Gastric network

The stomach has a pacemaker. Interstitial cells of Cajal generate a slow electrical rhythm at roughly 0.05 Hz (one cycle per 20 seconds, or ~3 cycles per minute) that coordinates peristalsis, and unlike the heart’s it never stops and never much changes rate. Rebollo et al. (2018) recorded that rhythm cutaneously with EGG during resting-state fMRI, and found that a distributed set of brain regions holds a constant phase lag to it.

That is a strange kind of network. Every other resting-state network is defined by regions correlating with each other. This one is defined by regions correlating with a stomach.

What is in it

After strict confound control the surviving map is:

• occipital cortex and precuneus
• bilateral insula
• pre- and postcentral gyri (somatomotor)
• cerebellar clusters

Rebollo & Tallon-Baudry (2022) characterize the network’s position on the cortical hierarchy: it sits in unimodal sensory and motor cortex rather than transmodal association cortex. That placement is theoretically loaded. If gastric coupling were a matter of interpreting bodily state, one would expect transmodal, anterior-insular or default-mode territory. Instead the stomach’s rhythm shows up where the brain does sensing and moving.

Note what this does not look like. It is not the lamina-i-spinothalamocortical-pathway interoceptive anatomy Craig described, and it is not the central-autonomic-network. Whether the gastric network is an interoceptive network at all, or a rhythm that entrains sensorimotor cortex for reasons unrelated to visceral perception, is unresolved — nobody has shown that anyone feels anything as a function of their gastric–brain coupling.

The size correction

The published extent of this network was, for five years, mostly artefact. Levakov et al. (2023) found that without regressing head motion and non-grey-matter signal, 19.89% of grey matter appears gastric-coupled; with them, 1.16% does. Prior work — Rebollo et al. (2018), Rebollo & Tallon-Baudry (2022), Choe et al. (2021), Müller et al. (2022) — used RETROICOR at best and none regressed motion.

The reason is not sloppiness but something more interesting: the confounds are themselves phase-locked to the stomach. CSF signal, global signal and three of six head-motion parameters synchronize with the EGG rhythm. The likely common cause is respiration, which moves the head and modulates the gastric signal at an overlapping frequency. In this literature motion is not noise added on top of signal; it is correlated with the thing being measured, which is the worst case.

Does the stomach cause it?

The best evidence is not human and not correlational:

• Vagotomy and hunger (rats). Cao et al. (2022) found a sensory/motor/limbic network coupled to gastric rhythm in head-fixed rats — motion cannot explain it — modulated by hunger state and substantially reduced by bilateral vagotomy. Cut the nerve, lose the coupling.
• Vagal stimulation (humans). Müller et al. (2022) found auricular taVNS increases stomach–brain coupling. See bioelectronic-medicine.

Together these make a reasonable case that the coupling is driven by ascending vagal traffic from the stomach rather than by shared physiological noise. It is the strongest causal story attached to any brain–body coupling measure in this wiki.

And yet it cannot be measured in a person

This is the tension the page exists to hold. Group-level, the network is real, replicable across labs and datasets, and causally supported. Individual-level, Levakov et al. found the coupling has no test–retest reliability (r ≈ −.14 to .08 across runs in the same people), while the EGG signal alone retested at r = .74 and the fMRI connectivity was strongly subject-identifiable.

A phenomenon can be robust at the group level and unmeasurable at the individual level, and the gastric network currently is both. That is fatal for the biomarker uses — predicting weight loss, eating pathology, IBS severity — and harmless for the existence claim. See is-brain-body-coupling-a-reliable-individual-difference.

Why the wiki should care about the stomach specifically

Three reasons, none of them about digestion.

1. It is the channel where cross-modal coherence was originally claimed. Herbert et al. (2012) linked cardiac accuracy to gastric detection; Whitehead & Drescher (1980) to voluntary control of gastric motility. is-interoception-domain-general names gastric as the missing third axis in the decisive test — and Banellis et al. did not measure it.
2. It is where spontaneous thought points. The brain-constrained CCA in banellis-2026-body-wandering put stomach at the top of its thought loadings (0.58), and body-wandering’s neural signature is thalamo-somatomotor — the same unimodal territory the gastric network occupies. Whether that is coincidence or the same phenomenon met from two directions is untested and cheap to test.
3. It is a rhythm, not an event. Cardiac interoception is studied through discrete beats and heartbeat-evoked potentials. The gastric signal is a slow continuous oscillation with a stable individual frequency, which makes it the natural substrate for phase-based questions about how visceral timing organizes cortical activity — the kind of question the ERP tradition cannot ask.

