Is the Bulford Post Alignment to the Midwinter Sunset?

Look along the posts: https://timdaw37.github.io/bulford-posts-3d/

Set Midwinter, Sunset, Full orb. Then Midsummer, Sunrise, First gleam. The question is which of those two pictures you trust.

Midwinter Sunset

Midsummer Sunrise

Stonehenge has the same argument. The axis works both ways, and for a generation the tourist photograph has been midsummer sunrise up the Avenue. Parker Pearson and others have put the weight on the other end: you walk in from the north-east and watch midwinter sunset drop through the Great Trilithon. Ruggles has been careful to say the architecture is precise enough to pick the solstice in space, and that first or last gleam is the closer convention, not that one festival owns the monument.

Harding, Leivers and Silva, in PAST 113, publish the Bulford posts as a solstitial pair — midsummer sunrise and midwinter sunset, about 120 m apart, around 2950 BC. Harding found the two structural pits first and only then asked what they pointed at. That is the right order. The question here is only which way the line is better.

The 3D model is the published plan, not the Wessex GNSS. From that plan the posts run 48.31° / 228.31° true. On Bulford’s own skyline, 2950 BC, the six like-for-like events are:

Event	Convention	Azimuth	Miss from the posts
Midwinter sunset	Full orb	228.55°	+0.23°
Midsummer sunrise	First gleam	48.90°	+0.58°
Midwinter sunset	Half orb	229.03°	+0.72°
Midsummer sunrise	Half orb	49.29°	+0.97°
Midwinter sunset	Last gleam	229.53°	+1.22°
Midsummer sunrise	Full orb	49.72°	+1.41°

Full orb at midwinter sunset is the closest of the six — a quarter of a degree, inside a solar radius. First gleam at midsummer is next, half a degree, one disc. The rest are worse.

Two warnings, which do not cancel the table.

You must not mix conventions. First gleam at one end and full orb at the other are different ways of watching. If the same limb is used both ways, winter last gleam is worse than summer first gleam. Reciprocity was never going to give a second bullseye on one straight line: refraction and the solar radius push rise and set apart, and the south-west horizon is not high enough to cancel that.

The picture also flatters the winter disc. The south-west skyline is a local rise at about 880 m, so the sun sitting on it looks planted. The north-east gleam is 5.7 km out. Your eye and the azimuth are not quite the same measurement.

Ruggles’ published rule of thumb is that claimed sightlines of half a degree and better in the old lists were the ones most easily explained as selection, that British prehistoric orientations show no evidence of precision much greater than about 1°, and that a couple of degrees is still a fair orientation in a group of sites (Ruggles 1984, 304–306; 1997, 207). On that scale the midwinter full-orb miss is allowed, the midsummer first-gleam miss is allowed, and neither is a prosecution. A single pair in a crowded pit field still has to beat coincidence.

We do not have the GNSS. Two independent digitising jobs of the PAST plan agree to 0.2 m, which is inside the pit and not the solar miss. The miss may move when the survey comes out. Until then, the model says: if you have to pick one, pick the sun standing on the south-west ridge at midwinter.

Have a go with the table in one hand and the model in the other.

Harding, P., Leivers, M. and Silva, F. 2026. A newly discovered solstitial post alignment in the Stonehenge landscape at Bulford. PAST 113, 2–5.

Ruggles, C.L.N. 1984. Megalithic Astronomy. BAR British Series 123. Oxford.

Ruggles, C.L.N. 1997. Astronomy and Stonehenge. Proceedings of the British Academy 92, 203–229.

3D plan: https://timdaw37.github.io/bulford-posts-3d/

The Bulford posts in 3D

Open this to understand this post: https://timdaw37.github.io/bulford-posts-3d/

Drag to rotate, right-drag to pan, scroll to zoom. Click a pit for its name and coordinates. This is a data heavy page so it won't work properly on mobiles and older machines, a very light version will be served instead.

Figure: live viewer, along the posts, midsummer, first gleam, 2950 BC.

In PAST 113, Phil Harding, Matt Leivers and Fabio Silva publish two Late Neolithic post-pits at Bulford, about 120 m apart, and argue that they mark the midsummer sunrise and midwinter sunset of around 2950 BC. Harding’s excavation is the reason we have the pits at all. Out of dozens of ordinary scrapes on that hilltop he picked the two that were actually built as structural post-holes — a metre across, thirty inches deep, ash in the fills — and only then asked what they pointed at. That is the right order. The PAST paper is a proper account of a real discovery, and Silva’s insistence that a claimed solstice has to be tested against the local skyline, not a round number from Stonehenge, is the right question.

