History of Architecture

Before you read

Two minutes with a biscuit will make the rest of this lesson land harder. Take a plain digestive or a shortbread finger, hold it at both ends, and press slowly in the middle with your thumb until it snaps. Now look at where the crack started. It did not start on the top face, under your thumb, where the pressure was. It started on the underside, opened upward, and the top face was the last part to let go. Hold on to that image. Stone does exactly the same thing at a scale of tonnes, and the first two thousand years of monumental architecture are builders working around it.

Twenty-five tonnes, and a beam you could lie down on

Between roughly 2600 and 2400 BCE, on chalk downland in what is now Wiltshire, people dragged in thirty shaped blocks of sarsen, a very hard silicified sandstone, and set them upright in a ring about 30 metres across. Each upright stands about 4.1 metres above ground, is 2.13 metres wide and 1.06 metres thick, and weighs on the order of 25 tonnes. Across the tops they laid a continuous ring of lintels, each about 3.2 metres long, 1 metre wide and 0.8 metres thick. Inside the ring they raised five trilithons in a horseshoe 13.7 metres across, the tallest almost 7.5 metres high, with uprights reaching perhaps 50 tonnes.

Look at the proportions of one lintel at Stonehenge. It is 3.2 metres long and 0.8 metres deep. That is a beam roughly four times as long as it is deep, made of one of the strongest building stones in Britain, and it carries nothing at all except itself. Four thousand five hundred years later, a builder handed a steel section of the same proportions would use it to span a garage. The people who quarried, dressed, moved and lifted 25 tonnes of sarsen were not short of ambition or of labour. They were short of tension.

Key idea: A lintel that short, on supports that big, is not a design decision. It is a material limit showing through the design.

Why stone refuses to be a beam

Squeeze a block of limestone and it will take something like 30 to 60 megapascals before it crushes; granite will take three or four times that. Pull the same limestone apart and it lets go at perhaps 2 to 5 megapascals. The ratio is the whole story: stone is roughly ten to twenty times weaker in tension than in compression. That is a consequence of how it is made. Stone is a mass of grains and crystals held by cementing minerals, threaded with bedding planes and microcracks; pressing the assembly together closes those flaws, while pulling it apart opens them, and a crack that opens has a sharp tip that concentrates stress and runs.

Now put that stone into a beam, which is what a lintel is. Support it at both ends, load the middle, and it bends. The top surface shortens: compression. The bottom surface lengthens: tension. Somewhere in between is a surface that does neither, the neutral axis. Your biscuit told you which face governs. So does every fallen architrave in the Mediterranean, cracked upward from the soffit. If you want this worked through with diagrams and hand experiments, it is the content of the lesson on loads, compression and tension in ARCH 101, Introduction to Architecture; here we want the number it produces.

The number, and where it comes from

Follow this arithmetic. It is four lines long and it explains three thousand years of floor plans.

A simply supported beam carrying a uniform load develops its largest bending moment at midspan: M = w L2 / 8, where w is the weight of one metre of beam and L is the clear span. That moment produces a stress at the bottom face of M divided by the section modulus, and for a rectangle of width b and depth d the section modulus is b d2 / 6. Substitute, then write the self weight as w = rho g b d, where rho is density. The width cancels:

sigma = 3 rho g L2 / (4 d)

Read what just fell out of the algebra. Widening a stone beam does nothing. A wider lintel has more capacity and precisely as much more weight, and the two cancel exactly. Depth helps, and only linearly. Span hurts, and it hurts as the square. Double the span and you quadruple the stress at the underside.

Put numbers in. Take a sound limestone at 2400 kilograms per cubic metre with a tensile strength of 3 megapascals, and make the lintel a full metre deep. Solve for the span at which self weight alone cracks it: L2 = 4 d f / (3 rho g), which gives about 170, so L is about 13 metres.

Thirteen metres, and then start subtracting. That figure assumed the beam carries nothing but itself, and a real architrave carries a frieze, a cornice, a roof and whatever the roof carries. It assumed the stone is flawless, and quarried stone is not; one bedding plane running the wrong way through the tension zone and the effective strength collapses. It assumed you are willing to build at the exact point of failure, which no one who has watched a lintel let go ever is, so divide by a factor of four or five. What survives the subtractions is a working clear span somewhere between three and six metres, with the boldest ancient stone lintels reaching eight or nine.

The point: The span limit of stone is not a vague notion that stone is heavy. It is a specific number, in metres, and you can walk into any pre-Roman monumental building and find that number written in the spacing of the supports.

Karnak: 134 columns because the ceiling would not reach

Go to the Great Hypostyle Hall at Karnak, begun under Seti I and finished under Ramesses II in the nineteenth dynasty, around 1290 to 1224 BCE. The hall covers about 5,000 square metres. Into that area the builders put 134 columns in sixteen rows. The twelve columns of the central nave rise about 24 metres and measure some 10 metres around, which is a shaft over three metres thick. The remaining 122 columns stand about 14 metres.

Stop and ask the obvious question, because most visitors do not. Why 134? A room of 5,000 square metres roofed with modern steel would need perhaps a dozen supports, and roofed with a long-span truss, none at all in the middle. The answer is in the paragraph above. Every roof slab and architrave at Karnak is stone, and stone will not reach. So the supports have to be close enough together that no beam ever exceeds its limit, and the columns have to be fat enough to look, and to be, entirely untroubled by the load they carry. The result reads to a modern eye as overwhelming forest-like grandeur, and it genuinely is overwhelming. But the forest was not chosen for the effect. The effect was the by-product of a span limit.

The one place where Egyptian builders spent their span allowance deliberately is the nave. Raising twelve columns to 24 metres while the neighbours stop at 14 opens a band of wall above the side roofs, and into that band they cut grilled stone windows. That is a clerestory: the only way to get daylight deep into a stone-roofed hall, and it costs a great deal of stone to buy. The Romans and then the Gothic builders would spend enormous ingenuity on the same problem, and you will meet their answers in Lessons 3 and 7.

What the limit felt like from inside

Structural facts become spatial experiences, and this is the lesson's real payoff. A building whose supports must sit four or five metres apart cannot give you a clear view across a room. At Karnak you stand in a hall the size of a cathedral and you cannot see the far wall; your sight line dies against the third column. Movement becomes a matter of threading between masses. Light arrives in slots. The building is experienced as a sequence of near things rather than as a single volume.

Compare that with what you already know from ARCH 101 about procession and compression and release. Egyptian temple architecture makes procession from necessity: the axis is the only direction in which the space opens up, because along the axis the columns line up and you can see through. A visitor walking the axis at Karnak is walking the one line the structure permits, and priests knew exactly what that was worth.

Worth holding on to: Post and lintel produces rooms full of supports, and rooms full of supports produce axial, processional, dim architecture. When you meet the arch in Lesson 3, watch what happens to all three at once.

The joints tell you these builders knew

It is tempting to file post and lintel under primitive and move on. The evidence at Stonehenge refuses. The sarsen lintels are held down by mortise and tenon joints, projecting knobs on the uprights fitting hollows cut in the underside of the lintels, and the lintels are locked end to end by tongue and groove. Those are carpentry joints, carved into 7-tonne blocks of a stone so hard that dressing it with stone mauls took an estimated many thousands of working hours. Nothing about lateral stability was left to friction.

Then look at the shaping. The uprights widen slightly towards the top, so that seen from below, where every viewer stands, they do not appear to taper. The lintels are not straight: each is curved on plan so that thirty of them make a true circle rather than a thirty-sided polygon. These are corrections applied to the appearance of a building by people who had studied how it would be seen. Twenty-one centuries later, Athenian masons would apply a comparable set of corrections to the Parthenon, and that is Lesson 2.

Common misconceptions

• Massive columns mean the builders were being cautious or ignorant. The columns at Karnak are not oversized for their load. They are sized for the beams they must catch, and the beam spacing is set by the tensile strength of stone. Slimming the columns would not have helped; the roof still could not span.
• A thicker lintel spans further. Only if thicker means deeper. Widening a stone beam changes nothing at all, as the cancelling b in the formula shows. This one is worth fixing firmly, because the same confusion produces wrong intuitions about floor joists and bridge girders for the rest of your life.
• Stone is weak, which is why the spans are short. Stone is extraordinarily strong. A granite column can carry a stress that would flatten most timber. It is weak in one specific mode, tension, and bending is the mode that puts a beam's underside squarely into it.
• Post and lintel was abandoned once the arch arrived. It never was. It is how nearly every timber house on earth is framed, and it is how a steel frame works. What changed is that timber and then steel have real tensile strength, so the same geometry buys a span of ten or twenty metres instead of four.

The workaround before the arch

Builders who wanted to cover more ground had one trick before the true arch, and it is worth naming now because you will meet it again outside Europe in Lesson 6. In a corbel arch, each course of masonry is laid slightly past the one below until the two sides meet. Nothing spans anything: every stone is a small cantilever, held down by the weight of the courses above it. The Treasury of Atreus at Mycenae, built around 1250 BCE, uses corbelling to roof a chamber 14.5 metres across and about the same in height, which is a far greater span than any Greek architrave. The price is a very steep profile and a colossal mass of stone piled over the haunches to hold each cantilever down. Corbelling buys span with material. The arch, when it comes, buys span with geometry, and that is a much better trade.

Recap

• Stone carries perhaps ten to twenty times more stress in compression than in tension, and a loaded beam puts its underside in tension.
• For a stone beam under self weight, sigma = 3 rho g L2 / (4 d): width cancels, depth helps linearly, span hurts as the square.
• The theoretical self-weight limit for a metre-deep limestone lintel is around 13 metres; real loads, real flaws and a real factor of safety bring the working span down to roughly three to six.
• Stonehenge builds 3.2 metre lintels on 25 tonne uprights; Karnak fills 5,000 square metres with 134 columns; the Greek temple keeps its architraves under about 4.5 metres. All three are the same number showing through.
• The consequences are spatial: forests of supports, axial views, slot light, and clerestories bought at great expense in stone.
• Corbelling evades the limit by turning every stone into a cantilever, at a heavy cost in material.

Sources

1. Wikipedia contributors. (n.d.). Stonehenge. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Great Hypostyle Hall. Wikipedia. en.wikipedia.org
3. Kostof, S. (1995). A history of architecture: Settings and rituals (2nd ed., rev. G. Castillo). Oxford University Press. Chapters 2 and 4 on Egypt and the megalithic world.
4. Ching, F. D. K., Jarzombek, M. M., & Prakash, V. (2017). A global history of architecture (3rd ed.). Wiley. Sections on 2500 BCE and 1500 BCE.

A building with almost no repeated blocks

Work began on the Athenian Acropolis in 447 BCE and the temple was substantially finished in 432, with the sculpture running on past 431. The architects were Iktinos and Kallikrates; Pheidias oversaw the sculptural programme. The stylobate, the top step on which the columns stand, measures about 69.5 by 30.9 metres. Eight columns stand across each end and seventeen along each flank, each roughly 10.4 metres tall and about 1.9 metres across at the foot, built up from drums of Pentelic marble quarried sixteen kilometres away. The accounts were published on stone for the citizens to read.

Here is the fact that should stop you. Almost none of those blocks is a repeat of another. The steps are not flat. The columns are not straight, not vertical, and not all the same diameter. The walls lean. The entablature above follows the curve of the steps. Every one of those departures had to be surveyed, drawn, cut and checked individually, on a building of some 13,400 blocks, at a cost in mason-hours that is difficult to overstate. Somebody decided that was worth paying for. This lesson walks the decisions one at a time, with the numbers, and then asks the harder question of what they were for.

Refinement one: the floor is a low dome

Stand at one corner of the stylobate and sight along the top step towards the far corner. It rises. The platform bulges upward at its centre by about 10.3 centimetres over the 70 metre length, a rise of roughly one part in 700, and it curves on the short ends too. This is not settlement and it is not sloppiness; the curve was set out deliberately, and it propagates upward through the whole building, because the architrave, the frieze and the cornice all follow it. A block sitting at the third position along the flank is a different shape from a block at the tenth.

Try the arithmetic of what that means for the workshop. Every step block, every column base, every architrave has a slightly different geometry from its neighbour, and each has to arrive on site correct, because a marble block that is wrong is scrap. The Greek masons cut the contact faces using a technique called anathyrosis, dressing a narrow smooth band around the edge of each joint face and hollowing the middle slightly, so the visible joint could be made hairline without having to true up the whole surface. Iron cramps set in molten lead tied the blocks laterally.

The upshot: The curvature is not a detail. It is a decision to abandon repetition, and repetition is the single greatest economy available to anyone building in stone.

Refinement two: the columns swell

A Doric column tapers from bottom to top. If the taper were a straight line in profile, the shaft would look pinched, as though its waist had been sucked in. The Parthenon's shafts instead follow a slight convex curve, so the profile bulges outward between the ends. This is entasis, and at the Parthenon its magnitude runs at about one part in 550 to 600 of the height, which on a 10.4 metre shaft is a deviation from the straight of under two centimetres.

Under two centimetres, spread over ten and a half metres, on a shaft two metres thick. You cannot see it as a curve. You can only see its absence, which is the entire point of the exercise: the correction is designed to be invisible while the failure to make it would not have been. Compare the heavy, obvious entasis of archaic Doric temples a century and a half earlier at Paestum, where the swelling is pronounced enough to read as a bulge. The Parthenon is what happens when a tradition has been refining one move for six generations.

Refinement three: nothing is plumb

The columns do not stand vertically. They lean inward, by something like seven centimetres over their height, so that if you extended their axes upward they would converge a couple of kilometres above the Acropolis. The walls of the cella lean inward too. The corner columns get two further adjustments: they are made slightly thicker than the rest, by roughly a fortieth of a diameter, and they are set closer to their neighbours than the standard spacing.

The thickening has a stated reason that you can verify with your own eyes on any bright day. A corner column is seen against the open sky on two of its four sides, while an intermediate column is seen against the shadowed wall of the temple behind. A bright background eats the edges of a silhouette, so an unadjusted corner column looks thinner than its neighbours. Make it thicker and it looks the same. Vitruvius reports exactly this reasoning in Book III of De architectura, written some four centuries later, and it is the only ancient explanation of the refinements we have.

The corner problem, worked

The narrowing of the corner spacing has a different cause, and it is worth doing properly, because it is the clearest case in ancient architecture of a rule system generating a contradiction that has to be settled by somebody's judgement.

The Doric order requires two things of the frieze above the columns. First, a triglyph, the block with the three vertical grooves, must sit centred over every column, and another must sit centred over every gap. Second, the frieze must end at the corner with a triglyph, not with a half-metope, because a metope sliced in half at the angle of a building looks like a mistake. Now try to satisfy both.

The Parthenon takes the last option and distributes the contraction, reducing the corner bay and, more gently, the one next to it. Look at any elevation drawing and count the interaxial spacings from the corner inward; they are not equal. This is what a mature architectural language looks like when its rules collide: not a rule that solves everything, but a convention about which rule bends and by how much.

Remember: The corner problem is not decoration going wrong. It is a formal system meeting its own edge case, and the Greek answer is to hide the discrepancy in the dimension people are least able to measure by eye.

What the refinements were for, and what we actually know

Vitruvius says the corrections exist because the eye deceives, and that the architect must add what the eye subtracts. That explanation held for two thousand years and it is still the one most books give. It is also, on its own, weaker than it looks, and an honest course says so.

Consider the curvature of the stylobate. If the aim were to stop a long horizontal appearing to sag, a rise of one in 700 is very small for the job, and there is no ancient text connecting curvature to sagging. A competing explanation is drainage: a curved platform sheds rainwater from a temple that stands in the open. That is true of the curve but does not explain why the entablature, forty feet up and shedding water perfectly well on its own, is curved to match. A third position, argued by several twentieth-century scholars, is that the refinements make the building look alive rather than inert, and that pursuing that quality is a sufficient reason without any theory of illusion at all.

What would settle it? Evidence of intention, which is what we lack. We have the building, the accounts of what was spent, and Vitruvius writing four centuries afterwards in a different language about a different building culture. We can measure the refinements to a millimetre and cannot read the mind that ordered them. That gap is normal in architectural history and it is better to name it than to paper it over.

The structure underneath all this

Now connect this back to Lesson 1, because the Parthenon is post and lintel throughout and is bound by exactly the limit we calculated. The columns on the flanks stand about 4.3 metres apart centre to centre, and each column is about 1.9 metres thick at the base, which leaves the architrave a clear span of roughly 2.4 metres. That is comfortably inside the working range for marble. It is also the reason the temple is a ring of columns around a box rather than a hall: the Greeks had no way of covering the interior in one clear volume, and the cella of the Parthenon needed an internal two-storey colonnade to carry its roof.

So the greatest refinement culture in the history of stone building spent its ingenuity on the appearance of a structural system it could not extend. Every millimetre of entasis is a decision taken inside a fixed span limit. Rome will lift that limit within four centuries, and it will do so with a technique the Greeks knew about and did not think monumental. That is the next lesson.

Common misconceptions

• There is not a straight line in the Parthenon. A good line, endlessly repeated, and false. Plenty of the building is straight: the column flutes, the joints between drums, most vertical arrises. The refinements are specific and enumerable, not a general ban on straightness.
• The refinements are proven optical corrections. They are measured facts with a contested explanation. Vitruvius offers optical correction, but he is late, and drainage and sheer vitality are live alternatives. Hold the measurements firmly and the explanation loosely.
• The temple was a study in white marble. The sculpture and much of the upper architecture were painted, in reds, blues and gilding, and the ancient viewer saw a coloured building. The white Parthenon is a modern taste projected backwards, partly by way of centuries of weathering.
• Entasis was invented at the Parthenon. It is far older and, in archaic temples, far more pronounced. The Parthenon's contribution is the restraint.

What to carry forward

• The Parthenon of 447 to 432 BCE measures about 69.5 by 30.9 metres on the stylobate, with columns some 10.4 metres tall and 1.9 metres thick at the base.
• Its measured refinements: a stylobate that rises about 10.3 centimetres over 70 metres, entasis of about one part in 550 to 600, columns leaning inward some seven centimetres, corner columns thickened by around a fortieth of a diameter, and corner bays contracted.
• The corner contraction resolves a genuine contradiction in the Doric rules between centring triglyphs on columns and ending the frieze with a triglyph.
• The refinements cost the builders repetition, which is the largest economy available in stone construction.
• Vitruvius's optical-correction account is the only ancient explanation and is not the only plausible one; drainage and the pursuit of a living surface both have advocates.
• None of it changes the structure: architrave clear spans of about 2.4 metres, and an interior that has to be subdivided because nothing can cover it.

