Supernanny: An Intraspeaker Study of Addressee Effects in the Speech of Jo Frost

Limited research exists evaluating the extent to which intraspeaker style-shifting is conditioned by addressee age and addressee nationality. The current study investigated Supernanny Jo Frost’s realisations of (t) as glottal or non-glottal towards British and American adults and children. Frost was found to style-shift to child-directed speech (CDS) when addressing children, which is interpreted in terms of communication accommodation theory as a communicative strategy to enhance clarity. Frost generally avoided convergence towards her American (and British) interlocutors, such that her style-shifting was found to be conditioned less by addressee nationality than by addressee age. I argue that this avoidance of convergence was motivated by her desire to construct an authentic, authoritative identity, which she achieved through her exploitation of particular indexical meanings associated with [t] and [ʔ].


Introduction
Individuals have the capacity to alter their speech in a number of subtle and more pronounced ways depending on the interactional context. This speaker-internal variation can be described as style-shifting, the systematic study of which allows sociolinguists to build up a picture of the structure of society as it is reflected in and created by language use. Limited research exists evaluating the extent to which style-shifting is conditioned by addressee age and addressee nationality. For this reason, we investigated whether Supernanny Jo Frost employed styleshifting in terms of her realisation of (t) as glottal or non-glottal towards British and American adults and children. 1 Frost's variety of English falls somewhere on the continuum between Cockney and Estuary English (EE), and Cockney has been found to favour glottalization to a greater extent than EE (Altendorf 2003). Frost was found to produce fewer glottal variants of (t) towards children, while no significant difference was found between British and American addressees. Thus, addressee age was found to have a more pronounced effect on intraspeaker variation than addressee nationality. The present study illustrates that different aspects of addressee identity vary in the nature and extent of their impact on intraspeaker variation. The systematic variation between Frost's speech towards children and adults is likely to reflect an accommodative shift towards child-directed speech (CDS) for the purpose of clarity (Giles et al. 1987). In speech directed towards British and American interlocutors, Frost's avoidance of convergence appears to be motivated by a desire to construct an authentic, authoritative identity. The present study posits that she achieves this by invoking particular indexical meanings through her use of glottal and non-glottal variants. This study highlights the need for more integrative approaches which are able to encapsulate broader listener-oriented shifts as well as nuanced speaker-initiated variation.

Previous Literature
Style-shifting research involves the systematic study of when, how, and why an individual's speech varies as a reaction to, or for the purpose of initiating, a change in a given interaction. Various theories have been invoked to account for style-shifting, including attention to speech (Labov 1972), communication accommodation theory (CAT; Giles et al. 1987), which motivated audience design (Bell 1984(Bell , 2001, and theories of speaker design (Eckert 2008, among others). The present study opts to frame our findings in terms of the more general theory which motivated audience design (i.e., CAT), as well as drawing aspects directly from theories of speaker design. In this paper, I implement an integrated model to analyse the current results, amalgamating ideas from theories which have typically been conceptualised as distinct.
CAT (Giles et al. 1987) refers to a speaker's adjustment for the purpose of converging towards or diverging from their addressee's style of speaking (Giles et al. 1991). Factors such as the interlocutor's sociodemographic traits, communicative goals, and the interactional context come to the fore as important predictors of style-shifting according to the CAT model. Speaker design is an approach to stylistic variation encompassing a range of theories, unified in their emphasis of speaker agency (Schilling 2013). Theories of speaker design posit that "social actors weave their meanings through language choices" (Coupland 1985:155). This implies an active role for the speaker in the construction of their identity, and thus suggests that variation may be influenced by an interaction between both the contextual environment and the speaker's conscious or subconscious interactional goals. The present paper adopts Eckert's (2008) perspective regarding the flexible and often varied social meanings associated with particular variants. Central to this view is her notion of an "indexical field", or a "constellation of ideologically related meanings, any one of which can be activated in the situated use of the variable" (Eckert 2008:454). For example, while released /t/ (roughly approximating our "non-glottal realisation") is thought to characterise "formal" or "articulate" speech, it has also been found to index being "prissy" and "exasperated" (Eckert 2008:469). Therefore, in order to explain patterns of intraspeaker variation, it is crucial to go beyond the view that variants have static, categorical meanings associated with simply convergence or divergence.
Style-shifting may be motivated by a variety of different reasons, including the following: "(1) to gain the approval of our interlocutor(s); (2) to maintain a positive social identity; (3) to promote communicative efficiency" (Levey and Harris 2002:19). Although these motives were developed in terms of CAT, the present study opts to use them as a basis for our hypotheses because they appear to capture some of the motivations which are most relevant. The following sections will explore existing literature pertaining to the influence of addressee age and nationality on intraspeaker style-shifting.

