1 Introduction
This articleFootnote 1 is an attempt to expand on a suggestion I made in Lodge (Reference Lodge2009: 134–5) with regard to the lexical identification of forms of the definite article in the North of England. The discussion revolves round the concept of lexical sameness as between dialect and standard forms, the latter of which also have standard orthographic representation. It is also an attempt to provide a phonological analysis of the forms under consideration that is not segmental, but polysystemic and declarative. The published accounts so far assume a segmental approach to phonology and this is enshrined further in the phonetic transcriptions that are used representing the data to be analysed. It is also the case that the local forms are treated as variants of standard the, but I would not want to refer to this phenomenon as Definite Article Reduction (DAR; as in, for example, Jones Reference Jones2002a and Reference Jones2007, though he makes no claim himself about any such synchronic relationship, using the acronym for convenience). The modern-day [θ]- and [ʔ]-forms are not reductions of anything in a synchronic grammar, so DAR can only refer to a historical process, as yet not fully understood. However, the nature and development of the form of the definite article in various parts of the North of England (see footnote 2) have received relatively little attention, despite the fact that it is a characteristic of millions of speakers. The most recent papers on the phenomenon have been a historical and contemporary survey (Jones Reference Jones2002a), spectrographic investigations (Jones Reference Jones2007) and sociolinguistic (Rupp & Page-Verhoff Reference Rupp and Page-Verhoff2005; Tagliamonte & Roeder Reference Tagliamonte and Roeder2009.). It is clearly a complex issue and has relevance to a number of theoretical questions.
To start with, then, let us consider some of the relevant questions relating to the phenomenon.
(i) What are the phonetic forms?
(ii) What is the phonological distribution of the forms?
(iii) What is the social distribution of the forms?
(iv) What is the historical development? Will this help with a synchronic analysis?
(v) How many lexical storage forms are there? Should any of them be identified with the lexical item normally written the?
(vi) Are they all related historically?
I shall attempt to answer all of these except (iii), so first of all some discussion of the forms is necessary.
2 The forms
In broad, segmental transcriptions the definite article is usually represented as [t], [ʔ] or [θ], or some combination of these. Possible examples would be as in (1)–(3).
(1) [ɪn t mɔːnɪn] in the morning
(2) [ʊndə ʔ maʔ] under the mat
(3) [pʊt ʊp θ ɛːɹɪəɫ] put up the aerial
Similar transcriptions can be found in Orton & Halliday (Reference Orton and Halliday1962–3 = SED (vol. I))Footnote 2 and Lodge (Reference Lodge1984). However, since these are broad transcriptions, a number of issues are ignored. For example, is the [t] in (1) released or unreleased? If the latter, it could also exhibit assimilation of place: [ɪm pð mɔːnɪn]. Also, what is the nature of the glottal constriction: is it a complete closure or a close approximation? Is there associated creaky voice in the articulation, as indicated in some of the transcriptions for Lodge (Reference Lodge1978) and discussed in detail by Jones (Reference Jones2007)? And is there an accompanying oral closure? Whatever other kinds of phonetic realizations occur, at the very least we are dealing with glottal activity and voiceless dental friction. While bearing in mind the limited phonetic information represented in such impressionisticFootnote 3 IPA transcriptions, I shall continue to use them and discuss phonetic details separately, as appropriate.
In terms of distribution there is some variation across speaker groups. In those dialects that have [θ] (associated with Lancashire today, rather than Yorkshire), the distribution is usually [θ] + vowel and [ʔ] + consonant, as in
(4) [θɛːɹɪəɫ] the aerial
(5) [θɑspitɫ] the hospital
(6) [ɪʔɡaːdn] in the garden
(7) [ɡɪvəsʔbɹʊum] Give us the broom.
(8) [avjəɡɑʔʔtaɛm] Have you got the time?
Note that (8) is different from (9) in that the glottal closure is longer in the former, and native speakers are sensitive to such timing differences.
(9) [avjəɡɑʔtaɛm] Have you got time?
It also occurs utterance-initially as a closed glottis onset to even a voiced stop, as in (10).Footnote 4 (An alternative form of bus is [bʊz].)
(10) [ʔbʊsɪzkʊmɪn] The bus is coming.
