Devil's Dyke, Devil's Punch Bowl, Devil's Arrows, Devil's Bridge, Devil's Kitchen: Britain's map is full of the Devil's property. I found 107 Devil names on the Ordnance Survey's open gazetteer and rivers, and 23 named Devil monuments in national records, and asked what he owns, where, and whether the saints own something different.
The Devil is the busiest landowner in British folklore. He dug Devil's Dyke above Brighton to let the sea drown the Christians of the Weald, threw the Devil's Arrows at Aldborough, leapt across the Devil's Jumps and left the Devil's Punch Bowl behind him. In the classic Devil's Bridge tale he builds a bridge in return for the first soul to cross it. And he owns a remarkable amount of furniture: a chair, a kitchen, a dining room, a dressing room, a frying pan, a pair of bellows.
It is often said that the Devil gets the strange old things in the landscape: the prehistoric stones and earthworks nobody could explain, and the steep, wild, unsettling ground, while the saints get the holy wells and the churches. That is a testable claim. So I asked the map: where does the Devil live in Britain, and does he really get the old stones while the saints get the springs?
I collected every feature on the Ordnance Survey's open gazetteer (OS Open Names) and its open river network (OS Open Rivers) with Devil, the Scots Deil or a Gaelic word for the devil in its name: 107 in all. I compared each with fifty ordinary features of the same kind in the same 50 km square, and did the same for 393 natural features named after a saint, such as St Peter's Stone or St Catherine's Hill.
The short answer: yes, roughly. Devil names sit near recorded prehistoric monuments much more often than matched places do, and the saints' names do not. The saints instead sit by churches and holy wells, which the Devil's names do not. The Devil's ground is a little steeper than the places it is compared with, but no wilder; and he shows no liking for county boundaries. The names, though, are mostly not old. Where a date can be found, the Devil arrives in the record in the seventeenth and eighteenth centuries, thousands of years after the stones went up.
Small marks are the gazetteer names, coloured by what the Devil owns. Large diamonds are named Devil monuments in national records: barrows, stones, a stone circle, a hillfort, bridges. Open rings in Ireland come from OpenStreetMap. Saint-named features, prehistoric monuments and holy wells can be switched on. Tap any mark for its details and distances.
The gazetteer Devils are spread across Britain: 63 in England, 37 in Scotland and seven in Wales. In Scotland 16 are the Scots Deil (Deil's Caldron, Deil's Kettle, The Deil's Heid) and two are Gaelic: Bod an Deamhain, which the OS also names The Devil's Point, and Druim an Diabhail. Relative to the number of named features, the Devil is commonest in the East of England (75 per 100,000 features) and the South East (48), and rarest in Yorkshire and the Humber (one) and the North East (none).
Read the second half of each name and a pattern appears. The commonest single thing is a den (10 of them), then woods and copses. But the largest group of names is domestic. 24 of the 107 names (22%) give the Devil a household or a trade: a kitchen, a chair, a dining room, a dressing room, a study, a frying pan, a kettle, a caldron, an oven, a pair of bellows, a mill, a limekiln, a barn, a peat stack, a threshing floor, a bowling green and a punch bowl.
The household names are natural features, not buildings. Twelve of the 24 are coastal rocks, coves and beaches (the Devil's Kitchen, Frying Pan and Bellows in Devon, Cornwall and Dorset; the Devil's Dining Room and Peatstack in Aberdeenshire); most of the rest are waterfalls, pools and hollows inland (Deil's Caldron, Devil's Punch Bowl at Devil's Bridge). They are on steep ground: their median slope is 5.7 degrees, against 3.4 for the matched places. The Devil's body turns up too, mostly as bends and drops: 12 names give him an elbow, teeth, a head, a mouth, footsteps or a slide.
Compared with everything else on the gazetteer, valleys and hollows make up 2.4 times as large a share of Devil names as of all named features, and coastal features 2.1 times; hills only 0.4 times. Saints are common on the coast too (2.0 times), where islands, bays and headlands are named for them, but scarce in valleys and hollows (0.6 times). The Devil lives in the hole; the saint stands on the island.
