The Diphthongs əi and aι in Scottish, Ulster-Scots and Canadian English*

Robert J. Gregg

The development of generative phonology to deal with the phonological component of transformational grammars has in recent years led to many fruitful insights in the analysis and description of synchronic states of languages. It was inevitable, however, that generativists should eventually turn their attention to historical linguistics and seek to re-interpret the older laws of sound change in terms of modern rules, recognizing specifically the effects of the addition of new rules, the extension of rules by the removal of constraints, and the deletion or re-ordering of rules.(1)

If now to the historical dimension we add the geographical, we may finally be able to break out of the post-Saussurean synchronic-diachronic dilemma and emerge with the concept of a four-dimensional, time-space language continuum in closer accord with linguistic reality than earlier scholarly abstractions were. Linguists now also generally accept the idea that their domain should include a — frequently vertical — sociological dimension and recognize that linguistic change may initially have a social motivation and momentum, spreading with variable dynamism through time and space, each change applying separately to each form concerned, each form thus ending up with its own history and its own geography.(2) Finally, psycholinguists, in their study of language acquisition, have claimed that children make generalizations and internalize rules to keep pace with the language as they acquire it, and that these rules are subject to change as the language-learning process develops from phase to phase.(3)

It would seem advisable, therefore, to bear in mind all of these contemporary developments: changes in scope and attitude, expansion of explanatory power, and especially attempts at developing a unified theory to explain the phenomena of linguistic change, if we are to lay the foundations for an explanation of such striking phenomena as the English Vowel Shift or any aspect of it, with all its variabilities through time and space, through social and regional dialects, and across generation gaps.

If, for example, out of the complex set of changes that constitute the English Vowel Shift, we focus our attention on one item, namely:

ME ῑ → ɑι

we find much enlightenment in a systematic study of the different reflexes still to be observed in the various regional forms of English currently spoken. To begin with, it is clear that the diphthongizations of ME ῑ and ū are separate phenomena, though parallel and related. This is demonstrated by the persistence of undiphthongized u: (or some fronted and perhaps shortened reflex thereof) in local dialects north of the river Humber. Clearly diphthongization of ῑ occurred first in time, beginning in the southern parts of England, and spreading northwards till it covered the whole territory and all dialects, as no undiphthongized ME ῑ's remain anywhere.

If, further, we were to judge the situation by the current Standard English of England and the U.S.A. there would be no way of guessing that there had been any transitional stage between ME ῑ as starting point and the modern ɑι reflex. It is thus perhaps not surprising that many American and other phonologists, encountering the phenomenon of the general Canadian (and occasional American) diphthongal alternants ɑι and əι and observing the strictly limited and conditioned distribution of the latter, wish to explain əι as an idiosyncratic, regional 'raising' of an underlying ɑι.

To avoid this limiting viewpoint it is necessary to take a wider overview of the regional types of English, dialectal or standard, especially the more conservative forms that persist in Britain, such as the Scottish and Irish, in which many obviously Elizabethan or earlier features survive.

Many dialects, notably the Anglo-Irish — in Ulster as well as in the South — throw light on the earliest stages of development, in that they have diphthongs of the type əi or ʌi (the latter frequently with some degree of lip-rounding: ɔ̈i, especially in the South) universally, in all positions, with no environmentally conditioned variants. This situation could be covered by a rule stated as follows:

Rule (1) ME ῑ → əi, etc.

A backward time limit is put on these reflexes, as settlement history establishes that these dialects were introduced around 1600 by Elizabethan and early Jacobean planters from England. Living side by side with the 17th century Ulster Anglo-Irish, however, contemporary Jacobean planters from southwest Scotland must have already been using the two diphthongal variants — əi and ɑι — characteristic of present day Ulster-Scots (US) as well as modern Scottish rural dialects.(4) The distribution of the latter diphthongs can be described in terms of environment as follows:

ɑι occurs in hiatus, in final open syllables, and before voiced fricatives and r;(5) əi occurs in all other environments.

