Same shape, different millennia

Ask anyone to picture a stone circle and they will probably see Stonehenge, or Castlerigg in the rain, or the Ring of Brodgar against an Orkney sky. Stone circles feel British and Irish. Archaeologists count about 1,300 of the Neolithic and Bronze Age kind in Britain, Ireland and Brittany, raised between roughly 3300 BC and 900 BC.

But the ring of upright stones is about the simplest monument there is. You need a field, some boulders and a centre. So I gathered every stone circle I could find in Europe's two big open datasets, OpenStreetMap and Wikidata, cleaned and merged them, and sorted them into the traditions that archaeologists recognise. Then I asked when each tradition was built, using published dates and radiocarbon dates linked to named circles.

The picture that comes back is not an island one. The oldest rings on the map are in the Alentejo in Portugal, begun around 6000 BC. The most numerous, by a wide margin, are the Iron Age domarringar, or judge's rings, of Sweden, built more than two thousand years after Britain's circle tradition began. And the latest are the Gothic circles of Pomerania, still being laid out in the second century AD. Same shape, different millennia.

01 · The mapEvery ring in Europe's open data

Each dot is one record: a stone circle, or a ship setting (an oval of stones in the outline of a boat, a close cousin found around the Baltic). Colours show the tradition. Switch traditions on and off, pick a country, or drag the year slider to see only the traditions that were being built in a given year. Tap a dot for its details and links.

02 · WhenA ribbon of dates

Each band below is one tradition. The pale band is the span its sources give; the strong band is the core period they single out. The dots are radiocarbon dates from the p3k14c database that I could link to a named circle in my map: one dot per circle, at the middle of its calibrated dates. Hover or tap any band or dot for the details and the source.

The ribbon splits into two blocks. The first is Atlantic and early: the Alentejo cromlechs from about 6000 BC, then the circles of Britain, Ireland and Brittany, from about 3300 BC to 900 BC, with most built between 3200 BC and 2000 BC. The second block is late and mostly northern: the little cremation rings of the Pyrenees from about 1200 BC, the ship settings of the Baltic from about 1000 BC, the Swedish and Finnish judge's rings from about 500 BC to AD 400, and the Gothic circles of Pomerania from the later first century AD.

The radiocarbon dates broadly agree with the published ranges, but they are thin outside Britain and Ireland. I linked 356 dates to 43 named circles. Of the 37 circles in Britain and Ireland with linked dates, 30 have a middle date inside 3300 BC to 900 BC. The others are mostly later activity at old sites: burials put into a circle, or a fire lit in it, long after the stones went up. The three dated Pyrenean circles fall between 890 BC and 804 BC, right where the excavators put them. Sweden's thousands of judge's rings have almost no dates in the database at all.

The circle is not one tradition. It is at least two clusters in time, separated by a gap of more than a thousand years in some places and by open sea in others.

03 · WhereMost of the records are Swedish

Counted country by country, Sweden has the most records by far: 2,172, or 48.7 for every 10,000 square kilometres of land, against 569 (23.3) in the United Kingdom and 96 (13.8) in Ireland.

That is partly real and partly the data. 95% of the Swedish records come only from Wikidata, where almost all of them are unlabelled items that look like a bulk import from the Swedish heritage register, one item per ring. Britain and Ireland have no such import, so their circles are in the data only where volunteers have mapped them. My merged map holds 683 circles for Britain, Ireland and Brittany against the roughly 1,300 that archaeologists list, so about 53%. The Scandinavian lead is real in kind, since judge's rings are genuinely common, but the size of the lead is partly an accident of who digitised what.

04 · TraditionsSame ring, different habits

The traditions do not just differ in date. They sit in different places and different company. The table compares them on things the data can measure: how close they are to the sea, and how often a circle has another within 100 metres (a sign of cemeteries of many small rings rather than single big monuments).