Existing Research into the Effects of Addressee Age on Intraspeaker Variation
Our present hypothesis that a child interlocutor would condition an accommodative shift towards CDS is motivated by existing research. CDS is a well-documented "style" in the language acquisition literature, though it is less commonly studied in variationist sociolinguistics. CDS tends to be hyper-articulated (Kuhl et al. 1997) and is "characterized by a reduction in use of phonetic forms with connotations of non-standardness, stigma, and/or local currency" (Foulkes et al. 2005:196). Varying systematically from adult-directed speech (ADS), the adaptations made by speakers in CDS are found to aid interaction with children through, for example, "facilitating the salience and intelligibility of words" (Soderstrom 2007:516). Fernald (2000:243) suggests that a main function of CDS is to "provide clearer exemplars for the inexperienced listener". We therefore predict that one of the motivations of speaking in CDS may be to promote communicative efficiency.
A related potential motivation for shifting to the use of a "standard form" associated with CDS is for the purpose of socialization, i.e., to provide children with a more socially valued exemplar. This was investigated by Foulkes et al. (2005), who reported that a group of mothers in Newcastle produced significantly more realisations of [t] in CDS than in ADS. The present study is motivated to test for similar effects in a speaker who is not related to the children in question, with the prediction that Frost would produce more realisations of [t] than [ʔ] towards children.

Existing Research into the Effects of Addressee Nationality on Intraspeaker Variation
Existing research into the effects of addressee nationality finds that speakers may be motivated to style-shift towards a more "standard" variant in order to gain approval or maintain a positive social identity. Levey and Harris (2002:18) note that the adoption of EE might constitute a "compromise" for speakers who want to avoid the popular and often negative meanings attributed to using variants which are deemed as either Received Pronunciation (RP) or Cockney. The present study does not claim to pin down the exact nature of EE as compared to Cockney and RP, but to highlight the definition reached by Levey and Harris (2002:18) that EE "may be seen as a halfway house between two phonetic extremes". Frost's variety appears to fall towards the "Cockney" end of the continuum (also here referred to as "Mockney", short for "media representations of Cockney") (Androutsopoulos 2014:155). For this reason, we might expect her, according to CAT, to converge towards EE when interacting with speakers of EE and perhaps also when talking to speakers of General American English (GAE) (Wells 1982). Moreover, GAE variants of (t) include [t] and [ɾ], so a Cockney/EE speaker might be motivated to converge to [t] since it is the "variant in common" (Cruttenden 2014:87) with GAE.
Indeed, previous studies do find that speakers converge towards the variety of their interlocutor in second dialect contexts. For example, in the United States, Love and Walker (2013) studied English and American fans of soccer and gridiron, reporting that speakers with positive attitudes towards gridiron were more likely to adopt features of the corresponding variety (i.e., American English). Similarly, Babel (2010) reported that participants from New Zealand varied in the extent to which they accommodated to an Australian model talker based on their positive or negative pre-existing social biases regarding Australia. The aforementioned studies are similar in their findings in terms of a positive affiliation with a certain place affecting the extent to which a speaker is likely to converge towards that variety. In other words, it appears that theories of CAT and speaker design can be complementary, since "speakers of language cannot help accommodating, but group-identity attitudes modulate this automatic process" (Babel 2010:453). In this way, it is equally possible that a speaker may refrain from converging or diverging to some extent in order to maintain an "authentic" identity.
The pressure for a speaker to converge in a given context may also be mediated by the social roles pertaining to that context. Since linguistic variants can be combined in different ways to invoke and constitute certain social roles, Schilling-Estes (1998:76) views style-shifting "as a way for speakers to maintain multiple roles". Shapp et al. (2014) analysed the speech of Ruth Ginsburg and found a clear binary distinction between Ginsburg's THOUGHT vowel height during two separate periods of her career. During her term as a lawyer, Ginsburg appeared to be motivated by a desire to gain the approval of her interlocutors in court, converging her speech towards a "standard", non-regionally marked pronunciation of the THOUGHT vowel. However, at a later stage when she held the more prestigious role of Supreme Court Justice, Ginsburg reverted to raising her THOUGHT vowel. According to Shapp et al. (2014), this serves to provide evidence that Ginsburg's position of authority exempts her from the pressure to converge towards the "standard" variant. The current study builds upon these findings, investigating whether Frost, in her role as an "expert" (as opposed to her role as a "nanny"), would be motivated to style-shift, and whether this effect would be mediated by her interlocutor's age or nationality.
We move forward, then, with an integrated view of CAT and speaker design, assuming that convergence and divergence are not fixed but are flexible and can be manipulated by speakers in different ways related to interlocutors, social roles, and interactional context.

