The world according to the largest medieval map that survives. I took the 1,025 names that two Victorian clergymen indexed on the Hereford Mappa Mundi, matched them to real places where I could, and set the map beside a modern one: Jerusalem at the centre, east at the top, Paradise at the edge of the world, and Britain drawn far bigger than its size.
In Hereford Cathedral hangs a sheet of calfskin about 158 cm tall and 133 cm wide, painted around 1300. It is the largest medieval map of the world still in existence. It is not a map you could travel by. Jerusalem sits at the exact centre. East is at the top, where the Garden of Eden floats in a ring of fire. The Mediterranean runs up the middle like the stem of a T, splitting Europe on the left from Africa on the right, and the far edge of the world is crowded with dog-headed men, one-footed Sciapods and people with faces in their chests. Yet it is also full of real places: Paris, Rome, Babylon, Samarkand, and dozens of English cathedral towns.
In 1873 two Herefordshire clergymen, W. L. Bevan and H. W. Phillott, published a book that went over the map name by name and ended with an index of 1,025 names. I used that index as the inventory of the map, matched the names to real places through the book's own identifications and the Pleiades gazetteer of ancient places, and located 164 of the clearest inscriptions by eye on Konrad Miller's 1896 facsimile. Click a place on either map below to see it on the other.
Left: Konrad Miller's 1896 hand-drawn facsimile of the map (the labels are clearer than on the worn original); switch to a photograph of the original in Map layers. Dots: the 164 inscriptions I located by eye, coloured by class. Right: the 143 of them that the 1873 book identifies with a real place, at that place, and (smaller dots, in Map layers) 194 more names matched in the Pleiades gazetteer but not located on the map. Grey arrows (Map layers): from each place to where the map would put it if it were an ordinary map, after the best rotation and scaling. Sources: Bevan and Phillott (1873); Wikimedia Commons (public domain images); Pleiades (CC BY 3.0); Natural Earth.
Bevan and Phillott walk around the map the way a medieval reader would: Asia first, from Paradise downwards, then Africa, then the seas and their islands, then Europe and, last of all, the British Isles. Their index of 1,025 names follows that tour. Asia, the top half of the map, has 371; Europe, the bottom left quarter, 324; Africa, the bottom right, 117; the seas and islands 105; and the British Isles, squeezed into the bottom left corner, 83. Most are settlements, regions or peoples (640); the rest are rivers (139), islands (108), mountains (69) and seas. Wikipedia's article on the map counts its pictures another way: about 420 towns, 15 Biblical events, 33 animals and plants, 32 people and 5 scenes from classical mythology.
Names on the map, by part of the world and where they come from
The 985 names in Bevan and Phillott's index, leaving out the winds, quarters, continent labels and real animals and plants of the frame. Classes as defined below. Hover or tap a bar.
Bevan and Phillott sort the map's contents into four kinds of material: Biblical, classical, legendary and what they call contemporaneous, the medieval knowledge of the mapmaker's own day, which they found "almost entirely confined to Europe". I sorted every name the same way, with a fixed rule set out in the methods below. By that rule 714 names, 72%, are classical: Greek and Roman places and peoples, most of them taken from late Roman compilers such as Pliny, Solinus, Orosius and Isidore, and from the Antonine Itinerary, a Roman road book. 93, 9%, are Biblical, from Eden's four rivers to the stations of the Exodus. 121, 12%, are medieval: Santiago de Compostela, Prague, Iceland, Hungary and the towns of Britain. And 57, 6%, are legend: Paradise, Gog and Magog, the Golden Fleece, the Minotaur's labyrinth, and the monstrous races.
The mix changes across the map. Asia holds every Biblical name (93) and most of the legendary ones (36); Africa's are almost all classical, with its marvels crowded along the southern rim; Europe's are mostly Roman, with a layer of medieval kingdoms and pilgrim shrines on top. Only in the British Isles is the map up to date.
The four kinds of material
Share of the 985 names in each class.
To compare the map with reality I measured, for each located place, its direction and distance from Jerusalem, on the map and on the globe. Directions hold up surprisingly well once the map is turned so that east is at the top. For the 140 real places I located, the median error in direction is 19 degrees, and 81% point within 45 degrees of the right way. London, for instance, is 3,611 km from Jerusalem on a bearing of 318 degrees; the map puts it at 306 degrees. Samarkand is off by just -2.
Which way from Jerusalem: map against reality
Each dot is a located place: its true compass bearing from Jerusalem (across) and its bearing on the map, reading east as up (up the chart). Dots on the diagonal point the right way. Hover a dot.
Distance is another matter. The order is kept (the rank correlation between true distance and distance on the map is 0.90), but the scale collapses towards the edge. Places within 500 km of Jerusalem, a median of 128 km away, sit a median 20% of the way from the centre to the rim. Places 3,000 km or more away, a median of 3,747 km, sit at 93%. They are 29.3 times as far in reality but only 4.8 times as far on the map. Bethlehem, 9 km from Jerusalem, is 6% of the way out; London, 401 times farther, is 90%. Everything far away is pressed against the encircling ocean.
