Goin’ Home: The Role of Vowel Raising in Indexing an Ethnic Identity

Vowel height can be seen as a feature indexing a particular ethnic identity or indicating a style shift between two such identities. This paper focuses on Bradford-born South Asian musician Zayn Malik, an interesting subject given his prominent status as one of the most influential Asians in Britain today (Parveen 2016). I consider the role of vowel raising in the GOAT lexical set (Wells 1982) over the period of 2010 to 2016 and in style-shifting between casual, interview, and performative styles. The results of my study indicate significant differences in F1 and F2 values between particular years, and differences in F1 values between performative and interview speech. I explore these shifts in relation to the hostility Malik has faced and study the causes of his shifting, particularly in terms of speaker agency and audience design (Bell 1984).


Introduction
While the relationship between style-shifting and ethnicity is a complex one, there are varieties with particular features that have been found to index ethnic identity.There has been some research (Watt andTillotson 2001, Wormald 2014) over the past 15 years that has looked at a few of the features of both the Anglo-Bradford variety of English and what I shall refer to as the Asian-Bradford variety, which encompasses a wide range of features.There is also a body of relevant research on other, similar varieties, such as British-Asian English varieties spoken in Southall (Sharma 2011) and in Glasgow (Stuart-Smith et al. 2011).My study focuses on one speaker, the Bradford-born third-generation South Asian Zayn Malik, and shows that he has generally shifted to producing a higher, more Asian-Bradford realisation of the GOAT lexical set (Wells 1982) over the course of his career.I also find that he uses this variant more in interview-style speech than he does in performative speech.
At 4.9% of the country's total population at the time of the last census in 2011, South Asians make up the largest ethnic minority in the U.K. today ("British Asian" n.d.), with concentrations in areas such as London, Manchester, and Bradford, the latter of which is often referred to as "Bradistan" because of its large population of people of Pakistani descent.As of the last census, 26.83% of the population of Bradford considered themselves Asian or Asian-British, second only to White (67.44%)("Demography of Bradford" n.d.).As noted by Wormald (2014:118), 4% of individuals recorded Panjabi as their main language in the same census, where "Panjabi-English is the term used to refer to the native-English variety spoken by second-and future-generation speakers with Panjabi language heritage".The Panjab region encompasses North West India and Northern Pakistan, which includes the area of Pakistan that Malik's parental family comes from.Even outwith the linguistic literature, the Anglo-Bradford and the Asian-Bradford varieties are clearly distinct, as can be noted, for example, in a BBC study conducted on language attitudes: "You can tell when people are from Bradford, especially from the Asian community.[…] I think a Asian-Bradford person and a Bradford-Bradford person has a different accent" (BBC Voices 2005).