Choice of Speaker
The speaker, Jo Frost, is a self-proclaimed parenting expert and the executive producer and presenter of her own television show, Supernanny, which has been filmed in the UK and the USA and broadcast on Channel 4/E4 (UK) and ABC/Style/Up TV (USA). She has had over 20 years of childcare experience as a professional nanny to date, through which she "hone[d] her successful methods of child-rearing with hands-on, real-life experience" ("About Jo" n.d., para 1). Frost represents an optimal choice for the particular social contexts which are of interest since she speaks to both adults and children in the UK and in the USA. The premise of Supernanny is that through spending a week with a family who have reached out to her for help, Frost is capable of solving a variety of their issues, from temper tantrums to bedtime battles. Despite her seemingly working-class beginnings, she has achieved worldwide recognition for her contributions to the world of parenting (Milsom 2011).
Broadly, Frost's speech variety falls somewhere along the continuum between Cockney and RP, and so could be compared to EE (Altendorf 2003). Impressionistically, Frost appears to utilise more of the features typically associated with Cockney, such as the replacement of /θ, ð/ with /f, v/ (Cruttenden 2014:90). Indeed, Frost's particular idiolect has often been the subject of wider commentary by the press, exemplified by an opinion article (based on an interview with Frost) which transcribes her "non-standard" variants orthographically: "What does she like about kids? 'Everyfing,' Frost replies instantly, her mockney accent, in the vein of Jamie Oliver, reflecting hands-on, working-class authority" (Milsom 2011, para. 3). In this way, it is interesting to examine Frost's Cockney-esque speech in terms of style-shifting, since factors such as addressee age and nationality may influence convergence towards a variety which is perceived to be more standard.

Procedure
The present study investigated the speech of Jo Frost towards British adults and children, and American adults and children. The data consist of 153 tokens of (t) taken from episodes of the UK and US seasons of Jo Frost's Supernanny in 2007 and 2005, respectively. Video recordings of the episodes were accessed from YouTube (Uganda Be Kidding Me 2014, Supernanny Fan 2016), and the digitisation of speech data was carried out using Audacity (Audacity Team 2015). Since this procedure relied on auditory coding of tokens using Praat (Boersma and Weenink 2015), two researchers cross-coded part of the data, with an inter-coder reliability rate of 87.5%.

