “needs to be treated with caution”; Bonaz et al. (Bonaz et al. (2021)), sharing an author, and the roadmap (Khalsa et al. (2018)) use resting cardiac tasks to establish clinical deficits without that caveat. Where flagged: Hugo D. Critchley, Interoceptive psychopathology. Status: open.
• Is less felt body good or bad? Volynets et al. (Volynets et al. (2020)) read weaker felt bodily emotion in ageing as benign, aiding regulation and life satisfaction; Lyons et al. (Lyons et al. (2021)) find the same reduction in depression and recommend increasing body awareness in treatment. Where flagged: Age-related interoceptive decline, Is more interoceptive awareness better?. Status: open.
• Are traumatic memories indelible? Van der Kolk (Kolk (1994)), following LeDoux, holds the subcortical fear trace is permanent and can only be inhibited; Payne et al. (Payne et al. (2015)) hold trauma is a stuck state that is in principle fully reversible. Where flagged: Are traumatic memories indelible?, Kolk (1994). Status: open.
• Polyvagal theory’s anatomy. Porges (Porges (2023)) holds that dorsal and ventral vagal branches control the heart distinctly and that the ventral system is a mammalian adaptation; Grossman (Grossman (2023)) cites ablation and optogenetic evidence that the nucleus ambiguus does essentially all cardiac vagal control, and comparative evidence that the ventral system is not uniquely mammalian. Where flagged: Is Polyvagal Theory scientifically valid?, Polyvagal theory. Status: open (the wiki weights Grossman on the anatomy; the parties partly dispute what the theory asserts).
• RSA and social functioning. Polyvagal theory (Porges (2023)) builds its ventral vagal story to explain social engagement; Graziano and Derefinko (Graziano & Derefinko (2013)) find RSA withdrawal unrelated to children’s social functioning, which they flag as counter to polyvagal theory. Where flagged: Is Polyvagal Theory scientifically valid?. Status: open.

Decisions and somatic markers

• Does heartbeat accuracy improve decisions? Werner et al. (Werner et al. (2009)) find good heartbeat counters choose more advantageous Iowa Gambling Task decks; Dunn et al. (Dunn et al. (2010)) find accuracy unrelated to decision quality (r = .08), mattering only in interaction with the body’s response. Where flagged: Werner et al. (2009), Does somatic feedback guide decisions?. Status: open (extreme-groups versus unselected design, the same split as the arousal entry above).
• Does cutting peripheral feedback impair the gambling task? Bechara and Damasio (Bechara & Damasio (2005)) say the body loop is engaged most under ambiguity, as in the Iowa task; the peripheral manipulations collected by Dunn et al. (Dunn et al. (2006)) show little or no deficit, and pure autonomic failure patients outperformed controls. Where flagged: Body loop and as-if body loop, Does somatic feedback guide decisions?. Status: open (Damasio’s 1996 statement predicted little peripheral impact, so the result favours the earlier version of the framework).
• Is advantageous play implicit? The Iowa programme (Bechara & Damasio (2005)) holds that somatic signals bias choice before reportable knowledge, then restates that non-consciousness was never essential; Maia and McClelland, as reported by Dunn et al. (Dunn et al. (2006)), find focused probes reveal accurate reportable knowledge whenever play is advantageous, and Dunn et al. read the restatement as a retreat. Where flagged: Does somatic feedback guide decisions?, Dunn et al. (2006). Status: open.
• Two versions of the somatic marker hypothesis. Damasio (Damasio (1996)) makes the as-if loop the default mechanism and casts the amygdala as a downstream effector; Bechara and Damasio (Bechara & Damasio (2005)) index the loops to decision type and cast the amygdala as an upstream trigger, and neither text acknowledges the change. Where flagged: Damasio (1996), Somatic marker hypothesis, Amygdala. Status: partly reconciled (both are read as accurate snapshots of a framework that changed).
• Convergent validity of the gambling task. The wiki, following Bechara and Damasio (Bechara & Damasio (2005)), said the task correlates with gambling and delay-discounting tasks; Dunn et al. (Dunn et al. (2006)) show the cited study found only a partial link, in cocaine-dependent patients. Where flagged: Iowa Gambling Task. Status: resolved (the wiki’s summary was corrected in place).
• Insula lesions and risk-taking. Clark et al. (2008), as reported by Jones et al. (Jones et al. (2010)), find insula lesions produce high betting even at unfavourable odds; Weller et al. (2009), reported in the same review, find reduced risk-taking even when risk pays. Where flagged: Does somatic feedback guide decisions?. Status: partly reconciled (Jones et al. read both as insensitivity to expected value).

