A Roman map of the roads from Britain to India, squeezed onto a strip of parchment 6.8 metres long and 34 centimetres high. I found 465 of its places by eye, set the scroll beside a modern map, and measured 1,767 of its road legs. The shape is wildly wrong. The distances, mostly, are right.
In Vienna, in the Austrian National Library, lies a medieval copy of a Roman map. It is made of eleven sheets of parchment glued end to end into a strip about 6.8 metres long and only 34 centimetres high. On it the whole Roman world, from the Rhine mouth to Sri Lanka, is pressed flat like a flower in a book. Seas become thin ribbons. Italy lies on its side. Rivers run along the strip instead of across it. Yet the map is covered in roads, drawn as red lines with a little kink at every stopping place, and beside each stretch a number: how far to the next station.
The first sheet, which showed Spain and most of Britain, was lost long ago. What survives is known as the Peutinger Table, the Tabula Peutingeriana, after Konrad Peutinger, the Augsburg scholar who owned it in the sixteenth century. Richard Talbert's modern edition identifies its places, and the OmnesViae project by René Voorburg turns it into a dataset of 2,762 stations and 2,971 legs on the surviving sheets, with the lost western end rebuilt from another Roman road book. I used that dataset, found 465 of the places by eye on Konrad Miller's 1887 facsimile, and set the scroll above a modern map. Click a place on either map to see it on the other.
Top: Konrad Miller's 1887 hand-drawn facsimile of the Tabula (public domain). Dots: the 465 places I located by eye, coloured by region. In Map layers you can bend the modern coastline onto the scroll, show the parchment joins, and keep the modern map following the scroll as you drag. Bottom: the 2,103 surviving stations and 471 reconstructed ones that OmnesViae places, and the Tabula's legs, coloured by how well the written distance matches the road (Map layers). Sources: OmnesViae (René Voorburg) after Talbert (2010); Itiner-e (CC BY 4.0); DARE provinces (CC BY-SA); Natural Earth.
The numbers on the Tabula are not labelled with a unit, and the OmnesViae dataset stores them as plain numbers. Scholars read most of them as Roman miles of about 1.48 km, and those in Gaul as Gallic leagues of one and a half miles. The map itself drops a hint: beside Lyon it says "Lugduno caput Galliarum usque hic legas", Lyon, capital of the Gauls, leagues up to here.
The measurements back this up. For the 1,488 measurable legs outside Gaul, each unit on the map covers a median 1.41 km as the crow flies and 1.64 km along the Roman road network mapped by the Itiner-e project. In Aquitania, Lugdunensis and Belgica, the three Gallic provinces beyond Narbonensis, and in the two Germanies, the same unit covers 2.15 km as the crow flies, 1.52 times as much: very close to the one and a half of a league. In Narbonensis, the old Roman province on the Mediterranean, the figure falls straight back to 1.44. Read as miles, Gaul's roads would come out 1.58 times longer than the figures say; read as leagues, the ratio is 1.05, right to within a few per cent.
How many kilometres one unit on the map covers, province by province
Median great-circle distance per unit written on the map, for provinces with at least 25 measured legs. Hover a dot.
Converted that way, the Tabula's distances hold up. Across the 1,558 legs where both stations sit on a mapped Roman road, the median road distance is 10% longer than the figure on the map, and the median straight line 4% shorter, so the written figures sit just where a road distance should. 53% of legs agree with the road to within 25%, and 70% to within 50%. The misses are the familiar faults of a copy of a copy: a numeral dropped or doubled, a station skipped, a road joined to the wrong town.
Each leg: the distance on the map against the distance along the road
1,558 surviving legs with a figure and a road distance (log scales). The diagonal is perfect agreement; the band is 25% either way. Hover or tap a dot.
Over long journeys the errors cancel. Adding up the Tabula's own figures from Rome, the road to Cologne comes to 1,487 km, against 1,444 km along the mapped Roman roads; to Vienna 1,120 against 1,038; to Constantinople 2,081 against 2,335. The farther east, the more the Tabula's totals run short of the mapped roads: by about an eighth on the way to Antioch (2,980 against 3,411) and Jerusalem (3,499 against 4,031).
The far east is different. In Persia and India, where the map goes beyond anything Rome governed, only 27 legs can be measured, and one unit stands for anything from 0.8 to 18.7 km. Some look like Roman miles, some like Persian parasangs of five or six kilometres, and the Indian ones fit nothing. The map's makers were copying figures they could not check.
The places I located run from Gesocribate (Brest) in the west to Palibotra (Patna) in the east, about 8,526 km, and from Lvgdvno on the Rhine mouth to Mvzeris in south India, about 4,676 km. The real world they cover is 1.8 times as wide as it is tall. The scroll is 20.9 times as wide as it is tall. So one centimetre of parchment stands, on average, for 12.6 km east to west but 137.5 km north to south: the world is squashed about 10.9 times more in one direction than the other.
Still, the order is kept. West to east along the scroll follows west to east on the ground almost perfectly: the rank correlation between a place's longitude and its position along the scroll is 0.98. North to south is much looser (0.74 for latitude against height on the scroll), because whole regions are turned on their sides to fit. Italy runs along the scroll from its north-west to its south-east, and so does North Africa, laid out beneath it like a second long peninsula.
How stretched each region is drawn
For each region, the best straight-line fit from ground to scroll: how many times more parchment a kilometre gets in the most-stretched direction than in the least. Regions with at least 8 located places. Hover a bar for the direction the scroll runs along.