The Social Variables
The families which featured on the episodes analysed were named "The Gormley-Brickley Family" (UK) and "The Jeans Family" (US). The Gormley-Brickley family consisted of a mother and father (both employed as university professors) and their two sets of twins, consisting of twin boys aged almost 3 and a boy and a girl aged almost 5. The Jeans family was made up of a mother, a father, a daughter aged 4, and twin girls aged 3. The dialect spoken by the Gormley-Brickley family can be described as EE, and the speech variety of the Jeans family, who are from Denver, Colorado, approximates GAE. The occupations of the parents of the Jeans family are not clear from the recordings, 2 so limited observations are made in the current paper regarding the influence of social class as a variable. These particular recordings were chosen because the children who featured in them are of roughly similar ages and are all young enough to be spoken to with at least some features of CDS (cf. Soderstrom 2007).
The study analysed recordings ranging from 4 minutes and 11 seconds to 8 minutes and 52 seconds of speech.

The Linguistic Variable
The linguistic variable we investigated was the realisation of (t) as glottalized or non-glottalized. This variable was selected because, as highlighted in commentaries (cf. Milsom 2011), Frost is known for her frequent glottalization of (t). Glottalized (t) is a salient marker of the Cockney dialect and has been understood to connote dishonesty, hardness, and a lack of education (Levey and Harris 2002:18), whereas released (t) is typically thought to portray formality, politeness, and carefulness (Eckert 2008). In the analysis that follows, I will argue that the indexical meanings associated with these variants are not limited to the aforementioned.
Realisations of (t) were coded in a binary manner, classifying tokens as glottalized /ʔ/ or non-glottalized /t/. The term "non-glottalized" refers here to any variant of (t) which was not clearly glottalized and therefore encompasses unreleased /t/ and an occasional tap /ɾ/ in addition to typical released /t/. In ambiguous cases where it was unclear whether the token was [ʔ] or unreleased [t], both researchers discussed what they heard and an agreement was reached as to how the token would be coded. Table 1 presents the linguistic environments which were examined in the current study, all of which consider only word-internal contexts. Since these environments are not mutually exclusive, given that (t) may occur postliquid and preceding a word boundary, tokens were coded in such a way that both the preceding and following environment were considered. The examples shown in Table 1 are tokens produced by Frost in the current data set. (t) following /m/, /n/ parent 2.
(t) following a vowel little 4.
(t) preceding word boundaries treat These linguistic environments were selected because (t) is found to be consistently glottalized in many of these contexts in both Cockney and EE, and in at least some of these environments in RP (cf. Matthews 1938, Altendorf 2003. Although the above environments allow glottalization, they do not all favour glottalization equally. Altendorf (2003:91) ranks a subset of our linguistic environments hierarchically in terms of how much they favour glottalization: "preconsonantal position (gaTwick > pre-vocalic (quite_easy) and pre-pausal position (quiTe!) > pre-lateral position (bottle) > intervocalic position (buTTer)". Although the hierarchy was conceived to apply to Cockney and EE, the extent to which these linguistic contexts actually condition the realisation of glottal /t/ varies between varieties. For example, intervocalic /t/ is reported to glottalize in Cockney (Matthews 1938) but it is controversial whether EE includes intervocalic glottalized /t/ (Altendorf 2003:13). It is widely found not to glottalize in RP.
Additionally, previous literature outlines that even in non-regional pronunciations "pre-glottalization and glottal replacement very commonly affect a group of high-frequency words, namely : it, bit, get, let, at, that, got, lot, not (and contracted forms: don't, can't, aren't, isn't, etc.), what, put, but, might, right, quite, out, about" (Collins and Mees 2013:88). Since these high-frequency words are often glottalized across the board by speakers of different varieties, they were excluded from the present quantitative study in order to focus on words which may distinguish Cockney from EE and RP and to avoid potentially confounding variation. It is important to note that, although perhaps to a lesser extent than in British varieties of English, American varieties have been reported to include glottal stops as a possible variant of (t) in word boundary, pre-vowel environments such as shout at (Eddington and Taylor 2009). These particular linguistic contexts were also, therefore, excluded from the present quantitative analysis. Our study included a measure of lexical frequency to examine the potential effect of words which appeared more than five times in the corpus (Hay et al. 2010).
To test for the effects of the preceding environment, following environment, and lexical frequency on the realisation of (t), a general linear mixed model regression was implemented using the R Project for Statistical Computing (R Core Team 2016) and lme4 (Bates et al. 2016) using the lmer() function.

