Toeing the Party Line: Indexicality and Regional Andalusian Phonetic Features in Political Speech

: Performative style is an important sociolinguistic variable among politicians, who accomplish agentive goals through speech. Examining 32 Spanish politicians, this article focuses on four Andalusian Spanish phenomena: the fronting of / > t S / and the deletion of coda /s/, resyllabiﬁed intervocalic onset /s/, and intervocalic /d/. The analysis ﬁrst looks at overall community production norms for the variants then turns to examine the style-shifting patterns of one individual who deviates from these norms. This individual is examined through a consideration of lectal focusing in interaction to track moment-by-moment variation. While coda /s/ and intervocalic /d/ deletion show usage patterns governed by regional and contextual factors, the deletion of onset /s/ and fronted / > t S / reﬂect social variation and style-shifting. While politicians do not blindly follow partisan norms, normative expectations exist at the regional level that they can choose to depart from due to individual motivations and political afﬁliation in order to carry out identity work. This study combines quantitative examinations of community and individual variation to contribute to our understanding of style-shifting behavior in political speech and how politicians use linguistic tools to take on oppositional identities in the public sphere.


Introduction
In recent years, increased sociolinguistic interest has been turned toward public speech as a context for tracking agency and intention, including public figures in the English context (Bell 1984;Coupland 2001;Hay et al. 1999;Sharma 2018), as well as politicians (e.g., Harrington 2007;Hall-Lew et al. 2012;Podesva et al. 2015;Kirkham and Moore 2016;Holliday 2017). This includes Sharma and Rampton's (2015) analysis of what they term lectal focusing in interaction (LFI), which permits an examination of individual variation across time. Recent advances in sociolinguist theory, such as that demonstrated in the work by Ahearn (2001) and Eckert (2008), suggest that speaker agency plays a role even at the phonetic level to further goals related to identity construction. While all speakers can style-shift between a multitude of linguistic features, including regional and normative variants and those that exist across a variety of registers and communities, public speakers in particular must appeal to listeners using carefully crafted personae based on linguistic choices.
In the realm of political speech, there have been interesting findings, particularly in Southern Peninsular Spanish, that have pointed to the ways in which stylistic decisions intersect with individual identity through the use of regional and normative features associated with southern and central varieties of Spanish, respectively. For example, the former president of Murcia, Spain, María Antonia Martínez, indexed solidarity with working-class speakers through what Hernández-Campoy and Cutillas-Espinosa (2010 have described as a "hyper-vernacular" use of regional features. The authors argued that this linguistic behavior allowed her to carry out identity work situating her as a socialist Languages 2023, 8, 196 2 of 24 and as a woman of the people. Pollock and Wheeler (2022) found something similar for the former Andalusian president, Susana Díaz, whose use of regional variants shifted over time as she established herself in her presidency. Interestingly, Díaz received considerable national attention for speech patterns described as uneducated and rural, despite only deviating in certain contexts from the Andalusian female speech norms established in this study. These deviations, however, were found to be highly salient. Both of these female politicians, holding some of the highest positions of power in the country, were nonetheless perceived negatively in the press and on social media based on their use of traditionally male-favored regional features (Pollock and Wheeler 2022) 1 .
Other researchers have approached the matter in the peninsular context using other approaches and across other parts of Spain. In a study of Extremadura, Spain, Fernández de Molina (2021) found that regional variant use differs across speakers, suggesting a highly individualized use of language that reflects differing political goals. Meanwhile, through a diachronic approach, Cruz-Ortiz (2019, 2022) compared Andalusian politicians across almost 90 years, finding that they tended to converge with Madrid linguistic norms while speaking in the capital, showing how speakers of regional varieties modify their phonetic features to conform to national norms. Both of these studies also emphasized the importance of individual agency and stylistic variation as an important differentiating factor in variable phonetic production.
The present study follows in this vein of research, expanding the lens from individuals to a synchronic community of speakers while still focusing on stylistic variation. The central part of the data analysis identifies political speech norms in Andalusian Spanish, setting it in contrast to the Spanish of speakers from the capital, Madrid. As Cruz-Ortiz (2022) and others have done previously, this provides insight into the norms of this peninsular community of practice. The second part of the study, however, focuses on the type of individual variation that has often been cited as a key factor in previous research, using quantitative and qualitative methodologies to determine how a single individual differs from the community, and what his motivations may be. As Hernández-Campoy and Cutillas-Espinosa's (2010) case study shows, Spanish voters can view hyper-vernacular usage among politicians as a sign of both uneducated status and solidarity, sometimes simultaneously. This study contributes to our knowledge of sociophonetic variation and third-wave linguistic theory by determining how politicians navigate this complex indexical field of social meaning through the implementation of regional phonetic resources and then tracking how this plays out for one politician at the individual level.
This article serves as a contribution to the existing literature on stylistic variation and political discourse, as it tracks variable regional norms within Peninsular Spanish speaker groups and probes the stylistic variation that politicians employ to perform identity.

