Vowel chain shifts

Why does one vowel drifting out of place drag a whole set of unrelated vowels along with it?

Sound changes are usually described one at a time, as though each vowel had its own private history. The well-documented cases do not look like that. In the Great Vowel Shift the long vowels of English did not drift independently; they moved as a set, and the set ended up rearranged rather than scattered. The same holds of the shift running through the inland northern cities of the United States.

The word chain is a claim, not a description — it says the movements are causally linked. Is that true, or a pattern imposed after the fact on changes that happened to coincide? And if true, what connects one vowel's pronunciation to another's, given that no speaker is tracking the system?

Start with what a vowel physically is. Unlike consonants, vowels have no contact, no closure, no landmark — they are distinguished mainly by where the tongue body sits and how the mouth is shaped, a continuous two-dimensional range with no natural divisions in it. A language does not carve the vowel space at joints; it places a handful of targets in an open field.

Now ask what constrains where those targets sit. A listener has to tell them apart. If two sit close together, the ordinary variability of speech makes some tokens of one land where the other lives, and the contrast starts failing. So there is pressure toward dispersion. André Martinet framed this as margins of security around each vowel, and it is the standard structural account.

Notice this is a pressure, not a rule. It operates through listeners: a misheard token is a token that fails, and over many exchanges the productions that stay recoverable are the ones reproduced.

Now perturb the system. Suppose one vowel starts drifting, for any of various reasons from social imitation to articulatory bias. Two things can happen.

If it drifts toward a neighbour, it crowds it. Tokens of the two start to be confused, and the neighbour's productions that stay furthest from the intruder are the ones that survive, so the neighbour is pushed away into whatever space it has. That is a push chain.

If instead it drifts away, it leaves a region unoccupied. A neighbour spreading into the vacated area no longer risks confusion there, so productions that would previously have failed now succeed, and the neighbour follows. That is a drag chain, or pull chain.

Either way the perturbation propagates, because each vowel's safe territory is defined by its neighbours' positions, and moving one redefines the boundaries of the next. The mechanism needs no coordination and no awareness whatever.

Can we test it rather than admire it? The strongest test is the drag case, which has a natural experiment attached. A merger creates a vacancy at a stroke: two vowels collapse into one and a region that supported a contrast now supports none. The account therefore predicts movement in a merger's vicinity. In North America, communities with the low back merger — those for whom cot and caught sound the same — show a retraction of the vowel of cat and further adjustments behind it, found across Canada and the western United States, tracking the merger's territory rather than any social boundary that would explain it otherwise.

What would refute the whole account? A community in which one vowel moves squarely into a neighbour's region, the neighbour stays put, the two merge, and nothing else moves at all. That is fatal to the dispersion story — and here is the awkward part: it is not rare. Mergers happen constantly and often go uncompensated, which tells us the mechanism is real but weak, one outcome among several rather than a law vowel systems obey.

Three refinements, and one belief worth naming as false outright.

First, on direction. The Great Vowel Shift is the textbook chain, and the conventional dating — broadly the fifteenth through seventeenth centuries — is itself an approximation over a long, uneven process. Its outline is agreed: the long high vowels became diphthongs, so that time, once pronounced with the vowel of modern team, acquired the one it has now, and the long mid vowels raised into the space above them. What is not agreed is which end pulled. Whether the high vowels diphthongised first and dragged the mid vowels up, or the mid vowels raised first and pushed the high vowels out, has been argued for a century and is not settled by the evidence we have, which is spellings and rhymes rather than recordings. The chain is well evidenced; the causal arrow inside it is not.

Second, on scope. The structural story is incomplete on its own. Chain shifts spread through communities socially — led by particular groups, carried along social networks, sometimes accelerating for reasons unrelated to the vowel space. Structure says which movements are available once one is under way; it does not say why the first started or how fast it travels.

Now the false belief. Vowel shifts are widely described, when noticed at all, as sloppiness — younger speakers getting the vowels wrong. Every part of that is a prescriptive judgement rather than a structural observation. The shifts are systematic, they preserve the number of distinctions rather than destroying them, and the vowel each complainant regards as correct is itself the output of an earlier shift. What is genuinely at stake is not care but distinctness: the system tolerates any arrangement of targets in which listeners can still recover which one was aimed at — and when it fails to, the outcome is a merger, which is a loss of contrast, and even that is not a loss of anything a speaker needed to be careful about.

• Why some mergers spread irreversibly while the distinctions they replace never come back.