“We Pray for Our Nation an(d) Our Worl(d)”: The Influence of Race and Audience Attitude on Coronal Stop Deletion at the Inaugural Prayer Services, 2001–2013

This paper examines the effect of race, context, and white public space on the extent to which speakers articulate, hyperarticulate, hypo-articulate, or glottalize word-final English alveolar stops -/t/ and -/d/ in the controlled environment of the quadrennial US Presidential Inaugural Prayer. It shows that African-American speakers hyperarticulated and articulated /t,d/ more frequently than the white speaker, who hypo-articulated and glottalized /t,d/ consistently, especially on words like God, Lord, and Christ. These results suggest that the highly formal context required African-American speakers to perform /t,d/ to index themselves as authorities to an unfamiliar, white audience, while the white speaker did not consider race to influence listeners’ judgements of him, allowing him to index familiarity and trustworthiness. 


Introduction
Many speech communities have /t,d/ hypo-articulation, or an absence of production, in their linguistic repertoire (Tagliamonte 2011, cited in Grieser 2014, but stop hypo-articulation and glottalization are well-attested features of African American Language (AAL) (Thomas 2007). 1 Hypo-articulation has an expansive indexical field across speech communities (cf. Eckert 2008) that includes competency, likeability, authority, and lack of intelligence, among others (Rahman 2008, Podesva et al. 2015, Dick 2017; however, the ways in which speakers can use language features are largely conditioned by context and identity. Despite evidence that speakers of AAL hypo-articulate /t,d/ at higher rates than white speakers (e.g., Holliday 2017), I argue that the audience of the highly formal performance of the Inaugural Prayer is the driver for why the three African-American speakers explored in this paper hyperarticulate, or exaggerate production, of word-final /t,d/. African-American speakers can hyperarticulate to index authority and competency against long-standing language ideologies that devalue Blackness and AAL. 2 Prayer is a specific speech context in the larger event of religious ceremony and provides a unique context for /t,d/ articulation. Black evangelical church services reflect the linguistic practices of their constituents (DeBose 2015), and there are preaching styles unique to Black preachers (Kortenhoven 2017). Research on linguistic performance by Black religious professionals focuses largely on style in Black churches (e.g., Sharma et al. 2019, DeBose 2015, Kortenhoven 2017. Thorough work has been done examining the Black preachers' personas that they can access when preaching to a predominately Black audience (Sharma et al. 2019, Kortenhoven 2017. Religious services remain largely segregated (Mohamed et al. 2021), but Black preachers still preach in white spaces and are subjected to racist language ideologies about Blackness, competency, and authority. Thus, there is a nuanced relationship among race, religion, and performance, specifically in how Black preachers can index Blackness, and affiliation with the Black church, linguistically in white public space. White listeners' language ideologies likely require African-American speakers to perform formality more than white speakers, thus begging the question of the constraints placed upon Black preachers that frame how they negotiate their affiliation with the linguistically creative Black church and these performances before a white audience.
This paper analyzes the ways in which racial identity, context, and white space (including language ideologies) constellate to influence speakers' realization of the English phonemes /t/ and /d/. It examines the ways that speakers realize word-final /t/ or /d/ as hyperarticulated, articulated, hypo-articulated, or glottalized, which provides insight into how speakers interact with white audiences' predicted language ideologies. I studied the opening prayer, the only prayer these speakers delivered, in four successive U.S. Presidential Inaugural Prayer Services (2001Services ( -2013 with largely white audiences. The three African-American speakers articulated and hyperarticulated /t,d/ at much higher rates than the one white speaker, who consistently hypo-articulated and glottalized /t,d/. I argue that African-American speakers realized /t,d/ to index authority and competency, and reacted against an unfamiliar, white audience's potentially racist presuppositions about speakers of AAL (e.g. Rahman 2008), while the white speaker did not need to prove himself to a white audience and could pursue an intimate, friendly persona to complement his audience-established authority.

Literature Review
There are many ways to describe the absence of final /t,d/, including when the speaker does not produce an audible sound for the final consonant (Rositzke 1943). Stop hypo-articulation is more likely to occur in monomorphemic words, such as mist, than polymorphemic words with inflectional -ed endings, such as missed. Missed carries information about the action and the time at which the action was completed, giving the same /t/ sound a higher "functional load" than in mist, so the sound is less likely to be hypo-articulated (Guy 1991:3). In addition, words used more frequently are expected to have higher rates of deletion, and vice versa (Podesva et al. 2015).

