The first bridges were found, not designed: a tree fallen across a stream, a flat stone laid between boulders, a ledge of rock that happened to reach the other bank. The oldest surviving type is the clapper bridge, slabs of stone resting on piled-stone piers, still carrying walkers on Dartmoor and in the Pyrenees. It is a beam bridge, and its limit was the length of stone one could quarry and drag: a few metres. Korea has the same idea at a larger scale in the Nongdari at Jincheon, twenty-eight piers of dry-laid river stone carrying slab beams for nearly a hundred metres, repaired every year for the better part of a millennium.

The oldest bridge that still stands as its builders left it is at Arkadiko in Greece: a Mycenaean culvert of about 1300 BC, its opening spanned not by a true arch but by corbelling, each course of stone overhanging the one below until the two sides meet. It was built for chariots and it is still in use by farmers. Corbelling needs no centring and no wedge-shaped stones, but it works only for spans of a metre or two.

Rope came next, and independently in several places. Communities in the Andes and the Himalaya wove suspension bridges from grass and vine; the Q'eswachaka bridge in Peru is still rebuilt by hand every year by the villages that use it, a living record of how a suspension bridge was made before iron. Chinese engineers in Sichuan were hanging bamboo-cable bridges of a hundred metres and more by the first centuries AD; the iron-chain bridges of the Tibetan borderlands are older than any in Europe by several centuries.

Armies solved the same problem with boats. Herodotus describes a bridge on stone piers at Babylon around 600 BC whose timber deck was taken up every night, and the pontoon bridge that carried Xerxes' army across the Hellespont in 480 BC: two lines of ships, more than three hundred in each, held by flax and papyrus cables and covered with brushwood and earth. Two thousand years later King Jeongjo of Joseon crossed the Han river to Hwaseong in 1795 on a bridge of boats designed by Jeong Yak-yong, and its manual, the Jugyo jinam, is among the earliest surviving engineering specifications for a bridge anywhere.

What none of these had was a way to span farther than a single piece of material could reach, or to stand in a river without being swept away. That took a new idea about shape, and a new material for foundations.

The type that carried the ancient world was not the arch but the timber beam bridge on stone or timber piers, and the Romans built the two most famous. In 55 BC Julius Caesar's legions bridged the Rhine in ten days on driven timber piles, angled against the current and braced so that the flow pushed them tighter together; he described the structure in enough detail for it to be reconstructed. In AD 105 Apollodorus of Damascus built Trajan's bridge over the Danube for the Dacian war: twenty stone piers in the river and timber arches of about 38 m between them, 1,135 m end to end. For a thousand years nothing longer was built anywhere.

These bridges established the division of labour that lasted until iron: the piers were stone, because stone survives water; the spans were timber, because timber is light, strong in bending and can be cut long. The whole difficulty was in the piers. Roman engineers drove timber sheet piles in a ring, pumped or bailed out the enclosure, and built the pier inside it, the first cofferdams, and where they could they used a concrete of lime and volcanic ash from Pozzuoli that set under water and, unlike anything else available, went on hardening for centuries. The same pozzolanic concrete built the harbour at Caesarea and the dome of the Pantheon; without it the Roman river pier is not possible.

The timber span had one weakness that outlasted Rome: it rots and it burns. Trajan's superstructure was dismantled within a generation. The piers stood in the Danube until the nineteenth century, and two of them can still be seen at low water.

The arch lets small pieces of weak material do what one long piece cannot. Wedge-shaped voussoirs lean on each other, so every stone is in compression and the load runs round the curve into the abutments. Stone is very good at compression, so an arch built well in the first century is still carrying traffic in the twenty-first: the Pons Fabricius in Rome (62 BC) has never been out of use.