Sources

1. Wikipedia contributors. (n.d.). Parthenon. Wikipedia. en.wikipedia.org
2. Vitruvius. (1914). The ten books on architecture (M. H. Morgan, Trans.). Harvard University Press. Book III on temple proportion and adjustment. gutenberg.org
3. Wikipedia contributors. (n.d.). Entasis. Wikipedia. en.wikipedia.org
4. Fletcher, B. (2019). Sir Banister Fletcher's global history of architecture (21st ed., M. Fraser, Ed.). Bloomsbury and the Royal Institute of British Architects. Volume 1, on Greek temple design.

A bridge that drops two and a half centimetres

Somewhere around 40 to 60 CE, Roman engineers finished a stretch of aqueduct across the gorge of the Gardon in southern Gaul. The Pont du Gard stands 48.8 metres above the river at low water and runs 274 metres along its top. It carries three tiers of arches: six in the bottom row, eleven in the middle, and forty-seven originally in the top, of which thirty-five survive. Its blocks of shelly limestone weigh up to six tonnes and were set, for the most part, without mortar. Total stone: an estimated 50,000 tonnes.

Now the number that should impress you more than the height. The water channel along the top falls 2.5 centimetres across the 456 metres of the crossing. The whole aqueduct feeding Nemausus, modern Nimes, runs about 50 kilometres and drops roughly 17 metres over that distance, an average gradient near one in 3,000. To hold that gradient the surveyors had to know where they were, vertically, to within centimetres, over tens of kilometres of broken country, using a levelling instrument called a chorobates and a great deal of patience. The arches exist to keep a nearly level line across a valley that is nowhere near level.

Why this matters: Roman arch building is not a stylistic preference for round openings. It is the structural technique that let engineers put a piece of infrastructure wherever the survey said it had to go.

What the arch actually does

Go back to the biscuit from Lesson 1. A beam bends, the underside goes into tension, and stone cracks there. An arch removes the possibility. Cut the span into wedge-shaped blocks, the voussoirs, and set them in a curve. Load the top and each wedge is squeezed against its neighbours. Nothing is being stretched anywhere. The material is used in the one mode where it is superb.

The arch charges two prices for this, and they govern everything Rome builds. The first is thrust. Loads in an arch do not travel straight down; they run round the curve and arrive at the springing pushing outward as well as down. Something has to stop that push, which is why Roman arcades have such heavy piers and why an aqueduct's end arches sit against solid rock or an enormous abutment. Remove the neighbour of an arch in a row and the row unzips.

The second price is temporary. An arch is not stable until the last voussoir, the keystone, is in place, so until then the whole ring must be held up by a timber former called the centering. On the Pont du Gard you can still see the projecting stone corbels left in the piers to carry that timber, and the builders sized many arches so the same centering could be struck, moved along and reused. Timber, not stone, is the scarce resource in an arched building programme. Keep that fact: it is the whole drama of Lesson 9, in Florence, thirteen centuries later.

Extrude an arch along a line and you get a barrel vault, which roofs a corridor but pushes outward along its entire length, so both side walls must be continuous and thick. Cross two barrels at right angles and you get a groin vault, and something valuable happens: the thrust collects into the four corners. Between those corners the wall carries nothing, so it can be opened for windows or removed altogether. A hall covered in groin vaults on piers is a hall with clear floor and daylight, and Rome built them at enormous scale in the imperial baths.

The material that made it cheap

The arch was old. The Etruscans built them, the Mesopotamians built them in mud brick, and the Greeks knew perfectly well how they worked and largely reserved them for drains and gates. What Rome added was opus caementicium, and the addition changed the economics of building.

The recipe is lime, water, volcanic ash, and rubble. Burn limestone to make quicklime, mix it with pozzolana, the ash from the volcanic fields around Pozzuoli on the bay of Naples, and the silica in the ash reacts with the lime to form a hydraulic binder, one that sets by chemical reaction rather than by drying and will set under water. Into that mortar the builders bedded caementa, chunks of broken stone or brick, by hand, in courses. That last point matters and is usually got wrong: Roman concrete was laid, not poured. There was no slump, no vibration, no formwork full of a flowing liquid. Men placed mortar and men placed stones in it, layer on layer, inside a permanent facing of brick or small stonework that also acted as the mould.

Because it was laid rather than cut, concrete work needed unskilled labour rather than masons. Because it was monolithic once set, a concrete vault does not depend on the joints between its blocks. And because the aggregate is chosen by the builder, the density of the finished material can be varied deliberately from one part of a structure to another. Hold that last property. It is the answer to the Pantheon.

The Colosseum: an arch used as a corridor

Between about 70 and 80 CE, Vespasian and then Titus built the amphitheatre we call the Colosseum: 189 by 156 metres on plan, 48 metres to the top of the outer wall, an arena floor 87 by 55 metres, over 100,000 cubic metres of travertine in the outer structure, held with some 300 tonnes of iron clamps. Ancient sources claim 87,000 spectators; modern estimates settle nearer 50,000.

The interesting engineering is not the seating bowl. It is that the building can be emptied. Eighty arched openings ring the ground floor, seventy-six of them for ordinary spectators and numbered so that a ticket, a numbered pottery shard, sent you to one specific entrance. From each entrance, barrel-vaulted radial passages climb between annular corridors and discharge through openings, the vomitoria, directly into the tier your seat was in. There is no lobby, no bottleneck, and no need for anyone to cross anyone else's path. Fifty thousand people could clear the building in a few minutes.

Read that as an architectural argument. The arch is not just a way of spanning a hole in a wall; repeated in three dimensions, it is a way of building circulation. Every corridor, stair and passage in the Colosseum is a vault, and vaults can climb, curve and cross without any of the beams that a stone-and-lintel building would need and could not have. When you meet the Gothic cathedral in Lesson 7 and the steel-framed department store in Lesson 12, you will be watching the same argument continue.

The Pantheon, and what it is actually made of

Around 126 CE, under Hadrian, the Pantheon was completed on the site of Agrippa's earlier temple, whose inscription it still carries. Its interior is a sphere caught inside a cylinder: the dome's diameter and the height from floor to the eye of the dome are both 43.3 metres, so a ball of that diameter would touch the floor and the crown. The oculus, the open hole at the top, is about 8.9 metres across, or thirty Roman feet, and has never been glazed. The portico's granite column shafts are 11.9 metres tall, weigh about 60 tonnes each, and were quarried in Egypt. The dome is still the largest unreinforced concrete dome in the world, nineteen centuries after it was struck.

It stands because of four decisions, and none of them is mysterious.

1. Graded aggregate. The concrete is not one material. Lower down, the aggregate is heavy travertine, around 2,200 kilograms per cubic metre. As the dome rises, the builders switched to lighter stone, then brick, and near the crown to alternating layers of tuff and pumice at roughly 1,350. The dome gets lighter exactly where extra weight would do the most harm.
2. Varying thickness. The structure is about 6.4 metres thick at the springing of the dome and thins to about 1.2 metres at the oculus. Total dome weight lands near 4,535 tonnes, and the mass sits low, where it acts as its own buttress against the outward thrust.
3. Coffering. The five rings of twenty-eight sunken panels are usually described as decoration. They also remove a large volume of concrete from the shell, and they remove it from the middle of the thickness where it contributes least, leaving a ribbed surface behind.
4. The oculus, and the wall below. A masonry dome wants to crack radially near its base, where the material is being stretched around the circumference. The thick lower zone and the eight massive piers hidden inside the drum, linked by brick relieving arches you can see in the exterior brickwork, take that hoop tension in compression instead. The hole at the top is not a weakness: in a dome the crown region is in compression in every direction, so removing material there costs very little.

In short: The Pantheon is a piece of graded, hollowed, carefully proportioned engineering, built by people who knew what they were doing and left the evidence in the fabric.

Common misconceptions

• The Pantheon is a mystery that modern engineers cannot explain. It is one of the best-understood buildings in the world. The aggregate grading has been sampled, the thickness measured, the relieving arches photographed, the cracking mapped. What is remarkable is the confidence of the design, not any secret in it.
• The recipe for Roman concrete was lost. The chemistry was never a secret and pozzolana is still quarried. What lapsed after the western empire was the organisation, the logistics and the demand. Recent work has added something genuinely new: Seymour and colleagues argued in 2023 that lumps of unmixed lime, long dismissed as sloppy work, come from hot mixing with quicklime and let the material reseal its own cracks. That is a discovery about durability, not the recovery of a lost formula.
• Roman concrete was poured. It was laid by hand in courses, with the aggregate bedded into mortar inside a brick or stone facing. Picturing a Roman site as a modern pour gets the labour, the pace and the joints all wrong.
• The arch made walls unnecessary. It did the opposite at first. An arch pushes outward, so early vaulted buildings need more mass at the sides, not less. Only the groin vault, by gathering thrust into corners, begins to free the wall, and the full release waits for the flying buttress.

What Rome paid for it

It is worth being clear about the cost, because the next thousand years of European building is partly a story of not being able to afford this. Roman vaulted construction consumed pozzolana, which is local to specific volcanic regions; enormous quantities of timber for centering and scaffolding; a standing supply of unskilled labour, much of it enslaved; and a state capable of moving 60 tonne granite columns from Egypt to the Campus Martius. Remove any one of those and the technique becomes impractical. When the western empire's logistics went, the buildings that could still be built were smaller, and it took until the twelfth century for European builders to be regularly vaulting large spaces again, by a completely different method.

Putting it together

• An arch replaces bending with compression by cutting the span into voussoirs; it charges thrust at the springing and centering during construction.
• A barrel vault pushes outward along its whole length; a groin vault collects the thrust into four corners and frees the wall between them.
• Opus caementicium is lime plus pozzolana plus rubble, laid by hand inside a brick or stone facing, setting by hydraulic reaction rather than drying.
• The Pont du Gard holds a fall of 2.5 centimetres over 456 metres, 48.8 metres above the river, as part of a 50 kilometre line dropping about 17 metres in total.
• The Colosseum's eighty ground-level arches and vaulted radial passages make circulation the point: fifty thousand people out in minutes.
• The Pantheon's 43.3 metre dome works by graded aggregate from travertine to pumice, thickness falling from 6.4 to 1.2 metres, coffers that remove weight, and a thick base that resists hoop tension.

Sources

1. Wikipedia contributors. (n.d.). Pont du Gard. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Pantheon, Rome. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Roman concrete. Wikipedia, summarising Seymour et al. (2023) in Science Advances on hot mixing and lime clasts. en.wikipedia.org
4. Lancaster, L. C. (2005). Concrete vaulted construction in imperial Rome: Innovations in context. Cambridge University Press.
5. MacDonald, W. L. (1982). The architecture of the Roman empire, volume 1: An introductory study (rev. ed.). Yale University Press.

The problem, stated bluntly

Here is a geometry exercise you should try to solve before reading on. You have a square room, say 31 metres on a side, defined by four great arches on four piers. You want to roof it with a dome. A dome springs from a circle. The top of your room is a square. A circle inscribed in the square touches it at four points and leaves four corners uncovered; a circle drawn round the square covers the corners but has nowhere to sit. There is a hole in your building, four times, and it is in the shape of a curved triangle.

Rome had mostly avoided the problem by building domes on round rooms. The Pantheon's 43.3 metre dome sits on a continuous cylindrical drum, and the drum is the reason the walls are 6.4 metres thick: every point around the circumference is doing structural work. That is a fine solution and it produces exactly one kind of interior, a single centralised volume with solid walls. If you want a dome over a hall that also has aisles, galleries, a nave running east to west and daylight from the sides, you need the dome to land on points, not on a ring.

What matters here: Every dome in this lesson and the next is an answer to one question, how a circle meets a square, and the answer you pick decides what the rest of the building can be.

Two answers: the squinch and the pendentive

The older answer is the squinch. Build an arch, or a series of corbelled courses, diagonally across each corner of the square. The square becomes an octagon, and an octagon is close enough to a circle to carry a dome, or you repeat the trick to get sixteen sides. Squinches are structurally straightforward and visually obvious: you can see the little arch bridging the corner. Sasanian builders in Persia used them, and they became standard across a large part of the Islamic world.

The other answer is the pendentive, and it is worth constructing in your head step by step, because once you have built it you will never mistake it for anything else.

1. Draw your square, and draw the circle that passes through all four of its corners. That circle's diameter is the square's diagonal.
2. Imagine a hemisphere sitting on that circle. It is too big: it overhangs all four sides of the square.
3. Slice it with four vertical planes, one standing on each side of the square. The slices cut away the overhang and leave four semicircular arches, one on each side.
4. What remains between those arches, at the corners, are four curved triangles of spherical surface, each hanging down to a point at its pier. Those are the pendentives.
5. Now cut the whole thing off horizontally at the level of the tops of the four arches. The cut leaves a circle, and that circle is smaller than the original hemisphere and centred over your square.
6. Build your dome on that circle.

The pendentive is a single continuous surface, so the transition looks seamless rather than patched, and it delivers the whole load of the dome down four narrow points. Between those points there is no structural obligation at all, which is why the sides of a pendentive-domed building can be opened into arches, galleries and windows.

Justinian's church, 532 to 537

In January 532 the Nika riots burned down the second church of Holy Wisdom in Constantinople. On 23 February, Justinian began a replacement, and he did not hire architects in the builder tradition. He hired Anthemius of Tralles, a mathematician who had written on conic sections and burning mirrors, and Isidore of Miletus, a geometer who had edited Archimedes. The building was dedicated on 27 December 537, five years and ten months later, which for a structure of this size is a startling pace.

The result is Hagia Sophia. The main dome spans somewhere between about 31 and 33 metres depending on which axis you measure, since it is not truly circular, and its crown stands 55.6 metres above the floor. Forty ribs run from the crown to the base, and between their feet sits a ring of forty windows, so the dome appears to be separated from everything beneath it by a band of light. Procopius, writing while the mortar was still young, said it seemed not to rest on masonry but to be suspended from heaven by a golden chain. That is propaganda, and it is also an accurate description of what the ring of windows does to your perception.

The windows are cheaper than they look. A ribbed dome sends its load down the ribs, and the windows sit in the panels between them. Undercutting the base with openings is a legitimate structural move, provided the ribs land on something solid, and it is a move Gothic builders will repeat at much greater scale.

7 May 558: the dome comes down

Earthquakes struck in August 553 and December 557. On 7 May 558 the eastern part of the dome collapsed into the church, taking with it the ambo, the altar and the ciborium. Nobody had done anything wrong in the sense of misbuilding; the design itself had been too ambitious in a specific and diagnosable way.

The original dome was shallow. It continued more or less the curve of the pendentives, so it was closer to what we would call a sail vault than to a hemisphere, and a shallow dome is a hard thing to hold. Think of the thrust line: the load has to travel from the crown to the supports, and the flatter the path, the more horizontal the force at the bottom. Roughly speaking, the horizontal thrust varies inversely with the rise. Halve the rise of a dome and you double the outward push at its foot. Meanwhile the great piers, carrying that push, had leaned outward measurably during and after construction, and the semi-domes to east and west had deformed.

Justinian sent for Isidore the Younger, nephew of the original geometer, and the repair tells you the diagnosis was correct.

The rebuilt church was rededicated on 24 December 562. The story does not stop there: an earthquake in 989 brought down the western dome arch, and the Armenian architect Trdat repaired it with fifteen ribs; in 1346 the eastern arch and part of the dome fell and took six years to make good; and in 1317 Andronikos II had already added four large external buttresses. A building of this ambition in an earthquake zone is not a finished object but a maintained one.

Bottom line: The 558 collapse is the clearest early demonstration in architecture that the geometry of a dome, not just its material, decides whether it stands, and that rise is the variable to buy.

Where the thrust goes

Follow one unit of load from the crown of the dome down to the ground, because the plan of Hagia Sophia is essentially a diagram of that journey. From the crown, load runs down a rib to the base ring. From the base ring it enters the four great arches and the four pendentives, and the pendentives funnel it to the four main piers. Those piers are enormous, built of ashlar rather than the brick used elsewhere, and they are the only parts of the building that go straight to the ground carrying everything.

The horizontal component has to be dealt with in two different directions, and Hagia Sophia handles them by completely different means, which is why the building looks the way it does.

• East and west: two half domes take the thrust and pass it on to smaller exedrae and then to further masonry, stepping the load down in stages along the length of the church. This is what produces the extraordinary interior, a nave that seems to expand outward in every direction as you look east.
• North and south: there are no half domes, so the thrust is taken by heavy buttress piers and, above the galleries, by masonry that reads externally as great blank shoulders. The tympanum walls under the northern and southern arches carry no dome load at all, which is exactly why they could be filled with windows.

Compare this with the Pantheon one last time. Rome put a dome on a wall and got a perfect, sealed, singular room. Byzantium put a dome on four points and got a room that opens outward, admits light at the base of its dome, and can be extended by adding more vaulted volumes on the axis. The pendentive is the reason the great Ottoman mosques of the sixteenth century, Sinan's Suleymaniye above all, could develop the same idea further still.

Common misconceptions

• A squinch and a pendentive are two words for the same thing. They are different geometries. A squinch bridges a corner with an arch or corbelling, converting a square to an octagon; a pendentive is a continuous spherical triangle that carries the load to a point. If you can see a small arch across the corner, it is a squinch.
• The dome you look at today is the one Anthemius built. It is not. The present dome dates from the repair completed in 562, with substantial rebuilding after 989 and 1346, and it is higher and more steeply curved than the original.
• The ring of windows is a structural miracle. It is a competent use of a ribbed dome. Load runs down the ribs; the windows sit in the panels between. The apparent weightlessness is designed, not defied physics.
• The building has stood untouched for fifteen centuries. It has been buttressed, re-vaulted, tied and repaired repeatedly, and much of what holds it up is later than Justinian.

The short version

• A dome springs from a circle; a room is usually square; the squinch and the pendentive are the two classical resolutions.
• A pendentive is what is left when a hemisphere on the square's diagonal is cut by four vertical planes and one horizontal one: four spherical triangles delivering load to four points.
• Hagia Sophia was built 532 to 537 by Anthemius of Tralles and Isidore of Miletus; its dome spans roughly 31 to 33 metres and rises 55.6 metres above the floor.
• The original shallow dome failed on 7 May 558 because a flat dome thrusts hard; Isidore the Younger rebuilt it about 6.25 metres higher, ribbed and lighter, and it was rededicated in 562.
• Thrust leaves the building east and west through half domes and exedrae, and north and south through heavy buttress piers, which is why the north and south walls can be glazed.
• Putting the dome on four points instead of a ring is what makes an aisled, galleried, side-lit interior possible at all.