African American Language and Stop Hypo-articulation
Stop hypo-articulation (Purnell et al. 1999, Thomas 2007 and glottalization (Thomas 2007) are documented phonological features of AAL. First-wave studies on urban African Americans' linguistic and social identities (Fasold 1972;Labov 1972, cited in Grieser 2014, and more recent scholarship (Grieser 2014, Holliday 2017, Rahman 2008 have found that professional African Americans tend to have a large linguistic repertoire that they can use to coordinate their racial and professional identities. A repertoire can consist of standardized white varieties, Black standard varieties, and more vernacular African-American varieties. Features of AAL may be selected intentionally and stylistically (Thomas 2007) to index traits depending on context (Eckert 2008, Grieser 2014. Speakers of AAL have been found to hypo-articulate at higher rates than white speakers, even in formal contexts (Holliday 2017), but there are other variables that affect how African-American speakers can perform /t,d/ (Podesva et al. 2015), such as recognizability and familiarity.
Studies have shown that phonetic variables carry different social meanings for different listeners, such as Campbell-Kibler's (2011) work on IN/ING and Podesva et al.'s (2015) work on medial and final /t/. With respect to final /t,d/, presence can index high articulateness (Podesva et al. 2015), formality, dependability, and competency (Dick 2017). Absence can index casualness (Rahman 2008), lack of intelligence, and relaxedness (Dick 2017). The meanings that hypo-articulation can index, however, change if race becomes a point of differentiation between speaker and audience. White listeners' judgements of white speakers' hypo-articulation include attributes like friendliness, genuineness, sincerity (Podesva et al. 2015), and casualness (Dick 2017). 3 Black listeners' evaluations of African-American speakers' hypo-articulation can index "black authenticity" (Holliday 2017:463). 4 White evaluations of African-American speakers' hypo-articulation can include lack of prestige, increased relatability (Holliday 2017), and lack of formality (Rahman 2008). If African-American speakers can recognize that AAL and those who speak it bear the brunt of racist assumptions on language and competency (Rahman 2008, Podesva et al. 2015, they may articulate or hyperarticulate /t,d/ to overcome those ideologies (cf. Labov 1966).
Such racialized ideologies devalue speakers of AAL systematically. Purnell et al. (1999) found that after hearing the word "hello", listeners were more likely to fulfill callers' requests if they perceived the caller to be white, as opposed to perceiving callers as speaking an ethnic or racial variety, such as AAL. Speakers' identity racial identityunderlies language and becomes the basis for discrimination. Devos and Banaji (2005) also found that race is highly influential when evaluating communities: white participants associated white Americans more closely with the identity of "American" than they associated it with Black Americans, but Black Americans viewed themselves and white Americans as equally "American". White English is readily accepted in white public space as an unmarked variety (Bucholtz 1999, Hill 1998. Blackness is marked for white audiences, so white speakers enjoy the privilege of using features that African-American speakers would be evaluated negatively for using. Speech acts where speakers of AAL interact with a white audience in historically white spaces are of particular interest. It is here where racist language ideologies may appear and influence the audience's expectation or evaluation of a speaker and thus influence a speaker's performance in response to such ideologies.

Prayer as Speech Context
Prayer is one of many speech events involved in religious ceremony. Public prayer is a particular speech environment that is ritualistic and performative, as a speaker invokes the divine first and foremost but also has to take into account how listeners who are present but not addressed directly will evaluate the speech act (Fitzgerald 2012). Prayer is "situated and situating discourse" (Fitzgerald 2012:12) in that it frames the situation in which it is performed but is also performed contextuallybecause it is a performance, it is creative and diverse (cf. DeBose 2015, Mohamed et al. 2021).
As other studies have explained (e.g., Grieser 2014, Rahman 2008, context and content influence the extent to which speakers will use features that index them before a speech community. Church communities can be seen as speech communities in that they are defined by a shared belief in the divine and shared language practice. Black preachers have a large repertoire of preaching styles, including a "teaching style" that is less emotional and a more emotional "call and response" style (Mohamed et al. 2021:148, cf. Gumperz 1968. Sermons in predominately Black churches can include "markedly Black language" (DeBose 2015:686) that indexes a speaker's affiliation with the Black community. In formal contexts, speakers can access the seminary-educated preacher persona (Kortenhoven 2017), a performance indexing institutional religious education and marked with formal language, consistent articulation, and attention to social and political issues.
The ways that speakers engage the divine and the physical audience largely depend on their professional identity in relationship to the listeners. It affects the range of meanings that African-American speakers can glean from a linguistic variable if they are speaking to a white audience. Even though there is a rich variety of personas into which Black preachers can tap, resistance to Blackness in the public sphere pushes Black preachers to "mute their voices" before white audiences (Gilbert 2020).
This study looks to further unpack the ways in which Black professionals, specifically Black preachers, use linguistic variables to move around white public space. Black preachers must be creative and navigate their religious, public-facing identity and affiliation with the Black community in different ways than they would if they were preaching to a Black audience. In this work, I analyze the differences between the white and African-American speakers' realizations of word-final /t,d/ in prayer, with particular attention to the larger relationship between race and realization.

