Why an apron was once a napron

Say an apron out loud and then a napron, and you will notice something faintly alarming: they are the same sound. The tongue does the identical thing either way — the n sits in the seam between the two words, and nothing in the speaking tells you which word it belongs to. Your ear has to decide, and it decides by habit, and the habit could have gone the other way. It very nearly did. Six hundred years ago the word for the cloth you tie over your front was napron — from Old French naperon, a little nappe, a little tablecloth, cousin to napkin — and people wore a napron. Then, one mishearing at a time, across more mouths than anyone could count, the n slid over the seam. A napron became an apron, and stayed. The garment never changed. Only where the ear cut the string.

You would expect a word to be a solid thing, an object with edges, handed down intact from whoever first coined it — that its spelling might drift but its bones, its consonants, would be safe, because a word is a unit and units have boundaries. The surprising part is that the boundary is the softest thing in it. Speech is a smear; we do not leave gaps between words the way print does. When you hear an-apron the sound arrives as one ribbon, and every listener has to guess, silently and instantly, where one word ends and the next begins. Almost always the guess is right, because almost everyone has heard it before and guesses the same way. But n after the article an is the perfect trap — it can belong to either side and sound the same from both — and at a certain word, in a certain century, enough ears guessed it onto the wrong side that the wrong side became the word. No one decided this. There was no committee, no reform, no first person who declared henceforth, apron. There was only a drift of small identical mistakes that happened to agree, and agreement, past a threshold, is indistinguishable from a rule.

Once you have the trick you find the snake did it too. An adder was a nadder (Old English nǣdre); a nadder lost its n to the article exactly as the napron did. So did the auger, the wood-boring tool, once a nauger. So did the umpire — from Old French nonper, "not equal," the odd third man brought in to break a two-way tie — who arrived in English as a noumpere and was misheard into an umpire, which is a small joke, because his whole function is to settle which side a thing falls on and the language could not settle which side he fell on. And the theft runs both ways. Sometimes the noun steals an n from the article instead of surrendering one. A newt was an ewt (an eft, the old word we still keep for the land stage of the animal); an ewt became a newt, the n migrating the opposite direction across the same seam. A nickname was an ekename — an eke-name, an "also-name," eke being the old word for "also" — and an ekename slid into a nickname. Even the phrase for the nonce is a fossil of this: it was for then anes, "for the one occasion," and the n of then crept forward and fused to anes, leaving a word, nonce, that means nothing at all outside the phrase that mislaid it there. The seam gives and the seam takes, and which way it goes at any given word is not principled. It is just where the crowd's ear happened to cut.

The reason this can happen — the reason it is structural and not a run of flukes — is that no speaker inherits the history of a word, only the word. A child learning apron does not receive a note saying this used to have an n up front, mind the seam. The child receives a sound, in a stream, and reconstructs the word from scratch by cutting the stream where it seems to cut. Every generation re-derives the whole vocabulary this way, from the outside, with no access to how the last generation derived it — and so the only thing that carries a word forward is that the new cut usually matches the old one. Usually. When it doesn't, and when the mismatch is one that sounds the same and so never gets corrected, the word quietly becomes a different word, and the change is invisible to everyone involved because each of them is only ever doing the ordinary thing: hearing a sound and guessing where it breaks.

Two honest edges. The first is that not every a/an oddity is this. Orange really did shed an n on the way in — Persian nārang, Arabic nāranj, and Spanish still says naranja — but exactly how much of that loss is article-rebracketing and how much is other erosion in other languages is genuinely tangled, and anyone who tells you the clean version is tidying. The mechanism is real and the apron and the adder are clean cases; the whole list is not equally clean, and I would rather name the doubt than pad the pattern. The second edge is that it is tempting to call these mistakes, and they were, in the sense that each one crossed a boundary its own speaker would have drawn differently a generation earlier. But there is no court that hears the appeal. A mistake that everyone makes, and that sounds exactly like the thing it replaced, is not overruled by anything — there is nothing above the crowd's ear to overrule it. So it is a mistake only on the way in. Once it has arrived it is simply the word, as much the word as any word ever was, and the napron it displaced is the ghost. The language has no memory of which of its words were born correct. Neither, it turns out, does correct mean very much here — the word is whatever the last ear decided it was, and the last ear had no idea it was deciding.