Now the main test. For each Devil name I measured the distance to the nearest recorded prehistoric monument: 12,387 barrows and burial cairns, 1,194 standing stones, stone circles and chambered tombs, and 3,610 hillforts. 31% of the Devil names lie within 1 km of one. Of the matched places (woods with woods, coves with coves, from the same 50 km square), 19% do, and of random land 13%. In 2,000 replicate draws of matched places, a gap this large turned up with chance below 0.001.
The saints' names show nothing of the kind: 18% lie within 1 km of a prehistoric monument, against 18% of their matched places. The Devil's excess comes from barrows (13% within 1 km, against 8%) and hillforts (28% within 2 km, against 22%) together; neither alone is beyond what luck can give with 107 names.
The Devil sits next to the barrows and the hillforts; the saints sit next to the church and the well.
Some of the closest cases are telling. In Deils Wood in East Renfrewshire a cairn, recorded as "Deils Wood, cairn 350m E of Bonnyton", lies 4 m from the gazetteer point. Devil's Tor on Dartmoor is 36 m from Beardown Man, a standing stone. On the Aberdeenshire coast the Devil's Dining Room, the Deil's Kettle and the Devil's Peatstack all sit beside coastal forts in the Atlas of Hillforts (Cullykhan, Dunnicaer and Cleaved Head). And Devil's Dyke above Brighton is a valley right below a hillfort, 541 m away.
The pattern is strongest in the north and on the coast. 39% of the 41 northern Devil names (North East, North West, Yorkshire and the Humber, Scotland) lie within 1 km of a monument, against 21% of matched places (chance 0.011); for the Scots Deil names it is 44% against 18%. Coastal Devils lean on promontory forts (48% against 30%). Inland the excess is smaller but still there (24% against 15%, chance 0.026), and in the south the barrows carry it: 20% within 1 km of a barrow, against 11%.
The saint-named features are the Devil's mirror image. Of the 310 that do not simply borrow the name of a nearby St Something village, 47% lie within 1 km of a church, against 31% of matched places (chance below 0.001), and 13% within 2 km of a recorded holy well, against 4% (chance below 0.001). The Devil's names are no closer to churches than ordinary places (34% against 31%) and not significantly closer to holy wells (6% against 3%).
So the old division holds up surprisingly well on the modern map. The Church's names cluster round the Church's places; the Devil's names cluster round the monuments that were already ancient when the first churches were built. At Devil's Dyke the two meet in a single story: the Devil dug his trench to drown the Christians of the Weald, and St Cuthman of Steyning beat him with a candle and a crowing cock.
Is the Devil's ground odd ground? I measured the slope at each place and the relief (the highest minus the lowest ground within 1 km) from a terrain model with cells of about 100 m. Devil names are on steeper ground than random land and a little steeper than matched places: median slope 2.9 degrees, against 2.2 for matched places (chance 0.022) and 1.8 for random land. But their relief is no greater than that of matched places (65 m against 68 m): most of the drama comes with the kind of feature, cliffs, coves and gorges, rather than from the Devil choosing the wildest examples.
The most dramatic Devils are in Scotland: the Devil's Punchbowl (Coire na Ciche) on Arran has 634 m of relief within 1 km, and The Devil's Point in the Cairngorms 459 m. Saint names, by contrast, sit on gentler, lower ground than their matched places (median relief 54 m against 60 m, chance 0.051).
Devils are sometimes said to haunt boundaries, the no-man's-land between parishes. I could not test parish boundaries, but I could test county ones (ceremonial counties in England, preserved counties in Wales, council areas in Scotland). There is nothing there: 11% of Devil names lie within 2 km of a county boundary, against 13% of matched places (chance 0.607).
Most of the famous Devils are not in the gazetteer at all. OS Open Names leaves out antiquities, so the Devil's Arrows, the Devil's Den and the Devil's Quoits are missing from it. I found 23 named Devil monuments and landmarks in other national open records: five barrows or cairns, three chambered tombs, three sets of standing stones, a stone circle, a hillfort, a linear earthwork, seven bridges and two natural features with Wikipedia articles. Thirteen of the 23 are prehistoric monuments. Only three have a Devil name in the gazetteer within 1.5 km.
