NEAR to a merger but not quite

In the English of southern Britain there are many words with a close, front i-type vowel quality followed by schwa, ə. These can be divided into two groups according to their historical origins.

One group stems historically from a vowel followed in the same syllable by an r-sound (phonetically ɹ) which survives in the spelling but was replaced in pronunciation by ə. Examples include near, beer, weird, Shakespeare and endearment. I’ll refer to this group as ex-rhotic, or the ‘iR’ set.

The other group stems from a sequence of two syllables, the first containing an i-type vowel and the second an unstressed vowel which was reduced to ə. Often there’s a morpheme boundary between the two vowels. Examples include readjust, European, Maria, envious and skier (the r on the end of the agentive suffix on ski-er is irrelevant). I’ll refer to this group as never-rhotic, or the ‘i.V’ set.

The iR forms were monosyllabic in origin and for many speakers have remained so. The i.V forms have always been subject to optional processes of syncope and compression, making them only potentially, and not necessarily, disyllabic (what John Wells has termed ‘varisyllabic’). Daniel Jones pointed out Shakespeare’s potentially disyllabic treatment of i.V forms like envious with these lines from Romeo and Juliet:

Arise | fair sun | and kill | the en | vious moon (-vious 1 syllable)
Be not | her maid | since she | is en | vi-ous (-vious 2 syllables)

Obviously the sounds which stemmed from iR and from i.V were rather similar, both of them exhibiting a movement from a close-front to a mid-central quality. And there have been tendencies towards merging them, ie towards a situation in which infants acquiring them assign them the same cognitive structure.

In fact there seem to have been two tendencies toward merger, one towards treating them all as of the monosyllabic iR type, and one towards treating them all as of the potentially disyllabic i.V type.

The most successful tendency was the latter, especially in the north of England. In Accents of English, John Wells described

disyllabic or varisyllabic pronunciations of the type [fiːə] fear, [biːə] beer, which is best regarded as a phonemic sequence /iːə/ (p. 361)

In this type of accent, almost all iR and i.V forms are merged to the potentially disyllabic i.V structure. As an ‘educated northerner’, I myself am a native speaker of such an accent. To be honest I’ve never been intuitively sure whether beer has one syllable or two, but to me it feels pretty much a rhyme with skier, just as to sear steak feels pretty much the same as to see a steak. The punning headline Who wants to beer millionaire? works perfectly for me (beer = be a).

(The main exception to this in my type of accent is a group of historically i.V forms exemplified by onion, spaniel and William, which now have unstressed jə, eg on[jə]n, span[jə]l, Will[jə]m, etc. These are frequent words with a preceding n or l. Compare the frequent Will[jə]m with the less common Gilli-am, which is trisyllabic. The words in this group usually have a preceding short vowel: compare span[jə]l with mani-a, which is trisyllabic. I can only think of three words with a preceding long vowel: behav[jə], sav[jə] and Californ[jə]. For me the entire remainder of the iR and i.V sets are at least potentially disyllabic.)

This varisyllabic treatment of historic iR forms is not a marginal regionalism, but can be heard in “standard” speech of the dictionary type. In an earlier post I presented the following clips from the Cambridge online dictionary of near~nearly and retire~retirement, to show how NEAR words often behave like WIRE and SOUR words: more disyllabic when strongly accented phrase-finally (eg near, retire), but more monosyllabic when they occur non-finally (eg nearly, retirement).

Here’s a comparable example, a speaker who’s not markedly ‘northern’ introducing the Samsung’s smart-watch, the Gear:

There the word gear is introduced in an apparently disyllabic form ge-ar, then on repetition is compressed (and smoothed) to monosyllabic gɪː. This is quintessential Wellsian compression, just as it applies to WIRE and SOUR words. (John Wells describes RP compression as more likely in frequent words, fast/casual speech and/or non-initial use of a word in discourse. For modern BrE compression, I think prosodic structure plays an important role. Note that in the above example, the uncompressed gear is both an initial use and prosodically phrase-final.)