Sources

1. Wikipedia contributors. (n.d.). Hagia Sophia. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Pendentive. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Squinch. Wikipedia. en.wikipedia.org
4. Mainstone, R. J. (1988). Hagia Sophia: Architecture, structure and liturgy of Justinian's great church. Thames and Hudson.
5. Procopius. (1940). Buildings (H. B. Dewing, Trans.). Loeb Classical Library, Harvard University Press. Book I, on the church of Holy Wisdom.

Three problems, three devices

This lesson is built around a table, and the table is the argument. Islamic builders working between the eighth century and the seventeenth faced three separate problems that had nothing to do with each other, and produced three devices that are usually filed under ornament. All three are structural or geometric solutions first, and their beauty is what the solution happened to look like.

Cordoba: what to do when your columns are too short

In 785 Abd al-Rahman I began a congregational mosque in Cordoba on the site of an earlier church. He had money, a capital to dignify, and a supply of building material lying around the province in the form of Roman and Visigothic ruins. What he did not have was tall columns. Spolia, reused columns, come in whatever length the previous building used, and the shafts available in Andalusia ran to roughly three metres. A prayer hall roofed at three metres is a crypt.

The solution at the Great Mosque of Cordoba is one of the cleanest pieces of structural thinking in the early Middle Ages. Set the reused column on its base. Spring a horseshoe arch from its capital, tying it to the next column, which handles the lateral stability of the whole grid at low level. Then, instead of stopping, carry a masonry pier up from the capital, past the horseshoe arch, and spring a second, semicircular arch at a much higher level between those piers. The roof sits on the upper arcade. You have doubled your ceiling height using columns you could not lengthen, and you have done it with two rings of small, cheap, repeatable arches rather than one enormous one.

The voussoirs alternate red brick and pale stone, which reads today as decoration and began as construction: brick and stone in alternating courses is a normal way to build an arch when your stone supply is irregular, and the striping makes the geometry legible from the floor. By the time al-Mansur finished the last enlargement in 988, the hall had grown to nineteen aisles, roughly 180 by 130 metres overall, with something like 850 columns still standing today.

Why this matters: The famous forest of striped arches at Cordoba is what a supply constraint looks like when it is solved brilliantly rather than accepted.

The ribbed domes of al-Hakam II, and a date worth remembering

Between 961 and 971 al-Hakam II rebuilt the area in front of the mihrab, the maqsura, and roofed its bays with ribbed domes. Look carefully at how the ribs run. They do not all pass through the centre, as radiating ribs in a Roman or Byzantine dome would. Instead pairs of ribs spring across the square, crossing each other to leave a clear square or octagonal space at the crown, into which a smaller scalloped cupola is set. The dome is carried on eight or more intersecting arches, and the shell between them is thin.

That is a rib vault, in the sense that the ribs are the structure and the panels are infill, and it is being built in Andalusia in the 960s. European rib vaulting begins at Durham around 1093, and Gothic rib vaulting proper after 1140. Whether there is a direct line of transmission is argued over and the evidence is thin, so do not claim one. What you can say confidently is that the idea of a vault whose load runs in identifiable lines, with light panels between, was in use in Islamic Spain well before it appears in northern France.

What a muqarnas actually is

A muqarnas looks, on first sight, like a mathematical hallucination: a honeycomb of hundreds or thousands of small concave cells stepping down from a vault, each catching light differently. The dome of the Hall of the Two Sisters at the Alhambra runs to several thousand cells.

Structurally, it is far simpler than it looks, and the honest description is worth having. A muqarnas is a set of small niche-shaped units, each with a curved profile, stacked in tiers so that each tier corbels inward a little from the one below. In plan the units follow a strict geometric scheme, usually a grid of squares and rhombs, so that the whole assembly closes exactly. Most muqarnas are made of plaster, or of brick faced in plaster, or of timber, and are hung or applied on a load-bearing core of ordinary masonry behind. Some early Iranian examples do work as corbelled squinches carrying real load; the great majority of what you will see, and virtually all of the Alhambra's, does not.

So what is it for? Three things. It converts one geometry into another by a great many small steps rather than a few large ones, which reads as gradual rather than abrupt. It multiplies surface, and therefore multiplies the edges that catch light, which is why a muqarnas vault appears to shimmer and dissolve rather than to enclose. And it is generated by a rule: given a unit and a plan grid, the entire vault can be laid out on paper and cut in the workshop, which is how something this intricate could be built repeatably at all.

The Taj Mahal and the problem of two audiences

Shah Jahan began the Taj Mahal at Agra in 1631 for Mumtaz Mahal, who had died in childbirth. The mausoleum was substantially complete by 1648 and the complex by 1653. Ustad Ahmad Lahori headed the board of architects. The tomb stands on a square plinth 95.5 metres on a side and six metres high, with four minarets over 40 metres tall set at its corners, each leaning very slightly outward so that if one fell it would fall away from the tomb.

The dome is the lesson. Its inner shell is about 18.4 metres across and rises about 23 metres above the chamber floor; outside, a bulbous onion dome sits on a twelve metre drum and carries a gilded finial 9.6 metres tall. The two are not the same surface. There is a void between them.

Ask why anyone would build a dome twice. A single-shell dome must satisfy two incompatible clients. Inside, the room wants a dome whose height is comfortable in relation to the floor, because a dome that soars from a small chamber makes the chamber feel like the bottom of a well. Outside, the building wants a silhouette that reads from a distance across a garden, and that means something tall, and preferably swelling, so it is not lost against the sky. Choose one and you lose the other. The double shell lets a designer choose both: a low, calm inner dome for the room, and above it a taller profile carried on its own structure, with the space between hidden.

The Taj is not the first double dome, and the idea travels. Timurid builders in Central Asia and Iran were doing it in the fifteenth century, and you will meet a European version in Lesson 9 where Brunelleschi uses a double shell in Florence for a different reason again, to reduce weight and to give his workmen somewhere to stand.

Remember: A double-shell dome is not a decorative trick. It is the recognition that a building has an inside and an outside and that these are separate design problems.

Common misconceptions

• Muqarnas hold up the vault. Almost never. They are usually plaster or timber cells applied to a masonry core that does the carrying. Calling them structural overstates them; calling them mere decoration understates the geometry. They are a system for making a transition legible and for turning surface into light.
• The horseshoe arch is an Islamic invention. It is not. Horseshoe arches appear in Roman-period and Visigothic buildings in Iberia before the conquest of 711. What Andalusian builders did was adopt it, refine its proportions, and make it a signature.
• The complexity of Islamic geometry is mystical rather than practical. The patterns are constructed with compass and straightedge from a small number of repeating units, precisely so that craftsmen could set them out on site and in the workshop without error. A rule that generates a pattern is a production technology.
• The Taj Mahal's dome is solid marble. It is a brick and rubble structure faced in marble, in two shells, with a void between them. Almost no large dome anywhere is a solid mass of the material you can see.

Reading the table back

Return to the three rows. Cordoba solves a materials problem by stacking arches; the muqarnas solves a geometric transition by multiplying small steps; the Taj solves a conflict between interior and exterior by building the dome twice. None of the three is decoration in origin, and all three became, in the hands of the people who used them, unmistakable. That is the general pattern worth carrying out of this module: in the history of architecture, the ornaments you remember are usually solved problems that have been made beautiful on purpose.

What to remember

• Cordoba's double-tiered arcade exists because reused columns were about three metres tall and the hall needed to be twice that; horseshoe arch below, semicircular above, on piers carried past the capitals.
• Alternating brick and stone voussoirs are a construction habit before they are a visual signature; the hall reached about 180 by 130 metres and roughly 850 columns.
• Al-Hakam II's maqsura domes of 961 to 971 use intersecting ribs that leave the crown clear, a rib vault in Andalusia well before rib vaulting appears in northern Europe.
• A muqarnas is tiers of small niche cells, usually in plaster or timber on a load-bearing core; it makes a transition gradual, multiplies light-catching surface, and is generated from a strict plan grid.
• The Taj Mahal, 1631 to 1653, has an inner dome about 18.4 metres across and roughly 23 metres high, and a separate outer onion dome on a 12 metre drum with a 9.6 metre finial.
• The second shell exists because the interior and the skyline want different domes, and a designer with two shells no longer has to choose.

Sources

1. Wikipedia contributors. (n.d.). Mosque-Cathedral of Cordoba. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Taj Mahal. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Muqarnas. Wikipedia. en.wikipedia.org
4. Ettinghausen, R., Grabar, O., & Jenkins-Madina, M. (2001). Islamic art and architecture, 650-1250 (2nd ed.). Yale University Press.
5. Ching, F. D. K., Jarzombek, M. M., & Prakash, V. (2017). A global history of architecture (3rd ed.). Wiley. Sections on 800 CE and 1600 CE.

An inscription on a beam, July 1937

In late June 1937, the architects Liang Sicheng and Lin Huiyin rode into the Wutai mountains in Shanxi looking for a Tang dynasty timber building, which most scholars of the day believed no longer existed in China. At Foguang Temple they climbed into the roof of the East Hall, through bat colonies and centuries of dust, and on 5 July they made out an ink inscription on a beam naming the donor. The hall dates to 857 CE. It is seven bays by four, 34 by 17.7 metres, under a great hipped roof, and its bracket sets stand about half as tall as the columns that carry them.

This lesson takes four building traditions that developed outside the Roman line and treats them as what they are: engineering. We will work one example carefully, joint by joint, and then change the inputs and see what happens to the answer.

Worked example: taking a dougong apart

Start at the ground and go up. A Chinese timber-frame hall has no load-bearing walls at all. Round timber columns stand on stone base blocks, not fixed into the ground, and the walls between them are infill: brick, mud, timber screens, whatever the region supplies. Remove every wall and the building still stands. That separation of frame from enclosure is what European architecture will not achieve reliably until the nineteenth century, and you will meet it again in Lesson 12 in Chicago.

On top of each column head sits the dougong, and here is how it assembles.

1. A large squared block, the dou, is seated on the column head. It is a bearing block: it takes point loads and spreads them onto the column below.
2. Into a slot cut across the top of that block, a bracket arm, the gong, is dropped. It is a short beam running crosswise, and its ends project past the column line.
3. On each end of that arm sits another, smaller dou. Into each of those goes another gong, running at right angles to the first.
4. Repeat. Each tier steps a little further out from the column than the tier below, and the whole assembly grows into a corbelled bracket cluster that can project a metre or more beyond the wall.
5. The purlins of the roof land on the outermost blocks. The roof load runs back down through the tiers to the column head, and the deep eaves are carried out at the same time.

Nothing in that assembly is nailed or glued. Every joint is a mortise, a tenon or a lap, cut so that the parts interlock and gravity holds them. That was a choice, not a shortage: Tang China had abundant iron and used it freely elsewhere.

Now name the four jobs the bracket set is doing at once.

That last point became official policy. In 1103 the state architect Li Jie published the Yingzao Fashi, a treatise setting out eight grades of the basic timber module, the cai, with every member's dimensions derived from whichever grade a building was assigned. Choose the grade, and the size of every bracket, beam and rafter follows. It is a building code and a parts catalogue eight centuries before either term existed, and the East Hall at Foguang corresponds closely to the seventh of its eight ranks.

Key idea: The dougong is not an ornamental cluster under a Chinese roof. It is a load spreader, a cantilever, a damper and a standard, and it is why one system of construction could serve an empire for a millennium.

Change one input: Japanese timber in an earthquake

Take the same tradition, move it to a more seismically violent archipelago with a wetter climate, and watch the answers shift. The pagoda and main hall at Horyu-ji near Nara, generally dated to around 700, are the oldest surviving wooden buildings anywhere on earth. Their survival is not luck.

A five-storey Japanese pagoda is not a stack of floors bolted together. Each storey is a largely independent timber box resting on the one below, tied loosely rather than rigidly, and up the middle runs a central pillar, the shinbashira, which in many pagodas hangs from above or sits on a base stone and carries little of the roof load. When the ground shakes, the storeys move out of phase with each other, one going left as its neighbour goes right, and the central pillar knocks against them and damps the swing. Engineers describe the behaviour as a snake dance. Very few pagodas have ever been shaken down.

Japanese carpentry pushed the joint itself further than anywhere else, developing dozens of named splices and connections that lock in tension as well as compression, cut to fit a specific pair of members and assembled without metal fasteners. And at Ise Grand Shrine the tradition takes a step that ought to reorganise your idea of what a building is: the shrine is demolished and rebuilt on an adjacent site every twenty years, to the same design, using the same techniques, and has been for some thirteen centuries. The building is a perishable object; the permanent thing is the knowledge of how to make it, which is transmitted by making it.

Change another input: no arch at all

Now remove the arch entirely and see what happens to interiors. Neither Mesoamerican nor most classical Hindu temple builders used the true arch, and the consequences are visible from the doorway.

The Kandariya Mahadeva Temple at Khajuraho, built around 1030, rises about 31 metres. Its shikhara, the tower over the sanctum, is built of dry-jointed sandstone with iron dowels and no mortar, and it is corbelled: each course steps inward over the one below, exactly as in Lesson 1. Around the main tower cluster smaller spirelets, the urushringa, which are not only a mountain-range silhouette but a way of piling mass around the base of the central tower where the corbelling most needs holding down. The sanctum itself, the garbhagriha, is a small dark cell. The tower is almost solid. A visitor's experience of the building is overwhelmingly of the outside: you circumambulate a sculpted mountain and then step into a cave.

In Mesoamerica the same constraint gives the same result by a different route. Maya builders roofed rooms with corbel vaults, and a corbel vault of practical mass rarely spans more than about three metres, so Maya palace rooms are long, narrow and dim, with walls far thicker than the space they enclose. The great platform pyramids are not enclosures at all. The Pyramid of the Sun at Teotihuacan, roughly 225 metres on each side of its base and about 65 metres high, is a rubble and adobe core faced in stone: a mountain built to raise a small temple into the sky and to give thousands of people something to look at and process around. When you ask where the architecture is, the answer is that it is the exterior, the stair and the plaza, and this is a coherent design tradition rather than an unfinished one.

Change the material: dry stone at Great Zimbabwe

Great Zimbabwe, built by ancestors of the Shona from the eleventh century and occupied into the fifteenth, is the largest stone structure in sub-Saharan Africa. The outer wall of the Great Enclosure runs roughly 250 metres and rises up to 11 metres, several metres thick at the base and tapering as it climbs, built of something like a million granite blocks laid without a gram of mortar. Inside stands a solid conical tower about 9 metres high and 5.5 metres across.

Ask how a dry-stone wall 11 metres high stays up, because there is no adhesive doing any work. Three things do it. The granite of the region weathers naturally into flat rectangular slabs along its joints, so the raw material arrives roughly brick-shaped and can be laid in true courses with a bedding face; coursed masonry interlocks and distributes load in a way that random rubble does not. The wall is battered, thick at the bottom and thinner at the top, which keeps the resultant of its own weight comfortably inside the base. And the wall is wide enough relative to its height that overturning is not a live threat: it is a gravity structure, stabilised by its own mass, exactly like a Roman abutment or a modern dam.

The core of it: Every tradition in this lesson is solving for the same four variables, the material available, the earthquakes expected, the rain expected, and the labour that can be organised. Different answers do not mean different degrees of sophistication.

Common misconceptions

• Chinese builders used no nails because iron was scarce. China was smelting iron on an industrial scale by the Han dynasty. Interlocking joinery was preferred because it can be cut in a workshop, assembled by unskilled labour on site, dismantled, repaired member by member, and because a joint that can move a little survives an earthquake better than one that cannot.
• These traditions were static for centuries. Chinese bracket sets shrink dramatically in relative size between the Tang and the Qing, becoming decorative as beam framing takes over their structural work. That is three-quarters of a millennium of visible change, and dating a hall by its brackets is a standard technique.
• Mesoamerican pyramids are tombs like the Egyptian ones. They are chiefly platforms to raise temples and to stage public ritual; some contain burials, and the whole is usually built up in successive shells over earlier structures.
• A tradition without the true arch is a tradition that had not yet worked it out. The corbel was known everywhere, the true arch was known in Mesoamerica in principle, and building traditions optimise for what they are asked to produce. Where the religious programme wants a solid sculpted mountain, an arch buys you nothing.

Where this leaves us

• The East Hall at Foguang, dated 857 by an inscription found in 1937, is a timber frame carrying a hipped roof on bracket sets half the height of its columns; walls do nothing.
• A dougong stacks blocks and arms in tiers to spread load, cantilever the eave, dissipate earthquake energy at its friction joints, and standardise production.
• The Yingzao Fashi of 1103 sets eight grades of a timber module from which every member's size follows.
• Japanese pagodas survive earthquakes by letting loosely stacked storeys move out of phase, damped by the central pillar; Ise is rebuilt every twenty years, making the technique rather than the object the permanent thing.
• Without the true arch, interiors stay small: a Maya corbel vault spans about three metres, and the shikhara at Khajuraho is a nearly solid corbelled tower over a tiny sanctum.
• Great Zimbabwe's dry-stone walls stand by coursed interlock, batter and sheer mass, using granite that weathers into slabs.

Sources

1. Wikipedia contributors. (n.d.). Foguang Temple. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Great Zimbabwe. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Dougong. Wikipedia. en.wikipedia.org
4. Liang, S. (1984). A pictorial history of Chinese architecture (W. Fairbank, Ed.). MIT Press.
5. Ching, F. D. K., Jarzombek, M. M., & Prakash, V. (2017). A global history of architecture (3rd ed.). Wiley. Sections on 800 CE, 1000 CE and 1200 CE.

Durham, 1093, and a problem with rectangles

On 11 August 1093 the foundation stone of Durham Cathedral was laid. The choir was standing by 1096, the nave walls by 1128, and the high vault was closed in 1135. What went up over those forty years included the earliest large-scale stone rib vaults in Europe and some of the earliest transverse pointed arches, and Durham is still a Romanesque building: massive, dim, with drum piers a couple of metres thick. Nobody at Durham was building a Gothic cathedral. They were solving a roofing problem, and the solutions turned out to be the components of something else.

Start with the problem, because it is not obvious until you try to draw it. You want to vault a rectangular bay, longer along the church than across it, with intersecting barrel vaults. A semicircular arch has exactly one possible height: half its span. So the narrow transverse arch across the bay rises to half the nave width, the longer arch along the nave rises to half the bay length, and the diagonal, which spans the bay's diagonal and is longer than either, rises higher than both. Three arches meeting at one crown, all wanting to be at different heights.

Romanesque builders had three bad answers. Make every bay square, which forces the whole plan. Stilt the shorter arches, standing them on vertical stubs until their crowns come up to the diagonal's, which looks strained. Or dome the vault up so the crown floats above all three, which is heavy and hard to build. Durham does versions of all of this, and the pointed arches in its transverse arcades are a builder reaching for a fourth answer.