In June, before PAST was out, I put down some sceptical notes on the news reports: a crowded pit field, a topographic grain that favours certain orientations, and not enough published geometry to tell coincidence from design. The article now gives the plan, the azimuths they used, and the argument. The next thing I wanted was to stand on that plan and look.

So here is a 3D model of the published figure. All 39 pits from the Wessex plan are marked. The two alignment posts stand in their holes, 8647 at 2 m and 9019 at 4 m, as the paper draws them. The ground is OS Terrain 50 with a 1 m LiDAR inner patch; you can switch to map or aerial. The sun sits on Bulford’s own skyline, with first gleam, half orb and full orb as alternatives, at 2950 BC, 2500 BC or today.

Figure: the Terrain 50 landscape with the post-line.

From the plan the posts run 48.12° grid, 48.31° true. On this horizon, midsummer first gleam in 2950 BC comes up at about 48.88°; midwinter last gleam goes down at about 229.5°. Neither sits on the timbers. First gleam is the closer of the two, about half a degree off. Sunset is further. We do not have the Wessex GNSS coordinates of the pit bases — they have not been released — and every position in the model is digitised from the published 1:1000 plan. Two independent readings of 8647 and 9019 agree to 0.2 m, which is good enough to look and not good enough to prosecute a fraction of a degree. The miss may be in the plan, in the skyline, or in my arithmetic. I would rather put the model out than pretend otherwise.  Ruggles’ working rule is that half a degree is as tight as a two-post solar line can be, about a degree is a fair prehistoric axis, and two degrees is the chance envelope

Have a go. Start with Along posts, Midsummer, 2950 BC, first gleam. Then Midwinter. If the sun and the posts do not agree on your screen either, we are looking at the same thing.

Harding, P., Leivers, M. and Silva, F. 2026. A newly discovered solstitial post alignment in the Stonehenge landscape at Bulford. PAST 113, 2–5.

Earlier notes: Bulford: pits, double henges and a claimed solstice alignment. Code: github.com/TimDaw37/bulford-posts-3d.

Terrain: OS Terrain 50 and EA 1 m LiDAR © Crown copyright (Open Government Licence). Map: © OpenStreetMap contributors. Aerial: ESRI World Imagery. Drafting with Grok; the errors are mine.

The 52 Hillforts of Wiltshire Map

My 2022 list of the 52 hillforts of Wiltshire now has a map. The 52 Hillforts of Wiltshire is a standalone interactive gazetteer: the same 52 sites, numbered north to south, with Oxford Atlas notes and 1 m LiDAR.

Each entry carries its Atlas classification — univallate, bivallate, multivallate; contour, promontory, hilltop — graded by how far the sources will bear. The class is morphology, not date. A single rampart and ditch, as at Oliver's Castle, is typical of the earlier Iron Age on the downs; the multiple circuits at Yarnbury, Battlesbury and Barbury are the regional centres of the middle and later Iron Age, their banks still walking as separate lines. Reuse is common. Old Sarum holds a Norman castle and a medieval city inside an Iron Age circuit, and Vespasian's Camp carries a Roman name it has no claim to.

The working part is the LiDAR. The layers are the Environment Agency LIDAR Composite 1 m hillshade, DSM and DTM. A swipe control sets Esri World Imagery against the DTM hillshade, stripping the tree canopy: Wick Ball Camp and The Earldoms, near-invisible on the ground and on satellite imagery, resolve as bare earthworks. Pick a fort, then drag the gold bar along the bottom of the map.

Also on the page: search by name, parish or grid; filters for type, confidence and access; GPX, GeoJSON and KML downloads; and links out to OS Maps, the Atlas record and the NLS six-inch.

The Standing Stones of North West Devon

There is a corner of Devon that the megalith books skip. Dartmoor gets its rows and circles, Exmoor gets its miniliths and stone settings, and between and west of them lies a stretch of country — North Devon and Torridge, the Hartland peninsula round to Ilfracombe, the Taw–Torridge estuary, Lundy — that barely appears at all.

I have put together a map and gazetteer of what is actually there: 

nw-devon-standing-stones.netlify.app

It is not a long list. That is the point.

