The Central Autonomic Network

THE CLAIM

Benarroch named the circuit that governs the body's inner state

In October 1993 the Mayo Clinic neurologist Eduardo E. Benarroch published The Central Autonomic Network: Functional Organization, Dysfunction, and Perspective in Mayo Clinic Proceedings, volume 68, issue 10, pages 988 to 1001 (Benarroch 1993). Fourteen pages, one argument. The structures that regulate heart rate, blood pressure, temperature, breathing, hormone release, pain, and defensive behavior are not a scattered collection of reflex centers. They are one network. Benarroch gave that network a name and an address list, and the name held. Three decades later the abbreviation CAN turns up in cardiology, psychiatry, pain medicine, and imaging neuroscience, and the trail runs back to that paper.

Ask what a claim like that commits you to. If visceromotor output, neuroendocrine output, pain modulation, and behavior are governed by one interconnected circuit, then they cannot be four independent systems that happen to share a skull. A shift in one is a shift in the state of the whole. That is the proposition this page rests on, and it is Benarroch's proposition, stated in his own abstract long before anyone drew a functional connectivity map of it.

THE WORD

Autonomic never meant autonomous

John Newport Langley introduced the term autonomic nervous system in 1898 (Langley 1898), and the word has misled readers ever since. Autonomic sounds like self governing. It suggests machinery running beneath awareness on its own authority, sealed off from cortex and thought. Langley's own usage was narrower. He was naming a peripheral division of efferent nerves supplying smooth muscle, cardiac muscle, and glands, marked out by their ganglionic relay. He was not claiming those nerves had no master. The word traveled anyway, and by the middle of the twentieth century a great deal of teaching treated visceral control as a basement operation, mechanical and reflexive, with the interesting parts of the brain located somewhere upstairs.

Benarroch's 1993 paper is the correction. The autonomic nervous system has a head, and its head is cortical, limbic, hypothalamic, and brainstem all at once, with the parts talking back and forth rather than passing orders down a chain of command. Ask what follows from that. If the insula and the amygdala sit inside the autonomic network rather than beside it, then threat appraisal, memory, meaning, and social context are not outside influences on autonomic function. They are components of it. Nothing has to cross a border to get in, because the border does not exist.

THE SEVEN

Seven structures gathered into one anatomy

The 1993 abstract lists the network's members in a single sentence: the insular cortex, the amygdala, the hypothalamus, the periaqueductal gray matter, the parabrachial complex, the nucleus of the tractus solitarius, and the ventrolateral medulla. Seven named structures spanning cortex, limbic forebrain, diencephalon, midbrain, pons, and medulla. Read the list vertically and you are reading the whole neuraxis. No level of the brain sits outside this network. That was the structural surprise of the paper, and it was not obvious at the time, when autonomic physiology was still largely taught as medullary reflex arcs under loose hypothalamic supervision.

Benarroch widened the roster in later work. His 1998 study with Elizabeth Stotz-Potter on fatal familial insomnia (Benarroch and Stotz-Potter 1998) adds the anterior cingulate cortex and names hypothalamic subdivisions specifically, including the paraventricular nucleus, the dorsomedial nucleus, and the lateral hypothalamic area. The 2021 review with Guillaume Lamotte and Kamal Shouman (Lamotte and Shouman 2021) adds the locus coeruleus and the rostral ventrolateral medulla as distinct players. The list grew because the evidence grew. The architecture never changed. Every addition arrived as another node in a network, never as a new independent controller.

FOUR PROPERTIES

Reciprocal, parallel, state dependent, neurochemically complex

One sentence in the 1993 abstract does most of the conceptual work. Take its four properties separately. Reciprocal means information runs both directions between nodes, so no node is purely upstream of another. Parallel means several routes reach the same output, so a lesion in one path degrades the function without abolishing it. State dependent means the same input produces different output depending on what the system is already doing. Neurochemical complexity means the signaling is not one transmitter with one effect but many transmitters with overlapping targets and different time courses.

