The Embodied Brain
The brain in a body
We ended the last chapter with a slogan: a patterned tube is a controllable tube. The neural tube is not a uniform sausage of cells; it is carved by regulatory genes into territories with addresses, and those addresses are the beginnings of structures that will do different jobs. That is where we left the brain — built, regionalized, ready.
Ready to do what, though? This is the question I want to take seriously in this chapter, because I think the standard answer is subtly misleading. If you open most textbooks at this point, the next move is a grand tour of the brain’s parts: here is the thalamus, here is the cerebellum, here are the four lobes, please memorize the gyri. We will take that tour — but in the next chapter, and organized by function rather than by name. Before we walk the rooms of the house, I want to deal with something that the room-by-room tour tends to skip over, and that I think is actually prior to it.
Here it is. A brain is a control system, and a control system is only as good as its connection to the thing it controls. A thermostat wired to nothing regulates nothing. The brain spends roughly a fifth of your entire resting energy budget — a staggering overhead for an organ that is about 2% of your body mass — and evolution does not subsidize expensive organs out of generosity. (Recall the framing from Chapter 1: the costly brain earns its keep by buying prediction, by forecasting a need before the error arrives rather than merely reacting once it has.) But you cannot predict what you cannot sense, and you cannot regulate what you cannot reach. So before the brain can be a predictive controller of the body, it has to be wired into the body — densely, bidirectionally, with dedicated channels carrying information up and commands down.
That wiring is the subject of this chapter. My claim — and I want to flag at the outset that the framing is a choice, an emphasis, not a discovery — is that the machinery usually treated as boring infrastructure (the peripheral nerves, the vagus, the blood supply, the barriers and the gaps in the barriers) is not infrastructure at all. It is the sensory and motor apparatus of a control system whose controlled variable is the internal state of the body itself. The brain is not a disembodied processor that happens to sit on top of a body and issue orders. It is a node inside a set of feedback loops that run out through the body and back. Most of what the brain regulates, it regulates through these loops, by sensing the body’s internal condition, integrating those signals, and acting on them.
There is a word for the sensing half of this — the perception of the body’s own internal physiological state. It is interoception, and it sits underneath this whole chapter [@craig2002; @critchley2013]. Hunger, thirst, nausea, breathlessness, the racing heart, the full bladder, the warmth of a fever, the malaise of an infection — these are not incidental feelings layered on top of cognition. They are reports from the viscera, and the brain’s reading of them is the substrate on which allostatic regulation is built [@sterling2012]. When Peter Sterling argues that the brain’s basic job is predictive regulation of the body, the interoceptive channels are how it gets the data to predict with.
A caveat I will return to, in the spirit of the last chapter: it is tempting to describe all this with the brain as a little commander “reading dispatches” and “issuing orders.” I will use that language because it is vivid. But hold it loosely. There is no homunculus at a desk. There are loops — afferent fibers carrying signals bodyward-to-brainward, efferent fibers carrying signals brainward-to-bodyward — and the regulation is the loop, not a decision made above it. With that warning posted, let us look at the loops.
This chapter uses directional language immediately, so here is the minimum vocabulary. We do not need the full anatomical tour yet (that is Chapter 4), but we do need to be able to point.
Afferent means carrying signals toward the central nervous system (body → brain); efferent means carrying them away from it (brain → body). A useful mnemonic: Afferent = Arriving, Efferent = Exiting.
Directional terms. Rostral (toward the beak/front) and caudal (toward the tail/back); dorsal (back) and ventral (belly); medial (toward the midline) and lateral (toward the edge). In the spinal cord, the dorsal side is sensory (signals arriving) and the ventral side is motor (signals exiting) — a division that, as we saw last chapter, traces straight back to the alar and basal plates of the neural tube.
The gross plan, in one line. The embryonic neural tube swells into three primary vesicles (forebrain, midbrain, hindbrain), which subdivide into five, which mature into the adult structures we will tour in Chapter 4. For now, two landmarks matter: the hindbrain, where the body’s sensory traffic first lands, and the hypothalamus (in the forebrain), where much of the body’s regulation is orchestrated.
I do not expect you to memorize this. Refer back when a direction word trips you up.
The nervous system has a peripheral half
We tend to say “the brain” and “the nervous system” as if they were the same thing. They are not. The central nervous system (CNS) is the brain and spinal cord — the part encased in bone. The peripheral nervous system (PNS) is everything else: the nerves threading out to muscles, organs, skin, and gut. The course is called The Human Brain, and I have already confessed to being conflicted about that title; here is another reason for the discomfort. You cannot understand what the central nervous system is for without the peripheral nervous system, because the PNS is the interface. It is where the control system touches the controlled.
The PNS divides into two arms, and the division is functional. The somatic nervous system handles the “external” relationship with the world: voluntary control of skeletal muscle (efferent) and sensation from the skin, joints, and muscles (afferent). When you decide to lift a cup, the command travels out through somatic motor fibers whose cell bodies sit in the ventral horn of the spinal cord, exit through the ventral roots, and innervate the muscle. When you feel the cup’s warmth, the signal travels in through somatic sensory fibers whose cell bodies sit in the dorsal root ganglia — little clusters of neuronal cell bodies just outside the spinal cord — and enter through the dorsal roots. Sensory in the back, motor in the belly: the alar/basal plate again, now grown up.
The autonomic nervous system (ANS) handles the “internal” relationship — the regulation of the viscera. This is the arm that matters most for our thesis, so it gets most of this section. But notice the shape of the whole thing first: the somatic system points outward at the world, the autonomic system points inward at the body, and both are loops with an afferent leg and an efferent leg. The brain’s relationship to your hand and the brain’s relationship to your heart are built from the same basic parts.
The autonomic nervous system, beyond “fight or flight”
Almost every introductory account of the autonomic nervous system reaches immediately for “fight or flight,” and I want to start by gently prying that phrase loose, because it has done more to mislead students than almost any other slogan in this part of the field.