What's on it

Forty-eight records. A handful are unambiguous: the Whitestone above Ilfracombe, the three scheduled Damage Barton stones on the coast, the Mattocks Down pair, nine granite uprights on Lundy, and the Yelland double row on Isley Marsh — excavated in the 1930s and buried under estuary silt since the 1990s, which is the only lowland stone row in the county and the one monument here that would be famous anywhere else.

Most of the rest are quartz. That is the local signature: vein quartz and white quartzite blocks, the Gospel Stone above Woolacombe Bay, the Konk Stone in the next field, the scatter along Lee Downs. Whether some of those are set stones or quarried-out bedrock left standing is a live argument, and I have not tried to settle it.

Then there is a long tail. M.G. Palmer walked the Ilfracombe district in 1937 and published a list of standing stones; the Ordnance Survey went back in the 1950s and decided a good many of them were rubbing-stones, bound-stones or hedge clearance. Both readings are on the map, graded. And there are stones that are not standing stones at all but keep getting counted as such — the Devil's Stone at Shebbear, which is a recumbent sarsen with a folk custom attached, and West Middleton, which is scheduled as forty-eight post-medieval staddle stones and has been mistaken for a stone row more than once.

Everything is graded: certain, probable, possible, doubtful, lost, folklore, rejected. The rejected records stay in deliberately, so nobody rediscovers them.

The tithe layer

The pale diamonds are something else again: field-names from the 1830s tithe apportionments. Long Stone, Grey Stone, Hoarstone, White Stone, Broad Stone, Quoit.

A name is not a date. Some of these will record a prehistoric stone that has since gone into a hedge; many will be medieval boundary markers; some will be nothing at all. But in a region this thin on surviving monuments, the place-name evidence for stones that are no longer there may matter as much as the ones that are. Four fields called Long Stone west of Georgeham, with Longstone Lane running south from them, and no upright to be found — that is worth recording even though it proves nothing.

Toggle the layer separately. Where a grid reference is known the diamond sits on the field; otherwise it sits at the parish centre, which means it is a label, not a location.

Why bother

Two reasons.

The first is that absence is data. If you want to understand what the megalith builders of the south-west were doing, it helps to know where they were not. Mattocks Down, on the last downs before Exmoor, has a fallen menhir, a standing companion, a stone row recorded by Westcote in 1630 and destroyed since, and lost circles planned by Chanter and Worth. There is nothing remotely like that anywhere west of the Taw. The blank is real and it has a shape.

The second is more practical. Scattered across the HER, Palmer's paper, National Trust records and nineteenth-century notes, this material is effectively invisible. On one map with consistent grading, you can at least see what the questions are — and most of the remaining questions need boots and a tape measure, not another database.

A hat-tip to the Megalithic Portal

Anyone who has looked for a stone in the last twenty years owes a debt to the Megalithic Portal. Andy Burnham and a couple of decades of contributors have built the standing general resource for prehistoric sites in these islands, and it is the place most people start — including me.

This gazetteer is not drawn from it. The records here come from Historic England, the Devon HER, National Trust heritage records, Palmer 1937, Chanter and Worth, and the tithe apportionments, which makes the two independent where they overlap. That is useful in itself: two catalogues built from different sources that broadly agree are worth more than one.

It is also a much narrower thing. The Portal covers the world and does what a general gazetteer should — photographs, visitor logs, condition and access ratings, comments from people who have actually stood next to the stone and argued about whether it is bedrock. This site does none of that. It covers one district, west of Exmoor, and it grades records by how much confidence the sources will bear. That is the specialist job, and it only makes sense on top of the general one.

So: complementary, not competitive. If you want to see what a stone looks like, who has visited it and what they made of it, go to the Portal. If you want to know what the documentary record will and will not support for this particular corner of Devon, this map is the narrower tool.

Caveats

Almost nothing here is dated. Historic England's formula for standing stones is Late Neolithic to the end of the Bronze Age, for the few excavated examples — and none of the few excavated examples is in this area. "Neolithic" on this map is a possibility, not a result.

Many of these stones are on private land. Inclusion is not an invitation; the Yelland stones are on a tidal RSPB reserve and are, in any case, under the mud.

Corrections and additions welcome. Particularly from anyone who has walked Palmer's list.

