How to spell fish with the wrong letters

Here is a small piece of vandalism that has delighted schoolchildren for over a century. You can spell the word fish as ghoti. Take the gh from tough, where it sounds like f. Take the o from women, where it sounds like the i in fish. Take the ti from nation, where it sounds like sh. String them together — gh-o-ti — and you have "fish," spelled entirely with the wrong letters, every one of them defensibly borrowed from a real English word.

The joke is usually credited to George Bernard Shaw, the playwright who left money in his will to reform English spelling. He almost certainly never said it — ghoti shows up in a private letter from 1855, when Shaw was not yet born, and was probably invented as a Victorian parlour game (the linguists at Language Log have traced it back past him). But the point survives whoever made it. A language whose letters can be shuffled into that kind of nonsense clearly is not spelling by any rule a learner could follow. So why isn't it? And — the more interesting question — how did it get this way?

One little word, seven sounds

Before the answer, one more exhibit, because it makes the mess impossible to wave away. Read these words aloud and listen to the four letters they share:

• through — rhymes with blue
• though — rhymes with go
• thought — rhymes with sort
• tough — rhymes with stuff
• cough — rhymes with off
• bough — rhymes with now
• thorough — ends in a soft uh

Seven words, one spelling — ough — and seven different sounds. (Add the old-fashioned spelling hiccough, pronounced "hiccup," and you get an eighth.) No one designed this. No committee sat down and decided that -ough should mean eight incompatible things. The linguist Arika Okrent gave her book on exactly these oddities the honest title Highly Irregular: Why Tough, Through, and Dough Don't Rhyme, and her answer, boiled down, is a piece of history rather than a rule. Every one of those spellings was once a faithful record of how the word sounded. Then the sounds changed and the spellings didn't.

The photograph

Think of a spelling as a photograph of a word — a snapshot of how it was said at the moment someone wrote it down. For most of English history the photograph was retaken constantly: scribes spelled by ear, and a single word might appear a dozen ways in a dozen manuscripts. Spelling drifted along with speech because nothing held it still.

Two things then arrived, within a few decades of each other, and pulled in opposite directions.

The first was the Great Vowel Shift — a slow, sweeping change in how English long vowels were pronounced, running roughly from the 1400s to the 1700s. The vowels crept "upward" in the mouth; the highest ones broke into two sounds. The word we now say as "name" had been closer to "nahm-uh"; "see" had been "say"; "time" had been "teem"; "house" had been "hoos." It is the single biggest reason Chaucer, writing around 1400, sounds like a foreign language read aloud while Shakespeare, born a century and a half later, sounds merely old. English changed its voice, gradually and almost completely, across those three hundred years.

The second thing was the printing press. William Caxton set his up near Westminster Abbey in 1476, and print does something scribes never could: it makes copies identical. To print a book you must choose a spelling, and once a printed spelling is common enough, it becomes the spelling — repeated, sold, and read across the whole country. Just as English pronunciation was in the middle of its greatest upheaval, the technology arrived that would freeze the written form in place.

The vowels went on moving for another two centuries. The spellings, pressed into type, stayed where they were. English spelling is a photograph taken mid-stride.

That is the heart of it. Knight is spelled the way it was genuinely said in 1400 — with the k sounded and the gh a throaty rasp, roughly "k-nikht," a cousin of the German Knecht. We kept the picture and lost the sound. Make, time, house: the silent final e and the "wrong" vowels are all fossilised pronunciations, frozen a heartbeat before the Vowel Shift swept the actual speech past them.

Caxton didn't help matters. There was a written standard forming — the so-called Chancery English that the king's clerks had been settling into since the 1430s — but Caxton layered his own habits on top, and the typesetters he hired came from the Low Countries and did not speak English. They set words the Dutch way. That is very likely why gost became ghost, an h smuggled in by a homesick Flemish printer that has stuck to the word for five hundred years.

Every silent letter has an alibi

The frozen photograph explains the vowels. But some of English's strangest spellings are stranger still: letters that were never pronounced, in English or anywhere else. They were added on purpose.

Take the b in debt. English borrowed the word from Old French dette — no b, none ever spoken. Then, in the 1500s and 1600s, scholars who admired Latin decided the word ought to look like its distant Latin ancestor debitum, and so they wrote a silent b into it to show off the pedigree. The same impulse put the unpronounced b in doubt (from Latin dubitare), the p in receipt (from receptus), and — most absurdly — the s in island, which is a native English word that never had a Latin ancestor at all; someone simply thought it resembled isle and added the s by mistake (Merriam-Webster tells the debt story in full). These letters are not errors of the ear. They are footnotes — tiny scholarly annotations about where a word came from, welded permanently into how we write it.

And then there is colonel, which we say "kernel." English couldn't decide between two forms of the borrowed word — the French coronel, with an r, and the more classical colonel, with an l — so in the end it kept the spelling of one and the pronunciation of the other. The written word and the spoken word are, quite literally, two different words wearing each other's clothes.

What this means if you are learning

All of which is a delight if you love words, and a warning if you are trying to learn them. Because it means the one thing you cannot do with an English word — and, to varying degrees, with words in French, in Danish, in any language whose spelling has drifted from its speech — is read it and know how it sounds. The letters are a photograph of a pronunciation that has quietly walked out of the frame.

This is why a word learned only from the page is a word half-learned. You can meet colonel a hundred times in print and still say "col-o-nel" out loud in front of someone, because the spelling never told you the truth. You can know through, though and tough on paper and still trip over them in speech, because the paper was lying to your ear the whole time. The gap between how a word looks and how it lands is exactly where learners fall.

The repair is old and simple: don't let a word live only on the page. Meet it in a voice, and in more than one. Say it aloud yourself, so your mouth learns the shape the spelling hides. Keep it in a real sentence, where its sound is carried along by the words around it, and return to it across days until the said word and the seen word finally fuse into one. Spelling can tell you a word's history; only the ear can tell you the word.

There is a quiet consolation in all of this. English spelling isn't a maze built to trip you — it is a keepsake, a language holding on to a picture of its younger self long after its voice has changed. You don't have to decode the keepsake to speak the language; children who cannot yet spell knight say it perfectly. They learned the sound first, from the world, and met the spelling later as a curiosity. That is the right order, and it is the order vocabla is built around — a word met in the ear and the sentence before it is ever trusted to the page, so that when you finally see ghoti, you can smile at the joke instead of falling for it.