1.1 Why Theorize the Third Wave?

The goal of this volume is to advance sociolinguistic theory. It focuses on a growing research area about how social information is encoded in language, known by some as the ‘third wave’ of variationist sociolinguistics (Reference EckertEckert 2012). While this approach has been growing in popularity in recent years, its aims and assumptions have not been explicitly spelled out. This book will provide empirically motivated, explicit statements about key concepts, their historical development, and their contemporary implementation in third-wave research. Each chapter takes up an important theme for theorizing the third wave and presents new empirical data that sheds light on the topic at hand. The papers collectively make a case for why attending to social meaning is vital to the study of variation while also offering a foundation from which variationists can engage with social meaning in productive ways. We examine variation at a variety of levels of analysis, including phonetics, morphosyntax, and semantics; from multiple analytical perspectives, including language production and perception; across a range of cultural contexts, including Europe, Asia, the Middle East, and North America.

Traditional first- and second-wave approaches have treated variation as a window into language change (Reference Weinreich, Labov, Herzog, Lehmann and MalkielWeinreich, Labov, & Herzog 1968), and have examined the stratification of variation according to both macrosocial categories (e.g., gender, class) and locally significant categories (e.g., jocks, burnouts). Proponents of a third-wave approach focus on linguistic variation as a resource for taking stances, making social moves, and constructing identity. These social practices are possible only because linguistic variants carry meaning and take on new meanings in situated interaction, and as components of styles. While this term has been used by numerous scholars to refer to a wide array of things (see discussions in Reference Rickford, Eckert, Eckert and RickfordRickford & Eckert 2001; Reference Schilling, Chambers and SchillingSchilling 2013), we follow Reference Irvine, Eckert and RickfordIrvine (2001) in viewing distinctiveness as its defining characteristic. Zhang (this volume) also takes this perspective in defining style as ‘an emergent system of distinction … constituted by linguistic and other semiotic resources and practices that make distinction meaningful’. Consequently, a key component of third-wave work is examining how meaning-making operates within stylistic practice.

In this chapter, we provide an overview of the third-wave approach, addressing what we see as its principles and aims. In doing so, we expand on the overview of the third-wave approach as presented in Reference EckertEckert (2012) and Reference EckertEckert (2019), while also offering new avenues for research. Our aim is to highlight what the third wave has to offer to a theory of social meaning in language, drawing upon the volume’s chapters to illustrate our discussion. As such, this introduction highlights the content of the volume’s chapters, but it also makes explicit the theoretical perspectives that unite scholars engaged in taking a social-meaning-based approach to variation as evidenced in current work in the field. We thus introduce the contribution of each chapter not through the traditional style (a final subsection with one paragraph dedicated to one or two chapters) but in an integrated way, referring to the chapters to demonstrate the growing diversity of the third-wave approach as well as the fundamental principles that unify their shared perspective.

The volume is structured around three sections. The papers in the first section, Where Is (Social) Meaning?, examine the question of how third-wave research characterizes meaning. It focuses on issues relevant at different levels of linguistic representation, including phonetics, phonology, morphology, syntax, semantics, and pragmatics itself. The second section, The Structure of Social Meaning, then takes up the issue of how meaning is structured, taking the indexical field (Reference EckertEckert 2008) as a starting point. It examines the structure of signs, how social meanings relate to one another, and how conflicts among meanings are resolved. The third section, Meaning and Linguistic Change, considers the role that social meaning might play in processes of language change. While an interest in social meaning developed largely independently of the issue of change (cf. Reference LabovLabov 1963), third-wave variationists have increasingly noted that attending to meaning can help elucidate the question of why linguistic change transpires in the way that it does. This final section will thus bring our volume back into dialogue with the broader field of variationist research. In doing so, it highlights the intersection of first-, second-, and third-wave research.

In the remainder of this chapter, we present the fundamentals of an approach to variation that centres on social meaning. We start by outlining what social meaning is and discussing how the connections between variable linguistic forms and their social meanings arise. Section 1.3 turns its attention to linguistic form. In particular, we underscore the centrality of markedness and offer thoughts on how a form’s domain of linguistic representation (e.g., syntax vs phonetics) may bear on its social meaning. While Sections 1.2 and 1.3 focus on meaning and form, respectively, Section 1.4 discusses the nature of the association between the two, highlighting the properties of underspecification and multiplicity. Section 1.5 theorizes how to conceptualize the emergence of meaning when considering multiple forms at once, as opposed to single linguistic features in isolation. Section 1.6 takes up the practical question of how one goes about studying social meaning; we review a number of approaches and identify their strengths and drawbacks. Finally, we conclude by offering some thoughts on where the study of social meaning might be headed. Our goal for this chapter is neither to enforce a status quo nor prescribe a path forward, but rather to establish a common ground from which a variety of sociolinguists can set out on new explorations.

1.2 What Is Social Meaning?

Social meaning is the set of inferences that can be drawn on the basis of how language is used in a specific interaction. That set of inferences may be linked to the pragmatic function of the utterance itself (Acton, this volume; Beltrama & Staum Casasanto, this volume). In the interaction above, Meg uses a right dislocated tag (the demonstrative pronoun these, which is co-referential with subject they in the preceding clause) and a tag question (‘aren’t they?’). We know that structures that occur at the right periphery of the clause, like right dislocation and tag questions, can be focusing and, as such, they can have expressive, evaluative and/or affective functions (Reference AshbyAshby 1988; Reference FretheimFretheim 1995; Reference Lambrecht, Haspelmath, Konig, Oesterreicher and RaibleLambrecht 2001). Research on tag questions has also suggested that their status – as part way between a declarative and a question – gives them a conducive function (Reference HudsonHudson 1975); they simultaneously express the speaker’s viewpoint, while encouraging a specific response to it (Reference KimpsKimps 2007: 272). So we might draw inferences based on the construction of the linguistic item we hear: in this case, that Meg is being evaluative and attempting to conduce agreement around the evaluation.

