A dark world that lights up as travellers wrote it down. How much of the Earth each tradition had recorded, from the Bronze Age to 1960, and how that depended on who was doing the looking.
There is a famous animation of world exploration from Europe's point of view. The world starts black, and as each ship and caravan sets out, the places it reached turn from dark to light. It is mesmerising, and it has two problems. It leaves Europe's own neighbours dark for centuries after Greek and Roman writers had described them, and it presents one tradition's point of view as if it were the only one.
This is my attempt at an honest version. I compiled 681 journeys, from Austronesian voyages out of Taiwan around 2200 BC to the last Antarctic and Arabian crossings of the 1950s, each with a dated route. A place lights up when a journey whose account reached its readers passed close enough to describe it. Europe's world starts from the Bronze Age Aegean, not from nothing. Because much of the world was mapped not by expeditions but by surveyors and settlers, Europe's map also has a second layer: 52 regions lit from the year their systematic survey was published or settlement became general, from the Mitchell map of the American colonies to the aerial survey of northern Canada. You can switch that layer off, and you can switch the point of view to the Islamic world, China or Polynesia and watch a different world light up.
Press play, drag the slider or jump to a moment. The time scale is uneven: the ancient centuries pass quickly and the years after 1400 slowly. Click a route for its story, or anywhere on the map to see who first recorded it and when.
Each grid cell (an equal-area hexagon of about 1,770 km²) lights up in the year the nearest point of a route passed within reach: 150 km of open sea and 40 km of coast for a ship, 120 km either side of a land, river or sledge route, 150 km for a flight, 250 km around the points of a starting world. Small enclosed pockets fill in. For Europe, a survey and settlement layer lights whole regions from the year they were systematically mapped or settled (switch it off for journeys alone). Knowledge flagged as lost (Vinland, Norse Greenland, the Phoenician voyage round Africa) goes dark again unless another journey covers it. Routes drawn as they happen, coloured by century. Base map, borders and names: Natural Earth; optional relief tiles © Esri.
For Europe, the first big step is Ptolemy. His Geography, written around AD 150, sums up what the Roman world knew, from Ireland to the edge of China, and on its own lifts Europe's recorded world from 7% to 15% of the land. Then, for more than a thousand years, very little: by 1400 the figure is 20%. In 1492 it is 20% of the land and 11% of the Earth's surface.
The ocean voyages change the sea first. Europe's known share of the Earth's surface passes half in 1809; the land takes until 1841, and three quarters of it until 1936. The nineteenth century adds the most, 23.0% of the world's land, as the crossings of Africa, Australia and Central Asia coincided with the surveys of India, the American West and the prairies; the fastest single decade after 1400 is the 1870s (5.9%). From journeys alone the curve is lower and flatter at the end: half the land by 1861 and 64% in 1960, against 80% with surveys and settlement.
Share of the world known, by point of view
Share of the world's land known to each tradition by year; for Polynesia, whose world was mostly ocean, the share of the Earth's whole surface. The dashed red line is Europe from journeys alone. The time axis is uneven, like the slider. Hover the chart.
Switch the point of view and the map changes. The Islamic world's written geography, from the Arab conquests through al-Masudi, al-Biruni, al-Idrisi and Ibn Battuta, covers 17% of the land by 1400, from al-Andalus to China and down the African coast to Sofala. China's, from Zhang Qian's mission to the west in 138 BC to Zheng He's fleets, reaches 12% by 1492. Polynesian and wider Austronesian voyaging, dated here by archaeology rather than writing, lit far more ocean than land: 3.2% of the Earth's surface by 1400, but only 0.3% of its land.
In this gazetteer Europe, starting from the Aegean and with the Greek and Roman geographers behind it, leads on land from 599 BC onwards and is never behind the Islamic world. That lead is partly a matter of definition: Ptolemy's Geography is counted as one starting world for Europe, and in the confirmed-only check below, where entries whose dates could not be confirmed are dropped, the Islamic world leads until 1535.
Much was known elsewhere first. Of the area China's travellers recorded, 56% was lit for China before it was for Europe, a median 616 years earlier, mostly within China itself, Mongolia and Russia. Polynesians settled Hawaii around 1219 and Cook arrived in 1778; Rapa Nui around 1250, and Europe's first sight was 1722; Aotearoa around 1281, against Tasman in 1642. Chinese records describe Japan from the third century, Europe's from 1549.
Known to another tradition before Europe
Area lit for each point of view before it was lit for Europe (or never lit for Europe), million km².
Measured by the new area each journey was the first to light for Europe, the biggest are ocean voyages: the second leg of Cook's second voyage, 1773 to 1775, round the high southern latitudes, lit 8.7 million km²; Magellan's crossing of the Pacific 6.9 million. On land, where a route lights a strip about 240 km wide, the biggest are long journeys across unrecorded country: John C. Fremont in 1843, Agents of Maes Titianus around AD 100, and Daniel Gottlieb Messerschmidt across Siberia in 1719.
The journeys that lit the most for Europe
New area first lit by each journey (sea and land, and land only), excluding starting worlds. Click a row to see it on the map.