Here is the standard picture, which is not wrong, just badly framed. The ANS has two major divisions that innervate the same organs with generally opposing effects. The sympathetic division is the accelerator: it speeds the heart, dilates the pupils and the airways, mobilizes glucose, and shunts blood toward muscle — the body configured for exertion. The parasympathetic division is the brake: it slows the heart, constricts the pupil, and promotes digestion — the body configured for maintenance. “Fight or flight” versus “rest and digest.” So far, so textbook.
The trouble with the slogan is that it casts the autonomic nervous system as an emergency system — something that switches on when a tiger appears and otherwise idles. That is almost the opposite of the truth. The autonomic nervous system is running right now, continuously, in both of us, with no tiger anywhere. It holds your blood pressure within a narrow band as you stand and sit, trims your heart rate to your metabolic demand breath by breath, manages your core temperature, your gut motility, your bladder, your blood glucose. The “opposition” of sympathetic and parasympathetic is not a toggle between calm and panic; it is the continuous push-pull of a regulator holding a controlled variable steady — and, when the system is working in allostatic rather than merely reactive mode, adjusting the set-point in anticipation of demand. Your heart rate begins to climb fractionally before you stand up, not after your blood pressure has already dropped. That is prediction, implemented in autonomic tone.
And the two divisions are not a simple seesaw. Plenty of coordinated behaviors require both arms active at once. Sexual arousal is the classic example — erection is largely parasympathetic, emission largely sympathetic — and you cannot describe it as either “fight or flight” or “rest and digest” without the description falling apart. The accelerator-and-brake metaphor is a fine first handhold, but a real driver uses both pedals together more often than you might think.
It is easy to picture the autonomic nervous system as purely efferent — orders going out to organs. That picture is half the system, and the less interesting half. The viscera report back constantly: baroreceptors in your arteries reporting blood pressure, chemoreceptors reporting blood gases, stretch receptors in the gut and bladder reporting fullness. Most of this never reaches consciousness, which is exactly why it is easy to forget it is there. But it is the afferent leg of the loop, and without it the efferent commands would be flying blind. Keep both legs in view. We are about to meet a single nerve that carries the body’s afferent traffic in extraordinary volume.
The enteric nervous system: a brain in the gut, with caveats
Before we get to that nerve, one autonomic subsystem deserves its own treatment, because it is genuinely strange and because it is the locus of some of the most overhyped claims in modern neuroscience. Lining the wall of your gut, from esophagus to rectum, is the enteric nervous system (ENS) — a mesh of something like 400 to 600 million neurons, more than the spinal cord contains, organized into two interconnected plexuses [@furness2012]. It is sometimes called the “second brain,” and for once the nickname is earned in a specific, defensible sense: the enteric nervous system can run the basic program of digestion — sensing the contents of the gut, coordinating the muscular waves of peristalsis, regulating secretion — without instruction from the brain at all. Sever its connection to the CNS and the gut keeps digesting. No other organ system has anything like this degree of local neural autonomy [@spencer2020]. It is, in a real sense, a peripheral nervous system that learned to think for itself.
A fact students enjoy, and which is true: a large majority of the body’s serotonin (often quoted as ~90%) and a substantial fraction of its dopamine are found in the gut, not the brain. This is real. What it does not license is the leap you will see in popular writing — that because “the happy chemical” is mostly in your gut, your gut controls your mood. Gut serotonin and brain serotonin are, for most purposes, separate pools; serotonin does not freely cross the blood–brain barrier, and the enteric serotonin is mostly doing local jobs (driving motility, signaling to sensory fibers) that have nothing to do with how you feel on a Tuesday. The fact is true and the inference is mostly false, and I want you to get used to holding those two things apart, because the gut–brain literature is full of true facts attached to overreaching inferences.
So let me try to draw the honest line, because this is precisely the kind of live, half-settled frontier where your instructor’s job is not to give you a tidy answer but to show you where the solid ground ends.
What is solid. The gut and the brain communicate, bidirectionally, through several channels we will meet in this chapter — chiefly the vagus nerve, but also hormonal and immune signals. Enteric sensory cells, including specialized enteroendocrine cells in the gut lining, detect the chemical state of the gut’s contents and relay that information to the brain. This much is not in doubt.
What is suggestive but unsettled. The claim that the community of microbes living in your gut — the microbiota — influences brain function and behavior. There is, at this point, a genuinely large body of evidence that the gut microbiota and the brain are connected [@cryan2019; @morais2021]. In rodents, the evidence for causation is in places quite strong: germ-free mice raised with no microbiota at all show altered stress responses and altered brain development, and — this is the striking one — transferring the gut microbiota from a depressed human into a microbiota-free rodent can transfer depressive-like behavior to the animal. That is a causal experiment, and it points somewhere real.
What is genuinely uncertain — and where the hype lives. Whether any of this scales to human mood, cognition, or psychiatric disease in the strong way the supplement aisle implies. The honest summary, and here I am leaning on a pointed review by Walter and colleagues with a title I admire — “Establishing or exaggerating causality for the gut microbiome” — is that the human evidence is overwhelmingly correlational [@walter2020]. We see that the microbiota differs between depressed and non-depressed people, between people with and without various conditions. We do not, for the most part, know which way the arrow points. Does dysbiosis contribute to depression, or does depression (via diet, stress, and altered gut physiology) reshape the microbiota? Cross-sectional differences cannot tell us, and the rodent models — for all their power — translate to humans poorly and inconsistently. Probiotic trials in humans have shown, at best, modest antidepressant effects, and the field has a replication problem it is still working through.
My own view, for whatever a neuroscientist’s view on a fast-moving field is worth: the microbiota–gut–brain axis is real, it is important, and it is one of the most oversold topics in contemporary neuroscience. Both halves of that sentence are true at once. Be excited and be skeptical. The excitement and the skepticism are not in tension; they are what taking a frontier seriously actually feels like.
Here is a case that shows both how seriously to take gut–brain communication and how carefully to state what we know.