Literature Review
The correlation between language and ethnicity is not a new topic.Fought (2006:21-3) outlines the various ways in which language can be used to index one, or multiple, ethnic identities, commenting that "some features are so closely tied to ethnic identity that a single use of that feature can serve to identify a speaker as belonging to a particular group" (2006:22).For example, Stuart-Smith et al. (2011) look at the GOAT lexical set in Anglo-Glaswegian and Asian-Glaswegian speakers with a focus on the second formant (fronting or backing).Tokens of the GOAT lexical set were seen to be further front in the vowel space in Asian-Glaswegian speakers than in Anglo-Glaswegian speakers, suggesting a positive correlation between vowel fronting and the indexing of the Asian-Glaswegian ethnic identity.
The phonological inventory of Bradford English, in general, is similar to that of most of Yorkshire and the North.Features include vowel reduction, non-rhoticity, t-glottalisation, and devoicing (Hughes et al. 2012:104-5).The vowels of Bradford English seem to have undergone a noticeable shift over the last 30 years.Summarising previous work (Wright 1892, Orton and Dieth 1963, Petyt 1985, Hughes and Trudgill 1996) on the GOAT vowel, Watt and Tillotson (2001:210) noted "a levelling or reduction of localised phonemic or sub-phonemic contrast in the vowel(s) of lexical items now apparently collapsed into a class corresponding to Wells' RP-based GOAT set".More recently, Hughes et al. (2012:105) describe the GOAT lexical set as having largely reduced from contrastive [ɔʊ] and [o:] to [o:] in all phonetic contexts.Similarly, the typical realisation of the FACE lexical set is monophthongal [e:].Both Watt and Tillotson (2001) and Wormald (2014) note the monophthongal quality in tokens of the GOAT lexical set in all varieties of Bradford English.
Between varieties, Wormald's study of women from Bradford shows that Bradford Panjabi-English speakers had "significantly lower F1s" (2014:118), i.e., higher GOAT vowels, than Bradford Anglo-English speakers.Wormald's (2014) study presents an interesting contrast to the work of Sharma (2011), who observed that females shifted almost exclusively to the diphthongal "British" variant, considered to have more prestige than its monophthongal "Asian" counterpart.While research into the phonological inventory of the Asian-Bradford variety is limited to specific lexical sets, Wormald (2014) considers the impact of Panjabi on the realisation of English phonemes for Asian-British speakers.She suggests that /e/ and /o/ in Panjabi may be relatively close in the vowel space, which could indicate some impact on the position of the GOAT and FACE lexical sets in the vowel space of the Asian-Bradford variety.
Analysing this variation at an intraspeaker level requires consideration of several possible theories.Schilling (2013:339) describes "speaker design", which focuses on "how speakers use variation to fashion themselves and their surroundings".Speaker design is a collection of approaches, some of which fall into what Eckert (2012) terms the Third Wave of sociolinguistics.From this perspective, we might say that the speaker studied in the present paper will be seen to fashion an identity for himself through the use of style-shifting.Other types of style-shifting could also be considered, such as age grading (Labov 1994): when a speaker shifts over time from one variant to another (perhaps simultaneous with speaker design effects).Typically, evidence of age grading is clearest for longitudinal studies.Given the short period of time over which data have been collected in the present paper (6 years, spanning 2010 to 2016), there may be no obvious evidence of this kind of shift.However, those 6 years include teenage years, which Wagner (2012:184) suggests is a period of significant variation in terms of language use: "adolescents in this life stage are now often forced to establish their identity in relation to new acquaintances and become more overtly aware of sociolinguistic variation".
In contrast to speaker design is Bell's (1984) audience design model, which frames style-shifting as a response to the speaker's audience, made up of addressees, auditors, and so forth, with the degree of effect lessening for each of those members the further the audience member gets from the speaker.Taking the audience into account in the present paper allows for an analysis of how and when the speaker may shift as a response to his addressees as a single, ratified person or, on a much larger, performance scale, where the audience is made up of thousands of people.
The contexts of the style-shifting being considered here are two-fold: the first is a contrast over the 6 years between 2010 and 2016; the second is a contrast between speech style, that is, casual versus interview versus performative, with a particular focus on the latter two styles.Casual style encompasses speech recorded between Malik and his family and him and his bandmates in a non-interview, non-performance setting (unsurprisingly, this is the category with the least amount of available data).Interview style encompasses all speech within available interviews (typically conducted with one or more of his bandmates present in the years up to 2015 and alone after that).Performative style refers to any speech in a performance setting, such as onstage or directed to the camera.
The results show that the speaker shifts towards a more ethnically marked variant of the GOAT lexical set, a higher vowel with a lower F1 value, over the course of his career.This may be related to his adjustment to new linguistic markets and to the scrutiny his career brings from the media and the public as a whole.Further, he displays a higher vowel in interview-style speech than in either casual speech or performative speech, which can be understood in terms of a shift across speech styles.Like Sharma (2012:127), I will argue that "agency in such variation cannot be assumed" and focus on aspects of audience design (Bell 1984) in my analysis.