In other terms, the older diphthong əi could here still be regarded as the underlying form, but a new rule has been added, namely: (6)

This rule applies nowadays not only to the rural Scottish and US dialects, but to Standard English as spoken in Scotland and the Scottish-settled parts of Ulster (Ulster-Scots Standard), although there have been some sub-regional developments that complicate the relatively simple picture presented above.

In the rural Scottish and US dialects, for example, ME diphthongs of the type ei and ai occasionally survive as diphthongs, in which case they have invariably coalesced phonetically with the əi from ME ῑ, thus giving rise to contrasts such as:

/əi/ always (< ON ei) versus /aι/ I (7)
/bəi/ bay /baι/ buy
/gəi/ very (lit. gay) /gaι/ guy
/məi/ May /maι/ my (8)
/pəi/ pay /paι/ pie
/stəi/ stay/steep /staι/ sty (for pigs)

Other dialectal forms with əi in open syllables include /kləi/ 'clay', /həi/ 'hay', /hwəi/ 'whey', etc. In the Scottish Standard and Ulster-Scots Standard, of course, these contrasts will not arise, since /əi/ 'always', /gəi/ 'very' and /stəi/ 'steep' would be marked as dialectal and therefore excluded, and all the other forms would have e:: be:, me:, pe:, ste: 'stay'.

It will be noted that the application of Rule (2) regularly produces alternations such as:

/fəif/ fife but /faιv/ five
/ləif/ life /laιv/ live (adj.)
/prəis/ price /praιz/ prize, etc.

both in the dialectal and standard speech of the areas in question. Yet in both types of speech there are certain definable sets or subsets of lexical items which have to be marked as not subject to Rule (2). This constraint applies, for example, to the subset of noun class words which in Standard English elsewhere undergo the voicing rule before adding the plural marker, but which have the following Scottish and US forms:

/ləif/ life plural /ləivz/
/nəif/ knife /nəivz/
/wəif/ wife /wəivz/(9)

The same constraint applies to the set of strong verbs, so we find that the weak verbs apply Rule (2): /əráιv/ 'arrive', /daιv/ 'dive',(10) /rəváιz/ 'revise', /praιz/ 'prise', etc; but, on the other hand, the strong verbs do not: /drəiv/ 'drive', /rəiz/ 'rise', /strəiv/ 'strive', /θrəiv/ 'thrive'. This constraint seems to have been lost in some Scottish dialects, but the differential patterning is so strongly established in US that the last verb cited above, namely /θrəiv/, will always retain əi when the past tense is /θro:v/, but with speakers who have moved this verb into the weak class, /θrəiv/ at the same time shifts to /θraιv/ (past tense /θraιvd/).

In all types of US speech the pronominal forms /maι/ my and /ðaι/ thy represent a normal application of Rule (2), but the vocalic nucleus in the derivatives /maιn/ mine and /ðaιn/ thine remains ɑι in spite of its being in a closed syllable whose final consonant is not a voiced fricative. Thus /maιn/ mine (< my) with derivational -n is in contrast with the solid morpheme /məιn/ mine (as in coal mine), where Rule (2) naturally does not apply.

With speakers whose variants əi and ɑι are triggered entirely by the phonological environment it is not surprising if they are unaware of this alternation in their speech pattern. With the Scottish and US rural dialect speaker, however, for whom the actual word forms əi and ɑι invariably convey different meanings, the first being 'always' and the second 'I' or 'yes', and so on for bəi and bɑi etc., the situation is different. These two diphthongs represent two different 'points in the pattern'.(11)

For speakers of the Ulster-Scots Standard the contrast has become even more highly functional as a result of dialect borrowings from the neighbouring Anglo-Irish dialects, especially in the urban setting where the two dialects are in intimate contact and where the prestige attached to Anglo-Irish has had many important sociolinguistic consequences. Thus, it is to be observed that US rural dialect speakers who have moved into an urban area within their own dialect zone have been forced, for the sake of intelligibility, to drop markedly dialectal items or dialectal pronunciations from their speech and replace them with forms borrowed directly from Anglo-Irish.

