Variation As An Interactional Resource

Lesley Milroy
University of Michigan


  1. Introduction
  2. Glottaling Versus Tyneside Glottalization
  3. The Prepausal Constraint
  4. Phonological, Lexical And Interactional Explanations
  5. Conclusions
  6. Notes
  7. References

1. Introduction

The change in progress in British English I want to discuss is the spread of the glottal stop as a reflex of /t/ at the expense of alveolar realisations, and I shall try to show that the resultant patterns of variation in the North East English city of Newcastle can be seen to constitute an interactional resource available to speakers which they exploit to structure their interactions. Specifically, I will argue that speakers use the resources provided by a variable phonological system as turn delimitative devices, which are read by interlocutors as a signal that the current turn is complete. The broader general issue I am addressing is the link between the interactional level where speakers use various systematic strategies to accomplish conversational goals, and what we might term the community level, where variation is regularly patterned in accordance with the linguistic and extralinguistic factors identified by sociolinguists, and where the focus of interest is the variable linguistic system.

The data presented here arises from a large study of change in contemporary British English in two English cities, Derby in the North West Midlands and Newcastle upon Tyne in the North East.1 Details of the research design are set out by Milroy, Milroy, Hartley and Walshaw (1994), who present evidence from a number of sources of a robust change in progress in contemporary urban British English, whereby /t/ is realised as a glottal stop with no secondary oral articulation, or as a glottalised stop with some kind of supra-laryngeal gesture.

Although there is glottalization of voiceless stops (particularly /t/) in some coda contexts in all accents of English, glottalization is particularly salient in Britain where it appears in recent years to have been spreading quite rapidly. Traditionally, glottalization has been heavily stigmatized but is currently evident in the speech of persons of high social status who would ordinarily be thought of as using `Received Pronunciation'; one prominent user of glottalized variants (specifically the glottal stop) is Princess Diana. Glottalization is one of the features that John Wells had in mind when he remarked that `mainstream RP is now the subject of imminent invasion by trends spreading from working class urban speech, particularly that of London' (1982: 106). Glottalization - especially the glottal stop - is a prominent characteristic of London vernacular, and in Britain generally it tends to be associated with urban dialects. Its traditional image, confirmed in systematic sociolinguistic surveys in the 1970s, is urban, male and working class (see for example Trudgill 1974: Macaulay 1977). Almost 25 years ago Roach suggests a real time anchor for the spread of the change, commenting that that it is difficult to find `English speakers below about forty years of age who do not have some type of glottalization' (1973: 21), and Trudgill (1988: 44) documents an increase in glottal stop usage in formal speech styles in Norwich between 1968 and 1983. Mees (1987) discusses the spread of glottalization to Cardiff, South Wales (where it was not formerly salient), and Holmes (1995) describes it in New Zealand. It appears to be spreading to Newcastle upon Tyne also, and we are concerned here with the social and interactional constraints on its spread to certain environments. In order to interpret this particular phenomenon in relation to the larger scale change currently in progress, we need to say a little more about their distinction between glottalling - the replacement of /t/ by a glottal stop and glottalization - the replacement of /t/ by some kind of glottalizes stop.

2. Glottaling versus Tyneside glottalization

Traditional Tyneside English is heavily glottalized, but it is more strongly characterized more saliently by glottal reinforcement than by replacement of the oral plosive by a glottal stop. This `combined glottal and oral plosive' (Wells 1982: 374) has been variously described, but we are inclined to agree with Wells characterization of his auditory impression; he suggests that the glottal follows the oral closure with `glottal masking of the oral plosive burst'. Glottalization is more widely distributed than in most other British dialects and is particularly salient in intervocalic position in words of the type butter. pepper, sticker. There are some indications from dialect geography that this feature may be of some antiquity, as it shows the characteristic distribution of a relic form in the rural hinterland of Newcastle, being found also in the extreme north of England, south-west Scotland and northern Ireland.