The Speaker
Zain "Zayn" Jawadd Malik was born January 12, 1993 in East Bowling, Bradford.His mum, Trisha, is English and converted to Islam when she married Yaser Malik, a second-generation British-Pakistani.Malik was brought up as a Muslim, speaking both English and Urdu, alongside his three sisters.He has a large extended family in England on his father's side.He has stated that he did not fit in at either of his first two schools due to his mixed heritage (Yeung 2015).
Malik was part of the popular boyband One Direction from July 2010 until March 2015, when he quit to pursue a career as a solo artist.Unlike the other members of One Direction, who are all White-British, he has faced extreme racism and Islamophobia over the years from the mass media and through social media: "Zayn's existence as a Muslim and Pakistani pop star has never been without commentary, from those both inside and outside his ethnic and religious communities" (Anifowoshe 2016).Despite this, he has frequently spoken of the importance of his background and religion to him.In April 2015, he received an award for Outstanding Contribution to Music at the Asian Awards, thanking God and his parents for making him Asian.His South Asian background has clearly had a large impact on his life.It has been said that, for third-generation immigrants like him, "South Asia functions not as 'home' but as a more abstract representation of ethnic identity.It defines their diasporic identity, their ethnicity within British society" (Hussain 2005:132).
Malik has displayed metalinguistic awareness of his own speech: in particular, an awareness of the fact that he cannot always be easily understood.On stage in Osaka, Japan, he observed: "People normally find it hard to understand what I'm saying anyway so you guys have absolutely no chance" (azu xo 2015).The Internet is full of examples of people struggling to make sense of what he is saying, including comments on videos about everything from general confusion over what he has said to "that…that wasnt english" (zaynft1direction 2015).The influence of Bradford on his accent has been described as "unmistakable" (Cooper 2015).
He also seems to be aware-if not of how or why-that he does style-shift.In a BBC Radio 1 interview discussing One Direction's documentary film This Is Us, he commented: "I noticed that, like, depending on who we were talking to in the film, our own accents changed a little bit.That was a bit strange.Like, when I was talking to mum, like, on the phone, my accent went loads more Yorkshire.[…] Bit more Bradford" (BBC Radio 1 2013).

Methodology
The data consist of recorded footage of Malik between 2010, when he first entered the public eye as a contestant on the U.K. reality television show The X Factor, to mid-2016.This footage encompasses various interviews conducted within both the U.K. and elsewhere with the other members of One Direction and by himself; footage from the 2013 documentary This Is Us; and Malik's portion of the 2014 audiobook Who We Are.The recordings were obtained from Internet sources and digitised using Soundflower (Ingalls 2012) and Audacity (Audacity Team 2013), with tokens of the GOAT lexical set analysed manually in Praat (Boersma and Weenink 2013).To ensure reliability of my data, all of the coding was cross-checked between three trained raters, with an overall inter-rater reliability of 80%, rounded to the nearest 10Hz.
The variety of footage allowed me to analyse the path of stylistic variation between 2010 and 2016.In selecting recordings from different contexts, I was also able to analyse the stylistic variation present in different kinds of speech in different environments.During analysis, I also coded for the referent of the token and whether the recording was made in the U.K. or not (if known), but these factors did not help explain the variation and so will not be discussed further.
There are a few points to note about the recordings used.The first is that many of the recordings are taken from a context where Malik is talking alongside his four bandmates, and there are often instances of him being spoken over, especially as the quietest of the five.Often his utterances were mumbled or hard to make out and, in some cases, accurate tracking of the F1 and F2 values was not possible.These tokens were omitted.In addition, all tokens with a duration of less than 0.05 seconds were omitted from the data.The second issue to note is the disproportionate nature of the data collected: far more data were found for 2012 than for 2010, for example, and a comparatively small amount of data were collected for the casual speech type.
The dependent variables coded for were the F1 (height) and F2 (backness) in tokens of the GOAT lexical set, which are continuous variables.In RP English, the GOAT lexical set is realised as the diphthong [əʊ] moving from the central schwa to the near-close near-back rounded vowel.The formant measurements were taken from the midpoint of the vowel.
In Bradford English, the GOAT lexical set is usually realised with a back monophthong, "such that boat, for example, may be homophonous (or at any rate nearly homophonous) with bought" (Watt and Tillotson 2001:270; italics in original).Almost categorically, Malik's tokens of the GOAT lexical set are realised as monophthongs, as I would expect given the previous research (Watt andTillotson 2001, Wormald 2014).On the very few occasions that a diphthong or something closer to a diphthong was produced, the results were omitted from my data set.

Results
The two independent variables coded for that yielded noticeable results were year of the recording in the 2010-2016 span and style, i.e., casual, interview, or performative.F1 values in the tokens range from 156.4Hz to 1355.3Hz;F2 values range from 741.1Hz to 2986.1Hz.Typically, a high vowel (e.g., /i/) is taken to have an F1 value of around 240Hz, while a back vowel (e.g., /u/) is taken to have an F2 value of around 595Hz (Catford 1988:161).
In coding for the linguistic environment, I found that the following environment did not have a noticeable effect on either the F1 or the F2 of the vowel being analysed, as shown in Table 1.  ).While neither formant displays a continuous trend in either direction, the data for F1 are a little more consistent than those for F2, which vary vastly from year to year.In performing one-way ANOVA tests on F1 and F2 values by year, both produced significant results (p < 0.05).For F1, there was a significant difference between 2012 and 2014, between 2014 and 2015, and between 2014 and 2016; for F2, there was a significant difference between 2012 and 2014 only.