Linguistic Style and Political Speech
As the field of sociolinguistics has developed, so too have conceptualizations of linguistic style. In its earliest formulation, Labov (1972) described it as the "degree of attention paid to speech", with more careful speech being less likely to carry vernacular variants. As time passed and new waves of thought came to the fore, style changed shape as well. For Bell (1984), under the audience design theory, stylistic decisions should also encompass tiered rings of intended and possible hearers, from addressees to eavesdroppers. Based on an analysis of news presenters speaking to different audiences, Bell theorized that differences in style can arise based on expected social values of speech phenomena and whether they converge or diverge from the norms of an expected audience. Speakers' stylistic variation is produced to account, to some extent, for these levels of interlocutors.
In more recent years, internal stimuli have become particularly important to the classification of a sociolinguistic understanding of linguistic style. Authors like Coupland (2001) and Eckert (2008) have discussed concepts related to speaker design and social practice that place the focus on individual motivations stemming from social meaning and the specific speech context in which language is produced. Similarly to Bell, Coupland Languages 2023, 8, 196 3 of 24 examined public speech, focusing on a Welsh radio DJ who uses stereotypical local linguistic variants to discuss familiar topics, stylized American English ones when cuing up rock songs, and standard productions in general speech. Rather than shifting based on expected changes in audience, as there are none, the speaker changes as a means of taking on a stance toward differing matters, revealing solidarity with his region, as well as a familiarity with rock culture and a professional radio demeanor. Eckert (2008), in turn, saw these stylistic decisions as agentive, serving as part of a performative process of identity formation that goes beyond language. Everything from language to clothing serves as part of the social meaning-enriched bricolage used by speakers to associate themselves with certain groups and diverge from others.
It is important, in the examination of political speech, to strike a balance in our understanding of linguistic style. While speakers from manifold walks of life have access to a variety of linguistic tools to employ in the construction of identity, politicians and public speakers are also beholden to the normative expectation of the media through which they speak. Hernández-Campoy and Jiménez-Cano (2003) showed how public-facing radio hosts speaking Peninsular Spanish are at times limited by these normative expectations, termed a "script", which presents them with an unmarked option of following a channel's norms without deviation, leading to potential blowback. Politicians are faced with a similar challenge, although they have incentives to at times set aside the script: while a misstep may hurt their image, they must also protect their voters' interests and develop a consistent identity through stance accumulation, which can, at times, clash with expected speech norms.
This tendency to set aside the script is clear in Hernández-Campoy and Cutillas-Espinosa's (2010, 2013) examination of the 1990s Murcian president María Antonia Martínez. In their 2010 study, the authors analyzed politicians from Murcia and Madrid, as well as non-politicians from the Murcian community, finding that Martínez's rate of using Murcian variants outweighed that of community members and politicians. For Murcian variants, including lenition of coda /s/ and intervocalic /d/, elision of coda /r/ and /l/, and consonantal assimilation, she used a "hyper-vernacular" pattern of production across four years of analysis. Hernández-Campoy and Cutillas-Espinosa (2010, p. 307) saw this behavior as a means for her to intentionally employ "local Murcian features to achieve a particular effect related with identity construction", reinforcing her identity as Murcian, formerly working class, and a socialist. In a follow-up interview with Martínez in 2013, the authors found that she was aware of her language use as a means of accomplishing political goals, including reaching left-leaning voters. Her discussion was contrasted with news media and politician comments describing her, at the same time, as uneducated and ill suited to give public speeches but also as hardworking and trustworthy. This tendency is frequent among varieties like Murcian (or Andalusian: Villena-Ponsoda 2013) that are outwardly stigmatized but possess covert prestige at the regional level. All told, these studies combine Murcian identity, rural speech, and socialist ideals, hinting at the rich indexical landscape confronting progressive female politicians, with which they index both positive solidarity and negative rurality.
In a follow-up study, Pollock and Wheeler (2022) examined how a former president of Andalusia, Susana Díaz, used regional variation to navigate southern peninsular indexicality. The authors determined that intervocalic /d/ and coda /s/ elision increased in scripted speeches late in Susana Díaz's presidency, deviating from her peers. Further, they drew on third-wave sociolinguistic research to discuss the dynamic nature of gender performance, extending it beyond a static category (e.g., Eckert 1989;Podesva 2007;Zimman 2017). In the particularly salient case of intervocalic /d/ elision in the participial ending <-ado>, Díaz is more extreme than politicians and community members of both genders, eliding almost categorically. The authors argued that the politician uses this context as a means of reflecting her militant status in the Socialist party and as the first female president of Andalusia. Díaz's self-representation both received broad support from constituents on social media while at the same time being the subject of mockery in terms of her perceived Languages 2023, 8, 196 4 of 24 hyper-vernacularity by members of the national political establishment, despite using comparable rates of regional variants in most cases to her Seville peers.
In her study of political speech by Andalusians from 1923 to 2011, Cruz-Ortiz (2022) presented a similar result for <-ado>, indicating that it has long been a context that favors /d/ elision among politicians. Furthermore, based on comparisons of speech contexts, the author determined that Andalusian politicians converge with Madrid linguistic norms while speaking in the capital, except in the case of the coda /s/, in which articulatory weakening predominates, suggesting that, for the most part, Andalusians modify their phonetic features to conform to national norms. However, she also found that the broader political climate in Spain (i.e., the alignment of the governing party) does not influence regional speech use by Andalusian politicians, with usage rising and falling over time.
Alongside Spanish, there have also been studies that examine political speech in the English context. With regard to the production of diphthongs by Scottish National Party politicians, Hall-Lew et al. (2017) found that speakers followed party norms, using Scottish variants rather than London ones with distinct F1 and F2 levels. The authors described this as a means for these politicians to distinguish themselves from their opponents in the Scottish Labor Party, who instead favored a more southern-sounding variant. Hall-Lew et al. (2012) found a similar trend in American politics: in the lead-up to the invasion of Iraq in 2003, US Democrats and Republicans differed in their production of <a> in the name of that country. Speakers were found to be divided by no other social measure than party line, with Republicans almost categorically using the "disparaging" [ae] production and Democrats favoring the "European" [a]. As these productions came from a series of speeches across a very short time frame, the authors noted just how remarkable it was that speakers from vastly different geographic and political backgrounds united behind these differing productions based on their party affiliation.
Others working with political speech in English include Holliday (2017), Kirkham and Moore (2016), and Podesva et al. (2015). Holliday has focused on coronal stop deletion (CSD) in Barack and Michelle Obama's speech. She found that Barack's CSD patterns more closely mirror Mainstream US English, while Michelle's are closer to African American Language (AAL), but both demonstrate stylistic variation that sets them apart from these varieties. The resulting "informal" production pattern that they share, Holliday argued, serves as a means of seeming approachable on television. Kirkham and Moore (2016) examined /t/ glottalization and verbal variation by former British Labour Party leader Ed Miliband. They found nuanced variation between speeches for two audiences, with differences corresponding to the ideological orientation of his listeners, demonstrating his awareness of ideological and linguistic differences between listeners. Meanwhile, Podesva et al. (2015) performed a production and perception study of released /t/ in the speech of 2000s-era US politicians, including John Edwards and Condoleezza Rice. In these English examples, linguistic variation occurred not due only to the region, gender, or age of the speaker but also following partisan lines and performative decisions.
Overall, public speech has played an important role in the advancement of sociolinguistic theory surrounding speech style. There has also been a growing number of studies on political speech in recent years, particularly in Spanish, which suggest that this context offers fruitful opportunities for continued examination. Beginning with the innovative research of Hernández-Campoy and Cutillas-Espinosa (2010, these studies offer insight into our understanding of identity work, public speech norms, and the implementation of style-shifting behavior in Spanish.

