01Executive summary
Four milestones in which builders working entirely without a theory of statics arrived, by observation and by failure, at a structural system of remarkable efficiency.
The ribbed vault appears in the eleventh century — the first significant medieval advance beyond Roman method. Part of the nave vault at Cluny collapsed in 1125 because lateral thrust had been assumed to act vertically. The pointed arch arrived from the East early in the twelfth century. And the flying buttress was used systematically at Notre-Dame in Paris by about 1180. Within two generations the builders of the Paris region could balance horizontal forces with real economy of material, and they did it without any of the mathematics in the following part of this series.
02The problem: a stone roof that will not burn
The early Christian basilica was a Roman form with a timber truss roof and, in northern Europe, a tall wooden spire. It burned. Repeated destructive fires drove builders from the eighth century onward to rebuild in stone with vaulted roofs, and the whole subsequent development follows from that requirement.
A stone vault is heavy, and it does something a timber roof does not: it pushes sideways. A barrel vault or a groined vault exerts a substantial horizontal thrust at its springing, and unless that thrust is resisted, the supporting walls are pushed outward until the vault falls.
- Narrow bays onlyThe earliest Romanesque vaults span little, because slender nave columns cannot support a stone vault at all.
- Piers replace columnsSubstituting rugged structural piers for columns allows the whole nave to be vaulted rather than just the aisles.
- Groined vaultTwo barrel vaults intersecting at right angles concentrate load at four corners and open the walls between them.
- Ribbed vaultRibs are built first, independently of the web they carry, controlling the sagging that afflicts wide groined vaults.
- Pointed arch and flying buttressThe thrust line is straightened and what remains of it is carried outside the building to the ground.
03Cluny, 1125: the failure that taught the system
The third abbey church at Cluny, dedicated in 1130, was an immense building — roughly 90 metres long with ribbed vaults some 30 metres above the pavement. Its aisle arcades already had the elements that would define Gothic construction: unit-element bays, pointed arch ribs, thin stone webs and screen walls.
The great nave vault, however, was set on walls that had not been designed to take lateral thrust. Part of it collapsed in 1125. The builders had apparently assumed the vault would bear down vertically on the inner face of the wall. Strong lateral buttresses were built after the event.
It is one of the earliest cases in this set where a structural failure is traceable to a specific, identifiable misconception rather than to poor workmanship or overload — and where the profession demonstrably learned from it. The Gothic builders working around Paris a decade later appear to have taken two things from Cluny: that the unit-element bay was a real structural advance worth adopting, and that lateral thrust is a first-order problem. By the end of the century they could balance horizontal forces economically. Tacoma Narrows, the Comet fuselage and the seven investigations in the last series of this set are all the same shape of event. The profession has been learning fastest from its failures for nine hundred years.
It is worth noting what they did not have. There was no theory of statics; Stevin's triangle of forces was still four and a half centuries away. What the master masons acquired was a body of empirical structural understanding, developed by building, observing and occasionally watching something fall down. That it produced a system as efficient as it did is genuinely remarkable.
04The system, described as structure
Stripped of its architecture, a Gothic cathedral of the Paris region is a clearly organised structural system, and describing it that way makes the logic obvious.
| Element | Structural function |
|---|---|
| Piers | Carry all vertical load to the foundations. They rise to 30 metres or more and are the primary compression members. |
| Screen walls | Carry only their own weight and are not built structurally into the piers. Most of their area can therefore be window. |
| Pointed arch ribs | Follow the line of thrust more closely than a semicircular arch, so internal stress is lower and smaller stones suffice. |
| Vault web | Spans between ribs. Where stone is accurately cut and fitted it can be very thin — some webs are around 150 millimetres. |
| Flying buttress | Takes lateral thrust at the vault springing and carries it down and outward to ground, outside the line of the walls. |
| Timber roof | A rain-and-snow shed above the vault, on the structural wall. Often completed before the vaults, to keep weather out during construction. |
Two points are worth drawing out. First, separating the load-bearing function into piers and the enclosing function into non-structural screen walls is exactly the frame-and-cladding logic of a modern curtain-walled building, arrived at eight hundred years earlier for the same reason: it lets the enclosure be light and mostly glass.
Second, the flying buttress does more than resist vault thrust. It also helps counter wind load and suction on a very tall, comparatively light structure — which matters more than it sounds for a building of this height and exposure.
05How the work was actually run
The men who directed cathedral construction were master masons, not engineers or architects in any modern sense. They were paid three or four times an artisan's wage and held a relatively high social position. Many were illiterate. All had learned by apprenticeship. A few could read manuscripts such as Vitruvius, and it is worth saying plainly that Vitruvius contains nothing like enough to direct the construction of a Gothic cathedral.
Their material knowledge was better than is often assumed. Hard stone was used for structural piers and softer stone for non-structural wall areas; weather-resistant stone was chosen for external facings; lime washing and other treatments were used against decay. Wrought-iron cramps, dowels, tie rods and occasional eyebar chains appear throughout, and ironwork and nails were boiled in tallow or tinned to prevent rusting — a deliberate corrosion protection system, which the Greeks reinforcing their stonework did not have.
Their mortar, by contrast, was poor. It sometimes took years to set and often lacked adhesion, which is a substantial part of why the system relies on geometry, mass and thrust lines rather than on the joints doing any real work. The pozzolanic cement of the Roman chapter had been lost.
06Takeaways for current practice
- Separate the carrying function from the enclosing function. Piers and screen walls are frame and cladding, eight hundred years early.
- Follow the thrust line. The pointed arch reduces internal stress by conforming to the load path rather than by being stronger.
- A misconception can survive until something falls down. The assumption that vault thrust acted vertically was not corrected by argument.
- Set out full scale where you can. Templates and full-size drawings remove scaling error and need no dimensional literacy on site.
- Protect embedded metal deliberately. Tallowed and tinned ironwork is a corrosion system; the Greeks had none and lost their reinforcement.
The modern descendants sit in AS 3700 for masonry and AS 1170 for structural actions. Cited by number for orientation only — verify currency.