Style
Table 2 shows the mean F1 and F2 values for each of the three styles of speech coded for, as well as the distribution of data across styles.

317
My expectation was to see a continuum, with the highest vowel in casual and the lowest in performative speech, i.e., that the speaker would index ethnic identity more when in a relaxed environment than when onstage in front of thousands of people.This cannot be seen in Table 2, but it is worth noting the contrast between the mean F1 for interview speech and the other two styles.This shows that the speaker uses a higher vowel overall in interview than he does in either casual or performative speech.The mean F2 shows very little difference between interview and performative styles but a much lower value for casual speech, which again is not what I would have expected, since a higher F2 indicates a further fronted vowel, which correlates with indexing of ethnic identity.As found with the anomaly of the 2015 data in 4.1, there is a very small amount of data for casual speech, which could mean it is less representative of that style than the others.
In performing one-way ANOVA tests on F1 and F2 values by style, I found there to be a significant result for F1 but not for F2.Within style, it was the contrast between performative and interview speech that was significant.The contrast between interview and performative styles is demonstrated in Figure 2: the two styles show a similar pattern of change over time, with interview data following after the perfomative data but with a noticeable difference in vowel height.There is an overlap in 2011 and 2013 where there is no difference in styles, but the two styles diverge in 2012 and again from 2014 onwards.

Discussion
In my introduction, I put forward the claim that higher tokens of Malik's GOAT lexical set would indicate his use of a more ethnically marked variant, which would be more like a feature of the Asian-Bradford variety.Over the course of the past 6 years, Malik has had to very quickly adjust to living under constant media scrutiny-arguably the most out of the band members, as the only non-White artist.
The subtle but steady shift between 2010 and 2014 away from the more Anglo-Bradford variant of the GOAT lexical set to a higher, arguably more ethnically marked variant, could be seen to follow with his adjustment to the level of scrutiny from the media and the public as a whole, and as a form of age grading.Despite his widespread popularity within the media, the amount of racism targeted at him has not lessened over the years.For example, his departure from the band in March 2015 led many news outlets to question whether he was leaving to join the terrorist group ISIS.However, his ability to handle this racist commentary has grown stronger.Rather than trying to blend in among his four White bandmates, it could be argued that "he's separating his identity from the pallid whiteness that previously bled out his other dimensions" (Roisin 2015).Therefore, it is possible that he has adjusted to being seen as a Muslim South Asian-British figure and indexed himself to that identity through his language use, whether consciously or not.
However, in 2015, his vowel is seen to lower again quite dramatically, which veers off from the pattern that can be seen through the first 4 years of the time period.The vowel raises again in 2016 but not so much as to return to where it was in 2014.There are various factors that could be considered in addressing this unexpected pattern.Firstly, the amount of tokens varies greatly from year to year, as can be seen in Table 3. Table 3 also demonstrates the highly disproportional spread of tokens across the three styles identified, which could also account for some of the unexpected results.The insufficient data for certain styles in certain years (e.g., casual and interview styles in 2010) make it impossible to draw any conclusions about style-shifting within those groups.Furthermore, the period of 2015 to 2016 represents a time during which the speaker underwent a massive change in terms of his career and personal life as well as how he was choosing to present himself to the media and to the general public.This could also have had some impact on his variation, whether at a conscious or subconscious level.This brings me to the idea that speaker design (Schilling 2013) may be in play: a set of theories centred around the notion of agency in a speaker's use of variants.The contrast between Malik's use of the higher GOAT variant in interview speech and the lower variant in performative speech could be seen as agentive from the point of view of speaker design: perhaps, Malik is focused on how he presents himself in performative contexts, highlighting his celebrity persona over his ethnic identity.However, in interview speech, he is almost always interviewed alongside at least one of his bandmates by an interviewer who is often someone he has worked with in the past.One might expect him to be less attentive of his linguistic performance in this type of intimate setting.
Style could also be analysed in terms of Bell's (1984) theory of audience design, according to which the addressee in interviewer speech would be the interviewer, while Malik's bandmates and members of their PR team in the room would function as auditors.Those viewing the interview broadcast would, in these terms, function somewhere between overhearers and eavesdroppers: they are unaddressed, unratified, and the speaker will be aware of them to some extent but not necessarily of how many of them there are or who they are.Similarly, in performative speech, the addressees could be abstractly seen to comprise a very wide range of people, such as an audience of 20,000 people.The difference, however, is that the addressees are present and observable, unlike the audience being broadcast to in interview speech.When presenting himself in a context with a much wider scope of possible addressees, it could be argued that the speaker shifts to using a less marked variant-the Anglo-Bradford lower variant of the GOAT lexical set.
While Sharma and Sankaran (2011) note more style-shifting in older men than in younger ones, they found that men who had experienced hostility or bullying, as Malik has and still does, style-shifted more.Those younger men that experienced hostility had experiences closer to those of the older men.They required both "an ability to downplay Indianness and to pass as British" when in predominantly non-Asian circles and competence in the use of ethnic variants to maintain their connection to their Asian roots for "social survival" (Sharma and Sankaran 2011:423).Malik's linguistic behaviour could be seen to shift in a similar way to meet the opposing pressures of the predominantly non-Asian industry he works in and the deeply Asian roots of his home community in Bradford.This could account for his use of various ethnolinguistic repertoires when speaking.
However, the results suggest that the speaker is comfortable shifting between a variant that is more Anglo-Bradford and one that may be more ethnically marked.He also evidently seems quite aware of the effect of his hometown-an area that is ethnically marked in itself-on him as a person, which appears to have left a lasting impact on his speech and style-shifting.Malik acknowledges the influence of his background, saying: "The whole vibe of Bradford is influential […] There's a lot of character there.There's a lot of strong family values.Everybody's very proud, and everybody's stuck in their ways.That rubbed off on me a little bit and made me a stubborn person, and made me very aware of who I was.If you weren't aware of that in Bradford, you kind of got left behind" (Cooper 2015).That his upbringing and background would have had an effect on his speech seems very plausible.