Andalusian Spanish Features
In this study, the Spanish capital of Madrid is treated as a control group representing aspects of Northern and Southern Castilian Spanish, here labeled Northern Central Peninsular Spanish (NCPS; as described by Moreno Fernández (2009), Henriksen (2014), and others). The capital variety tends to be rather conservative, although intervocalic /d/ elision is frequent and aspiration of the coda /s/ can be found in certain parts of  Vida-Castro 2020;Pollock and Wheeler 2022;Pollock 2023;etc.). In order to distinguish how AS politicians use their language, four variable contexts were identified that differ across regions and offer politicians a context for identity construction and style-shifting behavior.
The first of these is perhaps the most studied linguistically, as well as one of the phenomena most identified with AS: coda /s/ deletion. In his description of phonetic variation in Peninsular Spanish, Samper-Padilla (2011, p. 100) discussed three common allophones of /s/-the elided [Ø], the aspirated [h], and the retained [s]-and mapped them onto geographical areas. Southern AS favors elision, Northern AS and the Canary Islands have higher rates of aspiration, and NCPS shows high rates of retention. The behavior of the pre-vocalic coda /s/ also merits mention. Due to resyllabification in Spanish, a word-final /s/ in pre-vocalic position before a word beginning with a vowel can become an onset /s/ (e.g., los otros ("the others") [lo.so.tRos]; Torreira and Ernestus 2012;Hualde and Colina 2014, p. 74). Despite the onset position, pre-vocalic environment, and low tendency for lenition of these contexts in NCPS, elision is not infrequent in AS for this phenomenon.
At the city level in Andalusia, research has shown that aspiration and elision are characteristic of Granada (e.g., García-Marcos 1990;Rincón-Pérez 2015), Linares (e.g., Gómez-Serrano 1994), and Malaga (e.g., Vida 2004), among other areas. Fernández de , in an examination of Western AS from Merida and Badajoz, found aspiration to be most common in interviews, although elderly speakers favored elision. Fernández de Molina (2018) determined, in a diachronic follow-up study, that the younger generation favors elision more than in the past, suggesting an ongoing spread of the phenomenon. In Eastern AS, Tejada-Giráldez (2015) showed that Granada speakers from the PRESEEA corpus were four times more likely to elide than aspirate, with under one percent retention. This nearly categorical rate of elision means that it can often be difficult to determine social differences in AS coda /s/ production.
Madrid, the political center of the country and a representative of NCPS, has been identified with equal frequency as using the conservative variant, with various studies finding generally low rates of /s/ weakening (e.g., Blanco 2004;Gil-Peña 2004;Ruiz-Martínez 2003). Furthermore, in NCPS, Samper-Padilla (2011, pp. 106-7) and Ruiz-Domínguez (1997) suggested that women retain more and aspirate less than men and that working class and older speakers retain the least, implying a correlation between retention and social prestige in this variety. With respect to linguistic factors, Samper-Padilla (2011) and Martín-Butragueño (1995) found elision to be more likely in multisyllabic words, in word-final and pre-pausal positions, and when morphological redundancy permits deletion. Aspiration is expected in word-final position before a vowel and word-internally before consonants. Overall, it is expected that NCPS should show considerably higher rates of /s/ retention than AS, in which the innovative elided and aspirated variants are much more common (García-Marcos 1990;Vida 2004).
The second phenomenon that diverges regionally is the post-alveolar affricate / > tS/. Much discussion of variable use revolves around the de-affricated post-alveolar fricative [S], which is favored by elderly working-class and rural speakers in southern AS varieties, such as in the cities of Seville, Granada, and Malaga (Herrero de Haro 2017; Melguizo-Moreno 2006; Samper-Padilla 2011). Outside of this phenomenon, little acoustic analysis has been carried out for the normative peninsular affricate, which is generally considered stable (Samper-Padilla 2011). However, no instances of de-affrication have been found in the corpus of political speech; instead, an alveolar fronted variant [ > ts] has been identified that has received relatively little attention outside of Chile (Flores 2014(Flores , 2018, where it is explained by contact with aboriginal languages like Huilliche, which contrasts / > tS/  (2022), differs quantitatively in its center of gravity (COG, Hz) from that in post-alveolar production and could result from analogy to the affrication of /st/ clusters described by Del Saz (2019) in AS. This raises interesting questions regarding possible neutralization between minimal pairs (e.g., hecho ("fact") [e > tso] and esto ("this") [e > tso]). As of yet, there are no social or linguistic descriptions of this allophone, meaning that this study will provide insight into variation that exists between regional and national varieties.
The final phenomenon under consideration, intervocalic /d/ elision, is widespread throughout Spain (and the Americas), although it is seen as most common in AS. The Madrid norm, described by Gil-Peña (2004) and Ruiz-Martínez (2003), is the dental approximant allophone [ð fl ], although it is important to point out that Cruz-Ortiz (2022) and Wheeler (2022, 2023) have associated intervocalic elision with political speech norms throughout Spain, especially in the participial <-ado>. Molina-Martos and Paredes-García (2014) found elision in Madrid to occur most frequently in the suffix -ado ("past participle -ed"). The rate of elision was higher when speakers talked about personal topics, during exposition and narratives, in adjacent pairs, in colloquial speech, near the end of the interview, and among men and members of the youngest and oldest generations. Other sociolinguistic research has shown that women tend to be more conservative in Madrid in relation to this phenomenon, including in the Salamanca district (Gil-Peña 2004) and in Alcalá de Henares (Blanco 2004). Men, on the other hand, tend to have higher rates of elision, as documented in Linares (Gómez-Serrano 1994) and Toledo (Molina-Martos 1998). While elision of /d/ does also occur in other parts of NCPS, it is typically less frequent (Antón 1998).
Elision of the intervocalic /d/ is rather common in AS, such as in Jaén (Moya-Corral 1979) and Linares (Gómez-Serrano 1994). Villena-Ponsoda and Moya-Corral (2016) found that elision is also advancing in Granada and Malaga. Elision is favored by young speakers in western AS cities, like Malaga (Villena-Ponsoda et al. 2011) and Granada (Samper-Padilla 2011, p. 114). In general, speakers with less formal education and from lower socioeconomic backgrounds tend to elide the most (e.g., Gómez-Serrano 1994;Pérez-Martín 2003;Ruiz-Martínez 2003). Although common throughout the country, intervocalic /d/ elision appears to be an innovative vernacular form that exists throughout Spain but is most common in particularly southeastern AS (Samper-Padilla 2011).
Based on these previous studies, it is expected that these sibilant and stop phenomena will serve as viable representations of AS norms, with some more minor variation between eastern and western AS. Given the typical sociolinguistic profiles of vernacular variants (e.g., Labov 1972), the social meaning associated with these aforementioned phonetic studies would suggest that the elision of coda /s/ and intervocalic /d/ is tied to rural, working class, and male identities, although more recent research in political speech (e.g., Cruz-Ortiz 2022) and variationist analysis of Andalusian Spanish (e.g., Villena-Ponsoda and Vida-Castro 2020) indicates that there is considerably more nuance. In certain contexts, elision of the intervocalic /d/ and the coda /s/ has a wider variety of meanings that move beyond masculinity, rurality, and working class meaning. As politicians select variants most fitting for their identity goals, the overt meaning these forms possess serves to influence their choices in the construction of regional solidarity and identification.

Research Questions
Research into phonetic properties of Peninsular Spanish shows clear differences between central and southern Spanish variants 2 . As Bell's (1984) and Coupland's (2001) theoretical frameworks suggest, politicians are aware of these differences and adjust to their interlocutor based on external and internal pressures related to the public persona they target. Based in this foundation, two questions guide the analysis of regional and party norms in this study. First, does production of the four regional phenomena follow linguistic and social norms established in the previous literature? Hernández-Campoy and Cutillas-Espinosa (2010) have observed that progressive politicians use features associated with working-class speech in Murcia more often than conservatives. Meanwhile, Cruz-Ortiz (2022) showed that variation in regional usage can occur over time independently of the ideology of the governing party, suggesting that other style-shifting behavior may play a role in variation. Party and other social factors were expected to influence how politicians accommodate toward NCPS Spanish, while regional variation was also expected to have a strong effect on individuals' identity goals (e.g., a male Madrid politician would likely have little incentive to develop a southern identity, while an Andalusian politician may benefit from presenting themselves as more normative or standard).
Second, based on these quantitative tendencies and an individual analysis of LFI in the vein of Sharma and Rampton (2015), how is individual style-shifting used to differentiate speakers and carry out identity work? As politicians pursue linguistic goals, the four dialectal phenomena can provide insight into stylistic goals, as well as showing how social meaning is conveyed within the indexical field of AS and NCPS for Andalusian politicians.