In this case we are dealing with a stop mechanism during which the vocal cords change position; so, when the lips are closed together the glottis is closed, too, but during the closure the vocal cords start to vibrate so that the release is voiced. And there may even be both glottal and fricative realizations in some contexts:
(11) [wɪʔθɛːɹɪəɫ] with the aerial
(12) [ɪnʔθɑspɪtɫ] in the hospital
In such cases the glottal closure may be epenthetic, especially when there is a preceding consonant, as in (12).
For some older speakers the distribution is [ʔ] + consonant, [t] + vowel with no fricative realizations. There are yet other speakers who use some kind of glottal constriction in all cases, irrespective of the following context.
In the SED there are further transcriptions, namely [
], some being more common than others, but there is no attempt at a phonological analysis or a distributional statement, except the listings of each environment in Upton, Parry & Widdowson (Reference Upton, Parry and Widdowson1994: 480–1), which contains the following assertion: ‘[t] [is] the only non-Standard English form of the definite article recorded in all responses’ to the relevant questions. Hardly an adequate interpretation of even the transcriptions in the SED.
In addition to the dialectal forms many speakers use standard English [ðɪ]/[ðə] forms. For example, in Lodge (Reference Lodge1984: 36) speaker N makes the following utterances in close succession:
(13) [dæʊnʔsɛllə] down the cellar
[teːksθɛːɹeəɫ] takes the aerial
[wɛːjəɡɛʔðeɛːɹeəɫ] where you get the aerial
(The lengthened lateral in cellar is not a normal part of his phonology, but the relatively low realizations of the short vowels, e.g. [e] in aerial and the + vowel, are.) So there is considerable variation, even in one and the same speaker, which needs to be analysed in any phonological analysis. The assumption of a grammatical item that can be identified somehow with segmentable bits of the speech continuum hinders our understanding of the phonological structure of the dialects in question. I shall return to this in section 4.
3 Historical development
This has been fairly fully discussed by Jones (Reference Jones2002a) and references therein, but there are a number of extra points to make. The main question is whether the forms come from the ME definite article [θe] or the demonstrative [θat]. It is, of course, possible that it is a mixture of the two forms. The arguments surrounding this issue involve assimilation: is the [t/ʔ] form an assimilated form in the context of a preceding [t] or [d] extended to other contexts, or is it from the final consonant of [θat], with subsequent glottal reinforcement and dearticulation? But there is also the fact that dissimilation is involved in some cases. There is evidence that the ME t-form appeared after [s] as well as [t] and [d]. This can hardly be assimilation. There is a long-standing dissimilation process whereby the sequence fricative + [θ] became fricative + [t]; hence width, length, but height, thought (noun; OE geþoht), which in OE and ME had a velar fricative, of course. (There are also dialect forms with a ‘double suffix’: [(h)aɪtθ].) This dissimilation also affects other [θ]-initial words, e.g. thou, especially when it follows the verb ending [s], as in modern dialect forms such as [dʊstə], does thou. (Note that the [t]-ending of archaic standard comest, doest, livest, etc. comes from an epenthetic form where the verb occurs before thou with accompanying reanalysis of the word-boundary; cf. Hogg Reference Hogg1992: 149–50. The normal northern dialect ending for the second person singular is, however, s.) And it also affects other suffixes, such as -eth with a loss of the vowel, hence OE ciest [tʃiəst] ‘chooseth’, and the ordinal number suffix þ/t, as in feorþa ‘fourth’ but fifta ‘fifth’, which explains later (fifteenth- and sixteenth-century) forms such as fift and sixt for fifth and sixth. In fact, there is evidence from dialect writing of the eighteenth century that these generalized alternations with assimilation and dissimilation were still current at that time (see the discussion of Tim Bobbin below).
