A For travellers heading west out of China in the first millennium AD, the oasis of Dunhuang was the last place of real comfort before the desert. Founded as a military outpost under the Han dynasty in the second century BC, it stood close to the point where the caravan route divided, one branch skirting the northern edge of the Taklamakan Desert and the other its southern edge. Merchants, soldiers, monks and diplomats all passed through, and many of them left offerings at a complex of Buddhist shrines carved into a cliff some 25 kilometres south-east of the town. According to a later inscription, the first of these Mogao Caves was hollowed out in 366 AD by a monk named Lezun after a vision of a thousand glowing Buddhas. Over the following thousand years, local rulers and wealthy families paid for hundreds more, and 492 decorated caves survive today, containing some 45,000 square metres of wall paintings.

B By the late nineteenth century, however, the site had been largely forgotten, and drifting sand had buried many of the lower entrances. In 1900 a Taoist priest named Wang Yuanlu, who had made himself the unofficial guardian of the caves, was clearing one of these entrances when he noticed a crack in a painted wall of the passage. Behind it lay a small chamber crammed from floor to ceiling with bundles of paper scrolls, paintings on silk and banners. Estimates of its contents vary, but the figure most often quoted is around 50,000 items. The latest dated document in the chamber was written in the first years of the eleventh century, which suggests that the room was walled up not long afterwards and remained untouched for almost nine hundred years. Wang informed the regional authorities, but the provincial government, unwilling to meet the cost of moving such a quantity of material across hundreds of kilometres of desert, simply instructed him to look after it.

C Why the chamber was closed up in the first place remains a matter of debate. The room had originally been built in the ninth century as a memorial chapel for a senior local monk, Hongbian, whose clay portrait statue stands in it today. One long-standing explanation holds that the monks hid their library in haste when the region was threatened by invading armies. Yet the deposit does not look like a carefully chosen treasure. A large proportion of it consists of incomplete copies, damaged fragments and duplicates. This has led a number of scholars, notably the Japanese historian Fujieda Akira, to argue that the chamber served as a storeroom for worn-out texts which, being sacred, could not simply be thrown away.

D News of the discovery spread slowly, but in 1907 it reached Aurel Stein, a Hungarian-born archaeologist working for the British government in India. Wang was at first reluctant to show him anything. Stein, who could not read Chinese, won the priest’s trust partly by expressing his devotion to Xuanzang, the seventh-century Chinese monk who had travelled to India in search of Buddhist scriptures and whom Wang greatly admired; Stein presented himself as following in the pilgrim’s footsteps. Working through his Chinese secretary, he was allowed to examine bundles brought out a few at a time, and he eventually left with many thousands of manuscripts and paintings. In return he made a donation, which Wang spent on the restoration of the caves.

E A year later the French scholar Paul Pelliot arrived. Unlike Stein, Pelliot was fluent in Chinese, and he was permitted to enter the chamber itself. Over roughly three weeks, working by candlelight, he went through every bundle, setting aside those he judged most valuable along with a large number of texts in other languages. When he showed some of his finds to scholars in Beijing in 1909, they were alarmed to learn what had been leaving the country, and the following year the government ordered that the remaining manuscripts be transported to the capital. The order was carried out carelessly: officials along the route helped themselves to scrolls, and some were cut into pieces so that the total count would still match. In 1924 the American Langdon Warner removed sections of painted plaster from several caves.

F The contents of the chamber have transformed the study of medieval Central Asia. Most of the texts are Buddhist scriptures in Chinese, but documents also survive in Tibetan, Sanskrit, Khotanese, Sogdian and Old Uyghur, together with a Hebrew prayer that may have belonged to a travelling merchant. Equally valuable are the ordinary records that official histories ignored: contracts for the sale of land and livestock, letters, schoolchildren’s writing exercises, medical prescriptions and model letters showing how to compose a polite apology. One scroll, now in London, contains the oldest known complete map of the stars, drawn in the seventh century and showing more than 1,300 of them.

G Because the collection was divided between London, Paris, St Petersburg, Beijing and Japan, scholars long had to travel from one library to another to piece together documents that had once been rolled up side by side. In 1994 the British Library launched the International Dunhuang Project to photograph the scattered material and reunite it online, and hundreds of thousands of images can now be viewed by anyone. The caves themselves face a different threat. The moisture and carbon dioxide in visitors’ breath damage the fragile paintings, and the Dunhuang Academy, which manages the site, now limits daily admissions and shows many tourists high-resolution digital reproductions before they enter only a handful of caves.

