A Scattered across the territories of the former Roman Empire are concrete structures that have survived for roughly two thousand years: harbour walls still pounded by the Mediterranean, aqueducts that carried water across deep valleys and, most famously, the Pantheon in Rome, whose vast dome was completed in about AD 126, during the reign of Hadrian. Modern concrete, by comparison, is typically engineered to last between fifty and a hundred years, and many motorway bridges and multi-storey car parks erected in the twentieth century are already showing signs of decay. Why the ancient material should have proved so much more durable is a question that has occupied engineers, chemists and archaeologists for well over a century.

B The Romans did not invent the practice of binding stones together with lime; earlier Mediterranean peoples had used lime mortars for centuries. What the Romans achieved was to turn it into a systematic technology that could be reproduced on an imperial scale. The essential binder was quicklime, obtained by heating limestone in a kiln, which was combined with water and a volcanic ash later named pozzolana after the town of Pozzuoli, on the Bay of Naples, where particularly rich deposits occur. To this mortar, builders added fist-sized chunks of rock or broken brick, known as caementa. Crucially, the mixture was not poured into moulds as a liquid, as concrete is today. It was laid by hand in successive layers and tamped down firmly, often between two facing walls of brick or stone.

C Much of what is known about these methods comes from Vitruvius, an architect and military engineer who served under Julius Caesar and later Augustus. In De architectura, a ten-volume treatise composed in the final decades of the first century BC, he set out recipes in considerable detail. For ordinary walls, he advised, three parts of quarried sand should be mixed with one of lime; for structures built in the sea, two parts of pozzolana to one of lime. He also recorded that the powder found around Baiae and Mount Vesuvius would harden not only in air but under water. Vitruvius was, however, a practitioner rather than a scientist: he could describe what happened without being able to explain why.

D It was in the sea that Roman concrete achieved its most remarkable results. Pliny the Elder, a naval commander as well as a naturalist, claimed that concrete placed in seawater became a single stone mass, impregnable to the waves and growing stronger every day. Roman engineers valued pozzolana so highly that they shipped it across the Mediterranean; the great harbour at Caesarea, constructed for King Herod between about 22 and 10 BC, was built with ash transported from Italy. In 2017, a team led by the geologist Marie Jackson of the University of Utah demonstrated that Pliny had not been exaggerating. Analysing cores drilled from ancient piers, they found that seawater percolating through the concrete had gradually dissolved components of the ash, allowing new minerals – notably a rare crystal called aluminous tobermorite – to grow in the pores and reinforce the material. In ordinary modern concrete, by contrast, chemical reactions with seawater are usually destructive.

E A second line of explanation emerged in 2023, when researchers at the Massachusetts Institute of Technology, led by Admir Masic, examined samples from the city wall of Privernum, an ancient settlement south-east of Rome. Like much Roman concrete, these contained small white lumps of lime which had long been dismissed as evidence of careless mixing or poor-quality raw materials. Masic’s team proposed instead that the lumps were the product of ‘hot mixing’, in which quicklime was added directly rather than first being combined with water. The lumps, they argued, act as reservoirs of calcium: when a crack forms and water seeps in, the lime dissolves and then recrystallises as calcium carbonate, sealing the gap. To test the idea, they produced hot-mixed samples, cracked them deliberately and ran water through them. Within two weeks the cracks had closed completely, whereas in samples made without quicklime the water continued to flow.

F Knowledge of Roman techniques did not vanish entirely when the western empire collapsed in the fifth century AD, but the large-scale use of pozzolanic concrete declined sharply. Medieval builders relied largely on cut stone and simple lime mortar. Renaissance architects pored over surviving copies of Vitruvius, yet it was not until the eighteenth century that binders capable of setting under water were again investigated systematically. In the 1750s, the British engineer John Smeaton, seeking a mortar that could withstand the waves for his Eddystone Lighthouse off Plymouth, discovered that limestone containing clay produced a lime that would set in wet conditions. In 1824, a bricklayer named Joseph Aspdin patented ‘Portland cement’, so called because it resembled a pale building stone quarried on the Isle of Portland.

G Interest in the subject is not merely historical. Manufacturing Portland cement requires kilns heated to around 1,450°C, and the industry is estimated to be responsible for roughly eight per cent of global carbon dioxide emissions; Roman lime could be burned at substantially lower temperatures. Nevertheless, few engineers suggest that the ancient recipe could simply replace its modern successor. Roman concrete is considerably weaker under compression, cures slowly, and was never designed to work alongside the steel reinforcing bars on which contemporary bridges and tower blocks depend. The likelier outcome is that principles drawn from ancient chemistry – volcanic additives, self-healing lime, lower firing temperatures – will be incorporated into new mixes.