The changes in the distribution of the various forms are difficult to determine, but they seem to involve a redistribution of ‘allophonic’ variants and lexical reinterpretation. A consideration of dialect representations in literature may go some way towards shedding some light on the variation at least and suggest some of the development between ME and the present day. Of course, it has to be remembered that authors are not necessarily concerned with authentic, detailed representation of dialect forms; indeed, it could be argued that it does not help the general reader unacquainted with the dialect(s) in question to understand the speech of characters, if the dialect representation is too broad and distant from the conventions of standard orthography. (For a detailed discussion of this issue, see Page Reference Page1973; see also Wales Reference Wales2006.) So, if we are to look at written sources for additional data, we need to know something of the interests and background of the authors we choose. Dickens, for example, wanted to indicate non-standard speech forms in a general way, choosing popularly recognized forms for this purpose. On the other hand, Elizabeth Gaskell, who lived in Manchester for about twenty years, was married to a man who lectured on philology and dialects, so would have been aware of many of the details of the northern speech she often represented. Emily Brontë used broad dialect forms in the speech of the servant Joseph in Wuthering Heights, so much so that her sister, Charlotte, edited the text before the appearance of the posthumous second edition. And then there are the dialect writers, who tell a story completely in local dialect for a local market, such as Tim Bobbin, the pseudonym of John Collier (1708–86), a school teacher from the Rochdale area, though originally from Urmston near the Lancashire/Cheshire border. In such cases authenticity is striven for.
The first cited example of an orthographic representation of the forms under consideration is George Meriton's A Yorkshire dialogue of 1673/1683 (Cawley Reference Cawley1959; see, for example, Jones Reference Jones2002a and Wales Reference Wales2006), which uses t’, th’ and zero, and is claimed to be the earliest instance of vowel-less, non-fricative forms of the definite article in writing (Jones Reference Jones2002a: 328). Before that th’ is used only before vowel-initial words. Around the same time Thomas Shadwell wrote and published his play The Lancashire Witches, first performed in 1681, published in 1682 (see Summers Reference Summers1927). Though not a local, Shadwell, who was born in Norfolk, goes to some length to represent the speech of a number of characters: Clod, a servant, Thomas o Georges, a witness at the witch trial, and two drunken ne'er-do-wells from higher up in society. There are quite consistent indicators of dialect, e.g. who (for hoo = she), neeght (night, and similar gh-words), caw'n (call, plural verb), pleck (place). The definite article is always th’, but it occurs before both vowels and consonants, e.g. th'leeghts. There is one instance of dissimilation, but not in the definite article; what's ’tiss (what's this?).
I will now look at some further examples from Bobbin, Brontë, Ainsworth, Gaskell and another Manchester author, Mrs G. Linnaeus Banks.
3.1 Tim Bobbin
In his story Tummus and Meary, a comic dialogue in eighteenth-century dialect (first published in 1746; I am using an edition from 1821), Collier is clearly making a conscious effort to represent in writing what the dialect sounds like. Although there are inconsistencies, the following patterns appear with respect to the definite article and other th-words.
(i) The commonest pattern is th’ for the definite article. This is most likely to represent the voiceless fricative, given the modern variants.
(ii) Occasionally the occurs, especially in the dule (devil).
(iii) t’ occurs after coronal stops: sed t'felley (fellow); whot te plague.
(iv) The assimilation after [t/d] and dissimilation after [s] are not lexically restricted but apply to other th- words: mend tat (that), ot tis seme brig (at this same bridge), whot tey can, howd teh tung (hold thy tongue); as teaw (as thou), whooas tat? (Who's that?).
(v) There is one instance with no indication of the article: swallut poker (swallowed the poker). This could be a misprint or an indication that the pronunciation was [-ttp-] with assimilation and no schwa.
(vi) T'other always has t’ and never th’, which may suggest the origin (tha)t other (cf. Jones Reference Jones2002a: 326, referring to Barry Reference Barry and Wakelin1972).
This suggests that the phonological form of the article is /θ/ (with the possibility of an additional schwa in set phrases such as oaths?) and assimilation and dissimilation are generalized. This particular representation of the dialect was a serious attempt by Collier to exemplify and support it against criticisms of a prescriptive, standardizing nature, and demonstrates many features that had disappeared from standard and many other kinds of English by that period, for example, the plural verb marker -(e)n, retention of final /x/ (though not before /t/, e.g. reet (right), but saigh (saw)), hoo for she, ME /e:/ and /ɛ:/ kept distinct. They are also apparent in the seventeenth-century representations referred to above. So there is every reason to accept that it is a relatively accurate representation.Footnote 5
3.2 Emily Brontë
Another author who wished to reflect the local dialect accurately, West Riding of Yorkshire in this case, was Emily Brontë in Wuthering Heights (Reference Brontë1847). In this case it was for reasons of characterization in a novel otherwise written in standard English. The speech of the servant, Joseph, (amongst others) is intended to be broad and local. In most of the novel the pattern of the definite article is t’ + consonant, th’ + vowel and h, presumably indicating h-drop. Oddly, given the consistency of representation, there are examples of th’ + consonant from the same character. Whatever the reasons for this, it seems that 100 years later the features of assimilation and dissimilation found in Tim Bobbin have gone, except in the case of the single lexical item thou, e.g. wud tey. Of course, there is also a geographical difference in this case, Yorkshire rather than Lancashire, and there are other distinguishing features in the dialect forms, e.g. shoo for she, and no plural verb endings, but the non-appearance of general assimilation and dissimilation is significant. In this case, too, care has been taken by the author, albeit with some inconsistency, to represent the local dialect features in great detail (hence Charlotte's later editing). Clearly this represents a different kind of phonology with a regular alternation of forms of the definite article only.