The conventional RP-type view of NEAR as monosyllabic ɪə and WIRE as disyllabic aɪ.ə doesn’t explain why these two forms should show the similar patterns of compression so audible in the dictionary clips. If a word like gear is basically gɪə, as the RP story goes, then it’s already monosyllabic and hasn’t got anywhere to compress to. Nor can the conventional view explain the musical treatment of iR words as disyllabic, also discussed in the earlier post. WIRE and SOUR are often broken so that their ə is musically separate. The same can also happen to NEAR. For instance, the song I Did What I Did for Maria by Murray and Callander (both from the south of England) includes rhymes like:

And my mind | was so calm | and so cle-ar |
As I took | my revenge | for Mari-a |

This separation of ə can happen to WIRE, SOUR and NEAR, but never to SQUARE. The conventional RP view of NEAR and SQUARE as identically monosyllabic ɪə and eə doesn’t explain this. (The only example I know of a SQUARE word becoming varisyllabic is mayor, which younger speakers have switched from mɛː to mɛjə, presumably a one-off spelling pronunciation influenced by AmE.)

So much for the tendency to merge ex-rhotic iR forms into the i.V set. The converse tendency, transferring historic i.V forms into the ex-rhotic iR set, was a feature of Received Pronunciation, the elite southern British accent of the 20th century; and the merger seems to have increased during the first half or so of that century. Perhaps the most entrenched example is the word idea, which has long been treated as if it were not i-de-a but rather i-dear, giving in RP a 2-syllable word aɪdɪə. The pronunciation of idea as an iR word even extends to some rhotic speakers: many in Scotland and Northern Ireland actually say it as idea[ɹ].

Daniel Jones, the first of the three great UCL codifiers of RP, was particularly likely to treat i.V forms like iR forms if they were stressed. Thus his English Pronouncing Dictionary (11th ed, 1956) recorded mus[ɪə]m, diarrh[ɪə], F[ɪə]t, Kor[ɪə] (I’ve converted the transcription to Gimson’s more familiar system). For Jones then, Fiat was a one-syllable word. On the other hand, unstressed forms of the historic i.V set never merged to iR. Thus for him, words like envious, copier, area, barrier, myriad, idiot, hideous, trivial retained a distinct, and potentially disyllabic, structure. (Jones’s term for unstressed i.V forms was ‘rising diphthongs’, for which he used a notation ĭə, but he considered this distinct from both ɪə and jə, and he certainly conceded its disyllabic potential; see his Outline of English Phonetics, 9th ed, p. 119.)

Moving on to the descriptions of Jones’s successor Gimson (EPD 14th ed, 1977), we find an apparent increase of the RP merger of i.V forms into the iR set. Words of the type just mentioned – envious, copier, area, barrier, myriad, idiot, hideous, obvious, trivial, etc – were all transcribed by Gimson with ɪə. The word envious, for example, is transcribed as if it were en-vierce, env[ɪə]s. Users of Gimson’s EPD had to refer to its Introduction to find an acknowledgement of Jones’s phonetic distinction, which Gimson felt not significant enough to merit a distinction in transcription.

Others went further than Gimson. A milestone in transcriptional dictionaries was the Oxford Advanced Learner’s Dictionary of Current English (1974) in which Jack Windsor Lewis for the first time indicated both BrE and AmE pronunciations. Perhaps motivated by a desire to simplify for the learner, OALDCE merged virtually all i.V forms into iR, as presumably monosyllabic ɪə; near is the only keyword given, with only word-final linking (r) distinguishing some of the historic iR forms. So OALDCE recommends Europ[ɪə]n, auditor[ɪə]m, prev[ɪə]s, accord[ɪə]n, champ[ɪə]n, etc.

These transcriptions can all still be found today, in the online Collins dictionary. Whether such pronunciations were ever common in ‘standard’ speech, I think they would sound distinctly old-fashioned now. (When their online dictionary was in beta, I did email Collins — generally one of my favourite publishers — to point out the quaintness of its transcriptions, but evidently they didn’t take me sɪərɪəslɪ.)

Things change when we come to the last of UCL’s great RP-codifiers, John Wells. In his Longman Pronunciation Dictionary, John didn’t merely apply the brakes on merger, he set about reversing it. So words like museum, Fiat and agreeable, which Jones gave monosyllabic ɪə, are transferred back to potentially disyllabic i.V (in LPD’s transcription system, iː‿ə). Similarly, Gimson’s envious, area, copier, idiot, etc return to explicitly varisyllabic i.V (in LPD’s coding, i‿ə). Likewise, the iR mergers of OALDCE and Collins – European, previous, champion, etc – return to the historic potentially disyllabic family.

