The Golden Age

For its first three thousand years, the puzzle traveled hand to hand — papyrus to scroll, scroll to monastery vellum — at the pace of ink and luck. Then Europe built the printing press, and later the steam press, the penny newspaper, and the commuter railway, and the puzzle discovered what it could do with an audience of millions. Between 1612 and 1945 the genre acquired everything we now think of as modern: books and bylines, columns and crazes, prize money and plagiarism, and at last scientific respectability. It also staged two instructive spectacles — a fraud that fooled the world for a century, and a mass demonstration that effort cannot beat arithmetic. This chapter is the story of that age, told through the people who built it.

The Bookshelf

It begins, properly, with a book. In 1612 Claude-Gaspard Bachet published Problèmes plaisants et délectables — "pleasant and delightful problems" — the first true book of recreational mathematics: number tricks, Josephus-style elimination games, and a deceptively humble question about weights. With just four weights — 1, 3, 9, and 27 pounds — and a two-pan balance on which weights may sit in either pan, you can weigh every whole-pound load from 1 to 40. Inside that party trick hides ternary arithmetic, smuggled into the reader's head as shopkeeping. Bachet's deeper gift was canon-formation: puzzles now had provenance — who posed what, and when — which is precisely what the coming age would learn to fake. (The same Bachet edited the edition of Diophantus in whose margin Fermat would scribble the most famous note in mathematics.)

In 1694 Jacques Ozanam followed with Récréations mathématiques et physiques, four volumes — the encyclopedia stage, the reference that fed every nineteenth-century puzzle writer. The armory was stocked and shelved. What the genre lacked was a detonator, and industrial printing supplied it: cheap paper, mass literacy, and dailies hungry for anything that made a reader buy tomorrow's edition.

In the winter of 1880, the world found out what that combination could do.

The Craze

Around 1874, Noyes Palmer Chapman, the postmaster of Canastota, New York, built a small wooden tray holding fifteen numbered sliding blocks and one empty space. It traveled quietly for a few years — friends, family, a few copies — and then, in January 1880, it went off like a charge. For four months, roughly January to April, the 15 Puzzle was a world craze: pushed in parlors, offices, and lecture halls across America and Europe, denounced by editorialists, blamed for lost workdays. Chapman filed for a patent only that March, as the wave was already cresting.

The engine of the mania was a specific challenge. Set up the board in order, but with tiles 14 and 15 swapped; slide until everything reads correctly. A thousand-dollar prize was attached to it — an offer whose loudest and longest promoter was the American puzzle showman Sam Loyd, a name to hold on to. Millions pushed tiles. Fortunes of attention were spent. And the entire time, the door was already locked: in 1879–80, in the American Journal of Mathematics, two mathematicians named Johnson and Story proved that the challenge is impossible. Not unsolved — impossible, the way five can't be even.

It is the book's next ritual, and it earns its three stars honestly: the puzzle is not to find a solution but to see, with certainty, why none exists.

★★★ 38 · The 14-15 Puzzle

A 4×4 tray holds fifteen sliding tiles, numbered 1–15, and one empty cell. A move slides an adjacent tile into the empty cell. In the solved board, the tiles read 1 through 15 in order, left to right, top to bottom, with the empty cell last, in the bottom-right corner. You are handed the board almost solved: everything is in place except that 14 and 15 have traded positions. By sliding alone, restore perfect order. Sam Loyd offered $1,000 for a solution. Should you try?

Pause here. Decide, at least, what kind of evidence would settle the question.

Solution. Don't try. No sequence of moves works — and the proof never examines a single sequence.

Count the empty cell as a phantom sixteenth tile. Any position of the board is then an arrangement of sixteen pieces in sixteen cells, and every move is one swap: the blank trades places with a neighboring tile. Now track two bookkeeping quantities.

