Transactions of the Yorkshire Dialect Society xix.42-48

This is the Unicode version, for which you need a Unicode phonetic font and a Unicode-compliant browser. If the phonetic symbols do not display properly, try the SAMPA version.

Our changing pronunciation

John Wells

Professor of Phonetics, University College London

1. Introduction

I begin by thanking the Yorkshire Dialect Society for your kind invitation to address you today. I congratulate you warmly on your anniversary, and find it particularly appropriate that you have chosen to celebrate it here in Saltaire, with its connections with the great nineteenth-century dialectologist Joseph Wright.

As a northerner myself by birth and upbringing, I am always delighted to have an excuse to visit the north of England. I think that in this company I had better keep quiet about which side of the Pennines I hail from; but my late mother was unquestionably a Yorkshirewoman. She was born in Leeds and grew up there, and for the last third of her life she lived in Wensleydale. One of my earliest memories is of visiting relatives in Dent, where as a six-year-old I remember being addressed as tha rather than you by my great-aunts and great-uncle.

My topic today is our changing pronunciation. I want to consider briefly a number of respects in which typical northern pronunciation is revealed as more conservative than Received Pronunciation (RP), our polite standard; and then one or two typical southern characteristics that turn out to be less conservative, more innovative, than RP. Finally I shall present some evidence about ongoing changes in the pronunciation of certain particular words.

2. Conservative northernisms

The first pronunciation characteristic that I want to look at concerns words such as song, hang, ring - words which end in spelling with the letters ng. Now for most of us nowadays these two letters correspond to just a single sound, [ŋ], phonetically classified as a voiced velar nasal. Just as bag comprises three sounds, [bæg], so equally does bang, [bæŋ]. But it was not always so. Until perhaps about 1600, everyone pronounced words of this kind with a plosive, a [g]-sound, after the nasal: [bæŋg]. And of course you can still hear this pronunciation in Birmingham, Stoke, Manchester, Liverpool and even in Sheffield (though not, I think, in Leeds and Bradford).

The innovation of NG coalescence, the change whereby the two letters ng came to correspond to one sound rather than two at the end of a word, was a historical sound change that was resisted in these mostly western parts of the midlands and north, although it caught on everywhere else.

It has interesting consequences from the point of view of the phoneme system of the language, that is to say when we consider how many independent contrastive "sounds" we need to recognize in the language. As long as final postnasal [g] remains, the velar nasal can be considered a positional variant of /n/. Once the loss has taken place, /ŋ/ has to be recognized as a separate phoneme.

Note also what happens when a suffix is attached to a stem ending in historical ng. If we take the word singer, one who sings, we see that for most of us the stem-final [g] was lost just as in sing itself. But in finger, on the other hand, where the [g] is in the middle of a stem, it remains. Hence for most speakers singer [ˈsɪŋə] and finger [ˈfɪŋgə] do not rhyme. But in the local speech of Liverpool and Manchester they do, because in [ˈsɪŋgə] we get just the same sequence as in [ˈfɪŋgə]. The same thing applies if we compare kingly with singly, which rhyme in Manchester but not in London or Los Angeles.

Strangely enough, this matter of the variable [g] has a pretty low social profile. It is not perceived as a crashing local-accent feature which ambitious upwardly-mobile northerners might want to try to modify or eliminate. The next feature I want to look at, though, is generally perceived in exactly that way.

This is the matter of the vowel in words like strut, cup (and saucer), love and so on. As we all know, a typically northern pronunciation of these words is [strʊt, kʊp, lʊv], as opposed to the southern and RP [strʌt, kʌp, lʌv].

Notice, though, that southerners and RP speakers nevertheless have an [ʊ]-sound in items such as put and full, good and cushion. The historical development behind this is that the short u vowel inherited from Middle English underwent a split, perhaps 300 years or so ago. The mechanism of this split is not, I think, thoroughly understood; but the upshot was that in most kinds of modern English full does not rhyme with dull, nor put with cut. In my book Accents of English (Wells 1982) I called this split the FOOT-STRUT split, since FOOT and STRUT respectively were the keywords I used for the lexical sets affected. In terms of the phonemic system, English acquired an extra vowel contrast, leaving it with six stressable short vowels in place of the earlier five.