The Pyrenean rings are the clearest case of a different habit. 84% have another ring within a kilometre, and the typical gap to the next one is 200 m: they come in groups along the high pasture ridges, and almost none (1%) is within 5 km of the sea. Ship settings are the opposite: 63% are within 5 km of the coast, as you might expect of monuments shaped like boats. The circles of Britain and Ireland are more often single monuments: only 13% have another circle within 100 m. In Pomerania, 75% do, because the Gothic circles sit inside cemeteries.

Where OpenStreetMap mappers have tagged a period, they mostly agree with the published dates: 46 of 47 tagged Pyrenean rings say Iron Age, and all 21 tagged ship settings say Bronze Age.

05 · Famous ringsTwelve circles, side by side

Each tile is a 12 by 12 km window around a well-known circle, drawn from my data: every other record nearby is a dot. Some famous rings stand almost alone; others sit in crowds. Tap a tile to fly to it on the map.

06 · The questionAn island invention, or a shape people keep reinventing?

If the circle spread from one place, you would expect neighbouring traditions to touch in both space and time. So for each pair of traditions I measured the closest distance between any two of their records, and the gap (or overlap) between their published date ranges.

Three pairs touch. Britain and Brittany are 116 km apart across the Channel and share the same dates. The Swedish judge's rings and the Pomeranian circles are 186 km apart across the Baltic and overlap by 325 years; Wikipedia's article on Iron Age stone circles notes that the Pomeranian practice "suggests Norse influence". Ship settings and judge's rings sit in the same fields.

Other pairs do not touch. The nearest British circle to any Scandinavian one is Hjaltadans in Shetland, 411 km of open sea from Domsteinane på Sola in Norway, and the published ranges leave a gap of about 400 years between the end of one and the start of the other. The Pyrenean rings begin while Britain's tradition is ending, but the closest records are 796 km apart. And the Alentejo cromlechs end, on their sources' dating, some 1,800 years before the Pyrenean rings begin, though they are only 178 km apart.

So my modest answer is: probably not one invention. The map is consistent with at least two separate beginnings, an Atlantic one in the Neolithic and a northern one in the Iron Age, with the Pyrenees as a possible third. Data like these cannot prove that people reinvented the ring; ideas can travel without leaving circles along the way, and many circles are missing from the data. But the simplest shape in monumental architecture does seem to have been found more than once.

07 · The groundDo British and Irish circles seek a view?

For Britain and Ireland I have a terrain model, so I could test one old idea: that circles were placed for their views. For each of the 669 circles I compared the spot with ten random land points 1 to 5 km away.

The circles do sit a little higher than their surroundings. 65% of circles stand higher, relative to the ground within a kilometre, than the typical comparison point, and 28% are in the top quarter of their local ground, against 14% of comparison points. Their views are slightly wider: 60% see more of the ground within 5 km than their comparison points do. But the effect is gentle. The typical circle sees 23% of the ground around it against 21% for random spots nearby. Circle builders liked a shoulder or a low rise, not the summit.

08 · LimitsWhat this map cannot tell you

Coverage is uneven. Sweden's records come almost all from Wikidata, apparently a bulk import of the national heritage register; Britain's and Ireland's come mostly from volunteer mapping and hold about half the circles archaeologists list. Poland's Gothic circles are mapped one by one at some sites and not at all at others: Odry has twelve circles, ten of them whole, but only four are mapped. Counts across countries are counts of records, not of monuments.

One date per tradition, not per circle. Almost no record carries its own date. I give each tradition the range its sources give, which hides the variety inside it. The "Elsewhere" group (90 records, from Galicia and the Causses to Sardinia, Germany and Bulgaria) gets no date at all.

Radiocarbon dates are not building dates. A date at a circle may come from a burial, a hearth or a post-hole, from before or long after the stones were raised. The p3k14c coordinates are deliberately blurred, so I linked dates only by name and within 1 km, and I list every link for checking.

Classification is by place. I put a circle in a tradition by where it is and whether it is a ship setting. A Bronze Age circle in Sweden would be counted with the Iron Age judge's rings. Modern circles (millennium circles, Welsh Gorsedd stones, healing circles) are removed where their names or tags say so, but some may remain.