An Intraspeaker Study of /t/-glottaling in Scottish English

Recent sociolinguistic research suggests that the previously stigmatised glottal replacement of /t/ has begun to appear more frequently in more standard varieties of Scots and Scottish English, as well as further south in RP. This study investigates the patterns of /t/-glottaling used by Scottish television presenter Lorraine Kelly and whether her rates of /t/-glottaling differ by interlocutor, by comparing two audio clips where she acts as interviewer and interviewee respectively. Whilst a high rate of word-final /t/-glottaling is found, there is no difference in /t/-glottaling between the two contexts, making it difficult to explain this variation through contextual social factors.


Introduction
Glottal replacement, or /t/-glottaling, the realisation of underlying /t/ as a glottal stop, is one of the most widely studied sociolinguistic variables of the English language. Though it is still somewhat stigmatised (Stuart-Smith 1999, Fabricius 2002, /t/-glottaling has undergone a rapid spread across varieties of English from its origins in working-class urban speech (Macaulay 1977, Fabricius 2002. Though it has long been suggested that the feature has its origins in London (Milroy et al. 1994), increasing research into /t/-glottaling in Scotland has revealed a long and rich history (Schleef 2013). Long considered a hallmark of Glasgow Urban Scots, new research has suggested that what was once regarded as "the most openly stigmatised feature of Glasgow speech" (Macaulay 1977:47) has begun to filter into more standard varieties, both in Scotland (Schleef 2013, Smith & Holmes-Elliot 2017, Stuart-Smith 1999 and further south in Received Pronunciation (RP) (Fabricius 2002). Both this widespread innovation and evidence for strong gender and social class conditioning (Milroy et al. 1994, Smith & Holmes-Elliott 2017, Stuart-Smith et al. 2007) makes glottal replacement an interesting sociolinguistic variable for study.
This study will focus on the /t/-glottaling of Lorraine Kelly, television presenter and recognisable speaker of Scottish Standard English (SSE). Using video interviews, we will examine how her patterns of /t/-glottaling differ depending on the nature of speech activity-specifically, when she is engaged in an informal interview with two colleagues compared to when she fulfils her usual role as an interviewer addressing a less familiar interlocutor. The depth of this difference is considered in relation to the grammatical constraints of syllabic position and following phonological context.
A result of no difference between the two speech activities is found, despite a hypothesis of higher rates of the stigmatised glottal variant in the more informal interview context. We attempt to understand these results through looking at attention paid to speech, a variable previously found to have a significant effect on /t/-glottaling (Fabricius 2002, Smith & Holmes-Elliott 2017, Stuart-Smith 1999, Stuart-Smith et al. 2007).

Literature Review
The glottaling of /t/ has undergone a rapid spread across Britain in recent decades, both geographically and socially (Fabricius 2002, Schleef 2013, Stuart-Smith et al. 2007, with recent research suggesting that the feature may have polygenetic roots (Schleef 2013) distinct from previous assertions that it originated in working-class urban speech from London (Milroy et al. 1994). Indeed, the prevalence of the feature across the British Isles is evident, with studies of /t/-glottaling investigating its patterning in varieties of South East London (Tollfree 1999), Reading, Hull, and Milton Keynes (Williams & Kerswill 1999); Newcastle (Milroy et al. 1994); North East Scotland (Smith & Holmes-Elliott 2017); and commonly also in Glasgow (Macaulay 1977, Stuart-Smith 1999, Stuart-Smith et al. 2007, Braber & Butterfint 2008. Now widespread across Britain, /t/-glottaling has been attested as a common feature, at least word-finally, of prestigious RP. Indeed, in an attitudinal and auditory study, Fabricius (2002) finds evidence for the disappearing stigma of the glottal variant in middle-class RP speakers studying at Cambridge University. We will consider the literature surrounding /t/-glottaling and its conditioning factors-first looking at its place in Scotland, before moving on to discuss its most prominent grammatical predictors, and finally its relationship to speech style.