In the popular speech of the north of England, of course, this split was resisted, and that is why in these parts full and dull, pull and Hull are perfect rhymes, and so are put and cut, foot and strut. There are still only five stressable short vowels. This is something we find everywhere north of a line roughly from the Severn to the Wash, up to the Scottish border.

When we come to people of the kind exemplified by William Hague - a man who sounds very definitely northern in his speech, though considerably up-market from the local working-class accent of where he hails from - we usually get a kind of compromise vowel. It's not a [kʊp] of tea for him, and not a [kʌp] either, but rather some kind of [kəp]. More interesting than the question of the precise shade of vowel used is the question of whether pairs like put and cut rhyme in this kind of speech. Indeed they may well do so, because it's not just cut that has its vowel modified from [ʊ], but put as well. This is the origin of the northernism at which, I'm sorry to say, southerners often laugh: when they hear sugar pronounced as [ˈʃʌgə] or butcher as [ˈbʌtʃə]. This is what results when a northerner who basically has the five-term short-vowel system tries to imitate or acquire the RP/southern six-term system. It is a kind of hypercorrection. Unfortunately, the spelling doesn't help the aspiring northerner, since we find the same vowel letter in cut as in put, and the same letters in flood (RP [flʌd]) as in good [gʊd]. If it impresses people to use [ʌ] in percussion, why does the same adjustment sound ludicrous in cushion? Why are outsiders not impressed when the definitely dialectal [luːk] for look is replaced not by RP-style [lʊk] but by something they take as luck?

Attitudes are different again when we take another characteristic feature of northern pronunciation, namely the use of a short vowel in words such as bath, staff, glass and answer. This reflects yet another instance of northerners standing out against a sound change that took root in the south of England and in RPł the change I call BATH Broadening. For a thousand years or more in the history of English bath and so on had had a short vowel, just as in words like cat and trap. But by three hundred years or so ago London people were lengthening this vowel in the position before [f, θ, s], voiceless fricatives. So bath went from [baθ] to [baːθ]. In due course further changes meant that the long and short vowel qualities came to diverge rather noticeably, so that we now get [bɑːθ] with a very different vowel-sound from cat [kæt].

Some words resisted this change. And new words with the short vowel came into the language. As a result, broad-BATH speakers now have non-rhyming pairs such as gl[ɑː]ss but g[æ]s, c[ɑː]stle but t[æ]ssel, bath but maths, pass but mass, disaster but aster, answer but cancer. And they still haven't altogether made up their minds about plastic, graph, and substantial.

The short-vowel pronunciation in bath is not stigmatized in the north in the way that [ʌ] in cup may be. Of course the north is supported here by American and Canadian English: indeed, from a world perspective it is RP and the south of England that are out of line, not the north.

An matter where we clearly see northern conservatism oppposed to southern innovation is in the long mid vowels exemplified respectively in the keywords face and goat. The long monophthongs we get in many parts of the north - as in [feːs], [goːt], or rather opener versions tending towards [fɛːs, gɔːt] etc. - can be compared with the diphthongizing innovation represented by RP [feɪs, gəʊt] and still more as compared with the wide diphthongs [fʌɪs, gʌʊt], etc., resulting from the diphthong shift that has taken place in the speech of London or Australia.

Some northern speakers retain a formerly widespread contrast between monophthong and diphthong, so that, for example, words such as late and eight do not rhyme, being [leːt] and [Eɪt] respectively; or so that there is a difference between toe [toː] and tow [tɔʊ].