Corpus
In order to better understand political speech in Andalusian Spanish, 32 participants were selected who formed a sort of political community of practice, sharing similar positions in the upper echelons of national and regional government. These individuals included presidents, senators, and mayors who were vocal in Spanish news broadcasts, presenting themselves in varying ways based on their social and political backgrounds. The selected participants were balanced by gender and came from the two largest political parties in Spain and Andalusia, the progressive Partido Socialista Obrero Español (PSOE, "Spanish Socialist Workers Party") and the conservative Partido Popular (PP, "People's Party"). Individuals were from four cities: three in Andalusia (i.e., Malaga, Seville, and Cordoba), representing relatively eastern, western, and northern locations in the region, and Madrid, representing the NCPS variety. See Supplementary Table S1 for a full list of speakers and information about speech contexts.
The audio was collected between 2011 and 2019, during most of which time the PP was the national governing party and the PSOE was the regional one in Andalusia. Interestingly, however, both parties shifted around the same time, with a corruption scandal in the national PP leading the PSOE to take national control in mid-2018 and the elections in early 2019 electing an Andalusian President from the PP. Thus, despite relatively long-standing political stability, this corpus was collected from speakers in the lead-up to major political change across the nation. With the PSOE losing control in Andalusia for the first time in over three decades and populist parties like Vox on the rise throughout the country, this presents an interesting political environment for further linguistic study (Rama et al. 2021).
Politicians in the corpus were born within 150 miles of the city they represented and served in government between 2014 and 2019. Audio included three speech contexts: scripted speech, unscripted interviews with male interlocutors, and unscripted interviews with female interlocutors. Scripted speech was expected to be more carefully chosen than the more extemporaneous interview speech, and gendered differences were chosen based on previous findings in Spanish public speech analysis (e.g., Flores 2014) that speakers vary based not only on their own gender but also that of their interlocutor, with women using more prestige forms with other women and men using more innovative and vernacular forms with other men. To avoid the collection of overly formal speech common early in speech contexts, coding began four minutes into each audio file (Díaz-Campos et al. 2018).

Dependent Variables
Across the four phenomena that vary between NCPS and AS described in this article, Samper-Padilla (2011) has discussed a spectrum of normative to innovative productions  Table 1 for coda /s/ elision, wordfinal onset /s/ elision, affricate fronting, and intervocalic /d/ elision. These variables were coded acoustically using spectrographic data in Praat, including the absence of aperiodicity in the waveform for /s/, the spectral center of gravity (Hz) of frication in the affricate, and drops in intensity for /d/ (Boersma and Weenink 2022). Spectrographic representations demonstrate the ways by which tokens were identified and distinguished. Figure 1 shows a consistent amplitude and lack of breaks in periodicity across the vowels [i], [a], and [e] in comunidades ("communities"), which indicates elision of both instances of /d/. The aperiodicity seen at the end of the word in the range around 4000 Hz on the waveform indicates a fricative production of /s/ as [s].

Dependent Variables
Across the four phenomena that vary between NCPS and AS described in this article, Samper-Padilla (2011) has discussed a spectrum of normative to innovative productions from central to southern Spain. Examples are shown in Table 1 for coda /s/ elision, wordfinal onset /s/ elision, affricate fronting, and intervocalic /d/ elision. These variables were coded acoustically using spectrographic data in Praat, including the absence of aperiodicity in the waveform for /s/, the spectral center of gravity (Hz) of frication in the affricate, and drops in intensity for /d/ (Boersma and Weenink 2022). Spectrographic representations demonstrate the ways by which tokens were identified and distinguished. Figure 1 shows a consistent amplitude and lack of breaks in periodicity across the vowels [i], [a], and [e] in comunidades ("communities"), which indicates elision of both instances of /d/. The aperiodicity seen at the end of the word in the range around 4000 Hz on the waveform indicates a fricative production of /s/ as [s]. In podemos ("we can") in Figure 2, there is a clear lowering of intensity between the first two vowels of the word, indicating that /d/ was produced as the approximant [ð̞ ], while the lack of aperiodicity at the end of the word points to either elision of /s/ or potentially aspiration as a voiced [ɦ]. In podemos ("we can") in Figure 2, there is a clear lowering of intensity between the first two vowels of the word, indicating that /d/ was produced as the approximant [ð fl ], while the lack of aperiodicity at the end of the word points to either elision of /s/ or potentially aspiration as a voiced [H]. Finally, Figure 3 offers an example of variable production in the Spanish affricate from the phrase el muchacho ("the boy"). While the first instance of /t͡ ʃ/ has a high center of gravity, with noise beginning near 4000 Hz, the second is much lower, with noise beginning around 2500 Hz. Gordon et al. (2002), in an examination of typologically distinct languages, found /s/ on average to have a higher COG than /ʃ/. The considerable difference in this Spanish example indicates a difference in tongue position when producing the two affricates: while the first is produced nearer the alveolar ridge (i.e., [t͡ s]), the second is closer to the palate (i.e., [t͡ ʃ]).  Finally, Figure 3 offers an example of variable production in the Spanish affricate from the phrase el muchacho ("the boy"). While the first instance of / > tS/ has a high center of gravity, with noise beginning near 4000 Hz, the second is much lower, with noise beginning around 2500 Hz. Gordon et al. (2002), in an examination of typologically distinct languages, found /s/ on average to have a higher COG than /S/. The considerable difference in this Spanish example indicates a difference in tongue position when producing the two affricates: while the first is produced nearer the alveolar ridge (i.e., [ > ts]), the second is closer to the palate (i.e., [ > tS]). Finally, Figure 3 offers an example of variable production in the Spanish affricate from the phrase el muchacho ("the boy"). While the first instance of /t͡ ʃ/ has a high center of gravity, with noise beginning near 4000 Hz, the second is much lower, with noise beginning around 2500 Hz. Gordon et al. (2002), in an examination of typologically distinct languages, found /s/ on average to have a higher COG than /ʃ/. The considerable difference in this Spanish example indicates a difference in tongue position when producing the two affricates: while the first is produced nearer the alveolar ridge (i.e., [t͡ s]), the second is closer to the palate (i.e., [t͡ ʃ]).  Once tokens were coded based on production type, certain categories of variation were combined to permit binary quantitative analysis to compare the AS and NCPS varieties. In the case of the coda and word-final onset /s/, production of fricatives [z] and [s] was grouped into a normative category, while lenited allophones included [h] and elision. Only two productions of the affricate, both alveolar and post-alveolar, were identified; the post-alveolar [ > tS] was treated as the normative production, and the fronted alveolar [ > ts] was described as the regional variant. Finally, for the intervocalic /d/, only the normative dental approximant [ð fl ] and the elided [Ø], described as the regional variant, were identified.