But we do not just draw inferences by reading the pragmatics of a construction. We also rely on inferences about the sort of person who produces the utterance, the situation they are in, the nature of the relationship between interlocutors, the speaker’s orientation to the content of the talk, and more. So, in the interaction above, Meg’s utterance gains social meaning if you know that, while Kim and Meg have been friends since the first year of school, Meg has recently been spending time with another group of girls who have a reputation for being more wild, daring, and rebellious than girls like Kim. This has created tension between Meg and Kim as they begin to explore their emerging differentiation. This distancing is evident in the distinction between Meg and Kim’s social practice, but it’s also there in Meg’s orientation to her talk: in the deixis of they and these, in the labelling of Kim and her friends as ‘posh’ (a pejorative term linked to the practices of those associated with higher social classes), and, as mentioned, in Meg’s efforts to conduce the fieldworker to align with her viewpoint.

Unlike semantically based inferences, the inferences drawn about social meaning are inherently indeterminate (Reference PodesvaPodesva 2007; Reference EckertEckert 2008; Maegaard & Pharao, this volume; Gafter, this volume). While the syntax of right dislocated tags might make them focusing and/or evaluative, they are not inherently negatively evaluative or directly indexical of a rebellious style of speech. The ability to read Meg’s right dislocated tag in this way comes from its stylistic framing: the syntactic configuration, the lexis, the phonology, and everything that we know about the interaction and its participants. Yet, neither researchers nor language users attend to all components of social meaning at the same time: variationist studies often focus on how a single variant works in a specific social interaction, and there is evidence that the disparate experiences of individual language users makes them pay attention to different aspects of language (e.g., Reference Hay, Warren and DragerHay et al. 2006; D’Onofrio, this volume; Drager et al., this volume). The computational complexity of interpreting social meaning in any given interaction makes the enterprise of third-wave sociolinguistics challenging, but also an area rife with possibilities for new research directions.

Meanings are also made indeterminate by the simple fact that the kinds of inferences that might be drawn are manifold. For example, language users make inferences about stance, which refers to those meanings that are constructed around evaluation of some object of talk and the alignment between interlocutors (Reference Du Bois and EnglebretsonDu Bois 2007). So too do they make inferences about persona, or the characterological traits interactants evoke through situated linguistic practice. At the same time, social types are related to larger ideological constructs like class, ethnicity, and gender (see Reference Moore and PodesvaMoore & Podesva 2009).

It is also the case that different components of a construction (like right dislocation) might link to different levels of meaning. So, while Meg’s use of deixis in they and these clearly portrays her stance towards Kim (and the other girls in the interaction), the simple fact that right dislocation is more commonly used by kids who are engaged in anti-school practices at her school might cause us to note something about her persona (although, of course, how frequency interacts with meaning is a complex question; see Reference Hay, Jannedy and Mendoza-DentonHay, Jannedy, & Mendoza-Denton 1999; Reference SnellSnell 2010). Likewise, Meg’s pejorative use of the word posh might lead us to recognize disassociation from a higher-class social type. Note that all of these meanings are contextual: how we understand the deixis in they and these requires us to understand who Meg is talking about and to; linking right dislocation to a persona type relies on our knowledge of the system of distinction she is embedded in; and her use of the word posh is triggered by her alignment in a specific discussion about taste and economic capital.

Each of these linguistic artifacts (deixis, right dislocation, use of ‘posh’) function as signs by Reference SaussureSaussure’s (1916) definition, in that they encompass a linguistic form that is associated with a meaning. However, following Reference Peirce, Houser, De Tienne, Eller, Lewis, Clark and Bront DavisPeirce (1895), Reference Silverstein, Basso and SelbySilverstein (1976, Reference Silverstein2003), Reference Eckert and CouplandEckert (2016), and Reference Gal and CouplandGal (2016), we might highlight the interpretative element of this process: signs are associated with meanings, but the precise meaning is an artifact of an interpretative process, and one that is ideologically mediated. For example, Gafter (this volume) shows that pharyngeal phonemes in Hebrew can signify any one of a range of potentially conflicting social meanings (historical and prescriptive accuracy; Mizrahi ethnicity; low socioeconomic status; Arabic first language). The social meaning(s) that listeners arrive at, however vaguely, can only be determined in the moment of use, dependent on the particular ideologies made relevant in context. That is to say that, while all linguistic forms have the potential to signify social meaning, a form only does so when our system of ideas and beliefs creates a link between the form and a type of social meaning (such as stance, persona, or social type). This is the process of indexicality, as articulated in linguistic anthropology.

At its core, indexicality is a process of association, where a linguistic form points to some dimension of its conventional context of use (e.g., Silverstein 1976, 2003; Reference Ochs, Duranti and GoodwinOchs 1992). In many variationist approaches to indexicality, the relevant dimension of context is the typical user of a particular form. For example, in suburban Detroit, negative concord has the potential to index burnout identity, as burnouts use the form more often than their jock peers. But associations with social types or social groups is just one of many dimensions of context, any of which could emerge as relevant. Another relevant dimension could be the kind of stances typically taken while using a form. The same feature, negative concord, can be indexically associated with the rebellious stances that its users typically take while using it (Reference EckertEckert 2000; see also Moore, this volume). Whatever type of social meaning we study, indexicality stands as a core concept in theorizing social meaning, as it represents one of the primary means through which the connection between a linguistic form and its social interpretation arises.

