Are the origins of interoception social or biological?

Debate at a glance

• Summary: Whether the capacity to feel one’s own body — and the identity of core feelings like hunger, warmth or pain — is a phylogenetic endowment of the homeostatic afferent system that experience merely tunes, or is constructed developmentally out of a caregiver’s responses to the infant, such that the origins of core feelings are (in Fotopoulou & Tsakiris’s phrase) ‘social not biological’.
• Status: open

Positions in the field

• A. D. (Bud) Craig: Biological/phylogenetic. Interoception is a specific anatomical system — small-diameter afferents → lamina I → VMpo → posterior insula, re-represented anteriorly — that evolved as the afferent side of homeostasis and is present, with a known evolutionary trajectory, across primates. Hunger, thirst, warmth and air hunger are homeostatic sensations carried by identified pathways, not social achievements. Craig does not argue against social influence; he simply never treats development as the explanandum, and the identity of a feeling is fixed by its afferent channel rather than learned. (Craig (2002), Craig (2009))
• Aikaterini Fotopoulou & Manos Tsakiris: Social. ‘Mentalizing homeostasis’ (2017): because infants lack the motor repertoire to act on the world and thereby disambiguate the causes of their own interoceptive states, they cannot bootstrap a generative model of their viscera alone. The caregiver’s responses are constitutive of the model, not an input to a pre-existing one — hence the origins of core subjective feelings such as hunger and satiation, cold and warmth are ‘social not biological’. NOTE, since the Quigley et al. (2021) ingest: read first-hand, Tsakiris states the developmental claim as an untested hypothesis (‘questions of when and how interoceptive awareness emerges in early life remain unexplored’; researchers ‘can now begin to test’), and does not himself use the ‘social not biological’ formulation. The strong slogan is Oldroyd et al.’s quotation of the 2017 paper, not his framing here. (Quigley et al. (2021))
• Kristina Oldroyd, Monisha Pasupathi & Cecilia Wainryb: Social, empirically hedged. Adopt Fotopoulou & Tsakiris’s framing and supply the wiki’s first data: attachment style tracks interoceptive sensibility profiles, and maternal rejection of negative emotion predicts lower self-report/physiology congruence in youth. Their operative claim is weaker than the slogan they quote — that caregiving shapes how accurately and trustingly a child comes to read an existing bodily signal, not that it creates the signal. (Oldroyd et al. (2019))
• Anil Seth: Neutral on origins, but structurally hospitable to the social account. If interoceptive perception is inference over a generative model, the model’s priors must come from somewhere, and nothing in the framework says they must be innate. Seth does not address development, so his position here is inferred from the architecture rather than stated. (Seth (2013), Seth & Friston (2016))
• Lisa Feldman Barrett: Constructionist, and a natural ally of the social account one level up. Discrete emotions are not natural kinds but are constructed from domain-general core affect plus conceptual knowledge — and conceptual knowledge is learned, largely from others. Barrett’s target is the categories built on bodily sensation rather than the sensation itself, so her position bears on this debate by analogy rather than directly. (Lindquist et al. (2012))

Connections

• Related debates: inferential?, Do discrete emotions map onto distinct brain regions or networks?
• Related concepts: Social origins of interoception, Attachment style, Interoception, Embodied selfhood, Interoceptive inference

Opened by the ingest of Oldroyd et al. (2019), the wiki’s first developmental source.

Why this is not simply a contradiction

It would be easy to file Craig and Fotopoulou/Tsakiris as head-on opponents. They mostly are not, and saying so would overstate the conflict:

• Craig’s claim is about channel and pathway — that there exists a specific afferent system whose activity constitutes homeostatic feeling. Nothing in it denies that experience tunes the system, and Craig himself notes the AIC’s protracted development and invokes distorted interoceptive predictions in anxiety.
• Fotopoulou & Tsakiris’s claim is about content and calibration — how an infant comes to know which state a given afferent volley signifies, and whether to trust it.

Most of the time these pass each other by. Oldroyd et al.’s own findings are compatible with Craig without remainder: an intact lamina I pathway plus a caregiver who taught you to distrust its output yields exactly their avoidant profile.

