To ring every order of the bells once and come home, you must walk the symmetric group by single steps. English ringers were doing it by 1621 — forced into it by the weight of a bell.
A tower bell is not a toy. A ring of eight runs from a couple of hundredweight up to a tonne or more of bronze, and each is rung full circle — swung mouth-up, over the balance, through very nearly a complete turn and back. At that size the swing has its own clock. A ringer can hurry the bell a little or hold it a little, but cannot stop it, cannot double it, cannot make it leap. From one pull to the next a bell can shift its place in the order by at most one.
So the bells ring in a sequence — call one whole sequence a row, an ordering of all the bells from treble to tenor — and the rule that physics writes is severe: from one row to the next, no bell may move more than one place. Neighbours may swap — one pair, or several disjoint pairs at once — and that is all. You cannot send the tenor to the front. You can only ever exchange neighbours.
This is a constraint a mathematician would recognise instantly and might never think to impose. It says: the only move available is to swap neighbours — the adjacent transposition. And the ringers, who imposed nothing and only obeyed the rope, spent four centuries discovering what that single permitted kind of move can and cannot reach.
Ringing begins and ends on rounds — the plain descending scale, treble to tenor, 1 2 3 … n, highest bell first. Everything in between is the bells passing through other orders. The art, called change ringing, is the art of leaving rounds, moving through orderings by legal changes, and returning — without ever repeating a row.
Set the engine to Plain hunt and trace one bell. Every bell does the same thing, just out of step with the others: it works its way from the front of the row to the back one place at a time, rests a beat against the wall, then works back to the front. The path of a single bell — ringers call it the blue line, and they memorise it, because in the tower there is no score and no conductor reading the rows aloud — is a clean zig-zag, corner to corner.
Plain hunting is lawful and lovely, and it is not enough. Watch the verdict under the engine: a plain hunt on n bells comes back to rounds after just 2n rows, having touched only 2n of the orderings. On four bells that is eight rows out of twenty-four. The weave closes too soon. To ring every order, you need more than a weave; you need to break the pattern at exactly the right moments — and that is where the centuries of craft went.
The ringers' grail is the extent: ring every possible order of the bells once and only once, by legal changes, and come home to rounds. On n bells there are n! orders — 6, 24, 120, 720, 5040… — so the extent on six bells is 720 rows, and on twelve it is 479,001,600, a quantity no band has ever rung (it would take decades without sleep). The complete extent on eight, all 40,320 rows, has been rung: about seventeen hours, no rest, no repeat. It is one of the longest feats of memory and stamina in any human art.
Strip the romance and what is the extent, exactly? Make a graph. Put one dot for every ordering of the bells — n! dots. Join two dots whenever you can get from one to the other by a single legal change, i.e. by swapping two neighbouring bells. An extent is then a path that visits every dot exactly once and returns to its start.
An extent is a Hamiltonian cycle on the Cayley graph of the symmetric group, generated by the adjacent transpositions. The ringers were the first people to need one — and they found them by hand.
That graph has a name and a shape, and you can turn it in the next instrument. But first: what does a complete extent actually look like as a procedure? Set the engine to Plain changes and ring it. One bell hunts steadily out to the back and in to the front; each time it reaches an end, the bells waiting behind it make a single quiet swap, and that small swap is what tips the whole pattern into a fresh region of orderings it has not yet visited. Run it on four bells and the verdict confirms what your eye suspects: 24 rows, all 24 orderings, each exactly once, every change legal. A Hamiltonian cycle, rung.
Here is the turn this whole piece exists for. In the late 1950s and early 1960s, three people — Hugo Steinhaus posing it, Selmer Johnson and Hale Trotter publishing it independently in 1962–63 — gave computer science a clean algorithm for a natural problem: generate all n! permutations of a list so that each differs from the one before by a single adjacent swap. A minimum-change ordering — a Gray code for permutations. It is taught today as the Steinhaus–Johnson–Trotter algorithm.
It is the same thing the engine just rang. The constraint the algorithm imposes for elegance — change as little as possible, one adjacent swap at a time — is the constraint physics imposed on a half-tonne of swinging bronze. The procedure the ringers call plain changes, in which one bell hunts while the rest make a single swap at each turn, is Steinhaus–Johnson–Trotter, move for move.
This is not a loose analogy, and it is not my claim to make: it is Donald Knuth's. In The Art of Computer Programming, Volume 4A, in the section on generating all permutations, Knuth presents the algorithm under the ringers' own name — "Algorithm P (Plain changes)" — and traces it explicitly to English change ringing, noting that the method was known to ringers and described in print by the 1600s. The computer scientists rediscovered, and named after themselves, a procedure that had been walked by hand and ear in stone towers for three hundred years.
The earliest printed systematic treatments come from the circle around Fabian Stedman — Tintinnalogia (1668, largely written by Richard Duckworth and published by Stedman) and Stedman's own Campanalogia (1677) — and ringers were already ringing changes by method, not just rounds, in the early 1600s; plain changes on four bells are documented from around 1621, with a manuscript method by Peter Mundy from 1653. These books, written to teach a craft, contain some of the earliest sustained printed work anyone did on the systematic enumeration of permutations — not group theory (its language and abstractions were two centuries off) but the practical thing the theory would later be about, pursued by people who only wanted to ring every order of the bells and not get lost.