Conclusion
Through studying musician Zayn Malik's production of the GOAT lexical set, I have been able to consider when and why the speaker may index his South Asian ethnic identity.I have considered this variation in terms of the timespan during which he has been a successful figure very much in the media and the public eye and his different kinds of speech: predominantly, the contrast between interview and performative speech.Future research might analyse the ways in which Malik does or does not make use of various ethnolinguistic repertoires, as "a fluid set of linguistic resources that members of an ethnic group may use variably as they index their ethnic identities" (Benor 2010:159).The present paper shows some evidence of his intraspeaker variation.In future research, it would be interesting to undertake a longer longitudinal study with more second-or third-generation South Asian speakers from the Bradford area in order to observe style-shifting in to the linguistic marketplace, considering whether there is a significant shift in the period of their lives when they are beginning their career and forging new relationships, as opposed to their childhood and adolescence and their later, post-working years.

Figure 1
Figure1plots the formants of Malik's GOAT lexical set by year over the time period of 2010 to 2016.There appears to be an overall downward trend in F1 between 2010 (mean F1= 630Hz) and 2014 (mean F1 = 528Hz), indicating that the vowel was raising over this time period.In 2015, the F1 generally seems to increase suddenly (mean F1 = 722Hz), showing the vowel being realised as a lower vowel.However, this could, in part, be due to the limited amount of data available for 2015 (i.e., only two recordings, both of which were performative).The vowel is higher again in 2016 (mean F1 = 629Hz) with lower F1 values, but not as high as during the 2011-2014 period.Meanwhile, F2 shows a fronting trend between 2010 (mean F2 = 1204Hz) and 2012 (mean F2 = 1302Hz).Malik then produces the vowel further back in 2013 (mean = 1164Hz) and 2014 (mean F2 = 1156Hz), fronting the most in 2015 (mean F2 = 1442Hz), and then backing again in 2016 (mean F2 = 1233Hz).While neither formant displays a continuous trend in either direction, the data for F1 are a little more consistent than those for F2, which vary vastly from year to year.In performing one-way ANOVA tests on F1 and F2 values by year, both produced significant results (p < 0.05).For F1, there was a significant difference between2012 and 2014, between 2014 and  2015, and between 2014 and 2016; for F2, there was a significant difference between 2012 and 2014 only.

Figure 2 :
Figure 2: Mean F1 over time by type of speech.

Table 1 :
Mean F1 and F2 values by following linguistic environment

Table 2 :
Mean F1 and F2 values by style

Table 3 :
Tokens of GOAT lexical set by year and style