Indexicality is central to third-wave research. One of the foundational concepts is Reference SilversteinSilverstein’s (2003: 194) indexical order, which refers to the degree of ideological complexity of a linguistic form, where change to that complexity is ‘always already immanent’. Sociolinguists have built on the concept of the indexical order in different ways. On the one hand, Reference Johnstone, Andrus and DanielsonJohnstone et al. (2006) describe it as a hierarchical process such that the social meanings of language are initially unconscious but may, over time, become stereotypical. This process is imagined as operating as a movement through structured (and temporally dependent) orders of indexicality that can be related to Reference LabovLabov’s (1972) description of variables as indicators, markers, and stereotypes. Importantly, in this model of indexicality, the social meanings identified are almost exclusively related to the main correlations measured in first- and second-wave research (e.g., persona types, such as ‘Pittsburghers’, or social types, such as ‘working class’). In order to understand the ways in which persona and social type are embodied (and constructed) in interaction, others have invoked a more fluid indexical field (Reference EckertEckert 2008; Reference Moore and PodesvaMoore and Podesva 2009) – a semiotic space where potential meanings co-exist, but in which specific social meanings are only activated by the existence of co-occurring associations and/or certain ideological conditions. So, to go back to the previous example, and drawing on the relationship between stance and social meaning (see Reference Kiesling and JaffeKiesling 2009), if one is attending to Meg’s negative stance towards Kim, then this might facilitate an indexical link to a persona type which frequently makes negative and face-threatening evaluations (so a Townie persona in Meg’s school rather than a Popular one). But the indexical relationship might also operate from macro to micro. For instance, Meg’s use of right dislocation might be interpreted as constitutive of a working-class social type (by someone unfamiliar with the local dynamics of the discourse in which it appears). This might lead to certain stances being assumed which reflect ideologies about how working-class people present themselves (e.g., as brash and direct). As Reference Gal and IrvineGal and Irvine (1995: 995) note, there is no ‘view from nowhere’, and where we are standing determines what we perceive.

The indeterminacy of social meaning can lead to a number of different indexical paths. Indexical relationships can be transformed into iconic ones (Reference Irvine, Gal and KroskrityIrvine & Gal 2000) through a process of iconization, whereby the indexical link comes to be ‘as if the linguistic feature somehow depicted or displayed a social group’s inherent nature or essence’ (Reference Irvine, Eckert and RickfordIrvine 2001: 35). Iconic relations between form and meaning are grounded in local ideology, and, in so being, they can erase the indexical orders through which a form’s social meaning has developed. Reference ZhangZhang (2005; this volume), for example, details how rhoticity in Beijing Mandarin is iconized as an ‘oily’ sound befitting the ‘smooth operator’ persona and their slippery qualities. Similarly, a speaker’s small body is iconic of high fundamental frequency (f0, the acoustic correlate of pitch), given dominant ideologies of gender and age. In both of these cases, the forms (rhoticity and f0) can index a wide range of stances, but the iconic meanings serve to erase alternative inferences. Importantly, the process of iconization is culturally specific, as illustrated by Drager et al. (this volume), who argue for an iconic relationship between pitch and body size that challenges dominant Western ideologies. Their data from Hawai‘i demonstrate that a higher f0 is associated with larger body size, and that this association depends on the listener’s concurrent perception of the speakers’ ethnicity. This shows that even these iconic relations are ideologized and conventionalized.

Whereas indexical meaning points to dimensions of the context in which forms are used, and iconic meaning obscures the arbitrariness of such connections, semantic meaning refers to ‘the content conventionally associated with words’ (Beltrama & Staum Casasanto, this volume), presenting another path along which lexical and morphosyntactic forms can come to be connected to social meanings. In their work on the social meaning of the Italian intensifier, -issimo, Beltrama and Staum Casasanto show that listeners draw stronger social inferences when the intensifier is used on nouns (e.g., gelati-issimo). Importantly, listeners hear speakers as strikingly more outgoing, more excitable, and friendlier when they use -issimo in these contexts. Beltrama and Staum Casasanto suggest that these social meanings are not independent of the fact that the linguistic structure in question is an intensifier. In other words, the conventional semantic meaning of intensification bleeds into social interpretations of the speakers, or the social meaning.

Finally, the link between form and social meaning also derives from the points of contrast between alternatives, or systems of distinctiveness (Reference Irvine, Eckert and RickfordIrvine 2001). That is, the use of any given form may give rise to a particular interpretation specifically because it was used instead of another form that might have been used in the same discourse/pragmatic context. In our example, for example, Meg’s ‘these’ positions her very differently in relation to Kim and her friends than ‘these girls’ would have. In interpreting the utterance, Meg’s interlocutor might reasonably conclude that Meg is disaligning with Kim and her shopping preferences – not because ‘these’ on its own is inherently impersonal, but because it is less personal than ‘these girls’ would have been. Acton (this volume) explains that an approach to social meaning based on alternatives grew out of similar concepts in the fields of pragmatics and sociolinguistic variation, and goes on to offer an extended analysis of demonstratives like the one briefly discussed here.

Although we have identified four ways that the connection between a linguistic form and its social meaning(s) are forged, we do not wish to suggest that these operate independently of one another. Zhang’s (this volume) discussion of erhua and retroflex initials illustrates this point well; although there is something iconic about the surface realization of retroflexion, meaning also arises from listeners’ recognition of the fact that another form could have been uttered. And even when linguistic forms are highly iconized qualia, or sensuous qualities (Reference GalGal 2013: 32) ‘[s]uch terms always occur in contrast sets’. For instance, when certain types of political discourse are described as ‘plain’ speech, this is contrasted with more elaborate or florid styles, and when Beijingers are described as ‘oily’, this is contrasted with a more corporate style.

