What Is Identity? Your identity is, very simply, who you are. It has two basic aspects: your name, which serves first of all to single you out from others (the deictic function), and then that deeper, intangible something that constitutes who you really are, which we might think of as the meaning of your name (the semantic function). Names are obviously pieces of language, but less obvious is the role played by the language you speak, and how you speak it, in constructing that other, deeper aspect of your identity, both for yourself and others. Imagine a group of strangers at a bus stop. The bus they are waiting for drives past without stopping, and the following remarks ensue: A: Well fuck me. B: I say, wasn’t that the 23? C: I’m like not believing this. Picture in your mind what A, B, and C look like. You can probably say something about how they are dressed, their background, what they do, what they are like, and whether you would like them or not. It is extraordinary how much we are able to infer from what are, after all, a few squiggles on a page. We have an instinctive capacity to construct the identity of a whole person in our minds based on minimal input, and it is most effective when the squiggles represent something the person said. Selfidentity has long been given a privileged role in identity research. But the identities we construct for ourselves and others are not different in kind, only in the status we accord them. The gap between the identity of an individual and of a group – a nation or town, race or ethnicity, gender or sexual orientation, religion or sect, school or club, company or profession, or the most nebulous of all, a social class – is most like a true difference of kind. Group identities seem more abstract than individual ones, in the sense that ‘Brazilianness’ doesn’t exist separately from the Brazilians who possess it, except as the other time-budgeting placed a severe upper limit on the size of groups they could maintain’’ (Dunbar, 1996: 77). Language made it possible to increase group size without losing social cohesion or the time needed to gather and hunt food. Because language can be directed at several people simultaneously, the rate of grooming is increased manifold. Dunbar notes, ‘‘Being able to assess the reliability of a prospective ally becomes allimportant in the eternal battle of wits’’ (Dunbar, 1996: 78–79). On the one hand, language serves the purposes of the individual who is seeking to make an alliance: ‘‘It allows you to say a great about yourself, your likes and dislikes, the kind of person you are; it also allows you to convey in numerous subtle ways something about your reliability as an ally or friend’’ (Dunbar, 1996: 78). It also serves the person being courted as a prospective ally: ‘‘Subtle clues provided by what you say about yourself – perhaps even how you say it – may be very important in enabling individuals to assess your desirability as a friend. We get to know the sort of people who say certain kinds of things, recognizing them as the sort of people we warm to – or run a mile from’’ (Dunbar, 1996: 79). He concludes, ‘‘Language thus seems ideally suited in various ways to being a cheap and ultra-efficient form of grooming. [. . .L]anguage evolved to allow us to gossip’’ (Dunbar, 1996: 79). If we think about language from this perspective, its primary purpose is no longer necessarily restricted to one of the two traditionally ascribed to it, communication (by a speaker having an intention and wishing to transmit it to listeners) and representation (of the universe, as analyzed into the logical categories that languages are thought to contain). Before either of these, and in many regards enveloping them both, language exists, in this reversed perspective, for the purpose of reading the speaker. Sociolinguistic inquiry into identity and language is concerned with how people read each other, in two senses. First, how the meanings of utterances are interpreted, not just following idealized word senses and rules of syntax as recorded in dictionaries and grammars, but in the context of who is addressing whom in what situation. Secondly, how speakers themselves are read, in the sense of the social and personal identities their listeners construct for them based on what they say and how they say it (a complex process, because most speakers’ output is already shaped in part by how they have read their listeners). Every day each of us repeatedly undertakes this process of constructing our reading of people we encounter, in person, on the telephone, on the radio or the screen, or in writing, including on the Internet, based on their language – what they say and how they say it. Targeting Identity in the Analysis of Language Modern linguistics has moved slowly but steadily toward embracing the identity function as central to language. The impediment has been the dominance of the traditional outlook, which takes representation alone to be essential, with even communication relegated to a secondary place. This outlook was never the only one available, however, and when early 20th century linguists such as Jespersen and Sapir came to investigate how language functions to define and regulate the role of the individual within the social unit at the same time that it helps to constitute that unit, they were not without predecessors. It