The Fricative Manoeuvre : Dialect Style-Shifting between Castilian and Mexican Spanish in YouTube Interviews with Musician

This study examines style-shifting in the case of Paul Banks, lead singer of the indie rock band Interpol, who spent a substantial amount of time during his adolescence in Spain and Mexico and therefore has access to their corresponding dialects. Our results show that Banks uses the dialect of the viewers during interviews. We see this as a strategy to index locality and solidarity in both Spanish-speaking regions and to convey aspects of his past and identity in order to win over viewers and fans.


Introduction
Competent speakers of any given language are able to vary the way they talk according to factors like their relationship with the interlocutor, communicative goals, and sociodemographic traits (Van Herk 2012:30), to mention a few.This paper aims to study a specific case of intraspeaker variation, which manifests as switching between different varieties of a language in this particular study.We have found the terminology to be somewhat problematic, as we were unsure of how to refer to the phenomenon under examination since the subject uses different varieties of the same language, rather than different languages per se.Given that the fundamentally different aspects of these varieties are mostly phonetic, we have opted to use the term style-shifting.
The speaker in question is the British-American celebrity Paul Banks, who lived in different Spanishspeaking countries during his childhood and adolescence and has therefore come into contact and acquired different varieties of Spanish (FaceCulture 2012).We predict that the subject will use a different variety of Spanish depending on which public he is trying to appeal to, whether consciously or unconsciously.We track this styleshifting by analysing the speaker's use of two different unvoiced fricatives that vary from one variety of Spanish to another.