1.3 How Does the Form Affect Social Meaning?

What do we currently understand about how the form (or ‘sign vehicle’) itself affects the social meaning? As Acton (this volume) notes, the markedness of a form is key to social meaning. Here, markedness relates to occurrences of a form which are more noticeable – perhaps because they are less frequent, or because they violate dominant social norms, or require more interpretive effort. Hence, markedness alludes to asymmetry between linguistic variants. The markedness of a form is fundamental to understanding its social meaning, because more marked forms are more likely to accrue meaning than less marked forms. This is exemplified by Reference SilversteinSilverstein’s (2003) analysis of the tu/vous distinction. A speaker’s choice to use vous for ‘you’ in French is ideologized as ‘marked or elevated in value’, whereas tu is ideologized as ‘neutral’ (Reference SilversteinSilverstein 2003: 209). This markedness arises from a construal in person reference, from ‘literal’ (e.g., surface form of 2nd person plural used for 2nd person plural) to ‘figurative’ (e.g., surface form of 2nd person plural used for 2nd person singular). This creates an asymmetry, which is then ideologized as marking social value: first deference, then honorification. Markedness may also arise from a mismatch between underlying semantic meaning and grammatical formulation. For instance, Beltrama and Staum Casasanto (this volume) show that listeners have to rely on pragmatic inference to decode intensifiers like ‘totally’ when they are used with non-gradable predicates. Intensifiers typically have semantic meaning related to reaching the top, or at least the very high region, of a bounded scale. In Beltrama and Staum Casasanto’s study, listeners generally perceived speakers using totally as excitable. However the degree to which a speaker sounds excitable depends upon whether or not they use totally with predicates that are graded (e.g., the bus was totally full) or non-gradable (e.g., she was totally born twenty years ago). In this comparison, the non-gradable predicate is more marked because we have to work harder to understand the pragmatics: because we can’t be ‘a little bit’ born twenty years ago, we understand the use of totally to reflect the speaker’s attitude to being born twenty years ago, rather than the extent to which they were actually born twenty years ago.

The precise nature of the relation between markedness and meaning is yet to be fully theorized, in part because markedness itself may derive from different sources. For example, one way in which a variant might be more marked is if it occurs with low frequency. But what constitutes ‘low frequency’ differs with respect to the speaker, listener, or interactional context. For example, among all the readers of this chapter, there will be many who find the right dislocation in ‘They’re all posh, these’ to be marked, but there will be some for whom it is less marked than other syntactic alternatives. As another example, Hall-Lew, Cardoso, and Davies’ and Starr’s papers (this volume) both detail changes in markedness over the course of a community sound change, showing how frequency-based markedness can shift between generations, in dialogue with the community’s changing structure.

The markedness of form can be due to factors other than frequency, and these are also relative. Research like Drager et al.’s (this volume) shows how the connection between the sign vehicle (low pitch) and the sign object (sassy, arrogant) is mediated by the perceived persona (gay cis man; Hawaiian ethnicity). Of course, this perception might vary. D’Onofrio’s paper (this volume), also shows how the markedness of vowel perception depends on the listener’s cognitive representation of the kind of speaker they are listening to, and this representation may be dynamically updated based on immediately prior experiences. The markedness of a form will also be influenced by the markedness of the other forms with which it co-occurs, as demonstrated, among others, by Maegaard’s and Pharao’s analysis of segmental and prosodic variation (this volume). While this kind of markedness may be in part due to a kind of frequency effect – for example, the listener’s personal experience of a given frequency relative to a type of speaker, or the forms with which it co-occurs – markedness may also be related to features of the form itself. For example, Podesva’s analysis of vowel fronting (this volume) shows the greater stylistic potential of those variants that are realized with jaw lowering, perhaps because of the context-specific indexicalities of an open-jaw articulatory posture and the markedness of this posture. The manifestation of markedness may also differ, in part, between levels of linguistic representation, given differences between the frequency and distribution of phonetic and phonological variation on the one hand and morphological, lexical, and syntactic variation on the other, and their abilities to allow for and constrain different kinds of stylistic expression. The relevant aspects of linguistic structure that contribute to the markedness of phonetic or phonological features will typically differ from those that condition morphological or syntactic features. For example, Starr’s study (this volume) of low back vowel realization depends on marked differences between Singaporean forms and British form on the one hand and American forms on the other, along with the phonological contrasts between different low back vowels that differ by variety. In contrast, Acton’s analysis (this volume) depends on the relevant dimension of markedness for different demonstratives is whether or not anthropomorphization is part of the demonstrative’s semantic entailment.

Beyond differences in markedness, the effect of form on social meaning differs more generally with respect to the level of linguistic representation. One key difference is between variables capable of scalar variation, on the one hand, and variables that are only discrete. For variables with a scalar dimension, meaning may map onto only parts of the scale (Reference Podesva, van Hofwegen, Levon and MendesPodesva & Van Hofwegen 2016; Reference Dickson and Hall-LewDickson & Hall-Lew 2017; Reference D’OnofrioD’Onofrio 2018). Furthermore, that form–meaning relationship may also differ from, or augment, the form–meaning relationship for the same variable at a discrete level. Reference PodesvaPodesva (2006), for example, shows how presence of a released word-final /t/ is recruited for stancetaking in one domain (i.e., precision), while variation in the phonetic realization of that variant can be recruited for other kinds of stancetaking (i.e., only long, intense releases index prissiness). Debates about the semiotic potential of discrete (or ‘digital’) signs, on the one hand, and continuous (or ‘analog’) signs, on the other, goes back to at least Reference Lévi-Strauss, John and WeightmanLévi-Strauss (1969: 28; see Reference ChandlerChandler 2017: 184), with the classical distinction framing the continuous as more ‘natural’ or ‘authentic’ than the latter. Nowadays, the important semiotic distinction between the two is rather that discrete signs ‘impose digital order on what we often experience as a dynamic and seamless flux’ whereas continuous signs ‘can signify infinite subtleties’ which ‘blend into one another’ (Reference ChandlerChandler 2017: 184). Consequently, the socioindexical analysis that is possible for any given study is in part dictated by the linguistic variable in question.