The chalk has eaten my homework

Brian John has now put numbers on the claim that the Salisbury Plain we walk on is not the Salisbury Plain an Anglian ice sheet would have stood on. In today’s post the land surface is said to have fallen by 8–10 m over the last half million years. That piece sits on a run of earlier arguments: that the southern glacial limit is farther south than the mapped evidence (February 2022); that Salisbury Plain itself shows no traces of ice action; that sarsens and solution hollows are a long-standing dilemma (December 2016; August 2016); that periglacial processes have done much of the work (January 2023; Murton et al. in the Lyell volume); that Newall’s Mound is a solution hollow inverted into a hummock capped by 1.5 m of clay-with-flints, likened to a kame (28 August); and that the chalk downs are a “serene landscape but lethal geochemistry” in which any till would be destroyed.

The Paleogene silcrete that became the sarsen scatter is given 30–100 m of subsequent lowering, with a conservative 50 m used as a working figure. Quaternary rates are lined up from several places: 15–30 mm per millennium on modern UK chalk, 20–25 mm/ka from the Norber pedestals, 10–15 mm/ka in the Driftless Area, ~20 mm/ka on the Normandy chalk plateau. The conclusion offered is that “the landscape as it was at the time of the Anglian Glaciation … is simply not there any more. End of story.”

The rates are in the right order of magnitude. Chalk does dissolve. Solution pipes, coombe rock, head and clay-with-flints have been mapped across Wessex for more than a century. Jukes-Browne already knew that pure chalk does not contain enough insoluble residue to make the plateau clays we see. None of that is new, and none of it requires a Pleistocene ice sheet on the Wiltshire downs.

What the new numbers do not do is make a till sheet disappear without leaving the fraction that chalk chemistry cannot touch.

Alkaline water does not eat silicates

White Chalk is often more than 98% CaCO₃. Meteoric water charged with carbonic acid removes that matrix. The porewaters and rendzina soils that result are alkaline to circumneutral, buffered by excess carbonate. They are not an acid bath. Quartz, zircon, rutile, tourmaline and dense igneous clasts do not dissolve on Quaternary timescales. Frost shatters a dolerite. Periglacial attrition turns it into gravel, sand and heavy-mineral grains. It does not vaporise the silicate cargo into thin air.

Eight or ten metres of surface lowering on a carbonate plateau concentrates that residue; it does not erase it. Clay-with-flints is exactly that sort of remanié — chalk gone, insolubles left, later mixed with Palaeogene cover and cryoturbated. John now says as much of Newall’s Mound: an old hollow whose flanks have lowered while the fill stayed put, so the sink becomes a low hummock under 1.5 m of clay-with-flints. That is ordinary downland inversion. It is not a vanished Irish Sea till sheet, and dressing it as a kame analogue is a mountain made from a molehill. If ice had dumped even a thin and patchy till, the same logic still applies. After the carbonate matrix had gone you would still expect quartz, zircon, rutile, tourmaline, micro-lithics of dolerite and rhyolite, and far-travelled pebbles in residual soils, in those inverted caps, and in the river gravels that drain the Plain.

That test has been run.

The grains are local

Stream sands draining Salisbury Plain carry detrital zircon suites whose age spectra match recycled southern British Mesozoic and Palaeogene sources — the London Basin / Thanet Formation pattern — not a first-cycle dump from the Lower Palaeozoic of Wales or a crystalline northern shield. In the 2026 U–Pb study, 550 zircon analyses produced a single concordant grain at 464 ± 16 Ma, the Darriwilian age that characterises the Mynydd Preseli / Fishguard Volcanic Group sources of the Stonehenge rhyolites. One grain in hundreds is the opposite of the Welsh igneous signal required if ice had carried dozens of multi-tonne bluestones and ground the rest into rock flour. The apatite record is dominated by a Cenozoic reset (~60 Ma), not Palaeozoic basement.

Clast counts in the Wiltshire Avon terraces are the same story at pebble scale: local flint, sarsen and Greensand chert, not the exotic suite an Irish Sea or Welsh ice stream would impose. Field survey still fails to produce a bluestone erratic in an undisturbed Quaternary context away from the monument and its debitage. The Newall boulder matches Craig Rhos-y-felin rhyolite of the Stone 32d group and was moved by people. Coastal patches in North Devon or Somerset do not put a grounded lobe on the Wiltshire chalk plateau.

If 8–10 m of the old land surface has gone, those minerals and clasts are what should have been concentrated into the modern drainage. They are present. They do not show a Welsh ice sheet.