Overall Results
The results show that a high number (49%) of Frost's total realisations of (t) were glottalized. She also shows extensive intraspeaker variation with respect to this variable. Figure 1 presents the percentages of Frost's glottal and non-glottal realisations of (t) towards British and American adults and children. Our findings indicate an addressee effect in terms of age in the expected direction: adults heard more glottal realisations than children did. However, addressee nationality was not found to significantly affect the realisation of (t), although she does appear to produce slightly more glottal stops when addressing British children versus American children.

Statistical Analysis
In order to determine whether or not our observed addressee age effect was reliable as a predictor of glottalization, we used a general linear mixed effects model in R (R Core Team 2016). Word was included as a random effect. Fixed effects included addressee age, nationality, an interaction effect between age and nationality, and the linguistic environment following (t). The main effects are presented in Table 2, and the full model is presented in Table 3 (with significant factors and factors which neared significance in bold). (T)-realization across addressees Table 3. Levels of significant predictors

Addressee Age
The factor with the highest F value is addressee age ( Table 2; F = 17.2847) and within levels, relative to the intercept baseline, the addressee age is approaching significance (p = 0.096) as a predictor of glottal (t), as reported in Table 3. This suggests that Frost's productions of glottal and non-glottal (t) differ depending on whether the addressee is an adult or a child, with adults hearing more realisations of the glottal variant than children. Table 2 illustrates that, in our model, the following environment has a moderate impact on the realisation of (t) (F = 6.2569). Table 3 shows the levels of the factor "following environment" that differ significantly from the reference level, which is where (t) falls before a word boundary and subsequent consonant (as pre-vocalic instances of /t/ were excluded). Thus, intervocalic contexts significantly favour glottalization (p < 0.05) and contexts in which a vowel occurs after (but not before) (t) also significantly prefer glottal (t) (p < 0.05). Our result that a vocalic following environment was statistically significant as a predictor of glottalization is consistent with similar findings in the field (cf. Cruttenden 2014:90, Altendorf 2003.

Non-significant Factors
Addressee nationality was not found to be a significant factor in predicting Frost's (t) production. It is retained in the model because it was one of the main social variables considered in this study. The interaction effect between addressee age and nationality also did not reach significance. An evaluation of the possible reasons for this lack of significance will be undertaken in Section 5. Finally, Tables 2 and 3 omit preceding environment and lexical frequency. Neither factor was found to significantly correlate with glottalization. However, since a number of the "frequently glottalized words", which were purposefully omitted from our analysis, also happen to be highly frequent overall, it is possible that our measure does not reflect the true impact of lexical frequency on t-glottalization (Collins and Mees 2013:88).

