I Variationist approaches

Non-binary approaches to gender and sexuality

Penelope Eckert and Robert . Podesva (Part I leads)

Introduction

The study of variation is founded on replicable correlations with macrosocial categories, particularly with class, gender, age, and ethnicity. These correlations provide the larger general landscape of social variability - the distribution of variables across society, the social areas and kinds of distinctions that are at work in variation, and the nature of the networks that guide the spread of change. At the same time, we need to treat these correlations with caution, because categories are structural objects that abstract away from, and erase, the practice that produces them. They provide a global pattern, but no explanation. A purely structural perspective can all too easily lead to deterministic analyses - women talk the way they talk because they are women, working class people because they’re working class, gay people because they are gay. This approach guides the focus to the categories themselves as objects of indexicality, or to some stereotype associated with membership in particular categories. It also leads us to think purely in terms of the analytic structure of the categories - as continua (class, age, attention paid to speech) or binary oppositions (gender, sexual orientation, race, and even ‘above’ or ‘below’ the level of consciousness).

This is illustrated in the path that gender and sexuality studies in variation has taken. Labov’s (1966) analysis of variation and socioeconomic class was a radical move in the 1960s, opening the door to serious quantitative work in linguistics and introducing a solid basis for the social analysis of variability. But the robust finding that linguistic form correlates with class stratification established a class basis for subsequent interpretation of variation. The view of class as a continuum of global prestige was established as the source of meaning in variation, with variants viewed as carrying prestige or stigma, and patterns of style shifting as reflecting orientation to the prestige and stigma of class position. With class-based prestige and stigma as the fundamental terms of indexicality, correlations with gender were then taken to reflect binary orientations to prestige and to class position. Women’s greater use than men of stable variables was attributed to status consciousness (e.g. Trudgill 1972) - an attribution that has not held up well in the face of the regular finding that women commonly (but not always¹) lead in sound change. At the same time, the assumption that women are more status conscious than men has led class to bleed into qualities, with femininity associated with refinement and masculinity with toughness. And as attention has turned to sexuality, it has been all too easy to treat sexual orientation as a fractal (Irvine and Gal 2000) within the male/female binary, with the speech of gay women and men treated as masculine and feminine respectively. Just about any linguistic variable is likely to show a correlation with class and with binary gender because both are fundamental to the social order - in fact, they are so fundamental that correlations with them are often in themselves meaningless. People pursue their lives at the intersections of social categories, each intersection bringing forth its own conditions and situations, its own constraints and possibilities. It is in day-to-day practice at this juncture that variation becomes meaningful, and that correlations are produced.

Variation, style, and persona

We take as illustration the use of negative concord (e.g. 'I didn't do nothing') in the speech of students at Belten High School in the Detroit suburbs (Eckert 2000). Figure 2.1 shows a stark version of the typical gender pattern in the use of negative concord at Belten High - the boys use significantly more negative concord than the girls. But the picture shifts when we look beyond this binary, as participation in the class-based social categories that dominate the school social order yields yet another binary that interacts with gender. The ‘Burnouts' and the ‘Jocks’ (Eckert 1989, 2000) are communities of practice (Eckert and McConnell- Ginet 1992; Lave and Wenger 1991) constituting working-class and middle-class cultures respectively. Figure 2.2 shows that while the overall gender difference persists in each community of practice, it is far greater among the Jocks than among the Burnouts. It is important to note that the strong correlation is with the speaker’s participation in their peer-based communities of practice, and not with their parents’ socioeconomic class. This indicates that the use of negative concord is not passively acquired at home and in the neighbourhood, but is part of the stylistic construction of an adolescent self.

We note that the gender difference is far greater among the Jocks than the Burnouts, an interaction that clearly indicates that these categories are not simply additive, but

Figure 2.1 Percent negative concord by binary gender, (x² = 3124, p = .000).

