You find a flower in the hills whose throat grips a particular bee exactly, so that the bee walks in once and comes out dusted on the back. You think immediately: this shape is for that bee. You are probably right.
The trouble is you apply the same thought to everything — the human chin, why blood is red, why you had that dream last night. The longest methodological argument in evolutionary biology is not about whether selection exists; nobody has fought over that in a long time. It is about this: on what grounds do you say this particular trait was selected for?
Set a round dome on a square frame made of four arches and four curved triangular surfaces are left over at the corners. They have to be. It is geometry: build it this way and those four are there.
In St Mark's in Venice, the craftsmen painted an evangelist into each of those four surfaces, the figures fitted to the shape beautifully. Standing underneath and looking up, it is very easy to conclude that the surfaces were set aside so those four could be painted.
The order runs the other way. The surface was there first; the painting was filled in afterwards.
In 1979 Gould and Lewontin wrote a paper and used this for the title. Their point was plain: a trait existing, and looking apt, is not by itself evidence that natural selection picked it out. It may be the unavoidable companion of something else.
The cleanest example is blood. Yours is red not because red did your ancestors any good, but because the molecule that carries oxygen has an iron atom in it, and iron bound to oxygen absorbs part of the spectrum; what is left reaches your eye as red. What selection acted on was whether oxygen gets delivered. Red came along for the ride.
Here is one still being argued over: the chin. Humans are the only animal with a chin that juts forward — Neanderthals had none, chimpanzees have none. Decades of explanations have piled up: it braces the jaw, it helps speech, it is a mating signal. Another camp thinks it was never picked out on its own — the whole face and tooth row of our species shrank, the upper part pulling back harder than the lower, and the small piece left below sticks out. One study comparing skull shapes across humans and apes found clear directional change in things like a smaller face and a larger braincase, while most measurements around the chin showed no sign of being selected on their own. This is not settled — and the unsettledness is the point: looking useful does not close the case.
Why this happens mechanically: selection scores whole individuals — do they survive, do they leave offspring — not traits one at a time. A single change usually alters several things at once, and as long as the tag-alongs are not a drag, they ride along too. Nobody issues permits to by-products.
What Gould and Lewontin were actually aiming at was not the concept of adaptation but a habit of writing: assume every trait has a use, then reason backwards to a story that hangs together; when one story is knocked down, swap in another, and never require any of them to be tested.
The flaw is not that the stories are wrong. It is that this way of working always wins — and an explanation that always wins is risking nothing.
So how do you pay the debt? There are a few routes you can actually walk.
The hardest is to go and measure. In 1980 Endler dug ten ponds in a greenhouse, laid out to resemble stretches of Trinidadian hill stream, and split one batch of guppies among them. Into four he put a single cichlid — a fish that hunts adult guppies in earnest. Into four others he put several killifish, which take only fry. The last two got no predator at all, as controls.
Male guppies carry coloured spots. The showier they are, the more they attract females — and the more they catch a predator's eye. So the prediction was unambiguous: in the cichlid ponds spots should get fewer and duller; in the ponds with the fry-eater, adult males could afford to be gaudier.
He counted spots a few months in, and again twenty months in. The pattern went the way the prediction said. Why that counts: the prediction was written down before the result was seen, and there were controls. A story told afterwards cannot do that.
The second route is comparison: does the same trait keep turning up alongside the same living conditions in lineages that have nothing to do with each other? Once could be coincidence. Seven or eight independent times is harder to shrug off. The third route is to look for marks left on the genetic strand itself — a variant driven up fast by strong selection drags a whole stretch of its neighbours up with it, leaving a suspiciously uniform region, and that uniformity can be measured.
One thing gets read backwards a lot, so: this critique did not kill adaptationism. Most evolutionary biologists still hold that many traits are adaptations, and still go and study what they do. What changed is the bar — "it looks apt" was demoted from conclusion to hypothesis, and you still owe the evidence above.
There is a nerve that leaves your brain, travels down into the chest, loops under a great vessel above the heart, then turns around and climbs back to your throat. The thing it controls sits about a hand's width from where it started. It travels roughly a metre.
In a giraffe the same nerve takes the same route: down into the chest, around the vessel, then back up that long neck. Several metres out and back, in order to cover the last few centimetres.
This is not a design flaw, because there was no design step. In a fish the nerve runs straight: brain, gills and heart sit close together, the vessel is right there, one small loop and it arrives — too short for the detour to show. Then land animals grew necks joint by joint, the heart shifted back and down, and this nerve, pinned behind the vessel, was simply drawn out along with everything else.
Every step in evolution has to be alive. If an unlivable state sits in between, the gap does not get crossed, however good the far side looks.
The eye is the prettier case. Your retina is installed backwards: the light-sensing layer faces the back of your skull, nerve fibres and blood vessels lie across the path the light has to travel, and they have to gather into a bundle and pass out through the eyeball at one spot. That hole is your blind spot, patched over by two eyes and a brain, which is why you never notice it.
The octopus eye evolved separately, looks broadly similar, and is installed the right way round: the sensing layer faces the light, the wiring runs out the back, no blind spot.
Both work. The difference is not which is better but where each was continuing from. The vertebrate retina develops as an outpouching of part of the embryonic brain, and the moment it bulges out the orientation is set. Hundreds of millions of years later there is still no chance to flip it, because the intermediate stage is blind.
Jacob put it in 1977 in a phrase still in use: evolution is not an engineer, it is a tinkerer. An engineer can start over and draw plans first. A tinkerer has only the parts left from the last version, and has to keep the thing working while altering it.
So "why wasn't it built more sensibly" is the wrong question from the start. No version was drawn toward an endpoint. Each is only the next version reachable from the last one. The earlier something was fixed, the more got stacked on top of it, and the less it can move — that is the mechanical content of the phrase "historical baggage".