Another general difference between levels of linguistic representation is that while social meaning in the sound domain is likely to be iconic (either from the start or eventually), it has been argued that the meanings of morphosyntactic variation are typically understood in relation to standardness (e.g., delinquent, tough) (Reference EckertEckert 2019). Morphosyntactic variables (at least those typically studied by variationists) also tend to be more sharply stratified than phonological ones (Reference CheshireCheshire 1999: 61), and this patterning may limit the range of social meanings they can acquire. For instance, if a form – such as nonstandard were (e.g., ‘I were really happy’) – correlates strongly with working class speakers, then this may constrain the social meanings it can develop (such that its social meanings are linked to attributes which are also indexically linked to the working class, like ‘toughness’ or ‘resilience’). On the other hand, if a phonetic variable can be used in a scalar way, then there is more potential for it to signify a wider range of social meanings. So, while word-final /t/ release might correlate with higher social class groups, the ability to produce long, intense releases exploits articulatory movement to communicate affect. Reference EckertEckert (2019) describes a ‘cline of interiority’ with variables capable of indexing internal, personal affective states at one end, and those which most typically index external public social facts at the other. However, the extent to which this cline of interiority maps onto different levels of linguistic representation remains an open question, especially given that variationist research on the social meanings of syntax lags behind understanding of the social meanings of phonetic and phonology (see Acton, Beltrama, & Staum Casasanto, and Moore, this volume).

1.4 What Is the Nature of the Association between Form and Social Meanings?

The association between a form and its social meanings is indeterminate, by virtue of its underspecification (e.g., Reference PodesvaPodesva 2007) and multiplicity (e.g., Reference EckertEckert 2008). It is not, however, chaotic or unconstrained, but structured by historical possibility.

Underspecification refers to the insight that the form–(social) meaning relationship is inherently vague. It emerges in a stylistic context, rather than being necessarily predetermined. This is one of the distinguishing features of the third-wave approach to variationism (Reference EckertEckert 2012): while the form–meaning relationship in the first and second waves is observed to be something that could be uncovered by the analyst (e.g., a ‘working class variant’ or a ‘regional dialect feature’), the third-wave argument for underspecification posits that that form–meaning relationship is determined by how it is deployed in interaction.

The third-wave approach to variationism also shifts analytical focus towards the ambiguity and multiplicity of meaning: how a speaker navigates all the possible meanings available in a given interaction, and what those meanings are, and why. This mapping of one form to multiple potential meanings, or multiplicity, is perhaps most straightforwardly understood with reference to Reference EckertEckert’s (2008: 435) concept of the indexical field, ‘a constellation of ideologically related meanings, any one of which can be activated in the situated use of the variable’. The tag-cloud method used by Drager et al. (this volume), for example, effectively demonstrates the vast range of social meanings that might be indexed by a single variant. Maegaard and Pharao (this volume) argue that a variant can have multiple indexical fields, where the field that a listener orients to is influenced by the preceding speech. Gafter (this volume) shows how the meanings in a variant’s indexical field need not be uniform or compatible, and can even be contradictory, or exact opposites (see also Sharma, this volume). Similarly, underspecification and multiplicity also mean that competing variants might have complementary or orthogonal meanings rather than opposite meanings (Hall-Lew et al., this volume; Reference Campbell-KiblerCampbell-Kibler 2011b).

1.5 What Happens When Multiple Signs Come Together?

Thus far, we have been conceptualizing social meaning in terms of single signs in isolation, and indeed, many studies concerned with social meaning take single linguistic features as their objects of study. But in situated linguistic practice, any given sign is produced alongside a multitude of others. In the example at the start of this chapter, for example, Meg engages in lexical (posh), syntactic (right dislocation), and discourse (tag question) practices that – separately and together – convey social meaning. In this section, we consider the range of stylistic processes through which multiple signs conspire to produce meaning. As we will see, meaning emerges in a variety of complex ways.

Perhaps the most straightforward approach to theorizing how meaning emerges from the co-occurrence of linguistic features is the compositional approach. Such an approach begins with the assumption that each linguistic feature is associated with its own set of meanings. The meaning of a style consisting of several features can then be reliably derived by relating the meanings of the style’s component features. For example, Reference PodesvaPodesva (2008) argues that a caring doctor persona emerges when a medical student frequently releases word-final coronal stops (which indexes competence in a professional environment), produces rising intonational contours on assertive declaratives (which facilitates taking a talk-sustaining, non-threatening stance), and occasionally phonates in a weak falsetto just on discourse markers (which he argues indexes restrained expressiveness) when meeting with a patient. Here, the emergent social meaning is the union of the style’s component features’ meanings, since all such meanings are compatible with one another. In cases where meanings are incompatible, meanings are intersective. In Copenhagen Danish, for example, the meaning of fronted /s/ is ambiguous between sounding ‘homosexual’ and ‘immigrant’ (Reference Pharao, Maegaard, Møller and KristiansenPharao et al. 2014). But when fronted /s/ occurs in the context of ‘street’ prosody, its associations with sexuality disappear, and speech is interpreted as overwhelmingly ‘immigrant’. In this example, the meaning of one feature (‘street’ prosody) disambiguates the meaning of another (fronted /s/).

While a compositional view of social meaning is both useful and necessary – as any given linguistic feature is meaningful in its own right – it can offer only a partial account of the way that meaning emerges from co-occurring linguistic features. Compositional approaches assume that the meanings of features pre-exist and exhibit some degree of stability. But speakers constantly exercise stylistic creativity by lifting features out of familiar contexts and packaging them in new ways, through a process known as bricolage (Reference Lévi-StraussLévi-Strauss 1966; Reference HebdigeHebdige 1979). At its essence, bricolage is a process of recontextualization, a ‘transformational’ process whereby some aspects of meaning are retained from a feature’s earlier contexts and, crucially, new meanings emerge (Reference Bauman and BriggsBauman & Briggs 1990: 75). For example, when Chinese yuppies embed English technical terms (marketing, securities, proposal) in Cosmopolitan Mandarin speech, the terms acquire a host of new social meanings (e.g., a global orientation) that are not present in their conventional English-language contexts. The meaning(s) of any given sign can therefore not be interpreted independently from the meaning(s) of other signs.