This is the part to be careful about. A Devil name records a story, not a date. Where Wikipedia gives a date for the name, it is late: John Aubrey mentions the Devil's Arrows in 1687 (the stones are Stone Age), the English name Devil's Bridge is first recorded in 1734, and the Devil's Jumps first appear on John Rocque's map of 1765. The great Devil's Dyke near Newmarket was simply "the ditch" or the "great ditch" in medieval documents. For most of the smaller names I have no date at all, and they could be just as recent. What the map shows is where people, at some point, decided that only the Devil could have made this; not when.
The Republic of Ireland has its own Devils. OpenStreetMap has 15 named Devil places there once shops, car parks and duplicates are set aside: the Devil's Bit Mountain in Tipperary, the Devil's Glen in Wicklow, the Devil's Punchbowl on Mangerton, the Devil's Ladder on Carrauntoohil, and Devilsmother. They are on the map but not in the tests, because OpenStreetMap names are not a like-for-like gazetteer with OS Open Names.
I searched NAME1 and NAME2 of all 267,744 OS Open Names features (settlements, landforms, landcover and water) and the names of all named watercourses in OS Open Rivers for: Devil as a word, with or without a possessive or plural (so Kill Devil Copse counts, but Mandeville and Devilla Forest do not); Devils- fused to a following word (Devilsden, Devilsmouth, Devilspit, Devilsrest); the Scots Deil, Deil's or Deils as a word (not Welsh Deildre or Deilo); and the Gaelic Deamhain, Diabhail or Dhiabhail. This gives 101 gazetteer features and six watercourses. I set aside 11 near misses (six Mandevilles, Devilla Forest, Demoniac Plantation, two Deildres and Gallt Deilo). I did not search for Hell, Satan, Old Nick or the like, whose meaning is less certain. Each name's head word, the word after Devil, is sorted into a class by the image it uses; the classes are about form, not origin.
Saint names are OS Open Names features that are not settlements and have St, St. or Saint followed by a capitalised word in their name (393 features). 83 of them share their saint with a settlement within 5 km (St Agnes Wood near St Agnes, for example) and are likely to borrow a parish name, so the main comparison uses the other 310.
For every Devil and saint name I drew 50 ordinary OS names of the same local type from the same 50 km square (100 km or the whole country when fewer than 20 were available), leaving out all Devil and saint names, and 50 random points of land in the same square. Watercourses were matched with other named watercourses. Each test compares the real names with 2,000 replicate sets, each taking one control at random per name; chance values are two-sided. Distances are straight lines to: 12,387 barrows, burial cairns and mounds (OpenStreetMap and Wikidata, from Is the Barrow a European Thing?); 1,194 megaliths and stone circles (OpenStreetMap and Wikidata, from The Megalith Highway and Same Shape, Different Millennia); 3,610 hillforts (Atlas of Hillforts of Britain and Ireland, as used in Within Sight); 36,435 church buildings in Wikidata (from Whose Church Is It?); and 367 holy wells in Great Britain (from From Saint to Wish and Whose Church Is It?). Slope and relief come from the AWS Terrain Tiles elevation model at zoom 10. County boundaries are the ONS 2025 boundaries, as built for How Roman Is Your Road?
The names are few: 107 in all, and only a dozen or two in each class, so the class results are suggestive at best. Names record stories, not dates; many Devil names may be recent, and a name beside a barrow may have been given by someone who knew it was a barrow. OS Open Names leaves out antiquities and many small features, so the famous monument Devils are counted separately and not tested. The monument records are much fuller in some areas than others (holy wells especially are thin in Britain, at 367), and many barrows are lost. Coastal promontory forts drive much of the coastal result. Modern county boundaries are not parish or ancient boundaries. I ran 15 measures on many groups, so single chance values near 0.05 should be read with caution; the two results I would put weight on are the Devil's excess near prehistoric monuments and the saints' excess near churches and holy wells.