Figure 2.2 Percent negative concord by binary gender and social category.

intersectional. Most obviously, while Jock girls are considerably more constrained than boys to be conservative in their behaviour, Burnout girls are not (see Eckert 1989, 2000). There is considerable variety of behaviours among Burnout girls, ranging from fairly compliant with adult norms to quite rebellious. This range shows up in a friendship cluster of urban-oriented and rebellious Burnout girls who pride themselves on being the ‘biggest Burnouts’, commonly referred to by their classmates as the ‘Burned-out Burnouts’. These girls are a quite distinct friendship group from the bulk of burnout girls who embrace more working-class norms and who are less alienated from school, and less rebellious - the ‘Regular Burnouts’. Figure 2.3 compares several clusters of Burnouts and Jocks with the larger student body represented by not only Jocks and Burnouts but students who do not identify with either. As this figure shows, the Burned-out Burnout girls lead the entire student body in the use of negative concord. Meanwhile, those who use the least negative concord are the Jock boys who are engaged in student government (as opposed to those who are athletes only). In other words, we see that gender and class break down into the kinds of indexicalities that Trudgill raised, but not in the gender groups for which he raised them. Rather, the Burned-out Burnout girls are using negative concord to index an autonomous and anti-establishment stance, while the male student government Jocks are using standard variants in maintaining their corporate personae. When talking at the macrosocial level, it is reasonable to say that gender norms constrain women and men to engage in the world in such a way that it benefits them to use standard and nonstandard forms respectively. But these norms work not at the binary gender level but at the persona level as constrained, but not determined, by the gender binary. If negative concord indexes gender, it’s only indirectly through the relation between what it indexes directly (anti-establishment stance) and what it ‘means’ to be male or female.

Figure 2.3 Negative concord by subcategory. Shown as standard deviation from class mean, including Jocks, Burnouts, and In-Betweens.

In other words, as Levon (this volume, Chapter 3) says, we need to always ask the ‘other question'. In this case the question is: ‘When I see gender difference, where are the class interests and when I see class difference, where are the gendered interests?'. When we do this, we move on to consider the dynamics that unfold at the intersection of class and gender. And we experience gender - and class - not in isolation, but in the practices and personae that emerge at their intersection.

The Burned-out Burnouts and the Corporate Jocks are not passively responding to their place in the social order - in fact, nobody is. Social practices emerge as people in similar structural locations mutually engage in similar responses - responses that commonly reproduce the structure that gave rise to them, but that also carry the potential for change. Butler’s theory of performativity emphasises the dialectic between structure and agency in the production of the gendered individual. Category assignment places the individual in situations where they learn to engage in ‘a stylised repetition of acts through time’ (Butler 1988: 520), stylised acts that constimte gender as a central aspect of emergent identities. All aspects of identity are achieved performatively, as we engage with the affordances of our social world to construct a self with which we can take a place in that world. Central to Butler’s account of performativity is that in these repeated acts lies the potential for change:

If the ground of gender identity is the stylized repetition of acts through time, and not a seemingly seamless identity, then the possibilities of gender transformation are to be found in the arbitrary relation between such acts, in the possibility of a different sort of repeating, in the breaking or subversive repetition of that style.

(Butler 1988: 520)

Style is the key to performativity. It is the means by which we mark both pre-existing and emergent distinctions (Irvine 2001). Emergent distinctions might be particularly tentative, and a small stylistic move is sufficiently inexplicit to provide the possibility of denial, putting a verbal toe in the water. A young man who has been in the closet might want to make a tentative move towards sounding gay. Producing a slightly more fronted /s/ than usual would be both slightly perceptible and quite deniable, and depending on the reactions of others and his own embodied sensation in making that move, he might decide to move forward in or beyond that situation, or not. When we think of style we focus on the projection of persona to others, but style is also a bodily act and has a ‘feel’ that connects the speaker to the outward presentation. A stylistic move, in other words, is not just an external act, but a small - perhaps temporary - bit of personal transformation. And stylistic practice, the production of persona, is not a momentary event, but a continuous process that carries us through life. While there may be moments in which persona is foregrounded, and in which we make quite intentional moves, most of the time stylistic practice takes place quite unintentionally, even unconsciously. And it takes place not at the level of ‘gender’ or ‘sexuality’, but at the level of qualities, stances, momentary activities that are about the kind of person we want to be in the moment. While some of these qualities, stances, and activities may be overtly associated with gender or sexuality, most of them are quite indirectly (Ochs 1992) indexical of categories. And certainly, no single stylistic feature - no single variable - can work alone. Stylistic practice is a process of bricolage (Hebdige 1984; Levi Strauss 1962) in which meaningful elements are combined to constitute a higher meaning that is more than the sum of its parts.