A compositional approach is also incomplete on its own because clusters of linguistic features (rather than individual features alone) can themselves function as sign vehicles. We might label this a constructional approach to social meaning. African American drag queens, for example, can draw on obscenities and features of white women’s language to construct a (non-white) drag queen style (Reference Barrett, Bucholtz, Liang and SuttonBarrett 1999). Similarly, young people in Vollsmose, Denmark, are seen to combine ethnolectal and local regional (Funen) dialect features, and to do so in gender-differentiated ways to construct different persona (Quist, this volume). In both of these cases, the meaning that emerges from the co-occurrence of features cannot be straightforwardly derived from the meanings of individual features. To model the unfolding of meaning construction in real time, Tamminga (this volume) details a method of quantifying microcovariation between six different vocalic variables, theorizing the pairing of regression-based models and discursive evidence to push the ‘[q]uantitative frontiers for social meaning’.

The fact that meaning can arise in constructional fashion does not invalidate approaches that assume compositionality. Instead, it appears that individual features as well as unique combinations of features both contribute to the interpretation of a unit of speech, and an adequate theory of social meaning must permit both possibilities.

While a theory of social meaning that can handle the simultaneous and joint contributions of multiple signs has yet to be fully articulated, such a theory must include the following properties. First, individual signs differ in the degree to which their principal social meanings have been conventionalized, or enregistered (Reference AghaAgha 2003, Reference Agha2007). Often in these cases, the meanings of the most heavily conventionalized signs (what Reference Campbell-KiblerCampbell-Kibler 2011a compares to a red stiletto) are stable across very different social contexts and can neutralize the social meanings of other, less conventionalized signs. In cases of indexical shift, some signs may be undergoing de-enregisterment at the same time as new signs are becoming more conventionalized (Hall-Lew et al., this volume). Second, and in a related vein, the order in which signs are spoken in an utterance may influence the kinds of inferences listeners draw (Maegaard & Pharao, this volume; see also Reference Levon and FoxLevon & Fox 2014). Finally, levels of linguistic representation do not operate independently of one another. For example, Moore (this volume) illustrates the value of conceptualizing syntax as ‘housing’ and thereby constraining the meaning of phonetic features that appear within a particular syntactic structure. Along the same lines, it is reasonable to think that the iconic meaning of intensification (i.e., extremeness) can constrain the kinds of phonetic resources that are produced during the act of intensification; that is, low intensity, low pitch, and reduced vowels are generally incompatible with extremeness.

1.6 How Do We Study Social Meaning?

Thus far, we have taken for granted that social meanings can be observed or inferred by scholars of social meaning. In point of fact, the analytical practice of identifying social meanings is far from straightforward, and researchers have followed a number of strikingly contrastive approaches. While an exhaustive discussion of the full range of methods is not possible here, we briefly sketch out the dominant approaches and discuss some of the advantages and limitations of each.

The earliest approach taken in modern variationist sociolinguistics is the triangulation of meaning through a series of correlations. Reference LabovLabov (1963) found that the centralization of (ay) and (aw) was more common among fishermen (whose livelihood was most affected by the flourishing tourist economy), boys who are committed to staying on the island, and Vineyarders with positive orientations to the island. Taken together, we have strong evidence for Labov’s claim that centralization indexes a pro-island stance. No one of these correlations constitutes robust evidence for Labov’s interpretation on its own, but together they make a more compelling case. A major strength of the triangulation approach is that it can be taken with nearly any corpus, provided that there are sufficient numbers of speakers, and sufficient numbers of social factors.

In spite of the advantages of triangulation, the approach has a number of limitations. First, it is unclear what constitutes sufficient evidence. How many correlations must be shown to corroborate a claim about meaning? Second, meanings are unlikely to converge on a single social meaning (cf., e.g., Reference TrudgillTrudgill’s 2008: 244 myopic framing of identity as ‘national identity’). As the chapters by Gafter, Hall-Lew et al., and Maegaard and Pharao (this volume) illustrate, the indexical fields of some features might contain radically divergent, even conflicting, meanings. Finally, it is not always clear, a priori, which correlations should be considered. Traditionally, in the kind of variation studies Reference EckertEckert (2012) has labelled ‘first-wave’, linguists have sought correlations between sociolinguistic variants and membership in predefined demographic categories typical of the Western sociological tradition, like age, gender, race, and class. While correlations of this type reveal a great deal about the trajectory of linguistic change, they do not tell us much about the social meaning of variation; the most specific meaning we can attain under this type of analysis is a social address, which we might view as an age-gender-race-class coordinate.

The second approach to studying social meaning, ethnography, enables researchers to observe which distinctions matter in a community and also make connections with its members, who might shed light on the linguistic features we linguists are interested in. Moore’s (this volume) two-year ethnography with girls at Midlan High led her to identify four salient groups of girls at the school. One of these groups, the Townies, exhibit a strong anti-school orientation and partake in practices deemed dangerous by their non-Townie peers, such as drug-taking and sexual activity. Eden Village girls, by contrast, are the ‘good girls’ who wear pastels and glittery accessories. Towards the end of her fieldwork, Moore asked Eden Village girls to reflect on their use of tag questions, asking them if they say ‘innit?’ One girl, Leah, responds, ‘No, we don’t say “innit”. That’s Townies,’ and goes on to offer ‘int[t^h] it’ as the canonical Eden Village form. This example reveals not only that released /t/ is salient in this community, but that it serves to differentiate locally salient social groups: Townies and Eden Village Girls (see Reference Moore and PodesvaMoore & Podesva 2009).

The ethnographic approach is exemplified in a number of chapters in this volume, including Moore, Hall-Lew et al., Quist, and Zhang. While the ethnographic approach provides ethnographers with both access to and insight into communities under investigation, the analysis is reliant upon the researcher’s interpretation of the practice they observe and how it correlates with language in use: community members might not be able to articulate the social meaning of most types of linguistic features, particularly less salient features (Reference Eckert and CouplandEckert 2016). Models like Silverstein’s indexical order and Eckert’s indexical field clearly show the distributed, ideological (and therefore cultural) location of social indexicality, which necessitates that, by definition, most indexical relations lie below the level of conscious awareness.