Rome did not invent the arch. Mesopotamia and Etruria had used it in gates and drains, but Rome industrialised it: standard semicircular arches, built on timber centring that was struck and re-used at the next span, on piers a third as thick as the span so that each arch could stand on its own before its neighbour was built. Spans reached about 30 m; the Alcántara bridge in Spain (AD 106, by Gaius Julius Lacer, whose tomb stands beside it) has two spans near 29 m and rises 48 m over the Tagus. The Puente de Mérida runs 792 m on sixty arches. The Pont du Gard carried water, not people, three tiers of arches 49 m high, but it is the same engineering.

Why did it last? Three reasons that Chapter 1 has already given. The form puts stone in the one action it is good at. The massive spandrel fill and the thick piers make the dead load so large that no live load can move the line of thrust out of the ring. And the semicircle springs vertically from its pier, so the thrust of each arch is mostly carried straight down and the piers can stand alone, which is also why Roman piers are so wide and why Roman bridges block so much of the river. The price of that width was paid at every flood, and it was the medieval builders who first began to pay it down.

When the empire went, the knowledge went with it. Western Europe built almost no large stone bridges for six hundred years, and where it did, it often re-used Roman piers.

While Europe forgot, China advanced. The Anji Bridge at Zhaozhou, built by Li Chun in AD 605, is a segmental arch: not a semicircle but a shallow segment of a circle, 37 m across and rising only 7 m, with two small arches through each spandrel so that floodwater passes through the bridge rather than against it. It is lighter, flatter and cleverer than anything Rome built, and Europe did not match it for seven hundred years. It still carries traffic.

The Song dynasty then did something the rest of the world would not attempt again until the railway: it built long bridges out of enormous stone beams. The Luoyang Bridge at Quanzhou (1059, by the prefect Cai Xiang) crosses an estuary on forty-odd piers and 834 m of granite beams, some over 10 m long and weighing tens of tonnes, floated into place at high tide; the builders seeded the pier foundations with oysters so that the shellfish would cement the rubble together, which they did. The Anping Bridge (1151), further down the same coast, ran to 2,070 m and was the longest bridge in the world for most of a millennium. The Lugou (Marco Polo) Bridge outside Beijing (1192) put eleven segmental arches over the Yongding river and 500 carved lions on its parapets. A Song painting of the capital shows the rainbow bridge, a woven timber arch of interlocking beams with no fastenings, a type rebuilt in modern times and found to work.

Korea took the stone beam and the stone arch and made them precise. The twin stairway bridges at Bulguksa near Gyeongju (751), Cheongungyo and Baegungyo, spring from a true stone arch and are the country's oldest; the Seonjukgyo at Kaesong is a Goryeo stone beam bridge remembered for the assassination of Jeong Mong-ju on it in 1392; the Gwangtonggyo in Seoul (1410) carried the main road across the Cheonggyecheon on granite beams and stood, buried under the road for fifty years, until the stream was uncovered in 2005. Japan's answer was timber: the five arches of the Kintai Bridge at Iwakuni (1673), rebuilt to the same drawings every generation.

Persia built the most beautiful bridges of the age and made them do two jobs. The Si-o-se-pol at Isfahan (1602) is thirty-three arches of brick over the Zayandeh river with a promenade above; the Khaju Bridge (1650) downstream is a bridge, a weir and a pavilion at once, with sluices under its arches to hold the river up for irrigation. The Ottomans, heirs of both Rome and Persia, built the Stari Most at Mostar (1566, by Mimar Hayruddin), a single 28.7 m arch so slender its builder is said to have prepared for his own funeral on the day the centring was struck, and Mimar Sinan's eleven-arch bridge at Višegrad (1577).

Europe's return to bridge building came through the Church. In 1177 a shepherd named Bénézet began the Pont d'Avignon across the Rhône, twenty-two arches and 900 m; the lay brotherhood that finished it, the Frères Pontifes, went on to build bridges as an act of charity, and a chapel on the bridge collected the tolls. Old London Bridge (1176–1209, by the priest Peter of Colechurch) was the extreme case: nineteen pointed arches on piers so thick, and so widened by their protective starlings, that the Thames was narrowed to a fraction of its width and dropped like a weir between the piers. It carried a street of houses and shops, a chapel and a gate on which heads were displayed; it stood for six hundred years and its rebuilding in 1831 was as much about the river as the bridge.

