Evolution's Punctuation

For a hundred and fifty years we have assumed we knew how that run happened: slowly. Darwin was explicit — natura non facit saltum, nature does not make leaps. Change accumulates in increments too small to notice, and eventually there is a crow.

A team at the University of Michigan has now measured this properly for the first time, and nature makes leaps.

The measuring is the story, so start there. Bird skeletons live in museum drawers by the tens of thousands, and they have been sitting there, in some cases, since before the theory of evolution had finished being controversial. The bottleneck was never the specimens. It was that measuring a bird skeleton with calipers is soul-erasing labor, and a graduate career buys you maybe a few hundred specimens. So the field studied whatever was cheapest to measure — usually a beak, sometimes a wing — and reasoned about whole organisms from fragments.

Brian Weeks's lab at Michigan, working with David Fouhey's lab at NYU, built a computer-vision system called Skelevision: photograph the specimen against a measurement grid, let the model find twelve bones across the skeleton, move on. Forty-five seconds per bird. They ran it across more than fifteen thousand specimens from some two thousand species and came out the other side holding roughly a hundred and seventy thousand measurements. Jake Berv, the lead author, then wrote a statistical method — he called it bifrost — that could take the whole skeleton as a single evolving shape instead of a bag of unrelated bones.

Then they pointed all of that at forty-five million years and asked how fast body shape had changed.

The answer, published in Nature Ecology & Evolution, is: almost never, and then violently. Songbird body shape spends most of its history barely moving. The rate then spikes — and the largest spike in the record falls near the Eocene–Oligocene transition, roughly thirty-five million years ago, when the planet's climate reorganized, Antarctica started growing ice, and the warm greenhouse world ended.

The other thing the statistics found is stranger, and it is worth flagging early because it complicates the clean version of this story. Around fifteen million years ago, alongside another major geological event, the analysis picks up a cluster of evolutionary slowdowns — rates of body-shape change going down, not up. Not a second burst. The environment moved and the birds got more conservative. Hold onto that; it turns out to matter more than the tidy version would have.

How tightly those dates actually line up is worth holding loosely. Rate shifts inferred from a tree of living species, against a passerine fossil record as thin as this one, lean hard on molecular-clock calibration, and the error bars run wide. "Thirty-five million years ago" is a neighborhood, not an address. The stronger evidence is spatial, and it deserves more billing than it usually gets: bird communities at extreme latitudes, where seasonal temperatures swing hardest, contain species that evolve faster than those living near the equator. That is not a coincidence of timing. That is dose and response — more environmental variability, more morphological change — and it is what turns a suggestive date into something with a mechanism inside it.

So the trigger is not internal. Nothing in the birds decided it was time. Though "the environment moved" is covering two different stories here, and a rate curve can't fully separate them: the cold may have squeezed these lineages directly, or the same upheaval may have emptied enough niches that survivors were pulled into new shapes by opportunity rather than pushed by stress. Catastrophe and invitation leave the same signature. Either way the instruction came from outside — the shape of every skeleton in that world became, overnight in geological terms, the wrong answer to a question that had just been re-asked.

This is the old punctuated-equilibrium argument that Niles Eldredge and Stephen Jay Gould made in 1972 and that the field then spent five decades fighting about, largely because the fossil record is patchy enough to let everyone see what they came to see. What is new here is not the idea. What is new is that somebody finally counted.

And here is the part that should induce actual vertigo — with the caveat that this is the inference, not the measurement. What the study establishes is that the flatness is real. Why it's flat is a further step, and rate-of-change data can't take that step alone: strong stabilizing selection pinning a shape in place, nothing pushing at all, developmental constraint, a shortage of raw variation — all four look identical from here. But the climate signal argues for the first. If shape only moves when the environment moves, something was holding it while the environment sat still.

Which means the long flat stretches are not the boring part. We read stasis as nothing-happening, as a system idling between the interesting events. But stasis is what a well-fitted organism looks like from outside. Everything is happening. Selection is running full-tilt, every generation, and its output is keep it exactly the same. All that force, producing no visible change, because the environment is not asking anything new. The flatness is not the absence of pressure. It is pressure perfectly balanced — a held note, not a silence.

And that is where the fifteen-million-year slowdowns stop being an inconvenience and start being the most interesting number in the paper. If the environment were simply an accelerator — upheaval in, change out — then a major geological event should never show up next to rates going down. But if what an upheaval actually does is re-ask the question, then sometimes the new question has the same answer as the old one, and selection's whole job becomes holding the shape harder. An upheaval that clamps a lineage in place is stasis with the volume turned up: the same force, pushing the other way, finally leaving a mark. That reading is mine and not the paper's — a slowdown has as many candidate causes as a flat stretch does, and the rate curve won't adjudicate between them. But it is the reading that makes both halves of the record one phenomenon instead of two.

Which means the burst is not the arrival of something that was missing. It is the removal of what was holding it still.

There's a joke folded into this that I don't think anyone planned. The specimens that produced this finding had been in their drawers for over a century. The birds sat there through the entire punctuated-equilibrium war, holding the answer, while the argument raged in journals over fragments. Nothing about the evidence changed. What changed was that the cost of looking collapsed — a hundred and seventy thousand measurements arrived where a few hundred used to live — and the field lurched. A discipline in stasis, waiting on an environmental shift, punctuating the instant one arrived. The paper is an instance of its own conclusion.

But the sharper version of that joke isn't about timing. It's about what the instrument was doing to the theory the entire time.

For a century and a half the affordable instrument was a pair of calipers and a graduate student's patience, and that instrument could only produce one kind of evidence: a few hundred specimens, one or two traits, usually a beak. Small samples of low-dimensional data are precisely the conditions under which change looks gradual. You cannot resolve a burst you lack the sampling density to see, and a single trait averages away the whole-body reorganization that a burst largely consists of. Gradualism wasn't only a hypothesis winning on the evidence. It was, in part, the shape of what calipers could afford to say. Natura non facit saltum may be partly an artifact of measurement economics.

That is a live problem rather than a historical one, and it arrives with its own warning attached. Skelevision measures twelve bones. It measures them beautifully, at forty-five seconds a bird, at a scale the caliper era could not touch — and it will now generate twelve-bone-shaped theories with exactly the confidence the caliper era generated beak-shaped ones. The cost of looking collapsed. It did not thereby become unbiased. If cheap machine measurement is about to punctuate every empirical discipline at once, and it is, then the thing worth watching is not the answers arriving. It's which questions the new instrument makes cheap — because those are the ones that are going to look like discoveries.

So: you are almost certainly in a flat stretch. Most of anything's history is flat stretches. It does not feel like force perfectly balanced; it feels like nothing working. But the drawers are full, the pressure is on, and the shape you are holding is a correct answer to a question that will, eventually, get re-asked.

I should be honest about the consolation, though, because the record is offering less of it than I'd like. The lineages in that dataset are the ones that got reshaped. The ones that met the re-asked question and simply stopped are not in the drawers — which is precisely why the pattern reads as held pressure released into new form, rather than as a cull. Museums are built out of survivors. The burst and the extinction are the same event seen from the winning side.

That doesn't ruin it. It means the flat stretch isn't a promise. It's a position.

Nothing is happening.

That's the part that's working.