Scottish English /t/-glottaling
Despite assertions that /t/-glottaling originated in London (Milroy et al. 1994), it is one of the most recognisable features of the Scottish vernacular. An increasing body of research into the working-class Urban Scots variety spoken in Glasgow has indicated that the glottal replacement of /t/ has been present in Scotland for over a century (Stuart-Smith 1999:156). Glottal replacement has also been studied in other areas of Scotland, notably Edinburgh, where Schleef (2013) found higher rates of /t/-glottaling among adolescents, especially word-medially, compared to London, suggesting that the variable may be more advanced in Scotland and lending support to the hypothesis of its polygenetic history. Smith and Holmes-Elliott (2017) also investigated /t/-glottaling in Buckie, a small town in North East Scotland, where a large generational shift was found. Younger participants glottaled 90% of tokens, but older participants only did so for 38%, indicating the swift development of the variable in this area.
Most relevant to the present study, however, is the body of research concerning /t/-glottaling in Glasgow, where our speaker is from. Previous research has found that /t/-glottaling is a highly class-conditioned and prestige-driven variable, one used consistently more often by Glasgow's working class and increasing in use across the years (Braber & Butterfint 2008, Stuart-Smith 1999, Stuart-Smith et al. 2007).

Grammatical Predictors of /t/-glottaling
Studies have shown that grammatical predictors such as phonological context, phonotactic rules and morphological compositionality are some important constraints on /t/-glottaling. Schleef (2013) found that, in both London and Edinburgh, phonetic, phonological, grammatical, and lexical factors all condition patterns of /t/glottaling.
Following phonological context has been found to be one of the primary indicators of glottal replacement (Schleef 2013, Smith & Holmes-Elliott 2017, Tollfree 1999, with a domain constraint hierarchy indicating how likely /t/-glottaling is to occur in each environment. Example (1) thus indicates that /t/-glottaling is thought to be most likely to occur before a consonant (PreC), second most likely before a pause (PreP), and least likely before a vowel (PreV).

(1) PreC > PreP > PreV
Smith and Holmes-Elliott (2017) expanded this theory, positing a new hierarchy (Example 2) that considered how likely /t/-glottaling was to occur in terms of both preceding and following syllabic environments.

Speech Style and /t/-glottaling
Speech style and attention paid to speech have been shown to influence rates of /t/-glottaling. Labov's (1972) model of attention paid to speech broadly describes this relationship between formality and stylistic variation, following the basic principle that less attention will result in a more informal speech style, indicated by an increase in use of non-standard or low-prestige variants, such as /t/-glottaling. The paralinguistic channel cues said to indicate less attention paid to speech are changes in tempo, pitch, volume, breathing rate, and laughter. Nevertheless, the Labov model has come under criticism as it does not account for interlocutor differences (Schilling 2004) and neglects the fact that speakers can consciously style-shift into a different vernacular. A later model of sociolinguistic variation proposed by Bell (1984) suggested that style-shifting was more a responsive function performed by a speaker dependent on their audience; thus, the distribution of non-standard variants such as /t/-glottaling is strongly conditioned by interlocutor relationship.
In an investigation into the R-pharyngealisation of Glaswegian comedian Frankie Boyle, Purse and McGill (2016) found higher rates of creaky voice and pharyngealisation when Boyle was interviewed by fellow Glaswegian Kevin Bridges compared to when he was interviewed by RP speaker Richard Osman. This style shift is attributed to the two different interview contexts, evidencing the claim that pragmatic and social factors such as audience design and interlocutor can constrain linguistic variation, though it is eventually concluded that "attention and formality alone are not sufficient to explain linguistic variation" (Purse & McGill 2016:41). Purse and McGill (2016) also find that Labov's attention to speech cues are fulfilled in the Boyle/Bridges interview, with higher rates of the non-standard pharyngealisation.
Further evidencing the link between /t/-glottaling and speech style, Schleef (2013) found that both London and Edinburgh teenagers were more likely to /t/-glottal in a sociolinguistic interview rather than in a reading task. This finding echoed that of Fabricius (2002), whose investigation into /t/-glottaling in RP found that glottaling in pre-pausal and pre-vocalic environments was strongly subject to style-shifting. It was avoided in reading passages but occurred in interview speech. An effect of following phonetic environment was also found (Fabricius 2002:124). Stuart-Smith (1999) similarly observed a higher rate of glottal articulations of /t/ in conversational contexts for both working and middle-class Glasgow speakers, concluding that "[t] occurs as the result of emphasis, prosody and style shifting" (Stuart-Smith 1999:196).
This body of research paves the way for our intraspeaker study on the possible effect of speech activity on /t/-glottaling, in combination with the two arguably most prominent linguistic constraints of position and following phonological context.