The medieval bridge was a piece of the town. The Ponte Vecchio in Florence (1345, attributed to Taddeo Gaddi) is three segmental arches with shops on top, the flattest arches Europe had built since Rome; the Charles Bridge in Prague (begun 1357 under Peter Parler) is a fortified street with towers at each end; the Pont Valentré at Cahors (1308–1378) has three towers and a portcullis. The Rialto in Venice (1591, Antonio da Ponte) put a single 28 m stone arch on 12,000 timber piles driven into mud, an early lesson that a bridge is only as good as its foundation, and it has not moved.

Quietly, the proportions changed. Roman piers were a third of the span; medieval builders, who could not afford Roman piers and had learned what they did to floods, brought the pier down toward a fifth, and the pointed arch, which pushes less outward than a semicircle, let them do it safely. By 1750 Jean-Rodolphe Perronet in France would take the pier to a ninth of the span, but only because he had understood something the medieval masons never wrote down: that in a row of equal arches the thrusts cancel across each pier, and the pier need only carry weight, as long as every arch is built before the centring of any is struck. His Pont de Neuilly (1774) and Pont de la Concorde (1791) are the last great stone bridges and the thinnest.

Until the seventeenth century a bridge was proportioned by rule and proved by standing up. The change began with drawings and ended with equations.

1502

Leonardo da Vinci sketches a single stone arch of 240 m to cross the Golden Horn for the Sultan. It is never built, and a Norwegian footbridge to his geometry (2001) shows it would have worked.

1570

Andrea Palladio publishes truss bridges in his Four Books of Architecture: timber spans made of triangles, the first published trusses in Europe.

1638

Galileo's Two New Sciences asks how strong a cantilever beam is, gets the answer wrong by a factor of three, and gets the scaling right: strength grows with the square, weight with the cube (section 1.8).

1675

Robert Hooke, in a Latin anagram, states that the hanging chain inverted is the standing arch (section 1.6), and that stretch is proportional to load, the law that bears his name.

1744

Leonhard Euler derives the load at which a slender column buckles, using the beam theory the Bernoullis had built. It is a hundred years before anyone needs it for a bridge.

1773

Charles-Augustin de Coulomb's essay on statics gives the correct bending stress in a beam and the first real analysis of how an arch fails, by hinging, not sliding.

1747

France founds the École des Ponts et Chaussées, the first engineering school anywhere, on the corps that had run French roads since 1716. Perronet is its first director. The profession now has a name and a syllabus.

1826

Claude-Louis Navier publishes the beam theory engineers still use, and, three years earlier, the first theory of suspension bridges; his own Pont des Invalides in Paris fails at its anchorages during construction and is taken down. Theory and practice were not yet the same thing.

1847

Squire Whipple in New York publishes the first correct analysis of truss forces, and for the first time a bridge member is sized before it is built rather than proved by surviving.

1864–1886

Maxwell's reciprocal theorem, Culmann's graphic statics (1866), Mohr's circle, Castigliano's theorems (1873) and, in 1886, Heinrich Müller-Breslau's principle of influence lines, the method by which every moving-load calculation on this site is still done.

By 1890 the beam, the truss, the arch and the suspension cable could all be calculated, elastic analysis of continuous and indeterminate structures existed, and the profession could ask a new question: not whether the bridge would stand, but by how much. The safety factor, and eventually the limit state of Chapter 4, followed from that.

Timber was the long-span material of the eighteenth century, and Swiss carpenters were its masters. Hans Ulrich Grubenmann's bridge at Schaffhausen (1757) crossed the Rhine in two spans of 52 and 59 m under a roof, and his Wettingen bridge (1764) reached 61 m in a single arch-truss; both were burned by armies in 1799. In America Lewis Wernwag's Colossus at Philadelphia (1812) spanned 104 m over the Schuylkill, the longest timber span ever built, until it too burned in 1838. Fire and rot were timber's enemies, and the roof of the covered bridge was not decoration: it kept the rain off the trusses and roughly tripled the life of the bridge.