Literature review
This study builds on work on speaker design (Schilling-Estes 1998).The main idea behind this theoretical approach to intraspeaker variation is that speakers are constantly using language to shape their social identities and present themselves in different ways.For example, our speaker could be indexing locality by using different varieties of Spanish that are tightly linked to place (as a social construct) and culture, much like Becker (2009), who shows how Lower East Siders of New York use non-rhoticity to index belonging to that place and their identity when discussing neighbourhood topics.
Among other relevant studies is the one conducted by Sharma (2011), who looks at style-shifting in two groups of second-generation British South Asians.The older group lived in Britain during a period of overt hostility towards immigrants, while the younger group grew up in different social and historical contexts, as a result of which they did not face such hostility.Sharma looks at their use of variants that index ethnicity, framing style-shifting as ethnolinguistic style repertoires that individuals form and utilise in order to index different social meaning.The concept of repertoire, as used by Sharma, consists of a set of phonetic and stylistic variables that can be manipulated by speakers to index ethnicity to a greater or lesser extent, depending on the social situation.One of the more striking cases is that of a male from the older group.He uses different repertoires depending on the ethnic background of the interlocutor: when speaking to the Indian interviewer and Sri Lankan maid, he uses Indian variants at much higher rates compared to the other interlocutors; with the British Asian interlocutors, he still uses the Indian variants but at a much lower rate; with the Cockney mechanic, on the other hand, he only uses British variants.Another case is that of a female from the younger generation.There seems to be a more clear-cut distinction in her style-shifting, as she only uses the Indian variants with her Indian mother and British Asian brother.Males of the older generation make more use of different repertoires than the females, while in the younger generation, the gendered pattern is reversed.Furthermore, the older male and the younger female subjects' style shifts are quite noteworthy, showing "wholesale bidialectal alternation" (Sharma 2011:479) similar to our subject's combined use of fricatives and other resources.Despite the fact that the current paper does not look at ethnicity, the principle of using repertoires is applicable in this particular case.Fought (2006:21) defines an ethnolect as a "pool of resources from which members of a speech community draw the linguistic tools they need" (see also Benor's (2010) paper on ethnolinguistic repertoire).
Another article that is key to this study is Schilling-Estes's (1998) paper analysing the highly performative speech in Ocracoke English of her subject, Rex O'Neil.Schilling-Estes looks at various instances of the same sentence illustrating phonetic variants closely associated to Ocracoke English, namely, the pronunciation of /ay/ with a backed and raised nucleus, as well as /I/ raising.Rex is recorded in three settings: while speaking to his brothers, talking to a fieldworker, and in a performative context where he performed the phrase for an interviewer.With reference to the work of Silverstein, Schilling-Estes states: When speakers attempt to "put on" a dialect for an audience, they have available to them only the features they can perceive; further, there is evidence that the greater perceptual awareness speakers have of a given language feature (whether this awareness is at the conscious level or not), the greater the extent to which the feature will figure in their demonstrations and discussions of the language variety in question.(Schilling-Estes 1998:64, referencing Silverstein 1981) According to this perspective, perceptual salience is a factor that affects the use of certain styles, the prediction being that the speakers are more likely to use the features of a style that they find more noticeable when styleshifting.
Hernández-Campoy and Cutillas-Espinosa (2010) examine the speech of a politician, María Antonia Martínez, from Murcia, a region in the southeast of Spain.Of interest is that her broadcast speech includes many features typical of the local Murcian dialect.Her speech was compared with that of other female Murcian politicians, male Murcian politicians, male non-politicians, and male politicians from Northern Spain as a control group.The analysis also involved different contexts varying in degrees of formality for each of the informants.Results from the study showed that in all contexts, María Antonia uses more local dialect features than members of any other group, violating expectations for gender (Schilling-Estes 2002), social class, and occupation.Coupland (2007) states that speaker design must also be audience design (Bell 1984), due to the fact that identity is intrapersonal and conveyed in an unfolding discourse.That is to say, audiences influence what aspects of our persona we wish to convey; therefore, even if the style-shifting is initiative rather than reactive, the audience still plays an important role.We have also considered audience design in our analysis (Bell 1984), where style shifts are primarily motivated by the addressee, although overhearers may have some effect, but to a lesser extent.In Bell's (1984) article, this phenomenon is fundamentally reactive.If it is the speaker who initiates the styleshifting or style shift it is more likely that they are converging to an absent group.Rickford and McNair-Knox (1994) investigated the way addressee and topic influenced style-shifting by an African American girl named Foxy.The data consist of two interviews conducted by one African American and one European American interviewer.The results showed that Foxy used fewer vernacular forms and was more standard with the European American interviewer.At the time, the study was one of the few testing the effects of the interviewer on the speech of the interviewee; recently, these studies have been receiving much more attention and it is a growing topic in the sociolinguistic literature.Another study in style-shifting and audience design by Blom and Gumperz (1972) involved the social meaning affecting style-shifting in Norway between the northern standard, Bokmal, and the northern dialect, Ranamal.Members of the community in which the study was conducted, considered these to be two distinct dialects that are never mixed.In their everyday communication, they made the choice based on social interaction: the speakers know, based on the audience, which dialect is appropriate in the given situation.
There are many more instances of studies on audience design, such as white Americans using African American Vernacular English to associate with Hip Hop culture (Bucholtz 1999), whereby the subjects shape their narrative according audience, or the study of African American drag queens and their use of white woman's language with African American English, gay language, and third person singular -s absence (Barrett 1995).All of these studies demonstrate Coupland's (2007) point of view that speakers present themselves differently through their use of language according to the setting and addressee.
Given that he spent a substantial amount of time in both Spain and Mexico, Banks is able to use both national varieties of Spanish depending on the context.Our hypothesis is that Banks uses the Peninsular variety during interviews conducted in Spain and by a Spanish interviewer who speaks Castilian Spanish, and the Mexican (or Latin American) variety when being interviewed by a Latin American interviewer in Latin America.Both interview and setting potentially affect style-shifting, but as the nationality of the interviewer and the country where the interview took place coincided in all interviews, we have not been able to discern which one has a greater influence or if it is a combination of both factors that prime the style-shifting.