There is a hypothesis, originally from neuropathologist Heiko Braak, that some Parkinson’s disease may begin in the gut and ascend to the brain along the vagus nerve [@braak2003]. The reasoning starts from an autopsy observation: the misfolded protein that defines Parkinson’s pathology, α-synuclein, often appears early in the dorsal motor nucleus of the vagus and in the nerve plexuses of the gut — earlier, in some patients, than in the substantia nigra, the midbrain region whose dopamine-neuron loss causes the disease’s motor symptoms. Braak proposed that pathological α-synuclein might propagate, prion-like, from the enteric nervous system up the vagus to the brainstem and onward.
For years this was an intriguing pattern in postmortem tissue and not much more. Then, in 2019, Kim and colleagues did the experiment in mice: they injected pathological α-synuclein into the gut wall and watched it spread — first to the dorsal motor nucleus of the vagus, then up through the brainstem, and eventually to the substantia nigra, accompanied by dopamine-neuron loss and Parkinson-like motor and non-motor deficits. Crucially, cutting the vagus nerve (truncal vagotomy) blocked the spread [@kim2019]. And there is supporting human epidemiology: people who underwent truncal vagotomy for ulcer disease decades ago appear to have a somewhat reduced later risk of Parkinson’s.
This is a striking convergence of pathology, a causal animal model, and human epidemiology, all pointing along the anatomical route this chapter is about. So why am I filing it under “caution” rather than “established”? Because the strong claim — that Parkinson’s is, in general, a gut-origin disease — remains unproven in humans [@horsager2020]. The propagation looks bidirectional: there appear to be “body-first” cases that fit Braak’s gut-to-brain route and “brain-first” cases that do not, and we cannot yet say what fraction of disease each accounts for. The vagus is clearly a highway; whether it is the origin is unsettled.
Two lessons. First, this is what a maturing frontier looks like — not a binary “true/false,” but a hypothesis being decomposed into the cases where it holds and the cases where it does not. Second: be wary of the version of this you will encounter in popular health writing, which tends to flatten “Parkinson’s may have a body-first subtype that involves the vagus” into “Parkinson’s starts in your gut, so fix your gut.” The careful claim is exciting enough. It does not need the inflation. (You will also see the gut floated as the origin of ALS and autism. The evidence there is far thinner than for Parkinson’s, and I would not lump them together; the Parkinson’s case is in a different evidential class.)
Measuring the autonomic state: psychophysiology
One more thing belongs with the autonomic nervous system, and it is methodological. Because the autonomic nervous system drives peripheral organs in measurable ways, we can read out a person’s autonomic state from the outside, non-invasively. This is the basis of psychophysiology, and it is worth knowing both because the measures recur throughout the course and because they make the abstract idea of “autonomic tone” concrete.
The general logic is that sympathetic activation produces a family of bodily changes — and we can instrument each of them. The common measures:
- Pupillometry. The pupil dilates with arousal, attention, and cognitive load (and constricts with parasympathetic dominance). Tracking pupil diameter gives a surprisingly sensitive, fast index of mental effort.
- Electrodermal activity — the skin conductance response (SCR), historically the galvanic skin response (GSR). Sweat glands are under sympathetic control; sweaty skin conducts a small electrical current better than dry skin. A spike in skin conductance is a fairly direct readout of a sympathetic burst.
- Heart rate and heart rate variability (HRV). Heart rate rises with sympathetic drive; variability between successive beats tends to fall under stress and rise during relaxation. HRV in particular is heavily shaped by parasympathetic (vagal) influence on the heart, which is one reason it has become a popular — sometimes over-interpreted — index of “vagal tone.”
- Respiration. Breathing rate and depth shift with arousal.
You have met this technology in one famous and scientifically dubious application: the polygraph “lie detector” is essentially a multi-channel psychophysiology rig (skin conductance, heart rate, respiration) recording arousal during questioning. The reason polygraphy is unreliable is instructive and worth stating plainly: these measures index arousal, not deception. An innocent person terrified of a false accusation and a guilty person calmly lying can produce overlapping traces, because the autonomic nervous system does not know or care whether you are telling the truth — it responds to how threatened you feel. The instruments are real and useful; the inference from arousal to lying is the weak link. (Notice that this is the same lesson as the serotonin example and the microbiome example: a real signal, a tempting inference, and a gap between them. I promise I am not arranging these on purpose. The gap just keeps showing up, which is itself the point.)
The vagus and the afferent flood
I have mentioned the vagus several times now, deferring it, because it deserves to be the hinge of this chapter. If you take away one structural fact about how the brain is wired into the body, take this one.
The vagus nerve is the tenth cranial nerve. Its name comes from the Latin for “wandering,” and it earns it: from its origins in the medulla it wanders down through the neck and thorax into the abdomen, branching to innervate the larynx, heart, lungs, and most of the gastrointestinal tract. It is the longest and most widely distributed of the cranial nerves, and it is the principal conduit of the parasympathetic nervous system. So far this sounds like an efferent story — the brain’s long arm reaching down to slow the heart and stimulate digestion. And the vagus does carry those commands.
But here is the fact that reorganizes the whole picture, and that I find genuinely arresting every time I teach it. The vagus is overwhelmingly a sensory nerve. Roughly 80% of its fibers — the figure is sometimes given as 80 to 90%, and the exact number has a long pedigree in the physiology literature — are afferent, carrying information from the body up to the brain [@berthoud2000; @prescott2022]. Only about 20% are efferent. The nerve we describe as the brain’s great parasympathetic output is, in the brute accounting of its fibers, mostly an input. The body reports up far more than the brain commands down.
Sit with that, because it inverts the intuitive model. We imagine the brain as the body’s commander, issuing a high volume of orders to a passive periphery. The vagus says the reverse: the dominant flow of traffic, by a factor of four, is the periphery reporting its state to a brain whose main job is to listen and adjust. The heart, the lungs, the gut, the great vessels — all of them are continuously streaming status reports upward: blood pressure, oxygen and carbon dioxide levels, gut distension, the chemical contents of a meal, the presence of inflammation. The brain is, more than anything, an organ that is being constantly informed about the body. This is interoception with an anatomical address.