What turned carpentry into engineering was the patent truss. Theodore Burr's arch-truss (1804), Ithiel Town's lattice (1820, a web of planks pinned at every crossing, buildable by any carpenter), Stephen Long's panel truss (1830), William Howe's truss (1840, with wrought-iron verticals that could be tightened) and Thomas Pratt's (1844, diagonals in tension so that the long members could be iron rods) were all designed for the railway, which wanted stiffness and quick construction, and they were the first bridge types sold as designs rather than built as one-offs. The Warren truss of 1848, equilateral triangles with no verticals, was the last and simplest, and it is still the default form of a modern steel truss.

The railway's demand for hundreds of identical spans is what pulled iron into the truss, member by member: first the tension rods, then the whole thing. Timber lasted longest where it was cheapest, in the trestles of the American West, some of them a hundred metres high and a kilometre long, built in weeks and rebuilt every decade.

The Iron Bridge at Coalbrookdale (1779, Abraham Darby III and the architect Thomas Pritchard) is a stone arch translated into cast iron: 30.6 m, cast in pieces at the foundry that had invented coke smelting, and joined with carpentry-style dovetails and wedges, because nobody yet knew how else to connect iron. It worked, and it started the century in which the material changed faster than the theory. The Wearmouth bridge at Sunderland (1796) went to 72 m; Thomas Telford proposed a single cast-iron arch of 180 m across the Thames in 1801, and, when that was refused, built the Pontcysyllte aqueduct (1805) in a cast-iron trough on stone piers 38 m high.

Cast iron is strong in compression and brittle in tension, so the arch suited it and the beam did not. Used as girders it failed: Robert Stephenson's Dee Bridge at Chester (1847) broke under a train and killed five, and the inquiry pushed engineers toward wrought iron, which could be rolled, riveted and, above all, pulled. Telford had already shown what tension iron could do. His Menai Suspension Bridge (1826) hung a 176 m span from wrought-iron eyebar chains, twice anything before it, after James Finley in Pennsylvania (1801) and Samuel Brown's Union Bridge on the Tweed (1820) had shown the form in iron at smaller scale. In France the Seguin brothers built the first wire-cable suspension bridge at Tournon (1825), and Joseph Chaley's Grand Pont at Fribourg (1834) reached 273 m on wire. Suspension bridges also taught the first lessons in dynamics: Broughton (1831) collapsed under soldiers marching in step, and armies have broken step on bridges ever since.

Stephenson answered the Dee with the Britannia Bridge (1850): trains inside rectangular wrought-iron tubes, spanning 140 m over the Menai, proportioned by the tests of William Fairbairn and Eaton Hodgkinson on models, the first bridge designed by experiment and the ancestor of every box girder. Isambard Kingdom Brunel's Royal Albert Bridge at Saltash (1859) combined a wrought-iron tube in compression with chains in tension in each 139 m span, and his Clifton bridge (finished 1864 after his death) used chains from the Hungerford bridge he had built in 1845. In America John Roebling put a railway on a suspension bridge at Niagara (1855), 251 m with a deep stiffening truss, after Stephenson had written that it could not be done; and the Wheeling bridge on the Ohio (1849, 308 m, the longest in the world) was torn apart by wind in 1854, filmed by no one but described by a witness in words that could have been used at Tacoma ninety years later.

Steel arrived in bridges with James Eads' arch over the Mississippi at St Louis (1874), three spans of about 150 m in tubular chrome-steel ribs, built by cantilevering out from the piers so the river stayed open, and founded on pneumatic caissons sunk more than 30 m below the water, where the workers learned, at the cost of fifteen lives, what decompression does. Nine years later the Brooklyn Bridge (1883) spanned 486 m on galvanised steel wire cables spun in place; John Roebling died of an injury on the site before it started, his son Washington was crippled by the caissons, and Washington's wife Emily carried the work to the end. Between them the Roeblings had settled how suspension bridges would be built for a century. Gustave Eiffel's wrought-iron arches at Porto (Maria Pia, 1877, 160 m) and Garabit (1884, 165 m) were the last great works in the older metal.

