Interoception

visceral sensebody sensehomeostatic afferent processing
the idea

The body's sense of its own internal state — the stream of signals by which the brain tracks heartbeat, breath, hunger, temperature, gut, and pain, and turns them into something felt. It marks the difference between knowing you have a heart and feeling it beat, and the field's central claim is that emotion itself is built from this channel before any thought or words arrive. Whether the brain reads these signals from the bottom up or predicts them from the top down is the field's running disagreement.

The body's sense of itself: the afferent signal stream by which the brain tracks the physiological condition of the body — heart rate, breath, hunger, thirst, gut state, temperature, pain — and represents that state as a felt internal condition. Two definitions sit on either side of the field's internal disagreement: the Craig 2002 formulation (the sense of the physiological condition of the body, with a specific thalamocortical pathway terminating in the right anterior insula), and the predictive-processing formulation worked out across Seth 2013 and Barrett & Simmons 2015 (actively-inferred predictions about expected body state, with prediction error driving learning). The first reads the signal as bottom-up; the second reads it as top-down. The two are sometimes treated as variants of the same construct and sometimes as importantly different things.

Etymology§

The term interoception is from the British neurophysiologist Charles Sherrington's 1906 The Integrative Action of the Nervous System, where he proposed a tripartite division of sensory function: exteroception (the body's interface with the outside world), proprioception (sense of self-position and limb-state), and interoception (sense of internal visceral state). Sherrington's coinage sat largely unused for most of the twentieth century — visceral signals were studied piecemeal under labels like visceral afferents, vagal afferents, and homeostatic feedback. The modern revival is from A. D. Craig's 2002 Nature Reviews Neuroscience review, which collapsed two decades of his spinothalamic-tract work into Sherrington's old term and gave it a new anatomical specificity. The post-2002 literature uses the word in Craig's sense; the predictive-processing literature retains the term while changing the directionality of the causal story.

Interoception is the construct that names the felt internal state — the difference between knowing you have a heart in your chest and feeling it beat, between knowing you are hungry and feeling the gnaw. The argument of the field is that subjective emotional experience is interoceptive at base: that what we call a feeling is the brain's representation of a body state, and that the body and the conscious mind communicate through this channel before any cognitive or linguistic processing kicks in.

The literature is structured by a series of distinctions. The first is between Craig's direct-readout model, in which the right anterior insula re-represents afferent body-state signals as conscious feeling, and the predictive-processing model worked out by Seth and Barrett and Simmons, in which the same anatomy generates top-down predictions about expected body state and reads prediction errors as feelings. The second is between interoceptive accuracy, sensibility, and awareness — three distinct constructs disambiguated by Garfinkel et al. 2015 and canonicalised in the 2018 Khalsa Roadmap. The third is between the neuroanatomy programme that traces back to Augustine's 1996 review and the clinical-translation programme running through addiction, anxiety, eating disorders, and the post-Roadmap psychiatry literature.

The field has its own internal critique about operational drift: the worry that interoception in 2025 covers a wider set of phenomena than Craig 2002 specified, with single-channel work on heartbeat and breath standing in for a much broader claim about feeling. The three-way Garfinkel distinction was a methodological correction in this direction, and the predictive- processing reframing is another. The most useful entry to the literature reads Craig and the Seth / Barrett-Simmons predictive frame against each other before reading either alongside the clinical- translation work.

Discussed in§

you are here in the territory

8 direct·1 two hops·14 further·46 off-graphopen the full territory →