Discussion
In summary, while our results are limited by a skewed number of tokens per condition, it is possible to see a trend in that Frost produced more glottal realisations of (t) towards adults than towards children. In terms of whether the addressee was British or American, no significant difference was found. This section now presents a qualitative analysis to obtain a more detailed account of the apparent lack of "nationality effect". Qualitative analysis offers a parallel view to quantitative analysis, which contributes to and enriches our understanding of the nuanced motivations for intraspeaker style-shifting in context.
In altering the way she speaks to children, Frost may be invoking the "clear", "careful", and "effortful" characteristics highlighted as being associated with released /t/ (Eckert 2008:469). In terms of CAT, this style shift represents convergence to the perceived needs of an interlocutor for the purpose of clarity, which is a "natural tendency in many of us", according to Levey and Harris (2002:18). Instead of viewing this observed shift in terms of moving along the dialect continuum from Cockney to EE, we are motivated to analyse this variation as a shift between "conversational speech" and "clear speech", or, more specifically, CDS (Smiljanić and Bradlow 2008 (1) illustrates Frost's only realisation of [ʔ] in the entirety of her speech to American children in our recordings. The token occurs while Frost is in the process of simultaneously disciplining Andra and teaching her mother Barbara how it should be done. Perhaps Frost's general shift to CDS is momentarily overridden by the duality of purpose or the emotionally charged nature of the act of "disciplining".
Excerpt (1) 1 Frost Andra...Andra 2 Sit /ʔ/ down 3 Firm voice Barbara 4 Sit /t/ down Moreover, Excerpt (2) outlines the rhetorical question directed towards the British child, Aidan, as Frost disciplines him after his parents' numerous attempts have failed. Frost's use of the glottal stop here contrasts with her typical use of [t] towards children and seems to index something like "exasperation", which is arguably related to the fact that she is in the process of implementing a discipline technique. Note that the token of (t) in the word "wait" was excluded from quantitative analyses as it occurs in a context which permits glottali zation in GAE.
Excerpt (2) 1 Frost Are we just going to sit /ʔ/ here and wait /ʔ/ until you're ready then?
Overall, in her speech to American and British children, Frost switches to a CDS-like register, producing more non-glottal [t] than glottal [ʔ]. Since the exceptions to an almost complete usage of [t] over [ʔ] in CDS appear to be limited to contexts in which "discipline" is the main focus, I advocate for further study into the interaction of speaker's emotional states with clear speech registers such as CDS.
However, if ease of understanding were the only important factor in determining style shift, then one might ask why Frost does not also significantly reduce her production of glottals in speech towards American addressees. Since glottalization is less common in GAE than in British varieties, it may be less easily perceived by American interlocutors, for which reason it was predicted to occur less often when Frost was addressing the Jeans family. It is important to note one caveat here: the analyses of addressee nationality mainly reflect Frost's speech to adults since there were more tokens realised towards adults than children in our data set. Nevertheless, addressee age does approach significance as a predictor of glottalization. The present study argues that Frost's motives for producing a high number of glottal realisations of (t) to British and American interlocutors may reflect a more speaker-initiated maintenance of difference for the purpose of preserving an authentic, and perhaps also performative, identity.
Maintenance, defined as "the absence of accommodative adjustments by individuals", appears to account for the lack of addressee-nationality effect (Giles and Gasiorek 2013:158). In both the UK and the USA episodes of Supernanny, Frost is accepted as the "expert", which may serve to minimize the pressure for her to speak a "standard" variety. This interpretation is reminiscent of that which was implemented by Shapp et al. (2014), who found that once Ginsburg was a Supreme Court Justice, she reverted to raising her THOUGHT vowel. Like THOUGHT-raising, t-glottalling conveys a field of well-established social meanings related to the characteristics of people who come from a given place. For example, English people who use glottalized (t) typically (although not exclusively) come from London and the surrounding areas, and may be perceived to be "uneducated" (Levey and Harris 2002:18). Cameron (2000:4) notes that there is an "increasing systematization of knowledge and codification of procedures for talking to other people". In other words, one channel through which "experts" assert their authority is the mandating of particular ways of using language (Cameron 2000). This concept is illustrated in Excerpt (3), in which Frost is teaching the American mother, Barbara, how to use her voice in different ways to discipline her children. Frost's status as the "expert" is reinforced by and reflected in this interaction. This excerpt includes tokens of the variable occurring in "frequently glottalized words" (Collins and Mees 2013:88) in order to present a more complete picture of the variation. Most notably, despite the highly metalinguistic purpose of her speech, Frost uses both variants, maintaining glottal realisations of (t) in "frequent" words, and producing non-glottal [t] in other words such as "literally" and "clarity". While Frost avoids glottalizing content words, perhaps as a result of the "formal" metalinguistic nature of the speech (cf. Labov 1972), her continued use of the glottal variant in frequent words enables her to retain a "familiar" stance in what could otherwise be a demeaning exercise for Barbara.
Barbara has really been struggling with discipline, so I'm going to be teaching Barbara how to project and how to maintain a low tone voice. I'd like you literally /t/ to say "No, stop that /ʔ/" 4 Barbara "No, stop that" 5 6