True mergers are not neatly reversible. As a comparison, consider the PALM and START vowels, genuinely merged by SB speakers. For such speakers, Parma sounds exactly like Palmer; linking r is equally possible in Kandahar‿insurgency and Omaha‿insurance; and when such speakers attempt to Americanize their speech, they are liable to insert r not only into START words but also, erroneously, into PALM words — oblivious to the spelling which, of course, is taught after the phonological system has been acquired.

So it would be remarkable if SB speakers started to de-merge PALM and START forms consistently. And, to the extent that modern SB speakers differentiate the historic iR and i.V sets, then this suggests that the two sets never merged at all. That is, although there was a fashion among the RP elite to treat i.V words as if they were iR words (cf the RP i-laxing discussed in my preceding post), it seems that others in southern Britain didn’t do the same, and the contemporary standard accent tends to reflect this.

In two respects I wonder if John’s de-merger of iR and i.V goes quite far enough, as far as contemporary speech is concerned. Firstly, there are some historic i.V words like theorem, diarrhoea, Beatrice, which LPD insists are still iR-merged, ie that pronouncing them with varisyllabic iː‿ə is not ‘Received’. I tend to doubt there’s much point today in claiming that such words behave differently from varisyllabic the-ory, IKE-A, pe-ony.

My second point relates to the modern pronunciation of the iR set – real ex-rhotic NEAR words. When I wrote my earlier post on NEAR a year and half ago, I concentrated on the tensing of its first element and on the evidence for its varisyllabicity. Since then, I’ve been struck by the rapid spread of monophthongal ɪː in iR words. This seems to have been spɪːheaded by the super-frequent words here and year, which for many speakers are now hɪː and jɪː in all environments, even when strongly-accented phrase-finally. Here’s Rupert Grint saying Who’d bother crossing the Channel when you can come here?

The new ɪː for NEAR certainly extends beyond here and year. Here’s Daniel Radcliffe saying weird as wɪːd:

Unsurprisingly we also get ɪː in the word idea, which as we’ve seen has long been pronounced as if it were idear, even in some rhotic accents. Here’s the BBC’s political correspondent Ross Hawkins saying some id[ɪː]s:

However, my tentative impression is that the younger speakers who most use ɪː in iR words don’t tend to use it in historic i.V words (with the exception, of course, of idea). And this is my second point about LPD. Its transcription iː‿ə implies potential neutralization with NEAR words – potential merger with the iR set. I’m curious to know how true this is for younger speakers.

As John Wells has described, RP could compress i.V forms like European and Caribbean and make them sound the same as iR forms peer and beer. Similarly classic RP could optionally compress be a to bɪə, so that the pun Who wants to beer millionaire? could be as effective in classic RP as it is in my varisyllabic system (albeit with the merger going in opposite directions). But do the younger speakers who have pɪː for peer and bɪː for beer also say Euro[pɪː]n and Caribb[ɪː]n? Does IKEA really compress to IK[ɪː]? (Excepting very rapid speech, where all bets are off.)

As mentioned in my post on the demise of RP happY-laxing, there has been a trend since the heyday of RP for NEAR to be heard with a first element like FLEECE rather than like KIT. It could be that for many SB speakers this was only a conditioned variant, chiefly restricted to strongly-accented phrase-final position, so that beer came to sound like Caribbean in citation form, while in most other positions ɪə remained and was monophthonged (as has happened with the SQUARE vowel) to ɪː. This would be a matter not so much of compression when weak as of ‘de-compression’ when strong. It would entail no underlying merger of iR and i.V, so that younger speakers could subsequently generalize iR = ɪː to all positions.

A highly pertinent comment comes in an email from Tomasz P. Szynalski, who made a quick comparison of iR and i.V forms in both the citation forms and the illustrative sentences of the Longman Dictionary of Contemporary English (5th ed) and found them overlapping but distinct: i.V never impressionistically monosyllabic in the citation forms, iR never impressionistically disyllabic in the sentences.

Of course we need proper data, properly measured. I’ve discussed the topic with my UCL colleague Professor John Harris, and he’s already set the treatment of iR and i.V by younger SB speakers as a student data project. I’m very keen to see the results, and will keep you posted. (It would also be good to know what’s going on in contemporary northern systems.)

If it turns out for younger SB speakers that beer, near, gear etc have ɪː as their basic pronunciation, and that IKEA, skier, European etc remain resiliently distinct, then it would seem that the historic iR and i.V sets are more separate today than they ever were in descriptions of RP.