3. Innovative developments in the south

There are many changes which appear to have arisen in, or spread from, the south of England. In particular, a number of developments that have arisen in the last two decades or so are associated also with the rise of what has been dubbed Estuary English - a term coined by Rosewarne, 1984, after the Thames estuary, and implying influence of the southeastern part of England centred on London. From the phonetic point of view, Estuary English is supposed to comprise the middle ground between traditional RP on the one hand and Cockney (London working-class speech) on the other. It is best seen as a variety of Standard English, though spoken with a regional accent, just as Standard English may be expressed in a northern or Scottish or Irish accent. But since London is, as ever, the main source of new fashions, in pronunciation as in everything else, many of the characteristics of Estuary English are being, or are likely to be, gradually incorporated into RP. Estuary English is well described in the popular though well-informed book Do you speak Estuary? (Coggle 1993), with its subtitle 'The new Standard English-how to spot it and speak it'.

The final vowel in words such as happy, coffee, valley was traditionally identified with the /ɪ/ of bit. But many speakers nowadays identify it with the /iː/ of beat. In many recent works (e.g. Wells 1990, Roach 1991) the phonetic symbol /i/ is used, to denote this variable or intermediate quality, thus /ˈhæpi, ˈkɒfi, ˈvæli/. This notation reflects the fact that there is no actual opposition between /ɪ/ and /iː/ in these weak syllables (happy does not become a different word by switching from one vowel to the other); what has happened is a change in the preferred phonetic quality of the weak vowel. If our phonological theory is sufficiently sophisticated to recognize a distinction between a strong vowel system (used typically but not exclusively in stressed syllables) and a weak vowel system (used only in unstressed syllables), then we can place /i/ in the weak system. It is used not only word-finally, but also before a vowel as in happier /ˈhæpiə/, various /ˈveəriəs/, radiate /ˈreɪdieɪt/.

Increasingly in RP words such as fold, goal are said with a back rounded diphthong with a starting point comparable to the [ɒ] of lot. This new diphthong [ɒʊ] is found only before dark /l/, [ɫ], or the vowel that develops from it (see below). Speakers who do this are often quite conscious of the difference between their [ɒʊ] and ordinary [əʊ]. Through a process of morphological regularization, they may extend this to words where /l/ is morpheme-final but followed by a vowel, yielding occasional minimal pairs such as wholly /ˈhɒʊli/ vs. holy /ˈhəʊli/.

Traditionally classified as back and rounded, the vowels /uː/ and /ʊ/ are not only losing their lip-rounding but also ceasing to be very back. Thus spoon, conservatively [spuːn], may now range to a loosely rounded [spʉːn] or even [spɨːn], while good /gʊd/ is often pronounced with a schwa-like quality.

In various environments the consonant /t/ tends to be pronounced as a glottal plosive, [ʔ], rather than as the traditional alveolar [t]. This t-glottalling is by now normal before a following obstruent consonant in a different syllable or word, as in football [ˈfʊʔbɔːl], quite good [ˌkwaɪʔ ˈgʊd]. It is also frequent before a sonorant consonant in the same environment, as in witness [ˈwɪʔnəs], atlas [ˈæʔləs], network [ˈneʔwɜːk], quite wrong [ˌkwaɪʔ ˈrɒŋ]. London's second airport, Gatwick, for me has a careful variant [ˈgætwɪk] and a casual variant [ˈgæʔwɪk]. The phonetic environments for this development now, however, extend to word-final position even when the next word does not begin with a vowel, as in quite easy [ˌkwaɪʔ ˈiːzi], take it off [ˌteɪk ɪʔ ˈɒf], not only [ˌnɒʔ ˈəʊnli], or absolute-final (prepausal) right|| [raɪʔ]. Intervocalically within a word, as in city, water, glottal stops are still regarded as Cockney, thus [ˈsɪʔi, ˈwɔːʔə].