Those four words describe an oscillating regulatory system rather than a control panel. A network with reciprocal loops and state dependence has resonance. It has preferred operating points. It can be biased toward one configuration and then hold there, because the loops that maintain the configuration are the same loops that would have to change it. Ask what that predicts. It predicts that a nervous system can settle into a sustained pattern of output that no longer matches the situation in front of it, and that the pattern will present as the person rather than as a symptom. Benarroch did not phrase it that way. The properties he listed permit it.

THE INPUT

The solitary tract is where the body reports in

The nucleus of the tractus solitarius is the network's primary sensory gateway, and the 1993 abstract says so directly: inputs to the CAN are multiple, including viscerosensory inputs relayed on the nucleus of the tractus solitarius and humoral inputs relayed through the circumventricular organs. Two channels, two kinds of message. The neural channel carries traffic from baroreceptors, chemoreceptors, gut mechanoreceptors, and airway receptors, most of it arriving along the vagus and glossopharyngeal nerves. The humoral channel carries blood borne signals across the small set of regions where the blood brain barrier is deliberately absent.

Ask what the second channel implies. If part of the network reads the blood directly, then circulating state is a legitimate input to central regulation, on equal footing with nerve traffic. Inflammatory signaling, osmolarity, and hormone concentration are not background conditions. They are messages. Benarroch returned to this territory for the rest of his career, including a 2017 Neurology review he co-authored on the nucleus of the solitary tract and medullary reflexes (Cutsforth-Gregory and Benarroch 2017), and a 2019 review on autonomic and neuroimmune interactions (Benarroch 2019). The gateway is where the body's chemistry and the body's nerve traffic become a single conversation.

THE OUTPUT

One network, four output channels

The CAN does not have a single command line. Benarroch names four kinds of output: preganglionic sympathetic neurons, preganglionic parasympathetic neurons, neuroendocrine neurons, and respiratory and sphincter motoneurons. Smooth muscle, cardiac muscle, glands, hormones, breath, and continence, all controlled from one integrated source. The breadth is the point. When the network changes state it does not change heart rate alone. It changes heart rate, vascular tone, respiratory pattern, gut motility, hormone release, thermoregulation, and pelvic floor behavior together, as a coordinated pattern.

The same abstract dismantles the simple balance model. It states that the paraventricular and other hypothalamic nuclei contain mixed neuronal populations that control specific subsets of preganglionic sympathetic and parasympathetic neurons. Mixed populations. Specific subsets. That is not sympathetic on one pan of a scale and parasympathetic on the other. It is a network selecting particular combinations of output for particular situations. Sympathetic outflow to skin, to muscle, to kidney, and to heart can be dissociated from one another. Any teaching that reduces autonomic state to one dial running from stressed to relaxed is teaching something the anatomy contradicts.

THE CORTEX

The insula gives the viscera a cortical map

Cortical representation of visceral sensation is the finding that lifts the CAN out of the brainstem, and the evidence was fresh when Benarroch wrote. In 1987 David Cechetto and Clifford Saper published Evidence for a Viscerotopic Sensory Representation in the Cortex and Thalamus in the Rat in the Journal of Comparative Neurology, volume 262, issue 1, pages 27 to 45 (Cechetto and Saper 1987). Their conclusion was that the ascending visceral sensory pathway is organized viscerotopically at every level of the neuraxis, insular cortex included. Viscerotopic is the visceral counterpart of somatotopic. The organs have a map in the same sense that the skin has a map.

Five years later the human demonstration arrived. Stephen Oppenheimer, Adrian Gelb, John Girvin, and Vladimir Hachinski stimulated the insular cortex directly in five epilepsy patients during surgery and published in Neurology, volume 42, issue 9, pages 1727 to 1732, in September 1992 (Oppenheimer 1992). Left insular stimulation produced bradycardia and depressor responses more often than the reverse, right insular stimulation produced the opposite bias, and the asymmetry reached significance at p less than 0.005. Benarroch carried the clinical consequence into his own abstract: involvement of the insula and amygdala in seizures or stroke may produce severe cardiac arrhythmias. He kept returning to the region, including a dedicated Neurology review of insular cortex in 2019, volume 93, issue 21, pages 932 to 938 (Benarroch 2019).