was just that mainstream linguistics as it had developed within the 19th century was not inclined to see such questions as falling within its purview. An overview of the development of such inquiry within linguistics and adjacent fields can be found in Joseph (2004), from which some key moments will be excerpted here. The first is Labov (1963), a study of the English dialect of Martha’s Vineyard, an island off the coast of Massachusetts, where the diphthongs in words such as right and house are pronounced as [ey] and [ew] rather than [ay] and [aw]. This feature is not found in the dialects of the mainlanders who ‘summer’ on Martha’s Vineyard, and with whom the Vineyarders (year-round residents) have a complex relationship of dependency and resentment. ‘‘It is apparent that the immediate meaning of this phonetic feature is ‘Vineyarder.’ When a man says [reyt] or [hews], he is unconsciously establishing the fact that he belongs to the island: that he is one of the natives to whom the island really belongs’’ (Labov, 1963: 307). This is very much the sort of analysis of the effect of linguistic identity on language form that would characterize work in the 1990s and since, though it was sidelined in the mid-1960s by the statistical charting of variation and change. In the meantime, one particular identity focus – gender – led the way in directing attention to the reading of identity in language. Lakoff (1973) argued that, in both structure and use, languages mark an inferior social role for women and bind them to it. Gender politics is incorporated directly into the pronoun systems of English and many other languages, through the use of the masculine as the ‘unmarked’ gender (as in ‘Everyone take his seat’). Lakoff points to features that occur more frequently in women’s than in men’s English, such as tag questions, hedges, and pause markers, which as marks of insecurity and of the role women are expected to occupy are fundamental to maintaining the status quo in gender politics. As the notion of separate men’s and women’s language was accepted, the more general notion of the language–identity link was let in through the back door, leaving the way open for the study of group identities of all sorts beyond those national and ethnic ones traditionally associated with language difference. This was a challenge to a sociolinguistics that had been fixated on class differences. By the mid-1980s this shift was underway in the work of, for example, Gumperz (1982), Edwards (1985), and Le Page and Tabouret-Keller (1985), though it was really in the 1990s that it would come to occupy the mainstream of work in sociolinguistics and linguistic anthropology. This work also received significant input from social psychology, where one approach in particular needs to be singled out: Social Identity Theory, developed in the early 1970s by Tajfel. In the years following his death in 1982, it came to be the single most influential model for analyzing linguistic identity. Tajfel (1978: 63) defined social identity as ‘‘that part of an individual’s self-concept which derives from his knowledge of his membership of a social group (or groups) together with the value and emotional significance attached to that membership.’’ Within this simple definition are embedded at least five positions that in their time were quite revolutionary: that social identity pertains to an individual rather than to a social group; that it is a matter of self-concept, rather than of social categories into which one simply falls; that the fact of membership is the essential thing, rather than anything having to do with the nature of the group itself; that an individual’s own knowledge of the membership, and the particular value they attach to it – completely ‘subjective’ factors – are what count; and that emotional significance is not some trivial side effect of the identity belonging but an integral part of it. Beyond this, Social Identity Theory marked a break with other approaches in the fact that it was not concerned with analyses grounded in a notion of ‘power,’ but simply in the relative hierarchizations that people seem instinctively to impose on ourselves, most particularly in our status as members of ‘ingroups’ and ‘out-groups,’ which would come into even greater prominence in the ‘Self-Categorization Theory’ that developed as an extension of the original model, notably in the work of Tajfel’s collaborator Turner (see Tajfel and Turner, 1979; Turner et al., 1987; Turner, 1991). Partly under the influence of such work, sociolinguists were beginning to reorient their own object of investigation. Milroy (1980) reported data from studies she conducted in Belfast showing that the ‘social class’ of an individual did not appear to be the key variable allowing one to make predictions about which forms of particular linguistic variables the person would use. Rather, the key variable was the nature of the person’s ‘social network,’ a concept borrowed from sociology that Milroy defined as ‘‘the informal social relationships contracted by an individual’’ (Milroy, 1980: 174). Where close-knit localized network structures existed, there was a strong tendency to maintain nonstandard vernacular forms of speech – a tendency difficult to explain in a model such as Labov’s, based on a scale of ‘class’ belonging where following