Every critic of evolution eventually reaches for the same weapon: the gaps. Species show up in the rocks looking finished, sit around unchanged, and then get abruptly swapped for something else — with precious few “in-between” forms bridging the jump. Where, they ask, are the missing links?
Here’s the twist this lesson turns on. Under punctuated equilibrium, those gaps aren’t an embarrassment to be explained away. They’re a prediction. If rapid change happens in small, isolated populations over geological eyeblinks, then the intermediates were rare, local, and short-lived — exactly the recipe for not fossilising. The gap isn’t a hole in the theory. It’s the theory’s fingerprint pressed into the rock.
Before you read — take a guess
Before we dig in — the fossil record shows species appearing 'suddenly' with few transitional forms between them. Under punctuated equilibrium, what does that pattern mean?
The gaps are data
Darwin knew the gaps were a problem for him. In On the Origin of Species he devoted a whole chapter to “the imperfection of the geological record,” fretting that the smooth, finely-graded series his gradualism predicted simply wasn’t there. His escape hatch was preservation: the record is a book with most of its pages torn out, and the missing links are missing only because we haven’t dug them up yet.
That’s the gradualist read of the gaps: intermediates existed in a long, even chain, but erosion and bad luck destroyed most of them. Keep digging and the chain should slowly fill in.
Punctuated equilibrium offers a different read of the same rocks. Rapid change is concentrated at speciation events, in small peripheral isolate populations — a founder group cut off at the edge of a species’ range (Mayr’s allopatric speciation, from Lesson 2). Small population, short episode, one cramped corner of the map. Every factor there lowers the odds of leaving fossils. So under this model the scarcity of intermediates isn’t an accident to apologise for — it’s predicted. The gaps are data, and they say the same thing over and over: change was fast and local.
Worked example: a lineage that sits, then gets replaced
Take Phacops rana, a trilobite from the Devonian of North America, roughly 380 million years ago. Eldredge’s classic study read its eye lenses column by column up through the strata. What he found wasn’t a smooth drift in lens count. A given form sat unchanged through metres of rock — hundreds of thousands of years of stasis — and then, at a bedding plane, was abruptly replaced by a descendant form with a different, reduced number of lens files. No gentle ramp between them in that basin. The intermediate step had happened elsewhere, in a small isolated population, and only arrived in the main record already finished.
Read it as a table of expectations versus observation:
| What gradualism predicts here | What the rocks actually show |
|---|---|
| Lens count drifts smoothly, layer by layer | Lens count holds flat for long spans |
| A dense series of intermediate forms in the same section | An abrupt swap at one horizon, few intermediates |
| Gaps blamed on missing pages | Gaps predicted — change happened off-stage, in a small isolate |
Neither picture denies descent with modification. They disagree about tempo — how the change is distributed in time — and Phacops, like many lineages, votes for the staircase.
'Sudden in the rock' ≠ 'instantaneous in time'
When a palaeontologist calls an appearance “sudden,” they mean sudden relative to the rock column, not instantaneous in years. Sedimentation is slow and uneven: a single bedding plane a centimetre thick can represent ten thousand years — hundreds of human generations. A transition that looks like a knife-edge in the strata may have taken longer than all of recorded human history. “Rapid” is a geological verdict, not a stopwatch reading. Nothing here happens in one generation, and nothing here is anti-Darwinian.
The landscape reading: stuck on a peak
To see why stasis is so stubborn, borrow the fitness landscape from an earlier course. Picture every possible form of an organism spread out on a map, with height standing for fitness — how well that form survives and reproduces. Selection is gravity in reverse: it pushes populations uphill, toward higher fitness.
In this picture, stasis is a population parked on a local peak. It sits at the top of its own hill. And here’s the trap: every small step off the summit is a step downhill, into lower fitness. Stabilising selection — the conservative face of natural selection — immediately punishes those deviants and shoves the population back up to the top. Try to wander and you get returned to sender.