Experimental methods are well suited for identifying the meaning of less salient features. Among these, social evaluation methods are used most commonly in sociolinguistics, including by Maegaard and Pharao, Beltrama and Staum Casasanto, and Drager et al. in this volume. These typically take the form of a matched guise study, in which researchers obtain assessments of stimuli that differ in terms of a single linguistic feature. Often, guises are constructed through the acoustic manipulation of speech (e.g., Reference LevonLevon 2006; Reference Campbell-KiblerCampbell-Kibler 2007). Listeners are asked to evaluate speakers on a number of social dimensions (with the best studies basing these dimensions on concepts found to be relevant to the community under study), and differences in social evaluation are attributed to differences in guise. For example, Campbell-Kibler (this volume) shows that stimuli containing fronter /s/ are perceived as less masculine than otherwise identical stimuli with backer /s/.

Given the cognitive complexity of social evaluation tasks (i.e., listen to a form, decide what is being said, reflect on how it was said, and evaluate a speaker on the basis of how it was said), scholars have begun using more implicit approaches for accessing social meaning. One class of these is speech perception paradigms where researchers disclose social information about a speaker before asking listeners to identify what they heard (e.g., Reference NiedzielskiNiedzielski 1999; Reference Hay and DragerHay & Drager 2010; also Campbell-Kibler, this volume). For example, listeners who are told that a speaker is a business professional are more likely to classify a backed trap token in the frame blVk as black than block (Reference D’OnofrioD’Onofrio 2018). This classification sheds light not only on how speech is perceived, but also reveals that speakers conventionally associate backed trap with a business professional persona (one aspect of the vowel quality’s social meaning). Other, even more implicit, tasks include the false memory (D’Onofrio, this volume) and eye-tracking (Reference D’OnofrioD’Onofrio 2015) paradigms.

Experimental methods are attractive because they enable researchers to isolate specific linguistic features. They are particularly useful for hypothesis testing, in cases when researchers can independently motivate potential social meanings of the features in question. Conversely, they are not particularly well suited for exploratory work. In the matched guise paradigm, for example, the only social meanings that can be identified are those corresponding to the social attributes on which listeners are evaluating speech. Another limitation of most experimental methods is that subject pools are either too limited (i.e., college undergraduates) or too general (i.e., crowdsourcing platforms like Amazon Mechanical Turk or Prolific Academic). Given that some social meanings are specific to particular communities, it would be useful to administer experimental studies to members of the communities in question, as in Reference Flores-BayerFlores-Bayer’s (2017) study of the sh-ch alternation in East Austin, Texas, or Reference LawrenceLawrence’s (2017) study of goat and goose variation in York, England. For smaller communities (of practice), experimental methods are not viable options, as listeners as community members are likely to recognize the voices under analysis, which would compromise a considerable degree of experimental control.

While experimental methods are useful for establishing listeners’ connections between linguistic forms and their interpretations, they leave open the question of whether speakers exploit social meanings to achieve social ends. One means of investigating the strategic use of linguistic variants is to examine patterns of intraspeaker variation. For example, Reference PodesvaPodesva (2007) shows that Heath, a gay medical student, produces falsetto more often and with greater phonetic strength when talking with friends at a barbecue than when talking with a patient or his father. He argues that Heath recruits falsetto to construct a ‘diva’ persona at the barbecue – a situation where Heath has very different interactional goals from the other two situations. Apart from cross-situational analysis, intraspeaker variation can be examined in a variety of other ways. Several contributions in this chapter look intra-situationally to identify stretches of speech when speakers exhibit marked shifts in variation patterns, under the assumptions that these shifts in form indicate shifts in social meaning. Hall-Lew, Cardoso, and Davies (this volume) locate the timing of an indexical shift among those speakers who show topic-based style-shifting, as compared to the older and younger speakers, who do not. Sharma uses the Lectal Focusing in Interaction (LFI) approach (Reference Sharma and RamptonSharma & Rampton 2015; Sharma, this volume) to identify shifts in stance and footing, and Tamminga (this volume) advances a bottom-up approach to quantitatively identify when a particular linguistic feature exhibits peaks or valleys. Other contributions in the volume (Hall-Lew et al., Gafter, Starr) use word lists or reading passages (compared to interviews) to investigate what speakers’ orientations to read speech reveals about their language ideologies and participation in standard language markets. Importantly, these studies recognize that culturally specific, historically situated ideologies about reading can provide better explanations for reading styles than attention paid to speech. All of these intraspeaker studies assume that intraspeaker variation provides a window into speaker agency, where agency (or performativity) may operate either above or below the level of consciousness.

The intraspeaker approaches discussed can capture shifts in style where social meaning is relatively stable. In other words, shifts must last sufficiently long that they can be observed and identified using quantitative methods. We hasten to point out that shifts in linguistic form can be much more ephemeral, and stress that qualitative methods can be used to analyse such shifts. Although variationists are rarely trained in the methods of discourse analysis, it offers a principled, rigorous approach for analysing how linguistic variables are strategically deployed in interaction (e.g., Reference Schilling-EstesSchilling-Estes 2004; Reference CouplandCoupland 2007; Reference Kiesling and JaffeKiesling 2009; Reference Moore and PodesvaMoore & Podesva 2009). We feel compelled to point out here that discourse analysis is not equivalent to analysing the content of speech; it is a grounded approach that attends both to linguistic form and interactional dynamics. Under ideal circumstances, qualitative analysis will be conducted against the backdrop of a larger quantitative study, which can provide a sense of the range of variation possible in a community. Zhang (this volume), for example, deeply investigates the speech of Rebecca, a Chinese yuppie whose individual patterns are best understood in relation to the wider community patterns published in Reference ZhangZhang (2005). Also, although we present discourse analysis as a distinct approach from the examination of intraspeaker variation patterns, the approaches can be successfully combined. Sharma (this volume) uses LFI to identify moments where Indian English exhibits a spike and then correlates this with an interactional position (e.g., countering doubt) that can be independently motivated based on an analysis of the discourse.