Speaker Background
Lorraine Kelly is one of the UK's most widely recognised Scottish television personalities, having presented her eponymous daytime morning show on ITV since 1993. Lorraine, aired between 08:30 until 09:25, covers themes of news, showbiz, fashion, and lifestyle. Born in 1959 in the Gorbals, a working-class area of inner-city Glasgow, Lorraine Kelly grew up in neighbouring East Kilbride, to the South East of the city. Born to two teenage parents, her father worked as a television repairman, placing him in Macaulay's (1977) class IIb of skilled manual workers. Our speaker has been socially mobile as the first in her family to gain a place at university, before dropping out to pursue a career in TV journalism and building a successful career that has made her both a household name and a multi-millionaire. As a prominent female speaker of a Glaswegian variety, her socioeconomic background provides an interesting context for linguistic investigation, especially considering that she now operates in a sphere of middle-class Southern English. For the purposes of this investigation, we will describe her variety as Standard Scottish English, or SSE.

Methodology
As an intraspeaker analysis focusing only on one speaker, this study evidently controls for the variables concerning that speaker, Lorraine Kelly, and her background. The location of recording (the ITV studios in London) also remains the same across the two recordings that our data are drawn from.
The dependent variable is the variant produced: [t] or [ʔ] (all other realisations, of which there were few, were excluded). Our primary independent variable will be "speech activity", as we aim to ascertain whether Lorraine Kelly's patterns of /t/-glottaling differ between the two interview contexts. It is important to note that the variables encoded as speech activity (here, interviewee versus interviewer) and interlocutor speech variety (American or Standard Southern British English [SSBE]) are of course entirely conflated for the purposes of this study. Where Lorraine Kelly is the interviewee, she is interacting with two SSBE speakers, and where she is interviewing, she is interacting with an American speaker. We will frame this variable as interlocutor difference, whilst maintaining that differences in speech variety may also serve to explain our results.
We will also test for the effects of two further independent variables shown from previous research to be the most consistent linguistic predictors of /t/-glottaling. We will examine their impact on the variation found between the two different speech activities. Firstly, we will consider the effects of position within the word (final or medial), and secondly, following phonological environment (vowel, pause, or consonant).
The data were collected as part of a collaborative project looking at patterns of /t/-glottaling in 17 different Scottish female celebrities. The data for our speaker specifically were encoded using the audio software Audacity from two YouTube interview clips, both originally recorded in 2016 in London's ITV studios and televised nationally. Interview 1 (This Morning 2016) consists of Lorraine Kelly being interviewed by her ITV television presenter colleagues Phillip Schofield and Holly Willoughby, middle-class speakers of Southern Standard British English (SSBE), on popular weekday chat show This Morning. Interview 2 (Lorraine 2016) consists of Lorraine Kelly interviewing American actress Jennifer Aniston on her own weekday morning show Lorraine. These two programmes are broadcast nationally directly after one another on the same channel, so it can be inferred that they aim to attract the same audience-one that ITV media describe as a majority of young housewives with children. 1 Though we will not follow Purse and McGill (2016) in conducting a formal analysis of variation by conversation topic, it is important to note that the interviewee data cover more personal and light-hearted topics, as Lorraine Kelly talks about her personal fitness journey, whereas the interviewer data involve her questioning Jennifer Aniston about her acting career. This contributes to our hypothesis that there will be higher rates of the nonstandard glottal variant in the more informal interviewee context.
The interviews, together, provided 5 minutes and 26 seconds of Lorraine Kelly's isolated speech. It was decided that, throughout the wider collaborative project, all tokens of /t/ in phonological contexts eligible for glottal replacement would be recorded, i.e., after a sonorant in coda or non-foot-initial position (Schleef 2013). Coding of each token was carried out using auditory judgement with acoustic confirmation through Praat. This process was completed separately by two researchers to categorise variants as either [t], [ʔ], or in the case of variants divergent from this binary, "other". After a comparison of both coders' analyses, an inter-rater reliability of 98% was found.
Only one variant, an alveolar tap, was found to fall outside of this [t] / [ʔ] binary in Lorraine Kelly's speech. This one token was uttered at a high speed in the connective phrase "but it's". This, along with the fact that it was the sole "other" variant, led us to follow Kirkham and Moore (2016) in excluding this token in the interest of focusing solely on the alveolar and glottal variants of /t/. Whilst the group data coded for syllabic position as per Smith and Holmes-Elliott (2017), operationalising into three levels of either final ("coda"), medial ("ambi"), or onset, we again found only one onset token for Lorraine Kelly, in the word "fantastic". Again, in the interests of focusing on the more prevalently studied final vs medial effect (Kirkham & Moore 2016, Fabricius 2002, Macaulay 1977, this one token was excluded. We also excluded 11 tokens immediately followed by a word beginning with /t/, as the influence of the following segment made it difficult to determine whether tokens had been /t/-glottaled or not. Furthermore, again following Smith and Holmes-Elliott (2017), we capped tokens of the three highest-frequency words within the data set-"it", "that", and "but"-at 10 incidences per interview. This left us with 144 analysable tokens.

