Portugal, Galicia, Brittany, Ireland, Wales, Orkney and Scandinavia built recognisably the same great stone tombs. I put 13,280 megalithic monuments on one map and tested, three honest ways, whether the idea travelled by sea.
Walk into a passage grave in Brittany, then one in the Boyne valley in Ireland, then one on Orkney, in northern Portugal or on a Danish island. The details differ, but the idea is the same: a long, low passage of upright slabs, roofed with stones that took dozens of people to move, leading into a chamber for the dead. Dolmens, gallery graves and standing stones repeat the trick from Andalusia to Sweden.
Why would farmers a thousand miles apart build the same thing? One old answer is that the idea travelled by boat, along the Atlantic coast. In 2019 Bettina Schulz Paulsson modelled more than 2,400 radiocarbon dates and argued in PNAS that the first megalithic graves appeared in north-west France and that the tradition spread from there, in a few phases, by sea.
I wanted to see what open data can say about that story. So I gathered every megalithic monument that volunteers have mapped in OpenStreetMap, 13,280 of them after cleaning, and set them three tests. Do megaliths sit closer to the sea than chance, and closer than other prehistoric burial mounds? Do the oldest radiocarbon dates from megalithic sites sit on the Atlantic coast? And is the Atlantic facade really where the great stone tombs are, type by type?
The short answer: the sea-route story survives all three tests, but none of them proves it, and one of them is much weaker than it first looks.
Every dot is a monument tagged as megalithic in OpenStreetMap: 4,352 dolmens, 3,064 menhirs and standing stones, 1,345 stone circles, 1,142 nuraghi and talaiots (the stone towers of Sardinia and the Balearics, kept as their own class), 557 passage graves, 506 chamber tombs and chambered cairns, 188 long barrows, 184 alignments and stone rows, 90 gallery graves, 59 Irish court, portal and wedge tombs, and 1,617 megaliths whose type nobody has tagged. Filter by type, country or period tag, or switch to the radiocarbon view and press play to watch the dated tombs appear.
Coverage is uneven. OpenStreetMap has 4,063 megaliths in France but only 306 in Denmark, a country famous for its dolmens, where mappers have tagged 25,669 burial mounds as plain tumuli instead. Gaps on the map are often gaps in tagging, not in prehistory.
Monuments © OpenStreetMap contributors (ODbL), extract of 27 October 2026; outlines are placed at their first vertex. Dated sites: p3k14c, linked by name keyword or by a shared name within 1 km, calibrated with IntCal20.
If the idea came by boat, you might expect the monuments to crowd the coast. They do, compared with random ground. Of all 13,280 monuments, 29% stand within 10 km of the sea. Ten random points on land for every monument, in the same countries, put only 13% that close. That makes megaliths 2.3 times as likely to be near the sea as the land they sit in, and the pattern holds in 10 of the 13 countries with at least 50 monuments.
But random land is a soft target. People in prehistory did not live at random, and coasts and river mouths are good places to farm, fish and travel whatever you build for your dead. So I used a second, tougher yardstick: the 46,596 barrows and tumuli in OpenStreetMap that carry no megalithic tag, reweighted so that each country counts as much as it does among the megaliths. Against those, the edge shrinks: 29% of megaliths against 23% of barrows, a ratio of 1.3. Because monuments cluster, I resampled 50 km grid cells rather than single sites to get an honest range for the megalith figure, and it runs from 22% to 37%, wide enough to include the barrows.
How far from the sea?
Share of each group in each distance band. Barrows and random points are reweighted to the megaliths' mix of countries.
The detail matters. Tombs are the least coastal group: 20% within 10 km, against 16% of barrows. Menhirs, stone rows and stone circles are the most striking: 35% within 10 km, 1.7 times the barrow figure. The Sardinian and Balearic towers are coastal too, but less so than the tumuli mapped in Italy.
Within 10 km of the sea, by group
Dot: monuments, with a 95% interval from resampling 50 km cells. Diamond: barrows and tumuli, same country mix. Square: random land, same countries.
Country by country
Countries with at least 50 megalithic monuments in OpenStreetMap. Barrow figures only where there are at least 30 barrows.
Portugal is the clearest exception: only 8% of its mapped megaliths are within 10 km of the sea, below the 15% of random land, because the great dolmen fields of the Alentejo and the north lie inland.
Distance to the sea cannot tell us which way an idea moved. Dates can. I took the p3k14c database of archaeological radiocarbon dates, 77,532 dates in the map's frame, dropped 1,571 shell dates, and linked sites to megaliths by two documented rules: the site's name contains a megalithic word (dolmen, anta, mamoa, hunebed, long barrow, nuraghe and so on), or it shares a distinctive name with a mapped megalith within 1 km. That gave 254 monuments. A check of 50 links picked at random found 42 that point at the same monument or the same named group, seven that are unclear, usually a village name shared with a tomb nearby, and one that is wrong.
I calibrated every date with the IntCal20 curve and took each monument's earliest date between 5500 and 500 BC. One early date proves little: it can come from charcoal that was old when the stones went up, or from activity before the monument. So for each region I used a steadier marker, the date by which the earliest tenth of its dated tombs had begun, with a bootstrap interval, and only for regions with at least eight dated tombs.
When do the tombs begin?
The 10th percentile of each dated tomb's earliest calibrated date, by region (number of dated tombs in brackets). Lines are 95% bootstrap intervals.
