The Emerging Science of Interoception: Sensing, Integrating, Interpreting, and Regulating Signals within the Self

Study at a glance

• Authors: Wen G. Chen, Dana Schloesser, Angela M. Arensdorf, Janine M. Simmons, Changhai Cui, Rita Valentino, James W. Gnadt, Lisbeth Nielsen, Coryse St. Hillaire-Clarke, Victoria Spruance, Todd S. Horowitz, Yolanda F. Vallejo, Helene M. Langevin
• Year: 2021
• Journal: Trends in Neurosciences
• DOI: 10.1016/j.tins.2020.10.007
• Design: Narrative review and proposed research framework (no new data); Trends in Neurosciences 44(1) special issue on the neuroscience of interoception, pp. 3-16. Open access (CC BY). Authored entirely by NIH program staff across ten institutes and offices, building on the NIH Blueprint workshop ‘The Science of Interoception and Its Roles in Nervous System Disorders’ (Bethesda, 16-17 April 2019).

Key findings

• Proposes a revised description of interoception that explicitly includes DESCENDING body regulation: the processes by which an organism senses, interprets, integrates, and regulates signals from within itself. The authors state plainly that this significantly expands the traditional scope and may differ from much of the current literature.
• Offers a principled criterion for the interoception/exteroception boundary: assess whether signals and responses REPRESENT, rather than ORIGINATE FROM, the internal or external world.
• Applies that criterion to three contested cases: proprioception is a form of interoception; the vestibular system belongs to interoception despite sharing a sensory organ with audition; gustation is genuinely mixed, but gastrointestinal nutrient sensing is clearly internal.
• Supplies a naming scheme for the whole loop: interoceptive signals, interoceptors, ascending and descending pathways, central interpreters, central integrators, central regulators, and interoceptive effectors.
• Sorts interoceptive signals into three types — biochemical (inorganic ions to peptides), mechanical (stretch, tissue extension, cellular deformation), and thermal/electromagnetic.
• States that the biochemical identity of most interoceptors remains unknown, and makes their systematic identification the field’s first Outstanding Question.
• Records substantial insular species differences: von Economo and fork neurons in AIC/ACC of macaques, great apes, humans, elephants and whales; the ventral salience network largely homologous between monkey and human but the dorsal attentional network far more developed in humans; markedly different amygdalar organization in rodents versus monkeys.
• Names the lifespan measurement gap concretely: there is no measure of interoceptive sensitivity suitable for non-verbal populations such as infants or older adults with dementia.
• Concedes that for most diseases linked to interoception the connection is ‘largely symptomatic and descriptive’ rather than pathophysiological or causal.

Limitations

• Not a scientific result but a FUNDING AGENCY’S RESEARCH AGENDA — all thirteen authors are NIH program staff, and the framework signals where an institutional portfolio is heading as much as where the evidence points. Carries an explicit disclaimer that it does not represent official NIH views.
• The central move — folding descending body regulation into interoception itself — is a definitional proposal supported by argument, not evidence, and the authors acknowledge it departs from prevailing usage.
• Concedes it is ‘sometimes impossible to distinguish regulatory interoceptive signals from body regulation’, which is a boundary problem the expansion creates rather than solves.
• Review with no data and no systematic search; coverage is broad and shallow, and each section is a gesture toward a companion article in the same issue.
• Several sections cite the workshop’s own participants and the co-authors’ institutes, and the urinary example is sourced substantially to co-author Valentino’s own work.

Connections

• Methods: Heartbeat detection task, Microneurography, Skin conductance response (SCR), Multidimensional Assessment of Interoceptive Awareness (MAIA)
• Concepts: Interoception, Exteroception Boundary, Proprioception, Interoceptive sensors, Central autonomic network, Insular cortex, Von Economo neurons (VENs), Lamina I spinothalamocortical pathway, Interoceptive taxonomy (Farb et al. Box 1), Age-related interoceptive decline, Salience network, Gut-hippocampal memory, Urinary interoception (lower urinary tract), Craving, Interoceptive psychopathology, Multisensory integration, Homeostasis (and feelings as homeostatic sensations), Allostasis
• Debates: inferential?, Is the heartbeat-counting task a valid measure of cardiac perception?
• Related studies: Berntson & Khalsa (2021), Quigley et al. (2021), Bonaz et al. (2021), Weng et al. (2021), Petzschner et al. (2021), Khalsa et al. (2018), Quadt et al. (2018)

The framing paper of the Trends in Neurosciences 44(1) special issue, and reference 1 in every other article in it. The wiki reached it last, having already ingested all five companion reviews — neural circuits, functions, diseases, interventions, computational models — each of which cites it as the source of the framework they say they are elaborating.