norms of standard usage marked one as higher on the hierarchy and entitled to benefits that most people desire. Labov’s early work on Martha’s Vineyard had suggested that the answer lay in identity, specifically in the value of belonging to a group who, although not highly placed in socioeconomic terms, could nevertheless claim something valuable for themselves (in the Martha’s Vineyard case, authenticity). Milroy’s book provided statistical backing for such an explanation. Although the inner workings of the social network depend somewhat on the amount of personal contact, the essential thing is that its members share norms. As attention turned to understanding the nature of these norms, two much publicized views had an impact. Fish (1980) had devised the concept of the ‘interpretative community’ to account for the norms of reading whereby people evaluate different readings of the same text as either valid or absurd. An interpretative community is a group sharing such a set of norms; its members may never come into direct physical contact with one another, yet share norms spread by the educational system, books, or the media. Soon after, Anderson (1991, first published in 1983) proposed a new understanding of the ‘nation’ as an ‘imagined community,’ whose members, like that of the interpretative community, will never all meet one another let alone have the sort of regular intercourse that creates a ‘network.’ What binds them together is the shared belief in the membership in the community. Notably with the work of Eckert, sociolinguistic investigation of groups ideologically bound to one another shifted from statistically based examination of social networks to more interpretative examination of ‘communities of practice,’ defined as ‘‘an aggregate of people who come together around mutual 488 Identity and Languageengagement in an endeavor’’ (Eckert and McConnellGinet, 1992: 464). In the course of this endeavor there emerge shared beliefs, norms, and ideologies, including, though not limited to, linguistic and communicative behavior. The advantage of the community-of-practice concept is its openness – any aggregate of people can be held to constitute one, so long as the analyst can point convincingly to behavior that implies shared norms or, better still, elicit expression of the underlying ideologies from members of the community. This line of research is thus continuous with another one that has focused more directly on the normative beliefs or ideologies by which national and other group identities are maintained. Some early work along these lines was published in Wodak (1989) and Joseph and Taylor (1990), and subsequently a great deal more has appeared, e.g., in Verschueren (1999), Blommaert (1999), and Kroskrity (2000). Other features of recent work on language and identity include the view that identity is something constructed rather than essential and performed rather than possessed – features that the term identity itself tends to mask, suggesting as it does something singular, objective, and reified. Each of us performs a repertoire of identities that are constantly shifting, and that we negotiate and renegotiate according to the circumstances. Co-constructing National Identity and Language As mentioned at the outset, it is not simply the case that identity is built upon language. The reverse is also true. One of the first obstacles to be overcome in establishing a national identity is the nonexistence of a national language. The ‘nation–state myth’ – that basic view of the world as consisting naturally of nation–states – is bound up with an assumption that national languages are a primordial reality. Whatever difficulty we might have in determining the borderlines of who ‘the Germans’ are, whether the Germanborn children of Turkish immigrants are German for instance, or whether certain Alsatians are French or German, the German language is going to figure significantly in the equation. Hitler attempted to justify his initial invasions of neighboring countries on the grounds that these German-speaking peoples were inherently part of the German nation; and, as Hutton (1999) has shown, his policies of oppression and ultimately extermination of the Jews were underpinned by the argument that, although their language, Yiddish, was a form of German, their lack of a nation gave them the perverse racial peculiarity of not being able to have a ‘mother tongue.’ They therefore did not belong to the German body politic but were seen as a parasite within it. But whether Bohemian, Alsatian, and Yiddish dialects were part of ‘the German language’ were not facts given in advance, nor even ones that a linguist could establish scientifically. The ‘German language,’ like every national language, is a cultural construct. It dates from the 16th century and is generally credited to Luther, who, in translating the Bible, strove to create a form of German that might unite the many dialect groups across what until the late 19th century was a patchwork of small and large states, linguistically very diverse. (This story is itself a part of the cultural construct, and although not false, it is considerably oversimplified – in order to shape up as a proper ‘hero’ myth, it ignores or marginalizes the work of many other individuals and broader cultural changes.) The prototype of the modern national language was Italian, which may seem surprising, given that Italy did not become a political nation until 1860, with full unification coming in 1870, just a year before that of Germany. Or perhaps not so surprising – the political divisions of the Italian peninsula may have been precisely what motivated the creation of cultural unity through linguistic means. In the Romance-speaking world during the thousand years from the fall of the Roman Empire to the Renaissance, ‘language’ meant Latin, used for all official and written purposes, though what people spoke in nonofficial contexts was a local dialect, historically related to Latin though significantly different from village to village. There was, then, no ‘Italian language.’ That concept and its realization are credited – heroically, and again only semimythically – to Dante, author of the Divina Commedia. Dante’s treatise De vulgari eloquentia (c. 1306), not published until 1529, lays out the process by which he claimed to discover, not invent, the national language of a nation that would take five and a half centuries to emerge politically. To the modern reader this all seems a fiction, a pretense of discovery in what will actually be Dante’s invention of an illustrious vernacular – which will in turn camouflage how much of it is actually based on his native Tuscan. But Dante’s volgare illustre became the template on which other modern European standard languages were modeled. Once the national languages existed, however, their invention was promptly forgotten. The people for whom they represented national unity inevitably came to imagine that the language had always been there, and that such dialectal difference that existed within it was the product of recent fragmentation, when in fact it had preceded the unification by which the national language was forged. In the words of one prominent historian: Identity and Language 489National languages [. . .] are the opposite of what nationalist mythology supposes them to be, namely the primordial foundations of national culture and the matrices of the national mind. They are usually attempts to devise a standardized idiom out of a multiplicity of actually spoken idioms, which are downgraded to dialects [. . .]. (Hobsbawm, 1990: 51) By the early 19th century this ‘nationalist mythology’ would lead to strong Romantic theorizations of national political identity being grounded in a primordial sharing of language. One of the strongest expressions was that of Fichte: The first, original, and truly natural boundaries of states are beyond doubt their internal boundaries. Those who speak the same language are joined to each other by a multitude of invisible bonds by nature herself, long before any human art begins; they understand each other and have the power of continuing to make themselves understood more and more clearly; they belong together and are by nature one and an inseparable whole. (Fichte, 1968 [1808]: 190–191) Fichte was writing to rouse the ‘German nation’ to repel the advance of Napoleon. He was successful in doing so. However, in 1870 the shoe was on the other foot, when the Franco-Prussian War led to the German annexation of Alsace, a Germanspeaking province that had been part of France for more than two centuries and whose population was mainly loyal to France, despite their linguistic difference. This provoked a sharp turn away from the Fichtean view on the part of French linguists such as Renan (1882), who formulated a new view of national identity as based not on any primordially determining characteristic such as language, but on a shared will to be part of the same nation, together with shared memories. The nation, in other words, exists in the minds of the people who make it up. This is the conception that Anderson (1991: 6) would return to in defining the nation as ‘‘an imagined political community.’’ The ‘legacy of memories’ Renan pointed to would dominate future philosophical and academic attempts to analyze national identity. The other element, the collective ‘will’ of the people, would, however, have the deepest political impact, starting with the redrawing of the map of Europe at Versailles in 1919. It has continued to be the assumed basis for the legitimacy of the political nation up to the present time. Billig, a colleague and collaborator of Tajfel, has explored how the ‘‘continual acts of imagination’’ on which the nation depends for its existence are reproduced (Billig, 1995: 70), sometimes through purposeful deployment of national symbols, but mostly through daily habits of which we are dimly aware at best. Examples include the national flag hanging in front of the post office and the national symbols on the coins and banknotes we use each day. Billig introduced the term banal nationalism to cover the ideological habits that enable the established nations of the West to be reproduced. In Billig’s view ‘‘an identity is to be found in the embodied habits of social life’’ (Billig, 1995: 8), including language. Smith (e.g., 1998, Chapter 8) has emphasized how much of the effort of nationalism construction is aimed at reaching back to the past in the interest of ‘ethno-symbolism,’ and this can be seen particularly in the strong investment made by modern cultures in maintaining the ‘standard language,’ by which is meant a form resistant to