Materials
Data came from C-SPAN's video coverage of the Inaugural Prayer Services. The Inaugural Prayer Service is a quadrennial gathering at the Washington National Cathedral on the morning after a president's inauguration. The service consists of hymns, choral and instrumental music, readings of biblical passages, reading of Psalms, and prayers, the first of which is the opening prayer (Washington National Cathedral 2013). The main purpose of the Inaugural Prayer Service is to welcome and bless the sitting or incoming president and vice president. It also calls for national unity with reference to current events relevant to the United States (Gjelten 2021). I chose to analyze the opening prayer, which occurred between 10 and 15 minutes into the service and was the first prayer at the service. Each sample was designated as a prayer through the adjacency pair of "amen / amen" from speaker to audience. 5 The opening prayers range from 30 seconds to 4 minutes. Each speaker is involved in the same social and speech event. There are three African-American speakers and one white speaker. 6

Procedure
To map realizations of /t,d/, I transcribed each prayer and coded final stop realization as hyperarticulated, articulated, hypo-articulated, or glottalized. Hyperarticulated phones were defined as unusually voiced or audibly released, and articulated phones were those that were released but without those features. Hypo-articulated phones did not produce a sound, and glottalized phones were glottalized. I was most interested in the relationship between race and hyperarticulation.
While I still compared specific variants of realization, the broader relationship between race and tokens of presence and tokens of absence was of the most interest. I categorized articulation and hyperarticulation as "presences" and hypo-articulation and glottalizations as "absences". 7

Results
The relationship between speakers' race and their percentages of tokens of absence and presence was of the most interest, but morphemic status and word context were not found to influence the rate at which speakers would articulate or hypo-articulate /t,d/. The African-American speakers articulated and hyperarticulated /t,d/ at much higher rates than the white speaker, who hypo-articulated and glottalized /t,d/ consistently.

Morphemic Status and Word Context
Previous literature (Guy 1991) found that monomorphemic words were more likely to contain hypo -articulated tokens, and polymorphemic words were more likely to contain articulated tokens. The Chi-square test did not suggest that morphemic status had an effect on articulation (p > .05). There were 123 tokens on which there was a presence or absence of final stop /t,d/. Of those 123 tokens, 93 occurred on monomorphemic words (75.6% of total tokens). Audible /t,d/ was produced in 51.6% of monomorphemic words and 56.7% of polymorphemic words.
The second independence test examined if a word's context as lexical (open class) or function (closed class) was likely to influence a token of presence or absence. For the four speakers, lexical words were more likely to have articulated /t,d/ than function words (p < .05). When running a test for the three African-American speakers, the relationship reached significance. There was not a significant relationship for the white speaker and two African-American speakers (Revs. Baxter and Moss Jr.). 8

