Cryptograms

The puzzle that taught the world to count letters

Pick a sentence. Decide that every A will be written as Q, every B as F, and so on down a scrambled alphabet only you know. Write the sentence out that way, leaving the spaces and the commas exactly where they were, and hand it to somebody. You have made a cryptogram — and you have also, without meaning to, left them everything they need. The commas are still commas. The one-letter word is still one letter long. The letter you used most is still the letter used most. A substitution cipher hides the alphabet and leaves the language standing, and the language is the part that gives it away.

• 480 quotations and proverbs in the collection, every one hand-checked and attributed
• 16.7 different letters in an average puzzle — of the 26 there are, which is the whole game
• 0 letters that ever stand for themselves — enforced on every cipher, never left to chance

How it plays

Each square shows a cipher letter with a writing line above it. Click a square and type, and your letter lands in every square carrying that same cipher letter at once — because in a substitution cipher the unit you are solving is the letter, not the position. That is what makes a cryptogram cascade: one correct guess repays you five or six squares across the whole quotation, and one wrong guess spreads visible nonsense you can spot and take back.

Beside the board sits the letter key, listing each cipher letter with how many times it appears, commonest first. That list is not decoration; it is the tool. Underneath, the keyboard lights up every plain letter you have already spent, since a substitution is one-to-one and no two cipher letters may claim the same answer — if you assign one twice, both are flagged in amber straight away.

Check marks anything wrong in red and costs nothing. Hint hands over the letter that fills the most squares — free, but a hint-free solve is what sets a best time. Reveal is there when a puzzle beats you, and it voids that solve; the counters on this page only ever count honest ones.

Where the cryptogram comes from

Substitution ciphers are ancient, and for most of their life they were not a game. What changed them into one was the discovery that they leak. Around 850 CE in Baghdad, the philosopher al-Kindi wrote a treatise on deciphering cryptographic messages containing the oldest known description of frequency analysis: count the symbols, rank them, and match the ranking against how often letters appear in ordinary writing in that language. It is the earliest recorded use of statistical inference of any kind, and it quietly ended the era in which a substitution cipher could keep a secret.

Nine centuries later the technique became entertainment, largely through one enthusiastic American. In July 1841, Edgar Allan Poe published A Few Words on Secret Writing in Graham's Magazine and made a rash offer: send in a cipher of the kind described and he would break it. Readers sent hundreds. His cipher columns became one of the magazine's most popular features, he claimed a hundred solutions in six months, and he closed the run with two ciphers credited to a Mr. W. B. Tyler, whom later scholars have suspected of being Poe himself. The second of those two held out for over a century and a half. It finally fell in July 2000 to Gil Broza, a software engineer in Toronto, working for a $2,500 prize backed by Williams College — 159 years after Poe printed it.

“Human ingenuity cannot concoct a cipher which human ingenuity cannot resolve.” — Poe, in that same essay. Play long enough here and the line itself will come up as a puzzle.

The hobby organised itself in 1930, when the American Cryptogram Association was founded on the first of September, drawing several of its members and much of its vocabulary from the National Puzzlers' League. Its journal, The Cryptogram, has been printing challenge ciphers since February 1932 — and it is where the puzzle acquired its two formal names. The distinction matters more than it sounds.

• Aristocrat

  Normal word spacing kept. You can see how long every word is, where the one-letter words sit, and which shapes repeat.

  Everything on this page is an Aristocrat. The spaces stay.

• Patristocrat

  Spacing stripped out, letters run together in blocks. Brutally harder — word shape was most of what you were reading, and now it's gone.

  The ACA's journal still prints them. We don't.

Meanwhile the puzzle went mainstream in newspapers as the Cryptoquote and Cryptoquip from King Features, and as Celebrity Cipher from Andrews McMeel — where most people meet it for the first time.

What the letters really do here

Solvers are handed ETAOIN SHRDLU — the letters of English by frequency — and told to trust it. It is worth checking against the actual puzzles. Here is the textbook order set against the one measured across all 480 quotations in this collection, 24,082 letters in total.

That is 7 of the first nine in exactly the textbook position — close enough to lean on, different enough that you should treat the ranking as a shortlist rather than an answer key. The far end is where the real information is: Z · Q · J · X together account for a rounding error, so a cipher letter appearing once or twice in a quotation is far more likely to be one of those than to be a vowel.

Before writing anything, read the letter key. You won't know which is which yet — but you'll know the shape of the answer, and that's enough to start testing.

Six ways into a cryptogram

Roughly in the order you should try them — and every figure below is counted from the 480 quotations this page actually deals.

