Why buildings look the way they do
Stand in front of any old building and count the windows. Then look at how much wall sits between them.
In a medieval stone house the windows are small holes in a lot of wall. In a Victorian mill they're large and regular. In an office built after about 1955 there's barely any wall left at all. Nobody's taste changed that much in the meantime. What changed is what's holding the building up, and whether the wall still has a job.
Almost everything about how a building looks comes out of four questions. What can you get hold of, how far can it span, what's holding up the floor above, and what the law lets you do. Style sits on top of that. It rarely gets to overrule it.
Stone is brilliant and can't span anything
Stone is superb in compression. Pile it up and it'll happily carry the weight of a mountain. Ask it to bend and it's pitiful, because stone has almost no tensile strength — the underside of a loaded beam is being stretched, and that's the side where stone gives up.
So a stone lintel laid flat across two supports can only manage a short gap before it cracks in the middle. A few metres, and less if there's anything heavy sitting on it. That single fact writes the plan of every pre-Roman stone building.
Look at the hypostyle hall at Karnak: a forest of enormous columns crammed close together, with the visitor squeezing between them. It's not a design statement about the majesty of the gods, or not only that. The columns are that close because the stone beams overhead couldn't reach any further. Greek temples have the same constraint, which is why the spacing of a Doric colonnade barely varies from one temple to the next. The interior of a Greek temple is small and dark because there was no honest way to roof a big open room in stone.
Timber does better, because wood works in tension as well as compression, and medieval Europe used it to get spans that stone couldn't touch. Westminster Hall got its hammerbeam roof in the 1390s, spanning something over 20 metres in oak with no columns in the middle. That's a hall you can actually use. But timber burns, and it rots, and big straight oaks eventually ran out.
The arch is a trick for stopping stone from bending
The way round the problem is to arrange the stone so it's never asked to bend at all.
An arch does exactly that. Cut the stones as wedges, lean them against each other, and load travels down through the curve as pure compression from one voussoir to the next. Stone will do compression all day. Suddenly you can span serious distances with a material that snaps if you look at it sideways.
The Romans took the arch and ran. Extend an arch along a line and you get a barrel vault. Cross two vaults and you get a groin vault, which puts all the load onto four corners and leaves the walls between them free. Rotate an arch about its centre and you get a dome.
They also had concrete, made with volcanic ash, which sets under water and can be poured into shapes rather than cut. The Pantheon, finished around AD 126, is the demonstration piece: an unreinforced concrete dome 43 metres across, still the largest of its kind nearly two thousand years later. The Romans knew exactly what they were doing with it. The walls are enormously thick at the base and thin towards the top, the coffers cut weight out of the shell, and the aggregate changes as it rises — heavy basalt low down, light pumice near the crown. The 8.8-metre hole in the middle isn't a flourish either. Removing the material where the compression rings are weakest makes the whole thing more stable.
There's a catch with arches and it drives most of what happened next. An arch doesn't just push down. It pushes outwards at the bottom, and if nothing resists that thrust the supports splay and the arch drops. Roman practice was to resist it with sheer mass — colossally thick walls with small openings. That works, and it's why a Roman building is mostly wall.
Gothic cathedrals are an argument about thrust
Every dramatic feature of a Gothic cathedral is a move in the same engineering problem: how do you keep the thrust under control while removing as much stone as possible?
The pointed arch is move one. A steeper arch pushes down more and outwards less than a semicircular one of the same span, and it also lets you span unequal distances at the same height, which a semicircle can't. That freed the plan.
The ribbed vault is move two. Build a skeleton of stone ribs first, then infill between them with a thin web. The loads travel down the ribs to particular points rather than pressing evenly along the whole wall head, so you only need real structure where the ribs land.
The flying buttress is move three, and it's the one you can see from outside. If the thrust arrives at known points, you can catch it with a pier standing well clear of the building and throw a half-arch across the gap to reach it. The buttressing moves outdoors. What's left in between doesn't need to hold anything up, so it can be glass.
That's the whole reason Gothic interiors look the way they do. Chartres, Amiens and the rest aren't tall and glazed because someone wanted a mystical effect and then found a way; the structural system arrived first and the effect came with it. The cathedral at Amiens got its vaults over 42 metres up.
And they found the limit the hard way. Beauvais aimed for a choir vault of about 48 metres, finished it in 1272, and watched a large part of it collapse in 1284. The rebuild is heavier and more thickly buttressed, the church was never completed, and the crossing tower came down too in 1573. Medieval builders had no structural calculations. They had precedent, proportion and rules of thumb, and the way you found out you'd gone too far was that it fell over.
Iron and steel: the wall stops working
For thousands of years the outside wall did two jobs — keeping weather out and holding the building up — and everything about a façade was a compromise between them. A bigger window meant a weaker wall. Then that link was cut.
Cast and wrought iron came first, in mills and railway stations, where fire risk and enormous spans made masonry impractical. The Crystal Palace of 1851 is the shock exhibit: an exhibition building around 560 metres long, made from standardised cast-iron parts and glass, assembled in about nine months and then taken down and moved. It was prefabricated in a way that wouldn't seem out of place today.
Steel was better than iron in every direction and got cheap in the second half of the nineteenth century. Frame a building in steel and the loads run down the columns; the outside wall is now a curtain hung on the structure, keeping out rain and doing nothing else. The Home Insurance Building in Chicago, finished in 1885, is the usual candidate for the first of these, and Chicago had an unusual amount of rebuilding to do after the fire of 1871.
Once the wall is only a raincoat, the pattern of any modern façade follows: regular structural bays, big openings, thin spandrels, and no reason at all for the ground floor to be thicker than the top.