Independent Variables
Based on the descriptions provided by the previous literature, linguistic and extralinguistic variables were coded for the phenomena (Table 2). Following the spectrum of innovative productions described by Samper-Padilla (2011), city was selected as an extralinguistic variable, with the expectation that Madrid would represent the highest proportions of nationally standard forms, while Seville and Malaga (as representatives of western and eastern AS, respectively; e.g., Fernández de Molina and Hernández-Campoy (2018, p. 513); Herrero de Haro and Hajek (2022) in the south would have the highest rates of innovative regional AS features and Cordoba to the north would be medial to the other three, with a higher rate of normative features than other AS varieties (Samper-Padilla 2011). Next, speaker gender was selected, given the frequent correlation between male speech and innovative variants for /s/ (Ruiz-Domínguez 1997; Samper-Padilla 2011, pp. 106-7) and intervocalic /d/ (Gómez-Serrano 1994;Molina-Martos 1998) in comparison with the tendency in studies of peninsular political speech for female politicians to go against these trends (e.g., Hernández-Campoy and Cutillas-Espinosa 2013; Pollock and Wheeler 2022). Following that, political party was included based on the hypothesis proposed by Hernández-Campoy and Cutillas-Espinosa (2010) that left-leaning politicians use regional features as a means of indexing rural, working-class identity. Finally, the speech context was selected as an extralinguistic variable based on findings by Flores (2014) and Bell (1984), who have demonstrated how one's direct audience and environment can influence production. Direct speeches were expected to have some of the highest rates of normative features, given the national audiences these speakers were addressing, while interviews were directed toward male and female interlocutors, presented a more relaxed atmosphere, and tended to be produced on regional and local channels.
These linguistic variables selected in this study were chosen based on their mention in the previous literature. Instances of /s/ were coded for position, distinguishing between word-final intervocalic and coda positions, with the expectation that more elision would occur in the latter case, while a following vowel would be more likely to maintain [s]. Monosyllabic forms were expected to be less likely to be elided because of the reduced information they contain. The variables of preceding and following phone were included to track phonological processes that encourage elision, with elision expected to be most frequent in pre-consonantal position. Morphemic status was included, distinguishing contexts where /-s/ was used to mark plurals and verb endings from other contexts. Linguists studying variationism, such as López-Morales (1983), Samper-Padilla and López-Morales (1990), and Vida (2004), have studied these contexts and determined that /-s/ tends to appear more often when not semantically necessary, such as in interior position and in monomorphemic words.
Similar linguistic variables were selected for /d/ and / > tS/. This included the preceding and following heights of the phone, which were hypothesized to influence the likelihood of different production types. Molina-Martos and Paredes-García (2014) have shown that /d/ is more likely to be elided when following /a/ and preceded by /o/ in the suffix -ado ("(past participle) -ed")-as reaffirmed in political speech by Cruz-Ortiz (2022)which is demonstrated by vowel height. Although there is little discussion of Spanish affricate fronting, lower vowels were predicted to move the tongue's apex further from the alveolar ridge, potentially altering phonotactics. While lexical stress was analyzed for / > tS/, as stressed syllables were expected to be more enunciated and favor the normative production, /d/ was analyzed for lexical status, as previous studies have argued that participles and adjectives are more likely to favor elision than other contexts (e.g., Cruz-Ortiz 2022; Pollock 2023).
Finally, the politician was classified as a random effect in the mixed-effects logistic model to account for the expectation in this type of model that each token is independent. This allowed similarities across the productions of a single individual to be acknowledged in the model. The independent factors are presented in Table 2.

Data Collection and Analysis
Data were analyzed using a mixed-effects logistic regression model in the Rbrul software for R (version 4.0.2, Johnson 2009). Based in the variable-rule approach of sociolinguistics, mixed-effects models reduce imbalance across naturalistic data coming from multiple speakers producing multiple tokens, accounting for imbalances that fixed-effect models do not through the use of random effects (Johnson 2014). These models provide three types of data that describe how factors influence variation, as discussed by Tagliamonte (2012). First, factors that are significant predictors of variation that go beyond the level of chance are indicated by a p-value less than 0.05. Second, the magnitude of effect is shown through a representation of the factor-weight range based on the difference in logodds between factors. Third, the direction of effect is represented by both the factor weight and log-odds for individual variables, representing the degree to which variables favor one variant over another. Factor weights go from 0 to 1, with weights below 0.5 disfavoring use of the application value and those above 0.5 favoring it.
The models produced for each of the phenomena indicated social and linguistic factors that played a role in speaker behavior. The selection of these models was carried out through comparisons of AIC and log-likelihood values, favoring simpler models with fewer factors where possible. Following that, the next step in the analysis involved identifying and studying a single individual's linguistic behavior in greater depth, determining how they both followed and deviated from the norms established in their community. Given what previous studies investigating political speech have found (e.g., Hernández-Campoy and Cutillas-Espinosa 2010; Hall-Lew et al. 2012Pollock and Wheeler 2022), politicians were expected to trend along regional and political lines, with socialists in particular using linguistic features that help them indicate their solidarity with the working class.

Results
In total, 50 instances of coda /s/, 25 of intervocalic /d/, 25 of intervocalic onset /s/, and 10 of / > tS/ were coded for all 32 speakers in each of the three speech contexts (i.e., scripted speech, interview with male interlocutor, and interview with female interlocutor). Together, this yielded 10,010 tokens (i.e., 110 tokens × 32 politicians × 3 contexts). These tokens were coded dichotomously as "regional" or "normative" variants based on the identification and classification described in the methodology section to permit binary treatment in the logistic regression. To avoid token-based effects, no more than four repetitions of any token were coded for a given speaker. Figure 4 offers a brief look at the division of tokens based on the four dialectal contexts, as well as the tendencies for speakers to produce the normative versus regional variant. Intervocalic /d/ and / > tS/ were least likely to be produced with regional variants. By dint of their relatively infrequent use, this could also suggest that, when they are used, they have a higher degree of saliency in speech. When individual speakers used the elided or fronted variant, it provided vital information about the variable, whereas in the case of both the coda and word-final onset /s/, the categorical elision of some speakers reduced variation. Four mixed-effects logistic regression models were developed to determine which factors played a role in conditioning regional versus normative feature usage within the corpus.
weight and log-odds for individual variables, representing the degree to which variables favor one variant over another. Factor weights go from 0 to 1, with weights below 0.5 disfavoring use of the application value and those above 0.5 favoring it.
The models produced for each of the phenomena indicated social and linguistic factors that played a role in speaker behavior. The selection of these models was carried out through comparisons of AIC and log-likelihood values, favoring simpler models with fewer factors where possible. Following that, the next step in the analysis involved identifying and studying a single individual's linguistic behavior in greater depth, determining how they both followed and deviated from the norms established in their community. Given what previous studies investigating political speech have found (e.g., Pollock and Wheeler 2022;Hall-Lew et al. 2012; Hernández-Campoy and Cutillas-Espinosa 2010), politicians were expected to trend along regional and political lines, with socialists in particular using linguistic features that help them indicate their solidarity with the working class.

