The Problem Is Getting the Weight Down

Stone is roughly ten to twenty times stronger in compression than in tension. That sentence is the engine of several thousand years of building, and once you have it you can read almost any structure put up before steel.

Here is why. Put a stone slab across two supports and load the middle. The top of the slab is squeezed and the bottom is stretched. Stone does not mind being squeezed; it minds being stretched enormously. So the crack starts underneath, in the middle, and runs upwards, and the slab fails.

This is why Greek temples have columns so close together. The architrave is a stone beam, and the distance it can span before it breaks of its own weight is short. The famous rhythm of a classical colonnade is a tension limit made visible.

Now bend the slab into an arch.

Every block in an arch is squeezed by its neighbours. There is no stretching anywhere, if the shape is right, so the material is working in the only mode it is good at. An arch of ordinary stone can span many times what a beam of the same stone could, and a vault made of arches can roof a space no beam would cross.

But something has to be paid, and it is paid sideways.

An arch does not just press down on its supports. It presses outwards, continuously and for as long as it stands, because the thrust that travels round the curve has to go somewhere when it reaches the bottom. There is an Arab proverb that the arch never sleeps, which is the most accurate thing anybody has said about structures in one sentence.

Roman engineering mostly dealt with this by mass. Make the walls thick enough and the outward thrust is absorbed before it can do damage. It works, it is reliable, and it produces buildings with very little window in them, because the wall is doing structural work and cannot be perforated much.

The medieval solution was to stop fighting the thrust and start steering it.

Point the arch, and the thrust leaves the vault at a steeper angle, closer to vertical, which means less of it is trying to push the walls apart. Then catch what remains on a separate pier standing outside the building, and connect the two with a half-arch, which is the flying buttress. The thrust now travels out of the vault, across the gap, down the pier and into the ground, and it never touches the wall at all.

Which means the wall does not have to be a wall.

This is the whole of Gothic, structurally. Once the thrust is handled outside, the surface between the piers carries nothing but itself and can be glass from floor to vault. Chartres, Sainte-Chapelle, the clerestory at Amiens: these are not walls with windows in them, they are frames with glass in them, and the coloured light that people describe in religious terms is the engineering consequence of a solved thrust problem.

The silhouette follows too. Those flying buttresses, pinnacles and piers that read as decorative complexity from the outside are the load path, exposed. The pinnacle on top of a buttress pier is not an ornament; its weight pushes the resultant force back towards the vertical so the pier does not topple outwards. It is a counterweight with a hat on.

So the usual way of teaching this, Romanesque gives way to Gothic as taste changes, has the causation backwards. The taste is downstream. Somebody worked out how to take the thrust outside the building, and the moment that was possible, the interior filled with light and the exterior filled with structure, and the result got a style name afterwards.

The test is to walk into any pre-industrial building and ask one question: where is the weight going, and what is stopping this from spreading at the bottom? Columns close together means a stone beam. Thick walls and small windows means thrust absorbed by mass. Thin walls and enormous windows means the thrust left the building, and if you step outside you will find where it went.