Methods
For our data, we used five interviews found on YouTube (ieo18 2010, ieo18 2011a, ieo18 2011b, Jointheposse 2011, Rolling Stone Colombia 2015).All interviews were conducted in a non-formal style, which is indicated by the fact that all participants always used the second person non-formal pronoun tú rather than the formal counterpart usted.Two interviews were conducted in Spain and three in different countries in Latin America.In all of the instances the interviewer was local.We did not specify which countries in Latin America, as this had no impact on our data and Banks is not proficient in non-Mexican varieties of Latin American Spanish.We also analysed the effect of age and gender of the interviewers on Banks's style-shifting but they did not prove to be significant predictors, so these findings are not included in the results.The interviews were recorded in 2007, 2010, 2011, and 2015, with two interviews taking place in 2010.Each interview lasted between 2 to 12 minutes, and a short clip of Interpol playing live music interrupted some of the recordings.The total duration of the recordings was 29.5 minutes.
The first variable we examine in this study is the distinction between the voiceless interdental fricative /θ/ used in Castilian Spanish and the voiceless alveolar fricative /s/ present in Latin American Spanish.In Castilian Spanish, the voiceless interdental fricative is regularly used.However, in Latin American varieties, it does not exist; instead, the voiceless alveolar fricative /s/ is used.In other words, /θ/ is realised as /s/ in Latin American varieties of Spanish because they lack the unvoiced interdental fricative /θ/ in their phonological inventory (Navarro Tomás 1961:16).As a consequence, words that in Peninsular Spanish would be pronounced with the interdental fricative [θ] are pronounced with the alveolar fricative [s].The envelope of variation only includes tokens where there would be a difference between the Peninsular and Latin American variety of Spanish.For example, we looked at words such as hacer (meaning to do), since in the Mexican variety it would be pronounced as /aseɾ/ while in Castilian or Peninsular varieties it would be pronounced as /aθeɾ/, but we excluded words like son (third person plural of the verb ser, meaning to be) because there are no differences in pronunciation from one variety to another (apart from possible differences involving prosody).This variable was coded for auditorily by two coders, with an initial agreement of 96.16%.Upon closer inspection, both coders agreed on 100% of the tokens.
The second variable we focus on is the frontness of the voiceless velar fricative in Castilian Spanish /x/, as in some varieties of Latin American Spanish the voiceless velar fricative /x/ can be articulated further back or more to the front of the oral cavity, i.e., [x̙ ] or [x̘ ] (Vaquero de Ramírez 1998:43-45).In the case of Mexican Spanish, Moreno de Alba (1994:105) specifies that the voiceless velar fricative /x/ is generally realised as more fronted [x̘ ] than its Castilian counterpart, becoming postalveolar in some cases.To code for this variable, we used Praat (Boersma and Weenink 2015) to measure the intensity in frequency and the centre of gravity of the velar fricative.Figure 1 provides a sample of the process by which we extracted the centre of gravity, which is the frequency that halves the spectrum into high and low frequencies and represents their mid-point.By using this measure, high energy frequencies will have higher centre of gravity values.Therefore, the lower the value, the more fronted the articulation.One of the interviews was discarded due to background noise that interfered with acoustic measurements.Only tokens were counted in which the fricative in question was followed by a front vowel.

Results
The number of tokens in total are 77 for the /θ/-/s/ distinction and 20 for the velar fricative.For the first variable (instances of /s/ that would be pronounced as [θ] in Castilian Spanish), we show the findings in Figure 2, and for the second variable (frontness of /x/), we show the findings in Figures 3 and 4.  Figure 2 shows the frequencies of realisations of the different variants of the dependent variable with respect to the setting in which the interviews take place.The part of the bar coloured red represents the percentage of instances of the alveolar fricative [s], while the blue part of the bar represents the interdental fricative [θ].We can see that, for the most part, Banks uses the alveolar fricative [s] in Latin American contexts and the interdental fricative [θ] in Spanish contexts: he uses the alveolar fricative [s] 12.5% of the time in Spanish contexts and 97.96% in Latin American contexts, and he uses the interdental fricative 87.5% of the time in Spanish contexts and 2.04% of the time in Latin American contexts.While Banks robustly style-shifts between contexts, he is relatively more variable in Spain than in Latin America.The results for the second variable show that Banks's velar fricatives are articulated more to the front in Spanish settings (2700 Hz) than in Latin American settings (2300 Hz).There is also much more variation in production in the Spanish than in the Latin American contexts, as indicated by the length of the box in the boxplot (one standard deviation from the mean).It should be noted that the number of tokens for the second variable is quite small.As a result, it is quite difficult to conclusively determine a trend or a correlation between the setting of the speaker and the value of the centre of gravity of the fricatives he produces.

