Something thuds downstairs in the middle of the night. Before you have worked out what it was, your heart rate is up, your breathing has gone shallow, and you are lying perfectly still in the dark with your ears straining.
Half a minute later you remember: the cat. Your body does not come back with you — it takes several more minutes to settle.
The order is backwards. You do not get frightened and then your body follows; a whole set of settings flips first, and you find out afterwards. That by itself needs explaining: what is doing the flipping, and on what.
More than your heart rate changed in that half minute. Breathing, pupils and muscle tone all shifted; attention narrowed to a slit; memory got turned to the page marked danger, and you suddenly wondered whether you locked the door last week. You did not do these one at a time. They arrived together.
Ancestors ran into a small number of situations over and over: something is stalking you, what you swallowed was bad, the child is gone. Each one needs a different setting, and a fixed one — being stalked calls for tension and a narrow field of view, bad food calls for nausea and getting it back out, and mixing the two is useless. A variation that happened to appear wired one cue to the whole set at once. Individuals carrying that wiring were a little likelier to survive such a moment and raise their young; the ones who had to dial each knob on the spot were slow, and slow lost.
Treating emotions as "superordinate programs that run the other systems" (the phrase is Tooby and Cosmides') organizes the field well, but the framework itself is not a result — every specific emotion has to earn its own evidence.
One piece of evidence says the faces are not copied from anyone. In Athens in 2004, researchers photographed judo competitors at the Olympic and Paralympic Games in the instant a match ended. Athletes blind from birth, who have never seen a single human face, made the same winning and losing faces as the sighted ones.
The second piece is harder. A patient known as S.M. lost both amygdalae to a rare disease. Taken to a pet store to handle snakes, walked through a haunted house, she was not afraid; shown photographs of frightened faces, she could not tell what they were. Yet in 2013 researchers had her inhale air with 35% carbon dioxide, and she had a panic attack on the spot, clawing at the mask and saying she could not breathe. Two other patients with the same damage did the same. Guarding against threats out there needs the amygdala; noticing that something inside you has gone wrong runs on another route. One word had been covering at least two systems.
Sockets, traffic, pill bottles — the things that genuinely kill — take a hundred tellings, and a small child may still not take them in. But let him see an adult recoil from an insect once, and from then on he recoils too.
This has been done cleanly in the lab. Rhesus monkeys raised in Mineka's laboratory had never seen a snake and were not afraid of one. Show such a monkey a video of another monkey panicking at a snake and after a few viewings it is afraid — afraid enough to refuse to reach over the snake for food. The trick is in the editing: researchers spliced the very same panic footage onto artificial flowers and a toy rabbit, and the watching monkeys never learned to fear flowers; spliced onto a toy snake or a toy crocodile, they learned at once. The model monkey's expression is identical; all that differs is what it is aimed at.
So what gets passed on is not the content "snakes are frightening" but a bias: some pairings are unusually cheap to form. The bias does not guarantee you fear snakes, it only paves a few learning channels wider than the rest. That is why phobias are anything but randomly distributed: far more people fear snakes, spiders and heights than fear electricity, sockets or cars — and it is the second list that does the killing now.
Put the other way round: evolution does not "know" snakes are dangerous. Primates dealt with snakes for tens of millions of years, and the quick learners among them left slightly more offspring on average. Sockets have had about a century, which is nothing.
The monkey work is clean; the human work is not. Do pictures of snakes and spiders paired with a shock really produce a conditioned response that is harder to extinguish? A 2018 review pooled thirty-two such experiments and found the effect in only ten. This one holds up in monkeys and is unsettled in humans.
Write two lists. Things that disgust you: vomit, rotting meat, a suppurating wound, faeces, a cockroach crossing your food. Things that will actually kill you: cars, electricity, heights, the wrong pill. The lists barely overlap.
In 2004 someone put a questionnaire on the BBC's science pages; after implausible responses were dropped, just under forty thousand remained, from more than a hundred countries and territories. The items were paired images: one carrying a disease cue — lesions, parasites, rotting meat, bodily fluids — and one almost identically composed but recoloured so the disease reading disappears. The cue-carrying image was rated more disgusting in every region; women rated higher than men, and ratings declined steadily with age. This is correlation, not causation, but the pattern was predicted in advance rather than read off the data afterwards.
Pathogens are invisible. By the time there is a fever, they are already in, so the only thing that can work is a visible cue before contact — smell, colour, texture, whether something is crawling on it. Individuals who happened to carry "see this sort of cue, back away" got infected a little less and left a few more offspring. What sets it off is never the actual risk, it is the cue. Cars do not disgust anyone for exactly that reason: a car carries no cue any ancestor would recognize.
The same cue system also runs on people: individuals with visible signs of illness get avoided, and that shows up in experiments. Push one step further — claim the system explains hostility toward outgroups — and the evidence thins fast. A 2023 review rates that line of evidence as mostly indirect, and across dozens of countries the link between disgust sensitivity and prejudice is small. One thing has to be said up front: explaining where a reaction comes from is not endorsing it. This system was calibrated to cues from tens of thousands of years ago, and it gets people wrong all the time.
A baby wakes at night, finds the room empty, and cries until you have to get up. You know he is perfectly safe: the door is locked, the room is 22 degrees, nothing can get in. He does not know that.
Crying at being left alone shows up between six months and a year, on much the same timetable across very different cultures — it even appears on schedule in a rural Bangladeshi sample under heavy developmental stress. Set that back tens of thousands of years: leaving an infant who cannot walk alone overnight in a sealed room simply did not happen, and back then "no adult anywhere nearby" and real danger were close to the same thing. The trigger is set to that world's statistics, not to your house.
Why does it go off so easily? Because the two errors do not cost the same. A false alarm costs one wasted bout of crying; a miss can cost everything. As long as that asymmetry holds, the threshold gets pushed down until false alarms are routine. People call it smoke-detector logic: the one in your kitchen that screams at toast is not broken, whoever set it did the arithmetic.
This story is handy, handy enough to fit anything, which is exactly what the bar set out in the evolutionary psychology piece is there to stop: an evolutionary explanation counts only if it first says what should be visible somewhere else, then goes and looks, and falls over when the opposite turns up. The infant case clears part of that bar — the cross-cultural timetable can be predicted in advance and could have been refuted. "Modern anxiety is caused by mismatch" has no observation that could overturn it, so it is still a story rather than a result.