Where does all this afferent traffic land? Overwhelmingly at one structure: the nucleus of the solitary tract (NTS; also called the nucleus tractus solitarii), a column of cells in the medulla. About 80% of vagal afferent fibers terminate here, making the NTS the brain’s primary visceral-sensory relay — the great clearinghouse where the body’s reports first arrive in the central nervous system [@saper2002]. It is the visceral analogue of the way the thalamus relays the “external” senses, and it is far less famous than it deserves to be.
The NTS does not keep this information; it distributes it, and the projection targets are exactly the structures you would design a regulatory system to inform. It projects to the parabrachial nucleus (a major waystation in the pons that integrates visceral and other signals), to the hypothalamus — the master regulator of homeostasis, which we will return to repeatedly — and onward to limbic and forebrain regions involved in the feeling and the behavioral consequences of bodily states [@saper2002]. So the path is: body → vagal afferents → NTS → parabrachial nucleus / hypothalamus → integration and response. That is the afferent half of the loop, drawn end to end, and the hypothalamus sitting at its terminus is not an accident. The body’s status reports flow to the structure whose job is to keep the body in balance.
In keeping with this book’s cross-species habit: the NTS is not a clever human invention. It is a deeply conserved structure with recognizable homologues across the vertebrates [@nts_vertebrates2026]. The lamprey has one; the fish has a prominent one (in some fish the vagal and facial sensory lobes are among the largest structures in the brain, because taste and visceral sensation matter enormously to an animal feeding in water). A bird’s NTS contains a topographic map of the viscera. This is what we should expect. Sensing the internal state of the body and adjusting physiology to match is not a luxury that arrived with mammalian sophistication; it is among the oldest jobs a vertebrate brain has, because it is among the most necessary. An animal that cannot monitor and regulate its own viscera does not live long enough to evolve anything fancier. The elaborate human forebrain is a late and expensive addition built on top of this ancient body-regulating core — not a replacement for it. We are, here as elsewhere, a modified vertebrate, not a different kind of thing.
There is also an efferent immunological trick worth previewing here, because it shows the loop closing. The vagus does not only report inflammation upward; through its efferent fibers it can suppress inflammation downward, via what Kevin Tracey named the cholinergic anti-inflammatory pathway or “inflammatory reflex” [@tracey2002; @borovikova2000]. Efferent vagal activity, through acetylcholine, can dampen the production of inflammatory cytokines by immune cells in the body. So the vagus is a true two-way regulatory loop for the immune system specifically: afferents sense peripheral inflammation, the brainstem integrates it, and efferents can turn the inflammation down. Hold onto this; it links directly to the immune section later, and it is part of why “just stimulate the vagus” has become such an active (and, predictably, sometimes oversold) therapeutic idea.
Breaching the barrier: the circumventricular organs
The vagus is the brain’s neural line to the body — fast, wired, point-to-point. But the body also speaks to the brain in a slower, broadcast medium: the bloodstream, carrying hormones, metabolites, and immune signals. And here we hit an apparent problem, one that sets up the next two sections.
The brain, almost everywhere, walls itself off from the blood. The blood–brain barrier is the name for this: the endothelial cells lining the brain’s capillaries are stitched together with tight junctions, and unlike the leaky capillaries elsewhere in the body, they do not let molecules slip freely from blood into brain tissue [@daneman2015]. This is, on the whole, a feature and not a bug — the brain is a delicate electrochemical instrument, and you do not want every fluctuation in blood chemistry, every ingested toxin, every circulating hormone sloshing directly onto your neurons. The barrier keeps the brain’s internal environment tightly controlled.
But now consider the regulatory problem this creates. If the brain is supposed to keep the body in homeostasis, it needs to know the body’s blood chemistry — the osmolality (saltiness) of the blood, its glucose, its hormone levels, whether there are toxins or inflammatory signals present. How can the master regulator monitor the blood if it has deliberately walled itself off from the blood?
The answer is elegant: the brain leaves itself a set of windows. The circumventricular organs (CVOs) are small midline structures bordering the ventricles where the blood–brain barrier is deliberately absent — the capillaries there are fenestrated (leaky, windowed) rather than sealed [@miyata2015; @mckinley2003]. At these specific, limited sites, and essentially nowhere else, neurons are directly exposed to the chemical environment of the blood. They are sampling ports — places where the brain pokes a sensor through its own barrier to taste the blood directly.
It helps to split the CVOs into two functional groups, because they face in opposite directions:
- Sensory CVOs — chiefly the subfornical organ, the organum vasculosum of the lamina terminalis (OVLT), and the area postrema. These contain neurons exposed to the blood that sense its contents and project that information inward to the rest of the brain. The subfornical organ and OVLT monitor blood osmolality and circulating signals related to fluid balance and blood pressure — they are central to how the brain knows you are dehydrated and generates thirst. The area postrema, in the floor of the fourth ventricle, is a chemoreceptor trigger zone: it samples the blood for circulating toxins, and when it detects them it can trigger vomiting. (This is why the area postrema is sometimes called the “vomiting center,” and why a number of drugs cause nausea as a side effect — they are being detected at this deliberately unguarded window.) Tellingly, the area postrema projects directly to the NTS, knitting the humoral sensing of the CVOs into the same visceral-relay machinery the vagus feeds. The two channels — neural and bloodborne — converge.
- Secretory CVOs — chiefly the median eminence and the neurohypophysis (posterior pituitary), plus the pineal gland. These face the other way: rather than sensing the blood, they secrete into it. The median eminence is where hypothalamic releasing hormones are dumped into the special portal blood supply that carries them to the pituitary; the neurohypophysis releases oxytocin and vasopressin into the general circulation; the pineal secretes melatonin. These are the brain’s output ports to the bloodstream.