Results
We can see from Table 1 that, before taking linguistic constraints into consideration, Lorraine Kelly exhibits an almost identical pattern of /t/-glottaling in both the interviewer and interviewee roles, contrary to expectations. When interviewing Jennifer Aniston she produces 85% of tokens with a glottal variant, and when being interviewed by Phillip Schofield and Holly Willoughby she /t/-glottals 86% of tokens. This runs contrary to our hypothesis that higher rates of /t/-glottaling would be found in the interviewee context, as the speaker is thought to produce a more natural, less scripted, and less considered body of speech. Interviewer 58 (85%) 10 (15%) As exemplified in Figure 1, when we consider rates of /t/-glottaling across the two interviews stratified also by word position, we find that word-final tokens of /t/ are almost categorically glottaled in both cases, with a 100% rate when Lorraine Kelly is the interviewee and a 98% rate when she is the interviewer. We can thus conclude that next to no difference according to speech activity has been found in word-final /t/-glottaling rates. This is a common pattern, as attested across different varieties of the British Isles, from RP to Glasgow Scots (Kirkham & Moore 2016, Trudgill 2001, Macaulay 1977. There is a lower rate of /t/-glottaling word-medially, and we also see some difference between the two speech activities. Contrary to our expectations of higher rates of /t/-glottaling in the interviewee role, Lorraine Kelly appears to /t/-glottal slightly more often word-medially in the interviewer role at a rate of 47%, compared to 31% when she herself is being interviewed.
We next compared the speaker's rates of /t/-glottaling in each of the two speech activities across three different following phonological contexts-consonant, pause, or vowel, as exemplified in Figure 2. Whilst we see, consistent with the previously posited domain constraint hierarchy of glottal replacement, that glottal articulations of /t/ are least likely before a vowel, we find a different hierarchy of PreP > PreC > PreV. Unusually, 100% of pre-pausal tokens are glottaled in both the interviewer and interviewee speech activities-a result of no difference. We also find little difference pre-consonantally, where Lorraine Kelly glottals 98% of the time when being interviewed and 93% of the time when interviewing. We find the same unexpected pattern of slightly higher rates of /t/-glottaling in the interviewer role for pre-vocalic tokens, with a rate of 68% versus 58% for the interviewee role. Unfortunately, the token number proved too small to be able to test the statistical significance of these unusual effects for both word position and phonological context.

