Central autonomic network
The efferent counterpart to the interoceptive afferent system — the machinery that acts on the body once its state has been represented. Bonaz et al. (2021) put it at the centre of their account of how interoceptive dysfunction becomes physical illness, and it fills a gap the wiki had been working around: this material had been distributed across visceromotor-areas (the cortical issuers of allostatic prediction), hpa-axis (the endocrine limb) and amygdala (the trigger), with nothing naming the system they belong to.
What it is
Per Bonaz et al.’s glossary, the CAN “controls the activity of preganglionic sympathetic and parasympathetic neurons in the dynamic control of visceral function.” It is hierarchically organized and integrated with interoceptive representations, and it spans:
| level | structures |
|---|---|
| cortical/forebrain | anterior cingulate cortex, insula, amygdala |
| subcortical | hypothalamus |
| brainstem | autonomic nuclei (NTS, parabrachial, PAG) |
| spinal | preganglionic centres |
Its function is to orchestrate autonomic, endocrine, motor and behavioural responses together — including stress responses and immune reactivity — “for adaptative survival within an uncertain external environment” (Beissner et al. 2013). The wiki’s existing pages hold pieces of this: the insula and ACC as limbic sensory/motor pair (Craig), the amygdala’s central nucleus driving sympathetic and HPA output (LeDoux), the agranular visceromotor cortices issuing descending predictions (EPIC). The CAN is the name for their union.
Reflexes enslaved by policies
The architectural claim is the one worth keeping, and it is a strong one:
Lower-order homeostatic reflexes couple afferent interoceptive signaling to specific efferent physiological outputs (e.g., in baroreflex control of blood pressure), yet these reflexes are enslaved by descending signals from the CAN that reflect more general, higher-order ‘allostatic policies’.
So the baroreflex is not an autonomous loop that higher centres occasionally interrupt; its set-point is continuously written from above in service of anticipated demand. This is allostasis given a concrete efferent implementation, and it is the same picture Petzschner et al. (2021) describe control-theoretically as the difference between a reflex arc with a fixed set-point and flexible control that moves it. Bonaz et al. add that both reflexes and allostatic responses “mostly operate unconsciously,” with interoceptive sensations reaching consciousness — and attracting appraisal — “especially in pathological states.”
Why it carries the clinical argument
The CAN is what makes Bonaz et al.’s “dysfunction at any level” claim mechanically coherent. If the system is one hierarchy from belief to preganglionic neuron, then a disturbance introduced anywhere propagates in both directions — a cortical belief can produce a gut symptom, and a peripheral pathology can distort representation. That is the logic behind treating IBS, chronic-pain, pelvic pain and dysautonomia as interoceptive conditions rather than as separate specialties’ problems.
Dysautonomia — any disorder of autonomic function, increasingly including conditions defined by primary maladaptive autonomic reactivity (vasovagal syncope, hyperhidrosis, postural tachycardia syndrome) rather than by nerve damage — is the CAN’s own failure mode, and the review’s link between joint-hypermobility and affective symptoms runs through it.
The brainstem floor, in enough detail to intervene on
Weng et al. (2021) fill in the brainstem row of the table above, because that is where their interventions act.
- NTS is the main termination of vagal afference (~80% of vagal fibres are afferent, carrying signals from head, neck, thorax and abdomen), and relays onward to the monoaminergic neuromodulatory nuclei — locus coeruleus and raphe. That relay is why stimulating a peripheral nerve can plausibly change cortical excitability and mood, and it is the whole mechanistic basis of bioelectronic-medicine.
- The NTS also receives somatosensory afference from the auricular branch of the vagus (ABVN), which is what makes non-invasive ear stimulation a route into this hierarchy.
- The dorsal medullary vagal system operates in phase with breathing (respiratory sinus arrhythmia is its visible form): NTS second-order relay neurons receive pulmonary-stretch and aortic-baroreceptor afference primarily during inhalation, with proposed inhibitory input during inhalation and facilitatory input during exhalation from the ventral respiratory column.
That last point is the one with consequences beyond this page. The afferent limb of the CAN is not a steady channel — its gain is phasically modulated by the respiratory cycle, one synapse in. Any measure of interoceptive signalling that ignores respiratory phase is averaging over a systematic modulation, and any intervention can exploit it (see RAVANS on bioelectronic-medicine, and respiratory-interoception).
The baroreflex appears here on both sides of the wiki’s ledger: as the paradigm example of a reflex “enslaved” by descending allostatic policy (Bonaz et al., above), and as the target of a bottom-up behavioural intervention that improves its sensitivity (slow-breathing). Both can be true — a set-point written from above can still have its loop gain improved from below — but the two literatures describe the same reflex in opposite directions without meeting, which is worth noticing.