Tables 1 and 2 show how much more frequent glottal reinforcement is than glottal replacement in Newcastle and Table 1 shows that it is particularly preferred by males. Table 2 on the other hand associates the glottal stop particularly with younger speakers, with middle class speakers and females in the lead. Since this is the social trajectory of change reported in we are on reasonably firm ground in regarding glottalization (the Table 1 pattern) with its very different age, class and gender distribution as a traditional feature of the dialect rather than an innovation. While glottalization is associated with rural dialects the glottal stop appears to be chiefly an urban phenomenon (see further Milroy et al 1994).

Table 1: Percentage use of glottally-reinforced variants of /t/ (medial and pre vowel word final contexts combined)
males females N
males females
older WC 61 19 427 762
older MC 69 21 694 675
younger WC 52 41 421 647
younger MC 46 31 521 569


Table 2: Percentage use of glottal stop variants of /t/ (medial and pre vowel word-final contexts combined)
males females N
males females
older WC 9 9 427 762
older MC 3 3 694 675
younger WC 10 13 421 647
younger MC 19 30 521 569

Note: contrast between male and female patterns of use with respect to glottalized and glottal variants is highly significant: p<<0.001.

Note that Tables 1 and 2 refer to the distribution of variants in prevocalic contexts, both within words and at word boundaries, and that glottalized variants are very frequent indeed in these contexts, particularly in the speech of men. We shall see shortly that glottal variants also occur very much more frequently in these contexts than in prepausal contexts where they are almost entirely absent. This pattern of distribution is very different from that reported by Wells for RP, where the glottal stop is gradually moving into different places in the syllable: it occurs already in most coda contexts, but although it can occur word finally before vowels : (pick i[t] up ) or in absolute final position (Let's star[t])! Intervocalically within a word it remains firmly excluded from RP (cf. Cockney ci[t]y). Wells comments:

`the increased use of glottal stops within RP may be attributed to influence from Cockney and other working-class urban speech. What started as a vulgarism is becoming respectable.' (Wells 1994: 201)

The point I wish to emphasise here is that the ordering of contexts affected by the change seems to be different in RP and in Tyneside, and preliminary work on the urban dialects of Derby and Milton Keynes suggests that the RP pattern of glottaling /t/ in coda positions in preference to intervocallically within the word is rather general in other dialects also of British English. Bearing these different distributions in mind, we turn now to discuss a constraint on glottalization apparently peculiar to Tyneside which almost entirely inhibits the occurrence of glottal stops (or any glottalized form) in the coda in prepausal positions. This we have described as the (P)re(P)ausal (C)onstraint.

3. The Prepausal Constraint

Although, as we have seen, the dialect is heavily glottalised, glottal realizations do not occur (with a few exceptions to be discussed below) in pre-pausal positions in the discourse, e.g., at phrasal boundaries, utterance ends or at the ends of speaker turns. The preferred realisation in these positions is a fully released stop with no glottalisation, but with heavy affrication or spirantisation, auditorily similar to the breathy realisations associated with Liverpool and Anglo-Irish dialects of English (Knowles 1978:82) and contrasting markedly with glottals. Instrumental analysis suggests that in some instances at least this realisation is not a fully occluded stop consonant, but a fricative.2 Although variation between glottal and non-glottal must be viewed as a (variable) phonological phenomenon, an illuminating account takes us beyond phonology. This is because the most obvious constraint is not a property of the lexical item or the syllable nor the word-boundary: rather, variation is related to utterance or phrasal position. A lexical item that is glottalized in intra-utterance position will appear as fully released in pre-pausal or utterance-final position. Thus, in items like sat, sheet, bite, the coda is sometimes glottalized (i.e in phrase internal contexts) and sometimes fully released (i.e in prepausal contexts) in identical lexical items and in a way that appears to be rule-governed. Note that most accounts of glottalization of stops associate it generally with coda positions. In the following sections we explore a number of candidate explanations for this distribution, focussing particularly on interactional constraints using the analytic framework of Conversational Analysis.