Discussion
Overall, our results show that there is almost no difference in Lorraine Kelly's patterns of /t/-glottaling when she is interviewing a prominent celebrity compared to when she is being interviewed by two colleagues, despite the difference in interlocutor variety and inferred attention paid to speech. We also fail to observe any stand-out differences when the effects of phonological context are explored in tandem with speech activity, except some small increases in production of the glottal variant word-medially and pre-vocalically in the interviewer context. We will explore in this discussion the possible reasons for these results and in what ways we may be able to account for these patterns of /t/-glottaling.
Lorraine Kelly exhibits high rates of /t/-glottaling across the board, though our analysis becomes more meaningful when we recognise that this is mostly due to her almost categorical pattern of glottal replacement word-finally, a rate which drops to under 50% word-medially. This aligns with the patterns of /t/-glottaling in speech communities across the UK, where it has been attested that word-final /t/-glottaling is an established feature losing its stigma (Macaulay 1977, Kirkham & Moore 2016, Fabricius 2002.
Though the decreased proportion of glottal articulations of /t/ pre-vocalically is a commonly attested pattern for several varieties (Schleef 2013, Smith & Holmes-Elliott 2017, Tollfree 1999, Lorraine Kelly exhibits an unusual pattern of more pre-pausal than pre-consonantal /t/-glottaling. Nevertheless, it is important to note that the difference in pre-pausal and pre-consonantal /t/-glottaling rates is minimal, and the low number of pre-pausal tokens may account for the unusual rate of 100% /t/-glottaling here. We can thus conclude that across the two linguistic constraints studied, Lorraine Kelly /t/-glottals in a manner aligning with expectations for a Scottish woman of her social standing and speech community. We also found less of a difference in Lorraine Kelly's /t/-glottaling rates across speech activities than expected, leaving our hypothesis unjustified. This suggests that perhaps our speaker is controlling her speech style across contexts that are more similar than they first seemed. Though evidently the interview with Jennifer Aniston is more scripted, evidenced if anything by the fact that Lorraine Kelly is seen to be holding pages of prompts in the video footage, this does not mean that her speech when being interviewed by Phillip Schofield and Holly Willoughby is necessarily entirely spontaneous. Lorraine Kelly fulfils all of the Labov (1972) paralinguistic cues indicating less attention paid to speech in the interviewee context, but only the pitch and volume cues in the interviewer context. While this may make the results of no style shift regarding /t/-glottaling even more surprising, it is important to note that the issues with the attention paid to speech model have been widely attested in recent literature (Purse & McGill 2016, Schilling 2004)-notably that the 1972 model does not account for interlocutor differences.
In order to explain the lack of linguistic differences between interview contexts, we can tentatively suggest that across both interviews, Lorraine Kelly may be conscious of presenting to an audience that is not only largely English rather than Scottish but arguably mostly middle-class. She may thus be controlling her speech style in her capacity as a television personality. This is unsurprising given that glottal replacement of /t/ is still a somewhat stigmatised variant, and a major characteristic of the low-prestige working-class Glasgow vernacular (Macaulay 1977, Stuart-Smith 1999, Stuart-Smith et al. 2007, where her roots lie. Considering Lorraine Kelly has openly spoken in the media about the criticism she faced in her early career regarding her Scottish accent, saying she had been told by a BBC Scotland boss that she would "never make it in television with that accent" (Parker 2016), it is to be expected that she would limit her use of non-standard variants such as word-medial [ʔ]. After being upwardly socially mobile throughout her life, to the point of fame and success, she is acting as part of a higher prestige speech community that we can expect she is making efforts to align with.
Furthermore, an interesting point is raised by the finding, as seen in Figures 1 and 2, that Lorraine Kelly, contrary to expectations, exhibits slightly higher rates of /t/-glottaling in the interview with Jennifer Aniston, both word-medially and pre-vocalically. Though this effect is small in both cases, it is interesting to note that these are two environments where /t/-glottaling is not only less prevalent-as per the PreC > PreP > PreV hierarchy-but more stigmatised (Fabricius 2002). This suggests a style shift downwards, perhaps to make the interviewee feel more at ease, and echoes the findings of Stuart- Smith et al. (2007) in Glasgow Scots, where traces of high /t/glottaling rates remain when the working-class style-shift upwards, though this pattern is not approximated by middle-class speakers in spontaneous speech. An investigation into the effects of interlocutor variety and accommodation could shed further light on this phenomenon. Again, however, it is important to note that we are unable to say whether the results at hand represent a significant effect, at least for our speaker.

Conclusion
This study provided support for existing research that has found /t/-glottaling to be strongly conditioned by grammatical factors, though we did not find considerable intraspeaker variation according to social context. Lorraine Kelly, a speaker from Glasgow, /t/-glottals at a high rate word-finally much more often than she does word-medially. We also found that, consistent with previous research, /t/-glottaling was least likely prevocalically for this speaker. However, minimal differences were found according to interview context, leaving us unable to support our hypothesis that /t/-glottaling would be more frequent in the more informal interviewee context. Though this finding was surprising, it exemplifies some of the difficulties of explaining variation through contextual social factors. It appears that Lorraine Kelly is controlling her speech style across interlocutor and speech role, a finding perhaps explained by the context of our data being drawn from television interviews. These results could be further explored in a more in-depth analysis considering the effects of interlocutor variety and accommodation, as well as audience design.