Where the disagreement is real

The tension concentrates in one sentence, which Oldroyd et al. quote approvingly: that the origins of core subjective feelings such as hunger and satiation, cold and warmth are “social not biological.” These are Craig’s paradigm cases. He treats them as homeostatic sensations with identified afferent channels and an evolutionary history — a thermoreceptive afferent means cold because of what it is wired to, not because someone told the infant so. The strong reading of “social not biological” says the feeling of cold has no determinate identity prior to a caregiver’s mediation. That is a genuine incompatibility, and it is not resolved by either literature.

Three ways it might dissolve, none yet tested:

1. Scope. If “origins” means the origins of the concept rather than the origins of the sensation, the conflict evaporates — and this reading is charitable, since it makes the position Barrett’s, applied one level lower. Fotopoulou & Tsakiris’s rhetoric (“social not biological”) resists it; their argument may not.
2. Precision, not content. In a predictive frame, the caregiver may set the precision assigned to interoceptive prediction errors — how much the signal is trusted — while the signal’s identity remains anatomically fixed. This would preserve Craig entirely and still deliver Oldroyd et al.’s data. It is the reading their Trust finding most directly supports.
3. Both, at different levels. Innate homeostatic signal, socially learned interpretation. Almost certainly the eventual answer, but as stated it is a truce rather than a theory: it needs to specify which layer of the interoceptive hierarchy stops being innate.

The social side, heard directly (Quigley et al. 2021)

Until now every statement of the social position in this wiki was a quotation of Fotopoulou & Tsakiris (2017) by someone else. Quigley et al. (2021) gives Tsakiris the pen, and two things change.

First, the claim gets weaker in his own hands. He writes that “questions of when and how interoceptive awareness emerges in early life remain unexplored,” blames limited-range interoceptive measures (especially in children), and describes the recent work as making it possible to “begin to test” the hypotheses about interoception’s social origins. He does not use the “social not biological” formulation. Whatever the 2017 paper says, the position as stated here is: plausible mechanism, no developmental data, measures only now arriving.

That does not settle the debate, but it relocates it. The sharp incompatibility recorded above — between Craig’s anatomically-fixed feeling identities and the claim that cold has no determinate identity before a caregiver mediates it — is carried by a slogan that its originator does not press when writing at length. The wiki should attribute the strong reading to Oldroyd et al.’s use of it and hold Tsakiris at reading (2) or (3) from the dissolution list above.

Second, the biological side gains an item from the social side’s own author. Affective touch (C-tactile afferents): CT afferents in hairy skin are tuned to the force, velocity and temperature of primate grooming, are reported as pleasant, and project to insula rather than somatosensory cortex. That is a congenital afferent channel specialized for contact with a caregiver. It cuts both ways and the wiki should say so:

• For the social account: it is precisely the hardware a socially-constituted interoception would need. Caregiver contact is not merely an occasion for learning; it has a dedicated line.
• Against the strong social account: the specialization is anatomical and present before any learning. If the channel that carries social touch is innate and its pleasantness is a property of the fibre population, then at least one affectively-loaded interoceptive signal has its identity fixed the way Craig says signals get their identity.

Neither Quigley et al. nor anyone else in the wiki draws this connection; it is the wiki’s inference from two sections of the same review, and is recorded as such.

What would move it

The debate is currently under-determined by evidence, and the honest position is that no source in this wiki tests it. What would help:

• Longitudinal developmental data. Oldroyd et al. name this themselves as the field’s precondition: “a reliable method of quantifying interoception across the lifespan that will facilitate longitudinal developmental studies.” No such data appears in any source here — every claim on the social side is cross-sectional.
• Separating detection from report. The strong social claim predicts a deficit in feeling; the weak claim predicts a deficit in trusting or admitting the feeling. Every existing measure on the social side (Multidimensional Assessment of Interoceptive Awareness (MAIA), physiology congruence) routes through self-report and so cannot tell them apart.
• The severe-deprivation case. If early caregiving constitutes core feelings, severely neglected children should show anomalies in hunger/thermal feeling itself, not merely in reporting or trusting it. This is the sharpest available test and the wiki has no source on it.