A To anyone standing on its surface, a glacier appears as motionless as the mountains that enclose it. In reality, almost every glacier is on the move, typically at somewhere between a few metres and a few hundred metres a year, and a small number travel very much faster. That a body of solid ice could flow at all was by no means obvious to early naturalists. Alpine villagers had long been aware that glaciers advanced and shrank over the years, and that boulders dropped by the ice often lay many kilometres from the rock they had come from. What actually propelled the ice, however, remained a puzzle until the nineteenth century, when scientists first set out to measure glacial motion systematically.

B One of the earliest to do so was the Swiss naturalist Louis Agassiz, who in the early 1840s hammered a straight row of stakes across the Unteraar Glacier in the Bernese Alps. Within a few years the line had been bent into an arc bowing downhill, demonstrating that the centre of the glacier advanced more rapidly than its margins, which were retarded by friction against the valley sides. At around the same time the Scottish physicist James David Forbes was carrying out careful surveys on the Mer de Glace above Chamonix. Forbes showed that the ice kept moving through the night and in winter, and he argued that a glacier behaves like a viscous substance, capable of flowing slowly while remaining solid. Both men claimed credit for key observations, and their quarrel became notoriously bitter. Forbes’s theory was in turn challenged by the Irish physicist John Tyndall, who maintained that ice moves by repeatedly fracturing under pressure and then refreezing.

C A convincing physical account emerged only in the twentieth century. In the early 1950s John Glen, working at the University of Cambridge, subjected ice samples to controlled loads in the laboratory and recorded how quickly they deformed. He found that ice creeps, the layers within its crystals slipping past one another, and that the rate of deformation rises steeply as the stress increases: approximately with its cube, so that doubling the load makes the ice deform around eight times as fast. This relationship, now known as Glen’s flow law, explains why thickness matters so much. A glacier deforms chiefly near its base, where the weight of the overlying ice is greatest, whereas near the surface the ice is under little pressure and therefore behaves in a brittle fashion, splitting open into crevasses. These fissures rarely extend more than a few tens of metres down, because at greater depths the ice simply flows inwards and closes them.

D Internal deformation is not, however, the only mechanism at work. Where the base of a glacier is at its melting point, the ice can slide over its bed on a thin layer of water, and where it rests on soft sediment, that sediment may itself be smeared forward beneath it. The relative importance of these processes varies enormously. Glaciers in the coldest parts of Antarctica are frozen to the rock beneath them and move solely by creep, in some cases covering only a metre or two in a year. In temperate mountain glaciers, by contrast, sliding can account for more than half of all movement. The pressure of water at the bed is critical: when it approaches the weight of the ice above, the glacier is partly lifted off its bed and may accelerate markedly.

E This sensitivity to water became a pressing concern in 2002, when Jay Zwally of NASA and colleagues reported GPS measurements from a site in west Greenland showing that the ice sheet sped up each summer as surface meltwater drained through vertical shafts, known as moulins, to its base. Their finding raised the alarming prospect that a warming climate would lubricate the ice sheet, hastening its flow towards the sea. Subsequent research has complicated this picture. As melting intensifies, the network of passages beneath the ice can reorganise into large, efficient channels that carry water away rapidly, lowering the pressure and allowing the ice to settle back onto its bed. In 2015 a team led by Andrew Tedstone reported that, in one land-terminating region of south-west Greenland, ice flow had slowed by roughly 12 per cent over two decades, despite a marked increase in melting.

F Perhaps the most dramatic behaviour is shown by surge-type glaciers, which make up only around one per cent of the world’s total. After decades of sluggish movement, such a glacier may suddenly accelerate to ten or even a hundred times its normal speed for months or years. In 1953 the Kutiah Glacier in the Karakoram advanced more than 12 kilometres in about three months. When the Variegated Glacier in Alaska surged in 1982–83, a team including Barclay Kamb of the California Institute of Technology drilled boreholes to its bed and found that the usual system of tunnels had given way to a network of linked cavities that trapped water at high pressure. The surge ended abruptly when the tunnels re-formed, releasing a flood from the glacier’s snout. Why surges cluster in particular regions, including Svalbard, Alaska and the Karakoram, is still disputed, but satellite radar, first used to map ice motion in Antarctica by Richard Goldstein and colleagues in 1993, now allows scientists to monitor such events across entire mountain ranges.