So the circumventricular organs are the missing mechanistic link between two ideas that the room-by-room anatomy tour leaves disconnected: the brain’s “plumbing” (its blood supply) and the brain’s sensing. The blood is not just delivering fuel to the brain; at these windows, the brain is reading the blood as a sensory surface. And once you see that the brain has dedicated machinery for sampling bloodborne signals, the door opens to a question the older textbooks barely asked: what is in the blood that the brain wants to read? Hormones and salt and glucose, yes. But also — and this is where we are heading — the molecular signals of the immune system.
If you go looking, you will find the subcommissural organ listed as a circumventricular organ in some sources and pointedly excluded in others, because it lacks the fenestrated capillaries that define the group. The choroid plexus, which manufactures cerebrospinal fluid, sits in a similar gray zone — it has the leaky vasculature but not the neurons. I mention this not because you need to adjudicate it, but as one more instance of a theme from Chapter 2: every nomenclature has edge cases, and the edge cases are usually where something interesting is happening rather than where someone made a mistake. “Carve nature at its joints” is good advice right up until you discover that nature has more joints than your scheme anticipated, and some of them are ragged.
The vascular system: supply and signal
We have arrived, by the logic of sensing, at the brain’s blood supply — and I want to treat it in a way that the usual “here is the Circle of Willis, memorize the arteries” presentation does not. The vascular system has two jobs in this chapter, and the second one is usually invisible.
The first job: fuel
The first job is the obvious one. The brain runs on glucose and oxygen, both delivered by blood, and it runs on them with almost no reserve. Recall the budget from earlier chapters: the brain is about 2% of body mass but consumes roughly 20% of the body’s resting energy — on the order of several hundred kilocalories a day just to keep ~86 billion neurons supplied. (Herculano-Houzel’s figure of about 516 kcal/day for the human complement of neurons is the number we have been carrying.) Now add a crucial fact: unlike muscle, which stockpiles energy as glycogen, the brain stores almost none. It is a high-draw device living hand to mouth, utterly dependent on continuous delivery.
This is why interruption of blood flow to the brain is catastrophic on a timescale of seconds. Cut the supply and consciousness is lost within seconds; let the interruption persist and the deprived tissue dies. This is a stroke. In an ischemic stroke, the common kind, an embolus — typically a fragment of atherosclerotic plaque — travels along an artery until it lodges where the vessel narrows too far for it to pass, blocking flow to everything downstream. The tissue served by that vessel is starved of glucose and oxygen and becomes infarcted: it dies, and the functions it supported are lost — a paralysis, a speech deficit, a field of blindness, depending on what the blocked artery fed.
Stroke is a tragedy for the patient and, historically, one of the most important tools neuroscience has had. Because a stroke knocks out a specific, vascularly-defined territory and produces a specific loss of function, the correspondence between the two has been a primary source of our knowledge of what brain regions do. Much of classical localization — including the language regions we will study later — was first charted by correlating the site of a patient’s lesion (often a stroke) with the function they lost. The brain’s blood supply is, among other things, a map of how the brain can fail, and how it fails has taught us how it works.
The supply itself, in brief (we will not belabor the catalogue, since the functional tour is Chapter 4): blood reaches the brain by two routes. The paired internal carotid arteries ascend through the neck and feed mainly the anterior brain — the “anterior circulation.” The paired vertebral arteries ascend along the spine and fuse into the single basilar artery, supplying the brainstem, cerebellum, and the posterior brain including the occipital lobe — the “posterior circulation.” These two systems meet in a ring at the base of the brain, the Circle of Willis, from which the major cerebral arteries branch. The ring is a piece of engineering redundancy: in principle, if one feeder is compromised, the ring allows flow to cross over and compensate. In practice the compensation is partial and varies between individuals, but the design intent — a shared reservoir rather than independent end-arteries — is clear.
The second job: signal
Now the part that earns the vascular system its place in this chapter rather than the next. The blood is not only a fuel line. It is a communication channel, and the brain both reads it and writes to it.
We have already seen the reading machinery — the circumventricular organs, where the brain samples circulating hormones, osmolality, toxins, and (the next section) immune signals. And we have seen the writing machinery — the secretory CVOs, where the brain dumps hormones into the blood to act at a distance. Put those together and the vascular system becomes a third great input/output system of the brain, parallel to the neural one. The vagus and the spinal nerves are the brain’s wired connection to the body — fast, specific, addressed. The bloodstream is the brain’s wireless connection — slower, diffuse, broadcast. Both carry the body’s state up and the brain’s regulation down. A control system wired into its body by only nerves would be missing half the conversation; the body’s chemistry — its hormones, its fuels, its inflammatory state — is carried in blood, and the brain has evolved the apparatus to participate in that chemical conversation in both directions.
There is a beautiful consequence of the brain’s hand-to-mouth energy economy, and it is the reason we can image the living, working human brain at all.
When a patch of brain becomes more active, it demands more fuel, and local blood flow to that patch increases to meet the demand — a phenomenon called functional hyperemia, first noted by Roy and Sherrington in 1890 [@roy1890]. The local matching of blood flow to neural activity is called neurovascular coupling, and it is mediated by a little consortium of cells — neurons, astrocytes, and the vessels themselves — sometimes called the neurovascular unit [@iadecola2017]. The upshot is that where the brain is working, the blood follows. And that means if you can image local blood flow or blood oxygenation, you can infer where neural activity is occurring.
This is exactly what the two great functional-imaging methods do. PET (positron emission tomography) and, more importantly today, functional MRI (fMRI) both work by, in effect, following the energy — fMRI by detecting the blood-oxygenation changes (the “BOLD” signal) that accompany the hemodynamic response to neural activity. When you read that a study “showed activation” in some brain region, what was almost always measured was not neural firing directly but the blood-flow change that neurovascular coupling produces a couple of seconds later. We will take this up in detail when we cover imaging methods. For now, notice the double duty the vascular system is doing in this chapter: it is both an instance of body–brain signaling and the tool that lets us watch the brain signal.