The constraint has been investigated mainly by auditory analysis (with supplementary corroborative instrumental analysis) as part of a larger study which focuses on phonological variation and change in contemporary spoken British English. Fieldwork in Tyneside has produced recordings of 32 speakers distributed according to class, age and gender, yielding 4 speakers per cell as shown in Tables 1 and 2. Informants were recorded first in a (usually single sex) dyad for around 50 minutes. One young working class woman, K, was recorded twice, with interesting results as we shall see. Informants were then asked to read a word-list constructed to elicit citation forms. The list had been constructed with the PPC in mind, as the constraint had been noticed in preliminary listening to recordings that had been carried out for a smaller pilot project (Hartley 1992). It was hypothesised that single-word citation forms with final voiceless stops (e.g. sheet, gate) would be treated by wordlist readers as pre-pausal items and would therefore be realized as non-glottals, but that phrase-internal and word-internal /t/ would normally be glottalized. It was particularly important to include phrasal items such as I beat it, as it was expected that the same (or similar) words would be treated differently in phrase internal and pre-pausal positions. Table 3 shows a portion of one speaker's version of the wordlist focusing on /t/, which followed the predicted pattern. Tokens read as non-glottalized are in bold.

Table 3: The prepausal constraint on glottalization: the word list style of one speaker

Table 4 shows the glottalization patterns of all speakers with respect to this relevant portion of the wordlist. As predicted, the single-word items shown in Table 3 with final /t/ were read as non-glottal and fully released and the items with word-internal and phrase-internal /t/ were consistently near categorically by most speakers (except that a minority of speakers - all older women- tended to avoid glottalization here). The item it , even when not carrying primary stress, was also uttered with full release in the prepausal context (I beat it). For 31 out of 32 speakers, the PPC was applied categorically in all relevant items. The remaining speaker (young female) showed failure of the PPC, i.e., glottalisation, in two out of thirty relevant items: these were print and salt. These violations of the rule, summarized in Table 5, constitute approximately 0.2% of the total number of the tokens recorded. Thus, on wordlist data alone, we can conclude that the PPC in effect operates categorically in pre-pausal position.

Table 4: Word list style: the prepausal constraint on glottalization : all speakers
Total N Non-Glottalized Glottalized
N % N %
960 958 99.79 2 0.21

Speakers are however often found to behave differently in spontaneous conversation than in word list reading, and the question arises of whether the patterns shown in Table 4 are replicated in the conversational data. Investigating this is a much more difficult and time-consuming task replete with practical problems relating to the identification of pauses and final positions in speaker-turns, and the number of relevant sites that are identified naturally varies from speaker to speaker. In the event we came close to our aim of locating 30 sites per speaker (120 per cell), as the results displayed in Table 5 show.

Table 5 sets out for each speaker group the number and percentage of 'violations' of the constraint, i.e., the occurrence of glottals where non-glottals are predicted by the analysts (the speakers of course are not presumed to be violating anything - it is the analysts' tentative formulation of the constraint that is violated). Violations are more numerous than in wordlist style, but for 7 groups out of 8 they are rather rare, varying between 2% and 7% and averaging about 4% - i.e., 4 violations in every hundred relevant items. Indeed 11 of the 32 speakers never violate the constraint at all - they produce fully released stops in 100% of relevant cases, and the total number of violations is 62 out of 928 - approximately 6.79%. Considered from a variationist perspective such a low incidence of a variant is unusual and, taken together with the wordlist results, suggests that we are indeed dealing with what is for most speakers a categorical across the board constraint, the conditions for which we may be able to specify more closely. This involves accounting for the apparent violations - the exceptions that in Neogrammarian terms may be governed by another specifiable rule or rules, and it is to be expected that the young female group which shows 30% violation of the constraint will be of some interest here. I turn now to suggest some phonological and lexical characteristics of `exceptions' to the PPC, focussing finally on an account based on the resources for structuring conversation provided for interactants by this part of the dialect's variable phonology.