Reached from the other end: attend to your stomach and the network appears (Haruki & Ogawa 2023)

Everything above defines this network by coupling — regions that hold a phase lag to an organ, found by recording an organ. Haruki & Ogawa (2023) arrived in overlapping territory by a route with no organ recording in it at all: they asked 31 people to attend to their stomachs for ten seconds.

The largest cluster in their study — 3793 voxels, t = 8.42 — was occipitotemporal and medial visual cortex during gastric attention, with eyes fixed on a word, no food cue, no distension. They report that this medial visual territory largely overlaps the Rebollo & Tallon-Baudry (2022) gastric map (their Fig. 2B, using the same NeuroVault collection).

Why this is the most interesting thing to happen to this page. The first item under Open below asked whether gastric–brain coupling relates to anything a person experiences. It is not answered — nobody has correlated an individual’s coupling strength with an individual’s percept, which is the question, and the reliability null makes that study hard to run. But the gap has narrowed in an unexpected direction: the territory the coupling occupies is territory that deliberate attention to the stomach engages. Two methods with nothing in common — a phase lag to an unfelt 20-second rhythm, and an instruction — land in overlapping cortex.

Two brakes, and they are not small.

• The overlap is eyeballed from a figure. No Dice coefficient, no spin test, no null distribution — a weaker standard than the same comparison received in Levakov et al. the same year, on this same map, in a paper whose whole point was that this literature’s spatial claims had been undertested.
• Imagery is not excluded. People asked to attend to an interior they cannot clearly feel may picture their abdomen, and the same contrast recruited left frontal eye field and premotor cortex alongside the visual areas — an oculomotor/imagery signature. “The stomach is wired to visual cortex” and “people imagine their stomachs” predict much the same map, and the paper does not consider the second.

The convergent evidence the authors marshal for the first reading is real and independent of their own data: direct gastric stimulation evokes occipital activity in rats (Cao et al. 2019), cats (Pigarev et al. 2013) and humans (Mayeli et al. 2021). Their functional rationale — gastric state is for foraging, which needs visuospatial information yoked to energy status — is also supported within their own results by orbitofrontal, hippocampal and motor recruitment (see gut-hippocampal-memory). Held as a promising convergence with a soft spatial claim under it.

The bypass proposal, and why the network’s shape matters (Engelen et al. 2023)

The wiki met this network through the paper that trimmed it and the paper that reached it by attention. Engelen, Solcà & Tallon-Baudry (2023) is the review those two are arguing with — the statement of the network from the lab that discovered it (Rebollo et al. 2018), placed inside a general theory of interoceptive rhythms.

Its contribution to this page is a reason the network’s composition is theoretically loaded. The gastric network “includes all exteroceptive cortices as well as motor regions, some default-network regions but very few transmodal, cognitive regions.” From that shape the review floats a strong proposal: the gastric rhythm may orchestrate brain dynamics by bypassing higher-level regions — coordinating sensorimotor cortex directly rather than routing through association cortex. This is the review’s own hedge as much as its claim (“this hypothesis has not been experimentally tested”), but it sharpens the unimodal-placement observation already on this page: a rhythm that lands in sensory and motor cortex and skips cognitive prefrontal cortex is doing something other than informing interpretation of bodily state.

It also supplies the framing this page’s third “why care” reason needs. Cardiac interoception is studied through discrete beats; the gastric signal is a slow continuous oscillation. Under the review’s scaffolding hypothesis (see cardiac-cycle-effects and phase-locking-value), that slow stable rhythm is a natural carrier wave — the areas coupled to the stomach are “organized along a gradient of phases, more suggestive of a travelling wave than near-zero phase-lag synchrony,” unlike respiration’s tighter phase alignment. Whether a travelling gastric wave functionally coordinates cortex is exactly the untested bypass claim.

Open

• Does gastric–brain coupling relate to anything a person experiences or reports? No study has connected an individual’s coupling strength to a percept, a feeling, or a questionnaire. The nearest thing is the territorial overlap above, which is a fact about maps rather than about people.
• Is the visual-cortex component of this network about vision at all, or is it what a rhythm does to unimodal cortex generally? Both the coupling map and the attention map include it, by different routes, and neither explains it.
• Is the surviving 1.16% the network, or the part of the network that survives a preprocessing choice? Without global signal regression the map is larger and more similar to prior work.
• Is there a gastric analogue of interoceptive accuracy at all — a perceptual task on this channel, as opposed to a coupling measure? interoceptive-control notes you cannot ask someone to drive their gastric motility; detection is a separate question, and largely a 1980s literature. Haruki & Ogawa’s attention paradigm does not fill this in — it has no correct answer, so it yields a gastric signature and no gastric score. See is-interoception-domain-general.