The Stress Response

Stress is not one system but two

Imagine a car swerves towards you on a pavement. Something happens to your body immediately: you are already moving before you have consciously decided anything, your heart is hammering, your hands are cold and shaking. Something else happens to your body about twenty minutes later, when the car has long gone and you are sitting down with a cup of tea, still feeling strange, still not hungry, still slightly wired. Those two experiences are not the same physiology at different intensities. They are two different systems, and the second one had barely started when the first one had already finished its work.

The reason this distinction matters so much is that nearly every popular account of stress silently blends them. Articles talk about "stress hormones" flooding the body during a near-miss, when in fact the near-miss is handled almost entirely by nerves and by adrenaline, and the principal stress hormone, cortisol, arrives far too late to have helped. Articles then talk about "fight-or-flight" as the thing that damages you over years, when fight-or-flight is over in minutes and the damage story is about a different molecule from a different gland.

Both begin in the same place. Sensory information about a potential threat reaches the amygdala, a pair of almond-shaped nuclei buried in the medial temporal lobe and part of the limbic system. The amygdala performs a rapid, coarse appraisal of significance, and it does so partly on the basis of a fast subcortical route from the thalamus that bypasses the detailed processing of the cortex entirely. That is the neural reason you flinch first and identify the object second. The amygdala then signals the hypothalamus, and the hypothalamus is the point at which the road forks. Down one branch runs the nerve signal. Down the other runs the hormone.

Keep the fork in mind for the rest of the page. Everything that follows is either about one branch, the other branch, or the relationship between them.

The fast system: seconds, nerves, adrenaline

The sympathetic nervous system is the body's mobilisation network. Its neurons run from the spinal cord out to essentially every organ, and when it fires as a coordinated block, the effect is a whole-body reconfiguration for immediate physical action. This is Walter Cannon's fight-or-flight response, described in the early twentieth century and still the correct picture of the fast arm.

What makes it fast is that it is wired, not chemical, at least to begin with. The hypothalamus drives brainstem and spinal sympathetic outflow directly, and a nerve impulse travels the length of the body in milliseconds. There is no waiting for a molecule to diffuse into the blood and circulate.

That last card deserves emphasis, because it is where the fast system does something clever. Direct nerve wiring is quick but narrow: only the organs you have wired get told. By having the sympathetic system also fire onto a gland that squirts its transmitter into the blood, the body converts a targeted neural signal into a broadcast chemical one, without giving up much speed. Adrenaline reaching a tissue through the bloodstream arrives seconds rather than milliseconds later, which for a whole-body emergency is fast enough.

Two more things about the fast system. First, it is fast to switch off as well as on: adrenaline is cleared from the blood in minutes, which is why the shakiness after a near-miss fades reasonably quickly. Second, the noradrenergic arousal it produces in the brain, driven by the locus coeruleus rather than the adrenal gland, is part of why emotionally charged events are remembered so vividly. Arousal at the time of an event enhances its consolidation, which is adaptive: the situations that nearly killed you are exactly the ones evolution would like you to remember. See neuromodulation for how the noradrenergic system sets brain-wide gain.

The slow system: the HPA cascade

The second branch of the fork is a chain of three glands, each releasing a hormone that acts on the next. It is worth walking through it one link at a time, because the shape of the chain is the point.

1. Step one: the hypothalamus releases CRH

   Neurons in the paraventricular nucleus of the hypothalamus release corticotropin-releasing hormone (CRH). They do not release it into the general circulation, which would be wasteful. They release it into the hypophyseal portal system, a small private network of blood vessels running the short distance down the pituitary stalk. A portal system is a vascular link connecting two capillary beds directly, and its function here is to deliver a tiny quantity of hormone at high concentration to exactly one target, the anterior pituitary, without diluting it in the body's five litres of blood.

2. Step two: the pituitary releases ACTH

   CRH arriving at the anterior pituitary causes it to release adrenocorticotropic hormone (ACTH), and this one does go into the general circulation. ACTH now travels everywhere in the body, but only one tissue is listening for it, because only one tissue carries the receptor in quantity: the adrenal cortex.

3. Step three: the adrenal cortex releases cortisol

   ACTH stimulates the adrenal cortex, the outer shell of the adrenal gland, distinct from the medulla in the middle that handled adrenaline. The cortex synthesises and releases cortisol, a steroid hormone, which enters the blood and reaches essentially every cell in the body. Because cortisol is a lipid-soluble steroid, it passes straight through cell membranes and binds receptors inside the cell, which then act directly on DNA to change which genes are transcribed. This is why its effects take time and why they last.