3.3 William Harrison Ainsworth
Ainsworth was born in Manchester, though he later moved to London, and wrote several romantic novels, including The Lancashire Witches in Reference Ainsworth1849, which is based on the same story as Shadwell's play, but is a considerably longer text. He has several characters who speak dialect and there is variation in their use of the form of the definite article. Although the story is set in the sixteenth and seventeenth centuries, it is likely that the language represented is that of the nineteenth century. It is difficult to tell whether the differential use of the forms has any literary relevance in terms of designating particular traits of the characters, or whether there is any linguistically relevant distribution. Certainly in the introduction to the story, set in the 1530s around the time of the northern rebellions against Henry VIII known as the Pilgrimage of Grace (November 1536), the form is t’ before both vowels and consonants and there is evidence of assimilation and dissimilation of thou/thee/thy, as in (14).
(14)
But in the rest of the story, set in the early seventeenth century, there is more variation, sometimes in one and the same speaker, between th’ and t’, both before vowels and consonants. So we find, for example, buried th'ess [ash] at th'change o’ t'moon. There seems to be no relevance of the preceding environment in many cases either, so we find at th'feet, o’ th'furst (at the first), i’ t'cowld from Jennet, a young witch, who also uses the in many places. There is one instance of the earlier assimilation in howd te tongue, which could be seen as a set phrase (see also footnote 5). It would appear that variation between th’ and t’ reflected something of the rural dialect of the mid nineteenth century. Of course, we find same-speaker variation today, but it is unlikely that attention to such detail would have been a concern to Ainsworth (as opposed to Bobbin). Whereas his representations of Lancashire dialect are authentic enough, this is clearly not his focus; the representation of King James I's speech later in the novel is clearly a literary, conventional ‘Scots’.
3.4 Elizabeth Gaskell and G. Linnaeus Banks
Elizabeth Gaskell in Mary Barton (Reference Gaskell1848) and North and South (Reference Gaskell1855) was only concerned with fairly general indications of northern dialects. One of these features was, however, the definite article, invariably written th’. T’ is restricted to forms representing schwaless to. This may have become a literary convention for representing ‘northern’ (or, perhaps, more specifically, Manchester) speech. It is used later by Mrs G. Linnaeus Banks in her representations of Manchester speech of the time of the Peterloo Massacre in her novel The Manchester Man (c.Reference Linnaeus Banks1875). This would suggest a dialect with [θ] in all cases and no alternation, again without any assimilation and dissimilation. It suggests that an urban dialect had developed by the mid nineteenth century with levelled-out features. Certainly, still in the mid twentieth century there were deliberately old-fashioned recitations and poems in broad dialect with [θ] + consonant, as in (15), which reflects some of the SED material.
(15) [wɪə ɹɪit dæʊn ɪn θ kɔɪɫɔɪɫ]
We're right down in the coal-hole
3.5 The main changes and a suggested explanation
Between 1750 and 1850 there seems to have been a change in the phonological system of at least some of the dialects of Lancashire and Yorkshire. The system in mid-eighteenth-century Rochdale and surrounding area was that described in section 3.1 above: the definite article was in most environments a voiceless dental fricative and was subject to general assimilation to and dissimilation from certain preceding sounds. This reflects the situation with regard to any [θ] as it was in Old English and presumably it continued so at least in the North of England. This system was replaced by a range of systems, one in which there were stop/fricative alternations determined by the following environment, a fricative-everywhere system, or a stop-only system. By the twentieth century there were also glottal stop-only systems. In all cases there has been a generalization of what were assimilated forms or the lexical form to other and even all environments.