The pointed arch is not a style

Here is the property that matters, and it is purely geometric. A pointed arch is struck as two arcs whose centres sit on the springing line, offset from the midpoint. Move the centres further apart and the radius grows and the arch gets taller and sharper; move them together and it flattens toward a semicircle. That single degree of freedom means a pointed arch can be given any height you like for any span you like.

Suddenly the rectangular bay is trivial. Draw the transverse arch, the longitudinal arch and the diagonal all rising to the same crown, each with its own curvature. No stilting, no doming, no forced square plan. Bays can be any proportion the plan wants, including the awkward trapezoidal bays of an ambulatory curving round the end of a choir, which is precisely where the earliest Gothic vaulting appears.

The pointed arch pays a second dividend. For a given span, raising the crown steepens the path the load takes down to the springing, and a steeper thrust line means a smaller horizontal push. Compare two arches over the same 6 metre opening: a semicircle rising 3 metres, and a pointed arch rising 5. The pointed one delivers substantially less outward thrust at its feet, because the same load is travelling down a steeper route. Steeper is cheaper in buttressing.

The point: The pointed arch buys geometric freedom and lower thrust at once, and neither has anything to do with how it looks.

What a rib vault separates

A groin vault is a single continuous shell, and it must be built on continuous centering: a full timber surface, following the whole vault, holding every stone until the last one is set. Timber is the expensive, slow, reusable thing on a medieval site, as it was on a Roman one.

A rib vault splits that one object into two. First you build the ribs, arches of cut stone running along the groins and around the edges of the bay. Each rib needs only its own narrow centering, a light timber former the width of the rib itself, which can be lifted, struck and moved to the next bay. Once the ribs are closed they stand on their own. Then you fill in the four curved panels, the webs, between them, and the webs are thin, often a single course of light stone, laid on small formers spanning rib to rib. In some vaults the webs are so thin you can see daylight through a broken one.

Three consequences follow. The centering bill collapses. The vault becomes a set of lines rather than a surface, so the load leaves it at four identifiable points instead of along four edges, and those points can be caught by shafts running down the piers. And because each rib is drawn independently, the vault will now happily cover a bay of any shape.

Saint-Denis, June 1144

Abbot Suger of Saint-Denis, a few miles north of Paris, rebuilt the east end of his abbey church and had the new choir consecrated on 11 June 1144. What he built was a double ambulatory with chapels opening off it, vaulted with pointed ribs over irregular bays, and with the walls between the chapels reduced almost to nothing, so that the whole east end reads as a continuous band of glass. Suger wrote about the result himself, in two treatises on the administration and consecration of his church, and what he wanted was light, understood theologically: a continuous luminosity that would lift the mind from the material to the immaterial.

So the driver of the experiment is not structural ambition for its own sake. It is a theology of light, being handed to masons as a brief. The masons already had ribs, from Durham and from Norman and Lombard experiments, and pointed arches, from local use and from a Mediterranean world that had used them for centuries. What they had not yet solved was the wall.

Why the wall was the real problem, and what the flying buttress does

A high vault pushes outward at the top of the wall that carries it, at the level of the clerestory, thirty metres up. Something must resist that push. There are only a few ways.

The last is the flying buttress. It is a half arch, a prop, springing from a free-standing pier outside the aisle and landing exactly where the vault's thrust arrives on the nave wall. Its job is to catch a horizontal force at high level and walk it out and down. The pier it lands on is heavy, and the pinnacle usually set on top of that pier is not a spire for effect: added vertical load steers the resultant of thrust plus weight downward into the pier, keeping the line of force inside the masonry rather than letting it slide out through the side. A pinnacle is ballast, dressed up.

Why this matters: None of the three devices does the job alone. The rib collects the load into points, the pointed arch reduces the thrust and frees the geometry, and the flyer catches what is left at the exact height where it arrives. Together they abolish the load-bearing clerestory wall, and once the wall is gone it can be glass.

Chartres after the fire of 1194

On the night of 10 June 1194 fire destroyed almost all of Chartres Cathedral except the crypt and the west front. The rebuilding that followed is where the system arrives complete.

The high vault stands 37 metres above the floor over a nave 16.4 metres wide, and the elevation has three storeys, not four. The gallery, which in earlier great churches had been doing structural work as well as holding people, is simply deleted; the triforium is squeezed to a narrow band; and everything saved is given to the clerestory, whose windows now run nearly the full height of the arcade below them. That trade is only possible because flying buttresses, designed in from the start rather than added later in a panic, are taking the thrust the gallery used to take. The cathedral holds around 167 windows, including three roses, the west of about 12 metres across and the north and south around 10.5.

Stand in the nave and you are inside the argument. The piers are slender for their height because the load is collected, not smeared. The wall above the arcade is essentially absent. The light is coloured because there is enough of it to afford being filtered. Every one of those experiences is downstream of a geometry decision about how to draw an arch.

The competition went on. Amiens, begun in 1220, reaches 42.3 metres. And then a chapter at Beauvais decided to go higher still, which is the next lesson.

How the masons knew

None of these builders had structural analysis. They had geometry, templates, rules of proportion, and the memory of what had stood. A master mason set out a vault with compasses and a straightedge, using constructions such as taking a dimension from the diagonal of a square, and passed the proportions on as ratios rather than as forces. The sketchbook of Villard de Honnecourt, compiled around 1225 to 1235, shows exactly this world: plans, elevations, a mechanical saw, a lion drawn from life, and geometric constructions for setting out. Their rules encoded experience, and experience is a reliable guide right up until you leave the range in which the experience was gathered.

Common misconceptions

• Flying buttresses are decoration, or were bolted on later to save failing buildings. Some were retrofits and are recognisable as such. But at Chartres they are part of the original design, and they are the reason the clerestory could be glazed at all. Remove the flyers from a mature Gothic cathedral and the building comes down. They are the primary structure, standing outside where you can see them, which is what makes them look optional.
• Gothic is a style the builders chose and named. The word is an insult applied much later, in the sixteenth century, meaning barbarous and northern. The builders called what they were doing the modern work, opus modernum, or the French work.
• The pointed arch was imported from Islamic architecture. Pointed arches were in use across the Islamic world well before they appear in northern France, and contact through Sicily, Spain and the crusader states was real. But no documented line of transmission has been established, and pointed forms also arise locally out of vaulting problems. Note the parallel; do not assert the causation.
• Rib vaults are stronger than groin vaults. Not necessarily. Their decisive advantage was constructional, in centering saved and geometric freedom gained, and their loads being collected onto points that could be buttressed.

Pulling it together

• A semicircular arch has one possible height for a given span, which makes vaulting a rectangular bay awkward; a pointed arch can be given any height for any span.
• A taller arch over the same span drives less horizontal thrust, so pointed also means cheaper to buttress.
• A rib vault separates structure from infill: ribs on narrow reusable centering, then thin webs, with load leaving at four points.
• Suger's choir at Saint-Denis, consecrated 11 June 1144, applies both to an irregular ambulatory in pursuit of continuous light.
• The flying buttress catches the high vault's thrust at clerestory level and walks it out to an external pier, whose pinnacle is ballast steering the force line into the masonry.
• At Chartres, rebuilt after the fire of 10 June 1194, the gallery is dropped, the vault reaches 37 metres over a nave 16.4 metres wide, and about 167 windows fill a wall that no longer has to carry anything.

Sources

1. Wikipedia contributors. (n.d.). Chartres Cathedral. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Durham Cathedral. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Flying buttress. Wikipedia. en.wikipedia.org
4. Panofsky, E. (Ed. and Trans.). (1979). Abbot Suger on the abbey church of St.-Denis and its art treasures (2nd ed., G. Panofsky-Soergel, Ed.). Princeton University Press.
5. Fitchen, J. (1961). The construction of Gothic cathedrals: A study of medieval vault erection. Oxford University Press.

Before you read

Find a piece of dry spaghetti and stand it on end on the table. Press straight down with one finger until it fails. It does not crush; it bows sideways and snaps. Now break the strand in half and repeat with the short piece. It takes far more force, and the failure is different in character. You have just demonstrated that a slender compression member fails by going sideways, and that halving its unbraced length transforms its capacity. Keep that in mind. It is the single most useful idea in this lesson, and it is what the masons of Beauvais did to their own building after it fell.

The explanation everyone gives, and why it is not good enough

Here is the version you will find almost everywhere. The chapter of Beauvais Cathedral, begun in 1225 under Milo of Nanteuil, set out to build the highest choir in Christendom. They reached a vault of about 48 metres, against 37 at Chartres and 42.3 at Amiens. In 1284 the vaults came down. Therefore Gothic had a maximum height, Beauvais exceeded it, and the collapse marks the end of the experiment.

It is a satisfying story and it fails on the first piece of evidence you check. The choir was rebuilt at the same height. Not lower. The same. It has stood, with maintenance and alarm, for more than seven centuries since. Whatever failed in 1284, it was not the number 48.

So this lesson is a debugging exercise. We have a plausible diagnosis, a building that contradicts it, and a repair whose details tell us what the thirteenth century actually concluded. Let us trace the reasoning to the point where it breaks.

Step one: masonry does not fail by crushing

Work out the stress. A limestone pier at Beauvais carries the weight of the masonry above it. Even under a great cathedral, the compressive stress in the piers is on the order of one or two megapascals. Good building limestone crushes somewhere above thirty. The material is loafing, at perhaps five per cent of its capacity. No Gothic cathedral has ever failed because its stone was overloaded in compression.

This is the crucial reframing, and the engineer Jacques Heyman made it the basis of the modern analysis of masonry. Since the material has effectively unlimited strength for the stresses involved, and since mortar joints cannot carry tension, a masonry structure is governed by geometry, not by strength. Heyman's safe theorem states it cleanly: if you can find any line of thrust that lies entirely within the masonry and is in equilibrium with the loads, the structure is safe. It does not matter whether that is the actual line. Existence is enough.

Turn that around and you have the failure mode. A masonry building fails when no such line can be found inside the material any more: when the thrust line is pushed out through a face, a hinge opens there, and enough hinges turn the structure into a mechanism that can move. Cracks in a cathedral are not cracks in the sense a modern engineer fears. They are hinges, and they tell you where the thrust line has been leaving.

The core of it: Ask of a Gothic building not is it strong enough, but can a line of thrust be drawn inside it, and how much room does that line have to move before it runs out of stone.

Step two: what actually changes when you go higher

Now we can be precise about what building tall costs, and it is not the stress in the stone.

• The thrust line has less room. A taller vault delivers its thrust at a greater height, so the buttressing system has a longer lever to control and thinner sections at the top to control it with. Any disturbance, a gust, a settlement, a summer's thermal movement, shifts the line by a certain distance. What matters is that distance as a fraction of the available thickness, and taller buildings make the fraction worse.
• Slenderness grows. The piers at Beauvais rise, free-standing, further than any before them, and as your spaghetti showed, capacity against sideways instability falls sharply with unbraced length.
• Wind stops being negligible. Wind pressure acts on the great surface of the upper walls, roof and buttresses, at maximum height, and it acts sideways, which is the one direction masonry has no reserve in. Robert Mark and Maury Wolfe, modelling the structure photoelastically in the 1970s, concluded that wind loading on the tall, slender upper works was the most likely proximate cause at Beauvais.
• Construction disturbances accumulate. Stephen Murray's study of the building's history documents differential settlement and an interrupted, improvised construction sequence, with parts of the structure loaded before others were in place to restrain them.

Note what all four have in common. None of them is about the strength of stone. All of them are about the position of a line of force in a structure whose geometry no longer gives it much room to wander.

Step three: read the repair

The most valuable evidence about a failure is what the people who saw it did next. The reconstruction after 1284 kept the height and changed the bay structure. Intermediate piers were inserted into the choir and the chevet, roughly halving the bay length, and the vaults were changed from quadripartite to sexpartite, that is, divided into six panels rather than four by an extra transverse rib.

That is a diagnosis expressed in stone. They concluded the problem was the spacing and the slenderness, not the elevation, and the corrected building has outlasted the corrected diagnosis by seven hundred years.

Step four: the other collapse, and what it teaches

Beauvais fell twice, and confusing the two events muddles the lesson. In 1569 a crossing tower and spire were completed to about 153 metres, briefly making the church the tallest structure in the world. On Ascension day 1573 the tower and the crossing vaults came down.

This second failure has a much simpler explanation, and it is a managerial one. The crossing piers had been designed for a crossing, not for a spire of that mass, and the nave that would have braced the crossing from the west had never been built. A tower needs to be tied in all four directions; here it was tied in three. Contemporaries reported cracking and movement before the collapse, and workmen left the building before it came down, which is why the death toll was low. The lesson is not about the limits of Gothic. It is that adding a large load to a structure designed for a different one, while leaving a quarter of its bracing unbuilt, is a decision, and someone made it.

What Beauvais actually is now

The building you can visit is a choir and a transept and nothing else. The nave was never built; a temporary wall closes the west end and has done so for centuries. Inside, a lattice of iron and timber bracing, installed in the twentieth century after monitoring showed continuing movement, ties the transept together. The cathedral is under permanent observation. It is simultaneously the highest Gothic vault ever built and the clearest surviving demonstration that Gothic structure is a negotiation with geometry that never quite finishes.

Common misconceptions

• Beauvais proves Gothic builders had reached the limit of their system. They rebuilt at the same height and it stands. What they reached was the limit of a particular set of proportions, spacings and buttressing, and they corrected those.
• Medieval builders had no way to reason about structure, so failures were inevitable. They had no calculus, but they had proportional rules distilled from a century of buildings, and those rules worked reliably inside the range they came from. Beauvais failed because it stepped outside that range, which is a specific and recoverable kind of error rather than general ignorance.
• Cracks mean a masonry building is failing. In masonry, a crack is usually a hinge, and hinges are how these structures accommodate movement. What matters is whether enough hinges have formed to make a mechanism, and whether a thrust line can still be drawn inside the stone.
• The 1284 and 1573 collapses are the same story told twice. They are different failures with different causes: the first in the choir, about slenderness, spacing and probably wind; the second in the crossing, about hanging an enormous unplanned tower on piers and bracing that were never meant for it.

The takeaway

• Beauvais began in 1225, its choir was finished in 1272 at a vault height of roughly 48 metres, and part of that vaulting fell in 1284.
• The rebuilt choir kept the height and halved the bay by inserting intermediate piers, changing the vaults from quadripartite to sexpartite; it has stood ever since.
• Masonry piers in cathedrals work at a small fraction of the stone's strength, so failures are failures of geometry, not of material.
• Heyman's safe theorem: a masonry structure is safe if any equilibrium thrust line can be drawn wholly within the masonry, and it fails when no such line exists.
• Height costs you room for the thrust line to move, slenderness, and exposure to wind; Mark and Wolfe's modelling points to wind, and Murray's history to settlement and construction sequence.
• The 1573 fall of the 153 metre crossing tower is a separate event with a separate cause: an unplanned load on piers and bracing designed for something else, with the nave never built.

Sources

1. Wikipedia contributors. (n.d.). Beauvais Cathedral. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Amiens Cathedral. Wikipedia, for the comparative vault height of 42.3 metres. en.wikipedia.org
3. Heyman, J. (1995). The stone skeleton: Structural engineering of masonry architecture. Cambridge University Press.
4. Murray, S. (1989). Beauvais Cathedral: Architecture of transcendence. Princeton University Press.
5. Wolfe, M. I., & Mark, R. (1976). The collapse of the vaults of Beauvais Cathedral in 1284. Speculum, 51(3), 462-476.

A hole 44 metres across, open since 1367

The Florentines had a problem of their own making. The plan for Santa Maria del Fiore approved in 1367 called for an octagonal dome over the crossing spanning about 44 metres, springing from a drum whose base sits some 52 metres above the floor. The drum was built. The dome was not, because nobody in Florence knew how to build it, and the city had committed itself in advance, publicly, to a structure it could not yet make. For fifty years the cathedral had a hole in it, covered against the weather, while a succession of committees studied the question.

On 19 August 1418 the Opera del Duomo, the cathedral works board, announced a competition for the dome and its supporting apparatus. The winner, confirmed with Lorenzo Ghiberti attached as an unwanted co-appointee, was Filippo Brunelleschi, trained as a goldsmith, who had spent years in Rome measuring ancient buildings. Construction began in 1420. The dome was closed in 1436, and it took over four million bricks.

This lesson walks his solution in order, because it is a genuine procedure: a set of moves that only work together, each one answering a difficulty created by the last.

Why centering was not available

From Lesson 3 you know that an arch or vault needs centering, a timber former holding every unit until the structure closes. Scale that up to a 44 metre octagon whose springing is 52 metres in the air. The timber would have to be either raised from the cathedral floor, meaning a forest of scaffolding four or five storeys tall carrying an enormous curved deck, or cantilevered inward from the drum, which is a structural problem at least as hard as the dome itself. The quantity of large timber required was, in the Tuscany of 1420, a serious constraint in its own right, and the cost and fire risk of such a structure were obvious to everyone on the committee.

So the design brief was not really build a dome. It was build a dome that holds itself up at every stage of its own construction. Everything that follows is an answer to that sentence.

The solution, step by step

1. A pointed profile. The dome follows a curve the Florentines called the quinto acuto, the pointed fifth: each arc is struck with a radius of four fifths of the span, giving a steep, pointed section rather than a hemisphere. From Lesson 7 you know what that buys, a steeper thrust line and much less horizontal push at the base. It also reduces the hoop tension that tries to split a dome open around its lower circumference.
2. Two shells. Brunelleschi built an inner shell around two metres thick at the springing and a thinner outer shell, perhaps two thirds of a metre, with a gap between them and a stair running up inside it. The outer shell is a lightweight weather skin and profile-maker; the inner does the structural work. Notice that this is the same device as the Taj Mahal's double dome from Lesson 5, adopted for a different reason: here the second shell saves weight and gives the workforce somewhere to stand.
3. Ribs. Eight major ribs run up the corners of the octagon, with sixteen intermediate ribs hidden between the shells, and horizontal arches tie them together at intervals. The two shells and the ribs form a single stiff assembly rather than two independent domes.
4. Chains. A dome pushes outward at its base, and masonry cannot take tension. Four horizontal chains of sandstone blocks cramped with iron encircle the dome, with a fifth made of chestnut timber between the first two. They act as tension rings, doing at the base what the mass of the Pantheon's drum did by sheer thickness.
5. Herringbone brickwork. The bricks are laid in horizontal courses, but at regular intervals along each course a brick is stood on end, running vertically, so that the coursework reads as a herringbone, spinapesce, a fish spine. This is the move most often described as a mystery, and it is not one.
6. Machines. None of this happens without lifting several million bricks 50 to 100 metres. Brunelleschi designed a reversible ox-hoist, able to raise or lower without unharnessing and reversing the animals, and a slewing crane, the castello, to position loads at height. In 1421 the Florentine Signoria granted him a three-year monopoly on a river transport vessel he had designed, an early instance of what we would now call a patent.