A Sleep is, on the face of it, a hazardous habit. An animal that is asleep cannot feed, cannot search for a mate and, above all, is poorly placed to notice an approaching predator. If the state conferred no essential benefit, one might expect natural selection to have eliminated it long ago. Yet every animal that has been examined carefully appears to sleep in some form. The American sleep researcher Allan Rechtschaffen once remarked that if sleep does not serve an absolutely vital function, it must be the biggest mistake evolution ever made. Half a century later, scientists still disagree about what that function is – or whether there is only one – but research on species far removed from ourselves has transformed the terms of the debate.

B Studying sleep in animals requires a definition that does not depend on asking the subject how it feels. Researchers generally rely on behavioural criteria: a characteristic resting posture, a reduced response to stimuli, the capacity to be roused quickly, and a ‘rebound’ effect, whereby an individual deprived of rest compensates by resting more later. Using these tests, sleep has been identified in fruit flies, zebrafish and nematode worms. Most strikingly, in 2017 a team at the California Institute of Technology reported that Cassiopea, the so-called upside-down jellyfish, met every criterion, even though it possesses no brain, only a loose network of nerve cells. At night its pulsing slowed by roughly a third; individuals took noticeably longer to react when the surface beneath them was removed; and when they were disturbed by jets of water throughout the night, they were less active the following day. If so simple a creature sleeps, the behaviour may be far older than the brain itself.

C The most extraordinary variations, however, are found among animals for which losing consciousness entirely would be fatal. Dolphins, which must rise to the surface to breathe, cannot afford the deep, motionless sleep of land mammals. In the 1970s, the Russian neurophysiologist Lev Mukhametov, recording electrical activity in the brains of bottlenose dolphins, discovered that the animals allowed only one hemisphere to sleep at a time while the other remained alert. The halves took turns, and during each period the eye connected to the waking hemisphere stayed open. This ‘unihemispheric’ sleep enables a dolphin to keep swimming, surfacing and monitoring its surroundings. Its importance was underlined in 2005, when researchers observing captive bottlenose dolphins and killer whales reported that newborn calves and their mothers showed almost no sign of rest during the calves’ first month of life – a pattern that would cause serious harm in a land mammal.

D Birds offered the first experimental evidence that half-brain sleep might serve as a defence against predators. In a study published in 1999, Niels Rattenborg and colleagues at Indiana State University filmed and recorded mallard ducks resting in a row of four. The ducks in the two central positions, flanked by companions, slept mostly with both hemispheres. Those at the ends, however, increased their use of unihemispheric sleep by around 150 per cent, and consistently directed their open eye away from the group, towards the open space from which a predator would be most likely to approach. The ducks, in other words, were not simply sleeping lightly; they were regulating which side of the brain stayed on guard according to where danger lay.

E Some seventeen years later, Rattenborg, by then at the Max Planck Institute for Ornithology in Germany, addressed a long-standing puzzle: how do seabirds that remain aloft for weeks manage to sleep at all? With colleagues, he fitted great frigatebirds nesting in the Galápagos Islands with small devices that recorded brain activity during flights over the ocean lasting up to ten days. The results, published in 2016, showed that the birds did sleep on the wing, sometimes with one hemisphere and occasionally with both, mostly while circling in thermals – columns of rising warm air. The episodes were extraordinarily brief, averaging around twelve seconds, and added up to only about 42 minutes a day, compared with more than twelve hours a day when the same birds were back on land. While circling, the waking hemisphere was usually the one linked to the eye facing the direction of the turn, suggesting that the birds were watching where they were going.

F Such findings suggest that sleep is far more flexible than was once assumed. Northern fur seals, studied by Oleg Lyamin and Jerome Siegel, sleep with both hemispheres on land and display the rapid eye movement (REM) stage familiar in humans; in the water, however, they switch to unihemispheric sleep and REM almost disappears, sometimes for weeks, apparently without ill effects. This challenges theories that treat REM as indispensable for mammals. Equally surprising, a 2017 study led by Paul Manger of the University of the Witwatersrand, which tracked two wild African elephants in Botswana, found that they slept for only about two hours a day, often standing up. None of this means that sleep is unimportant. Experiments on mice by Maiken Nedergaard and colleagues in 2013 indicated that during sleep the spaces between brain cells widen, flushing out waste products that accumulate during waking hours – a function that may be shared by many species. What the animal kingdom reveals is not that sleep is optional, but that evolution has found remarkably varied ways of accommodating it.

