Two in the morning, thermometer reads 38.6. You take a paracetamol tablet. Underneath that reflex sits an unspoken judgement: the fever is part of the illness, so pushing it down is a step in the right direction.
But think back to the half hour before it climbed. You were shivering, hunting for blankets, complaining the room was cold. A broken air conditioner stops cooling; it does not read thirty degrees and keep heating. This looks more like somebody turned the thermostat up a notch, with the body dutifully chasing the new number.
"Why do we get sick" does not have one answer. One of them is: the thing you took to be the illness is not the illness.
There is a small region under your brain called the hypothalamus, and it holds a set point — the temperature the body is supposed to maintain. During an infection, signalling molecules released by immune cells act on that region and push the set point higher. Everything after that follows: measured against the new number you are now "too cold", so your blood vessels constrict, your muscles shiver, you go looking for blankets, and you keep heating until you catch up. The body is doing all of it on purpose, and it burns real extra energy to do it.
The cleanest evidence that the response actually helps comes from an experiment by Kluger's group in 1975. Desert iguanas cannot generate their own heat; they set their body temperature entirely by moving into sun or shade. After being injected with a pathogenic bacterium they chose warmer spots and ran about two degrees hotter than usual. The decisive step was next: infected lizards were confined to cool, medium or warm environments with no choice, and the cool group died at markedly higher rates. Temperature had been isolated as a single variable, so the conclusion holds.
There is corroborating evidence in humans. A 1973 study gave an antimotility drug to volunteers infected with Shigella, a bacterium that causes dysentery. The diarrhoea was indeed stopped; the price was that fever lasted longer and the bacteria took longer to clear. Plug the thing that flushes outward, and the mess stays inside for extra days.
One more sentence has to follow, though: a defence being useful is not the same as suppressing it being harmful. In 2015, a 700-patient randomised trial across more than twenty Australian and New Zealand intensive care units gave paracetamol to febrile patients with suspected infection, and found no difference in mortality or in days spent out of intensive care. The two statements look like they are fighting. They are not.
Your kitchen smoke alarm shrieks whenever you pan-fry fish, and you have probably thought about pulling the battery out. The manufacturer could make it duller but must not: the two kinds of error cost wildly different amounts. Alarm without fire costs you five annoyed minutes. Fire without alarm costs the house. Even if the real fire has a probability of one in a thousand, setting the threshold at "better to shriek too often" is still the better deal.
The responses that make you feel awful run the same arithmetic. Vomiting costs one meal and half a miserable day. Failing to vomit something genuinely poisonous can cost you every meal after that. So the threshold that got filtered through must lean towards "better to throw up a few times too many". Nesse named this the smoke detector principle. Nobody is weighing anything here — it is only that individuals with looser thresholds left more offspring on average, so loose became the default.
A direct consequence: the vast majority of the fevers, nausea, anxiety and pain you experience in a lifetime really were suffered for nothing. That is not a counterexample; it is exactly what a well-calibrated system looks like. Demanding that it fire only when something is truly wrong is demanding that it accept misses — and the missed cell outweighs the other three.
The two seemingly contradictory conclusions at the end of "fever is not a breakdown" dissolve right here. Suppressing one episode of a defence usually costs nothing, because that episode was not saving you; occasionally one of them does cost, and you cannot tell in advance which. The intensive care trial measured an average over a large group — and an average showing no difference is perfectly compatible with the defence mattering in a small minority of cases.
Myopia makes a good example because the numbers are so stark. Among people finishing secondary school in East Asia, 80–90% are short-sighted, and one to two in ten are highly myopic — a degree that really does damage the retina. All of that rose within a few decades, and a few decades are nowhere near enough for genes to shift that far.
The eyeball does not grow to a blueprint in childhood; it calibrates its own length against what it sees as it grows. Under bright outdoor light with distant scenery, that calibration produces an eye focused correctly at distance; move it into indoor lighting dominated by screens and books at arm's length, and the same rule overshoots. The rule did not break. The input changed.
This is not a guess made after the fact. A randomised trial in Guangzhou in 2015 took twelve primary schools and roughly 1,900 first-graders, and gave half the schools an extra forty minutes of outdoor class a day. Three years later, cumulative myopia incidence was 30.4% in the extra-outdoors group against 39.5% in the controls. Randomised, three years of follow-up, that sample size — it is among the hardest evidence the field has.
Ailments of this shape — the trait did not get worse, the statistics of the environment changed — are called mismatch. The concept needs a guardrail, though: it is a superb generator of hypotheses and also the easiest thing in biology to spin stories with. Any modern complaint can be fitted with an offhand "our ancestors didn't live like this", and it will sound plausible every time. The just-so story critique from the evolutionary psychology topic is the same failure mode. There is exactly one test: can you name the specific environmental parameter that changed, and can you change it back and watch the outcome move? Myopia passes (outdoor hours, changed back, incidence genuinely fell). Most popular claims do not.
Two more kinds of ailment are neither defences nor things you can dodge by changing the environment.
The first kind comes bundled. Sickle cell disease is a severe blood disorder caused by two copies of one variant, yet in regions with heavy malaria that variant stays at a substantial frequency. The reason is that people carrying a single copy do not get the disease but resist severe malaria strongly — a 2005 Kenyan study put the protection at close to 90%. In that environment, single-copy carriers left more offspring on average, so the variant grew more common; and as long as it is common, every generation necessarily produces some children with two copies. Nobody weighed anything, and nobody traded anything away "for malaria resistance" — the ledger simply came out positive, and the attached cost stayed on with it.
The second kind has nowhere to go. Your food and your air share a stretch of throat, which is why you can choke, and occasionally die of it. This is not a step that went wrong; it is the result of fish anatomy from hundreds of millions of years ago being modified one step at a time, with no option to start over. As the evolutionary developmental biology topic put it, evolution is a tinkerer, not an engineer — it reworks whatever is at hand. Choking is the price of that history.
Set the four bills side by side: it is a defence, not damage (fever, coughing, vomiting); the adversary turns over far faster than you do (bacteria can run dozens of generations in a day, you take twenty-odd years for one, so in this tug-of-war you are always catching up — it is why flu vaccines have to be reformulated every year); the environment's statistics changed (myopia); bundled costs and roads never built (sickle cell, choking). One question, four entirely different explanations.
One last thing to nail down: these explanations answer why the machine tends to break in this particular way, never whether to fix it. Knowing fever is a defence does not mean you should grit your teeth through it; knowing a sickle variant had benefits in some places does not mean it should go untreated. Sliding from "this is how it came about" to "therefore this is how it ought to be" is the most common and most harmful misreading in the whole field.