Why a closed ring cannot fall

Now the central idea, and it is worth building carefully because everything else hangs on it.

Think of one horizontal course of brickwork running all the way round a dome. While it is incomplete, it is a curved cantilever of brick leaning inward over nothing, held only by mortar that has not set, and it will slide or fall. The moment it closes on itself, the situation changes completely. A complete ring that is being squeezed inward cannot get smaller without the bricks in it being compressed, and brick in compression is exactly what does not yield. The ring has become a compression ring, and it is stable without any support from below at all.

So a dome can be built in horizontal rings, each one closed before the next is begun, and at no point does it need centering, because at every stage the finished part is a stack of closed rings. This works for a circular dome directly. Brunelleschi's is an octagon, which is awkward, since a polygon's courses want to behave like eight separate arches; the ribs and the horizontal connecting arches are partly there to make the octagon behave more like a ring.

Which brings us back to the herringbone. The vulnerable moment is not the finished ring, it is the ring in progress: bricks laid on a sloping bed, with wet mortar, that want to slip down the incline before the ring closes. The upright bricks interrupt that slope. They key into the course below, they break each length of coursework into short segments that cannot slide as a unit, and they hold newly placed bricks in position while the mortar takes. The herringbone is a temporary works solution built permanently into the fabric.

The upshot: Brunelleschi did not defy gravity. He found a construction sequence in which the unfinished building is always a stable object, and then invented the brickwork detail that gets each stage safely to the next.

What it cost and what it meant

The dome was closed in 1436 and consecrated on 25 March of that year. Brunelleschi died in 1446, and the lantern that weights and closes the crown, itself a necessary piece of structure since it loads the compression ring at the top, was completed by Michelozzo in 1461. The whole rises about 114.5 metres. Not one workman is recorded as having died in the main construction, which for a sixteen-year project at that height is remarkable and is usually credited to Brunelleschi's insistence on safety arrangements including netting and diluted wine.

The building also changes what an architect is. Brunelleschi was not a master mason who had come up through the lodge. He was a goldsmith who had studied Roman ruins, who won a competition on the strength of a proposal, who designed machines, who quarrelled with his co-appointee until Ghiberti was sidelined, and whose name is attached to the result. Within a century, Alberti will have written the theory that makes that role respectable, and Palladio will have published the drawings that make it exportable. That is the next lesson.

Common misconceptions

• Nobody knows how the dome was built. A great deal is known: the accounts of the Opera del Duomo survive in quantity, the fabric has been surveyed and monitored for decades, and the herringbone, the chains and the double shell are all directly observable. Specific questions remain open, particularly the exact geometric method used to guide the courses, but the structure is not a riddle.
• He built it with no support at all. The claim is narrower and more interesting: no centering spanning the void. There was scaffolding, there were platforms hung from the masonry as it rose, and there were guides for setting out. What there was not was a timber dome underneath the brick one.
• The dome is a hemisphere. It is a pointed octagonal vault on a quinto acuto profile. The pointing is structural, not stylistic, and the octagon is inherited from the 1367 design.
• The lantern is decoration. It is 750 tonnes or so of masonry deliberately loading the top of the dome. Compressing the crown ring helps hold the whole assembly together, which is why the dome was left with an opening until the lantern could be built.

What you now know

• The 1367 design committed Florence to a 44 metre octagonal dome springing 52 metres above the floor, and it stood open until the 1418 competition.
• Conventional centering was impractical at that span and height, so the brief was really to design a construction sequence that is stable at every stage.
• Brunelleschi's answer: a pointed quinto acuto profile, two shells with ribs between them, four stone and iron chains plus one of chestnut, herringbone brickwork, and purpose-built hoisting machinery.
• A closed horizontal ring of masonry is self-supporting because shrinking it would require compressing brick; the herringbone stabilises each course during the interval before the ring closes.
• Construction ran from 1420 to 1436 and used over four million bricks; the lantern, structurally necessary as a load on the crown, was finished by Michelozzo in 1461.
• The project also invents a role: an architect who wins a commission with a proposal, designs the machines, and gets the credit.

Sources

1. Wikipedia contributors. (n.d.). Florence Cathedral. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Filippo Brunelleschi. Wikipedia. en.wikipedia.org
3. King, R. (2000). Brunelleschi's dome: The story of the great cathedral in Florence. Chatto and Windus.
4. Prager, F. D., & Scaglia, G. (1970). Brunelleschi: Studies of his technology and inventions. MIT Press.

A book finished around 1452, printed in 1485

Leon Battista Alberti was a papal official, a cryptographer, a writer on painting, on the family and on horses, and an illegitimate son of an exiled Florentine banking family. Around 1452 he finished De re aedificatoria, ten books on building, and presented it to Pope Nicholas V. It was printed in Florence in 1485, three years before the first printed Vitruvius, which makes it the first printed book on architecture in history.

That fact deserves a moment. Brunelleschi's dome was a solved problem in a specific city. Alberti's book was a portable, reproducible argument about what architecture is, and print meant it could be in Antwerp and London and Krakow without anyone having to travel to Florence. From here on, architecture has a literature, and the literature is often more influential than the buildings.

Alberti's definition, and the trap inside it

Alberti defines beauty as a reasoned harmony among the parts of a body, such that nothing may be added, taken away or altered except for the worse. He calls the quality concinnitas. Note two things about the definition. It is objective: beauty is a property of the object's proportions, not of your reaction to it. And it is complete in itself: if the harmony is right, nothing else is needed.

Then he separates ornament from it. Ornament, he says, is a kind of auxiliary brightness, something attached and added rather than inherent. A beautiful building remains beautiful if you strip its ornament; an ugly one is not saved by adding any.

Look at what that commits an architect to. The real work is proportion, geometry, the relation of part to part. Decoration is second order. In 1908 Adolf Loos will write an essay arguing that ornament is a waste of labour and should be discarded altogether, and he will be treated as a revolutionary. He is finishing a sentence Alberti started, and you will meet him in Lesson 13.

Key idea: By defining beauty as proportion and ornament as an addition, Alberti made it possible, four and a half centuries later, to propose subtracting the ornament and keeping the architecture.

Alberti building

He was not only a theorist, and the buildings show a mind solving problems with the classical vocabulary rather than copying it.

• Palazzo Rucellai, Florence, from about 1450. He takes the device the Colosseum used, superimposed orders of pilasters marking each storey, and applies it to the front of a merchant's house. A Roman public monument's grammar becomes domestic, and the palace facade acquires a system rather than merely windows.
• Santa Maria Novella, Florence, 1458 to 1470. A genuine composition problem: a tall nave with a much lower, wider aisle either side, so the front is a narrow tall block flanked by low sloping shoulders. Alberti squares the lower part, puts a temple front on the upper, and joins the two with a pair of great scrolls. Volutes flanking a church front become standard for the next three hundred years because he solved that geometry once.
• Sant'Andrea, Mantua, from 1472. The facade is a Roman triumphal arch; the interior is a single enormous barrel-vaulted hall with chapels instead of aisles. It is the Roman vaulted interior recovered, and it points straight at the Baroque.

Palladio, and a book you can build from

Andrea di Pietro della Gondola was a stonecutter in Vicenza until the humanist Gian Giorgio Trissino took him up, renamed him Palladio, and took him to Rome to measure ruins. In 1570 he published I quattro libri dell'architettura, the Four Books, in Venice.

Its novelty is not the theory. It is that Book Two publishes Palladio's own designs as measured plans, elevations and sections, with dimensions written on them, drawn to a consistent convention, in a book a builder could carry. Alberti wrote a treatise with no illustrations at all. Palladio published a catalogue you could work from. That difference is why Palladian buildings exist on four continents and Albertian ones do not.

Book One contains the rule set, and it is worth having concretely. Palladio recommends seven room shapes in plan: circular, square, the square plus a third at 4 to 3, the square plus a half at 3 to 2, the square plus two thirds at 5 to 3, the double square at 2 to 1, and the rectangle on the diagonal of a square. Room heights are then derived from the plan dimensions by taking the arithmetic, geometric or harmonic mean of length and breadth. So a room 12 by 6 might stand 9 high by the arithmetic mean, about 8.5 by the geometric, or 8 by the harmonic. The rules are simple enough to apply on site and rich enough to produce variety, which is exactly what a transmissible system needs.

The Rotonda, and one honest correction

The Villa La Rotonda, begun in 1566 outside Vicenza for a retired churchman and finished after Palladio's death by Scamozzi, is a square block with a domed circular hall in the middle and four identical temple porticoes, one on each face, each reached by a broad flight of steps. It sits on a hilltop and addresses the landscape in four directions equally.

Here is the correction. Palladio put temple fronts on houses because he believed ancient houses had them, reasoning that the temple form must have descended from domestic architecture. He was wrong: Roman houses turned inward to courtyards and did not present porticoes of this kind. The most consequential motif in three centuries of domestic architecture, the columned portico on the front of a house, rests on a mistaken piece of archaeology. It is worth knowing, because it is a reminder that traditions can be founded on errors and remain perfectly good architecture.

Note also what most of Palladio's villas actually were. Not follies: working farms for Venetian landowners investing in agriculture on the mainland, with long arcaded barns, the barchesse, attached to a dignified central house. Villa Emo and Villa Barbaro are granaries and threshing floors with a piano nobile in the middle.

The afterlife, in one table

The transmission mechanism is printed books, not travel. Colen Campbell published Vitruvius Britannicus from 1715, Giacomo Leoni issued an English Palladio from 1715, and Isaac Ware another in 1738. Thomas Jefferson owned Leoni and used the Four Books as a working manual at Monticello, which he built, tore down and rebuilt across four decades.

Two things should be said plainly about that last row. First, Jefferson's most explicitly Roman building is not Palladian at all: the Virginia State Capitol of 1785 to 1788 is modelled directly on the Maison Carree at Nimes, a Roman temple, chosen because he wanted a republic to look like a republic. Second, Monticello and the plantation villas that followed it were farm estates in the Palladian sense, and the farm was worked by enslaved people, whose quarters and workrooms Jefferson arranged into wings tucked below the terraces so that they were invisible from the house's principal views. The architectural device and the social arrangement are the same decision.

Common misconceptions

• Palladio invented the portico on a house. He introduced it on the mistaken belief that ancient domestic architecture had used it. Roman houses did not.
• Palladian architecture means symmetry and columns. The columns are the visible part. The transmissible content is the proportional system: seven room shapes, heights derived by means, and a hierarchy of rooms about a central axis.
• Palladio's villas were aristocratic pleasure houses. Most were working farm centres for landowners moving capital from Venetian trade into mainland agriculture, with barns attached.
• Alberti's separation of beauty and ornament was a modest bit of theory. It is the hinge on which a great deal of later argument turns, up to and including the Modern movement's decision that ornament could simply be removed.

Summing up

• Alberti's De re aedificatoria, finished around 1452 and printed in 1485, is the first printed book on architecture and defines beauty as a reasoned harmony from which nothing can be altered but for the worse.
• Separating inherent beauty from added ornament makes proportion the architect's real subject and makes later subtraction of ornament thinkable.
• Alberti's buildings solve problems with the classical vocabulary: superimposed orders on a palace, volutes reconciling nave and aisle, a triumphal arch as a church front.
• Palladio's Four Books of 1570 publish his own designs as dimensioned plans, elevations and sections, which is why the system travelled.
• Its rule set is concrete: seven room shapes, heights by arithmetic, geometric or harmonic mean, rooms ordered about an axis.
• The Villa Rotonda's four porticoes rest on a mistaken belief about ancient houses, and the type reaches Kent by 1725, London by 1729 and Virginia by 1809, where the villa becomes a plantation house.

Sources

1. Wikipedia contributors. (n.d.). Andrea Palladio. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Villa La Rotonda. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). De re aedificatoria. Wikipedia. en.wikipedia.org
4. Alberti, L. B. (1988). On the art of building in ten books (J. Rykwert, N. Leach, & R. Tavernor, Trans.). MIT Press. (Original work completed c. 1452, printed 1485.)
5. Wittkower, R. (1971). Architectural principles in the age of humanism (3rd ed.). Alec Tiranti.

A church that would fit inside a pier of St Peter's

In 1634 the Discalced Trinitarians, a poor order, gave a difficult little corner site at the crossroads of the Four Fountains in Rome to Francesco Borromini, then in his mid thirties and working without a reputation of his own. The church of San Carlo alle Quattro Fontane was built between 1638 and 1641; its facade came much later, from 1662, and was finished after Borromini killed himself in 1667. Contemporaries said the whole church would fit inside one of the piers that carry the dome of St Peter's, and they were not far wrong.

On that site Borromini did something the Renaissance had no vocabulary for. He drew the plan from two equilateral triangles set base to base, making a rhombus, and inscribed circles in it, and from that geometry generated an oval. Then he made the walls undulate: the columns stand on a wall that curves in and out between them, so the boundary of the room is never flat and never still. Above, an oval dome is coffered with crosses, octagons and hexagons that shrink as they climb, which makes the dome read as far deeper than it is. The facade, when it came, alternates concave and convex bays, so the front of the building appears to breathe.

Why this matters: The Renaissance treated the circle and the square as perfect and static. The Baroque treats geometry as something to set in motion, and the room as something that acts on the person standing in it.

Bernini, and space that grips you

Gian Lorenzo Bernini was everything Borromini was not: a sculptor of enormous fame, socially fluent, papally favoured, and Borromini's lifelong antagonist. Between 1656 and 1667 he laid out the piazza in front of St Peter's, and it is the clearest demonstration of Baroque spatial rhetoric anywhere.

The problem was that the basilica's front is very wide and comparatively low, and it stands at the end of a long approach. Bernini enclosed the space with two curved colonnades, four rows deep, 284 columns, sweeping out from the church to embrace an oval piazza some 240 metres across before returning. He described the colonnades himself as the arms of the church, reaching out to receive the faithful. Walk it and the rhetoric works on your body: you are gathered in, funnelled, released into the open oval, then compressed again into the narrower trapezoid before the steps. Bernini also built the Scala Regia, a stair rising from the piazza into the Vatican that narrows and lowers as it climbs, so that a figure descending it appears to grow. Perspective is being used as a tool, like brick.

Versailles: a plan as a political instrument

In 1682 Louis XIV moved the court of France permanently to Versailles, out of Paris and its mobs. The palace had begun as his father's hunting lodge, which survives inside the fabric, and was expanded first by Louis Le Vau and then by Jules Hardouin-Mansart. The Hall of Mirrors, built between 1678 and 1684, is 73 metres long, 10.5 wide and 12.3 high, with seventeen tall windows on the garden side and seventeen matching arches opposite, each filled with mirror glass, 357 panes in all. Mirror was then a Venetian monopoly and ruinously expensive; making it in France was itself a piece of state policy.

Read the plan politically, because that is what it was for. The king's bedchamber sits on the central axis, at the point where the whole composition converges, and the daily rituals of rising and retiring were public events with a hierarchy of who might attend. Andre Le Notre's gardens run that axis out for kilometres, past a Grand Canal about three kilometres end to end, with the landscape itself clipped, levelled and made to obey. Several thousand courtiers lived in the building, competing for rooms whose distance from the axis was a public measurement of their standing.

The whole thing is a machine for concentrating attention on one person, expressed in geometry. And it worked well enough that Caserta, Schonbrunn, Peterhof and a dozen smaller courts copied it.

Worth holding on to: Baroque planning gives architecture a new claim: that a plan can organise behaviour, and that the organisation is the point.

The turn: what happened after 1750

Then the argument reverses, and it reverses because of shovels. Excavation began at Herculaneum in 1738 and at Pompeii in 1748, and for the first time Europeans could see Roman domestic life rather than infer it from ruins and texts. In 1762 James Stuart and Nicholas Revett published the first volume of The Antiquities of Athens, measured drawings of Greek buildings, and Greek architecture stopped being a rumour. Johann Joachim Winckelmann's histories of 1755 and 1764 argued that Greek work was the summit and Roman work a derivation, which reorganised the whole hierarchy.

Alongside the archaeology came a theory. In 1753 the abbe Marc-Antoine Laugier published his Essai sur l'architecture, with a frontispiece showing a primitive hut: four tree trunks, beams, a pitched roof of branches. Everything necessary, said Laugier, is there. Columns, entablature, pediment. Everything else, pilasters, pedestals, broken pediments, niches, is licence. It is a demand that architecture be reduced to what can be justified, and it lands on a generation that has just been shown the actual buildings.

So neoclassicism is not simply more classicism. It is archaeological where the Baroque was inventive, reductive where the Baroque was elaborative, and it prefers the plain wall, the unfluted column, the primary solid.

Boullee's sphere and Ledoux's semicircle

Etienne-Louis Boullee took the reduction to its logical end and stopped building. He taught, and he drew: enormous projects in pure geometry, cubes, cylinders, pyramids, and above all the cenotaph for Isaac Newton of 1784, a hollow sphere roughly 150 metres across, its shell pierced with small holes so that daylight would fall inside as a field of stars, with a great lamp hung at the centre to serve as a sun at night. It could not have been built. It was not really meant to be. Boullee's manuscript, an essay on art, went unpublished until 1953, and his influence travelled through his drawings and his students.

The idea both he and his contemporaries were pursuing is usually called architecture parlante, architecture that speaks: a building whose form announces its purpose without inscriptions, so that a monument to the man who explained the heavens is a model of the heavens.

Claude Nicolas Ledoux built. The Royal Saltworks at Arc-et-Senans, constructed between 1775 and 1779, is an industrial complex laid out as a half circle: eleven buildings on the arc and the diameter, with the director's house at the centre of the straight edge, from which every workshop, store and dwelling is visible. Salt was a royal monopoly and heavily taxed, so a saltworks was a fortified revenue plant. Ledoux made the surveillance the plan. He went on to imagine an entire ideal city, Chaux, radiating from it, and to build the toll gates that ringed Paris for the tax farm from 1785, monumental little temples on the boundary where duty was levied on everything entering the city. They were widely detested, and several were attacked in July 1789.

In Virginia, Thomas Jefferson was making the same move in a republican key. The Academical Village he laid out for the University of Virginia from 1817 puts a library, not a chapel, at the head of a lawn, in a Rotunda modelled on the Pantheon at about half scale, flanked by ten pavilions each in a different order so that the buildings themselves teach the classical language. It is architecture parlante with a curriculum.