I should add a note of the honest-uncertainty kind, because this is a place where the textbook story is cleaner than the science. The cellular mechanism of neurovascular coupling — in particular, exactly how much of the signal runs through astrocytes versus directly through neurons, and whether the increased flow is really “for” delivering oxygen at all — is still genuinely debated [@howarth2014; @drew2019]. There is a respectable line of argument that the flow increase overshoots what the metabolic demand strictly requires, which would mean we do not fully understand what functional hyperemia is for. This matters practically: every fMRI result rests on assumptions about what the blood-flow signal means, and those assumptions are still being worked out. It is worth knowing that one of neuroscience’s most-used tools sits on a foundation that is still, in part, under construction.
The immune–brain axis
We come now to the part of this chapter I most wanted to write and was most worried about getting wrong — so let me tell you up front how I am going to handle it. This is the newest material in the book, much of it from the last decade, some of it from the last few years. Fast-moving fields are exciting precisely because they are unsettled, and unsettled fields are where it is easiest to overstate. So I am going to build this section on a solid anchor — a piece of immune–brain biology that has been well established for decades — and then walk outward toward the frontier, flagging clearly where the ground gets softer. When we reach the genuinely new and exciting findings at the end, you should hear them as promising and provisional, not as settled fact. That is not me hedging. That is the actual epistemic status of the work, and learning to track it is part of the training.
For most of the twentieth century, the brain was considered immune-privileged — walled off from the immune system, sitting behind the blood–brain barrier, lacking lymphatic drainage, essentially a separate country with closed borders. The dogma had real evidence behind it. It was also, we now know, substantially wrong, or at least far too strong. The brain and the immune system are in constant communication, in both directions, and the channels of that communication are the subject of this section. Let me organize them as three routes — neural, humoral, and what I will loosely call the border-tissue route — moving roughly from oldest-and-most-solid to newest-and-most-provisional.
The solid anchor: sickness behavior
Start with something you have experienced. When you come down with the flu, you do not merely have a fever and aches. Your behavior changes, in a characteristic and coordinated way: you lose your appetite, you withdraw from social contact, you become sleepy and listless, you lose interest in things that normally motivate you. We tend to think of this as just “feeling bad” — an unfortunate by-product of being sick. It is not a by-product. It is a program, organized by the brain, and it has a name: sickness behavior [@hart1988; @dantzer2008].
Here is the chain of events, which is well established. When you are infected, cells of your innate immune system release signaling molecules called pro-inflammatory cytokines — interleukin-1β, TNF-α, interleukin-6, and others. These cytokines are the immune system’s local messengers, coordinating the inflammatory response in the body. But the brain detects them, and in response it orchestrates the whole suite of sickness behaviors [@konsman2002]. The lethargy, the anorexia, the social withdrawal, the fever itself — these are brain-driven outputs, triggered by the brain’s reading of immune signals.
And — this is the conceptual payoff, the reason this section sits where it does in the book — sickness behavior is adaptive. It is not the infection disabling you; it is your brain reallocating your resources to fight the infection. Fever is metabolically expensive and is run deliberately because many pathogens replicate poorly when hot. Anorexia and lethargy conserve the energy that fever and immune activation demand, and reduce your exposure to new pathogens and predators while you are compromised. Social withdrawal may reduce transmission to kin. The whole package is a coordinated, evolved, brain-mediated shift in physiology and behavior in the service of survival. Which is to say: sickness behavior is allostasis — predictive reallocation of the body’s resources in anticipation of need — implemented in response to an immune challenge. The immune system reports a threat; the brain reconfigures the entire organism to meet it. This is the thesis of the chapter, written in the language of the immune system: the body signals its state, the brain integrates the signal, and the brain acts back on the body. The control loop runs through the immune system just as surely as it runs through the heart and the gut.
Now, how do the cytokines reach the brain? This is where the three routes appear, and notice that the first two are channels we have already built in this chapter:
- The neural route. Vagal afferents detect cytokines in the body’s tissues and relay the inflammatory signal up to the NTS — the same fast, wired pathway we traced earlier. This is the rapid arm: the brain learns of peripheral inflammation through its sensory nerves, the same way it learns of a full stomach or a drop in blood pressure. (And recall the efferent counterpart, the cholinergic anti-inflammatory reflex: having sensed inflammation, the brain can signal back down the vagus to restrain it. The loop closes.)
- The humoral route. Cytokines circulating in the blood act at the circumventricular organs — the windows in the barrier we just met — and drive the local production of further signals (such as prostaglandins) that propagate the inflammatory message into the brain proper. This is the slower arm, and it is exactly the “reading the blood as a sensory surface” idea from the CVO section, now with immune molecules as the thing being read.
So sickness behavior is not some exotic new system. It is the afferent/efferent machinery of this entire chapter — vagus, NTS, CVOs, hypothalamus — being used to sense and respond to the immune system. That is why I can call it solid: the routes were independently established, the cytokines are identified, the behavioral program is reproducible, and the adaptive logic is coherent. This is forty years of work, and it holds.
There is a sober and important extension, which I will state carefully because it shades from solid toward active research. If acute immune signaling to the brain produces adaptive sickness behavior, what happens when immune signaling becomes chronic — in persistent infection, autoimmune disease, chronic stress, or the low-grade inflammation associated with obesity and aging? The hypothesis, associated especially with Dantzer and colleagues, is that sustained inflammatory signaling to the brain can tip sickness behavior over into something that looks like clinical depression — and that inflammation may therefore be one contributing pathway to depression in some people [@dantzer2008]. The overlap between sickness behavior and the symptoms of depression (anhedonia, fatigue, social withdrawal, appetite and sleep disruption) is hard to ignore. There is supporting evidence: patients given inflammatory cytokines as therapy (interferon-α for hepatitis or cancer) develop depression at strikingly high rates, and meta-analyses find elevated inflammatory markers in depressed populations. I find this genuinely persuasive as one pathway among several — and I want to be careful with exactly that qualification. Inflammation is not “the cause of depression”; depression is heterogeneous, and most depressed people are not detectably inflamed. The honest claim is that there appears to be an inflammation-linked route into depressive states in some individuals, which is both scientifically important and a long way from the headline “depression is an inflammatory disease.” Same discipline as before: take the real finding, refuse the inflation.