The cortical top row, traced rather than inferred (Berntson & Khalsa 2021)
The table above puts “anterior cingulate cortex, insula, amygdala” in the cortical/forebrain row on the strength of imaging, lesion and meta-analytic evidence. Berntson & Khalsa (2021) fill that row in with a method that follows actual connections: retrograde transneuronal rabies virus tracing, injected into an organ and allowed to walk backward through a countable number of synapses.
The CAN’s own definition, which the review quotes from Benarroch (1993), is worth having verbatim: “an integral component of an internal regulation system through which the brain controls visceromotor, neuroendocrine, pain, and behavioral responses essential for survival.”
Adrenal medulla, monkey (Dum et al. 2016). Three distinct cortical networks on the lateral and medial prefrontal surfaces reach the gland in three synapses, each with a human equivalent involved in movement, cognition, and affect. So cortical control of the body’s main stress effector is not one pathway but three, differentiated by function.
Adrenal medulla, rat (Dum et al. 2019). Far more restricted — mainly primary/secondary motor cortex, with lesser input from primary somatosensory cortex and insula. The cortical areas that are the major source of cognitive control in the monkey are absent from the rat’s adrenal network. The review’s conclusion: “the network of brain–body connections is more widespread in primates.”
Stomach, rat (Levinthal & Strick 2020). The two autonomic limbs have different cortical origins: parasympathetically-linked neurons project from rostral insula and medial prefrontal cortex via three synapses; sympathetically-linked neurons from primary/secondary motor and primary somatosensory cortex via four. One organ, two cortical systems, and gut regulation therefore executed by “a complex set of sensorimotor feedback loops between the brain and periphery.”
The species finding deserves weight beyond this page. Most of the wiki’s circuit-level evidence is rodent, and most of its theoretical load — allostatic policy, appraisal, belief-driven regulation — is carried by exactly the cognitive-control cortex that the rat’s adrenal network lacks. A rodent null about cortical visceral control is weak evidence about humans; a rodent positive about brainstem circuitry likely transfers better. This is the general form of the caution Petzschner et al. raised for parabrachial→insula projections.
The NIH gives the efferent limb a vocabulary (Chen et al. 2021)
Chen et al. (2021) do something none of the sources above do: they name the parts of the descending limb, and then argue the whole limb belongs inside the definition of interoception rather than beside it.
- Regulatory signals of interoception — signals generated by the CNS to regulate interoceptive processes, often in response to interoceptive input but also to cognitive and exteroceptive factors
- Central regulators — the CNS neurons that generate them
- Interoceptive effectors — the target organs, whose changed function alters the signals subsequently sent back up
They stress, as this page’s framing does, that the descending limb is not only neural. Regulatory signals travel by vascular and lymphatic routes to interact directly with non-neural cells, and the two systems interleave: a signal may be delivered humorally in order to act on the neural pathway before reaching its final effector.
On the neural side they place the effector neurons more peripherally than this page had them. Spinal (sympathetic) efferents exit the cord and synapse onto either the sympathetic chain ganglia or the prevertebral ganglia, which then project directly onto peripheral organ cells; vagal/cranial (parasympathetic) efferents synapse onto parasympathetic ganglia located near the target organs. Those ganglia — not the cortical or brainstem sources — are the effectors in their scheme.
The definitional move is recorded at interoception as its seventh framing. The concession attached to it belongs here: because a regulatory signal’s impact is measured by the target organ’s response, the authors admit “it is sometimes impossible to distinguish regulatory interoceptive signals from body regulation.” This page exists because the efferent limb was real and unpaged; that is a different claim from the one Chen et al. make, which is that the efferent limb is interoception. The wiki holds the first and records the second.
The immune efferent limb
The CAN’s third output, alongside autonomic and endocrine. Berntson & Khalsa trace the reflex: peripheral inflammation → vagal afferents → NTS, which projects up to hypothalamus, amygdala, striatum, cingulate and insula, and laterally through intermediaries to the dorsal vagal motor nucleus, nucleus ambiguus and RVLM as the efferent limb. Stimulating mouse RVLM or vagal efferents suppresses innate immune responses and downregulates splenic proinflammatory cytokines by a cholinergic mechanism.
Two honest limits from the same authors: the precise role of those ascending targets in regulating immune function “is unclear,” and the account is “a limited view of the neural regulation of immune function.” See sickness-behaviors, bioelectronic-medicine.
Relation to polyvagal theory
Worth stating because the two are easy to conflate. The CAN is the mainstream, uncontroversial description of central autonomic control, resting on lesion, stimulation and imaging meta-analysis. polyvagal-theory is a specific and contested proposal about the evolutionary organization of the parasympathetic branch. Nothing on this page depends on the three-division scheme, and Bonaz — a vagus researcher — does not invoke it.