This affects the incidence and the contrastive possibilities of əi and ɑι very considerably. For example, in spite of Rule (2), which calls for ɑι in open syllables, many monosyllables in Ulster-Scots Standard have əi because, for rural dialectal forms that were unacceptable in urban speech, forms with əi were simply borrowed from Anglo-Irish to replace them. In this way, and entirely for sociolinguistic reasons, we find the following replacement pattern:

Rural US USS (from AI)
/di:/ 'to die' is replaced by /dəi/
/fli:/ 'fly' /fləi/
/hix:/ 'high' /həi/
/i:/ 'eye' /əi/
/li:/ 'lie' (falsehood) /ləi/
/sæ̈x/ 'sigh' /səi/
/θi:/ 'thigh' /θəi/

Even etymologically unrelated replacements may occur in Ulster-Scots Standard:

/ble:t/ 'shy' is replaced by /ʃəi/
/hwʌ̀t fʌ̀r/ 'why' /hwəi/

although the latter form is obviously affected by the minor constraint (not mentioned above) on Rule (2) whereby əi does not go to in any type of US if the preceding segment is w. Thus letter 'y' is called /wəi/, and 'wise' is /wəiz/, even in spite of the final voiced fricative z. The forms /wəiər/ 'wire', /kwəiər/ 'choir'/'quire', /ənkwe̍iər/ 'inquire', /rəkwe̍iər/ 'require', /məgwe̍ər/ 'Maguire', show the same constraint, these latter being conspicuously deviant from the norm whereby the diphthong ɑι always precedes r, but is separated therefrom by a transitional ə, for example: 'dire' /daιər/, 'fire' /faιər/, 'hire' /haιər/, 'lyre' /laιər/, 'mire' /maιər/, 'pyre' /paιər/, 'sire' /saιər/, 'tire' /taιər/ (in both senses, i.e., including British tyre).

These Ulster-Scots Standard monosyllables are now, of course, completely integrated and stabilized in current speech patterns along with the normal, native, US monosyllables with regular ɑι such as /baι/ 'buy'/'by', /kraι/ 'cry', /daι/ 'dye', /fraι/ 'fry', /gaι/ 'guy', /aι/ 'I', /laι/ 'lie' (recline), /maι/ 'my', /paι/ 'pie', /raι/ 'rye'/'wry', /staι/ 'sty', /taι/'tie', /traι/ 'try', /vaι/ 'vie'. It will be noted Ulster-Scots Standard speakers have in this way əi versus ɑi contrasts, different from the US or Scottish rural ones:

/əi/ 'eye' versus /aι/ 'I'
/dəi/ 'die' /daι/ 'dye'
/ləi/ 'lie; to tell a lie' /laι/ 'lie' (recline)

but they also, like the rural speakers, have the contrast:

/məin/ 'mine' (e.g., coal mine) versus /maιn/ 'mine' (belonging to me)

Viewed against these Anglo-Irish, US and Scottish backgrounds with their patterned variations in time and space, based on Rules (1) and (2), producing first a universal əi and then an innovation by which əi → ɑι in certain conditions and with intricate constraints, the patterning of the Canadian English (Canadian) diphthongs əi and ɑι simply calls for the expansion of Rule (2), widening its scope so that it now applies in the environment of all voiced consonants, not simply the voiced fricatives, in addition, of course, to the other environments already specified in Rule (2). This generalized rule would thus read:

This rule will give the same output as the earlier Scottish and US rules with voiced fricatives, hiatus and word final position, but will apply where the earlier rules did not, namely before voiced plosives b, d and g, the nasals m and n, and lateral l. Thus Canadian agrees with Scottish and US in alternations such as:

/fəif/ 'fife' /faιv/ 'five'
/prəis/ 'price' /praιz/ 'prize'
/pəip/ 'pipe' /paι/ 'pie'
/səit/ 'site'/'cite' /saιər/ 'sire'

but disagreements between Canadian and the others could be tabulated as follows:

Scots and US Canadian
/dʒəib/ 'jibe' /dʒaιb/
/təid/ 'tide' /taιd/
/təigər/ 'tiger' /táιgər/
/təim/ 'time' /taιm/
/ləin/ 'line' /laιn/
/məil/ 'mile' /maιl/

A detailed examination of the incidence of əi and ɑι in Canadian as compared with Scottish or US would thus show it to be unlikely that the Canadian əi was evidence for linguistic influence exerted by the large numbers of Scottish and Ulster-Scots immigrants to Canada over the years. What might be claimed at most would be a reinforcement, encouraging the retention of the Canadian əi/ɑι dichotomy, as it would have been relatively easy for the Scottish and US incomers to extend and simplify their rule to cover the same scope as the Canadian with its innovation, that is, to remove constraints so that the whole class of voiced consonants trigger ɑι rather than əi.(12)

As the case of the constraint affecting Scottish or US /ləivz/, /nəivz/, etc., discussed above, so the occurrence of Canadian forms such as /rə̍idər/ 'writer' and /wə̍idər/ 'whiter', arising from the application of the medial voicing rule to the regular, underlying /rə̍itər/ and /wə̍itər/, may best be explained by appeal to the theory of rule ordering. If we assume that there are two rules, one involving voicing and the other the selection of diphthongal variants, then clearly in the Scottish or US output /ləivz/ the choice of diphthong, əi, is determined on the basis of /ləif/ as the underlying form, and the voicing rule applies after the choice of the diphthong.(13) In this instance the Canadian forms /ləif/ plural /laιvz/ shows the opposite ordering. With the Canadian derivatives of /rəit/ 'write' and /wəit/ 'white' on the other hand, the choice of diphthong is made first, with /rə̍itər/ and /wə̍itər/ as the underlying forms. The voicing rule is applied second, so that the output is in contrast with the regular forms /ráιdər/ 'rider' and /wáιdər/ 'wider'. In these examples rule ordering is clearly crucial, and, as is shown by the differing surface forms in Canadian as compared with Scottish and US, the ordering is not predictable and may vary from region to region and also, we presume, from period to period.

In conclusion, then, if our aim is to produce phonological theories and explanations that will embrace all types of English or at least all the major varieties of spoken Standard English, we should not in the instance of the diphthongal forms discussed above start from a standard American or Southern British basis and assume that Canadian or Scottish and US variations are somehow deviant, late derivatives. A simpler explanation is reached by postulating that these latter dichotomous forms are historically older and that the other standard forms represent a simplification by the complete deletion of all contraints on the basic rule that əiɑι, the historical stages being represented in the current geographical distributions as follows:

where Rule (1) is represented by Anglo-Irish, Rule (2) by Scottish and Ulster-Scots, Rule (3) by Canadian, and Rule (4) by Standard Southern British and Standard American. In this perspective then, the occurrence of əi does not represent a 'raising', but rather the occurrence of ɑι represents a 'lowering' which ultimately takes over the whole territory.

It would undoubtedly be best to leave for another article any discussion of the following relevant matters:

— whether əi would be better than the ῑ (14) (suggested by Chomsky & Halle) as the underlying representation of the stressed vowel in 'divine' and 'divinity', the rule əiɑι giving the normal output in dəváin, and dəvιnəti being the output of a trisyllabic laxing rule which involves prior schwa deletion instead of shortening.

— the promising possibility of tying in the diphthongs əi and ɑι with a feature [±length].(15)

— the question of whether ME ῑ and ū were first lowered and then centralized, or vice versa, or lowered to mid height before being centralized.(16)

— the parallel but somewhat different case of ME ūəu in Scottish and US, but ɑω as well as əu in Canadian.


* Originally published in the Canadian Journal of Linguistics 18:2 (1973), 136-145.