How did this restructuring come about? It is difficult to see how a straightforward language-internal historical development could account for this change. The details are unlikely to be reconstructable but in general terms population movement and mixing seem to be likely factors, as far as the urban dialects are concerned. During the Industrial Revolution, from about 1750 to 1850, population migration from rural to urban locations was continuous. To take a major modern city, Manchester, the increase in population from 1700 onwards was spectacular. In 1700 it was c.9,000 people; by 1800 it was 89,000; it had doubled by 1820 and doubled again by the census of 1851. Its population then was something under 360,000. Until the census of 1851 the origins of the incoming population are unknown, though it is usually assumed to be from not too far away.Footnote 6 There are, of course, exceptions, such as the large influx of Irish immigrants in the 1840s. So it is reasonable to assume that Manchester speakers had a Lancashire base with additional elements in varying quantities from further afield. This is, in fact, a new-dialect situation, such as those that have been studied in detail, e.g. Milton Keynes (Kerswill Reference Kerswill, Melchers and Johannesson1994, Reference Kerswill, Graddol, Swann and Leith1996; Kerswill & Williams Reference Kerswill and Williams2000) and New Zealand (Trudgill Reference Trudgill2004). Without any firm knowledge of all the contributory linguistic elements it is difficult to give any details of how the grammatical changes came about, but I would like to suggest one possible contributory factor.
Irish speakers of English in the nineteenth century would presumably have had one phonological feature the same as today: the use of [
] (and [
]) rather than [θ] (and [ð]). Today many speakers from Liverpool, another city with Irish input, use alveolar [d] rather than [ð] in word-initial position, often affricated. If the definite article in Manchester was [θ] (usually without any schwa), then Irish speakers at least would replace this with [
]. Over the generations this could have become the target for some speakers, and a shift to alveolar from dental contact would bring it in line with current articulations. By the early twentieth century, when glottal reinforcement seems to be on the increase in various British English accents, the glottal closure would have occurred in the appropriate environments, especially before consonants. Dearticulation to [ʔ] began to occur fairly soon thereafter, though its spread to all environments took several decades (see, for example, the generational differences in this regard discussed in Lodge Reference Lodge1984: 22). This glottal reinforcement and dearticulation applied to all instances of lexical /t/ in the appropriate contexts, not just the definite article.
It has been pointed out by both referees that [t]-forms were already in place in the Lancashire area as a result of assimilation and dissimilation of [θ], and that the SED shows a distribution in which stops are less common as one moves westwards. (For a detailed discussion of the distribution of forms in the SED material, see Jones Reference Jones2002a.) Neither of these facts can be disputed. However, I am not claiming that the sole origin of [t]-forms in Manchester is due to Irish influence, but such forms would be reinforced as a target by the additional imported forms. As regards the SED material, this does not reflect the twenty-first-century nor the seventeenth-century distributions and it is ill-advised to equate earlier written records with even mid-twentieth-century distributions, as pointed out by Jones (Reference Jones2002a: 341). My discussion of written forms has been an attempt to tease out some of the details of the intervening periods. So, I am not suggesting that all speakers suddenly stopped using fricatives, otherwise there would be no variation today, and indeed the retention of [θ] by some speakers is significant here, but I am merely putting this forward as one of a number of possible changes that affected some of the population to a sufficient extent that it spread to more people as a grammatical target. Similarly, if there were glottal forms back in the seventeenth century, dearticulation would reinforce such forms as a target in further environments.
The spread of a context-specific form to other environments is common enough and is often apparent in child acquisition. For instance, many English children generalize one of the assimilated forms of /ð/ to other word-initial positions, e.g. [nɪs ɪz naɪs] this is nice; [a wþnʔ sʌm ə laʔ] I want some of that. So, a levelling-out of, for instance, the assimilated/dissimilated form to all environments does not seem an unreasonable development to postulate.