The frontier: the brain’s borders are immunologically busy
Now we step out toward the genuinely new, and the tone shifts with the evidence. Everything from here is real, published, and exciting; much of it is also recent, derived substantially from rodents, and still being worked out in humans. Calibrate accordingly.
The old “immune privilege” dogma rested partly on a specific anatomical claim: that the brain, unlike every other organ, has no lymphatic drainage — no network of lymphatic vessels to carry fluid, waste, and immune cells out to the lymph nodes where immune responses are organized. In 2015, two groups independently overturned this. Louveau and colleagues, and Aspelund and colleagues, described meningeal lymphatic vessels — genuine lymphatic vessels running in the meninges, the membranes surrounding the brain, draining cerebrospinal fluid and its contents (including immune cells) to the deep cervical lymph nodes in the neck [@louveau2015; @aspelund2015]. The brain, it turns out, does have a lymphatic system; we had simply missed it, hidden in the dura.
I want to flag the word “rediscovery,” because the history here is itself a lesson. There were nineteenth-century anatomical reports of lymphatics around the brain that the twentieth century forgot or dismissed. The 2015 papers are sometimes written up as a “(re)discovery” for this reason. That a major anatomical feature of the human brain could be described, lost, and re-found two centuries later should be humbling — and it should make you cautious about treating any current “the brain definitely lacks X” claim as final. Absence of evidence has a way of turning into evidence of absence prematurely.
What has happened since 2015 is a fast-growing field, and here is roughly where it stands. The meningeal lymphatics have been shown to matter functionally: impairing them in mice slows the clearance of macromolecules from the brain and worsens cognitive performance, and their function declines with aging — with suggestive links to the accumulation of the amyloid that characterizes Alzheimer’s disease [@damesquita2018]. Critically for the “is this just a mouse thing?” worry, the vessels have been visualized in living humans and non-human primates by MRI, so their existence in our own brains is not in doubt [@absinta2017]. What remains genuinely uncertain is the magnitude of their role in human disease — whether meningeal-lymphatic dysfunction is a cause, a consequence, or a bystander in Alzheimer’s and other conditions. The existence is established; the clinical significance is a live question.
This connects to a parallel and equally new idea, the glymphatic system — a proposed route by which cerebrospinal fluid flushes through the brain tissue along the spaces around blood vessels, clearing metabolic waste (including amyloid), with the flushing apparently enhanced during sleep [@iliff2012]. The meningeal lymphatics would then be the downstream drain that carries this fluid out of the skull. I will flag plainly that the glymphatic model, while influential and attractive, is more contested than the meningeal-lymphatic anatomy — the existence of the vessels is settled in a way that the details of the bulk-flow clearance mechanism are not. But the broad picture emerging is striking: the brain is not the sealed, immune-isolated organ the textbooks described. It has drains. It clears waste through them, perhaps especially while you sleep. And those drains empty into the immune system’s filtration network.
The frontier, continued: the skull as an immune organ
Here is the most recent and, to me, the most surprising thread — and it is the one the brain–body framing of this chapter makes sense of, where a nomenclature tour would leave it stranded as a curiosity.
The bones of the skull are not inert armor. Like other large bones, they contain bone marrow — the tissue that manufactures blood cells, including the white cells of the immune system. For a long time the reasonable assumption was that immune cells responding to a brain injury or infection arrived from the general circulation, summoned from marrow throughout the body. But in 2018, Herisson and colleagues found something unexpected: there are direct microscopic vascular channels running through the inner table of the skull, connecting the skull’s bone marrow straight to the meninges [@herisson2018]. In mouse models of stroke and meningitis, immune cells (neutrophils) preferentially took this shortcut — emerging from the skull marrow directly into the inflamed brain coverings, rather than traveling the long way around through the blood. Roughly twice as many of the responding neutrophils came from the adjacent skull as from a distant bone. And — again addressing the rodent worry — analogous channels have been found in human skull samples, though whether they function the same way in us is not yet established.
The field has moved quickly since. Skull and vertebral marrow are now understood to be specialized local reservoirs supplying immune cells to the meninges and the brain’s borders [@cugurra2021]. And the traffic appears to run both ways: cerebrospinal fluid has been shown to access the skull marrow through the same channels, carrying signals from the brain to the marrow that can shape immune-cell production [@mazzitelli2022]. Read that again in the light of this chapter’s thesis: a bidirectional channel, afferent and efferent, between the brain’s fluid and the factory that builds its immune cells. The brain can inform the marrow, and the marrow can supply the brain. It is the afferent/efferent loop again, in a place nobody thought to look until a few years ago.
There is one more turn, and it is the specific point I most wanted to reach, because it ties the immune frontier back to the autonomic nervous system where this chapter began. Bone marrow is not merely near nerves; it is innervated — and that innervation regulates the immune system at its source. Sympathetic nerve fibers run into the marrow alongside its blood vessels, and they control the release of hematopoietic stem cells and their progeny into the circulation. The control is exquisitely timed: noradrenaline released from sympathetic terminals acts on the marrow’s stromal cells in a circadian rhythm, so that the production and release of blood cells oscillates over the day [@katayama2006; @mendezferrer2008]. (Pain-sensing nerve fibers innervate the marrow too, and modulate this mobilization as well [@gao2021].) The nervous system, in other words, has its hand directly on the lever of blood-cell production — the autonomic efferent arm reaching all the way down to regulate the immune system’s manufacturing.
In keeping with our cross-species, comparative habit, here is a detail I love: this circadian control of blood-cell release exists in humans too, but it is phase-shifted relative to nocturnal rodents — our stem-cell mobilization peaks toward the evening, theirs toward their active period. The mechanism is conserved; the timing is tuned to each species’ daily schedule. That is exactly the signature of an old, shared regulatory system adjusted at the margins by each lineage’s way of life — the same lesson the conserved Bauplan taught us in Chapter 2, now playing out in the dialogue between brain and immune system.
