Where Feeling Lives

may 21, 2026

(on A.D. Craig 2002, How do you feel? Interoception: the sense of the physiological condition of the body, Nature Reviews Neuroscience 3:655-666)

Craig’s 2002 paper makes a claim that sounds modest and isn’t. He says interoception - the brain’s ongoing representation of the body’s physiological condition - is what subjective feeling is made of. Not informational input that becomes feeling somewhere downstream. The felt material itself.

The anatomy he marshals is specific enough to be worth taking seriously even if you reject the philosophy. A lamina I afferent pathway in primates carries signals about temperature, pain, itch, muscular ache, distension, sensual touch, oxygen status, blood pH, and a dozen other interoceptive channels up through distinct thalamic relays (VMpo, MDvc) and into posterior, then mid, then right anterior insula. The right anterior insula is, in Craig’s reading, where the represented body becomes felt. Lesion the pathway and you get specific disruptions in the felt quality of bodily states - not just in knowing-about-the-body, but in how the body shows up as something to oneself.

This is a stronger claim than it sounds.

The standard philosophical positions on emotion mostly differ in what they do with the body. James and Lange said feelings ARE bodily-state perceptions: the heart races, then you feel afraid, in that order. Schachter and Singer said the bodily state is undifferentiated arousal and cognition does the interpretation. Lisa Feldman Barrett’s constructionism treats interoception as one ingredient in a categorization-and-conceptualization process that’s what really generates emotion. Damasio’s somatic markers are body-state representations that bias decision-making without necessarily being identical to feeling.

Craig threads through all of these. He’s a Jamesian in spirit - feeling is bodily-state-becoming-conscious - but with the anatomical machinery James didn’t have. There’s a specific afferent system. A specific cortical destination. A specific evolutionary lineage: the lamina I pathway is poorly developed in non-primate mammals and reaches its peak elaboration in humans. If Craig is right about what this pathway does, then “feeling” in the strong sense may not be uniformly distributed across vertebrates the way folk intuition assumes. A rat has nociception; whether a rat has Craig-feelings is genuinely uncertain on his account in a way it isn’t on a constructionist one.

For the philosophy of embodiment specifically, the implications are strange in two directions at once.

The extended-cognition tradition - Clark, sometimes Varela, the various “4E” cognition writers - tends to push the cognitive unit outward. The body engages tools, tools engage environment, the boundary of the mind gets fuzzy and porous. Craig’s account refuses the move where feelings are concerned. Feelings are radically interior. They depend on an afferent system whose endpoints are specific cortical tissue. You can’t extend a feeling into a hammer the way you can extend a calculation into one. Whatever else embodiment turns out to mean, on Craig’s account the felt body has hard edges.

And yet Craig’s feeling-of-bodily-state is as embodied as anything gets. There’s no Cartesian residue. The right anterior insula isn’t a non-physical theater; it’s tissue, with cells, with a developmental trajectory, with lesion-sensitivity. You can stain it. You can image it. You can disrupt it. The account is anti-extended and maximally embodied at the same time. People who think those positions exclude each other are usually conflating “embodied” with “outward-facing.” Interior tissue is still tissue.

Where I’d push back on Craig is on the constitution claim, which is where the argument outruns the anatomy.

Craig shows that the lamina I → insula pathway is necessary for the felt quality of interoceptive states. Lesions there disrupt feeling in specific, characteristic ways. That’s strong. He hasn’t shown that the pathway is sufficient, and he hasn’t shown that the felt quality is exhausted by what happens there. The right anterior insula could be the readout point of a feeling-process that’s distributed more broadly across cortex and brainstem, with the insula serving as the place that integration becomes legible to the rest of cognition. Anatomical specificity isn’t the same as constitutive sufficiency, and Craig keeps quietly conscripting the specificity into doing sufficiency work. The lesion data narrows the candidate set; it doesn’t close it.

Predictive-processing accounts complicate Craig in a productive way. Anil Seth, Karl Friston, and Barrett-in-some-moods argue that what’s felt isn’t the raw afferent signal but the inferred state: the brain’s best guess about its body, updated by interoceptive prediction error. On this view the interoceptive pathway is where the error signals arrive; the felt quality is a property of the inference, not the afferent per se. This vindicates Craig’s pathway as central while denying his account of what makes the pathway felt. You can take Craig’s anatomy seriously without accepting that “interoceptive afferent representation” is identical to “feeling.” The afferent is the input to a process; the feeling is what the process delivers.

Against the constructionists, though, Craig is more right than they acknowledge. The anatomy is too clean and the lesion data too sharp for interoception to be merely one ingredient among several in a categorization process. The pathway has a specificity that “raw material for concept-laden categorization” doesn’t predict. If the constructionist were correct that the felt quality is mostly categorical construction over generic arousal, lesions to lamina I should disrupt that construction diffusely, not in the specific modality-preserving ways Craig documents. The data don’t quite fit the constructionist’s loose-ingredient story.

So a defensible position, sitting between the two: Craig is more right than the constructionists about there being a privileged neural pathway that traffics in felt material, and more wrong than he claims about that pathway being constitutive of feeling rather than tightly necessary for it.

What stays open is the question Craig’s anatomy can’t answer and the constructionists evade: why this particular tissue, this particular pathway, produces felt quality at all rather than just signaling. That’s the hard problem in its interoceptive register, and Craig neither solves it nor pretends to. He locates the place where the felt quality reliably shows up. That’s a major contribution. It’s also less than he sometimes implies. Knowing where feeling lives is not yet knowing what makes any of it feel like anything.

if it stayed with you, write to me.