Results
In total, 50 instances of coda /s/, 25 of intervocalic /d/, 25 of intervocalic onset /s/, and 10 of /t͡ ʃ/ were coded for all 32 speakers in each of the three speech contexts (i.e., scripted speech, interview with male interlocutor, and interview with female interlocutor). Together, this yielded 10,010 tokens (i.e., 110 tokens × 32 politicians × 3 contexts). These tokens were coded dichotomously as "regional" or "normative" variants based on the identification and classification described in the methodology section to permit binary treatment in the logistic regression. To avoid token-based effects, no more than four repetitions of any token were coded for a given speaker. Figure 4 offers a brief look at the division of tokens based on the four dialectal contexts, as well as the tendencies for speakers to produce the normative versus regional variant. Intervocalic /d/ and /t͡ ʃ/ were least likely to be produced with regional variants. By dint of their relatively infrequent use, this could also suggest that, when they are used, they have a higher degree of saliency in speech. When individual speakers used the elided or fronted variant, it provided vital information about the variable, whereas in the case of both the coda and word-final onset /s/, the categorical elision of some speakers reduced variation. Four mixed-effects logistic regression models were developed to determine which factors played a role in conditioning regional versus normative feature usage within the corpus.

Coda /s/
The tendencies for coda /s/ elision included a more marked difference between AS and NCPS. As a regional identifier of southern Spanish on the peninsula, it occurred frequently in political speech, potentially as a highly salient means of expressing regional belonging, although less frequently in public speech than has been found in previous studies examining semi-directed private interviews among Andalusian speakers. These results are presented in Table 3. While no social group of NCPS speakers showed less than Languages 2023, 8, 196 13 of 24 80% retention of coda /s/, Andalusians tended to retain /s/ in fewer than 20% of cases. Conservatives in Cordoba and Seville retained less than socialists, while the opposite was true in Malaga. Women retained more in Madrid and Cordoba, while men tended to retain more in Malaga and Seville. While regional norms hold true, political affiliation presents a less clear image, with conservatives in Cordoba and Seville using higher rates of regional features than socialists, despite the expectation that this behavior would correlate with working-class norms and progressive politics. In the mixed-effects logistic model, four factors were determined to condition variable use of coda /s/ (Table 4). First, the city was most predictive of variation, with the highest range in factor weights: speakers from Madrid favored retention, whereas those from the three Andalusian cities disfavored it markedly. Next, for following phone type, pre-pausal /s/ and /s/ preceding stops favored retention, while those following sonorants disfavored it. Third, morphology was predictive of retention in that coda /s/ marking the morphemic plural or verbal second person (e.g., chicos ("kids") or tienes ("you have")) disfavored retention, while all other contexts (e.g., más ("more")) favored it. This reaffirms previous findings (e.g., López-Morales 1983; Samper-Padilla and López-Morales 1990; Vida 2004) suggesting that retention is more common in contexts where morphemic meaning is not conveyed. Finally, the speech context also conditioned use, with scripted speeches favoring retention and interviews favoring elision.

Word-Final Onset /s/
These contexts for /s/ showed greater variation within regions. Tendencies are presented in Table 5. Socialist females from Seville, for example, retained around half of the time, while those from Malaga retained in less than 10% of productions. Regionally, the NCPS variety represented by Madrid showed near total retention, while northern AS speakers in Cordoba had a nearly medial rate of retention and southern AS speakers in Malaga and Seville had lower rates, reflecting the spectrum of innovative usage based on geography described by Samper-Padilla (2011). This increased variability may suggest stylistic usage of this feature. It is important to note, however, that trends by political affiliation were similar to those present for coda /s/, with conservatives from Cordoba and Seville retaining less, the opposite of what would be expected based on previous research. Given the recent rise of the alt-right party Vox in Spain and increased moves toward populism by the political right across the Western world, it is possible that this shift from previous norms is part of a process of realignment on the part of conservative Spanish politicians, who are trying to reach out to working-class speakers by indexing what had previously been seen as aspects of left-wing identity. The mixed-effects logistic regression model of best fit included only two predictive factors ( Table 6). The first variable, city, followed a similar trend to coda /s/: Madrid showed high levels of retention, followed by the three Andalusian cities. Second, for grammatical function, the same result held in this context as with coda /s/, such that /-s/ in plural and verbal markers disfavored retention, while all other cases favored it.

The Affricate / > tS/
The Spanish affricate [ > tS] exhibited comparable allophonic variability across regions, although there was considerable social variation by political party and speaker gender. These tendencies are displayed in Table 7. Although several social intersections used the normative variant categorically, socialist female politicians in Malaga and Seville had a lower rate of normative post-alveolar production, around 60 to 70 percent. In fact, conservative speakers of AS overall used more of the normative variant than socialists, and men tended to use it more than women (with the exception of socialists in Madrid).
The fronted variant, which has received little sociolinguistic description, appears to have a degree of social prestige given its more frequent appearance in female speech, and its use by politicians on the left places it in opposition to the other phenomena under consideration in this study. In the mixed-effects logistic regression, five factors were found to predict variable use (Table 8). Gender was the strongest predictor based on the range in factor weights, with men favoring the normative [ > tS] and women disfavoring it. Next, for phonological context, both high and low preceding vowels and high and mid following vowels (e.g., mucho ("a lot")) favored the normative variant, while mid preceding vowels and low following vowels disfavored it (e.g., ocho ("eight"), muchacha ("girl")). For speech context, the normative postalveolar variant was favored in interviews with opposite-gender interlocutors, whereas both interviews with same-gender interlocutors and scripted speeches (where intervocalic /d/ and coda /s/ were more likely to be retained) showed higher use of the fronted variant. Finally, for lexical stress, stressed syllables with the affricate (e.g., [mu.'cha.cho] ("boy")) favored the normative production, while unstressed syllables with the affricate (e.g., ['mu.cho] ("a lot")) favored the regional production.

Intervocalic /d/
Allophonic production of intervocalic /d/ varied based on social and linguistic factors, although the standard deviation was the lowest among the four phenomena. The tendencies of retention are presented in Table 9. Speakers from Madrid had the lowest total level of retention, although tendencies varied markedly by social group. In Cordoba and Malaga, socialists retained /d/ more than conservatives, while in Seville and Madrid, the opposite was true. Women from both parties tended to retain /d/ more than men, with the exception of conservatives from Malaga, where retention was higher among male politicians, and among Seville socialists, where retention was comparable between genders.
Although Samper-Padilla's (2011) metareview of variation in Peninsular Spanish pointed to geographical differences mapping onto a spectrum of elision for /d/, differences in production were marked by politician gender rather than city of origin. The mixed-effects logistic regression, treating retention as the application value, included five factors significant in predicting variation across productions. This is shown in Table 10. For intervocalic /d/, speech context had the largest factor weight, with scripted speeches favoring retention and both interview contexts disfavoring it. Second, preceding and following vowel heights indicated the importance of linguistic environment for /d/ elision. Preceding high and mid vowels, as well as following high vowels, favored retention (e.g., podido ("could have")), while preceding low vowels and following low or mid vowels favored elision (e.g., trabajado ("worked")). Fourth, monosyllabic words favored retention (e.g., pero de ("but of")), while polysyllabic words favored elision (e.g., partido ("party"), sido duro ("been difficult")). Finally, intervocalic /d/ in adjectival morphemes (e.g., bien razonada ("well-reasoned")) favored retention, while participles (e.g., ha votado ("has voted")) and all other contexts disfavored it.