Speakers' Race
The relationship between race and realization is significant (p < .001). Figure 1 compares each speaker's percentages of tokens of presence and absence. The three African-American speakers' articulation rates were extremely close, within 2.5 percentage points of each other. Rev. Moss Jr. had the highest rate of articulation (67.4%), and Rev. Williams-Skinner had the lowest rate (65.0%) for the African-American speakers. Rev. Graham had an extremely high percentage of tokens of absence (80.6%). His percentage of tokens of absence was higher than the percentage of tokens of presence for the other speakers, which showed that he was more likely to hypo-articulate or glottalize /t,d/ than the other speakers were likely to articulate or hyperarticulate them. Overall, race was likely to affect the ways in which speakers realized /t,d/. 5 Rev. Williams-Skinner was the only speaker not to include the traditional "amen". 6 When I analyzed the data, I did not use phonetic software or other listeners to check the accuracy of the ways that I coded for hypo-articulated and articulated stops, so there is inevitable human error. Because this is a small sample size (n = 4 speakers), it may strengthen my conclusion if I repeated the study, or a similar study with formality and context controlled, with a larger sample size. 7 See Guy (1991) for his explanation of why he included glottalization as a non-absence. 8 The test for Rev. Williams-Skinner's prayer did reach significance, but her prayer was significantly shorter than the other prayers and thus has higher standard error. Her prayer was 87 words, compared to the second shortest prayer, which was Rev. Baxter's at 160 words. Rev. Moss Jr. spoke for 335 words, and Rev. Graham spoke for 425 words.  Figure 2 presents the average percent realization for the four speakers. The three African-American speakers did hyperarticulate /t,d/ at a much higher rate (50.6% of total realizations) than the white speaker (5.6% of total realizations). African-American speakers were much less likely to hypo-articulate (13.8%) and glottalize (19.5%) than the white speaker (50.0% hypo-articulation, 30.6% glottalization). Individual speakers' percentages of realization by variant are given in Figure 3. Rev. Graham had significantly more tokens of absence than other speakers and the lowest percentage of any realization across speakers (5.6% for hyperarticulation). The African-American speakers (Baxter, Moss Jr., and Williams-Skinner) articulated and hyperarticulated /t,d/ at various rates, with a range of 13.1 percentage points for hyperarticulation and 15.5 percentage points for articulation. As both part of the larger category of "presences", the greater the percentage of stops were articulated, the lower the percentage of stops were hyperarticulated. Rev. Williams-Skinner had the highest rate of articulation of the three speakers (25.0%) and the lowest rate of hyperarticulation (40.0%), while Rev. Baxter has the highest rate of hyperarticulation (57.1%) and the lowest rate of articulation out of the four speakers (9.5%).