The railway wanted stiffness more than span, and got it from the cantilever. The Forth Bridge (1890, John Fowler and Benjamin Baker) crosses two spans of 521 m in 54,000 tonnes of steel, with tubes big enough to walk through and members sized for a wind of 2.7 kPa, because the Tay Bridge, 60 km up the coast, had blown into the river in 1879 with a train and 75 people on it, its designer having allowed almost nothing for wind. Baker explained the cantilever to the public with two men on chairs holding a third between them on a plank; it is still the clearest picture of the form. The Quebec Bridge collapsed twice during construction, in 1907 when its compression chords buckled under a dead load its engineers had underestimated, and in 1916 when the suspended span fell while being lifted, before its 549 m cantilever opened in 1917. It is still the longest cantilever in the world, and the first collapse is the reason Canadian engineers wear an iron ring.

The steel arch reached 298 m at Hell Gate in New York (1916, Gustav Lindenthal) and 500 m at Bayonne (1931) and Sydney Harbour (1932), whose 503 m arch was erected as two cantilevers held back by cables until they met. But the twentieth-century record belonged to the suspension bridge, and to a piece of theory: Josef Melan's deflection theory of 1888, applied by Leon Moisseiff at the Manhattan Bridge (1909), which showed that the cable's own stiffness carried much of the load and let the stiffening truss shrink. Othmar Ammann's George Washington Bridge (1931) doubled the record to 1,067 m and was so much stiffer than the theory required that it opened with no stiffening truss at all. The Golden Gate (1937, 1,280 m) followed. Then the theory was taken one step too far: Tacoma Narrows (1940), 853 m long and only 2.4 m deep, twisted itself apart in a 68 km/h wind four months after opening, on film, and aerodynamics entered bridge design for good.

The workshop changed as much as the drawing office. Riveting gave way to welding in the 1930s and to high-strength bolts in the 1950s; welded steel brought a new failure, brittle fracture, to King's Bridge in Melbourne in 1962; and fatigue, which the railway had met first, got its own clauses in every code.

Concrete came back as a modern material once iron could be put inside it. Joseph-Louis Lambot built a rowing boat of mortar on wire mesh in 1848; Joseph Monier, a gardener, patented reinforced flowerpots in 1867 and built the first reinforced concrete bridge at Chazelet in 1875, a footbridge of 16 m. François Hennebique turned it into a construction system in 1892, with stirrups, bent-up bars and a licensing network that put up thousands of structures; his bridge at Châtellerault (1899) has three arches, the centre one 50 m. In America Josef Melan's system encased a steel arch in concrete, which sounds like a compromise and built hundreds of bridges.

Robert Maillart made it an art, and a structure. His Tavanasa bridge (1905) cut away the concrete where the thrust line said it was not needed; his Salginatobel Bridge (1930), a three-hinged hollow-box arch of 90 m in a Swiss gorge, put the hinges where the thrust must pass (section 1.6) so that temperature and settlement could not stress the arch; his deck-stiffened arches at Schwandbach (1933) made the arch a thin curved slab and gave the deck the job of resisting unbalanced load. Every one of them was the cheapest tender.

Eugène Freyssinet learned the material's secret the hard way. His bridge at Le Veurdre on the Allier (1911), three flat concrete arches of about 70 m, began to sag in the months after it opened; he went out at night with the jacks he had used to lift the arches off their centring and jacked them back up, and understood that concrete keeps deforming under load. He had discovered creep, which no code yet mentioned and which would shape everything he did afterwards. At Plougastel (1930) he built three arches of 186 m over the Elorn estuary, then the longest concrete spans in the world, on a single timber centring floated from one span to the next. The Sandö Bridge in Sweden (1943) reached 264 m; Gladesville in Sydney (1964) 305 m, built from precast hollow voussoirs; Krk in Croatia (1980) 390 m; Wanxian over the Yangtze (1997) 420 m, on a steel tube skeleton filled with concrete; and in 2024 the Tian'e Longtan Bridge in Guangxi brought the concrete arch to 600 m.