WHEN IT BREAKS

Central autonomic disorders are the proof of the map

Lesion evidence is what turns a proposed network into an accepted one, and Benarroch had a clinic full of it. In the same year as the CAN paper he published Central Autonomic Disorders with F. L. Chang in the Journal of Clinical Neurophysiology, volume 10, issue 1, pages 39 to 50 (Benarroch and Chang 1993). Multiple system atrophy attacks preganglionic autonomic, respiratory, and neuroendocrine output together, producing orthostatic hypotension, urinary incontinence, laryngeal stridor, and sleep apnoea inside one progressive disease. Hypothalamic lesions produce hypothermia or hyperthermia. Medullary lesions produce orthostatic hypotension, paroxysmal hypertension, and sleep apnoea. Subarachnoid hemorrhage can produce arrhythmia, myocardial injury, hypertension, and pulmonary edema at once.

The sharpest case is fatal familial insomnia. Benarroch and Stotz-Potter reported severe neuronal depletion in the mediodorsal and anteroventral thalamic nuclei alongside exaggerated sympathetic activation with preserved parasympathetic drive to the cardiovascular system. Sleep, circadian rhythm, and autonomic regulation fail together because the same integrative machinery serves all three. Ask what that pattern tells you. If one focal degeneration disrupts sleep, cardiovascular control, and hormone rhythm simultaneously, then those functions were never separate to begin with. They were outputs of a shared regulatory state.

STRESS

The same network carries the stress response

In 2021, twenty eight years after the original paper, Benarroch returned to the CAN with Guillaume Lamotte and Kamal Shouman in Autonomic Neuroscience, volume 235, article 102870, under the title Stress and Central Autonomic Network (Lamotte and Shouman 2021). Their opening claim is that the CAN plays a critical role in the stress response, which is triggered either by challenges to homeostasis or by unpleasant social and environmental situations. Note the pairing. A hemorrhage and a humiliation enter the same network. The review assigns roles: the insula integrates physiological signals, the anterior cingulate generates predictive responses, the amygdala triggers survival responses through hypothalamus and brainstem, the hypothalamus patterns the response according to stimulus type, and the periaqueductal gray initiates the motor and autonomic strategy.

That last structure deserves attention. The periaqueductal gray does not simply raise arousal. It selects a strategy, and one of the strategies available is immobility rather than activation. Fight, flight, and playing dead are all outputs of the same node. Ask what that means for reading a body in front of you. High output and low output are not opposite ends of a single axis running from unwell to well. Both can be defensive selections made by an integrating network under load, and neither by itself tells you whether the system is regulating accurately.

THE IMAGING

The network became measurable

The CAN moved from anatomy to measurement once resting state functional MRI could be paired with beat to beat cardiovascular recording. In 2020 Ding, Tarumi, Wang, Vernino, Zhang, and Zhu published in Brain Structure and Function (Ding 2020) using 3 tesla resting state fMRI in 22 healthy older adults with a mean age of 68 plus or minus 6 years. Left amygdala connectivity with medial frontal gyrus, bilateral postcentral gyri, and paracentral lobules correlated with baroreflex sensitivity and R to R interval variability, with correlation coefficients running from 0.663 to 0.703. Right amygdala patterns differed, clustering instead in anterior cingulate cortex.

In 2024 Valenza, Di Cio, Toschi, and Barbieri went further in Imaging Neuroscience (Valenza 2024), analyzing 34 Human Connectome Project participants aged 22 to 36 across two sessions, using separate sympathetic and parasympathetic activity indices derived from the pulse signal. The sympathetic index correlated positively with the blood oxygen level dependent signal across insula, prefrontal cortex, cingulate, temporal and occipital cortex, amygdala, hippocampus, thalamus, basal ganglia, cerebellum, and brainstem. The parasympathetic index correlated negatively across largely the same territory. Same regions, opposite signs. Benarroch's reciprocal interconnections, made visible in a scanner three decades after he named them.