A Of the roughly 7,000 languages spoken in the world today, around 40 per cent are classed as endangered, many of them with fewer than a thousand speakers. In 1992, the linguist Michael Krauss of the University of Alaska published a paper in the journal Language predicting that, by the end of the twenty-first century, as many as 90 per cent of the world’s languages might be either extinct or doomed. In the same article he rebuked his own profession for its complacency, warning that linguistics might go down in history as the only science that had looked on while most of what it studied vanished. Krauss’s figure may yet prove too pessimistic – the rate of loss is notoriously difficult to measure – but the direction of travel is not in serious doubt.

B Those who argue for preservation usually begin with heritage. Each language, on this view, is a unique cultural achievement, containing stories, humour and habits of thought that cannot be fully translated. Kenneth Hale, a linguist at the Massachusetts Institute of Technology who spent decades working with Aboriginal Australian and Native American communities, is widely quoted as likening the loss of a language to dropping a bomb on the Louvre. The image is vivid. Yet the analogy is imperfect in one important respect. A painting has no say in whether it survives, whereas a language lives only for as long as people choose to speak it and pass it on to their children. To treat a language as an exhibit is, in a sense, to forget the speakers.

C A second argument concerns knowledge. K. David Harrison, in his book When Languages Die (2007), points out that many small languages encode a detailed understanding of the local environment: the behaviour of animals, the medicinal uses of plants, the signs of changing weather. The Tofa, reindeer herders in Siberia, have terms that classify a domesticated reindeer by its age, sex, fertility and whether it can be ridden, so that a single word conveys what English needs a whole phrase to express. There is a scientific case, too: every language is a test of theories about what human grammar can and cannot do, and a theory built only on the world’s major languages rests on a dangerously narrow sample. Of the arguments for preservation, this seems to me the more persuasive, since it rests on evidence rather than on sentiment.

D Not every linguist agreed. In a response to Krauss and his colleagues, published in the same journal that year, Peter Ladefoged of the University of California, Los Angeles, warned against a paternalistic attitude towards communities that were abandoning their languages. Speakers, he observed, may have sound reasons of their own for adopting a more widely used tongue, and it is not for outsiders to tell them otherwise. In his view, the responsible linguist should document a language as fully as possible, but should not campaign for its survival. A blunter version of this position holds that linguistic diversity is simply an obstacle, and that a world with fewer languages would be one of greater harmony. This last claim, at least, is easily dismissed: some of the bloodiest civil wars in history, from the American to the Spanish, were fought between people who spoke the same language.

E Ladefoged’s warning deserves to be taken seriously; it would be absurd for academics to insist that a community remain loyal to a language that offers its children few opportunities. But the notion of free choice is harder to sustain when one examines how most languages have actually declined. In nineteenth-century Wales, pupils caught speaking Welsh in some schools were made to wear a wooden board known as the ‘Welsh Not’; in Canada, Australia and the United States, indigenous children were sent to residential schools where their languages were often forbidden. A decision taken by parents who were themselves punished for speaking their mother tongue can hardly be described as neutral. Much apparent ‘choice’ turns out, on closer inspection, to be the long shadow of coercion.

F Nor is decline irreversible. Hebrew, for centuries confined largely to prayer and scholarship, became an everyday language again, though in unusual historical circumstances. More instructive are recent efforts. In New Zealand, Māori ‘language nests’, pre-schools in which young children are cared for entirely in Māori, have operated since 1982. In Hawaii, where by the mid-1980s fewer than fifty children could speak the indigenous language, similar immersion pre-schools opened in 1984, and their former pupils have gone on to Hawaiian-medium schools and university courses. What these cases have in common is that the initiative came from the communities themselves, with public funding following later. Recordings and dictionaries, however valuable, cannot do this work alone: an archive preserves a language much as a museum cabinet preserves a butterfly.

G It does not follow that every language can, or should, be rescued by outsiders. Where a community has genuinely decided to move on, the most useful contribution linguists can make is a thorough record. But where speakers wish to keep their language, governments should remove the obstacles that remain, rather than treating disappearance as a natural and inevitable process. A common objection is that maintaining a minority language holds children back. Research points the other way: children are entirely capable of growing up with two or more languages, and there is no sound evidence that doing so impairs their development. Supporting a language, in the end, does not mean preserving it in amber; it means giving its speakers a genuine choice.