There are, however, some aspects of the changes that are explicable in terms of language-internal development. For instance, some of the supraglottal and glottal stop realizations can be seen as epenthetic, e.g. [ɪntθɑspɪtɫ], [ɪnʔθɑspɪtɫ]. In such cases, despite the segmental transcriptions of the kind found in some of the SED material, the stop phase is a retiming of the raising of the velum during the alveolar or dental contact and it is likely that the glottal closure occurs during the supraglottal closure, i.e. [ɪnʔtθɑspɪtɫ] for the latter example. (In the SED transcriptions glottal reinforcement is usually represented with a liaison mark: [
] or [
].) Also, any [θ] following an alveolar contact could assimilate to [t], as in [ɪntjaːd] in the yard, [ɔːɫttaɛm] all the time, in line with the assimilation after /t/ and /d/ referred to above. These would be further influences on the development of the system of stop realizations of the definite article. Likewise glottal reinforcement and dearticulation are seen as general, language-internal developments. Exactly when they first started to operate is difficult to determine.
4 The grammars of today
Any attempt to treat the glottal forms of the definite article as a separate segment, whether an allophone of /t/ or a separate phoneme in a monosystemic analysis of the phonology, will run into a number of problems. If it is a separate phoneme, it is a very odd one, since it is restricted to one lexical item only. Of course, if an appeal is made to biuniqueness (for a discussion of the role of this concept in phonological theory and its relationship to monosystemicity, see Lodge Reference Lodge2009), then all other instances of glottal closure will be classified as belonging together as an allophone of the same phoneme. Putting aside the fact that in some of the modern accents under consideration [ʔ] alternates with [k] as in like/liking, respectively, the glottal stop alternates with [t] at the end of morphemes, as in light/lighting, respectively. This would then mean that article [ʔ] was phonemically /t/ and yet [ʔ] and [th] are in contrast in syllable-initial position: [ʔɛn] the hen: [thɛn] ten. Furthermore, in those accents where the article is subject to [ʔ] ~ [θ] alternations, it is not clear how to resolve the conflict between alternating [ʔ] ~ [θ] in the article and [ʔ] ~ [t] in morpheme-final position in a monosystemic approach based on biuniqueness: how can the glottal stop belong to two different phonemes (three, if [ʔ] ~ [k] is included)?
So, the kind of phonology that I envisage is based on that presented in Lodge (Reference Lodge2003, Reference Lodge2007, Reference Lodge2009); it is non-segmental, polysystemic and declarative.Footnote 7 Phonological forms are represented as attribute-value matrices (AVMs) or equivalent tree-diagrams. Phonological features are monovalent and relate to particular phonetic parameters such as phonation or place of articulation. They can attach to any level of phonological structure, e.g. foot, syllable, onset, rhyme. Phonological structure is layered, one for each parameter, giving the general structure to the AVMs such that each layer is an attribute of the structure and each mutually exclusive feature on that layer is a value. This is represented in (16) from Lodge (Reference Lodge2009: 103) with added phonation values.
(16) lips: {spread, neutral, round}
resonance: {front, retracted, central, advanced, back}
pharynx: {wide, normal, narrow}
velum: {oral, nasal}
phonation: {voiced, voiceless, closed glottis, breathy, creaky}
nucleus manner: {high, mid, low}
manner: {stop, fricative, lateral, trill, approximant, high, mid, low}
place: {bilabial, labiodental, alveolar, palato-alveolar, dorsal, uvular}
Although some features under some circumstances may be implicitly interpretable as phonetic features, there has to be a set of specific statements relating phonological structures to utterances, the phonetic interpretation component (cf. Lodge Reference Lodge2009: 107–9).
Syllable structure is assumed to be as in (17), based on the principle of headedness (see Anderson & Ewen Reference Anderson and Ewen1987; Ogden Reference Ogden1999; Lodge Reference Lodge2009).
(17)
Another issue that I raised in the Introduction above is the notion of panlectal grammars (see Lodge Reference Lodge2009: 120–42). It is not difficult to demonstrate that native speakers of language X are not aware of the grammars of other variants (see, in particular, Trudgill Reference Trudgill1983). That there may be considerable overlap of grammatical knowledge in some instances is not in dispute, but a suitable starting point is an investigation of separate grammars to account for any observed variation rather than an a priori assumption of panlectalism. Today there are basically three grammars as regards the definite article (in addition to the possible borrowing of standard forms of the), which I will deal with in turn. In all three types of grammar the definite article is part of the first onset of the NP, whether the word involved is an intensifier, an adjective or a noun. Whether this means it should be viewed as a clitic rather than a separate word is not an issue I wish to address here.
