Documented collapses look alike: complexity that costs more than it brings in, then an ordinary shock that turns fatal.
The picture of a civilisation destroyed by a single event — an invasion, an earthquake, a drought — is almost always wrong. The same shocks regularly hit societies that absorb them without trouble. What sets apart the ones that fall isn't how violent the shock was: it's how much margin they still had.
The central mechanism
A society solves its problems by adding complexity: administrations, specialists, infrastructure, rules. Every addition works, and every addition costs.
The problem is that the returns on complexity shrink. The first irrigation canal transforms a region. The twentieth gains a few percent. The first level of administration makes decisions possible; the fifth slows them down. There comes a point where maintaining the system consumes more than it brings in, and where the whole society works to pay for its own structure.
At that stage, it isn't in danger yet. It simply has no margin left. An ordinary shock — a bad harvest, an epidemic, a border revolt — then becomes a fatal one, not because it's stronger, but because nothing is left to absorb it.
That's why collapse, seen from the inside, rarely looks like a catastrophe. It looks like a string of reasonable give-ups: a service nobody restores, a road nobody repairs any more, a garrison nobody replaces.
What comes back in every documented case
Interdependence spreads the shock. The trade system of the late Bronze Age linked around ten powers through flows of metals, grain and craftsmen. Each was richer thanks to the others, and each had become unable to live without them. When a few nodes gave way, the whole thing came apart within a few decades. A supply chain is an efficiency gain in normal times and a channel of contagion in a crisis.
Knowledge disappears faster than buildings. It isn't held in stones but in people who practise it and pass it on. Break one generation of transmission, and a technique mastered for centuries is lost, even while its products are still visible everywhere. Societies have lived among ruins they no longer knew how to build.
Writing dies first. It belongs to a small group of specialists that only an administration can support. When that administration stops paying, writing stops — and that's exactly why periods of collapse are so poorly known to us: they left nobody to describe them.
The local takes over again. Scale collapses before life stops. Large systems break up into units that feed themselves. It isn't the end of everything: it's a brutal change of scale, where survival is played out within a day's walk.
Decline is slower than people think and faster than they prepare for. It's counted in decades, not days. Slow enough for each generation to think it's normal; fast enough that none has time to rebuild what the previous one lost.
Three documented cases
The Late Bronze Age collapse, around 1200 BCE in the eastern Mediterranean. Within a few decades, several powers disappear or shrink sharply, palaces burn, trade networks stop, and Linear B writing disappears from Greece. The proposed causes combine: a long drought recorded in sediment cores, population movements, the breakdown of tin supply routes — tin being essential for bronze — and a chain collapse of interdependent palace economies.
The lowland Maya cities, between the eighth and tenth centuries. The great southern centres are gradually abandoned, monumental building and dated inscriptions stop. Palaeoclimate data point to several episodes of severe drought, in a region that depended on reservoirs and intensive farming.
The Western Roman Empire, from the third to the fifth century. It doesn't vanish overnight: the coinage loses value, trade routes shrink, towns contract, large building projects stop, and the administration withdraws province by province. Analyses of ice cores show a clear drop in lead pollution, a marker of mining and metalworking, well before the political end.
What archaeology consistently observes
Complexity drops before population does. Fewer administrative levels, less craft specialisation, less long-distance trade, a return to simpler local pottery.
Writing dies early. It depends on an administration that supports scribes; when that stops paying, writing stops. That's why these periods are poorly known, as L'écriture est née des comptes, pas des poèmes shows.
Buildings outlive know-how. People have lived among ruins they no longer knew how to build, or maintain.
Scale shrinks before life stops. Large systems break up into self-sufficient local units. The same sequence, observed in modern technical networks, is documented in Les réseaux, la puissance et la fragilité.
Decline is counted in decades, sometimes centuries, never days. That's what contemporary accounts fail to notice: each generation sees a situation that looks normal to it.