Freyssinet's bigger idea was patented in 1928. Concrete cracks in tension; so pre-compress it with high-strength steel wire before the load arrives, and in service it never sees tension at all. The trick that had defeated earlier attempts (Jackson in the United States had tried it in 1886) was creep and shrinkage: the concrete shortens over the years and the prestress leaks away. Freyssinet's answer was steel so strong, and stretched so far, that the loss was a fraction of the total, and a material that strong did not yet exist when he began. By 1946 it did, and his bridge at Luzancy on the Marne, 55 m of precast prestressed segments assembled on site, was followed by five sisters in four years. In Belgium Gustave Magnel built the first continuous prestressed bridge at Sclayn (1948) and, having lectured on it in America, the first prestressed bridge there, Walnut Lane in Philadelphia (1950).

Post-war Europe needed thousands of bridges and had no money for the falsework that concrete required, and prestressing solved that too. Ulrich Finsterwalder built out from the pier in segments, each one prestressed back to the last, first over the Lahn at Balduinstein (1951) and then across the Rhine at Worms (1953, 114 m) and Bendorf (1964, 208 m): balanced cantilever construction, the method that Eads had used in steel in 1874, and the river stayed open. Fritz Leonhardt and Willi Baur pushed a whole deck out from one bank over temporary supports at the Rio Caroni in Venezuela (1964), which became incremental launching. Jean Muller cast segments in a yard against each other, so that they matched exactly, glued them with epoxy and stressed them together at Choisy-le-Roi (1962) and across the Oléron strait (1966, 2.9 km); precast segmental construction crossed to America with him, span by span at Long Key and Seven Mile in the Florida Keys (1982) and in balanced cantilever around Grandfather Mountain at Linn Cove (1983). Chapter 8 covers each of these methods.

The quieter revolution was the standard precast girder: the AASHTO I-girders of the 1950s, the bulb-tees and NU girders of the 1980s, Australia's Super-T, cast in a yard, trucked to site and lifted in overnight. Most of the bridges built in the world since 1960 are of this kind, and most of the bridges on this site's Budget estimator are too. At the other end of the scale the concrete box grew: 260 m at Brisbane's Gateway Bridge (1986), 301 m at Stolma in Norway (1998, with lightweight concrete at midspan), 330 m at Shibanpo (2006). Christian Menn's Ganter Bridge in the Alps (1980), with its stays buried in concrete walls, and Jacques Mathivat's Odawara Blueway Bridge in Japan (1994) began the extradosed type, a prestressed box that borrows a short tower from the cable family.

After Tacoma, the suspension bridge was rethought in the wind tunnel. Deep trusses came back for a generation (Mackinac, 1957, 1,158 m, whose designer David Steinman had warned about Tacoma; the Verrazzano-Narrows, 1964, 1,298 m, Ammann's last). Then the Severn Bridge (1966) replaced the truss with a streamlined steel box only 3 m deep, with inclined hangers, and the aerodynamic box deck became the European standard: the Bosporus (1973), the Humber (1981, 1,410 m, the record for seventeen years) and the Great Belt East Bridge in Denmark (1998, 1,624 m). Japan took the other road: the Akashi Kaikyō Bridge (1998) reached 1,991 m on a deep truss built for typhoons and earthquakes, and the Kobe earthquake of 1995 moved its towers apart by a metre while it was under construction. In 2022 the 1915 Çanakkale Bridge crossed the Dardanelles with a 2,023 m span on a twin box deck, the first past 2 km.