BENARROCH AND THE MODEL

The central autonomic network is the anatomy the Unified Model of Tone uses

Benarroch's central autonomic network, named in 1993, gives the architecture of the Unified Model of Tone its most complete anatomical description. The model reads regulation as one variable, tone, held across the whole nervous system rather than issued from a center. A claim that size needs a circuit with the reach to carry it. The network has that reach, as Benarroch and his successors describe it. Its members run from cortex to medulla: insular cortex, anterior cingulate cortex, ventromedial prefrontal cortex, amygdala, hypothalamus, periaqueductal gray, and brainstem nuclei including the rostral ventrolateral medulla and the nucleus of the tractus solitarius. That circuit integrates sensory, emotional, cognitive and homeostatic information into coordinated autonomic output. Tone is set by the whole network, with the brainstem as the final common pathway.

Emotional state, cognitive load, interoceptive accuracy, and postural input all modulate heart rate, airway caliber, blood pressure, and gut motility. They do that because they are inputs to one integrated circuit, not because four separate systems happen to move at the same time. The prefrontal cortex, the central autonomic network, the brainstem RVLM and nucleus of the tractus solitarius, the vagal afferents, the cardiac pulse, the respiratory rhythm, and the interoceptive stream form a single regulatory circuit. The variable that circuit regulates is tone.

Within the brainstem the rostral ventrolateral medulla generates sympathetic vasomotor tone, supplying continuous excitatory drive to spinal sympathetic preganglionic neurons. When RVLM activity becomes pathologically elevated through oxidative stress, neuroinflammation, or altered afferent input from higher centers of the network, the result is a sustained sympathetic bias expressed throughout the system. That is the neural substrate for what clinicians recognize as being stuck in fight or flight, and it is specific, measurable, and anatomically localized. The phrase names a level of drive in the medulla rather than a mood.

Then comes the model's sharp qualification, and it is the model's own step. Naming these structures is not the same as locating regulation inside them. What the network holds is a level of readiness that can be raised and lowered rather than a function stored at an address. The same circuit produces defense in one hour and digestion in the next with nothing structural having changed. A regulatory function is a tunable state rather than a fixed place, which is why an intervention that alters no anatomy can alter everything functional.

Say plainly where the reading goes past the source. Benarroch wrote clinical neurology. He described the network, named its four properties, and catalogued the disorders that break it, and he made no claim about tone as this library uses the word. The model sets his central autonomic network beside Sherrington's account of integration and Cannon's homeostasis, then states its own position in its own name. Each researcher established what their own work established, and the integration of their separate findings into one variable is the model's. The anatomy is his. The variable is ours.

WHAT THE RECORD SHOWS

The central autonomic network in seven dated findings

• 1993. Benarroch named the central autonomic network in Mayo Clinic Proceedings, volume 68, issue 10, pages 988 to 1001 (Benarroch 1993), gathering seven structures from insular cortex to ventrolateral medulla into one anatomy of internal regulation.
• Four properties. The 1993 abstract characterizes the network as reciprocal, parallel, state dependent, and neurochemically complex (Benarroch 1993). State dependence is the property that lets one circuit settle into a configuration and keep holding it.
• 1987. David Cechetto and Clifford Saper reported viscerotopic sensory representation in cortex and thalamus in the rat, Journal of Comparative Neurology, volume 262, pages 27 to 45 (Cechetto and Saper 1987). The organs have a cortical map in the same sense the skin does.
• 1992. Stephen Oppenheimer and colleagues stimulated the insular cortex in five epilepsy patients, Neurology, volume 42, pages 1727 to 1732, and found left and right responses biased in opposite directions at p less than 0.005 (Oppenheimer 1992).
• 1998. Benarroch and Elizabeth Stotz-Potter reported fatal familial insomnia with neuronal depletion in the mediodorsal and anteroventral thalamic nuclei alongside exaggerated sympathetic activation (Benarroch and Stotz-Potter 1998). Sleep, circadian rhythm, and cardiovascular control failed together.
• 2020. Resting state fMRI at 3 tesla in 22 older adults, mean age 68 plus or minus 6 years, tied left amygdala connectivity to baroreflex sensitivity with correlation coefficients running from 0.663 to 0.703 (Ding 2020).
• 2024. In 34 Human Connectome Project participants aged 22 to 36, the sympathetic index correlated positively and the parasympathetic index negatively across the same cortical and brainstem territory (Valenza 2024).