Reading it last turns out to be the right order, because on its own the review is thin. Its survey sections are one-paragraph gestures at articles this wiki has already read in full. What it uniquely contains is in two places: a definitional proposal and a boundary criterion.

What kind of document this is

Worth establishing before the content, because it changes how the claims should be weighted. All thirteen authors are NIH program staff — NCCIH, OBSSR, NIMH, NIAAA, NIDA, NINDS, NIA, NIDDK, NCI, NIDCR — with Helene Langevin, the NCCIH director, as senior author and Wen Chen, also NCCIH, corresponding. It arises from an NIH Blueprint workshop and its Outstanding Questions are recognizably a portfolio: what are the molecular entities of interoceptors, how can assessment be made objective and quantitative, what technologies are needed to modulate interoceptive processes.

This is not a criticism of the science, and the paper carries the standard disclaimer that it does not represent official NIH views. But a framework proposed by the agencies that fund a field is a different sort of object from one proposed by a laboratory: it is partly a description and partly an announcement of what will be fundable. The wiki should read the expansion of the definition below with that in view — noting, without insinuating, that widening interoception to include body regulation and non-neural (vascular, endocrine, immune) signalling brings a great deal of NCCIH’s complementary-and-integrative remit inside the term.

The definitional proposal: descending regulation moves inside

Every definition the wiki holds treats interoception as afferent — a sensing-and-representing process, whatever else is then built on it. Craig calls it the afferent side of homeostasis. The roadmap says senses, interprets, integrates. Quadt et al. stack four levels, all of them afferent or interpretive.

Chen et al. propose adding a fourth verb:

interoception includes the processes by which an organism senses, interprets, integrates, and regulates signals from within itself

with sensing denoting communication from the periphery to the CNS via ascending pathways and regulating denoting communication from the brain to other physiological systems via descending pathways. They name the parts accordingly: regulatory signals of interoception (the descending signals), central regulators (the CNS neurons generating them), and interoceptive effectors (the target organs, whose changed state alters what is subsequently sensed, closing the loop).

Two honest concessions come with it. First, that this “significantly expands the traditional scope of interoception and may differ from descriptions used in much of the current literature” — they know they are legislating, not reporting. Second, and more interesting, that because the impact of a regulatory signal is measured by the target organ’s response, “it is sometimes impossible to distinguish regulatory interoceptive signals from body regulation.” That is a boundary problem the expansion creates. If everything the brain does to the body counts as interoception, the term stops dividing anything.

How this sits with what the wiki already holds. It is a codification of a direction the field was already moving, not a break with it. Bonaz et al. already ran interoception as a bidirectional neuraxis with an efferent limb (Central autonomic network); Berntson’s reciprocity thesis already argued the unit is a circuit rather than a channel. What Chen et al. add is the decision to put that inside the definition, and the argument for doing so is pragmatic rather than theoretical: including body regulation “enables the development of more innovative methods to probe the functional impact of interoceptive processes.” It is a definition chosen for what it makes measurable.

Recorded on Interoception as its seventh framing, and not treated as overriding the six before it.

The boundary criterion: represent, not originate

The more durable contribution, and the one the wiki has needed for a long time.

The interoception/exteroception line has conventionally been drawn anatomically — the skin is the border, interoception is what happens below it. Chen et al. point out this fails in both directions and propose a functional replacement: assess whether the signals and responses represent, rather than originate from, the internal or the external world.

Three cases follow, and each of them changes something here:

• Proprioception is interoception. Neural activity in subcutaneous tissues, muscles and connective tissues that contributes to proprioception is, on this criterion, a form of interoception (citing Tuthill & Azim 2018). The wiki has been using the word for two ingests without a page — see Proprioception, now created — most consequentially in Payne et al., where “interoception and proprioception” are named together as the therapeutic lever of Somatic Experiencing (SE) with no account of why the pairing is natural. On Chen et al.’s criterion the pairing is not a pairing at all.
• The vestibular system is interoception. It sits in the same sensory organ as audition and is conventionally filed as exteroceptive, but it represents the organism’s own balance — internal state, external organ. This is the criterion’s cleanest win, because no anatomical rule gets it right.
• Gustation is genuinely mixed, sensing both worlds at once, but gastrointestinal nutrient sensing “is clearly more an indicator of internal status than a representation of the external world.”

That last case is the one to hold alongside its neighbours, because taste is now the third distinct proposal in this wiki for where the line falls, all published within a year of each other:

source	criterion	verdict on taste
Berntson & Khalsa	an act — the swallow marks the transition	undecided; gustatory afferents reach NTS like other visceral afferents
Petzschner et al.	the inferred state the signal informs, not the channel	not addressed directly
Chen et al.	what the signal represents, not where it originates	mixed at the tongue, internal in the gut

These are compatible and nobody adjudicates between them. Consolidated on the new Exteroception Boundary page, which exists because this question has been recurring on a dozen pages with nowhere to live.