4.1 [θ] + vowel, [ʔ] + consonant
In a declarative account there are no statements that change features (derivational rules). So in no sense can a feature array for [θ] be transformed into an equivalent for [ʔ] (or vice versa). Lexical entry forms are highly underspecified so that in many cases alternation of realization can be captured in terms of supplying the appropriate feature(s) in the relevant phonological environments, as in the case of coda obstruents in German, for example, which are lexically specified as [voiceless], if they are non-alternating, while in the alternating cases they are unspecified. The latter are subject to the same voicelessness constraint as the non-alternating ones in coda position, whereas in ambisyllabic position they are subject to the general default statement in (18)
(18) PHONATION → [voiced]
This is to be interpreted as: ‘If the phonation layer is unspecified, supply the feature [voiced].’ (For a more detailed discussion of German obstruents, see Lodge Reference Lodge2009.)
But in this case [θ] and [ʔ] have nothing in common except orality. Now it would be possible to use [oral] as the lexically specified feature for the definite article, except that there are environments where nasality goes through the glottal closure, as in (19).
(19) [ɪmʔmɔːnɪn]
in the morning
Of course, it makes little difference whether the velum is lowered or closed for the duration of a glottal stop, but it would require a somewhat perverse phonetic interpretation of [oral], if we had a phonetic interpretation statement to the effect that in the case of a glottal closure surrounded by nasality [oral] is implemented with a lowered velum! However, the particular alternation under consideration is lexically specific, so we are losing nothing by having an ad hoc phonetic implementation of a single morphemic feature, namely, definiteness in a NP. The feature [definite] attached to a NP will be interpreted as an onset voiceless dental fricative in any case where the first word, irrespective of grammatical category, is vowel-initial. In all other cases it will be interpreted as a glottal closure. The glottal closure is represented on the phonation layer as [closed glottis]. In the case of the onset feature it will appear with a following oblique to indicate that it shares the onset position with another phonation feature, [voiced] or [voiceless]. (For similar use of the oblique in phonological representations, see Lodge Reference Lodge2007 and Reference Lodge2009.) So definiteness in a NP with a consonantal onset will be represented as in (20).
(20) [closed glottis/]
The representation of the phonation of the initial consonant of the pot will be as in (21), of the bus, as in (22). (The higher node in each case is N.)
(21)
(22)
If the analysis of onset consonants presented in Lodge (Reference Lodge1992: 29) is appropriate for these accents too, then [voiced] is not specified lexically, but is the default specification of onset phonation.
This treatment does not interpret onset [ʔ] as a segment; it is the realization on the phonation layer of part of the onset syllable place. The duration of the different phonation features and their synchronization vis-à-vis the other onset features, e.g. the oral closure, will be accounted for in the phonetic interpretation component. (I shall discuss the issue of creak in section 4.4.) We should note that this bit of the phonology is specific to one lexical item only and is not generalizable.
4.2 [θ] only
The [θ]-only dialects may be extinct by today, though they certainly existed between fifty and a hundred years ago. In this case the lexical entry form is straightforward as /θ/. (The use of such IPA symbols within slant lines is not intended to signify a fully specified segment.) Which features are specified at the lexical entry level depends on what can be predicted and what cannot in the system of onsets. If place is constant in the realizations of this type of dialect, then [dental] is sufficient specification. The other features will be determined by means of predictive statements such as (23).
(23) [dental] → [fricative]
This is interpreted as: ‘If [dental] is lexically specified, then [fricative] automatically occurs as a predicted feature.’ Lip position and resonance are determined (and represented) at syllable level. The phonation feature [voiceless] would be specified lexically in other lexical items (cf. Lodge Reference Lodge1992: 29), but since dental friction before another consonant can only be voiceless in this type of dialect, it is not necessary to specify it in this case. Note that the phonotactics of onsets in these dialects is different from what is found in most other dialects.
4.3 [ʔ] only
As with [θ]-only dialects, [ʔ]-only can be treated as (part of) the onset of the first word of a NP. It will be represented lexically as [closed glottis] and be attached to consonants as described in section 4.1 above. Before vowels the onset will be just [closed glottis] with no shared phonation feature. Note that in this case, as in the grammar under section 4.1, the phonology is specific to the definite article only.
In (24) I give the lexical entry forms for a number of definite article + noun onset combinations.