By the halfway measure, the last large countries to be recorded for Europe are Papua New Guinea (half its land lit by 1933), Canada (1921, its north mapped from the air) and Namibia (1910). After 1900, 19.7 million km² of land was still being lit for the first time, most of it in northern Canada, Antarctica, the Brazilian interior and Alaska.
The map never goes fully light, and that needs saying plainly. In 1960 29.4 million km² of land is still dark, 10.4 million of it in Antarctica. Much of the rest is not land nobody had described but land that neither a journey in my gazetteer nor my deliberately coarse survey layer reaches: parts of the Sahara and the Sahel, Cameroon and the Central African Republic, the outer islands of Indonesia, the Amazon, northern Siberia and the Greenland ice. Read the late years as a lower bound.
72 of the journeys name a local guide, pilot or interpreter in their notes, and many more relied on one unnamed. A few: the Gedrosian pilot Hydraces, who guided Nearchus's fleet along the coast of Pakistan; the Indian sailor who showed Eudoxus the monsoon route; Ma Huan, Zheng He's Muslim interpreter; the Gujarati pilot who took da Gama from Malindi to Calicut; Malintzin, Cortes's interpreter; Enrique, the Malay interpreter of Magellan's fleet; Tupaia, the Tahitian navigator who sailed with Cook and drew him a chart of the islands; Matonabbee, who led Hearne to the Arctic coast; Sacagawea and the Shoshone guide Old Toby with Lewis and Clark; Kadu, the Marshallese navigator on the Rurik; Sidi Mubarak Bombay, caravan leader to Burton, Speke and Stanley; Kehu in New Zealand; and the Bedouin Salih bin Kalut, who took Bertram Thomas across the Empty Quarter.
The gazetteer. I compiled 681 journeys in eight regional files: Greek and Phoenician, Roman and medieval, the Atlantic and Indian Ocean, the Americas, the African interior, overland Asia, the Pacific and the poles, and the other points of view. 571 are Europe's, 42 the Islamic world's, 43 China's and 25 Polynesian or Austronesian. Each has a leader, sponsor, years, mode (sea, land, river, sledge, air, or a starting world), a dated route of 16,038 waypoints in all, a one-line summary, sources and notes. Routes follow coasts and named stops; 181 are marked low confidence because the route is reconstructed or disputed. Duplicates across files were merged (3) or renamed (2). Polynesian dates are settlement dates from archaeology and radiocarbon, not written accounts.
Sources checked. Every journey carries a source status, shown in its popup. 27% have their start and end years confirmed in the text of the cited Wikipedia article by an automatic check; 60% were taken by the compiler from the pages they cite; 11% cite an article that does not contain one or both years, or have dates partly from general knowledge; and 2.5% cite a page that could not be read or found. I ran every result again with only the first two groups (587 journeys) as a check.
The model. The Earth is divided into 288,122 equal-area H3 hexagons. Routes are cut into points every 8 km with years interpolated between dated waypoints, and a cell is lit by a journey when its centre is within reach of a point (distances above), in the year of the nearest point. A cell's first year is the earliest over all journeys of a point of view. An unknown pocket of fewer than 150 cells (about 266,000 km², roughly the size of the United Kingdom) wholly ringed by known cells counts as known from the year its rim was complete: this fills 8.5% of Europe's known area. A cell is land if its centre is on Natural Earth land, including Antarctic ice shelves.
Surveys and settlement. Journeys cannot capture the township surveys of the United States and the Canadian prairies, the trigonometrical survey of India, the Russian land survey or the settlement of the Pampas and of south-eastern Australia. For Europe only, I added 52 regions, defined by modern first-level divisions or countries and sometimes clipped to their settled part, each lit in one year between 1734 and 1955: when its systematic survey was published in Europe or settlement became general. They add 16.3% of the world's land by 1960. Their dates are mine, from general knowledge of the history of cartography, and are marked as not confirmed; they are coarse by design, and the page and data give every result with and without them.
Sensitivity. The radii, the fill and the survey layer are choices, so I varied them. Without the survey layer Europe's 1960 land share is 64%. Halving every radius brings it to 72%; increasing them by half raises it to 86%; no fill gives 76% and a fill of up to 400 cells 86%. Keeping only entries with confirmed or compiler-read sources, which also drops the survey layer, gives 58% and moves the year half the land was lit from 1841 to 1877. The shape of the curve, the slow millennium, the jump after 1492, the nineteenth-century surge, survives every variant; the exact percentages do not, so read them as indications.
Sensitivity of the main results
Europe's share of the land lit, by year, and the year a quarter and half of the land was lit, under each variant.
What this is not. It is not a measure of what any society knew. It measures how much of the Earth lies close to the journeys in one gazetteer, plus a coarse survey layer for Europe, compiled by me with AI assistance from mostly English-language sources, which know Europe's explorers far better than Arabic, Persian, Chinese or Polynesian traditions; the other points of view are therefore thinner than they should be. Trade, missions and much colonial mapping are missing, and the other traditions have no survey layer. Dates are the years a journey took place, not the years its account was published. Routes are approximations, and a fixed reveal distance treats a desert crossing and a forest river alike.