Intra-Speaker Variation: The Case of José Antonio Pérez Tapias
These results present certain broad norms in this political community of practice; namely, that word-final onset /s/ and / > tS/ tend to follow the hypothesized trends in terms of political party, while coda /s/ is regionally conditioned and intervocalic /d/ is contextually conditioned. The next step was to use these data to examine a speaker who deviates from the established social norm (e.g., Hernández-Campoy and Cutillas-Espinosa 2013; Pollock and Wheeler 2022).
The Granada politician José Antonio Pérez Tapias was selected based on differences between his production of coda and word-final onset /s/ and the norms of other politicians. Pérez Tapias is a university professor and politician in Granada who was born and raised in Seville and was part of the PSOE from 1993 to 2018. His speech and interviews come from the period leading up to his run for PSOE leadership, when he vied for the position of general secretary in 2014. After his failure to gain a plurality of the votes, he published a novel critical of social democracy in 2017 and left the PSOE in 2018, founding his own party, "Socialism and Republic". With his female interviewer, he discussed problems within his own party, while with the male interlocutor, he described the current PSOE leadership and his run for general secretary of the party. Finally, in the scripted speech, he laid out his run for office, mentioning various aspects of his platform.
Despite growing up in Seville and spending his working years in Granada, we can see that Pérez Tapias' rate of sibilant elision and affricate fronting does not approximate that of his peers (Table 11). Instead, his rates are considerably higher than any but the Madrid speakers for /s/, and higher than any other group for / > tS/. Pérez Tapias' behavior is unlike other individuals from his party in south Andalusia and differs from the baseline established by Andalusian members of the PSOE in Malaga and Cordoba as well. Only in terms of intervocalic /d/ deletion does he approximate his Sevillian peers 3 . This identification of individual differences opens the door to further ethnographic study of stylistic variation, allowing us to determine the performative intention he may have had in making use of normative variants to this extent. Another way to represent variation over time is through analyzing lectal focusing in interaction (LFI), as described by Sharma and Rampton (2015) and employed with public speech by Sharma (2018). In an analysis of LFI, timestamps of variants are recorded, and the production of regional variants is compared with normative ones over time to determine how regional variation differs from moment to moment. A total of 240 tokens of coda and word-final onset /s/, intervocalic /d/, and affricates were coded for the interview with the female interlocutor, and blocks of 20 s were selected to ensure an average of 13 words per block, with a standard deviation of 6, and no fewer than 3 tokens per block. This analysis is shown in Figure 5, with three peaks of regional production highlighted in which Pérez Tapias reached at least 80% regional production.
published a novel critical of social democracy in 2017 and left the PSOE in 2018, founding his own party, "Socialism and Republic". With his female interviewer, he discussed problems within his own party, while with the male interlocutor, he described the current PSOE leadership and his run for general secretary of the party. Finally, in the scripted speech, he laid out his run for office, mentioning various aspects of his platform.
Despite growing up in Seville and spending his working years in Granada, we can see that Pérez Tapias' rate of sibilant elision and affricate fronting does not approximate that of his peers (Table 11). Instead, his rates are considerably higher than any but the Madrid speakers for /s/, and higher than any other group for /t͡ ʃ/. Pérez Tapias' behavior is unlike other individuals from his party in south Andalusia and differs from the baseline established by Andalusian members of the PSOE in Malaga and Cordoba as well. Only in terms of intervocalic /d/ deletion does he approximate his Sevillian peers 3 . This identification of individual differences opens the door to further ethnographic study of stylistic variation, allowing us to determine the performative intention he may have had in making use of normative variants to this extent. Another way to represent variation over time is through analyzing lectal focusing in interaction (LFI), as described by Sharma and Rampton (2015) and employed with public speech by Sharma (2018). In an analysis of LFI, timestamps of variants are recorded, and the production of regional variants is compared with normative ones over time to determine how regional variation differs from moment to moment. A total of 240 tokens of coda and word-final onset /s/, intervocalic /d/, and affricates were coded for the interview with the female interlocutor, and blocks of 20 s were selected to ensure an average of 13 words per block, with a standard deviation of 6, and no fewer than 3 tokens per block. This analysis is shown in Figure 5, with three peaks of regional production highlighted in which Pérez Tapias reached at least 80% regional production.  As Sharma (2018) has described, these moments of high regional usage, reflecting a speaker's first-learned variety, provide a moment for one's "real me" to shine through. These moments emphasize deeply held beliefs or allow speakers to engage in humor or sarcasm as a means of reducing social distance through a show of familiarity with the interlocutor. In the first peak for Pérez Tapias at 60 s, the politician confided in the interviewer regarding his private conversations with current Prime Minister Pedro Sánchez, who won the PSOE elections in 2014. In the next peak at 180 s, he described the mismatch between voters' preferences and actual representation in Congress under the new administration. Finally, leading up to the last peak at 660 s, an interviewer referenced the need to "clean house" and addressed corruption in Andalusia, to which Pérez Tapias responded affirmatively, joking that that a broom would be necessary to address a lot of domestic matters.
While the first two contexts demonstrate instances of the politician using more personal language to confide in his interviewer with his opinions, the third instance demonstrates a moment of levity and humor combined with a critique of his party.

Social and Linguistic Norms in Andalusian Spanish
While previous research has suggested a correlation between political party norms and regional speech use, this trend was not borne out in the logistic regression models. An examination of descriptive tendencies showed that socialists use the fronted affricate more across Andalusia, elide coda and word-final onset /s/ more in Malaga, and elide intervocalic /d/ more in Seville. However, socialist politicians from Andalusia are similar to their conservative peers in other contexts, and those in Madrid are generally close to conservatives. Meanwhile, regional variation was found to condition production of both coda and word-final onset /s/, while gender was a predictive social factor for the affricate, and the social context governed variation for three of the four phenomena.
This last factor merits further attention, as it points to the degree of accommodation that speakers undertake based on the scripted or unscripted speech they produce. Notably, both coda /s/ and intervocalic /d/ perform similarly, with scripted speeches favoring retention and unscripted ones instead favoring elision. Intervocalic /d/ elision has been described in previous variationist research (e.g., Ruiz-Martínez 2003; Villena-Ponsoda and Moya-Corral 2016) as being associated with speakers from lower socioeconomic backgrounds, although the loss of /d/ in the participle -ado is an exception that has also been documented among educated speakers (e.g., Cruz-Ortiz 2022). Coda /s/ elision, on the other hand, is frequent throughout AS as a signifier of regional belonging (Fernández de Molina 2018;Vida 2004). This follows from the research by Villena-Ponsoda and Vida-Castro (2020), who described an intermediate variety between Standard Castilian and AS that is enjoying increased usage among young, middle-class, urban speakers. Both of these variants fit a similar profile and are notably reduced in more formal speech contexts, such as scripted speeches, in favor of a variety closer to the national standard.
The affricate had a different social meaning than these phenomena, with the normative variant occurring most in unscripted speeches with different-gender interlocutors, while the fronted regional variant was favored in scripted speeches and unscripted speeches with same-gender interlocutors. Particularly given the tendency for women to favor use of the regional variant, this suggests that women are most likely to produce the variant in speeches and when speaking with other women, while they are less likely to produce it in mixed company with male interlocutors. Similar to what Flores (2014) finds for the affricate in Chile, where it possesses social prestige, public speakers differ in their production of the variable based on the gender of their interlocutor.
With respect to regional variation, Samper-Padilla's (2011) description of a continuum of elision, increasing as one moves south from Madrid, was not fully borne out by the data. While Madrid was clearly the site of the most conservative behavior for /s/ and / > tS/ amongst socialists, it also had high rates of /d/ elision. Furthermore, rather than representing a medial tendency between Madrid retention and Seville and Malaga elision, politicians from Cordoba elided at comparable rates to those of other politicians, with the exception of word-final onset /s/. Based on the strength of the claims seen in the work of Hernández-Campoy and Cutillas-Espinosa (2010, it was thought that regional norms might take a back seat to stylistic goals. However, at the community level, that was not the case. In particular, it seems that variable intervocalic /d/ elision has stabilized among politicians, as Cruz-Ortiz (2022) and Pollock and Wheeler (2022) have shown for other Andalusian politicians.
Finally, with respect to linguistic variation, findings from the previous literature were reflected by those in the current study. Retention of /s/ was favored in nouns and before high vowels, and it was more frequent in word-final onset position than in coda position. For retention of /d/, elision was favored by participles and in the contexts /a_o/ and /a_a/. Meanwhile, / > tS/ presented several interesting results meriting further research, including a tendency to be fronted following mid vowels and sonorants, preceding low vowels, and in unstressed syllables.