First, the arrangement's parity. A basic fact about arrangements: a single swap of two pieces always flips an arrangement between "even" and "odd" — where even means reachable from the solved order by an even number of swaps, odd means odd. (Order matters not; the count's parity is fixed.) Since every move is exactly one swap, the arrangement's parity flips on every move.

Second, the blank's distance from home, counted in grid steps (rows plus columns away from the bottom-right corner). Every move drags the blank one step up, down, left, or right, so this distance changes by exactly ±1 each move: its parity also flips on every move.

Two lights, blinking in perfect sync. So consider their sum. Each move flips both parities at once, and flipping both leaves the sum's parity unchanged — the sum modulo 2 is an invariant: no legal move, and therefore no sequence of legal moves however long or ingenious, can ever alter it.

Now evaluate the two boards. The solved board: zero swaps from itself (even), blank at home, distance zero (even). Even plus even — the invariant reads even. The prize board: exactly one swap from solved — 14 and 15 traded — so its arrangement is odd, while its blank sits at home (even). Odd plus even — the invariant reads odd. Even can never become odd under moves that preserve the sum. The prize position is unreachable. Every one of the millions of solvers was pushing tiles around one half of a universe whose other half they could never enter: exactly half of all arrangements are reachable from the solved board — 16!/2 of them, about ten trillion — and the theorem checks exhaustively in miniature: on a toy 2×2 tray, exactly 12 of the 24 arrangements can be reached.

Loyd's thousand dollars was perfectly safe. He knew it.

The Move — Impossibility by invariant. Before grinding at any transformation problem — tiles, a twisted cube, a database migration — hunt for a quantity that every legal move conserves, and check whether start and goal agree on it. If they disagree, you are done, gloriously: ten minutes of invariant-hunting can refute ten thousand hours of trying. This is the Lo Shu's forced 15, promoted from arithmetic to dynamics, and it may be the single most valuable reflex in this book.

The public of 1880 did not have the proof, and pushed on for months — the first mass collision between honest effort and an invariant, and effort lost, as it always does. The lesson was expensive, was not learned, and will be offered again: more than one of this book's hardest puzzles is solved by refusing to start.

The Showman

Sam Loyd was the greatest puzzle showman who ever lived, and that sentence needs both halves. An American already famous in his teens for chess problems, he had a genius for the puzzle as spectacle. He sold his Trick Donkeys puzzle to P. T. Barnum — cut three cards, rearrange, and two riders somehow mount three donkeys. His 1896 masterpiece, Get Off the Earth, put thirteen warriors on a rotating disc; turn it slightly and count again — twelve. No warrior is hidden. Each figure has quietly donated a sliver of itself to the others, and the thirteenth exists only as the sum of slivers. The vanishing warrior still teaches the right question — watch what is actually conserved — and Loyd's posthumous Cyclopedia (1914) piled up a mountain of such material, not all of it his: he lifted work freely, including from the English puzzlist you are about to meet.

Then there is the fraud. From 1891 until his death, Loyd claimed — in print, in interviews, decade after decade — that he had invented the 15 Puzzle and engineered the craze of 1880. The claim entered the reference books and sat there, repeated in good faith, for over a century. It was demolished only in 2006, when Jerry Slocum and Dic Sonneveld published an archival investigation, The 15 Puzzle Book, tracing the puzzle's actual paper trail — to the postmaster of Canastota, Noyes Palmer Chapman. Loyd had not even entered the story until the mania was old news. Note the chronology with appropriate awe: the impossibility proof (1879–80) predates Loyd's claim of invention (1891). He spent his fame promoting a prize he knew mathematics had already voided, for a puzzle he did not create.

Genius and fraud in one biography — and, for the reader, a permanent calibration: this genre's two recurring cons are false attribution and fake difficulty, and they are as old as the business itself. You have now seen the first in its purest form; the "Einstein's riddle" nonsense of a later chapter is the same trick with a borrowed corpse. Check provenance. Even on puzzles. Especially on puzzles.