(24)
4.4 Creak
We need to consider one aspect of what we have been symbolizing as [ʔ] that is dealt with in detail by Jones (Reference Jones2007: 69–72): creaky voice. Creak may be used exclusively by some speakers as the realization of the definite article during the offset of the preceding vocoid articulation, in which case [creaky] can be used in the representation instead of [closed glottis], as in (25).
(25)
The duration of creak is part of the phonetic implementation, along with its location as part of the preceding vocoid articulation. Jones (Reference Jones2007) looks at controlled instances of -V(/)s- only, so the extent to which creak is a consistent realization of the definite article only needs further investigation as well as its occurrence in preceding contoid articulations. In the case of the teenage informant for Lodge (Reference Lodge1978) there are examples of creak following the glottal closure and in the vicinity of any [ʔ], that is the realization of what is usually phonemicized as medial or final /t/ in any word, as well as the definite article. So creak, at least for some speakers, may be an effect of any glottal closure/approximation (even lexical /k/, as in like, for example).
4.5 Standard ‘the’ and its relationship to the northern forms
In sections 4.1–4.3 I have assumed that each variety has its own grammar, at least as far as the definite article is concerned. What I have not assumed is a single underlying northern phonology; nor do I see the forms with initial [ð] as part of these grammars. That (some of) the forms are historically related to [ð]-initial forms is not in question, only the assumption that all these forms are necessarily to be derived from a form represented in standard orthography as the in a synchronic grammar. It is also the case, as we have seen, that many northern speakers use [ð]-forms alongside the local variants. Why this should be and under what circumstances has to be considered in a wide-scale, sociolinguistic study and is not the focus of this article. However, as a general mechanism, we can see this importation of a standard form as grammar-switching (for whatever reason it is triggered). Borrowing between dialects is commonplace and in some cases causes lexical and grammatical changes. This has happened with many northern speakers who only use [ð]-forms. (This was the case with the female informant for Lodge Reference Lodge1966, the grandmother of the informant for Lodge Reference Lodge1978.) Such changes via borrowing are often word-specific. For example, the traditional forms of father in Lancashire and Yorkshire are [faðə] and [fɛɪðə] (with rhoticity in parts of Lancashire), but probably the majority of speakers in these areas today use [faːðə], a borrowing from standard English (possibly via the church). (The last of these forms was the only form used by three generations of the family, some of whose members were informants for Lodge Reference Lodge1966, Reference Lodge1978 and Reference Lodge1984.) In a similar way [ð]-forms of the definite article have been imported from speakers of other dialects, including standard English. The division I wish to establish between the different grammars justifies my rejection of the usual designation of these phenomena as Definite Article Reduction, referred to in the Introduction.
There is one particular difference between the treatment of [ð] in the definite article and the local form [θ] that needs to be noted. The former is subject to assimilation after a sibilant, as in (26) and (27). There are many examples of this in Lodge (Reference Lodge1984), not just the Stockport data.
(26) [mɪs zə] miss the
(27) [wəz zə] was the
Again, these transcriptions ignore some of the important phonetic detail, but they serve to indicate assimilation. On the other hand, [θ] (of the definite article at least) is not subject to such alternations: *[wəz○ sɑspɪtɫ] is not a normal pronunciation of was the hospital.
5 Concluding remarks
This survey of northern forms of the definite article and their historical development has attempted to show that grammars may vary across speakers even in just one or two ways; that lexical entry forms should be non-segmental and highly underspecified; that a monosystemic approach to phonology is unhelpful; and that there must be a separate component of the grammar dealing with phonetic implementation.
The most recent variety that I can find which has a [θ]-form subject to generalizable variable realizations is that represented by Tim Bobbin from the mid to late eighteenth century. The changes which produced the later and present systems can be explained only by recourse to the sociohistorical background of population migration as opposed to a gradual, internally motivated set of purely phonological and phonetic changes on their own. From a historical point of view it may well be that the voiced fricative variants of the definite article are a purely southern British phenomenon. Southern in this context includes the counties of Shropshire, Staffordshire and Lincolnshire. Further investigation is needed to determine which form(s) of the definite article align(s) with which other local forms in the varieties under consideration, for example, we need to know whether the [θ]-form always goes with [ɛɪ] in weight, [ɔʊ] in thought and similar realizations of gh-words.