Models are not deposits

Numerical ice-sheet models are not a substitute for sediment on the ground. A model can be made to grow ice almost anywhere if the climate forcing, sliding law and bed friction are generous enough. That is why older, lightly constrained experiments are still cited as showing it was “perfectly possible” for ice to reach east of the chalk scarp. Possibility in a sensitivity run is not a reconstruction.

The reconstructions actually tied to mapped margins and dates — BRITICE-CHRONO and the later ensemble tests against flow geometry and deglaciation timing — put Devensian Irish Sea ice to the Celtic Sea shelf edge and onto parts of the north Devon and Somerset coasts. They do not put a grounded lobe over the Wiltshire chalk plateau. The Anglian ice sheet, the one usually invoked for a 450-ka till on Salisbury Plain, is even less a creature of those high-resolution Devensian models. Its limit of preserved evidence remains well north of Wessex.

And even if a future ensemble were forced across the downs, the prediction would still be a first-cycle Welsh and Irish Sea mineral load in the Avon sands and terrace gravels. That load is not there. A model that contradicts the grains is a discarded run, not a hidden ice sheet.

The loop has a new rung

The argument has retreated through a closed set of auxiliary hypotheses:

1. Missing moraines and till → cold-based ice.
2. Need large megaliths moved → warm-based basal plucking.
3. Missing gravel-sized erratics across the Plain → corrosive chalk chemistry (27 August).
4. Why the bluestones survive at one later monument → left unexplained.
5. New: the land surface itself has been lowered by 8–10 m (29 August), so the original deposits “must long since have disappeared.”

Each failed prediction is reclassified as a feature of the model. The fifth step is the same move as the third, only taller. Surface lowering is not in dispute. What it does not licence is the claim that durable silicates vanished with the carbonate. After 150 years of stratigraphy, terrace clast surveys and grain-scale U–Pb geochronology, the record across Salisbury Plain is not silent. It is negative.

A stranger in the family grave

Thirty new genomes from southwestern Britain: Bronze Age newcomers burying their dead in thousand-year-old Neolithic tombs — and why we still can't say whose ancestry the last seven per cent was.

---

Around 3200 BC four people were buried together at Monkton-up-Wimborne in Dorset. Close family — two first-degree pairs, one mitochondrial lineage between them. Either four siblings, or a woman and three children aged about five, nine and ten. The grave was sealed.

Fifteen hundred years later somebody opened it, laid a man in the middle of the earlier cut and capped him with flint. Not related to them. Not descended from them. From a population that didn't exist in Britain when they were buried.

The same at Sisters Long Barrow, where a young woman was placed in a scoop cut into the top of the mound two millennia after the three men beneath her — and at Sale's Lot, where the secondary burial is also a woman. Two of the three reused monuments in the new paper by Vuković and colleagues received women. The authors are right that three burials prove nothing. It is a well-timed three, six weeks after aDNA showed the Upton Lovell "shaman" — goldworking toolkit, ceremonial cloak, battle axe — was female.

(a) Map of southwestern England showing the location of each of the investigated burial sites. Black square on the inset map indicates the location of the study area. (b) Radiocarbon-dated individuals in the study per site (y-axis). For each individual, the 95.4% calibrated probability interval is shown in grey. Individual labels on the x-axis are coloured blue (male) and red (female). The orange dotted vertical line represents the point estimate for the Amesbury Archer; its 95.4% calibrated probability interval is 2470–2239 BCE. - Fig 1 from Vuković et al.

The paper covers thirty individuals from twelve sites in Gloucestershire and Dorset, c. 3800–1400 BC. Everyone before about 3100 BC looks like a European Neolithic farmer; everyone after about 2550 BC looks Bell Beaker. The archaeological reading — monuments keeping their meaning for people with no ancestral claim on them — is careful and, I think, right.

I'm interested in a different part.

The question they tried to answer

British ancestry changed by something like ninety per cent between 2450 and 2000 BC (what that number measures). The interesting remainder is the seven to nine per cent of Neolithic-farmer ancestry in English Chalcolithic and Early Bronze Age people.

Everything turns on where it came from. Descended from Neolithic Britons, and some Neolithic families had descendants. Carried across the Channel by incomers who already had farmer ancestry, and it says nothing about British survival at all. Same number, opposite meanings — at length and formally.