Distinctions of voice quality, segmental phonetics, prosody all have meaning potential, but potential that is underspecified, and unrealised in isolation. Itl release in flap position has been found in the speech of Orthodox Jews (Benor 2001), geek girls (Bucholtz 1996), and gay men (Podesva et al. 2002). And while this single feature can be said to index each category, it does so only in combination with the other elements that constitute the larger style in which it is embedded. Podesva (2004, 2007) documented the stylistic changes of a young gay doctor, Heath, as he moved from the clinic to a barbeque with his friends. In the clinic, his ‘caring doctor’ persona feamred a good deal of N release, indexing a formal and articulate style. In his ‘gay diva’ persona at a barbeque with his friends, he used fewer stop releases, but they had significantly longer and more intense bursts, giving them a parodic quality. And while he used falsetto in the clinic, but almost exclusively on discourse markers (‘okay’, ‘alright’), his gay diva persona made copious use of falsetto with greater duration, greater fO range and maximum. The /t/ release alone would not have contributed much to his gay diva persona - in fact, without other elements such as falsetto, the Ш release might have just sounded weird. It is in this stylistic practice - the combination and recombination of resources - that variation takes on meaning. And that meaning is not fixed, but changes as our needs to make meaning change.

Indeed, a fundamental problem with the purely structural view is that it implies a kind of permanence. Class, gender, and sexuality are certainly here to stay as categorisation schemes, but the strucmre of the schemes themselves changes with time. And this change does not happen in the abstract, but is inseparable from each individual’s trajectory through life, as each generation hopes to move beyond the previous one - to be more ‘modem’, more ‘advanced’, to move on from the past. New ideas emerge, new issues, and with them new ways of being in the world, and bricolage makes it possible to use old material to bring these new ways into the social landscape.

Meaning and use: creaky voice

In the remainder of this chapter, we illustrate the ideas above in an extended discussion of creaky voice, which has undergone massive indexical change. This single linguistic feanire also illustrates some of the pitfalls of the binary thinking that underlies variationist socio- linguistic research. We will show that creaky voice is both ideologically associated with femininity, and more prevalent in the speech of speakers who identify as women. In our view, these facts are trivial, as they tell us nothing about why women exhibit a preference for creak, nor do they tell us anything about gender. We advocate for an alternative approach that is concerned less with the question of who creaks the most and more with the question of what creak means. An approach centred around social meaning, we argue, can better explain gendered distributions of sociolinguistic variables while also providing insight into the social construction of gender.

The other chapters in this section similarly illustrate the impoxlance of taking a meaningcentric approach to the study of gender and sexuality. Zimman, for example, highlights the limits of conceptualising gender as having an atomic meaning. His study shows that gender identity generally predicted the realisation of Is/, but gender role (‘socially and/or institutionally recognised gender categories within a given culture') - a dimension of gender distinct from gender identity - was necessaiy to account for outliers. Levon emphasises that sexuality is constructed by means of linguistic features whose meanings are variable depending on speakers' gender presentation. Generally, as the degree to which female speakers identify as ‘lipstick’ increases, so too do their mean pitch levels. But different speakers exploit higher pitch in distinct ways, with ‘lipstick’ speakers using higher pitch when discussing gay topics and ‘butch’ speakers using higher pitch for non-gay topics. Campbell-Kibler and miles-hercules also stress that meanings relating to gender and sexuality are not inherent to linguistic features, but rather vary as a function of contextual factors, such as who the listener is. Using perceptual methods, they show that in spite of a dominant ideology whereby gayness and masculinity are negatively correlated among men, this association does not hold across the listener sample. Together, these studies indicate that ideologies associating linguistic forms with gender and sexuality are at once too general (linguistic practice cannot be explained by unanalysed, monolithic notions of gender) and too specific (gender is seldom articulated independently of other dimensions of identity), and fail to capture the fact that gender and sexuality are emergent.