Discussion
We can see that there is very clear intraspeaker variation in Banks's use of fricatives, given that he almost exclusively uses the interdental fricative [θ] in Spanish settings (87%), while he reserves the alveolar fricative [s] for Latin American settings (98%).This difference unsurprisingly obtained statistical significance (chi-squared = 54.972,df = 1, p < 0.001).While there is clearly a significant change in fricative production brought on by a change in setting, from the available data it is unclear whether it is a change in interviewer, location, public audience, or a combination of all three factors.The intriguing aspect of this analysis is how to account for sustained use of Castilian Spanish throughout an entire interview but the use of Latin American Spanish in a different interview.Obviously, it is not a matter of fluency in either of the two varieties, due to Banks having acquired both before the interviews were conducted and having been exposed to academic and social contexts in which he had to make use of them to study and interact with others.While this does not guarantee fluency, it is clear from the interviews that Banks has quite a high level of Spanish, and switching to one variety or another should not, therefore, be problematic.
Even if we tried to account for Banks's style-shifting using Labov's (1966) attention-paid-to-speech model, we would find that Banks is converging to a style of speech that seems to be affected by his setting.This suggests that there is more to his style-shifting than the mere dimensions of formality.This leads us to believe that audience design (Bell 1984) and speaker design (Schilling-Estes 1998) are both more suited for explaining the phenomenon at hand, especially taking into account Coupland's (2007) view on both models that audience design has to be speaker design.Hernández-Campoy and Cutillas-Espinosa (2010) state that their subject of study, María Antonia Martínez, uses Murcian dialect features in order to index social and political values, given that the Murcian dialect carries covert prestige and is a marker of solidarity and locality.Perhaps the reason why Banks would use different varieties of Spanish depending on context is to index similar values.
The style-shifting can be explained if we take speaker design theory into account (Schilling-Estes 1998).By using a different variety of Spanish in a given context, Banks is conveying different aspects of his identity and past, as well as indexing locality and solidarity with his viewers in order to reach them on a more familiar level.It is interesting to observe that during one of his interviews in Spain, Banks discusses and openly reflects on the time he spent in Madrid and the Spanish city of Alicante while using the Castilian variety of Spanish.
As stated, we looked for two different variables to exemplify this.The first was the use of /s/ and /θ/.In this case, our results were significant: Banks almost exclusively used /s/ in the Latin American contexts and /θ/ in the Spanish contexts.For the second variable, the point of articulation of the voiceless velar fricative /x/, the distinction is not as robust: setting does not always predict variation.This could be due to the fact that this variable is not as perceptually salient to the speaker or the hearer as the first one, which could be considered one of the most perceptually salient features of both Castilian and Latin American Spanish, for which reason it represents the most significant or obvious choice to display convergence towards the target audience (Schilling-Estes 1998).
The combined use of these features, including but not limited to the ones analysed in this study, points to a repertoire employed by Banks to index locality and solidarity with viewers and/or his interlocutor, much like the participants in the Sharma (2011) study.This could also explain why the second variable was not as robust as the first one: perhaps Paul Banks only modifies it in some of the interviews, thus skewing our results.Aside from the phonetic features mentioned, Banks also uses occasional expressions that are more associated with one of the varieties of Spanish.A particularly salient example is the use of the expression tomar una foto, which literally means to take a picture and is often used in Latin American Spanish, contrasting with the Castilian hacer una foto, which literally translates as to make a picture.
It is also interesting to see that in an interview attended by a bandmate who does not speak Spanish, Banks still prefers to speak in Spanish, despite his bandmate not understanding it.Perhaps he finds it easier to accurately express his thoughts to a Spanish-speaking interviewer in Spanish than in English, since different languages underlay different meanings (Hyltenstam and Obler 1989).In this interview, in particular, Banks happened to use English words while speaking in Spanish, which did not occur in any other interview.This could possibly be due to the presence of his bandmate, who under Bell's (1984) model functions as an auditor.