// all n! perms, each one // adjacent swap from the last give every value a direction (←) repeat: find the largest value that is mobile — its neighbour in its own direction is smaller if none: stop swap it that way (one place) reverse the direction of every value larger than it emit the new arrangement
// ring every order once, home // to rounds; one bell "hunts" the hunt bell works to one end one place per row (a neighbour swap each row) when it reaches the wall and lies still for a blow, the bells behind it make a single swap (the next "place") then the hunt turns and works back the order quietly advances
Return to the graph: a dot for every ordering, an edge for every legal swap. It is the Cayley graph of the symmetric group Sn on its adjacent-transposition generators — and it has a beautiful geometric body, the permutohedron. Drag to turn it.
On three bells the body is a flat hexagon: six orders, each joined to two others, a single ring. There is essentially one extent and it simply goes round. On four bells the body lifts into three dimensions as the truncated octahedron — 24 corners, 36 edges, the Archimedean solid that also happens to tile space. Every corner is one ordering of the four bells; every edge is one legal change; and an extent is a closed tour along the edges through all 24 corners. The plain-changes extent is one such tour, lit on the solid below.
No one in a seventeenth-century tower used the words group, generator, Hamiltonian cycle, Gray code. The symmetric group would not be named for another two hundred years; graph theory and the permutohedron later still. And yet the thing itself — the structure — was fully present, because the ringers were standing inside it, pulling on it. A band ringing a long extent is a distributed computer with no central clock: each ringer holds only their own bell's blue line in memory, listens, and obeys the change. Out of n people each running one local rule comes a single global walk that touches every arrangement exactly once and returns. That is the symmetric group, executed by humans and bronze, as sound, in time.
It is one of the cleanest cases I know of a deep mathematical object arriving through the body before it arrives through symbols — discovered not by a mind reaching for abstraction but by hands obeying a weight. The computer scientists who later wrote the algorithm down were not wrong to be proud of it; it is a good algorithm. They were simply not first. The ringers were first, and the ringers did not know they had found anything but a way to ring all the bells.
The bell could only ever swap with its neighbour. Everything else — the whole symmetric group — follows from refusing to repeat.
This is a piece at the join of campanology, combinatorics, group theory, and acoustics. Two rules govern everything in the Wasteland: it must be interesting on its own terms, and it must never lie about anything real. So here is exactly what is computed, what is asserted, what is cited, and where the claims stop.
Every row is generated live, not stored. Plain hunting is generated by alternating disjoint adjacent swaps; the full extent is generated by the plain-changes / Steinhaus–Johnson–Trotter procedure (largest-mobile-element method). The green verdict under Instrument A is computed in your browser at run time: it independently confirms that the run contains exactly n! rows, that they are all distinct (so every order appears once and only once), and that every change moves each bell by at most one place — i.e. it is a legal change. Nothing is taken on trust; the page checks itself in front of you. Instrument C runs both characterisations and asserts only what it has verified: that the plain-changes sequence and a from-scratch generation agree row for row.
The audio is synthesised, not recorded. Each strike is built from the five partials a well-tuned English bell is tuned to carry — hum, prime (fundamental), tierce, quint, nominal — at the classic ratios ≈ 0.5 : 1 : 1.2 : 1.5 : 2 of the prime. The tierce sits a minor third above the prime, which is the real reason bells carry a faintly mournful, minor colour; the nominal is the octave above the prime. The pitch you actually name the bell by — its strike note — is a virtual pitch: an octave below the nominal (usually coinciding with the prime), heard by the ear rather than present as its own partial. This five-partial true-harmonic scheme is associated with the principles championed by Canon Arthur B. Simpson in the 1890s ("On Bell Tones," 1895–96) — a re-derivation by rule of what the 17th-century Hemony founders had reached by ear. I have used these ratios because they are the honest physics of a bell; a real ring has further inharmonic partials and a far richer decay than these few sine components reproduce. It is a sketch of a bell's spectrum, and labelled as one.
This layer opens a new vein in the ground — Pattern: combinatorics and craft, order made audible, structure that arrives through the hands before it arrives through symbols. It is kin to the Number seam (this is, underneath, group theory and the enumeration of permutations) but it is not about a number refusing to resolve; it is about a human craft that turned out to be an algorithm. Future Pattern layers might run through weaving and braid groups, knots, the combinatorics of bell-ringing's named methods, tilings, or any place where people built a deep structure with their hands before anyone wrote its theory.
Verified before publishing by an adversarial fact-checking subagent over the historical and mathematical claims (the SJT/Knuth naming, the Stedman dates and authorship, the permutohedron facts, the bell-partial ratios, the ringing terminology), and by live in-browser computation for every row, every extent, and every "legal change" check. Where a source disagreed or a claim risked overreach, the phrasing above was pulled back to what holds.