Individual Style-Shifting in Andalusian Spanish
The case-study analysis of Pérez Tapias presents several suggestions for further research. His deviation from regional tendencies in AS seems to have served as a stylistic response to the specific, embattled political situation in which he found himself in 2014. Bucholtz and Hall (2005, p. 595) have described stance as the means by which a speaker describes their relationship with a certain concept using language and stylistic variation. While a speaker's stance may shift from moment to moment, the combined speech decisions contribute to a single unified whole that goes on to help inform further choices the speaker makes in a process called stance accumulation. Pérez Tapias appears to have been involved in just this behavior across all three interviews, as well as in the audio selected for LFI analysis, positioning himself in opposition to the corruption and secrecy of the current establishment.
Part of Pérez Tapias' stylistic effort can be seen as a deviation from the expected AS speech norm. By using normative variants marked as part of NCPS, particularly retention of /s/, he seems to have been doing the opposite of María Antonia Martínez, the former Murcian president studied by Hernández-Campoy and Cutillas-Espinosa (2010). As a candidate for a national position at the head of the PSOE, Pérez Tapias was targeting not only an Andalusian audience but also a national one. Instead of using working-class linguistic features to develop a relationship with local voters, Pérez Tapias used normative features to separate himself from the Andalusian PSOE political establishment and align himself with a national identity. As a transplant in Granada, he was also following local norms that favor greater rates of NCPS variation while simultaneously diverging from PSOE norms in the region.
Alongside his language use, his final showing in the polls provides some perceptual insight into his message. Pedro Sánchez, a Madrid politician, won the 2014 election, while Pérez Tapias came in third. These results raise a question for future investigations about the way that Andalusian and national voters perceived his linguistic performance. Villena-Ponsoda (2013) has drawn on geolinguistics to argue that urban hubs present prestige patterns that have social power relative to their distance from the speaker (i.e., a rural speaker would be most affected by local patterns, followed by those of the nearest large city and, finally, by national normative expectations). By using less locally valued national prestige norms, Pérez Tapias may have been perceived as overly academic or rejecting local values, the opposite of community reactions in Hernández-Campoy and Cutillas-Espinosa's (2013) findings for Martínez. Furthermore, given his intermediate tendency of regional production, placing him between AS and NCPS, it is possible that he was not close enough to either variety to win support from either subset of voters.
This analysis also demonstrates the value of examining discursive and topical differences related to style-shifting. While other research into peninsular political speech has described variation writ large at the level of individual speeches or in certain modes of speech (e.g., Hernández-Campoy and Cutillas-Espinosa 2013; Pollock and Wheeler 2022), emphasizing the role of individual decisions in stylistic variation (e.g., Cruz-Ortiz 2022), LFI analysis allows visualization of these trends at the individual level. As Sharma (2018) has argued for Fareed Zakaria, an Indian American news anchor, this analytical approach offers a better understanding of how a speaker varies in the moment, varying in regional production in the process of identity construction. Pérez Tapias, despite having lower rates of regional production than his peers, still uses these variants intentionally to emphasize moments of confidentiality and humor, creating rapport with his interviewer and his audience, as a means of emphasizing Andalusian solidarity even while constructing a more nationally relatable image.

Conclusions
This study contributes to the fields of Peninsular Spanish sociolinguistics, stylistic analysis, and political discourse through a sociophonetic lens. The quantitative results show differences in the linguistic resources available to politicians in four Spanish cities that can be used for style-shifting, establishing a baseline norm within this political community of practice. As both social and linguistic factors govern production of the four regional phenomena examined, it would behoove future researchers to include additional AS regional phenomena with variable levels of social acceptance that could be employed in style-shifting behavior, such as seseo and ceceo, liquid neutralization, and /x/ weakening.
Shifting from a macro view of community variation, the micro-level case study of individual stylistic choices by a single politician showed how speakers follow broad norms in a community of practice, as well as how linguistic tools in a speaker's repertoire can be employed to construct a consistent identity in opposition to the mainstream. Overall, this article contributes to our understanding of Andalusian Spanish and style-shifting in political speech by offering a comprehensive vision of regional variation and performativity, capturing the degree to which politicians have agentive command over regional and normative features as a means of controlling their public image.
Funding: This research received no external funding.
Institutional Review Board Statement: Ethical review and approval were waived for this study due to the public nature of the audio recordings used in the analysis.
Informed Consent Statement: Informed consent was waived as the audio data were drawn from pre-recorded, publicly-available sources.
Data Availability Statement: Data available on request from the author.

Conflicts of Interest:
The author declares no conflict of interest.

1
Additionally, albeit anecdotally, the former Spanish Prime Minister Felipe González (1982González ( -1996 is often referenced by Andalusians as one of the first major politicians to use Andalusian phonetic features in oral political discourse, granting them a higher degree of acceptability and prestige. 2 However, see Villena-Ponsoda and Vida-Castro (2020) for a description of an intermediate form blending Standard Castilian and AS in a type of koine mainly used by young urban speakers from the middle class. This shows an ongoing shift resulting from national education norms and regional varieties, yielding a novel variety gathering popularity across Andalusia and gaining acceptance particularly in Seville. 3 A reviewer notes that this may be the result of Pérez-Tapias' time in Granada, a region where convergence with NCPS is more likely than in the west of Andalusia. Given his early exposure to Sevillian Spanish, connections to a cohort of politicians across Andalusia, and ongoing contact with the eastern variety through his career, this presents an interesting case study to see how these various factors in the politician's life and background combine to inform his linguistic behavior and output.