RP is traditionally described as having two main allophones of /l/: clear [l] used before a vowel and dark [ɫ] used elsewhere. It is the dark allophone that is now undergoing a process of l-vocalization (becoming a vowel): ɫ → o. Thus in a word such as milk, traditionally [mɪɫk], the tongue tip may nowadays make no contact at all with the alveolar ridge: instead we have a new kind of diphthong, [mɪok]. Similarly shelf becomes [ʃeof], tables [ˈteɪboz], apple [ˈæpo]. The position where this development is most favoured is adjacent to a labial, as in the latter examples; but it is no longer restricted to this position. When it applies to cases such as middle, little a natural consequence is that the lateral release found in conservative speech ([ˈmɪdɫ, ˈlɪtɫ]) is replaced by an ordinary median release, [ˈmɪdo, ˈlɪto].

English has long had a tendency to convert /tj/ into /tʃ/, /dj/ into /dʒ/. We see this in the history of words such as nature, where the earlier /t/ plus /j/ has long ago been replaced by an affricate, /ˈneɪtʃə/. During the course of the twentieth century this process has continued apace. Jones pronounced actual as /ˈæktuəl/, a variant that nowadays would be perceived as mannered or indeed artificial: we say /ˈæktʃuəl, ˈæktʃl/. For me, perpetual and to graduate have formal, extremely careful forms /pəˈpetjuəl, ˈgrædjeɪt/, but everyday forms /pəˈpetʃuəl, ˈgrædʒueɪt/. These are all words, you will notice, in which the new affricate is followed by a weak (unstressed) vowel. The process of yod coalescence continues to widen its scope, extending now to stressed syllables. This makes Tuesday, conservatively /ˈtjuːzdi/, begin /ˈtʃuːz-/, identical with choose /tʃuːz/. Tune and duke become /tʃuːn, dʒuːk/, and reduce comes to have a second syllable identical with juice. I like to think of the avoidance of this development as a touchstone of RP (as against EE, which clearly accepts it); but I am not sure that this claim can really be maintained.

4. Specific lexical items

The changes we have been discussing up to this point have all been general ones, applying in an environment that can be specified phonetically. There are other changes, though, that are lexically specific: they involve just a single word that has changed its shape. Thus for example nephew, which at the beginning of the century was usually /ˈnevjuː/, is now mostly pronounced /ˈnefjuː/. This is not part of a general trend affecting /v/ between vowels, but something affecting just this word.

My data comes from the survey of pronunciation preferences that I carried out for the Longman Pronunciation Dictionary (Wells 1990). In the dictionary I reported the polling results relating to close on a hundred words in which speakers were known to disagree about the pronunciation. These results were pooled for all respondents. What I have done now is to analyse the results by respondent's age. In some cases this reveals no difference at all between the old and the young: for example, in chrysanthemum the pronunciation with /s/ is preferred over the form with /z/ by a margin of approximately 60% to 40% by all age groups. In other cases there is sharp age grading, such that one can see a clear trend as a newer pronunciation, preferred by the young, comes to predominate over an older form.

Thus in nephew the /f/ form, preferred by 79% of all respondents, proves to be the choice of a mere 51% of those respondents born before 1923, but of as many as 92% of those born since 1962. There is a clear trend line, showing that the /v/ form (which happens to be the one I prefer myself) is due to disappear entirely before very long.

Similarly, the percentage preferring /suːt/ over /sjuːt/ in suit has risen from 47% among the oldest group to 92% among the youngest. In deity /ˈdeɪ-/ (as against /ˈdiː-/) has risen from 40% to 98%. In zebra /ˈzebrə/ is preferred over /ˈziːbrə/ by 65% rising to 96%. In applicable stress on the second syllable is preferred over initial stress by 59% of the oldest, but by 91% of the youngest. In primarily, antepenultimate stress (/praɪˈmerəli/ and the like) is preferred over initial stress (/ˈpraɪmərəli/) by 51% of the oldest but 77% of the youngest.


  • Coggle, Paul, 1993. Do you speak Estuary? London: Bloomsbury.
  • Roach, Peter, 1991. English phonetics and phonology. Cambridge University Press.
  • Rosewarne, David, 1984. Estuary English. Times Educational Supplement, 19 October.
  • Wells, J.C., 1982. Accents of English. Cambridge University Press.
  • Wells, J.C., 1990. Longman Pronunciation Dictionary. Harlow: Longman.