Nobody climbs eleven flights
A frame will let you build high. It won't make anyone go up there.
Before mechanical lifts, buildings topped out at around six or seven storeys, and the value of a floor fell as you climbed. The good rooms were on the first floor. The attic was for servants and storage, because a landlord can't charge much for a room reached by ninety steps.
What changed that was a safety device rather than a lift. Hoists already existed; the fear was the rope breaking. Elisha Otis demonstrated a braking mechanism in 1854 that gripped the guide rails if the rope let go, and did it in public by having the rope cut while he stood on the platform. Passenger lifts followed within a few years. The economics of a tall building inverted — top floors became the expensive ones — and they've stayed inverted ever since.
Services shaped the plan just as much. An office lit only by daylight can't be much more than about 12 metres deep before the middle goes gloomy, which is why late-Victorian and Edwardian commercial buildings are so often narrow wings, E and H shapes, or blocks wrapped around a light well. Reliable electric lighting and then air conditioning removed that constraint. Willis Carrier's work on controlling humidity dates from 1902, and once you could light and cool the middle of a deep floor, buildings could be fat, sealed and windowless in the centre. The blank, deep-plan office slab isn't a fashion. It's what happens when daylight stops being the limiting factor.
Reinforced concrete and one lucky coincidence
Concrete has the same problem as stone: strong in compression, feeble in tension. Steel is the reverse. Put steel bars where the tension is and you get a material that does both, which is the idea behind reinforced concrete, developed commercially through the 1880s and 1890s, with François Hennebique's system among the most widely used.
Two bits of luck make it work at all. Steel and concrete expand at nearly the same rate when heated, so they don't tear each other apart through a normal temperature swing. And fresh concrete is strongly alkaline, which passivates the steel inside it and stops it rusting. Take that protection away — through cracks, or carbonation, or chloride from road salt and sea air — and the steel corrodes, swells, and blows the concrete off in sheets. Every crumbling 1960s car park is that reaction running to completion.
What concrete gave architects was continuity. A steel frame is an assembly of separate members; a concrete frame is poured as one piece, so floors can cantilever past their supports with nothing underneath the edge. Le Corbusier set the consequences out plainly in the 1920s: lift the building on columns, put the columns inside the perimeter, and both the plan and the façade come free. The horizontal ribbon window that reads as pure style is a straightforward result — if the outside wall carries nothing, a window can run the entire length of the building without stopping for piers.
Glass got cheap, and rather suddenly
Large flat glass used to be a luxury item. Plate glass was cast and then ground and polished on both faces, which was slow, wasteful and expensive, and cheaper sheet glass came with visible distortion. That's why nineteenth-century shopfronts show off their big panes: the pane was the point.
Pilkington's float process, made public in 1959, changed the price. Pour molten glass onto a bath of molten tin, let it spread to a perfectly flat sheet, cool it and cut it. No grinding at all. Large, flat, optically decent glass became an ordinary building material within about a decade.
The fully glazed building arrives at that moment for that reason. Whether you like the result is a separate conversation, and it isn't one the technology was having.
What was nearby, and what the weather was doing
Before railways, heavy materials came from close by, because carting stone overland was ruinous. That single constraint produced most of what people now think of as regional character: honey-coloured limestone in the Cotswolds, knapped flint where the chalk sits under the fields, brick wherever there was decent clay and fuel to fire it, granite in Aberdeen. Roofs were thatched, or tiled, or covered in whatever local stone would split thin enough.
Railways dissolved that. Welsh slate could reach the whole of Britain cheaply from the 1840s, and slate roofs went up in places with no slate for two hundred miles. Regional building traditions started to fade the moment the freight rate fell, which is much the same story as the shipping container, a century earlier and on land.
Climate wrote the rest of the vernacular, and the logic is usually visible once you know what you're looking at. Thick mud or adobe walls in hot dry places work because of thermal mass — the wall soaks up heat all day and releases it into the cold night, flattening a swing of 20 degrees or more into something liveable. Small windows in the same climate keep the sun out. Whitewash reflects it.
Hot and wet needs the opposite. Thermal mass is useless if the nights are warm too, so you want air movement instead: light walls, large openings, deep shading overhangs against sun and driving rain, and floors raised clear of damp ground and vermin. Cold places want compactness, small openings and steep roofs to shed snow. Windcatchers in Iranian towns pull air down a tower and through the house. Igloos use snow as insulation and the dome as a compression structure, which is the Pantheon's trick in a much less permanent material.
The rules that shaped the skyline
A surprising share of what buildings look like is written in regulations, usually after something went badly wrong.
London's Rebuilding Act of 1667, passed after the Great Fire, required brick or stone rather than timber, standardised storey heights and set out permitted classes of house. Georgian London's uniformity is partly a fire code. Party walls carried up above roof level, staircase widths, the number of exits from a floor — all of it descends from particular disasters.
New York's zoning resolution of 1916 is the clearest case of a rule you can see from the street. The Equitable Building had gone up as a sheer mass and thrown a vast shadow over its neighbours, so the city required towers to step back as they rose, in proportion to the width of the street. The result is the wedding-cake silhouette of Manhattan between the wars. Architects didn't decide the setbacks were beautiful. They were drawing the envelope the law allowed and then making the best of it.
Height limits did the same elsewhere. Paris held its buildings to a narrow range of heights with rules on street width and roof angle, which is why the centre reads as one continuous six or seven storey wall of stone with zinc roofs on top. London capped ordinary buildings around 80 feet for decades, a limit tied to what fire ladders could reach.
None of this settles whether a building is any good. It does mean that when a building looks odd, the question worth asking first isn't what the architect was thinking. It's what was holding it up, what could be got to the site, and what the rules would allow.