So the picture that has assembled itself, in just the last decade or so, is this. The brain senses the immune system (cytokines, via vagus and CVOs) and runs adaptive programs in response (sickness behavior). The brain’s borders have drains (meningeal lymphatics) that empty into the immune system’s network and may matter for clearing the brain’s waste. The skull’s marrow is a local immune depot wired directly into the brain’s coverings by physical channels carrying cells one way and signals the other. And the brain’s autonomic output reaches all the way to the marrow to regulate, on a daily clock, the very production of immune cells. The immune system is not a separate country with closed borders. It is one more body system that the embodied brain senses, integrates, and regulates — through dedicated afferent and efferent channels, exactly as it regulates the heart and the gut. That is the thesis, and the immune system may be its most dramatic vindication.
Let me end this section honestly. Almost everything in these last two subsections is younger than the students reading it, and the history of “the brain definitely lacks X” should make us humble about how this will read in twenty years. Some of what I have described will be deepened; some will be revised; the human relevance of much of it is still being established. I have tried to mark the solid anchor (sickness behavior) clearly and to label the frontier as a frontier. If you take away one durable thing, let it be the organizing idea — that the immune system is wired into the brain as another body-interface, sensed and regulated like the rest — rather than any single one of the newest findings, which are exciting precisely because we do not yet know which of them will last.
Coda: the machinery prediction runs on
Step back and look at what we have built. We started with a claim that might have sounded like a slogan: the brain is not a disembodied controller but a node in feedback loops that run out through the body and back. We have now seen the actual hardware of those loops.
The body reports its internal state upward through dedicated afferent channels — the vagus above all, carrying its afferent flood to the nucleus of the solitary tract; the spinal visceral afferents; the windows of the circumventricular organs where the brain tastes the blood directly; the cytokine signals of the immune system read through those same channels. These reports converge on integrating structures, the hypothalamus chief among them. And the brain acts back on the body through dedicated efferent channels — the two arms of the autonomic nervous system continuously tuning the viscera, the hormonal outputs secreted into the blood, the cholinergic reflex restraining inflammation, the sympathetic fibers reaching into the bone marrow to clock the immune system’s output. Sense the body’s state; integrate it; act on it; sense the consequences. That loop, in all its parallel channels, is the embodied brain.
And now recall why this matters for the spine of the whole unit. The brain is expensive because it buys prediction — because allostasis, the anticipatory regulation of the body, is worth more than mere reactive homeostasis. But prediction requires data. You cannot forecast a need you cannot sense, and you cannot pre-empt an error in a variable you cannot measure. The afferent/efferent machinery of this chapter is precisely the apparatus that makes allostatic prediction possible: it is how the brain gets the continuous, high-bandwidth stream of bodily data that any predictive regulator must have, and how it reaches back to act before the error arrives. The interoceptive channels are not a side-system. They are the sensory and motor surface of the body-regulating control system that the expensive brain exists to be. The prediction the brain is buying has to run on something, and this is the something.
Which brings us, finally, to the structures themselves. We have spent this chapter on how the brain talks to the body — the channels in and out. We have deliberately deferred the brain’s internal organization: what the hypothalamus actually contains, how the cortex is arranged, what the cerebellum and basal ganglia and amygdala are doing, how the great fiber tracts wire the regions to each other. That is the subject of the next chapter, Chapter 4 — an anatomical tour of the human brain, organized by function. Now that we understand the brain as an embodied controller — wired into the body it regulates — we are ready to tour the machinery that does the regulating, and to ask of each structure not merely “what is it called?” but “what is it for?” That second question is the only one worth asking, and it is the one we have been preparing to answer all along.
In the spirit of the last two chapters, here is an honest ledger for this one.
What we’re sure of.
- The brain is densely, bidirectionally wired into the body. The afferent/efferent organization is not in doubt; the somatic and autonomic divisions and their basic anatomy are textbook-solid.
- The vagus is predominantly an afferent nerve (~80% of fibers) and its afferents land chiefly in the nucleus of the solitary tract, which relays visceral information to the parabrachial nucleus, hypothalamus, and beyond. Solid.
- The circumventricular organs are real windows in the blood–brain barrier where the brain samples (or secretes into) the blood. The sensory/secretory division and the major members are well established.
- The brain’s energy economy (≈20% of resting metabolism, negligible storage) and the consequent catastrophe of stroke are firmly established, as is neurovascular coupling as a phenomenon and the basis of fMRI.
- Sickness behavior is a real, adaptive, brain-organized response to immune signaling, communicated by well-characterized neural (vagal) and humoral (CVO) routes. Forty years of work; it holds.
What we’re not sure of (and where the hype lives).
- The microbiota–gut–brain axis is real but oversold. Rodent causal evidence is strong; human causal evidence is largely correlational, and the translation is poor. Be excited and skeptical at once.
- The Braak “gut-first” hypothesis for Parkinson’s has real support (a causal mouse model, vagotomy epidemiology) but is unproven as a general human account; the disease appears to have body-first and brain-first subtypes in unknown proportions.
- The mechanism of neurovascular coupling (astrocytes vs. neurons; what the flow increase is even for) is genuinely debated — which matters, because every fMRI result depends on it.
- The inflammation–depression link is persuasive as one pathway in some people, not as a theory of depression in general.
- The meningeal-lymphatic and skull-marrow-channel findings are real and confirmed to exist in humans, but most of the mechanistic and disease-relevance work is recent and rodent-derived. The organizing idea is more durable than any single finding; expect revision.
If there is a meta-lesson threaded through this chapter, it is the one that keeps recurring in the body of it: a real signal and a tempting inference are different things, and the gap between them is where careful science lives. Serotonin in the gut, arousal on a polygraph, a microbe in the stool, a cytokine in the blood — each is real, and each invites an overreach. Learning to feel that gap, and to sit in it without rushing to fill it, is not a failure of nerve. It is the skill.