This last example illustrates perhaps the most important point of this section, which is that methods will ideally be combined to provide the fullest analyses of social meaning. Some methods are better suited than others for hypothesis generation, while others are better suited for hypothesis testing. We think there is a need for more hypothesis testing in third-wave variationist work (see Acton, Beltrama and Casasanto, and D’Onofrio, this volume, for examples). Some methods are better suited than others for analyses of community-specific meanings, while others are better suited for analysing more heavily conventionalized social meanings that are shared across communities. So even while we advocate for more hypothesis testing, we simultaneously underscore the need to examine how social meaning plays out in specific communities (see Hall-Lew et al., Moore, and Quist, this volume). And finally, some methods are better suited for identifying meanings that listeners infer, while others are better suited for identifying meanings that speakers exploit.

1.7 Looking Forward

The intellectual tradition of a variationist approach to social meaning organically favoured community studies of speech production, with a focus on ethnographic fieldwork and attention to patterns in variable realization, especially in English. This history has resulted in an imbalance of evidence and areas of theorization, and points to several clear areas of development in the field of social meaning and linguistic variation.

There continues to be a dearth of third-wave research on non-English varieties, relative to other subfields of linguistics. While the current volume includes analyses of Danish, Hebrew, Italian, and Mandarin, and while there are other notable examples of non-English work (e.g., Reference StanfordStanford 2010), an obvious way forward for our field is the incorporation of a much greater diversity of language varieties, especially those from the Global South and non-WEIRD (western, educated, industrialized, rich, democratic) sociocultural contexts.

Starr (this volume) and Hall-Lew, Cardoso, and Davies (this volume) further show how the components of indexical fields reconfigure at a broader timescale, from generation to generation, relative to the wider sociohistorical context. These analyses were driven, however, by a desire to accurately model the association between the form and its social meanings, rather than by an a priori drive to model change in social meaning. Attending to the nature of the form–meaning link makes visible the fact that variationism’s focus on changes in form has been to the detriment of simultaneous changes in meaning, even though both kinds of change (potentially) happen together. Here we suggest that variationist sociolinguistics should do better to centre questions of meaning (e.g., Reference WongWong 2005; Reference McConnell-GinetMcConnell-Ginet 2011). For example, while both Beltrama and Staum Casasanto’s paper on intensifiers (this volume) and the paper by Acton on definites (this volume) are essentially synchronic studies, they inspire variationists to think of new ways of theorizing change in social meaning, perhaps through drawing clearer connections to research from historical linguistics such as reanalysis and grammaticalization, or research from computational linguistics on sentiment analysis.

A more radical approach to centring meaning would be to dispense with the notion of the linguistic variable (i.e., more than one way of saying the same thing), in recognition of the fact that sometimes/often/always using a different linguistic form results in saying something rather different. As Moore (this volume) notes, Reference CheshireCheshire (1987: 257) has argued that misplaced focus on the linguistic variable has ‘prevented any real progress being made in our understanding of syntactic variation’, but it’s possible that the notion of the linguistic variable has also retarded our understanding of the social meaning of language variation at all levels of linguistic representation. Rather than organizing an analysis around a linguistic variable (like intensifiers), one might centre the meaning of intensification and examine the totality of linguistic features recruited at moments of intensification. Such an approach would force us to take a stylistic approach that considers how multiple features work together and yield a more explanatory theory about how different levels of linguistic representation relate to one another.

One of the more undertheorized and exciting corners of the third-wave enterprise is a focus on the relationships between linguistic variation, social meaning, and cognition. While work by Campbell-Kibler (e.g., this volume) and D’Onofrio (e.g., this volume) is beginning to query a cognitive representation of indexicality, there remains a great potential for developing the field as an area of cognitive science. For example, Campbell-Kibler’s paper argues that a variety of linguistic practices in production and perception rely on different cognitive representations of the sign; this general and yet precise hypothesis has the potential to motivate new studies that could interface with the existing fields of language attitudes and psycholinguistics.

Another clear area of advancement for the study of linguistic variation and social meaning is a better engagement with parallel theoretical traditions. Reference SoukupSoukup (2018), for example, makes a strong case for situating third-wave work (more) explicitly with respect to interactional sociolinguistics, since both appear to draw on essentially an agentive, dialogic model of communication (Reference Bakhtin, Emerson, Holquist and McGeeBakhtin 1986 [1953]). Soukup argues that a shift in focus in third-wave theory towards the interaction order (Reference GoffmanGoffman 1983; Reference SchiffrinSchiffrin 1994) and contextualization (Reference GumperzGumperz 1982; Reference Bauman and BriggsBauman & Briggs 1990) may aid the field’s current analytic goals, such as a strong focus on the contribution of the listener to the meaning-making process. While there is no question that variationist studies of social meaning have been influenced more or less by interactional sociolinguistics, it is also the case that the field can benefit from deeper engagement as it continues to grow.

The three editors of this volume are friends, who were fortunate to have been mentored by Penny Eckert – the person most closely associated with the third wave. We have similar intellectual influences and similar perspectives on what the interesting questions are about linguistic variation. In the process of writing this introduction, however, we nonetheless encountered areas of disagreement where we discovered diverging understandings of some key concepts. Are all forms always already signs? Are all forms always potential signs, but not signs unless used? Are there some forms that are not ever signs? Each of these three perspectives was taken up by one of the editors of this volume, in spite of our having all been mentored by Penny Eckert. Indeed, most of the chapters in this volume, diverse as they are, have been authored by Eckert’s students, and all chapters were inspired by the contribution Eckert has made to the field of variationist sociolinguistics. The fact that the editors of the volume disagree indicates that there is still much work to be done. Consequently, one vision we very much share is that this volume will attract the interest of a diverse range of scholars who, inspired by Penny Eckert’s work, and what they read here, will share our goal to further increase our understanding of the relationship between social meaning and linguistic variation.