The Craftsman

England's answer to the showman was a quiet, self-taught puzzle writer named Henry Ernest Dudeney, and the contrast between the two men is the moral spine of this era. Where Loyd optimized for spectacle, Dudeney optimized for elegance — and he made elegance an explicit, almost mathematical criterion: the fewest pieces, the shortest path, the cleanest idea.

His masterpiece began as a challenge to his readers. On 6 April 1902, in his Weekly Dispatch column, Dudeney posed the haberdasher's problem: cut an equilateral triangle into as few pieces as possible that reassemble into a perfect square. That such a dissection exists was old news — a theorem of Wallace, Bolyai, and Gerwien (1807–1833) guarantees that any polygon can be cut into pieces that form any other polygon of equal area. Existence was free; the game was minimality. Four weeks later, on 4 May 1902, Dudeney published his answer: four pieces. Better: the four pieces can be hinged in a chain, so that the whole assembly swings one way into the triangle and the other way into the square, like a wooden animal turning inside out. In 1905 he showed a polished brass-hinged model to the Royal Society. It entered his classic collection The Canterbury Puzzles (1907), joined over the years by the spider-and-the-fly (1903), Amusements in Mathematics (1917), and — in the Strand, July 1924 — the puzzle that became the archetype of an entire genre, SEND + MORE = MONEY. Several of these are lying in wait for you later in this book.

The haberdasher's problem also carries this chapter's longest fuse. Was four actually the minimum, or had Dudeney merely failed to find three? Nobody could say — for one hundred and twenty-two years. In December 2024, Erik Demaine, Tonan Kamata, and Ryuhei Uehara proved that no three-piece dissection exists (arXiv 2412.03865): Dudeney's four pieces are optimal, the first such optimality certificate ever obtained for a nontrivial dissection. Sit with that asymmetry. Finding the beautiful answer took one Edwardian columnist four weeks; proving that nothing better exists took mathematics twelve more decades. Doing and proving you can't do better are different sports, played on different timescales.

The two careers intertwined, not always with consent — Loyd reprinted Dudeney's inventions under his own name, one more entry in the provenance ledger. But the deeper legacy is Dudeney's standard: that a puzzle's answer can be not merely correct but right — minimal, surprising, and worth building in brass.

The Don at Play

While the showman and the craftsman fought for the newspaper public, an Oxford mathematics don was smuggling logic into the nursery. Charles Lutwidge Dodgson — Lewis Carroll to everyone since — treated the puzzle as a literary form. His Doublets, launched in Vanity Fair in 1879 (the idea, he said, came to him on Christmas Day 1877), asks the solver to turn one word into another, changing a single letter at a time, every intermediate rung a real word. It looks like a parlor game. It is training in making moves that are individually trivial but must remain globally legal — and, a century later, computer scientists would recognize word ladders as a textbook exercise in searching a graph. Carroll had invented an algorithms assignment before there was a discipline to assign it.

A Tangled Tale (1885) wrapped mathematics in fiction; The Game of Logic (1886) turned the syllogism into a children's board game. And in 1895, in the philosophy journal Mind, Carroll published the deepest four pages in this whole golden age: "What the Tortoise Said to Achilles." Zeno's two runners return — you met them a chapter ago — but this time the race is an argument. The Tortoise accepts every premise of a simple deduction but refuses to accept the conclusion until the rule licensing the inference is itself written down as a premise. Achilles obliges. The Tortoise then requires the rule licensing that — and so on, forever. The regress never ends, and its moral has never been improved on: the rules of inference cannot all be premises; justification must stop somewhere. Logic, the discipline that studies proof, cannot itself be proved all the way down. Working logicians still cite the children's author, without irony, on exactly this point.