This paper tests it better than anyone has. The usual weak point is the local Neolithic reference, pooled from wherever data happen to exist. Here it is nine individuals from the same regional transect, in some cases the same monuments as the targets.

What came back

The earliest Bronze Age individual — the Sale's Lot woman, c. 2620–2460 BC, buried within a generation or two of the Amesbury Archer — comes back entirely continental. Most of the rest are estimated at zero, or at a positive number whose error bars include zero. Likewise across a further 118 English Bronze Age individuals from the published record.

One exception, which the paper flags as its clearest positive: a woman from Fir Tree Field Ring Ditch, at 45.3 per cent local Neolithic ancestry. The confidence interval runs from 4.2 to 86.3 per cent.

The strongest evidence in the study for Neolithic British ancestry surviving cannot distinguish a trace from most of a genome. It rests on fewer than fifty thousand markers. And she lived around 1400 BC, in the window where farmer ancestry across Bronze Age Britain rises again — a rise generally put down to renewed contact with the continent.

The part that matters

These are whole genomes, shotgun-sequenced and publicly deposited. Every downstream analysis then reduces them to single random base calls at 1.2 million pre-chosen positions — the 1240K panel, the field's common currency for a decade.

Not laziness: it is the only way to talk to the comparative dataset. But the panel discards the rare variation carrying information about specific recent shared ancestors, which is the one thing that could separate a Dorset Neolithic grandmother from a Rhineland one. Sequenced, then thrown away at harmonisation, in a paper published three weeks ago.

The honest qualification is that coverage is thin — median 0.27×, range 0.014–0.78× — so the better analysis probably wouldn't have run anyway. Which makes the constraint structural rather than accidental. (These libraries also mean English Chalcolithic–EBA shotgun data now exist, where when I wrote there weren't any.)

Where it leaves things

The arithmetic is unchanged: seven to nine per cent, well measured, unaddressed. What has changed is that the failure has been reproduced independently, with a better local reference than published data allow, by people with a laboratory.

The man in that Dorset grave wasn't descended from the family whose bones he was laid among. That much is solid, and striking. Whether anyone in Bronze Age Britain descended from Neolithic Britain, we still can't say.

---

Vuković, N., Bernhardsson, C., Edlund, H. et al. Diachronic reuse of Neolithic burial monuments by Bronze Age newcomers in Southwestern Britain. Sci Rep 16, 26819 (2026). https://doi.org/10.1038/s41598-026-66094-z

Rock on a Rope, how to level the lintels

 

In an earlier post on setting out the Stonehenge sarsens I suggested a simple way of establishing a horizontal: https://www.sarsen.org/2014/08/how-to-construct-level-sarsen-circle.html , place two posts either side of a small pool of water (an “Aurochs skin full of water”) and sight across their tops. Adjust the posts so their tops both have the same length of post above the water surface, the line of sight is level and can be transferred around the circle. 

A complementary approach starts from a different but equally basic physical fact: a free-swinging pendulum released from rest reaches essentially the same height on the opposite side of its arc. A tall central timber mast, a long fibre rope, and a heavy stone or timber bob are all materials the Late Neolithic builders already used. Release the bob from one marked position and it will rise to the matching height on the far side; adjust a second post until the bob just reaches it. For more accuracy, you release the pendulum from post A and it marks on post B the end of the first swing. When it comes back to post A the second time, it's going to be just a little bit lower because of air resistance. Halfway between the first and the second mark on post A will be the same as a mark on post B. Subtle and easy to get an accurate reading. Repetition is quick, free and easy so a reliable and repeatable measure can be recorded.  Rotate the plane of swing or shift the mast a short distance and a small set of levelled reference posts can be established around the intended circle. Once three or four reliable points exist, ordinary sighting or taut-string transfer finishes the job.

The two methods form a small family of “gravity levels”:

• Water-pool / skin method – immediate visual reference, excellent for a local datum, but needs a still surface and a reliable container.
• Pendulum method – needs no liquid, generates widely spaced equal-height points from one central mast, and is cheap to repeat when wind deflects the swing.

Neither is attested archaeologically. Both are simply possible with the technology of the time and address the practical problem of creating a consistent horizontal on sloping chalk so that the sarsen uprights could finish with level tops for the continuous lintel ring. And avoid the need for a water filled pig intestine or portable troughs as are often suggested as alternatives.

Pendulum method being tested