Region by region, the squeeze varies. North Africa is the most stretched (55 times), a long ribbon of coast with almost no depth; the Danube lands come next (21 times), then Egypt, Greece and Italy. Asia Minor (7.9) and the Levant (5.8) keep more of their shape, and Mesopotamia (2.1) and Persia and India (1.7) least of all, but only because there the map hardly tries: the far east gets a couple of sheets of loosely placed names.
Turn on "Modern coastline, bent to fit" in the scroll's Map layers to see this. I fitted a smooth surface through the 465 located places and used it to bend Natural Earth's coastline onto the scroll. The real Adriatic and Tyrrhenian coasts of Italy land almost exactly on the Tabula's grey-green seas.
Giving each part of the scroll to its nearest located place, Italy and its islands take 17.3% of the surviving parchment, though they are only 6.6% of the land of the empire's provinces: 2.6 times their share. Mesopotamia and Armenia get 2.3 times theirs, Greece and Thrace 1.6, the Levant 1.3. Gaul (0.89), North Africa (0.82) and Asia Minor (0.99) get about their fair share. Egypt gets far less (0.24), but most of its province is empty desert, and the Tabula, like the Romans, cared about the Nile valley.
Share of the scroll against share of the land
Share of the surviving scroll nearest to each region's located places, against the region's share of the land of the provinces of about AD 117 (DARE) and its share of the Tabula's stations. Hover a bar.
Britain and Spain barely appear: the sheet that held them is lost, leaving only a corner of Kent and a sliver of Gaul's Atlantic coast. OmnesViae rebuilds that missing end from the Antonine Itinerary, another Roman road book, adding 560 stations and 552 legs. I keep those separate throughout. Their distances behave just like the Tabula's (median straight line 0.97 of the written figure), as you would expect from a list of the same roads.
Rome sits enthroned in a gold circle, one of only three cities drawn as a person, with Constantinople and Antioch. 12 roads leave it. But on the surviving scroll it is not in the middle: it lies 35.4% of the way from the left edge, in Miller's segment V. The middle of the surviving scroll falls in the Balkans, near Skopje. Add back one lost sheet of average width at the western end, and Rome moves to 40.7%. It would sit at the exact centre only if about 3.3 sheets were lost; with 3, it lands at 49.0%. Some scholars, Richard Talbert among them, argue that the map was designed with Rome at its centre. If so, the measurements show how much is missing: about three sheets, not one.
Where Rome falls along the scroll
The eleven surviving sheets to scale, with Rome marked, and the sheets that would be needed at the western end to put Rome in the middle. Choose how many are lost.
Adding up every figure on the surviving sheets gives 51,359 units over 2,720 legs: 43,584 Roman miles, 4,598 Gallic leagues, and 3,177 units of unknown length in the far east. Converted, that is about 79,414 km of road, roughly 27% of the 299,171 km of Roman roads that Itiner-e has mapped. Another 251 legs on the surviving scroll have no figure at all, and the lost west would add about 17,313 km more.
Walking all of it, at the 20 to 25 km a day that Wikipedia gives for a traveller on foot, would take about 3,530 days, close to 10 years. A courier riding the imperial relay of horses, about 80 km a day, would need 993 days.
Days from Rome, by the Tabula's own figures
Shortest route over the Tabula's legs, adding up the written distances (a few legs without a figure are filled in, see methods). No sea crossings: the Tabula draws none. Choose a way of travelling.
The Tabula draws no sea routes, so its traveller goes the long way round. Rome to Carthage, a few days by ship, is 6,561 km by the map's roads, through the Balkans, Asia Minor, Syria and Egypt and along the whole of the African coast: 292 days on foot.
The regions, in numbers
Located places, share of the scroll and of the land, stretch, and kilometres of road written on the surviving scroll. Click a row to see the region on both maps.
The data. OmnesViae's export of the Tabula: 3,322 stations (one listed twice) and 3,523 legs, with the name as written on the map, Talbert's identification and coordinates. I used its coordinates as given. Checking them against the Pleiades gazetteer of ancient places, 94% of the 2,004 stations whose name matches a Pleiades place lie within 10 km of it. East of the Euphrates many identifications are uncertain.
Places on the scroll. There is no table of where each station sits on the Tabula, so I found 465 by eye on Miller's facsimile, reading the labels at full resolution and marking the town's picture or the start of its label: 433 read clearly and 32 with some doubt. They cover every sheet, at least 13 per sheet, but favour legible and well-known places.
Units and distances. A leg's figure is converted at 1.48 km per Roman mile, or 2.22 km per league in Aquitania, Lugdunensis, Belgica and the two Germanies. Straight-line distances are great circles between OmnesViae's coordinates. Road distances follow the Itiner-e network of Roman roads (from my earlier roman-roads study): each station is placed on the nearest road within 5 km, and the shortest route between the two is kept when it is no more than twice the straight line plus 10 km. That gave 1,558 legs. Itiner-e's roads are digitised with their bends, and the route between two stations can take a detour where the network has a gap, so the road figures lean long.
Shape and space. Stretch is the ratio of the two scale factors of a straight-line (affine) fit from ground to scroll, region by region; the fits are rough, since nothing on this map is straight. Share of the scroll gives each pixel of the painted strip to the region of its nearest located place, so it depends on where I located places. Land shares use the provinces of about AD 117 from the Digital Atlas of the Roman Empire, and are not defined beyond the frontier. Sizes in centimetres assume the surviving parchment is 6.75 m long, the length usually quoted.
Journeys. Shortest routes over all legs, weighted by the written distance; 249 legs with no figure are weighted by the road distance, the straight line times 1.13, or a typical leg of 23.7 km. Days use Wikipedia's speeds: 22.5 km a day on foot, 45 by cart, 80 by courier relay.