change, hence harking backward. Hobsbawm placed great stress on the fact that enthusiasm for linguistic nationalism has historically been a phenomenon of the lower middle class. The classes which stood or fell by the official use of the written vernacular were the socially modest but educated middle strata, which included those who acquired lower middle-class status precisely by virtue of occupying non-manual jobs that required schooling. (Hobsbawm, 1998: 117) These are also the people who become the mainstay of nationalism – not just by active flag-waving on symbolic occasions, but daily in the banal ways pointed to by Billig, including their use of ‘proper language’ and their insistence on its norms, for instance in conversation with their own children. Hobsbawm has suggested that, in Victorian times, the pattern was established whereby the lower middle classes (artisans, shopkeepers, and clerks) enacted their national belonging by showing themselves to be ‘‘the most ‘respectable’ sons and daughters of the fatherland’’ (Billig, 1998: 122). In other words, although their real identity was that of a social class, they masked it for themselves and others in a nationalistic guise. The mask was double-sided: in their obsession with ‘speaking properly’ as a mark of respectability, they were contributing to the linguistic construction of their nation. At the start of this article, I wrote that identities are, first and foremost, names, and that the work of identity construction consists of supplying meaning for the name. In the case of national identity, language has traditionally been a key ingredient in this process for at least five reasons: 1. Groups of people who occupy contiguous territory and see themselves as having common interests tend to develop, over long stretches of time, ways of speaking that are distinctive to them, marking them out from groups who either are not geographically adjacent to them or else are perceived 490 Identity and Languageas having different, probably rival, interests. In other words, language does tend to mark out the social features on which national belonging will come to be based – but it is only a tendency, because it also happens very frequently that the same way of speaking is shared by people with very different interests (religious ones, for instance), and that markedly different ways of speaking exist among a group of people who nonetheless see themselves as part of the same nation. 2. The ideology of national unity has favored a view that nations are real because those within them share a deep cultural unity, and this has coexisted with a widespread – indeed, nearly universal – belief that deep cultural unity is the product of a shared language. This is what Fichte meant by the ‘invisible bonds’ by which nature has joined those who speak the same language. Again, as with (1), it cannot be more than a tendency, because it is not the case that people who identify themselves as belonging to the same culture or nation think identically. Yet language is central to the habitus (an ancient term revived by Bourdieu): the fact that we spend our formative years attending long and hard to the task of learning words and their meanings from those around us results in our acquiring tastes, habits, and ways of thinking from them that will endure into adult life. The language does not somehow transmit culture and identity to its speakers – rather, the language is the text through the constant interaction with which older speakers transmit culture and identities (local and personal as well as national, ethnic, and religious) to the young. (In many cases the young will want an identity of their own and will attain it first of all by resisting the imposition of culture upon them by their elders.) 3. In addition to being the text of cultural transmission, the language is the principal medium in which texts of national identity in the more usual sense will be constructed. It is not the only such medium, nor the only powerful one, as Billig’s exposition of ‘banal nationalism’ has shown. But the particular concepts that constitute a national identity correspond to words in the national language, embodied in ‘sacred texts’ of the nation such as a constitution or key works of the national literature, including the national anthem. 4. As universal education is adopted throughout the nation, inculcating standards of ‘correct’ language assumes a central role. Overtly, this is from a perceived duty to maintain the culture. However, as Hobsbawm has shown, such is the force of the language-culture-nation-class nexus that, especially for the upwardly mobile members of the lower middle class, being a ‘proper’ citizen and member of the community is inseparable from using ‘proper’ language. 