So why doesn’t the population just climb to a higher peak nearby, if one exists? Because peaks are separated by valleys — bands of low-fitness intermediate forms. To reach a taller hill, the population would first have to march down into the valley, through forms that are worse than what it has now, before it could start climbing the far side. Selection won’t allow the descent. It’s like being trapped on a good-but-not-best hilltop because every route to the better hilltop runs through a swamp, and something drags you back up the moment you step toward the water.
Why big populations are the most stuck
Counter-intuitively, a large, thriving population is often the hardest to move. Two forces pin it: stabilising selection is strong when numbers are high (deviants are reliably outbred), and gene flow — constant interbreeding across the whole group — averages out any local experiment before it can spread. Size is stability. The valley walls are steepest for exactly the populations that look most successful. That’s the hindsight trap coming: the flattest, healthiest-looking line is often the most locked-in.
The valley-crossing
If selection resists every step downhill, how does a population ever leave its peak? A punctuation is a valley-crossing, and there are two ways to make one happen.
Route one — a small isolate slips across. This is the Lesson 2 story, now drawn on the landscape. Cut a tiny founder group off at the edge of the range and the arithmetic of selection changes. In a small population, genetic drift — random sampling of who breeds — and founder effects can overpower weak stabilising selection. The little group can wander downhill, into the valley, in a way a big population never could, because with few individuals chance beats the gentle push of selection. Once it drifts past the low point, the upslope of a new peak takes over and ordinary selection hauls it up the far side. Small N is what lets the swamp get crossed. Then the newcomer, now a distinct form on a new peak, can spread back into the main range and appear — suddenly — in the record.
Route two — the landscape itself deforms. The map isn’t fixed. When the environment shifts — climate, predators, a new food source, a mass extinction clearing the neighbours — fitness gets recomputed, and the whole surface warps. The old peak can sink (yesterday’s ideal form is now mediocre) or a new, higher peak can rise right next door. A population that was hopelessly “stuck” a moment ago suddenly finds itself on a slope, with uphill pointing somewhere new. It didn’t move; the ground moved under it. Now selection, the same force that pinned it, drives it rapidly toward the new summit — a punctuation with no drift required.
Notice the two routes need different triggers but produce the same fossil signature: a form that was flat for ages, then a fast shift to something new. Whether the population crossed the valley (small isolate) or the valley moved to it (environmental shift), the rock records the same thing — stasis, then lurch. That’s why the pattern is robust even though the mechanism behind any single case can be hard to pin down.
Drive it: watch the landscape inset
Below is the stasis-and-lurch lab. The main panel plots a lineage’s trait against time; the small landscape inset on the side is the payoff for this lesson — it draws the population as a ball sitting on a peak, with a dashed arrow showing the valley-crossing to a new peak.
Watch the inset as you move the sliders. Turn the constraint up high and the ball is pinned on the old (local) peak — the trait line goes long and flat, because stabilising selection returns every deviation. Now add a shock: the environment deforms the landscape, the valley becomes crossable, and the ball lurches across to the new peak — a sudden vertical jump in the trait line. High constraint plus a shock is the whole model in one gesture: held on the peak, then flung across the valley.
Stasis, then lurch
Pinned on a peak until a shock lets it cross the valley
A lineage’s trait sits almost perfectly still for long stretches, then jumps in a sudden burst when a shock finally breaks its constraints. Set how often the environment shocks it and how locked-in it is, and watch the flat lines punctuated by lurches.
Where it sits: pinned on a peak until a shock lets it cross the valley
With shocks at 4/10 against a constraint of 6/10: 4 punctuations across the horizon, the trait sits in stasis 93% of the time (longest still spell 17 units), and the biggest single lurch is 33 points — long equilibria punctuated by fast bursts — stable for ages, then all at once.