Interventions

• Is interoception the mechanism of mindfulness benefits? Farb et al. (Farb et al. (2015)) frame contemplative training as working through interoception, with self-report measures expected to capture its effects; Eusebio et al. (Eusebio et al. (2022)) find MAIA rose equally with or without interoceptive instruction, and Rai et al. (Rai et al. (2026)) find a training-induced MAIA rise mediated nothing. Where flagged: contemplative interoceptive training for well-being, contemplative training increase objective interoceptive accuracy?. Status: open.
• Do meditators perceive their heartbeats better? Practitioners report enhanced body awareness, and Farb et al. (Farb et al. (2015)) record the natural expectation of better accuracy; Khalsa et al. (2008), as recorded on contemplative training increase objective interoceptive accuracy?, find experienced meditators no better at heartbeat detection. Where flagged: contemplative training increase objective interoceptive accuracy?. Status: partly reconciled (training may change attention and regulation rather than accuracy).
• Does habitual exercise train or dull interoception? Wallman-Jones et al.’s 2021 review, as recorded on Physical activity, sedentary behaviour and interoception, treats activity as training the pathway and reads athletes’ attenuated insula response as efficiency; Wallman-Jones et al. (Wallman-Jones et al. (2023)) find habitual activity predicts lower self-reported interoception and read the same attenuation as habituation. Where flagged: Physical activity, sedentary behaviour and interoception, Amie Wallman-Jones. Status: open.

Theory

• Social or biological origins of core feelings. Craig (Craig (2002)) treats hunger, thirst and warmth as homeostatic sensations fixed by identified afferent pathways; Fotopoulou and Tsakiris, as quoted by Oldroyd et al. (Oldroyd et al. (2019)), call their origins “social not biological.” Where flagged: Are the origins of interoception social or biological?, Oldroyd et al. (2019). Status: partly reconciled (Tsakiris, writing in Quigley et al. (2021), states only an untested developmental hypothesis without the slogan).
• Is Craig’s model bottom-up? Seth (Seth (2013)) characterises Craig’s account as largely bottom-up; Craig (Craig (2009)) has the anterior insula represent “predictions of future feelings.” Where flagged: inferential?. Status: partly reconciled (the disagreement narrows to whether prediction sits at the top of an ascending hierarchy or is the architecture itself).
• Defining interoception by pathway. Craig (Craig (2002)) identifies interoception with homeostatic afferent pathways; Desmedt et al. (Desmedt et al. (2023)) show non-homeostatic Aβ fibres also process internal states, so no pathway is necessary and sufficient, and propose a phenomenon-based definition instead. Where flagged: Interoception, Desmedt et al. (2023). Status: open.
• Allostasis as premise or mechanism. Barrett (Barrett (2017)) makes allostasis the premise from which interoception and affect derive; Petzschner et al. (Petzschner et al. (2021)) treat it as subordinate to homeostasis, a way of moving the set-point. Where flagged: Allostasis. Status: partly reconciled (one is a claim about what brains are for, the other about what the set-point does).
• How autonomous is homeostasis? Quigley et al. (Quigley et al. (2021)) let homeostasis run independently with the brain intervening when needed; Bonaz et al. (Bonaz et al. (2021)) have homeostatic reflexes “enslaved” by descending allostatic policies. Where flagged: Homeostasis (and feelings as homeostatic sensations), Allostasis. Status: open (a difference of emphasis that decides whether dysfunction can be peripheral).
• Is all motivated behaviour allostatic? Barrett (Barrett (2017)) makes body budgeting the brain’s core task; Farb et al. (Farb et al. (2015)) argue some behaviour, such as roller-coasters and spicy food, deliberately sacrifices physiological balance. Where flagged: Allostasis. Status: open.
• What allostatic load is. McEwen and Barrett (Barrett (2017)) treat load as the cost of chronic over-regulation, with allostasis the primary frame; Payne et al. (Payne et al. (2015)), following Levine, treat load as a symptom of a stuck, reversible autonomic state. Where flagged: Allostasis. Status: partly reconciled (the stuck state could be relabelled as chronic allostatic over-activation, though Levine’s reversibility claim resists that).
• Autonomic specificity of emotion. Friedman (Friedman (2010)) reads six decades of psychophysiology as supporting emotion-specific autonomic patterns; Barrett (Barrett (2017)) doubts such signatures exist independent of conceptualisation, and Scarantino (Scarantino (2018)) accepts the negative record. Where flagged: Do basic emotions have distinct autonomic signatures?. Status: open.
• Can a variable emotion category be causally potent? Scarantino (Scarantino (2018)) keeps emotions as latent variables that cause expressions and autonomic change; Barrett (Barrett (2017)) holds categorisations are the events, with nothing underneath doing the causing. Both start from the same anti-essentialist premise. Where flagged: Basic emotions and natural kinds, Population thinking, Do discrete emotions map onto distinct brain regions or networks?. Status: open.
• Does the defensive state drive behaviour? Scarantino (Scarantino (2018)) grants the global organismic state causal power over behaviour; LeDoux (LeDoux (2012)) is more tentative, by Scarantino’s own account. Where flagged: Global organismic state. Status: open.
• What individuates emotions beyond valence and arousal? Tong (Tong (2015)) finds appraisal profiles separate 13 positive emotions; Barrett (Barrett (2017)) assigns the work to situated conceptualisation. Both agree the affective dimensions are insufficient. Where flagged: Tong (2015), Are appraisals causes of emotion, or descriptions of it?. Status: open.
• Do body maps record feeling or concept? Nummenmaa et al. (Nummenmaa et al. (2014)) and Volynets et al. (Volynets et al. (2020)) read bodily sensation maps as felt bodily states; the constructionist reading, given measured support by Ferré et al. (Ferré et al. (2024)), is that word-cued maps can reflect a shared concept of emotion that is itself largely interoceptive. Where flagged: Lauri Nummenmaa, Ferré et al. (2024). Status: open.