A Few problems trouble established democracies more persistently than empty polling stations. Turnout at national elections has declined in many wealthy countries since the 1980s, and the fall has not been evenly spread: those who stay at home are disproportionately young, less affluent and less formally educated. One remedy, adopted in roughly two dozen countries though enforced in only a minority of them, is to make voting a legal obligation. Discussion of the idea tends to be conducted as though the only issue at stake were individual liberty. That framing, in my view, obscures most of what makes compulsory voting worth examining, which is what it reveals about the purposes elections are supposed to serve.

B The best-known case is Australia, which made voting in federal elections compulsory in 1924 after a private member’s bill passed through parliament with remarkably little debate. At the election of 1922, fewer than 60 per cent of registered voters had taken part; in 1925, the first contest held under the new rules, the figure exceeded 91 per cent, and it has never since fallen much below 90 per cent. What is striking is how mild the sanction is. A first offence at federal level attracts a fine of 20 Australian dollars, and a reasonable excuse is usually accepted. A penalty of that size is hardly a deterrent in itself, which suggests to me that the law works less by threatening citizens than by establishing voting as something everyone is expected to do. It helps, too, that elections are held on Saturdays and that early and postal voting are widely available.

C Supporters argue that the effects reach well beyond the turnout figures. In his 1997 presidential address to the American Political Science Association, Arend Lijphart contended that unequal participation produces unequal influence, since politicians respond to those who vote, and that compulsion was the only practical way to bring participation close to universal. Evidence of the kind of difference this can make came from Anthony Fowler, who in 2013 exploited the fact that Australian states introduced compulsory voting at different times. He estimated that the reform raised turnout by about 24 percentage points and increased the Labor Party’s share of the vote, and he linked the change to a subsequent expansion of national spending on old-age pensions. Such findings are suggestive, but I would caution against assuming that a reform introduced in a young federation a century ago would have the same consequences in a modern, highly polarised society.

D Opponents raise objections of two different kinds. The philosopher Jason Brennan argues that citizens have no general duty to vote at all, only a duty to vote well if they choose to do so; compelling the indifferent and the poorly informed to take part, he suggests, may lower the quality of collective decisions rather than improve it. Annabelle Lever takes a different line. She doubts whether the evidence for the benefits claimed by supporters is anywhere near strong enough to justify coercion, and she defends a right to abstain: staying away from the polls, she argues, can itself be a legitimate form of political expression.

E Defenders reply that compulsion demands surprisingly little. Lisa Hill, a political scientist at the University of Adelaide, notes that because the ballot is secret, nobody can be forced to choose a candidate: a voter may leave the paper blank or deliberately spoil it. In elections to Australia’s House of Representatives, around five per cent of ballots are typically informal. Hill also presents compulsory voting as a solution to a collective action problem. Everyone gains from an electorate that is representative, yet each individual has little reason to bear the small cost of voting, so a modest legal requirement ensures that the shared benefit is actually achieved. The so-called ‘donkey vote’, in which a disengaged voter simply numbers the candidates in the order they appear, is often cited by critics, but most estimates put its effect at around one or two per cent, and two Australian jurisdictions remove even that advantage in their own elections by printing ballot papers with the names in rotating order.

F What happens when compulsion is withdrawn offers a further test. The Netherlands abolished compulsory voting in 1970, and at the general election held the following year turnout dropped from roughly 95 per cent to about 79 per cent. In Switzerland, by contrast, the canton of Schaffhausen has retained a small fine for non-voting, and its turnout has consistently stood well above the national average, which in most federal votes is below 50 per cent.

G My own conclusion is that compulsory voting is a blunt instrument, but a defensible one. It is remarkably effective at reducing the gap between those who vote and those who do not, and the burden it places on individuals is, in most systems, trivial. What it cannot do is produce the informed engagement on which good democratic decisions depend; a citizen who attends the polling station under protest is no better informed for having done so. Compulsion should therefore be regarded as a starting point rather than a solution — worth adopting only alongside serious civic education, convenient voting arrangements and a clearly marked option allowing voters to reject every candidate on the ballot.