North-west France comes first, at about 4885 BC from 13 dated tombs, which fits Schulz Paulsson's result. Ireland follows at about 4667 BC, then England and Man, Portugal at about 4190 BC, the Low Countries and north Germany at about 3896 BC, the rest of France at about 3764 BC and southern and eastern Spain last, at about 3290 BC. Taken together, the 110 dated tombs in Atlantic-facing regions start earlier than the 33 in Mediterranean and inland ones: median 3465 BC against 2590 BC.
Now the caution. The intervals are wide and overlap: north-west France could plausibly begin anywhere between about 5145 and 4382 BC on this evidence, and Ireland between about 5445 and 3554 BC. And within regions, tombs nearer the coast are not earlier than tombs inland: the rank correlation between a tomb's start and its distance to the sea is 0.00 across 118 linked tombs, which is nothing. If boats carried the idea, they carried it between regions, not visibly from the beach inland.
The earliest single dates
The twelve earliest monument start dates in the linked set. Treat single dates with care: several are known to be debated or may predate the stones.
The third test asks where each kind of monument lives. I split Europe into three zones: within 50 km of the Atlantic and northern seas (including the Channel, the North Sea and the Baltic), within 50 km of the Mediterranean and Black Sea, and inland. Random land in the same countries puts 27% of points in the Atlantic zone; 47% of megalithic monuments are there.
Atlantic, Mediterranean or inland?
Share of each type within 50 km of each sea, or more than 50 km from any coast. Baselines at the bottom.
Some types are emphatically Atlantic: 90% of long barrows, 85% of Irish court, portal and wedge tombs, 79% of stone rows and alignments and 73% of stone circles. But dolmens, the commonest tomb, lie about half inland (49%), and tombs as a whole are in the Atlantic zone no more often than ordinary barrows in the same countries (41% against 42%). The tombs' Atlantic flavour, in other words, is largely the Atlantic flavour of where Neolithic and Bronze Age people buried their dead at all.
Each test leans the same way. Megaliths are more coastal than the land around them; the earliest dated tombs cluster in north-west France and the Atlantic regions start earlier than the rest; and the most distinctive types, stone rows, circles, long barrows and the Irish court tombs, belong to the Atlantic fringe. That is what a sea-borne idea would leave behind.
But it is also what you would see if Atlantic Europe simply had more good stone, more people near the coast and more archaeologists with radiocarbon budgets. Megalithic tombs are only modestly more coastal than other burial mounds, and the regional dates rest on a few dozen tombs per region at best. The data fit the highway. They do not prove it was used.
Monuments. OpenStreetMap objects tagged as megaliths (archaeological_site=megalith, megalith_type, or historic=menhir, standing_stone, stone_circle or dolmen), plus stone circles found by a multilingual name search and checked against their tags and names. Points within 15 m of another of the same type were merged, which removed 2,809 duplicates, mostly a node and an outline for one monument or the separate stones of one row. Outlines are placed at their first vertex. Only countries Natural Earth assigns to Europe are kept. Period labels come from OpenStreetMap tags and, where an object links to Wikidata, from Wikidata; 9,136 monuments have none.
Coast. Distance to the Natural Earth 1:10m coastline, in an equal-area projection. Random points: ten per monument, uniform on land within the same country. Barrows: OpenStreetMap tumuli without a megalithic tag. Intervals resample 50 km grid cells.
Dates. p3k14c (uncalibrated), marine shell removed, calibrated one by one with IntCal20 on a 5 year grid; I use the median. Links: a megalithic keyword in the site name, or a shared distinctive name with an OSM megalith within 1 km. Adding sites that only lie within 150 m of a megalith (a looser rule) keeps north-west France first. Regions are drawn with simple boxes and country borders, set out in the code.
Coverage is uneven. OpenStreetMap reflects where volunteers have mapped and how they tag. France has 4,063 megaliths here, Denmark only 306 and Sweden 265; Denmark instead has 25,669 mounds tagged as plain tumuli, some of which will be megalithic tombs, and those sit in the barrow baseline. That is why every comparison is made within countries. Types are tagged inconsistently: one mapper's dolmen is another's chamber tomb, and 1,617 monuments have no type at all.
Radiocarbon coverage is uneven too. Britain and Ireland are dated far more densely than Iberia or Scandinavia, and Sardinia, Italy and south-east Europe have almost no dated tombs in the linked set. A region with few dates will look later than it was simply because its earliest tombs have not been dated. Earliest dates can come from material older than the monument; a 10th percentile softens this but does not remove it.
Coasts move. Sea level and shorelines have changed since the Neolithic, especially around the North Sea and the Baltic, so present-day distances are only a guide.
What this cannot show. Distance and dates are consistent with a maritime spread but cannot separate it from other explanations, such as people independently building in stone where good stone lay near the coast. This is a test of open data, not a substitute for the detailed regional chronologies archaeologists build site by site.
Two companion studies use the same OpenStreetMap and Wikidata pull: Stone Circles and Barrows of Europe.
Sources: OpenStreetMap contributors, extract via the QLever OSM planet endpoint, 27 October 2026 (ODbL 1.0). Bird, D. et al. 2022. p3k14c, a synthetic global database of archaeological radiocarbon dates. Scientific Data 9, 27. Reimer, P. et al. 2020. The IntCal20 Northern Hemisphere radiocarbon age calibration curve (0 to 55 cal kBP). Radiocarbon 62. Schulz Paulsson, B. 2019. Radiocarbon dates and Bayesian modeling support maritime diffusion model for megaliths in Europe. PNAS 116 (9), 3460 to 3465. Wikidata (CC0). Natural Earth (public domain).