Frost
Okay, that /ʔ/ was good. That /ʔ/ was good for a first time. Just bring the voice out /t/ with more clarity /t/. 7 Barbara "No, stop that" 8 9

Frost
Do a happy, praise voice, so bring your voice up high-"Andra, that /ʔ/ is really fantastic" 10 Barbara "Andra, that is fantastic" 11 Frost "No, that's not acceptable" 12 Barbara "That is not acceptable" 13 Frost "That /ʔ/ is very good" 14 Barbara "That is very good" 15 Frost "Behave yourself" 16 Barbara "Behave yourself" 17 Frost "I don't want you to do that /ʔ/" 18 Barbara "Do not do that" 19 Frost <laughing> So brilliant /ʔ/!
In the UK recording, despite the fact that the British adults are PhD-educated university lecturers, and could thus be thought of as belonging to a higher social class than Frost, who started out as a working-class nanny, Frost retains an "expert" role. This position of authority appears to afford her the license to speak in her own way, thus freeing her from the overarching need to converge to a perceived standard . One exception to this apparent "freedom", however, is illustrated in Excerpt (4), which is taken from one of Frost's feedback meetings with the British parents.
Excerpt (4) 1 Frost The children's delayed speech development is due to the fact [that when they...

3
Stephen [So do you think it is delayed for 2-and-a-half-year-olds? 4 Frost Yes, I do think it is slightly /t/ behind, yes.
This particular turn involves Frost responding after having been challenged regarding her statement about the children's development. In this instance, it appears that Frost is utilising the "formal" connotations of a released [t] (Eckert 2008:469) to re-assert her authority. It is possible that Frost becomes momentarily aware that the father who challenged her point is a university professor with a PhD.
As noted by Eckert (2008), any linguistic variant may have a whole range of associated meanings depending on factors such as the interlocutors and the particular interactional context . Frost appears to exploit a variety of different meanings in the respective indexical fields of [ʔ] and [t] from familiarity, approachability, and exasperation (associated with [ʔ]) to carefulness, formality, and politeness (generally associated with [t]). Frost's maintenance of the Cockney/EE [ʔ] variant, as opposed to convergence to the supra-local standard [t] (Foulkes et al. 2005), is viewed here as a strategy to portray an "authentic" persona. Therefore, the non-effect of interlocutor nationality appears, at least in part, to be explainable as speaker-initiated avoidance of convergence for the purpose of constructing an authentic, authoritative identity, strengthened by her role as the "expert".
By the time the UK episode was filmed, Frost was well established in the UK as a parenting expert, whereas in the USA episode (which is the first in its season), Frost could be seen to be asserting her status as the "expert" to a new audience of viewers. This has implications for how we might interpret Frost's realisations (as well as her "failure to realise") (t) as glottal. Nevertheless, in both episodes Frost's on-screen identity is tied to the premise of the show, which is that she is capable of taming even the most unruly of children. For this reason, it is likely that the way she speaks to her interlocutors during the episodes, and thus her usage of particular variants, is an extension of that persona.

Conclusion
In conclusion, as was hypothesized, Frost was found to produce fewer glottal variants of (t) towards children. However, contrary to our initial predictions, no significant difference was found between British and American addressees. Frost's style-shifting to CDS has been interpreted in terms of CAT as a communicative strategy to enhance clarity. The lack of addressee-nationality effect appears to reflect avoidance of convergence for the purpose of constructing an authentic, authoritative identity. Frost achieves this through her exploitation of particular indexical meanings, more specifically, the "carefulness" and "formality" associated with [t] and the "approachability" and "exasperation" associated with [ʔ]. Through this research, we aim to have contributed to understandings regarding the extent and nature of the influence of addressee age and nationality on intraspeaker variation. The present study also highlights the need for more integrative approaches which are capable of encapsulating both broader listener-oriented shifts as well as more nuanced speaker-initiated variation. Future studies may wish to examine the effect of speakers' emotion on CDS in order to further test our tentative findings regarding "non-standard" pronunciations during the administering of discipline.