Table 5. Conversational style: pre-pausal position: number (N) and percentage of glottal or glottalised tokens
class group tokens N %
WC Fem. 45-65 120 2 2
Male 45-65 111 2 2
Fem. 16-25 101 30 30
Male 16-25 120 6 5
MC Fem. 45-65 120 7 6
Male 45-65 116 2 2
Fem. 16-25 120 5 4
Male 16-25 120 8 7

4. Phonological, Lexical And Interactional Explanations

Exceptions to the PPC occur in short vowel items; in long vowel items such as great, meet, it is effectively categorical. Thus, we may be tempted to suggest that vowel-length is an important factor or at least that long-vowel environments block the operation of the constraint. Amongst the short vowel class, however, certain items (e.g. that, get, it) occur very frequently. Glottalised forms appear very commonly in these items, such that we might hypothesise that PPC violations are in the main restricted to them. If glottalisation is spreading into pre-pausal and turn-final environments, it appears to be spreading via these lexical items in conversational style, but not yet in careful wordlist style (as we have seen). In traditional phonological terms, then, this process of glottalisation might be well described as operating by a process of lexical diffusion, with frequently occurring items in the vanguard of the change.

The most remarkable difference in PPC application in comparison to other groups is displayed by the young working-class females. The high glottalization scores for this group in the main conform to the lexical patterns that we have just mentioned. However, the conversational behaviour of three of the four speakers in the group, H, K and L, is markedly different from that of the other subjects in the study, and we shall consider the patterns of variation in prepausal /t/ contexts of each of these. Speaker H (who speaks much less than her partner on the tape) produces glottal variants in 10/21 (52%) pre-pausal tokens. The glottals tend to occur in common items (got, that, it, out, about) and in tag items which are clearly turn-final (all right, but - the latter used adverbially in Tyneside in the sense of 'though').

Speaker K's behaviour, as summarized in Table 6 below, is however particularly striking, and gives a strong indication that exceptions to the PPC can illuminatingly be explicated in terms of the function of this tiny variable phonological subsystem as an interactional resource. From this point I shall develop the argument chiefly with reference to K's behaviour.

Table 6: The prepausal constraint on glottalization in conversational style: distribution of variants in the speech of K
Tape A: K in conversation with a female friend:
Total N Non-Glottalized Glottalized %Glottalized
45 21 24 53.1
Tape B: K in conversation with her brother:
Total N Non-Glottalized Glottalized %Glottalized
34 30 4 11.8

K participated in two recorded conversations, first with her brother, and later with a female friend, L. As shown in Table 6, the glottalized variants turn up most frequently in the conversation with L, but in the conversation with her brother K's pattern of PPC application is comparable to that of other informants: just 4 out of 34 (11.8%) of the pre-pausal and turn-final tokens are glottalised. K's violations of the PPC occur overwhelmingly on the sentence tag and that (e.g. you just miss your friends and that), and this suggests that the frequency of violations may depend to a great extent on the occurrence of sentence tags in particular conversations. This tag, which is one of several commonly occurring tag items in British English, is in turn related to the age, class and sex of the speaker, a matter which I cannot pursue here. However, I can comment that tags seem to be associated particularly with young speakers, especially women. For the tag and that alone, K in her same gender conversation uses a glottal in pre-pausal position 16 times out of 24 (67%). The same tag occurs only 6 times during K's entire conversation with her brother, and in 4 of these cases (67%) we find glottalization. Thus, even where tags occur infrequently, we find a similar proportion (around two thirds) of glottalized tokens accounted for by tags: this seems to link exceptions to the constraint in an interesting way with tags, since other tag items such as isn't it, occur more rarely, but are also glottalized in K's conversation, and in others.

Speaker L (K's female interlocutor) produces 6 pre-pausal glottal tokens, 4 of which occur on the tag and that. This tag is much rarer in the speech of other informants, but other tags which terminate in /t/ such as isn't it do appear occasionally to attract glottalised forms, especially in the speech of younger people. For example, the young middle class male P produces 4 pre-pausal glottals, 2 of which fall on the tag isn't it. Similarly, another young middle-class speaker R produces his only pre-pausal violation of the PPC on the tag wasn't it. Speakers who use few tags may also glottalize pre-pausal but, about, right, got and a limited number of other items, but these violations tend again to be strongly associated with young female speakers.