5. Insofar as nations are not the historical essences they purport to be, but are constructs that inevitably involve a certain amount of arbitrary and even capricious divisions and classifications, when a nation wants to control who can live in it, vote in it, and enjoy state benefits, language can appear to be the most obvious test for deciding whether particular individuals belong to the nation or not. Most nations no longer have laws based on racial classification – which are rarely easy to apply in any case – yet many do require cultural qualifications to be met, which are likely to include language either overtly or indirectly. Each of these factors has reinforced the others in giving the national language the force of a cultural– historical ‘ethno-symbolic’ myth as suggested by Smith (see earlier discussion). Within each, too, there is a contradiction or a caveat that has periodically pendulum-swung to prominence, such that the loss of belief in the national language and all it stands for is always potentially there and is bound to come to the fore at least periodically. The question that is unanswerable for now is whether national languages, together with the nations to which they are attached, represent a historical phase that is now on a course of decline heading for eventual disappearance, to be replaced with ‘glocalization’ – a term that has been coined to denote the combination of globalization with the resurgence of local, subnational sites of belonging – or whether they are an invention that has proven too useful for human social organization to be given up. See also: Communities of Practice; Ethnicity; Ethnolinguistic Vitality; Gender; Interactional Sociolinguistics; Language Attitudes; Multiculturalism and Language; Social Class and Status; Sociolinguistic Crossing. Reflections of Identity in Language Use The main focus of studies of language and identity in sociocultural anthropology (LISA) is on the interplay between global cultural categories (‘census identities’ such as gender, race, class, or others that may be culturally salient, such as kinship identities), and how these identities are constructed, performed, or challenged locally in speech events through linguistic forms. One of the distinguishing characteristics of the LISA approach is that rather than taking census and other identity categories as pre-given, it investigates how these categories are made relevant by and for language users. It is important to understand that LISA research does not lose sight of the larger picture; it attempts to discover how speakers’ ideologies about census identities shape, and are shaped by, interaction. Identity as viewed in this tradition is therefore emergent but constrained: identities are created and re-created when speakers are actually talking to each other, but the way these identities emerge is contingent on the speakers’ sociocultural discourses and ideologies. Identities are named ‘relationships’ in cultural ideologies and interactions; identity categories such as census identities are a shorthand for identifying – and thinking about – the many ways that a person relates to others in the community. Identity in Sociocultural Anthropology and Language 495For example, by referring to gender or drawing on a gender identity in interaction, a person calls up a bundle of culturally-shared relationships to other people in their culture. A man may call up solidarity with another man through cultural ideologies of masculinity that identify men as a class of people who share certain social traits. In this sense, identities are ‘imaginary’ (cf. Anderson, 1991 on imagined communities and nationalism) or ‘figured worlds’ (Holland et al., 1998). Because identities are relational, a person has no single fixed identity, only identities constructed and contextualized in interaction (and to the extent that an identity is psychologically real, it is based on the self’s conception of its place in psychologically idealized models of interaction). LISA studies investigate the tension between ideologies of identity categories and their emergence in interaction through language and other communicative modes. LISA does not study the motivation of speakers to join identity categories, but it does show how such categories are reproduced in ideologies and cultural discourses through talk. We might suppose that the most straightforward way of claiming an identity is to speak the language of that identity: one is Tongan when one speaks Tongan, Uzbekh when one speaks Uzbekh, etc. But this view is destabilized when speakers are multilingual (which seems to be the larger portion of the world’s population). Does a Black-AmericanHeterosexual-Woman become Dominican when she uses Dominican Spanish, or an Australian when she uses Australian English? By investigating just these kinds of situations (Bailey, 2001) LISA researchers have shown that the ‘languageequals-identity’ model is too simple to represent actual interaction. This research challenges the notion that languages are single, invariant entities, and has shown that a language is full of many different voices: registers, styles, varieties, and even other languages. At the same time, LISA analyses incorporate speakers’ own ideologies of language. For although we might objectively observe that language use is never pure, the subjective ideologies of language users is often quite different. Language users’ understandings of ‘‘how language is supposed to be’’ affect the way they use it: they bias performance in certain directions, as sociolinguistic variationists have been able to measure. The terms most used for such a mixed view of language in LISA are heteroglossia and polyphony, from the work of Russian literary theorist Mikhail Bakhtin. Bakhtin (1981) argued that words (and by extension other linguistic forms) carry the ‘voices’ of previous uses, into current contexts of use. That is, words have