Common misconceptions

• Baroque means overdecorated. The word began as an insult, from a term for an irregular pearl. The content of the style is spatial and geometric: ovals, undulating walls, diagonal views, hidden light sources, and plans that move you. A Baroque church can be quite plain and still be Baroque.
• Neoclassicism was a return to the Renaissance. It was a return past the Renaissance to the excavated evidence, with a strong preference for Greek over Roman that the Renaissance never had, and a reductive temper the Renaissance would have found bleak.
• Versailles was built as a palace from nothing. Louis XIII's hunting lodge is still inside it. The palace grew by accretion around a small brick and stone building the king declined to demolish.
• Boullee's projects were fantasy with no consequences. They defined a way of thinking, geometry as meaning, that ran through the nineteenth-century schools into twentieth-century modernism. When you see a building that is a pure cylinder or a stark cube, you are downstream of these drawings.

Looking back

• Borromini's San Carlo, 1638 to 1641, generates an oval from paired equilateral triangles, undulates the wall between its columns, and forces perspective in the coffering of its dome.
• Bernini's piazza at St Peter's, 1656 to 1667, uses 284 columns in four rows around an oval about 240 metres wide to gather, compress and release the approaching visitor.
• Versailles concentrates a court on one axis, with the Hall of Mirrors of 1678 to 1684 running 73 metres and holding 357 mirror panes, and gardens by Le Notre extending the axis for kilometres.
• After 1750 excavation at Herculaneum and Pompeii, Stuart and Revett's Athens of 1762 and Winckelmann's histories replace invention with archaeology, and Laugier's primitive hut of 1753 demands reduction.
• Boullee's Newton cenotaph of 1784, a sphere some 150 metres across, is architecture parlante at its purest, and was never intended to be built.
• Ledoux built the same idea: a semicircular saltworks of 1775 to 1779 in which the director sees everything, and the Paris toll gates that were attacked in 1789.

Sources

1. Wikipedia contributors. (n.d.). San Carlo alle Quattro Fontane. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Palace of Versailles. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Royal Saltworks at Arc-et-Senans. Wikipedia. en.wikipedia.org
4. Laugier, M.-A. (1977). An essay on architecture (W. Herrmann & A. Herrmann, Trans.). Hennessey and Ingalls. (Original work published 1753.)
5. Summerson, J. (1980). The classical language of architecture (rev. ed.). Thames and Hudson.

The problem: how do you bridge a gorge in an ironmaking district?

The Severn gorge at Coalbrookdale in Shropshire is steep, the river is busy with barges that must not be obstructed by scaffolding, and by the 1770s the valley was the most concentrated ironmaking district in the world. Timber was expensive and stone would need centering standing in a working river. What the district had, in quantity, was iron.

So the members of The Iron Bridge were cast at Abraham Darby III's foundry, five sectional ribs spanning 30.63 metres, nearly 1,700 individual components, some 385 tonnes of cast iron, and on 2 July 1779 the arch first spanned the river. It opened on 1 January 1781.

Look at how it is put together, because it is the most revealing thing about it. The foreman Thomas Gregory drew the details, and he detailed them as a carpenter would: mortise and tenon joints, dovetails, wedges. There is no riveting, no bolting in the modern sense, no welding. Cast iron was a new structural material and nobody had a language for joining it, so they used the language they had. That is what the first generation of any technology looks like.

The upshot: A new material does not immediately produce new forms. It first produces old forms made of the new material, and the forms change only when someone works out what the material is actually good at.

Three metals, three jobs

Everything in this lesson follows from a table you should learn.

Read the nineteenth century off that table. Early iron buildings use cast columns and wrought ties, because that is what each metal will do. Cast iron beams in mill floors cracked and dropped, sometimes fatally, which is exactly what a brittle material in bending does. Once Bessemer made steel cheap, the distinction dissolves and the frame becomes possible.

A greenhouse builder wins a competition

In 1850 the Royal Commission for the Great Exhibition had a site in Hyde Park, a deadline of May 1851, no acceptable design out of 245 entries, and an obligation to leave the park's mature elms standing. Joseph Paxton, head gardener at Chatsworth and a builder of very large glasshouses, sketched a solution on a sheet of blotting paper and got it accepted.

The Crystal Palace was 1,851 feet long, a number chosen to match the year, about 564 metres, roughly 124 metres wide, with a barrel-vaulted transept 33 metres high specifically so the elms could stay inside it. It went up between September 1850 and May 1851 and held some 300,000 panes of glass, every one the same size, because that size was the largest sheet the glassworks could then make.

That last sentence is the whole point, and it is why this building matters more than its style. The design is governed by the manufacturing module. Everything is set out on a grid derived from the pane; the cast iron columns are identical; the gutters double as structural members and as the rails along which glaziers ran their trolleys; every joint is bolted, so the building can be taken down and it was, moving to Sydenham in 1854, where it burned in 1936. Nobody had previously designed a great public building around what a factory could repeat and what a wagon could carry.

Say plainly what the Crystal Palace was not. It was not the first iron and glass building: conservatories, arcades, market halls and railway sheds had been going up in both for decades, and Paxton's own Great Conservatory at Chatsworth was a direct rehearsal. What was new was the scale, the speed, and the fact that a procurement process, not a proportional system, generated the form.

Eiffel, and a curve that is a calculation

For the Paris exposition of 1889, Gustave Eiffel's company built a tower 300 metres high, from 18,038 pieces of puddled wrought iron held by two and a half million rivets, weighing around 7,300 tonnes, in two years, two months and five days. Eiffel chose wrought iron rather than steel, which existed and was cheaper, because he trusted its consistency and had built his bridges from it.

The famous curve of the legs is not a taste decision. Eiffel described the principle himself: the shape is chosen so that the wind's overturning effect at any height is balanced by the geometry of the structure below, giving a profile that flares at the base and rises almost vertically at the top. The Eiffel Tower is a diagram of wind loading, executed at 300 metres. In February 1887 a group of writers and artists published a protest against it as a monstrous and useless lamppost. It was to be demolished after twenty years and was saved because it turned out to be an excellent radio antenna.

Chicago: the wall gives up

The fire of October 1871 burned out the centre of Chicago. What followed was a rebuilding boom on the most valuable land in the country, and two technologies made height worth having: the passenger lift with a safety brake, and the telephone, which let a firm's staff be stacked instead of spread. The obstacle was the wall.

Look at what a load-bearing wall costs when you go high. The Monadnock Building of 1891, by Burnham and Root, is sixteen storeys of pure brick, with no frame at all, and its walls are about 1.8 metres thick at street level. That thickness is unrentable floor area on the most expensive ground in Chicago, and it is buying you sixteen storeys. The masonry wall had reached the point where every extra floor cost more ground floor than it added at the top.

The Home Insurance Building of 1885, by William Le Baron Jenney, is routinely called the first skyscraper. Treat that carefully: it combined masonry piers with an internal metal frame, and the claim depends on where you draw the definition. The honest statement is that it is an important step in a sequence rather than a first.

The sequence arrives at the Reliance Building. Root designed the ground floor and basement in 1890; Charles Atwood completed the upper floors in 1894 and 1895; the steel framing for the top ten floors went up in fifteen days, between 16 July and 1 August 1895. The steel structure weighed about a third of what an equivalent stone one would have. The exterior is white glazed terracotta and plate glass, and the glass covers most of the surface, in the wide central pane with narrow opening sashes either side that Chicago made its own.

Stand in front of a photograph of the Reliance Building and cover the cornice. What is left could have been built in 1955. The wall has become a curtain: it hangs off the frame, carries only itself and the wind, and is therefore free to be almost entirely window. Everything in Lesson 14 depends on this having happened.

Why this matters: Between 1885 and 1895 the wall stopped holding the building up. Every argument about what a facade should look like from then on is an argument about a surface that has no structural obligations at all.

Common misconceptions

• The Home Insurance Building was the first skyscraper. It is a contested claim resting on a definition. It used masonry piers alongside its metal frame, and framed construction had precedents in mills and warehouses. Say important step, not first.
• The Crystal Palace invented iron and glass architecture. Conservatories, arcades and railway sheds got there first. Its innovation was designing at that scale around a manufactured module and a demountable bolted assembly.
• Cast iron is simply a weaker version of steel. It is excellent in compression and dangerous in tension, and the difference killed people in mill floors. Nineteenth century structural practice is largely the discipline of putting each metal only where it belongs.
• The frame arrived because architects wanted glass facades. It arrived because land in central Chicago was expensive, lifts made height usable, and thick masonry walls consumed the floor area they were built to sell. The glass facade is a consequence, not a motive.

The short version

• The Iron Bridge of 1779 spans 30.63 metres in some 385 tonnes of cast iron, jointed with mortise, tenon and dovetail because nobody yet knew how to join iron.
• Cast iron suits compression, wrought iron suits tension and bending, and cheap steel after 1856 does both, which is what makes a frame possible.
• The Crystal Palace of 1851 is 564 metres long, holds around 300,000 identical panes, and is designed around a manufacturing module and a bolted, demountable assembly.
• The Eiffel Tower of 1889 uses 18,038 wrought iron pieces and 2.5 million rivets, and its curve is a response to wind rather than a matter of taste.
• The Monadnock Building of 1891 shows the limit of the load-bearing wall: sixteen storeys bought with 1.8 metres of brick at street level.
• The Reliance Building's top ten floors were framed in steel in fifteen days in July and August 1895, and its terracotta and plate glass skin carries nothing but itself.

Sources

1. Wikipedia contributors. (n.d.). The Iron Bridge. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Reliance Building. Wikipedia. en.wikipedia.org
3. Chicago Architecture Center. (n.d.). Reliance Building. Buildings of Chicago. architecture.org
4. Giedion, S. (1967). Space, time and architecture: The growth of a new tradition (5th ed.). Harvard University Press.
5. Condit, C. W. (1964). The Chicago school of architecture: A history of commercial and public building in the Chicago area, 1875-1925. University of Chicago Press.

Before you read

Find something decorated within reach: a patterned mug, a phone case, a carved chair leg, a printed shirt. Spend one minute on three questions. Who made the decoration, a person or a machine? Roughly how many minutes of somebody's life went into it? And if the pattern went out of fashion, would you throw the object away sooner than you would throw away a plain one? Those three questions are, almost exactly, the terms of the argument that ran from 1853 to 1930 and that this lesson lays out.

The case for ornament: Ruskin, and a house at Bexleyheath

In 1853 John Ruskin published the second volume of The Stones of Venice, and its central chapter, The Nature of Gothic, made an argument nobody had made in quite that way. Look, he said, at the carving on a Gothic capital. It is uneven. Some of it is clumsy. That roughness is the evidence that a person was thinking while they cut it, allowed to make their own decisions and their own mistakes. Then look at a perfectly executed classical moulding, machine-precise, every unit identical. That perfection required a workman to be reduced to a tool. Ruskin's conclusion was moral: you cannot ask for perfect finish and free workers at the same time, and if you must choose, choose the workers.

William Morris read that chapter as a student and spent his life acting on it. In 1859 he commissioned Philip Webb to build him Red House at Bexleyheath: red brick left visible inside and out, an L-shaped plan following the rooms rather than a symmetrical facade, steep roofs, and every fitting, tile, window and hanging designed and largely made by Morris and his friends. In 1861 the firm that became Morris and Company followed. The Arts and Crafts movement grew out of it, and its principles were consistent: honest materials, visible construction, the designer and the maker as the same person or in the same room, and decoration understood as the trace of that person's work.

Morris was also the first to state the problem with his own position, and he stated it more bluntly than any critic. Handwork is slow, slow is expensive, and expensive means only the rich can buy it. He was a socialist who wrote that he had spent his life ministering to the swinish luxury of the rich, and he never resolved it.

What matters here: The case for ornament is not that decorated things look nicer. It is a claim about the conditions under which people work, and it stands or falls on that.

Art Nouveau: the same case, new materials, fifteen years

In 1893 Victor Horta completed the Hotel Tassel in Brussels for a scientist friend. The structure is iron, and Horta does not hide it: an exposed iron column stands in the middle of the stair hall, sprouting tendrils, and the whipping curve of that ironwork is repeated in the mosaic floor, the painted wall, the balustrade, the door handles. Every element is drawn by the architect and made for that house.

Art Nouveau accepts what the nineteenth century built with, iron and glass, and refuses what the nineteenth century did with them, which was to disguise them as stone. That is a real advance. But hold on to the sentence about every element being drawn by the architect, because it is the movement's death certificate. A style in which nothing repeats cannot be built cheaply, and by about 1910 Art Nouveau had exhausted itself commercially almost everywhere it appeared. It lasted roughly fifteen years, which for a movement of that visibility is very short, and the reason is arithmetic.

The case against: what Loos actually wrote

Adolf Loos delivered the lecture that became Ornament and Crime in 1910; it was printed in French in 1913, and it is often dated 1908. Almost everyone knows the title and almost nobody knows the argument, so here it is.

1. Ornament wastes labour and material. A decorated object is tied to a fashion. When the fashion passes, the object is discarded while it is still perfectly serviceable. A plain object stays usable. So ornament shortens the life of things, and the labour poured into it is destroyed.
2. The ornament worker is exploited. Because the ornamented object is not worth proportionally more, the carver and the embroiderer are paid less per hour than the workers who make plain things. Loos turns Ruskin's argument on its head: for Ruskin ornament liberates the craftsman, for Loos it underpays him.
3. Modern people have outgrown it. This is the part that has aged worst. Loos frames cultural development as an evolutionary ladder and compares tattooed Papuans with modern Europeans in a way that is straightforwardly racist. It is not an incidental blemish; it is load-bearing in his argument, and you should say so when you cite him.

Notice what is not in the list. Loos never says ornament is ugly, and he did not build stripped, poor interiors. The Looshaus on the Michaelerplatz in Vienna, of 1910, has a plain upper facade that scandalised the city and a ground floor clad in figured green marble, and his houses are full of mahogany, mirror, brass and stone. His objection is to applied ornament, pattern added to a surface, not to material richness. If you take the slogan and skip the essay, you get this exactly backwards.

Wright's third position

In 1901 Frank Lloyd Wright gave a lecture at Hull House in Chicago called The Art and Craft of the Machine, and it is best read as a direct reply to Morris. Wright's claim is that the machine is not the enemy of art, that fighting it is futile and beneath an artist, and that the machine has its own nature which a designer should learn: it excels at clean, repeated, precise work in wood, glass and metal. The task is not to keep the machine out of ornament but to design ornament the machine can make well.

So Wright keeps ornament and changes where it comes from. It should not be applied to a surface; it should be generated from the geometry of the building itself and from the nature of the material. The art glass in the Robie House of 1909 to 1910 is straight-line geometry in clear and coloured glass, cut and leaded by machine-assisted methods, and its pattern is the same family of lines that organises the plan. The house itself is a Prairie manifesto: a long horizontal mass, the roof cantilevered far past the walls, a central hearth as the fixed point, and rooms that flow into each other rather than sitting as boxes off a hall.

Wright therefore agrees with Morris that ornament matters and with Loos that hand production is a dead end, which is why his position outlasted both.

What would actually settle it

Three of the claims in this dispute are empirical, and it is worth noticing that they have rarely been tested.

What is not empirical is the underlying disagreement about what a building is for. Ruskin and Morris think a building is, among other things, a record of the labour that made it. Loos thinks it is an instrument that should not squander life. Wright thinks it is an organism whose parts should all express one idea. No amount of wage data adjudicates between those.

In short: This is not an argument about taste. It is an argument about labour, obsolescence and what buildings are for, conducted in the language of taste, and the winner in 1930 will win partly because it is cheaper.

Common misconceptions

• Loos said ornament is a crime. The title is a translation and a provocation. The argument is economic and cultural: ornament wastes labour, ties objects to fashion, and underpays the people who make it. He also builds in marble and mahogany, which the slogan cannot accommodate.
• Arts and Crafts was simply anti-machine nostalgia. Morris was a revolutionary socialist whose target was the condition of the worker, and in later lectures he accepted machines for work no one should have to do by hand. The nostalgia is real but it is not the argument.
• Art Nouveau failed because tastes changed. It failed because its production model, every element individually designed and made, cannot scale. Taste followed cost.
• Wright was a modernist who rejected ornament. He designed ornament all his life and defended it, and he attacked the modernists for supposing that a stripped surface was a solution rather than an evasion.

Where this leaves us

• Ruskin's Nature of Gothic in 1853 grounds the case for ornament in the freedom of the maker, and Morris builds Red House on it in 1859 while conceding that handwork prices out everyone but the rich.
• Art Nouveau, exemplified by Horta's Hotel Tassel of 1893, accepts iron and glass and insists on unique design for everything, and burns out in about fifteen years for reasons of arithmetic.
• Loos's Ornament and Crime, lectured in 1910 and printed in 1913, argues that ornament wastes labour, accelerates obsolescence and underpays craftsmen, and rests part of its case on an explicitly racist evolutionary claim.
• Loos objects to applied pattern, not to material luxury: the Looshaus is plain above and figured green marble below.
• Wright's answer, from his 1901 lecture onward, is ornament generated from the building's own geometry and made by machine, as in the Robie House art glass.
• Several claims in the dispute are testable and largely untested, but the deepest disagreement is about what a building is for, which no data settles.

Sources

1. Wikipedia contributors. (n.d.). Ornament and Crime. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Red House, Bexleyheath. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Hotel Tassel. Wikipedia. en.wikipedia.org
4. Ruskin, J. (1853). The nature of Gothic. In The stones of Venice (Vol. 2). Smith, Elder and Co.
5. Frampton, K. (2020). Modern architecture: A critical history (5th ed.). Thames and Hudson. Chapters on Arts and Crafts, Art Nouveau and Adolf Loos.

A ground floor curved to fit a car

In the spring of 1928 Pierre and Eugenie Savoye asked for a weekend house on a meadow at Poissy, north west of Paris. The brief was short: a summer house, space for cars, an extra bedroom, a lodge for a caretaker. What went up between 1928 and 1931 was a white box lifted clear of the grass on thin columns, and underneath it a ground floor whose glass wall bends in a long, shallow arc. That arc is not a flourish. It is the turning circle of a motor car: you drive in under the house, follow the curve round, set your passengers down at the door and carry on into the garage cut into the same curve. The plan of the Villa Savoye starts from a vehicle radius, which should tell you immediately that this house is not organised around a hearth, a staircase or a front door in the old sense.