This frequent association of PPC violations (which are in any case rare) with tags suggests that an interactional explanation of the PPC and its apparent exceptions should be considered. In a dialect with heavy use of glottals, interlocutors appear to be oriented to a fully released variant of /t/ as a turn-delimitative signal (i.e. that a speaker is prepared to yield the floor). There is in fact a small but suggestive literature on such interactive functions of phonetic cues, constituting a field described by Local, Kelly and Wells (1986) as `phonology of conversation'. French (1988), presents evidence that post-vocalic [r] is used in this way in a Yorkshire dialect; Local, Sebba and Wells (1985) smilarly examine a range of phonetic procedures for turn delimitation in London Jamaican, while Local et al (1986), document a range of phonetic signals, but make observations similar to our own on the turn-delimitative function of fully released /t/ in Tyneside.

In addition to phonetic cues however, grammatical elements such as tags of various types may additionally (or alternatively) constitute turns as complete; for example, Local et al (1985) discuss this function of the tag `you know'. (`Yeah, they thump him them thump him good an'proper you know'.) A range of prosodic and non-verbal cues fulfil a similar function. With respect to the Tyneside data presented here, the turn-delimitative function of tags may thus suggest a motivation for the absence of the phonetic cue - the fully released variant of /t/ - on words such as it and that, which, as we have already noted, are the short-vowel items most likely to be realised as glottal variants. Younger working-class females use by far the highest number of tags (this may itself be a gender-related feature) which largely explains why they have much the highest rate of exceptions to the PPC.

Exceptions may thus be plausibly be accounted for with reference to the turn-delimitative functions of the PCC, being redundant when it co-occurs with a sentence tag which is a particularly salient turn-delimitative cue. We need to look further however at the extent to which such an interactive account is adequate. Particularly, we have found that the PPC generally operates before mid-turn pauses, even when it seems clear that the turn is not constituted as complete. Examples include the fact tha[t] # the kids are a lot more streetwise, the da[t]e # was that day. Our findings in this respect do not at present support those of Local et al (1986) who note that where `delimitative features are present at a potential transition point but no transition occurs, current speaker frequently displays in his or her subsequent talk a desire to relinquish the turn' (p. 416). Such evidence may be found in subsequent attempts to achieve turn transition to another speaker `by the production of a tag-question followed by a brief pause' (p. 432). It is not entirely clear at present how far the instances of full release without glottalization in intra-turn contexts in our own data may plausibly be projected as potential opportunities for turn handover to which the current speaker is oriented; one confounding factor is the apparent stylistic function for some speakers of the fully released variant to mark emphatic stress. Nevertheless, despite these problems the disproportionate tendency of utterance final tags to be realised with glottal or glottalized tokens requires some explanation, which cannot be offered by a phonological account, and I suggest here that the variant realisations of /t/ permitted by the phonological system provide conversationalists with a resource which is exploitable for interactional purposes.

5. Conclusions

This two-level analysis has wider implications, first with respect to the function of phonological units. Variation in this sub-system is apparently being employed in stylistic and interactional functions and while such variation is certainly systematic, it seems unlikely that it can be governed purely by phonology, given the goals traditionally assumed by phonologists. An account of the kind I have sketched out here thus assumes that a system of phonological rules such as those which constrain alternative realisations of /t/ provides speakers with a specification of unmarked or normal behaviours which they might then manipulate in order to achieve specific interactional goals (cf. Mülhaüsler and Harré 1990, 101-2). I would suggest that it is this kind of constant stretching of rules, of normative limits, in everyday interactive contexts where we must surely look for the continuous implementation of linguistic change.

This leads us to contemplate the implications of these findings for the actuation of linguistic change and the mechanisms of its diffusion through the linguistic system - an aspect of what Weinreich, Labov and Herzog (1968) have described as the embedding problem. I have identified a linguistic context (prepausal) which is resistant to an ongoing change, and shown that this change is apparently led by young women. The grammatical tags, which are also used heavily by young women, may be viewed as the Trojan horse which is carrying the change into the last relevant context of the system which has hitherto resisted it. I have suggested an interactional motivation for the implementation of the final stage of this change, linking the interactional level of analysis with a more abstract analysis of phonological systems and sociolinguistic patterns. Surely a characterisation of this micro-macro link which enables us to relate abstract variable patterns to speaker behaviour is an important part of the variationist's research agenda.