The cable-stayed bridge is younger, though the idea is old: a Venetian carpenter drew one in 1617, several were built in the 1810s and two of them fell down, and Roebling added diagonal stays to the Brooklyn Bridge to stiffen it. Franz Dischinger, who had shown in 1938 that stays only work if they are pulled tight, built the Strömsund Bridge in Sweden in 1956, 183 m with two pairs of stays; the Theodor Heuss Bridge at Düsseldorf (1957) and the Severin Bridge at Cologne (1959) followed in a Germany rebuilding its Rhine crossings with the least steel possible. Hellmut Homberg's Friedrich-Ebert Bridge at Bonn (1967) hung the deck from many closely spaced stays instead of a few, which made the deck slender and the construction simple, and every cable-stayed bridge since has been built that way. Riccardo Morandi's Maracaibo bridge (1962) and the Brotonne Bridge in Normandy (1977, 320 m) proved the deck could be concrete; Alex Fraser in Vancouver (1986, 465 m) made it composite; then the span doubled and doubled again: Yangpu in Shanghai (1993, 602 m), Normandie (1995, 856 m), Tatara (1999, 890 m), Sutong (2008, 1,088 m), Russky Island (2012, 1,104 m) and, in 2025, the Changtai Yangtze River Bridge at 1,208 m, carrying road above and rail below. It is the default form between 300 and 1,000 m because it can be built by cantilevering, needs no anchorages and can be made stiff.

The same decades produced the very long bridge as opposed to the very long span. The Confederation Bridge (1997) crosses 13 km of the ice-choked Northumberland Strait on 250 m precast cantilevers; the Øresund (2000) links two countries with a rail-and-road cable-stayed span and a tunnel; the Bang Na expressway in Bangkok (2000) runs 54 km on precast segments; Millau (2004) carries a motorway 270 m above the Tarn on seven cable-stayed spans from pylons 343 m high; Rion-Antirion (2004) stands on gravel-reinforced seabed in an earthquake zone and lets its deck float on fuses; the Hong Kong–Zhuhai–Macau crossing (2018) is 55 km of bridges, islands and tunnel. Chapter 3 describes each family, and Chapter 11 the records.

Since 1990 the centre of long-span bridge building has moved to East Asia, first to Japan, then to Korea and, overwhelmingly, to China. More than half of the world's hundred longest spans of every type have been built there since 2000.

Japan connected its islands. The Honshu–Shikoku project built three routes across the Seto Inland Sea between 1978 and 1999: the Seto-Ōhashi (1988), six long bridges carrying road and rail including the 1,100 m Minami Bisan-Seto suspension span; the Akashi Kaikyō (1998); the Tatara cable-stayed bridge (1999) and the three Kurushima suspension bridges in a row on the western route. Japanese practice, forged by typhoons and earthquakes, set the standards for wind-tunnel testing, cable protection and seismic isolation that the rest of the region adopted.

Korea began with the Hangang Railway Bridge (1900), the first modern bridge in the country, and the Hangang road bridge (1917), and then built at speed after 1970: the Namhae Bridge (1973), a 404 m suspension span and the first in Korea; the Jindo Bridge (1984), the first cable-stayed; the Seohae Grand Bridge (2000), 7.3 km with a 470 m cable-stayed span on the west-coast motorway; the Yeongjong Grand Bridge (2000), a self-anchored suspension bridge carrying road and airport railway on two decks; the Gwangan Bridge at Busan (2003, 500 m); the Incheon Bridge (2009), 18 km across the tidal flats to the airport with an 800 m cable-stayed span; the Yi Sun-sin Bridge at Yeosu (2012), 1,545 m, the first Korean-built suspension bridge of that scale and its towers among the tallest in the world; and the Ulsan Harbour Bridge (2015), a single-span suspension bridge of 1,150 m. The Seongsu Bridge collapse in Seoul (1994) belongs in this list too, because the inspection and maintenance regime of the country's bridges was rebuilt after it.