One drawing from 1914 does all the work

Fourteen years before Poissy, Le Corbusier made a diagram he called Dom-Ino, intended as a system for rebuilding houses in the war zone of Flanders. It shows two flat reinforced concrete slabs and a roof slab, carried on six slender columns, with a stair at one side. Look at what is missing. There are no downstand beams under the slabs. There is nothing whatever in the perimeter. And the columns are set in from the edge of the slab, so that each floor plate runs past its supports and stops in mid air.

Everything in the checklist falls out of those two decisions. The columns carry all the load, so no internal wall holds anything up and partitions can go wherever the rooms want them: the free plan. The slab cantilevers past the outermost column, so the outside skin carries nothing either and can be glass, render, anything: the free facade. Since the skin carries nothing, a window can run continuously along it, because the strip of wall between two windows has no job to do. With no walls at ground level, the building can stand on columns with open ground beneath. And a flat slab roof, unlike a pitched timber one, can be occupied.

The point: The five points are not five ideas. They are one idea, a frame with its columns pulled in from the edge, counted five different ways.

The five points, taken one at a time at Poissy

Le Corbusier published the list in 1927. Here it is against the building, with the structural condition each one depends on.

Two details are worth having. The column grid started at 5 metres and was cut to 4.75 metres, along with moving the master bedroom down a level, purely to shrink the volume after the Savoyes objected to the cost. The cost rose anyway: an estimate of about half a million francs in February 1929 became something like 900,000 by the time the changes had worked through.

Notice the ribbon window that runs across the open terrace. Behind that stretch of glazing there is no room, only sky. It is continuous because the argument requires continuity, not because anyone inside gains. That is fair to hold against the building, and it is also the clearest demonstration that the facade has stopped reporting the interior.

What actually happened to the house

It leaked. Water came through the roof terrace and around the skylight from the first winter, the Savoyes complained repeatedly, and the family barely lived in the house before the war made the point moot. It passed to the French state in 1958, was listed as a historic monument in 1965 while Le Corbusier was still alive, and was restored between 1985 and 1997.

Do not make that a joke about flat roofs. The flat roof is the thing that makes the fifth point available, and in 1930 the waterproofing needed to make it reliable lagged the structural idea by decades. That gap, between a structural possibility and the secondary technology needed to inhabit it, is about to appear again.

Dessau, 1926: the same argument, built as a school

While Le Corbusier was arguing in Paris, Walter Gropius was building. The Bauhaus building at Dessau, put up in 1925 and 1926 for the school Gropius directed, is a reinforced concrete frame arranged as five functional parts: a trade school, a workshop block, a five-storey studio block called the Prellerhaus with 28 rooms in which students lived and worked, an auditorium and canteen between them, and a two-storey bridge carrying the administration and Gropius's own office over a road. There is no front. Walk round it and no elevation announces itself as the main one.

The famous part is the workshop wing. Its columns are set back from the perimeter, exactly as at Dom-Ino, so a glass curtain runs uninterrupted across all three floors and the full length of the block. It looks like a factory, which was the point: a school of design should look like the industry it was training people for. Gropius had tried the move with Adolf Meyer at the Fagus Factory at Alfeld in 1911, whose corners are glass with no visible support because the steel sits inboard.

And it did not work as an environment. Single glazing over three storeys facing the sun cooked the workshops in summer, and the only remedy available was curtains, which destroyed the transparency the whole wall existed to produce. In winter the same glass shed heat so fast that the building needed heavy heating, and ventilation depended on elaborate mechanically operated louvre windows. A glass curtain wall is comfortable only with double glazing, solar control coatings and mechanical cooling, which is to say from roughly the 1950s onward. At Dessau the idea arrives about thirty years before the equipment that makes it liveable, and the honest way to read the building is as a bet that the equipment would come.

So what?: Twice now, at Poissy and at Dessau, the structural move is sound and the building is uncomfortable, because the secondary technologies are late. Judge a building against what its own century could actually supply.

Barcelona, 1929: walls that hold nothing up

Ludwig Mies van der Rohe, working with Lilly Reich, had less than a year to design the German national pavilion for the 1929 International Exposition at Barcelona. The German republic wanted a self-portrait: democratic, progressive, unmilitary. Mies gave them a building with nothing in it. No exhibits, no offices, one figure by Georg Kolbe, and furniture designed for the room, including the chair still sold under the city's name.

Look at how the Barcelona Pavilion is made. The whole thing sits on a plinth of travertine. A regular grid of cruciform steel columns, clad in chrome so that they reflect and half disappear, carries a thin flat roof. Between the columns, and quite unrelated to them, stand independent planes of stone and glass: veneers of Tinos verde antico marble, a wall of golden onyx, sheets of grey, green and translucent glass. They slide past one another so the space narrows and widens as you move, and they steer you: up a few steps, turn, turn again, and out at ground level on the far side, the slope having delivered you downward while you were looking at the water.

That is the free plan taken to its conclusion: the columns hold the roof, the walls only divide. The pavilion is not quite as pure as the diagram, since some of those planes do help carry the roof, and Robin Evans noticed that the shiny columns read less as supports than as thin objects struggling to hold a floating plane down. The building was demolished in early 1930, less than a year old, and reconstructed on its original footings between 1983 and 1986.

Park Avenue, 1958: what happens when the box has to be a business

Now change the input. Instead of an empty pavilion for six months, give the same architect a Manhattan block, a corporate client, thirty-eight floors of lettable office and an American building code. The Seagram Building at 375 Park Avenue, finished in 1958, is 157 metres tall. Mies got the job because Phyllis Lambert, daughter of the company's chief executive, wrote to object to the first design her father had commissioned and was then put in charge of choosing an architect; Philip Johnson worked with Mies on the canopies, lobby, lighting and restaurant.

Two decisions carry the building. The first is the plaza. Mies set the tower back about 90 feet from Park Avenue behind an open square of pink granite with two fountains, giving up a great deal of rentable ground floor in exchange for a place from which the tower can be seen whole. It worked so well that New York's 1961 Zoning Resolution offered developers extra floor area in return for building plazas, which produced a great many plazas nobody wanted to stand in. One good building became a citywide rule, and the rule was worse than the building.

The second decision is why this building closes the lesson. Mies wanted the steel frame expressed on the outside, and New York's fire code required structural steel to be encased, so the actual frame is invisible. His answer was to apply bronze I-sections to the face of the curtain wall, running the full height, in front of the encased steel they stand for. The mullions are bronze, the spandrel panels Muntz metal, and the visible order of the facade is a scale model of a structure you are not allowed to see.

You can call that honest or you can call it ornament. What you cannot call it is structure. The most severe glass box on Park Avenue carries an applied order derived from a frame, which is very nearly what a classical pilaster is, and the man doing it was the one who said less is more. Hold on to that: the postmodern attack, two lessons from now, starts exactly here.

Bottom line: By 1958 the expressed structure of a modern building is a representation of structure, not the thing itself. Everything after this argues about what that representation should say.

Common misconceptions

• The five points are a style, a look of white walls, flat roofs and strip windows. Four of the five are direct consequences of one structural arrangement, columns set in from the edge of a slab. Copy the look onto a load-bearing masonry building and none of the five are actually present.
• Modernism meant cheap, stripped, unornamented building. The Barcelona Pavilion is travertine, golden onyx and chrome. The Seagram Building is bronze, Muntz metal, travertine and granite, and was among the most expensive office towers of its day. Plainness of form and cheapness of material are unrelated.
• The vertical members on the Seagram facade are the building's structure showing through. They are applied bronze sections in front of fireproofed steel that the code required to be hidden.

What to carry forward

• Dom-Ino, 1914: slabs on columns set in from the edge, no perimeter structure. Every one of the five points is a restatement of that.
• Villa Savoye, 1928 to 1931: all five points on a 4.75 metre grid, a ground floor curve set by a car's turning circle, and a final cost near 900,000 francs against an estimate of half that.
• Bauhaus Dessau, 1925 to 1926: five functional wings with no principal front, and a three-storey glass curtain wall that overheated in summer and bled heat in winter because the glazing lagged the structure by decades.
• Barcelona Pavilion, 1929: cruciform chrome columns carry a thin roof while independent planes of onyx, marble and glass only divide space; demolished in 1930, rebuilt 1983 to 1986.
• Seagram Building, 1958: 157 metres, 38 storeys, set back 90 feet behind a granite plaza that the 1961 Zoning Resolution then turned into a formula.
• Because code required the steel to be encased, Seagram's bronze mullions represent a structure they are not. Expressed structure has become a language rather than a fact.

Sources

1. Wikipedia contributors. (n.d.). Villa Savoye. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Bauhaus Dessau. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Seagram Building. Wikipedia. en.wikipedia.org
4. Le Corbusier. (1931). Towards a new architecture (F. Etchells, Trans.). John Rodker. Internet Archive
5. Frampton, K. (2020). Modern architecture: A critical history (5th ed.). Thames and Hudson. Chapters on Le Corbusier, the Bauhaus and Mies van der Rohe.

A famous time of death, and why it is wrong

Charles Jencks opened his 1977 book on postmodern architecture with a sentence that has been quoted ever since: modern architecture died in St Louis, Missouri, on 15 July 1972 at 3.32 in the afternoon, or thereabouts, when the Pruitt-Igoe housing project was dynamited. It is a wonderful line. Nearly every part of it is wrong, and working out how it is wrong is the most useful thing you can do with this subject.

Start with the date. The first building at Pruitt-Igoe was demolished with explosives on 16 March 1972. More followed on 21 April, on 9 June, and then on 15 July. Jencks's precise moment is the fourth demolition of a complex of 33 buildings, and it is nowhere near the end: the federal housing department announced the decision to clear the rest in August 1973, the last tenant moved out in May 1974, and the site was not fully cleared until 1976, at a demolition cost of about 3.5 million dollars. A story that needs a single instant to work has already been fitted to the argument rather than to the evidence. So let us do this properly, and to do it properly we first need the buildings the story is really about.

Marseille, 1947 to 1952: one building, two legacies

After the war Le Corbusier finally got to build the housing type he had been drawing since the 1920s. The Unite d'Habitation at Marseille, finished in 1952, is a single slab on massive pilotis containing 337 apartments in 23 different layouts across 12 storeys. The flats are two levels each and they interlock: a corridor runs along the centre of every third floor, with the flats on one side reaching down below it and those on the other reaching up above it, so two thirds of the floors have no corridor at all and their flats run right through the building. Le Corbusier called those corridors streets in the sky. On the roof he put a running track, a shallow paddling pool, sculptural ventilation stacks and a nursery school.

Two things came out of that building, and it matters that they are separable. The first is the type: a self-contained vertical neighbourhood standing free in open ground, which is the built form of the Ville Radieuse that Le Corbusier had proposed for cities in 1930. The second is the surface. The hoped-for steel frame proved too expensive in the post-war shortage, so the building went up in beton brut, raw concrete left exactly as it came out of the timber shuttering, with the grain of the boards and the joints between them printed on the finished wall. Walter Gropius, at the opening on the roof in 1953, said that any architect who did not find the building beautiful had better lay down his pencil.

Remember: Beton brut is French for raw concrete, and that phrase, not the English word brutal, is where brutalism gets its name. The style is not called after cruelty. It is called after a finish.

What brutalism actually claimed

The critic Reyner Banham gave the movement its English label in a 1955 article on the New Brutalism, and the architects at its centre in Britain were Alison and Peter Smithson. Their programme had three parts: show the materials as they are, show how the building is put together, and let the plan be legible from outside. That is a moral position about honesty, inherited from Ruskin by way of the argument you met two lessons ago and applied to concrete instead of carved stone.

There was also an economic argument that gets forgotten. In-situ concrete needs no cladding trade, no facing brick and no applied finish, so for a public authority building at scale on a fixed budget, leaving the structure as the finish removed an entire package from the contract. The Barbican Estate in London, Boston City Hall, the National Theatre on the South Bank and the Smithsons' own Robin Hood Gardens of 1972 all belong to that convergence of an ethic and a budget. Robin Hood Gardens was demolished from 2017 onward, over the objections of a long list of architects, which tells you the argument about these buildings is not settled.

The claim we are going to debug

Here is the standard story, stated as strongly as it deserves. Pruitt-Igoe was modernist doctrine built at full size: towers in a park, communal galleries, a clean break with the street. Within a decade it was violent, half empty and physically wrecked. It was blown up on television. Therefore the doctrine was wrong, and the profession that held it should have known.

The facts in that paragraph are mostly true. The inference is where it breaks. Take it in four steps.

Step 1: what was built was not what was designed

The complex was 33 identical eleven-storey blocks on a 57-acre site on the north side of St Louis, 2,870 apartments in total, housing more than 10,000 people at full occupancy, opened in 1954 and 1955. The architect was Minoru Yamasaki, early in a career that would later produce the World Trade Center.

His first scheme was not 33 identical towers. It mixed walk-ups, mid-rise blocks and high-rises. That proposal exceeded the cost ceiling imposed by the federal Public Housing Administration, and the agency intervened and imposed a uniform height of eleven floors on everything. Material shortages from the Korean War tightened the screws further. The landscaping that was supposed to make this towers in a park was cut from the final plan and few trees were ever planted. The ground floor shops went too, and there was no public mailbox anywhere in the complex. Density was set at 50 units to the acre, higher than in the slums being cleared. Yamasaki himself said at the time that low buildings at low density were unquestionably more satisfactory.

So the first move in the standard story, treating the built project as a pure specimen of architectural doctrine, is already false. It is a specimen of value engineering under a federal cost cap.

Step 2: the early numbers do not fit

If a design is inherently uninhabitable, it should be visibly failing from the first week. Occupancy at Pruitt peaked at 95 per cent and at Igoe at 86 per cent. In the sociological study conducted between 1963 and 1966, by which time the decline was well advanced, 78 per cent of residents said they were satisfied with their apartment and 80 per cent said the project met their needs better than where they had lived before. One early tenant described her eleventh floor flat as a poor man's penthouse.

Now put the vacancy against the city. St Louis was projected to grow from 850,000 people in the 1940s to a million by 1970. Instead it lost about 30 per cent of its residents in that period to suburbanisation and white flight, along with 11,000 manufacturing jobs. A city shedding a third of its population does not fill 2,870 new subsidised flats on its north side, whatever their plan looks like. By 1969 occupancy at Pruitt was 57.1 per cent and at Igoe 48.9 per cent.

Step 3: the rule that guaranteed the spiral

This is the part most accounts leave out, and it is the one that does the most work. The Housing Act of 1949 gave local housing authorities federal money to build, and then required them to meet operating and maintenance costs out of tenants' rent. Capital was funded; upkeep was not.

Follow the loop. Vacancy rises, so rent income falls. The authority cannot cut a mortgage, so it cuts maintenance: in 1958 it reduced maintenance by 10 per cent. Lifts break and stay broken, corridors go unlit, and the place becomes less desirable, so vacancy rises again. Meanwhile the minimum rent had to climb: 20 dollars a month in 1952, 32 in 1958, 43 in 1962, 58 in 1968, until some families were paying as much as three quarters of their income in rent and were being charged separately for replacing a fuse or a door lock. In 1969 the tenants struck for nine months, and the strike ended with the entire board of commissioners resigning. Eugene Meehan, who studied St Louis housing policy at length, called the whole arrangement a set of policies programmed for failure.

The core of it: A building that cannot pay for its own caretaker will decay whatever shape it is. The financing rule, not the massing, is the mechanism that turns a half-empty estate into a ruin.

Step 4: what the design criticism gets right

Debugging is not the same as acquittal, and there is a real architectural finding here. It concerns shared space.

The lifts were skip-stop: they served only the first, fourth, seventh and tenth floors, an economy measure meant to reduce congestion. That forced a large share of residents onto the stairs every day. The stairwells were enclosed, unwatched and, once maintenance failed, unlit, and muggings in them were frequent. The anchor floors carried long communal corridors called galleries, intended as social space, which were delivered unpainted, unfurnished and dimly lit. Nobody could see into them, no household could claim any part of them, and they became the property of whoever was prepared to stand there. A 1959 audit and a 1967 report both found that a large proportion of the crime was committed by people who did not live in the complex.

Oscar Newman built his theory of defensible space partly on this case, and his argument has a control group, which is what makes it worth taking seriously. Carr Village stood immediately next door: low-rise, with a very similar tenant population, and through the same years it stayed fully occupied and largely trouble free. Same city, same funding regime, same demographics, different building form, different outcome. Within Pruitt-Igoe itself the same pattern appeared in miniature, since flats grouped around small landings shared by two families, where tenants cleaned and watched their own common area, did comparatively well.

That is a finding about surveillance and ownership of shared space. It is not a finding about flat roofs, concrete, pilotis or the twentieth century.

Why the wrong explanation was so attractive

In 1991 Katharine Bristol published an article called The Pruitt-Igoe Myth which asked a sharper question: not whether the architectural explanation is true, but who it serves. If bad design destroyed Pruitt-Igoe, then architects are the decisive actors in public housing, and the discipline is central rather than marginal. The story flatters the profession even while appearing to condemn it. Bristol puts the weight instead on chronic underfunding, on racial discrimination, and on the collapse of the city's economy.

One small fact is worth adding, because it turns up in almost every retelling and is not true: Pruitt-Igoe never won an architectural award. Architectural Forum praised Yamasaki's original mixed-height proposal in 1951, before the federal agency flattened it into 33 identical slabs.

Common misconceptions

• Modern architecture died at 3.32 pm on 15 July 1972. The first demolition was on 16 March 1972; 15 July was the fourth. The last tenant left in May 1974 and clearance ran to 1976. Jencks wrote a good sentence, not a finding.
• Brutalism is named for its brutality. It comes from beton brut, raw concrete, the finish Le Corbusier used at Marseille because steel was unaffordable after the war.
• Pruitt-Igoe was an award-winning showpiece of modernism. It won nothing. A magazine praised the scheme that was never built.
• The buildings were hated from the day they opened. Occupancy peaked at 95 per cent in Pruitt, and most surveyed residents preferred the flats to their previous housing well into the decline.

The takeaway

• The Unite d'Habitation, 1947 to 1952, gave the post-war world two separable things: the free-standing slab of 337 flats with interior streets, and beton brut as a finish.
• Brutalism is an ethic of showing materials and construction, reinforced by an economic fact: leaving the concrete as the finish deletes the cladding trade from the contract.
• Pruitt-Igoe as built was not the scheme designed. A federal cost cap imposed 33 identical eleven-storey blocks, and deleted the landscaping and the ground floor shops.
• Occupancy peaked at 95 and 86 per cent, and most residents preferred the flats to what they had left. The collapse tracks the city losing 30 per cent of its population and 11,000 manufacturing jobs.
• The Housing Act of 1949 made local authorities pay maintenance out of rent, which converts any vacancy into decay. Minimum rent tripled between 1952 and 1968 and the tenants struck for nine months in 1969.
• The surviving design criticism is narrow and evidenced: skip-stop lifts, unlit stairwells and unclaimable galleries removed ordinary supervision from shared space, with low-rise Carr Village next door as the control case.