• AI Start with the one-letter words English has exactly two in ordinary use, and here they appear in 38% of the quotations. A outnumbers I by roughly 3 to one, so a lone letter is a weighted coin. Two different ones in the same puzzle settles it: you have both, and only have to decide which is which.
• THE Hunt for THE The commonest word in the language, and 5.3% of every word here — 313 appearances across the collection. Confirming it hands you three of the top six letters at once, and after THE, AND, YOU, NOT, ARE the three-letter candidates thin out fast.
• IS Mine the short words generally 47% of the words in these quotations are three letters or shorter, and English keeps a very small stock of them. The two-letter words that turn up most here are IS, TO, OF, IN, IT, BE — short enough to try exhaustively once one letter of the pair is known.
• 'S Follow the apostrophes The most generous punctuation mark in the puzzle, because what follows it is almost never a surprise. Of the 11 here, 100% are followed by S. One apostrophe usually gives you a letter free, and often the word around it too.
• LL Watch for doubles Two identical cipher letters side by side narrow English to a shortlist: here the commonest are LL, EE, OO, SS, TT, and the top three alone are 59% of every double on the page. Never one of Z or Q.
• _H_T Then read the shapes Once four or five letters are placed, stop counting and start reading — a word showing _H_T with a known H and T is a very short list. This is where a cryptogram stops feeling like arithmetic and starts feeling like the sentence talking back. It's the best part; be patient for it.

Why the short ones are the hard ones

Every instinct says a long quotation must be harder work than a short one. In a cryptogram the opposite is true, and it is worth knowing why: you are not fighting the length, you are fighting the silence. A substitution cipher is broken by repetition — the same letter turning up again and again in patterns English permits and forbids — so more text means more evidence. Squeeze the same twenty-odd letters into a forty-letter epigram and each one appears twice instead of six times, the frequency ranking becomes noise, and there is nothing to check a guess against.

So difficulty here is measured rather than declared. Each quotation is scored on how much purchase it gives a solver: how often its letters repeat, whether it has a lone A or I, whether it has an apostrophe or a cluster of short words, and how many of its letters appear exactly once and therefore tell you nothing. The three bands are the terciles of that score — nobody sorted them by length. Then look what fell out.

• Easy 60 letters
• Medium 50 letters
• Hard 40 letters

The hard band is the shortest — 40 letters against easy's 60. Nothing under 32 letters is dealt at all: below that a cryptogram stops being deduction and becomes a guessing game about which famous sentence you happen to be thinking of.

How these puzzles are put together

The cipher is built fresh in your browser each time, so the same quotation dealt twice is a different puzzle. Building it comes down to three small refusals.

1. No letter may stand for itself. The ACA's Aristocrat rule, and a practical one — a fixed point is invisible to you but quietly poisons the "this can't be E, I've spent E" chain every solver reasons along.
2. The alphabet may not come out as a shift. A cipher that happens to be a plain Caesar is breakable by trying 25 offsets rather than by reading English, so those are rejected outright.
3. Starters are chosen by effect, not by count. Handing over a fixed number of letters is wildly uneven — three rare ones might light up four squares while three common ones give away half the quotation. The generator samples possible sets and keeps the one uncovering closest to a fifth of the text, and it never gives you the puzzle's commonest letter: finding that is the classic opening move, and it would be a poor gift to take away.

The quotations themselves are hand-picked and hand-checked, from 126 distinct sources — proverbs from a dozen traditions, and writers from Aesop and Shakespeare to Poe, Austen, Twain, Dickinson and Marie Curie. The rule for the collection is that a line goes in only with wording and an attribution we can stand behind; a great many well-travelled internet quotations were left out on those grounds, since a puzzle whose answer is a misattribution is worse than no puzzle. The build refuses to ship a quotation that is too short to solve, too long to enjoy, or that offers a solver no foothold at all.

On paper

Cryptograms were a printed puzzle for a century before they were a screen one, and they lose nothing on the way back. The pack button above builds a PDF of fresh puzzles with a writing line above every cipher letter, an optional frequency table for solvers who would rather not count by hand, a name-and-date line for classrooms, a large-print option, and an answer key that prints the full letter map as well as the finished quotation — because the map is the part you actually spent the afternoon on. For a teacher, a cryptogram is a stealth lesson: letter patterns, common words, spelling and a little statistics, all disguised as a code to break.

If you're enjoying this

The closest cousin on this site is codewords, which is the same substitution idea poured into a crossword grid — numbers instead of letters, and every entry crossing two others so a guess is tested from several directions at once. If you like the letter-frequency instinct but want it faster, the word guessing game runs entirely on it, and anagrams ask you to rearrange letters rather than decode them. For something quieter, there are crosswords and word searches by the hundred. All free, no account, no daily ration.