In 1921, Alfred Watkins looked at a map of Herefordshire and saw it: standing stones,
burial mounds, moats, old churches, crossroads — strung along dead-straight lines he
named leys. By 1969 the lines had been reborn as channels of earth energy.
It is the perfect statistics test case, because deliberate ancient alignments are
actually real, parts of the siting history are
genuinely open, and the planetary energy grid is an
overclaim — and the whole dispute turns on a number you can
compute: how many straight lines should pure chance draw?
Watkins even left us a killable prediction — his rule that four points make a ley.
Let's run it.
Fig. 1 — Random "sacred sites" on a 100 km map sheet. Every 4-point (cyan) and 5-point (magenta) alignment within the corridor is found by exhaustive search, not by eye. Right: alignments found on this scatter vs. the number chance predicts.
Take the honest core first. Ancient builders really did align things on purpose —
Stonehenge's axis points at the solstice sunrise, Egyptian pyramids square to true
north within arcminutes, Roman roads run straight for days. The question was never
whether straight lines exist. It is whether the alignments Watkins found among
unrelated sites need any explanation at all.
So state the null hypothesis and make it compute. Scatter $n$ sites at random on an
$L\times L$ map sheet and call $k$ of them "aligned" when they fit inside a straight
corridor of width $w$. For a random pair a mean distance $\bar d$ apart, each extra
site lands in that corridor with probability $\approx \bar d\,w/L^2$ — so alignments
of every size arrive on schedule:
Drop in numbers a 1920s ley hunter would recognize — 150 candidate sites on a
100 km Ordnance Survey sheet, a 200 m corridor — and a purely random map
carries about 1,300 three-point alignments, roughly 120 four-point leys, and a
handful of five-pointers. Watkins's four-point threshold is met a hundred
times over before archaeology enters the room. (The formula slightly overcounts,
since every 4-point ley contains four 3-point ones; the ▶ button below runs the exact
null — 100 fresh random maps through the same detector.)
This is not a hypothetical. In 2010, mathematician Matt Parker ran the exercise on the
807 former sites of the Woolworths supermarket chain and found "uncanny" triangles and
alignments of the kind attributed to ancient wisdom. Anything with enough dots aligns:
supermarkets, pizzerias, phone boxes — and 150 random points, below.
The same claim, read by parallax — from the measured alignments to the energy grid.
actually real
Ancient builders aligned things deliberately.
Signal. Stonehenge's solstitial axis; Newgrange's roof-box catching the midwinter sunrise; the Giza pyramids squared to true north within a twentieth of a degree; Roman roads surveyed straight across whole counties. Individual, purposeful alignments are documented archaeology.
Kill test. Does the alignment encode astronomy or geometry beyond chance? Predict the azimuth, then measure it. At the famous monuments the prediction survives — that's why they're famous.
genuinely open
How much old siting hides under the new?
Signal. Continuity of sacred sites is real: Pope Gregory's letter of 601 AD ordered pagan shrines rededicated, not razed, and some English churches demonstrably sit on prehistoric earthworks — at Rudston a Bronze Age monolith still stands in the churchyard.
Unresolved. How systematic that reuse was — which medieval churches sit on older sacred ground, and whether any pre-Roman long-distance track lines survive in later road and parish geometry — is a live question in landscape archaeology.
Kill test. Excavation and dating beneath candidate sites, case by case. Slow, diggable, and genuinely undecided in the aggregate.
real kernel · overreached
"Leys are prehistoric straight trader tracks."
Signal. Watkins was a skilled amateur and some of his alignments are real to the eye; old straight stretches of track do exist. The Old Straight Track (1925) is careful, sincere fieldwork.
Noise. His site classes — churches, moats, mounds, crossroads, "old stones" — put thousands of candidate points on every map sheet. At that density, the simulator above manufactures his leys from noise; and a trade route that ignores bogs, rivers and cliff faces to stay straight is a strange way to carry goods.
Kill test. Do real site sets beat density-matched random scatters? Williamson & Bellamy ran it (1983), statisticians formalized it (Broadbent 1980; Kendall & Kendall 1980): no excess over chance.
Signal. The mood is honest — standing inside Avebury's circle at dusk feels charged, and John Michell's 1969 rebranding of Watkins ran on that feeling. The feeling is real; feelings are data about people.
Noise. The "energy" has no measurable field, no units, no dose, and no meter that finds it; it is detected only by dowsing, which fails blinded tests. A current that no instrument can read forbids no observation.
Kill test. Double-blind dowsing trials — most famously the Munich experiments (1987–88) — come out at chance level. The claim survives only by never specifying what would count as a miss.
Notice the sift: the real layer survives a measured azimuth; the open layer names
questions a trowel can settle; the overclaims die by counting. Watkins was right
that the map is full of lines — he was wrong about who drew them. Chance is a prolific
draughtsman, and it works cheap.