What does cortisol do once it arrives? Its central job is to make energy available, on the reasonable assumption that a threatened organism needs fuel. It raises blood glucose, promotes the breakdown of protein and fat into substrates the liver can turn into glucose, and reduces the uptake of glucose by tissues that are not urgently needed, so that what glucose there is goes to the brain and the muscles. In parallel it damps down expensive, deferrable projects: aspects of immune function, inflammation, growth, digestion, reproduction. It also raises blood pressure and increases the sensitivity of the vasculature to adrenaline, which is a neat interaction between the two axes.

Notice the timing. All of this begins minutes after the threat and unfolds over tens of minutes. Cortisol is useless against a car swerving at you. It is exactly right for a three-day famine, a wound that needs healing, a fight for dominance that lasts a season. The slow system evolved for slow problems.

Why a cascade rather than a nerve

A reasonable person, seeing this three-step chain, asks why the body bothers. The hypothalamus already has direct neural control over the adrenal medulla. Why not simply wire it to the cortex too, and skip the pituitary?

There are three good answers, and together they explain why endocrine cascades appear again and again in physiology.

Amplification. Each link in the chain is a biochemical multiplier. A small number of CRH molecules, delivered at high local concentration through the portal system, cause the release of a much larger number of ACTH molecules, each of which drives the production of a much larger number of cortisol molecules. The hypothalamus does not have to shout. It whispers, and the cascade does the amplifying. This is metabolically cheap and it makes the system exquisitely sensitive to small changes at the top.

Broadcast and persistence. This is the deeper point. A nerve can only speak to the organs it is wired to. A hormone in the bloodstream speaks to every cell that carries the receptor, everywhere in the body, in a single stroke, and it keeps speaking for as long as it remains in circulation. If the problem you face is not "move now" but "reallocate the whole organism's energy budget for the next several hours", then a broadcast, persistent, systemic signal is precisely the right tool and a nerve impulse is precisely the wrong one. The architecture fits the problem.

Regulation. A three-step chain has three points at which a signal can be turned down. That sounds like a design cost, but it is actually the whole point, and it takes us to the most important section on this page.

The off switch is the whole story

Here is the claim that reorganises everything else you know about stress: a healthy stress response is defined not by how well it turns on, but by how reliably it turns itself off.

Cortisol, once released, does not simply act on the periphery. It acts back on the brain. Glucocorticoid receptors sit in the hypothalamic paraventricular nucleus, in the anterior pituitary, and, densely, in the hippocampus. When cortisol binds them, the effect is inhibitory: the hypothalamus releases less CRH, the pituitary releases less ACTH, and the hippocampus exerts an inhibitory influence on the hypothalamus. The output of the system suppresses its own input.

Understand this and a great deal falls into place. The stress response is not a switch that someone has to remember to flip back. It is self-limiting by construction. Cortisol rises, cortisol shuts down its own production, cortisol falls. In an intact system, a stressor produces a pulse, and the pulse ends.

Which means the interesting pathology is not an overactive on switch. It is a failing off switch. In chronic stress, the feedback loop is what degrades. Sustained high cortisol can reduce the number and sensitivity of glucocorticoid receptors in the very structures that are meant to be sensing it, so the brake becomes less responsive to the signal that should apply it. The hippocampus, which is one of the principal sensors, is also one of the structures most sensitive to prolonged glucocorticoid exposure, which sets up a loop with an unpleasant logic: the response damages the brake, and the weakened brake permits more response.

The brake: prefrontal control of the amygdala

There is a second brake, and it operates upstream, at the level of appraisal rather than hormones.

The amygdala is fast, coarse and biased towards false alarms, which is exactly what you want in a threat detector: the cost of flinching at a stick that turns out not to be a snake is trivial, and the cost of failing to flinch at a snake is total. But a system tuned that way needs supervision, or it will keep the organism in a permanent state of alarm over shadows. The supervision comes chiefly from the prefrontal cortex, and in particular from ventromedial and dorsolateral prefrontal regions that exert top-down regulatory influence over amygdala reactivity. This is the neural substrate of the second look, the reappraisal, the thought that begins "hold on, that is not what it seemed".

Now the consequence. Under sustained stress, this top-down control weakens. Human neuroimaging associates chronic stress and high cortisol exposure with reduced prefrontal engagement and heightened amygdala reactivity, and prefrontal function is unusually sensitive to catecholamine levels: the same noradrenaline and dopamine that in moderate quantities support working memory and control degrade prefrontal function when they are too high. In other words, the very arousal the stress response produces is corrosive to the structure that is supposed to regulate it.