That creak has been ideologised as a (young) feminine speech feature is apparent in media discourse surrounding the feature, also referred to as vocal fry. Creaky voice has been described as ‘the new way young women talk’ (Weiss 2013) and a ‘female fad’ promulgated by celebrities like Britney Spears, Kim Kardashian, and Zooey Deschanel (Steimnetz 2011). In an opinion piece for The Guardian, author Naomi Wolf (2015) goes so far as to issue the plea: ‘Young women, give up the vocal fry and reclaim your strong female voice’.

There is reason to think that the ideology of the young female creaker is based - at least in part - on patterns of use. A number of studies have documented the prevalence of creak in the speech of young women (e.g. Wolk, Abdelli-Beruh, and Slavin 2012) and shown that women produce more creaky voice than men (Podesva 2013; Yuasa 2010). These previous studies are limited in that they have employed auditory methods, considered limited samples of speakers, and, in some cases, been based on analyses of read speech, which may not satisfy the social conditions required for the extensive production of creak. To address this issue, Callier and Podesva (2015) carried out a larger-scale acoustic study of creaky voice among 93 white, cisgender speakers from inland California. About half of the speakers self-identified as women, and both women and men spanned the entire adult life course, from 18 to 93 years old. Data were approximately hour-long sociolinguistic interviews. Every vowel was labelled as either creaked or not (using Kane, Drugman, and Gobi’s 2013 model, which classifies speech on the basis of acoustic properties) and measured for a variety of acoustic parameters that capmre different dimensions of voice quality variation, including spectral tilt (e.g. Hl*-H2*), or the relative power of higher frequencies compared to lower frequencies in speech, and periodicity (e.g. CPPS), or how regularly the vocal folds vibrate. The results are in many ways consistent with previous studies, as women produced significantly more creaky voice than men.

While we wouldn’t question the robustness of this overall pattern, it is important to look beyond binary gender difference to identify two other robust patterns. Figure 2.4 plots Hl*-H2* (an acoustic measure of spectral tilt that capmres the degree of glottal constriction during phonation) as a function of age for both female and male speakers. Lower values of Hl*-H2* indicate a great degree of glottal constriction, thus creakier phonation. The first pattern to note is that women exhibit a curvilinear pattern across the age span, that is, younger women creak the strongest, middle-aged women creak the weakest, and older women produce stronger creak than middle-aged women. This is a robust quantitative pattern, as a quadratic term for age emerged as significant for female speakers in a mixed-effects linear model, which tests for the effects of social factors while controlling for the influence of other social and linguistic factors, as well as random variation from one speaker to the next. So statements like ‘women creak the most' erase intragender differences and falsely characterise women as exhibiting uniform patterns. To return to our earlier

Figure 2.4 HI *-H2* (dB) as a function of speaker age for women (F) and men (M). Lower values indicate a greater degree of glottal constriction (creakier phonation).

discussion, we need to ask the other question, ‘When I see gender difference, where is age?’ We also note that even though women produce creakier phonation than men overall, young men in particular produce relatively strong creaky voice. Statements like ‘women creak the most’ incorrectly erase young men from the narrative of who creaks more and why. Callier and Podesva (2015) suggest that the noteworthy point about creak isn’t so much that women are doing it more, but that young people - women and men alike - are doing it more, because the domain of creak has expanded from its canonical phrase-final position to earlier in the phrase. A focus on gender alone would have obscured this pattern, yielding misleading insights about gender and failing to accurately characterise its linguistic patterning.