(1) King, Robert D. Historical Linguistics and Generative Grammar (Englewood Cliffs, 1969), 39-63.

(2) Malkiel, Yakov, 'Each Word has a History of Its Own', Glossa 1 (1967), 137-149.

(3) King, Robert D., op. cit., 85.

(4) Gregg, Robert J., 'Notes on the Phonology of a County Antrim Scotch-Irish Dialect, Part 1: Synchronic', Orbis 7 (1958), 392-406; Gregg, Robert J., 'Notes on the Phonology of a County Antrim Scotch-Irish Dialect, Part 2: Diachronic', Orbis 8 (1959); Gregg, Robert J., 'Scotch-Irish Urban Speech in Ulster' in Adams, G. Brendan (ed.), Ulster Dialects: An Introductory Symposium (Holywood, 1964), 164-192; Gregg, Robert J., 'The Scotch-Irish Dialect Boundaries in Ulster' in Wakelin, Martyn (ed.), Patterns in the Folk Speech of the British Isles (London, 1972); Gregg, Robert J., 'Synchronic Reflexes of Middle English ῑ in the Scotch-Irish Dialects', Proceedings of the Zweite Internationale Phonologie-Tagung (Vienna, 1972); Grant, William, The Pronunciation of English in Scotland (Cambridge, 1913); Grant, William, and James Main Dixon, Manual of Modern Scots (Cambridge, 1921); Wilson, James, Lowland Scotch (Oxford, 1915); Dieth, Eugen, A Grammar of the Buchan Dialect (Aberdeenshire) (Cambridge, 1932); Jones, Daniel, The Pronunciation of English 4th edition (Cambridge, 1966); Wells, John C., 'A Scots Diphthong and the Feature "Continuant"', Journal of the International Phonetic Association 1 (1971), 1.

(5) It should be noted that ɑι in a final open syllable does not shift to əi when an inflexional morpheme -d is added, e.g., tɑι 'tie', pret. tɑιd 'tied', with which latter form compare the solid morpheme təid, 'tide'.

(6) The class specified in this rule as [+cons +cont +voice] includes all the voiced fricatives as well as r, but not l, the latter being considered as a noncontinuant. (See Chomsky, Noam, and Morris Halle, The Sound Pattern of English [New York, 1968], 317-318; and Wells, op. cit.).

(7) In these rural dialects ɑι 'I' frequently goes to ɑ:, and mɑι 'my' to mɑ:, but this levelling occurs only in the two forms cited.

(8) Ibid.

(9) See below for comment on rule ordering.

(10) The verb dive is always weak in these dialects.

(11) Sapir, Edward, Language (New York, 1949), 56.

(12) This particular phenomenon of rule simplification I have observed in the speech of myself and my family. The shift to the Canadian pattern took place spontaneously and naturally on the part of the children, soon after our move to Canada almost twenty years ago from Northern Ireland, and was therefore established from the beginning in the speech of the grandchildren.

(13) This explanation is valid, even historically, as many Scottish and US dialects (especially the rural ones) still have the forms: nəif 'knife', plural nəifs 'knives', etc. The voicing rule for them has really nothing to do with the ME medial, intervocalic voicing. It is simply a straight replacive change, triggered by the standard language, when the plural form shifts to nəivz, etc. Compare also the aberrant forms mentioned as occurring in Buchan (Dieth, Eugen, A Grammar of the Buchan Dialect [Zurich, 1932], 53), saιθ for saιð 'scythe', twəiz for twəis 'twice', where a voiceless fricative has replaced a voiced and vice versa without altering the previously established diphthong.

(14) Chomsky, Noam, and Morris Halle, op. cit.

(15) Chambers, Jack, 'Canadian Raising', Canadian Journal of Linguistics 18 (1973), 113-135.

(16) Wolfe, Patricia M., Linguistic Change and the Great Vowel Shift in English (Berkeley, 1972), especially Chapter VI.