Dropping the Central Trilithon: a physics toy

 

Dropping the Central Trilithon: a physics toy for Stonehenge's oldest cold case

Open it here

Nobody recorded the tallest stones at Stonehenge coming down. Stone 55 of the Central Trilithon — the pair that once framed the winter-solstice sunset, and which most books will insist on calling the "Great" Trilithon (a friend who knows these stones far better than I do assures me the name properly belongs elsewhere, and I've learned not to argue with him) — fell and snapped in two at some unrecorded moment in antiquity, taking lintel 156 with it. By the time anyone drew the monument reliably, the wreckage was already lying where it lies today: 55a, 55b and the lintel sprawled across the recumbent Altar Stone, pinning it to the ground. Its partner, stone 56, leaned on alone until it was hauled upright and set in concrete in 1901.

That wreckage poses a question archaeologists still argue about: was the Altar Stone standing when the trilithon fell on it — or was it already lying flat? The Altar Stone has never been fully excavated, partly because the fallen sarsen pieces sit on top of it. So the resting positions of four stones are nearly all the evidence there is.

Which makes it a perfect problem for a physics sandbox. If every hypothesis has to end with the stones where they actually are, you can throw hypotheses at a rigid-body simulator all day and see which ones survive.

The toy

The simulation below (or [here], if the embed misbehaves) is a 2D cross-section along the solstice axis, built on the Matter.js physics engine at true scale — the uprights are 6.7 m proud of the ground, the lintel a metre-thick slab perched at over 7 m. Stone 56 is fixed, as the evidence demands. Everything about stone 55 is a slider:

• how far it leaned before letting go, and how sharp the subsidence "kick" was;
• how deep its socket was — 56's was measured at a famous 2.4 m in 1901; 55's is thought to have been embarrassingly shallow, which is presumably why it's the one that fell;
• how strong the sarsen was (how hard an impact snaps it in two);
• how tenaciously the mortise-and-tenon joints dragged the lintel along before letting go;
• and, in the standing scenario, how firmly the Altar Stone was bedded.

Dashed outlines mark where the pieces lie today. After every run, a verdict card checks the outcome against them, and a "trench notebook" narrates the collapse: the base kicking out of its socket, the tenons shearing, the fracture, the strike. Three preset buttons reproduce the runs discussed below, so you don't have to take my word for any of it.

What the model says

The lazy expectation is that one scenario matches and the other doesn't. That's not what happens, and the way it fails is the interesting part.

Fall onto an already-recumbent Altar Stone, and everything works on nearly the first try. With middle-of-the-road parameters, stone 55 tips, its upper end strikes the raised edge of the flat slab at around 4 m/s, and it snaps over that edge like a bar over a fulcrum — the butt end flat at about 2 m, the upper half flat at about 5 m across the altar, the lintel coming to rest on top of the pile. That is, uncannily, the arrangement in the ground.

Fall onto a standing Altar Stone, and the knock-over itself is easy — almost too easy. A ~30-tonne slab sweeping through its arc fells a half-metre-thick pillar without much argument. But here's the result I didn't see coming: the collision absorbs so much of the fall's energy that stone 55 tends to land in one piece. The standing altar acts as a crumple zone. And an unbroken 55 contradicts the most solid fact we have — the real stone broke. To get the full observed sequence in the standing scenario (altar felled flat and 55 snapped), I have to dial the sarsen strength down to a conveniently flawed stone. Push the other way — bed the altar firmly — and you get a third outcome the ground flatly rules out: a permanent stalemate, with 55 propped against a still-standing Altar Stone like a failed game of dominoes.

So the model can't prove which history happened; with enough slider-turning, both scenarios reproduce today's arrangement, and that underdetermination is honestly the deepest lesson in it. But the two scenarios are not on equal footing:

Falling onto a recumbent Altar Stone breaks stone 55 for free — the slab's edge is exactly the anvil the fracture needs. Felling a standing Altar Stone instead cushions the fall, and the break has to be bought with an extra assumption about a weak stone.

Call it an argument from parsimony, delivered by a physics engine. It lines up with where much of the archaeology has been drifting anyway: many researchers suspect the Altar Stone was placed recumbent by design, a threshold rather than a pillar. The simulation adds a small, mechanical voice to that side of the debate: the flat-altar story needs nothing special to be true.

The Tenon Problem and an Insight

A real difficulty is that the mortise-and-tenon joints at Stonehenge were remarkably tenacious. Elsewhere on the monument we can see the evidence: stones 6 and 7 of the outer circle, one leaning out and the other in, still carried their twisted lintel until a scaffold was put up in 1881; observers at the time noted that without the joints that section would have collapsed long before. Stone 56 of the Great Trilithon itself stood for centuries at a lean of something like 12–15° after its partner and lintel had already fallen. A simple model that treats the joints as springs, or even as a clean kinematic coupling that releases after a few degrees of lean, cannot reproduce this. The joints were “sticky”; they held under conditions that look precarious to us.

What made the Great Trilithon different was almost certainly not a sudden failure of the tenons at the top, but the progressive failure of the shallow socket at the base of stone 55 — the heel being pushed or washed out. Once that footing gave way, the upright was free to rotate past the point of no return while the still-engaged joints dragged the lintel with it. Our 2-D simulations are limited because they cannot properly separate those two mechanisms: a base that can soften and release, and an upper joint that remains tenacious until the geometry finally forces it apart.

Even so, the simplified model yields two clear archaeological implications. First, trilithons fail from the base, not from the top; it is the heel coming out that allows the structure to fall. Second, the observed breakage of stone 55 is far more consistent with an already-recumbent Altar Stone acting as an impact surface than with a still-standing one. More elaborate modelling of the joints can wait; these two results are already visible.

Caveats, cheerfully admitted

This is a toy, and it wears its assumptions on its sleeve. It's 2D, so both uprights live in the same cross-section and the lintel's three-dimensional tumble is collapsed onto a plane. Fracture is a threshold on impact velocity at the contact point, not real crack mechanics; the tenon coupling — genuinely unknowable — is a parametrized "how long did the joints hang on" knob rather than simulated joinery. The stones are rigid rectangles on level ground. None of the slider ranges pretend to precision; they pretend to span the plausible, which is all a hypothesis machine needs.

The right way to read it: not "this is how it happened," but "here is the space of ways it could have happened, and notice which corner of that space doesn't need a coincidence."

Go push the stones over yourself. Start with the "already recumbent" preset, then try to make the standing scenario work without weakening the stone. That failure is the most informative thing in the whole toy.

---

Built with Matter.js. The trilithon geometry follows the standard published dimensions; socket depths per Gowland's 1901 excavation of stone 56. Nomenclature corrected under expert duress. All remaining errors of physics and prehistory are mine.