At this point, we can say on the basis of apparent time data that creaky voice is on the rise, and that young women (and young men) are leading the change. The question here is why. To arrive at an answer, we argue, we need to consider its social meaning. Though creaky voice was once more prevalent in the speech of men (Esling 1978; Henton and Bladon 1988), it is unlikely that women are using it to borrow on men’s authority. Apart from being easy and far-fetched binary thinking, it fails to consider that it has had a pragmatic use independent of gender for some time. Although media treatments of creak characterise it as functionally useless at best and physiologically harmful at worst (Garfield 2013; Steinmetz 2011), social evaluation studies have shown that creak is not always evaluated negatively. Eckext (2019), for example, has shown that younger listeners are more forgiving than older speakers in their assessments of very creaky speech samples from National Public Radio news personalities. Yuasa (2010) further reports that listeners judge speakers who produce relatively creaky speech as sounding professional. These studies hint at the possibility that creaky voice could serve a useful interactional or stylistic function.

Work in discourse analysis suggests that speakers can strategically use creaky voice to show disengagement or express negative affect. Lee (2015) shows that speakers commonly slip into creaky voice when they go off topic for a bit, suggesting that creak distances parenthetical speech from the main thread of a conversation. She goes on to argue that speakers can draw on this conventional function of creaky voice to distance themselves from the issue under discussion, even when speech is on topic. Grivicic and Nilep (2004) advance a similar claim in their analysis of the word ‘yeah’ in telephone conversations. They argue that creaky ‘yeah’ expresses either a disalignment between interlocutors or a dispreference to continue on the current topic. Zimman (2017) shows that a transmasculine speaker draws on creak to index ‘a stance of disaffectation, an aloof persona, or a kind of emotional stoicism'. In all of these studies, creaky voice distances the speaker from either their addressee or the topic of discussion. The recurrence of this function of creak across studies suggests that it may cany a more stable, conventionalised meaning, just like pitch and the realisation of /s/, as discussed in all three of the other chapters in this section.

In contrast to most discourse analytic work, one of the goals of variationist sociolinguistics is to look beyond immediate interactions and draw generalisations across a community. Yet it is difficult to generalise about the ways that groups employ linguistic features (as opposed to simply the frequency with which they employ them) without attending to the interactional concerns at the heart of discourse analytic work, such as stance-taking. To repeat a phrase from earlier in this chapter, meaning lies in use. Of course, generalisation requires the analysis of relatively large amounts of data, which preclude line-by-line analyses of turns at talk. Is it possible to operationalise interactional analysis in a quantitative variationist paradigm?

Podesva (2018) advocates for an approach that quantifies affective stance on the basis of its expression across modalities. Such an approach makes it possible to test the hypothesis that creaky voice conveys negative, disengaged affect. Starting with the most direct expression of affect, we can consider speakers’ own assessments of interactions. We should observe an inverse correlation between how much speakers enjoy interactions and how much they creak. Second, we can examine the sentiment conveyed through the lexical or semantic material of speech. Begitming with the lexicon, if creak conveys negative, disengaged affect, we should observe more creak on words that convey negative valence, low arousal, and low dominance. Finally, we can consider the least linguistically explicit, though arguably the most transparently readable, expression of affect - embodied affect. If creak indexes negative affect, we would expect to see higher rates when people are not smiling. And if creak conveys disengagement, we would expect to see higher rates when people are moving their bodies less.

These predictions were tested in a corpus of audio-visual recordings collected in the Interactional Phonetics Laboratory at Stanford University. The lab’s ‘Living Room’ has the acoustical specifications of a sound booth (thus enabling the collection of high-quality audio recordings), but it is staged like a living room (to enable speakers to relax into the environment) and features inconspicuous video cameras (to enable the analysis of embodied practices). Approximately 150 speakers - mostly a diverse group of Stanford undergraduate and graduate students - were recorded in dyadic unscripted conversations. Of these speakers, 42, all from the Western United States, were chosen for analysis. Data were acoustically analysed using the methods described above for Callier and Podesva’s

(2015) study. The valence, arousal, and dominance for each word was coded using the lexicons published by Mohammad (2018). Finally, computer vision methods were used to identify when speakers were smiling, using the method in Podesva, Callier, Voigt, and Jurafsky (2015), and how much they were moving, using the method in Voigt, Jurafsky, and Podesva (2013).