Sources

1. Wikipedia contributors. (n.d.). Pruitt-Igoe. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Unite d'habitation. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Brutalist architecture. Wikipedia. en.wikipedia.org
4. Bristol, K. G. (1991). The Pruitt-Igoe myth. Journal of Architectural Education, 44(3), 163-171.
5. Newman, O. (1972). Defensible space: Crime prevention through urban design. Macmillan.

A book by an architect who had built almost nothing

In 1966 the Museum of Modern Art published a short book by a 41-year-old Philadelphia architect whose complete works amounted to a house for his mother and an apartment block for the elderly. The book was Complexity and Contradiction in Architecture, most of it written in 1962 on a Graham Foundation grant, and its first chapter is called Nonstraightforward Architecture: A Gentle Manifesto. Vincent Scully rated it the most important architectural text since Le Corbusier's book of 1923. Philip Johnson later said Robert Venturi had untied the chains with one stroke.

What follows is a dispute, not a verdict. Two positions, each with people who built real buildings and could argue for them, and a set of questions at the end about which parts of the argument could actually be decided.

The first position, in its own words

Venturi's target is named in the book: Mies van der Rohe's less is more, which you watched turn into bronze mullions on Park Avenue in the lesson before last. Venturi answers with two lines that became slogans. More is not less. And, more famously, less is a bore.

Underneath the slogans is a specific claim. Modern architecture had achieved clarity by excluding things: excluding history, excluding decoration, excluding contradiction, excluding anything the diagram could not carry. Venturi argues that the buildings people actually love are full of contradictions held together, and he takes his evidence from everywhere: Michelangelo, Frank Furness, Edwin Lutyens, Alvar Aalto, ordinary commercial streets. He wants what he calls the difficult whole, an architecture of both-and rather than either-or, in which an element can do two jobs at once and a facade can be at odds with the plan behind it without that being a failure.

You can see it in the house he built for his mother between 1962 and 1964. The Vanna Venturi House at Chestnut Hill has five rooms and stands about nine metres tall, but the front reads as monumental, because Venturi manipulates every cue you use to judge a building's size: an enormous gable split by a gap down the middle, a flat arch applied to the wall that carries nothing at all, and windows punched as holes rather than framed as compositions. Scully thought it the biggest small building of the second half of the century. It is a house that argues.

Worth holding on to: The founding text of postmodernism is not about ornament. It is about complexity, contradiction and the right of a building to be several things at once.

The duck and the decorated shed

The second book matters as much. In 1972 Venturi, Denise Scott Brown and Steven Izenour published Learning from Las Vegas, based on a studio in which Yale students surveyed the commercial strip with the seriousness normally reserved for Rome. Out of it came the most useful pair of terms in the whole argument.

The provocation is the last row. Venturi and Scott Brown are saying that a heroic modern building whose form is meant to express its function is doing exactly what the roadside duck does, only without admitting it, and that the ordinary decorated shed is cheaper, more adaptable and more candid. This is an argument about communication, and it is worth noticing that they did not praise Las Vegas so much as analyse it. Denise Scott Brown's role in the work was routinely written out of the credits at the time, which is its own comment on the period.

The second position arrives as a building: Portland, 1980

Doctrine becomes a public commission in Oregon. Portland ran a design competition for a municipal office building, with Philip Johnson on the three-member jury. Johnson removed Gunnar Birkerts's entry for not being postmodern enough, and the winner was Michael Graves. The mayor, Frank Ivancie, was among those who thought that glass and steel had made American downtowns boring.

The Portland Building opened in 1982: fifteen storeys, 29 million dollars, a cube in teal and cream with small square windows, a huge painted keystone motif, and stucco garlands hung on the side. In 1985 a hammered copper figure called Portlandia was set above the entrance. It is generally counted the first major postmodern office tower, opening before Philip Johnson and John Burgee's AT&T Building at 550 Madison Avenue in New York, which was completed in 1984, rises 197 metres in 37 storeys of pink granite, and finishes with a broken pediment carrying a circular opening that everyone immediately compared to furniture.

Both buildings do the same thing structurally: an ordinary frame, with an argument applied to the outside. They are decorated sheds, exactly as advertised.

The objections, at full strength

Now the other side, and it is not one objection but four.

The technical one. A facade made of applied historical elements is a cladding problem, and cladding problems leak. Structural and water infiltration faults at the Portland Building appeared almost at once. By 1990, eight years after opening, the lobby and food court needed rebuilding. In 2014 some city commissioners argued for demolition. The reconstruction that ran from 2017 to 2020 replaced the tile, the stucco garlands, the windows and effectively the entire envelope, at a construction cap of 140 million dollars plus up to 55 million in other costs. Set that against the 29 million it cost to build. The people who worked in it, meanwhile, described it as cheaply built and hard to work in, and the windows are small partly because the composition wanted them small.

The critical one. Kenneth Frampton argued in 1983 for what he called critical regionalism: postmodernism has not escaped the universal commercial culture it complains about, it has merely swapped one international language for another, and both are scenography. What resists, on his account, is architecture built out of a particular place: its light, its climate, its topography, its available construction. Judge a building by whether you could put it anywhere.

The one inherited from Loos. The ornament on these buildings is not carved by anyone. It is thin stone veneer, stucco on lath, painted concrete, cut by machine and hung on a frame. If your objection to modernism was that it starved the surface, an applied simulation of ornament does not answer it, and by the argument of Lesson 13 it is precisely the sort of fashion-tied surface that gets stripped off in twenty years. Which, at Portland, it was.

The one about jokes. A broken pediment on a skyscraper is a wink at the observer. Winks work once. Nothing dates faster than a joke about seriousness, and by the early 1990s the style was commercially finished, which is a shorter run than Art Nouveau managed.

The upshot: The strongest case against postmodernism is not that it was ugly. It is that it made a claim about communication and then bet it on a cladding detail with a twenty year life.

What would actually settle any of this

Notice how few of these have ever been tested at scale, which is the same finding as in Lesson 13. Arguments about buildings are conducted in the language of taste because the measurements are expensive and nobody commissions them.

Common misconceptions

• Postmodernism means sticking classical details on a modern building. It began as an argument about complexity, contradiction and communication. The pediments are a late and partial consequence, and in 2001 Venturi said flatly that he had never been a postmodernist.
• Learning from Las Vegas was a celebration of Las Vegas. It was a survey. The thesis is about the decorated shed as a building type, and about the honesty of admitting that most architecture is one.
• Less is a bore was an attack on all restraint. It answers one specific slogan of Mies van der Rohe's, and Venturi's own buildings are mostly plain boxes with one loud gesture.
• The Portland Building failed because postmodernism does not work. What failed was an envelope, on a tight budget, in a wet climate. The fair version of the charge is narrower: the style put its argument in the cladding, which is the shortest-lived part of any building.

Putting it together

• Complexity and Contradiction, 1966, argues for the difficult whole: both-and rather than either-or, contradiction held rather than resolved. Its evidence is Michelangelo, Lutyens, Furness and the commercial street.
• The Vanna Venturi House of 1962 to 1964 makes a nine metre, five-room house monumental by manipulating scale cues, with an applied arch that carries nothing.
• Learning from Las Vegas, 1972, splits buildings into ducks, whose form is the message, and decorated sheds, which apply the message to an ordinary structure, and argues that most architecture is the second while pretending to be the first.
• The Portland Building, 1982, 15 storeys and 29 million dollars, is the first major postmodern office tower; the AT&T Building follows in 1984 at 197 metres with a broken pediment.
• The technical objection is documented: Portland's envelope failed and its 2017 to 2020 reconstruction was capped at 140 million dollars, several times what the building cost.
• Frampton's critical regionalism reframes the whole quarrel as two international styles arguing, and asks instead whether a building could stand anywhere.
• Three of the four live claims are testable and largely untested; the fourth is about what a building is for, and no survey settles that.

Sources

1. Wikipedia contributors. (n.d.). Complexity and Contradiction in Architecture. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Portland Building. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). 550 Madison Avenue. Wikipedia. en.wikipedia.org
4. Venturi, R. (1977). Complexity and contradiction in architecture (2nd ed.). Museum of Modern Art.
5. Venturi, R., Scott Brown, D., and Izenour, S. (1977). Learning from Las Vegas: The forgotten symbolism of architectural form (Rev. ed.). MIT Press.

Six hundred and eighty-one entries, and two architects who expected to lose

In 1971 a jury sat down in Paris to judge an open competition for a new arts centre on the Plateau Beaubourg. It was the first time foreign architects had been allowed into a French state competition, and 681 entries arrived. The jury included Oscar Niemeyer, the engineer Jean Prouve and Philip Johnson. They chose a scheme by Renzo Piano and Richard Rogers, with Su Rogers and Gianfranco Franchini, both in their thirties, who by their own account entered believing they had no chance and therefore nothing to lose.

The Centre Pompidou opened on 31 January 1977 with everything on the outside. The steel structure, the escalators, the ducts and the pipework are all hung on the two long elevations, and colour-coded: green for water, blue for air handling, yellow for electrical services, red for circulation and safety equipment. The cross frames rest on cast steel rocker brackets, the gerberettes, which push the columns clear of the floor plates.

Every one of those decisions buys the same thing. With structure and services outside, the interior is a stack of enormous unobstructed floors that can be re-partitioned for any exhibition anybody thinks of in the next fifty years. Paris hated it and then made it one of the most visited buildings in Europe. It closed in 2025 for a renovation running to 2030, largely to remove asbestos, upgrade the technical systems and improve its energy performance: park that fact until the end of the lesson.

High tech is a maintenance argument, not a look

The obvious reading of these buildings is that the architects liked the look of machinery. The better reading is duller and more convincing: different parts of a building wear out at wildly different rates.

Rogers built the argument twice. At the Lloyd's building in London, designed from 1978 and finished in 1986, the lifts, stairs, lavatory pods and service risers sit outside in six towers, so the short-lived layer can be replaced without touching the long-lived one, and the underwriting room is left as a clear volume under a barrel vault.

Norman Foster took the same logic to Hong Kong. The HSBC Main Building, built between 1981 and 1985, hangs its floors from steel masts rather than standing them on columns, leaving the ground open as a public undercroft you can walk straight through, and delivers its servicing in modules built off site. At about 5.2 billion Hong Kong dollars, roughly 668 million United States dollars, it was the most expensive building in the world.

Key idea: Putting the services outside is not decoration and it is not honesty for its own sake. It is a bet that a building will be worth keeping longer than its equipment lasts.

Foster's later 30 St Mary Axe, completed in December 2003, is 180 metres and 41 floors on the site of the Baltic Exchange that the 1992 bombing wrecked. Its plan narrows at the base to give back street space and reduce the downdraughts flat-sided towers throw at pedestrians, narrows again at the top, and is braced by a diagonal steel grid on the perimeter rather than an internal frame.

Bilbao, 1997: what the software actually changed

In 1991 the Basque government offered to fund a Guggenheim museum in Bilbao's dying port district: 100 million dollars for construction, a 50 million acquisitions fund, a 20 million fee and an annual subsidy. Frank Gehry got the job with instructions to be daring, and the Guggenheim Museum Bilbao opened on 18 October 1997: titanium, limestone and glass on a 32,500 square metre riverside site, 24,000 square metres of building, 11,000 of it exhibition space in nineteen galleries, the largest 130 metres by 30.

Here is the part usually missed. It was delivered on time and on budget, at about 89 million dollars of construction, which for a shape like that is close to unheard of. Gehry has explained how, and none of the reasons are aesthetic. He kept the design team in control through construction so the scheme was not eroded by other interests; he had a realistic cost estimate before starting; and he modelled the surfaces in CATIA, the software Dassault wrote for designing aircraft, then worked directly with the trades from that model.

That is the real change. Software of that kind does not draw a curve; it defines a surface precisely enough that a fabricator can cut a unique panel from it and know it will fit. The line running from Brunelleschi's purpose-built hoists through Paxton's 300,000 identical panes reaches, here, a machine that can make every piece different at nearly the price of making them the same. The museum was quickly credited with reviving the city, and the Bilbao effect became a policy dozens of places copied, mostly without the same result, because a landmark is the visible part of an urban strategy and not a substitute for one.

Deconstructivism is the label attached to this work after a 1988 exhibition in New York grouped Gehry with Zaha Hadid, Rem Koolhaas, Peter Eisenman and Daniel Libeskind. Gehry has never accepted it, and the grouping is a curatorial convenience rather than a movement anybody joined.

Parametricism, and the bill for a one-off panel

Patrik Schumacher, of Zaha Hadid Architects, named parametricism in 2008 and claimed it as the successor to modernism. The method is real: instead of drawing a shape, the designer writes the relationships between elements, sets the parameters and lets the geometry follow, so changing one input reshapes the whole model consistently. Hadid's Heydar Aliyev Center in Baku, opened in 2012, is the showpiece: a continuous folded surface with no visible column in the public halls.

Now apply the test from Lesson 13. Art Nouveau died because every element was individually designed and could not be repeated. Parametric design revives that condition and pays for it with computer-controlled fabrication rather than a craftsman's hours. The bill still arrives, in bespoke moulds, in extra structure behind a surface that carries nothing, and in the fact that these buildings tend to be commissioned by clients with unusual budgets.

The argument that changes the subject

Everything so far in this course has been about how to build. The last twenty years have added a prior question: whether to.

Operational carbon, the emissions from heating, cooling and lighting a building once it is running, has fallen for decades as insulation, glazing and plant improved. Embodied carbon, the emissions from making and transporting the materials and putting them together, has not, and it is spent before anybody moves in.

The numbers are blunt. Making cement produces on the order of 900 kilograms of carbon dioxide per tonne, about half from the chemistry itself, since making lime means driving carbon dioxide out of limestone, and about 40 per cent from burning fuel to reach temperature. Cement alone accounts for up to 8 per cent of human-made carbon dioxide, and concrete overall for between 4 and 8 per cent of global emissions. Structural concrete works out at roughly 410 kilograms of carbon dioxide per cubic metre, and replacing 30 per cent of the cement with fly ash brings that to about 290.

Sit with that last pair. A 30 per cent substitution in one ingredient cuts the emissions of a cubic metre of structure by nearly a third, a larger saving than most of what a building's occupants will ever be asked to do. It is also a decision made by an engineer in a specification, years before the first tenant arrives, and invisible in every photograph of the finished building.

What matters here: The carbon accounting moves the decisive moment of a building from its operation to its procurement. What matters most is now the mix design, the structural quantity, and whether the thing gets built at all.

Two answers: grow the structure, or keep the building

The first answer is to build in wood. Cross-laminated timber, made by gluing layers of boards at right angles into large panels, and glued laminated timber for beams and columns, reach heights nobody attempted in wood before. Mjostarnet, at Brumunddal in Norway and finished in March 2019, is eighteen storeys and 85.4 metres, about 11,300 square metres of hotel, flats and offices, with glulam carrying the loads and cross-laminated panels forming the shafts and balconies. It was the tallest timber building in the world until Ascent MKE in Milwaukee, 25 storeys and about 87 metres, was completed in August 2022.

Be precise about the caveat, because it is structural. The top seven floors of Mjostarnet have concrete slabs. Timber is light, and lightness is a problem at height: a slender light tower accelerates in the wind enough for people inside to feel it, so mass was added where movement is greatest. The carbon claim also depends on what happens to the forest and to the building afterwards, since wood stores carbon only while it remains wood and while the trees are replaced.

The second answer is to stop demolishing. Anne Lacaton and Jean-Philippe Vassal won the Pritzker Prize in 2021 for the opposite of a landmark: taking 1960s social housing slabs scheduled for demolition and bolting prefabricated winter gardens and balconies onto them, so every flat gained several square metres of usable space and much better light while the tenants stayed in place. Their position, that the existing building should never be demolished, is the sharpest reply to the whole tradition this course has followed. If a sound frame exists, the carbon in it has been paid, and a replacement must earn back its own embodied emissions through decades of better operation before it is even level.

The upshot: The greenest structural decision available to most architects is to keep a building that already stands and make it good. That is a strange note on which to end a history of building, and it is where the argument actually is.

Common misconceptions

• A green building is one with a good energy rating. Ratings measure operation. Embodied carbon is paid up front, is invisible afterwards, and on a well-insulated new building can outweigh decades of running emissions.
• Timber towers are carbon negative. They beat concrete equivalents under reasonable assumptions about forestry and end of life, and they still contain concrete and steel, as Mjostarnet's top seven floors show.
• Computers let architects build any shape they want. The software controls fabrication, which is why Bilbao came in on budget, but every unique component is still made once and paid for once.
• High-tech buildings expose their services for effect. Services wear out three or four times faster than structure; put them where they can be reached and the building outlives its plant.

What to remember

• The Pompidou Centre, chosen from 681 entries in 1971 and opened in 1977, puts structure and colour-coded services outside so the floors inside stay clear and re-usable. Lloyd's and the HSBC building apply the same reasoning.
• Bilbao, 1997, cost about 89 million dollars and came in on time because of detailed estimating, direct work with the trades and aerospace software that defines a surface precisely enough to fabricate from.
• Parametric design revives the Art Nouveau condition, everything unique, and pays for it with machine fabrication rather than hours of handwork; whether that cost falls far enough is unsettled.
• Cement emits about 900 kilograms of carbon dioxide per tonne and accounts for up to 8 per cent of human-made emissions; structural concrete runs about 410 kilograms per cubic metre, or about 290 with 30 per cent fly ash.
• Mass timber reached 85.4 metres at Mjostarnet in 2019 and about 87 at Ascent MKE in 2022, with concrete added high up because a light tower moves too much in wind.
• Lacaton and Vassal's Pritzker in 2021 rewarded not demolishing: the carbon in a standing frame is already spent, and a replacement must earn back its own before it is even level.

Sources

1. Wikipedia contributors. (n.d.). Centre Pompidou. Wikipedia. en.wikipedia.org
2. Wikipedia contributors. (n.d.). Guggenheim Museum Bilbao. Wikipedia. en.wikipedia.org
3. Wikipedia contributors. (n.d.). Environmental impact of concrete. Wikipedia. en.wikipedia.org
4. Wikipedia contributors. (n.d.). Mjostarnet. Wikipedia. en.wikipedia.org
5. Frampton, K. (2020). Modern architecture: A critical history (5th ed.). Thames and Hudson. Chapters on high tech, critical practice and the contemporary condition.