Discussion
All three African-American speakers' rates of articulation and hypo-articulation were very close, with a range of 2.4 percentage points. Hyperarticulation occurred most frequently, followed by articulation. It does not appear that speakers attempted to hyperarticulate most times but forgot to do so in other times, because articulation and hyperarticulation occurred consistently throughout the prayers. Notably, there was no difference between the speaker who spoke at Bush's inauguration versus the two who spoke at Obama's. As literature and this research has suggested, speakers of a marked variety may be under more pressure to "unmark" their speech (Bucholtz 1999, Hill 1998 for non-intimate audiences in order to be evaluated positively. I argue that racism against African-American speakers, regardless of whether or not they were using features of AAL, pushed them to tap into the range of indexical meanings for speakers who articulate /t,d/ consistently to present themselves as competent, prepared, appropriate for the formal tone of the Inaugural Prayer Service, and authorities in white public space. While variation in final stop realization is not unique to AAL, speakers of AAL are already socially devalued. I argue that strict constraints for how African-American speakers can index formality and competency for a white audience may overwhelm other meanings into which speakers can tap through variant /t,d/ realizations. Racial identity is indexed differently if speakers before a white audience are Black because Blackness is marked and hyperracialized. Frequent tokens of articulation and hyperarticulation contradict Holliday's (2017) finding that the Obamas hypo-articulated /t,d/ at higher rates than white speakers including in formal contexts, but the difference may correlate to the extent to which African-American speakers are evaluated as members of the Black community and as Black religious professionals with whom a white audience may not be familiar. While these African-American speakers are likely well-respected and highly regarded among their congregations, evangelical churches are still largely segregated along racial lines (Mohamed et al. 2021). Because the audience for the Prayer Service is predominately white, Black reverends are less likely to have preached in white spaces and are thus less likely to have established their ethos for a white audience. I argue that the social pressures to perform for a very public, unfamiliar audience (Bell 1984) would have encouraged speakers to use a variety and praying style that are associated more with formal office and not an intimate religious community.
Context aside, the influence of white privilege cannot be overstated: in the U.S., whiteness is neither marked nor racialized (Bucholtz 1999). The very large majority of the audiences are white, which means that Rev. Graham's social and religious status is more likely to influence his listeners' perceptions of his competency than his race. Especially among conservatives and evangelicals, Rev. Graham was a massively popular figure with an incredibly expansive religious footprint. His persona was built around approachability and authenticityfor 60 consecutive years, he was listed as one of the "Ten Most Admired Men in the World" in a nation-wide survey (Graham Jr. 2018). His whiteness likely allowed him to capitalize on the social meaning of stop hypo-articulation as indexing intimacy, as opposed to proving his competency. I therefore argue that Rev. Graham's heavy hypo-articulation of /t,d/ was not a variable indexing informality, but approachability and likeability, which is a testament to how white privilege allows speakers to access particular meanings for the realization of a variable that African-American speakers may not be able to access before a white audience.
For all speakers, I argue that their experiences preaching influenced the ways in which they index professionality for a religious context. Rev. Baxter prayed in his home cathedral. He presented himself not only as an authority before the audience, but as a representative of the Washington National Cathedral for a visiting audience. Revs. Moss Jr. and Williams-Skinner prayed before an unfamiliar, white audience and tapped into the persona of a seminary-educated preacher (Kortenhoven 2017): with enhanced attention to language, articulation, and indexing of sophistication. Revs. Baxter, Moss Jr., and Williams-Skinner can access other personas with which to index affiliation with community, "experiential authority" (Kortenhoven 2017:21), and humilityamong other characteristicsbut I argue that the unfamiliar white context inhibited the personas that they were able to access. Importantly, the personas that Kortenhoven (2017) outlines index authority in the Black church and engage with features of AAL diversely. However, I argue that a white audience is more likely to judge Black preachers negatively for performing an alternate preacher personausing a racialized variety of English in white space (cf. Hill 1998, Rahman 2008)than they are to judge a white preacher for using a non-standard variety of English. Comparatively, Rev. Graham is not associated with Washington National Cathedral, but he had an extensive friendship of 20 years with the president for whom he delivered the prayer, George W. Bush (Bush 2018). He was friendly with the administration. Examining prayer gives insight in the ways that religious professionals can align themselves with particular personas depending on the context and audience. Religious affiliation is an exceedingly common aspect of social identity, but comparatively little has been done to examine how religious identity co-indexes alongside linguistic features. The most available data set for examining professional religious identity may be prayer.
For all four speakers, and for the subset of three African-American speakers, lexical words were more likely to have tokens of presence than function words, and function words were more likely to be monomorphemic. Lexical words that were hyperarticulated the most consistently were God, Lord, Jesus Christ, and president/President, all of which referred to a figure with power and influence. These speakers were tasked with delivering the opening prayer, arguably the most important performance, as it made the first appeal to and invocation of the divine as pertaining to a particular topic. It is not surprising that lexical words with religious connotations were hyperarticulated most consistently. Hyperarticulation on God or Lord might reinforce to the audience that the speaker also felt that those words were powerful (Fitzgerald 2012). It also becomes a way for African-American speakers to perform various intonation patterns characteristic of Black churches while maintaining their seminary-educated Black preacher persona (Kortenhoven 2017) before white audiences that likely push African-American speakers to hyper-perform in order to prove their authority in white space.
In this study, the percentages of tokens of absence and presence were higher on monomorphemic words (77.6% tokens of absence, 73.9% of tokens with presence), which is consistent with Guy (1991). Lexical words indexing religiosity (e.g., God, Lord) were low-frequency and very likely to be hyperarticulated, which is consistent with Podesva et al. (2015). Subsequent research could examine more intensely the relationship between word frequency and tokens of presence or absence, especially in contexts where low-or high-frequency words index meaning that is pertinent to how a speaker constructs a persona for an audience.

Conclusion
This research investigated the relationship between race and word-final /t,d/ realization (hyperarticulation, articulation, hypo-articulation, glottalization) by four speakers in four successive Inaugural Prayer Services. The data on race and realization concorded with previous literature on stop realization and race, suggesting that language ideologies influence speakers' ability to access the range of meanings for a variable. Language and practice continue to pander to white spaces and are neither random nor dissociated from identity and privilege.
The range of meanings for /t,d/ realization also underlies any influence that context may have on speakers' linguistic choices. Data suggests that African-American speakers responded in advance to language ideologies that systematically devalue AAL and tapped into linguistic features from their religious personas and racial identities in order to be indexed positively by their white audiences. Identities constellate and are more complex than what linguistic variables can index alone. Prayer is a genre of speech that is inherently creative. It is worth investigating how religious speakers can be intentional and linguistically savvy while remaining faithful to the genre constraints of prayers.