China built its first great river crossing without foreign help at Nanjing in 1968, a double-deck steel truss over the Yangtze, and then, from the 1990s, most of the world's records: the Humen Bridge (1997, 888 m), the Jiangyin (1999, 1,385 m), the Runyang (2005, 1,490 m) and the Xihoumen (2009, 1,650 m) suspension bridges; Sutong (2008) for cable-stayed; the Chaotianmen arch at Chongqing (2009, 552 m); the Beipanjiang Duge Bridge (2016), a cable-stayed span 565 m above the river; the Hong Kong–Zhuhai–Macau crossing (2018); the Changtai Bridge (2025); and the Huajiang Canyon Bridge in Guizhou (2025), a 1,420 m suspension span 625 m above the water, the highest bridge ever built. The scale is in the numbers: China now completes more long-span bridges in a year than Europe did in the second half of the twentieth century, and the engineering questions of this decade, fatigue of orthotropic steel decks, cable replacement, stay vibration, deep-water foundations, are being answered there first.

Every rule in a bridge code was paid for. The list below is not complete, but each entry changed the way bridges are designed, built or inspected, and each is covered in more detail in this site's Failures section.

Dee, 1847

Cast-iron girder broke under a train. Cast iron leaves the beam; wrought iron takes over.

Tay, 1879

Seventy-five dead in a storm on a bridge designed for almost no wind. Britain sets a design wind pressure; the Forth is built to it.

Quebec, 1907

Compression chords buckled under an underestimated dead load, 75 dead; the consulting engineer's warning had not been acted on. Built-up compression members get proper theory, and the responsibility of the engineer gets a ritual.

Tacoma Narrows, 1940

Flutter destroys a slender deck. Aerodynamic testing becomes part of long-span design.

King's Bridge, Melbourne, 1962

Brittle fracture of welded high-strength steel on a cold morning. Fracture toughness, weldability and notch tests enter the steel specifications.

Silver Bridge, Point Pleasant, 1967

One eyebar with one stress-corrosion crack, 46 dead. The United States creates the National Bridge Inspection Standards (1971): every bridge inspected at least every two years, fracture-critical members by hand.

Milford Haven, West Gate, Koblenz, 1970–71

Three steel box girders buckle during erection within eighteen months. Britain's Merrison rules and later codes rewrite the design of thin-walled steel and the checking of construction stages.

Mianus River, 1983

A corroded pin-and-hanger drops a span. Non-redundant details are named and hunted down; drainage becomes a structural matter.

Schoharie Creek, 1987

Flood scour undermines a pier footing, 10 dead. Scour evaluation of every bridge over water; the HEC-18 method on this site is its result.

Seongsu, Seoul, 1994

A suspended span of a Gerber truss falls at rush hour after years of unrepaired weld defects, 32 dead. Korea rebuilds its inspection law and its maintenance profession.

Injaka, South Africa, 1998

A launched deck collapses during construction. Temporary works and launching stages get the same scrutiny as the finished bridge.

I-35W, Minneapolis, 2007

Gusset plates half as thick as they should have been, carrying added deck weight and construction material, 13 dead. Gusset plates are load-rated; added dead load is tracked.

Polcevera (Morandi), Genoa, 2018

Corrosion inside the concrete-encased stays of a 1967 bridge, 43 dead. Ungrouted or inaccessible tendons everywhere are re-examined; replaceability becomes a design requirement.

Francis Scott Key, Baltimore, 2024

A container ship strikes a pier of a 1977 continuous truss and the whole main structure falls, 6 dead. Ship-collision protection, designed into bridges since the 1980 Sunshine Skyway disaster, is reviewed for the size of modern vessels.

The pattern is the same each time: a new material, a new scale or a new method outruns what its users understand, the structure finds the gap, and the code closes it. The engineer's job is to know where the gap is now.

Record spans by form, from the first to the present. The dates are openings; the spans are main spans in metres.

Read down the last column and the history of bridges is a history of four things arriving in turn: a form (the arch, the truss, the cable), a material (stone, iron, steel, concrete, high-strength wire), a method (centring, cantilevering, spinning, launching, match-casting) and a theory (Hooke, Euler, Navier, Melan, the wind tunnel). Each record was set the moment the fourth of these caught up with the other three, and each failure in section 2.14 happened when one of them ran ahead.