  1. Research supported by the UK Economic and Social Research Council (Grant no. R000234892 ``Phonological variation and change in contemporary spoken British English''). I am grateful also to my co-researches James Milroy, Paul Foulkes, Gerry Docherty and Penny Oxley for their input to this work.
  2. As Beal (1993) has pointed out, Irish immigration to Tyneside in the 19th century was extensive; in 1851 one person in 10 in Newcastle is reported to have been born in Ireland, and there are several features present in Tyneside English which seem to be of Irish origin. Since Liverpool English also is heavily influenced by this and other Anglo-Irish dialect features, the fully releaseed stop in prepausal positions in Tyneside may be considered to be a `borrowed' phonological rule, as discussed by Thomason and Kaufman (1988:16).




  1. Beal, Joan. 1993. The grammar of Tyneside and Northumbrian English. In J. Milroy and L. Milroy (eds) Real English: the grammar of English dialects in the British Isles. London: Longman.
  2. French, Peter. 1988. Word final /r/ in a northern English accent: an interactional account of variable production. In Auer, Peter and di Luzio, Aldo Variation and convergence. Berlin: de Gruyter.
  3. Hartley, Susan. 1992. A study of the effects of sex and age on glottalization patterns in the speech of Tyneside schoolchildren. Unpublished undergraduate thesis, University of Newcastle upon Tyne.
  4. Holmes, Janet. 1995. Glottal stops in New Zealand English: an analysis of variants of word-final /t/. Linguistics 33: 1-31.
  5. Kerswill, Paul. 1987. Levels of linguistic variation in Durham. Journal of Linguistics, 23: 25-49.
  6. Knowles, Gerard O. 1978. The nature of phonological variables in Scouse. Sociolinguistic patterns in British English, ed by Peter Trudgill. London: Arnold.
  7. Local, John Kelly, and William G.H. Wells. 1986. Towards a phonology of conversation: turntaking in Tyneside. Journal of Linguistics 22: 411-437.
  8. Local, John Kelly, William, G.H. Wells and Mark Sebba. 1985. Phonology for conversation: phonetic aspects of turn delimitation in London Jamaican. Journal of Pragmatics 9, 309-350.
  9. Macaulay, R. 1977. Language, Social Class and Education: a Glasgow study. Edinburgh: Edinburgh University Press.
  10. Mees, Inger. 1987. Glottal stop as a prestigious feature in Cardiff English. English World-Wide 8 (1): 25-39.
  11. Milroy, James, Lesley Milroy, Susan Hartley, S. and David Walshaw. 1994. Glottal stops and Tyneside glottalization; competing patterns of variation and change in British English. Language Variation and Change 6: 327-57.
  12. Milroy, James. 1982. Some connections between Galloway and Ulster speech. Scottish Language, 1: 9-20.
  13. Mülhaüsler, Peter and Harré, Rom. 1990. Pronouns and people. Oxford: Blackwell.
  14. O'Connor, Joseph D. 1947. The Phonetic system of a dialect of Newcastle-upon-Tyne. Le Maître Phonetique 87.6.
  15. Roach, Peter J. 1973. Glottalization of English (p), (t), (k) and (tsh) - a re-examination. Journal of the International Phonetics Association 3: 10-21.
  16. Thomason, S. and T. Kaufmann. 1988. Language contact: creolization and genetic linguistics. Berkeley and Los Angeles: University of California Press.
  17. Trudgill, Peter J. 1988. Norwich revisited: recent changes in an English urban dialect. English World Wide 9 (1): 33-49.
  18. Trudgill, Peter J. 1974. The social differentiation of English in Norwich. Cambridge: CUP.
  19. Weinreich, U. W. Labov and Herzog, S. (1968) Empirical foundations for a theory of language change. In Lehman W.P and MAlkiel (eds) DIrections for Historical Linguistics, 95-118. Austin, University of Texas Press.
  20. Wells, John C. 1994. The Cockneyfication of RP? In Melchers, G. and Nils-Lennert Johannesson Nonstandard varieties of language. Stockholm: Almqvist and Wiksell International.
  21. Wells, John C. 1982. Accents of English (3 vols). Cambridge: Cambridge University Press.