Ley Lines, Tested: Are Ancient Sites in Britain Lined Up More Than Chance Allows?

Alfred Watkins argued in 1925 that old sites in the British landscape fall into straight lines, or leys. Statisticians have long replied that a dense scatter of points will always contain alignments. I revisit the question with open data: 50,822 churches (of all dates), barrows and cairns, megaliths, stone circles, hillforts and holy wells in Great Britain from Wikidata, OpenStreetMap and the Atlas of Hillforts. In four landscapes that ley hunters made famous I count every set of three to six places that lie within a 100 m corridor between two end places 1 to 20 km apart, and compare each count with 300 random versions. Real sites form 1.5 to 5.4 times as many lines of four as points scattered evenly over the same land, but when the random points keep the real clustering (each place moved up to 1 km), the real counts are 0.94 to 1.14 times the random average and within the 95% range in all four landscapes. The same holds for prehistoric sites alone. A simple formula shows why: expected lines of $k$ points grow with about the kth power of density. The Saint Michael's line, drawn from St Michael's Mount through Glastonbury Tor, passes 393 sites within 1 km against 344 for moved copies of the line (the 74th percentile); prehistoric sites and churches dedicated to St Michael come near, but not past, the 95th percentile. Alignments are what the density and clustering of Britain's old sites predict. Interactive version: https://tristan-hotham.github.io/ley-lines/. Data and code: https://github.com/Tristan-Hotham/ley-lines. Blog post: https://blog.tristanhotham.com/2026/10/27/ley-lines/.