One idea, three views
The landscape reading lets you translate between three vocabularies for the same phenomenon. Read across each row:
| Fossil-record term | Landscape picture | What you see in the trait line |
|---|---|---|
| Stasis | Population on a local peak | A long, flat span |
| Punctuation | A valley-crossing to a new peak | A sudden vertical lurch |
| Stabilising selection / constraint | Steep valley walls holding it in | Why the flat spans stay flat |
| Small isolate or environmental shift | What lets the crossing happen — drift across the valley, or the ground deforming under it | The trigger for each lurch |
Once you can move fluently across a row, you can start at any column. See a flat line? Ask what peak the system is stuck on and how steep the walls are. See a sudden jump? Ask which valley just got crossed, and what let it — a small isolate, or a shifted landscape.
Match each concept to its landscape-and-record meaning.
Pick a term, then click its definition.
A flat line is not 'nothing happening'
The most seductive mistake in this whole model is reading a flat trait line as an empty one — a system idling, waiting, doing nothing. It isn’t. That flatness is stored pressure: a population held on its peak by stabilising selection actively swatting down every deviation, generation after generation. “No visible change” is the output of enormous, ongoing work, not the absence of it. In hindsight the long stasis looks like dead time and the lurch looks like the only “real” event — but the stasis is precisely what was loading the spring. Mistake the calm for emptiness and every punctuation will blindside you.
When to reach for the landscape reading
The peaks-and-valleys picture isn’t always the right tool — it’s a specific lens with a specific job. Reach for it when you need to explain why a system resists improvement it could clearly benefit from — when the better state is visible but the path there runs downhill first. It’s unbeatable for that “stuck despite a better option nearby” puzzle, because it names the culprit: the valley in between.
Reach for something else when the terrain is the wrong shape. If fitness rises smoothly with no intervening valley, there’s no barrier to model and plain directional selection covers it. If the “peaks” can’t be meaningfully ranked by a single fitness measure, you’re forcing a landscape onto something flatter than that. And remember the map is a cartoon: real fitness has thousands of dimensions, and a “valley” in two is often a mountain pass in a third. Use the landscape to build intuition about why things stick and how they unstick — not as a literal survey map.
Check yourself
A mature company is stuck on a profitable but ageing product — a clear 'local peak.' A better model exists, but reaching it means a painful stretch of worse performance first (the valley). Which move is most likely to actually get it across?
The whole lesson, mapped
Big picture
Gaps and peaks
- Why the record looks gappy
- The gaps are data
- Darwin's worry: imperfection of the record
- Gradualist read: missing pages, keep digging
- Punctuated read: rapid change in small isolates → few fossils, gaps predicted
- Sudden in rock ≠ instantaneous in time (1 cm ≈ 10,000 yr)
- Stuck on a peak = stasis
- Local peak: every step off is downhill
- Stabilising selection returns deviants
- Big populations most stuck (selection + gene flow)
- Valley-crossing = punctuation
- Small isolate: drift + founder effects cross the valley
- Environment shifts: landscape deforms, old peak sinks / new one rises
- Same fossil signature either way: flat, then lurch
- Traps
- Flat line ≠ nothing happening — it is stored pressure
- Landscape is a cartoon: many dimensions, use for intuition
- The gaps are data
Where this goes next
You now have the model’s hardest technical claim in hand: the gaps are a prediction, stasis is a population pinned on a local peak, and a punctuation is a valley-crossing that only a small isolate or a shifted landscape can pull off. That’s the biology, complete.
The surprise is how little of it is biology. The very same shape — long stasis on a local peak, snapped by a rapid crossing when a constraint breaks — turns up in technologies that plateau then leap, in companies that converge for years then reorganise overnight (Tushman and Romanelli built a whole theory of “punctuated equilibrium” for organisations out of exactly this), in Kuhn’s scientific paradigms that hold until they shatter, and in your own stubborn habits. Next lesson we take the crab’s lesson and read it across all of them.