What the survey sections are worth

Briefly, because most of it the wiki holds in better detail from the companion reviews:

Anatomy. Two ascending routes (cranial/vagal via nodose and jugular ganglia to NTS; spinal via dorsal root ganglia), with the hypothesis that vagal afferents primarily carry mechanoreceptor and chemosensory signals while spinal afferents carry temperature, pain and tissue injury — and the suggestion, worth flagging, that vagal and spinal afferents may represent opposing parasympathetic and sympathetic signals that inhibit each other within interoceptive brain regions. The naming convention “parasympathetic afferents” and “sympathetic afferents” is recorded on Lamina I spinothalamocortical pathway; the authors concede much remains to be studied about the differences between the two.

Insula. Penfield’s stimulation work, the viscerotopic map, the posterior-to-anterior topography consistent from rodent to human, thalamic relay to posterior insula, integration with exteroceptive sensorimotor and proprioceptive information posteriorly and centrally, AIC connected to OFC and ACC. Two things the wiki gains: the open question of whether there is a single body representation in the insula or multiple overlapping maps, and a compact statement of the species differences — see below and Insular cortex, Von Economo neurons (VENs).

Species. Von Economo and fork neurons in AIC and ACC of macaques, great apes and humans, plus elephants and whales; the ventral salience network largely homologous between monkeys and humans while the dorsal attentional-control network is much more developed in humans; markedly different amygdalar nuclear organization between rodents and monkeys; insular properties and cell types differing enough across species to limit “the applicability of some findings from rodents to human interoceptive health.” This is the second independent statement of the cross-species problem Berntson & Khalsa made the field’s structural gap, and it lands on Salience network as well: the network the wiki treats as one thing is homologous in its ventral half and not in its dorsal half.

Function. Two categories by direction of flow. Body-to-brain, where the exemplar is the gut-vagus-hippocampus memory work (Gut-hippocampal memory) held up as “an elegant example of how one could begin to address this challenge” — and where the authors concede it is “mostly unknown” how affect is influenced by interoception. Brain-to-body, where the exemplar is voluntary urination: CRF-labelled neurons in the pontine micturition centre (Barrington’s nucleus) whose stimulation causes immediate urination in mice and whose silencing inhibits it — sufficiency and necessity in one preparation (Urinary interoception (lower urinary tract)). Note that Rita Valentino, a co-author here, produced the foundational Barrington’s/CRF work being cited.

Disease. A long list — migraine and chronic pain, substance use, anxiety, depression, PTSD, OCD, autism, eating disorders, somatic symptom disorders, stroke and neurodegeneration — followed by the concession that matters more than the list: in most of these conditions “the links to interoceptive dysregulation are largely symptomatic and descriptive.” The wiki’s Interoceptive psychopathology page should be read with that sentence attached, from the funder’s own review. The Craving material (ghrelin predicting alcohol relapse; habenular TCF7L2 linking nicotine addiction to type 2 diabetes) duplicates what Weng et al. and Bonaz et al. carry in more detail.

Measurement. A short and rather bleak inventory of what human interoception research can currently do: heartbeat measures, skin conductance, self-report accuracy and awareness scales — “largely limited to a handful of approaches and remain mostly correlational” — plus Microneurography as the one recent addition offering objectivity. This is the second consecutive TICS review to nominate microneurography as the way out, after Quigley et al.; two of five companion papers pointing at the same underused technique is worth noting on that page.

The lifespan gap, stated as a measurement problem

The one gap section that adds something the wiki did not have in this form. Dynamic changes in interoceptive pathways across the lifespan “remain largely unexplored,” and the obstacle named is not funding or interest but instrumentation: there is no measure of interoceptive sensitivity suitable for non-verbal populations — infants, or older adults with dementia.

That is a precise diagnosis of why two of this wiki’s threads are thin at exactly their ends. Social origins of interoception rests on a developmental claim about infants whose interoception nobody can measure well (Maister et al. 2017 is the cited attempt); Age-related interoceptive decline extends into neurodegeneration where the same instrument fails from the other side. Both depend on heartbeat-counting-descended measures that require a verbal report of a counted quantity. See Interoceptive taxonomy (Farb et al. Box 1).

Placement

Read this as the field’s charter document rather than as a review: it is where the interoception research programme was given an official vocabulary, an official set of open questions, and an official widening to include the efferent limb. Its survey chapters are superseded by its own companion articles on nearly every point, and the wiki should cite those instead.

The two things to come back here for are the represent-not-originate criterion (Exteroception Boundary) and the explicit inclusion of regulation in the definition (Interoception). Both are proposals rather than findings, and both are worth holding precisely because the rest of the field now cites this paper as though they were settled.