This is not an abstraction. It is why a stressed person is more reactive and less deliberate. It is why decisions made under sustained pressure tend towards the habitual and the impulsive rather than the goal-directed and the flexible. It is why executive function, which is prefrontal work, is one of the first things to degrade when someone is under chronic strain. The stress response does not just make you feel bad. It shifts which parts of your brain are in charge.

Be careful with the strength of the claim, however. The direction of these effects is well replicated and the mechanism in animals is reasonably clear. In humans, much of the evidence is correlational and cross-sectional: stressed people show these patterns, but stress and the patterns may share causes, and demonstrating that stress produced the change requires longitudinal or experimental work that is harder to do and less abundant. The picture is credible. It is not the same as proven.

Acute stress helps, chronic stress harms

It is worth stating plainly, because the wellness literature makes it hard to hear: acute stress is good for you. It is not a malfunction or a design flaw or an evolutionary leftover. It is one of the most useful pieces of physiology you own.

An acute stress response mobilises stored energy, raises cardiac output, dilates airways, sharpens sensory processing, narrows attention onto what matters, and enhances the consolidation of memory for the event that triggered it. Every one of those is an advantage when facing a genuine challenge. Athletes, surgeons and people sitting examinations all perform better with some of this than with none of it. An organism incapable of mounting an acute stress response would not survive a week.

The damage story is entirely about chronic stress: the response mounted repeatedly, or mounted and never fully terminated. And here the honest reporting requires care, because this is a domain where the popular literature routinely converts animal findings into human certainties.

A rule for reading anything written about stress and the brain: when you see a claim that stress "shrinks the brain" or "kills brain cells", check whether the underlying study was done in a rodent. Very often it was, and very often the human evidence is a good deal more equivocal than the headline suggests.

Allostatic load: the cost of the response itself

Bruce McEwen introduced the concept that best organises this whole subject, and it turns on a distinction that sounds pedantic and is not.

Homeostasis is the maintenance of the internal variables that must be held within narrow limits for life to continue: blood pH, core temperature, blood oxygen. Allostasis is something different. It is the achievement of stability through change: the active adjustment of blood pressure, heart rate, hormone levels and metabolism to meet an anticipated or actual demand. Allostasis is not a failure of homeostasis. It is how homeostasis is defended in a world that keeps making demands.

McEwen's framing, set out in his 1998 New England Journal of Medicine paper on protective and damaging effects of stress mediators, identifies several distinct ways the load accumulates, and they are worth distinguishing because they are different failures.

1. Too many hits

   The response is mounted correctly and terminated correctly, but far too often. Each individual episode is healthy; the frequency is not. This is the physiology of a life with one crisis after another.

2. Failure to habituate

   A normal system dampens its response to a repeated, familiar, harmless stressor. A public speaker who has given a hundred talks mounts a smaller response to the hundred-and-first. Some people fail to habituate, and continue to mount a full response to a stressor that has been shown a hundred times to be safe.

3. Failure to terminate

   The response is switched on appropriately but does not switch off, so cortisol remains elevated long after the stressor has gone. This is the feedback failure of the previous section, and it is arguably the central mechanism of allostatic load.

4. Inadequate response

   The mirror image, and the one that proves the concept. If cortisol were simply harmful, then having too little of it would be protective. It is not. An inadequate glucocorticoid response permits other systems, notably inflammatory ones that cortisol normally restrains, to run unchecked. Under-response causes its own damage, which is exactly what you would predict if cortisol is a regulator rather than a poison.

That fourth item is the intellectual heart of McEwen's argument and the best single reason to abandon the "cortisol is bad" framing entirely. The system is damaging when it is dysregulated, in either direction. A response that is too small is not safety; it is a different disease.

The inverted U, and why it is a heuristic and not a law

The relationship between arousal and performance is often drawn as an inverted U: performance rises with arousal up to some optimum and then falls away as arousal continues to increase. The idea traces back to a 1908 paper by Yerkes and Dodson, who studied how mice learned a discrimination under different intensities of electric shock, and found that the intensity that produced fastest learning depended on how difficult the discrimination was.

The curve is genuinely useful. It captures something real, which is that a performer with no arousal at all is flat, unmotivated and slow, while a performer overwhelmed by arousal is rigid, narrow and error-prone, and that somewhere between the two lies a better place to be. It also captures the point that the optimum shifts with task difficulty: a simple, well-practised task tolerates high arousal, and a complex, novel one does not. Anyone who has watched a nervous student produce a fluent answer to an easy question and then fall apart on a hard one has seen it.

But it is a heuristic, and it is routinely over-applied, in ways worth naming:

Myths about stress

Related topics