We can see this phenomenon as Banks indexing locality by switching into the variety of Spanish spoken by the public.We can see other individuals applying different strategies to achieve the same effects, as seen in Becker (2009), where Lower East Siders of New York used rhoticity, and in Hernández-Campoy and Cutillas-Espinosa (2010), where María Antonia Martínez used features typical of the local Murcian dialect.Different individuals or groups use the means that are available to them to index different social meanings and shape their social identities.
Even though our main focus is on speaker design, it is possible to account for the phenomenon with audience design (Bell 1984), in which the audience influences style-shifting or code-switching, such as in Rickford and McNair-Knox (1994), where their informant, Foxy, used more standard forms with a European American interviewer than with an African American one.Banks, as a famous singer, is aware of the kind of personality he projects and thus chooses the local variety to seem more authentic or sympathetic.However, if we take this explanation into account, Banks could be converging to the interlocutor-that is to say, the interviewer-or the viewers watching him from their homes, or both.It is unclear from the data we have which of these is the significant predictor or if it is a combination of the two.
Another possibly interesting aspect of our results is that Banks atypically uses the alveolar fricative [s] in Castilian contexts more than he atypically uses the interdental fricative [θ] in Mexican or Latin American contexts.While this could be caused by an inclination to use the Latin American variety of Spanish, it is difficult to determine due to the low number of tokens.
When it comes to studies regarding famous personalities, the question of authenticity and nature of their speech is highly scrutinized.Whether the subject is making conscious effort or not to accommodate to the audience is controversial, because even if they are converging or diverging consciously, we cannot be sure this is the case.Another problem could be the question of the variables selected for studies and why we should be concerned with them.In our case, in one of the interviews, Banks makes a direct reference to fame, being famous, and being always in the spotlight.It is noteworthy to see him talking about fame while converging to the regional variety, as this could show his awareness of his speech and the effort of projecting a likeable persona by adjusting his speech towards the interviewer.Alternatively, he could find it easier to use the same variety as the interviewer or he may require little effort to adjust his dialect because he is able to smoothly switch from the one to the other.His awareness of his fame and status, and by extension of how other people perceive him, might not necessarily be reflected in his speech.

Conclusion
To summarise, in this study we have examined how a non-native speaker of Spanish changes his variety of Castilian or Mexican Spanish according to context with respect to velar voiceless fricative frequency.Our results have shown that he makes a clear distinction between /θ/-/s/ according to the interview.He may do so to highlight aspects of his identity and past in order to better appeal to the public.For /x/-[x̙ ]/[x̘ ], the results are not as clearcut: using the centre of gravity, we observed that the distinction is much less than expected and on a much smaller scale when compared with the first variable.We believe that this is due to the difference between varieties of Spanish being of lesser salience than in the case of the first variable.
What would strengthen our study, ideally, would be to have more data for both of the variables examined.The availability of more data could aid us in teasing apart which factors specifically condition his style-shifting.Other contextual information from which we could gain insight would concern his personal background, as it is clear here that his upbringing has affected the way he uses both Castilian and Mexican Spanish to converge to the local variant of the interviewer.
Most research in this area is focused on speakers shifting in one target language.It would be interesting to see more studies of bilinguals and their change of accents in different languages and what forces drive them towards shifting or not shifting in a different language.

Figure 3 :
Figure 3: Centre of gravity values in Spanish settings.

Figure 4 :
Figure 4: Centre of gravity values in Latin American settings.