Results largely confirm the hypotheses. Creaky voice is more prevalent in interactions that speakers rate as less enjoyable. It is also used more commonly on words that convey less dominance. Finally, creaky voice correlates with both forms of embodiment. For female speakers, creaky voice is more common in phrases where speakers are not smiling. And regardless of gender, speakers creak more at moments of time when they are moving less. This recalls a similar finding in Pratt’s (2018) study based on a year-long ethnography of students at a high school for the performing arts. She found that students in the school drew a distinction between kids who were considered ‘chill' and those who were considered higher-energy ‘louds’. She found that ‘chill’ students not only shifted their sitting positions less over the course of sociolinguistic interviews with her, but they also used higher rates of creaky voice.

The data largely support the hypothesis that creaky voice conveys negative, disengaged affect, which enables us to finally return to the question of why young women (and young men) do it more often, why the feature is on the rise. In short, creaky voice enables speakers to disengage or take negatives stances without saying so overtly (cf. Besnier 1990). Eckert

(2016) has argued that a design feature of language is being able to convey social meaning, including affect, without expressing it through lexical content. The social meanings conveyed by creaky voice, and all sociolinguistic variables for that matter, do not alter an utterance’s referential meaning. As a result, it would be infelicitous to respond to someone who is creaking by asking, ‘Why are you disengaging from me?’ The disengagement is in the room, but it isn’t on the table, so to speak. The plausible deniability of disengagement is useful, perhaps most useful to women in interactions replete with ‘mansplaining’ and ‘manterruptions’.

Conclusion

Gender is the product more than the explanation for linguistic variation. Viewing gender through a performative lens encourages variationists to situate gender locally, in interaction, and in dialogue with other dimensions of identity. In spite of ideologies that cast gender as a rigid binary structure, gender is constructed in practice in non-binary ways. The remaining chapters in this section offer ways of conceptualising and analysing the non-binarity of gender and sexuality. Zimman deconstructs gender into four potentially distinct dimensions - gender assigned at birth, gender identification, gender role, and gender presentation - and argues that doing so is necessary not only for adequately characterising the linguistic practices of transgender individuals, but also for explaining intragroup diversity among members of any gender category. Levon surveys a body of work examining the diversity of linguistic practices among individuals that comprise half of the sexuality binary (‘LGB’ or 'queer') and another body of work that considers how sexuality is articulated in conjunction with other mutually constitutive identity categories. Campbell-Kibler and miles-hercules quantify how strongly listeners’ masculinity ratings of hypermasculine speech performances correlate with their gayness ratings. They find not only that listeners vary significantly with respect to the strength of this correlation, but that the correlation is weaker among African American listeners. All three chapters emphasise the intersectionality of gender and sexuality. In the present chapter, we have argued for an approach to variation that focuses on personae, as speakers use linguistic features to perform gendered personae (e.g. Eckert 2006 [1996]; Podesva 2007), and listeners draw on and update representations of these personae when evaluating and perceiving speech (e.g. Campbell-Kibler 2007; D’Onofrio 2015, 2018). All of this is possible because variation is meaningful. So variationists must continue to drill down, past gender, to locate the meanings from which gender is constructed, meanings that represent how gender is experienced.

Note

1 We note, also, that it has become commonplace to take women’s lead in the use of a variant as diagnostic of sound change - yet another example of reliance on a shaky binary.

References

Benor, S. (2001) ‘Sounding learned: phonological variation in orthodox Jewish English’, in Sanchez, T., and Johnson, D. E. (eds) Penn working papers in linguistics: selected papers from NWAV 29. Philadelphia: University of PA, Department of Linguistics, pp. 1-16.

Besnier, N. (1990) ‘Language and affect’, Annual Re 19, pp. 419-451.