Black Squares, Two Proofs

On 21 December 1913, Arthur Wynne, a journalist assembling the Christmas fun page of the New York World, published a diamond-shaped grid of interlocking words he called a "Word-Cross." The typesetter's transposition — cross-word — stuck. For a decade it remained a minor feature; then the craze of 1924–25 made it an industry, and the crossword settled into the position it has never surrendered: the first puzzle in history to become a daily habit for millions, puzzling as ritual rather than event.

Britain, characteristically, complicated it. From 1926, the setter Torquemada pioneered the cryptic crossword, where each clue is a tiny machine of misdirection; four decades later Ximenes codified its rules of fair play (1966). The cryptic's great invention is structural: every fair clue contains both a definition and an independent wordplay route to the same answer — two separate proofs, required to agree. The solver doesn't merely find an answer; the answer arrives pre-verified, redundancy built into the object itself. Engineers, accountants, and cryptographers had discovered the value of independent confirmation in their own trades; the cryptic crossword taught it to the breakfast table.

Into the Laboratory

Three quiet events, clustered near this age's end, changed what puzzles were for.

In 1901–02, in the Annals of Mathematics, a Harvard mathematician named Charles Bouton did something to the matchstick game of Nim that no one had ever done to any game: he finished it. A complete theory — from any position, who wins with best play, and how, no cunning or psychology required, the winning method a matter of holding a hidden balance invariant every turn. Strategy had become mathematics. The proof and the practical weapon wait in Chapter 9; what matters here is the precedent: a game could be solved, closed, known — which raised, faintly and for the first time, a question this book will meet again: what, then, was the human player contributing?

Between 1917 and 1945, psychology claimed the puzzle as an instrument. Wolfgang Köhler's apes stacked crates to reach hanging fruit (1917), and the moment of sudden reorganization he observed needed a name that wasn't "trial and error." Karl Duncker supplied the framework (in German in 1935; in English, posthumously, 1945): insight is the restructuring of a representation — the problem does not change; your description of it does — and its great enemy is functional fixedness, the inability to see past an object's usual role. The uses you know mask the uses you need. Duncker's most famous experiment — a candle, a box of tacks, a bare wall — appears as a puzzle in Chapter 4, so nothing more will be spoiled here. Max Wertheimer's Productive Thinking (1945) completed the movement. After three thousand years of sorting minds, puzzles were now measuring them.

And around 1945, a puzzle swept the wartime scientific world hard enough to become a security joke: given twelve coins, one of them counterfeit — heavier or lighter, unknown which — find it, and its direction, in three weighings on a balance. It devoured so many research-hours in Allied common rooms that colleagues joked about dropping it over Germany to paralyze the enemy's scientists. Freeman Dyson published the definitive analysis in the Mathematical Gazette in 1946. Its deep content was a new way of counting — not coins, but possibilities against answers: how much can three verdicts of a balance possibly distinguish? That is information-theoretic thinking, practiced fluently two years before Claude Shannon gave information its mathematics and its name (1948). The full twelve-coin fight is a boss-room encounter in Chapter 7. Bring the counting.

The Age Closes

Run the arc once more. A French scholar gives puzzles provenance (1612); an encyclopedia gives them memory (1694); the penny press gives them the masses (1880); the masses are publicly defeated by an invariant they cannot see; a showman steals the credit and, more than a century on, a pair of historians take it back; a craftsman proves elegance is a standard, and mathematics spends twelve decades certifying him right; a don shows that logic itself can be played; the crossword makes puzzling a daily meal, with verification baked in; and in the last stretch, a game is solved outright, the "aha" becomes a laboratory measurement, and a coin puzzle teaches physicists to weigh information itself.

The puzzle entered this age as entertainment. It left as an instrument — for testing minds, measuring them, and asking what thinking is. The natural next step was audacious: stop studying the puzzles and start studying the solver. That project — thinking about thinking — is the next chapter, and the twentieth century pursued it with checklists, card-files, magazine columns, plastic cubes, and, at the very end, a riddle aimed at a new kind of traveler at the gate.