Distance from Jerusalem: map against reality
True great-circle distance (across, log scale) against distance on the map as a share of the way from Jerusalem to the edge of the world (up). Hover a dot.
To see which regions are drawn big, I compared distances between pairs of places within each region on the map and on the ground, against the same ratio for the map as a whole. The Holy Land is drawn 4.5 times bigger than the map's average; Egypt and Sinai 1.9 times; Mesopotamia and Persia 1.6; the British Isles 1.5. India and Central Asia are squeezed to 0.71 of the average, and Iceland, Norway and the Faroes are crammed into the rim so tightly that they barely register. The map spends its space on Scripture and on home.
How big each region is drawn
Median of map distance over true distance, for pairs of located places in the region at least 100 km apart, relative to all pairs on the map (1 = average scale). Regions with at least five located places.
The other big distortion is the Mediterranean. On the map it is a straight channel running from the Pillars of Hercules at the bottom to the Holy Land at the centre, so North Africa slides round towards the south: its places point a median 69 degrees off from Jerusalem, and Egypt's 72. If the whole map is laid on the globe with the best rotation and scaling, the rotation that fits best turns the map's top to a bearing of 98 degrees, a little south of due east, and the median place lands 516 km from where it really is. Turn on the grey arrows in Map layers to see where each place would fall.
The British Isles are a sliver of the world the map covers: 0.4% of the land of Africa, Asia and Europe, by Natural Earth's outlines. They have 83 names, 8.1% of the index, about 21 times their share of the land. Bevan and Phillott noticed: "The British Isles are drawn on a scale out of all proportion to the other parts of the map." Here, for once, the mapmaker was working from his own knowledge rather than from books. The towns are mostly cathedral cities and abbeys, Lincoln with its cathedral on a hill, Durham on its rock, Ely on an island in the fens, Glastonbury, Kirkham Priory; the rivers of the Humber, they wrote, are drawn "with a degree of accuracy which implies personal acquaintance with that part of the country".
It is not flawless. Caen in Normandy has been carried across the Channel into Devon, "by a strange blunder", perhaps because its two great abbeys held so much English land. Hereford itself is only a faint drawing of a cathedral, worn, as Wikipedia notes, perhaps by centuries of pointing fingers. But measured from Jerusalem the British towns point the right way to within a median 8 degrees, among the best on the map.
The located places, region by region
Number located, median true distance from Jerusalem, median error in direction, and how big the region is drawn relative to the map as a whole. Click a row to see the region on both maps.
The inventory. The index of Bevan and Phillott's Mediaeval Geography (1873), read from the Internet Archive's scan, is the list of names: 1,025 entries, each with the page where the book discusses it. The book is arranged by part of the map, so each name's page gives its continent. The scan's text is rough, and the map's own words, printed in black-letter type, come out as gibberish ("ColcegtriH" for Colcestria); I corrected the index entries the scanning had mangled, and the full text of every page is in the data, for checking.
The classes. I followed the book's own four kinds of material. Legendary: monstrous peoples, fabulous beasts and the places and objects of legend and myth, from a list I drew up. Biblical: every name in the book's sections on the rivers of Paradise, Palestine and Arabia, and the Exodus, and a few Biblical names elsewhere (Babel, Ur, Nineveh, Noah's Ark). Medieval: in the British Isles every name except a handful that come from Roman writers, and in the rest of Europe a list of post-Roman places, regions and peoples. Classical: everything else. The winds, quarters, continent labels and real animals and plants are left out. These are judgements, made name by name; the lists are in the code.
Real places. For the 164 inscriptions I located on the facsimile I read the book's identification (Zimaris is "now probably Mangalore", Babylon in Egypt is "the city of Cairo") and placed it on the modern map. For the other names I looked for the same name in the Pleiades gazetteer of ancient places, allowing for simple medieval spellings, in the right part of the world, and checked every match by hand, rejecting 78 that pointed to a different place with the same or a similar name. In all 337 of the 928 non-legendary names, 36%, have a real-world position: 143 from the book and 194 from Pleiades. That is a lower bound on what can be identified: the book names many more rivers, peoples and towns that I did not try to place by hand.
Positions on the map. There is no coordinate dataset for the Hereford map, so I located 164 well-known, legible inscriptions by eye on Miller's facsimile, at the town picture or, failing that, the middle of the label: 119 read clearly and 45 with some doubt (worn, crowded or ambiguous). They are carried onto the photograph of the original by an affine fit on 9 features clear on both, with a typical error of 44 pixels, about 1.4% of the map's width: the facsimile is a hand-made copy and the vellum has shrunk and buckled. Directions are measured with east as up and from the centre of the walled circle of Jerusalem; distances on the map as a share of the way from Jerusalem to Paradise or the Pillars of Hercules (1,536 pixels on the facsimile).
What this is not. A sample of the map, not all of it: the located places are the legible and well-known ones, so the distortion figures describe the map's real towns, not its whole surface. The 1873 identifications are the Victorians' and some have been revised since; the book itself calls a few names unknown. For a complete modern edition of every inscription, see Scott Westrem's The Hereford Map (2001) and the Virtual Mappa project.