meanings beyond pure denotation and carry the residue of earlier situations of use. Heteroglossia, as Bakhtin uses the term, refers to the tension between the pure official language and more vernacular forms of language existing in a language community and used in a single text. Polyphony refers the use of different languages or varieties that meet on an equal footing (cf. Morris, 1994: 248–249). Bakhtin’s focus was the novel, but LISA researchers have shown that heteroglossia and polyphony obtain in everyday conversation, and these properties of language help people to create identities in interaction. This view destabilizes both the simplistic view of ‘a language,’ and the assumption that by using a particular linguistic feature (e.g., word, pronunciation, or syntactic construction), one can perform an identity connected with a culturally-relevant population that is thought to use that linguistic feature. Meanings that rely on the current or previous context, or both, are known as indexical meanings. All languages have such indexical meanings as part of their grammar. In English, the meanings of words such as I and you are associated with actual people only in the moment of speaking. Thai pronouns indicate not only speaker and addressee, but also their relative ages and genders (Simpson, 1997). Most indexical meanings – whether in a grammar or not – relate to the aspect of context that provides information about the speaker’s identity. A linguistic feature indexes an identity because someone of that same culturally-relevant identity category has used it before, or has been represented as using it before (e.g., in media). Speakers rarely assert directly that they are, for example, a woman, or white, or Bosnian (although they do at times, and these moments are all the more interesting for their rarity). Rather, speakers more often rely on the processes of indexicality to create their identities. Theorizing indexical meaning as it relates to identity has been the focus of much of the work of Michael Silverstein (1976, 1996) and Elinor Ochs (1992, 1993), who have made a number of important distinctions among indexical meanings that are crucial for understanding how identity and language are theorized. The first distinction is between direct and indirect indexicality (Ochs, 1992). Direct indexicality is a meaning relationship that holds directly between language and the stance, act, activity, or identity indexed. For example, in English, using certain syntactic forms such as imperatives will generally directly index a relationship of power. The imperative is associated with the context of a speaker who has the power to force the addressee to perform the action stated in the imperative. Indirect indexicality arises when the social relationship (in this case, power) is further indexed to an identity category, such as Identity in Sociocultural Anthropology and Language masculinity. Thus we might notice that men use more imperatives than women, or there might exist (as in the United States) a cultural discourse such that men are supposed to be more authoritative than women. When this supposed connection is discovered, we might be tempted to say that imperatives index masculinity; however, it would be more accurate to say that imperatives indirectly index masculinity because they directly index power, and power is in turn associated with masculinity. An illustration of this process can be seen in the American English address term dude (Kiesling, 2004). Dude is used most frequently by young white men, and indexes a stance of casual solidarity: a friendly, but crucially not intimate, relationship with the addressee. This stance of casual solidarity is a stance habitually taken more by young white American men than other identity groups. Dude thus indirectly indexes young, white, masculinity as well. Such descriptions of indexicality are abstract, however, and do not take into account the actual context of speaking, such as the speech event and the identities of the speakers determined through other perceptual modes, such as vision. One of the most difficult aspects of indexicality is that indexes not only call up the previous contexts in which a linguistic feature has been used, but what previous context they call up can be influenced by the current context. Indexical meanings thus interact with the context of speaking, such that the (assumed or ‘pre-existing’) identity of the speaker can affect the indexicality of a form even as that form is seen as performing an identity. For example, in the United States, often when a women uses an imperative, a particular kind of authority – motherhood – is indexed: when a woman tells a friend (particularly a man) to ‘‘put on your coat,’’ the response is often sarcastically, ‘‘Yes, mother.’’ In this example, the use of an imperative directly indexes a stance of authority. In the dominant American cultural discourses of femininity, authority is indexed only in the context of motherhood, so this feminine role is indirectly indexed and available for comment by the addressee of the imperative. Such a response is not made when the speaker is known to be a man (unless the addressee wishes to refigure the speaker as a woman for some reason, in which case the layers of indexicality begin to become thick and tangled). One might argue that the content of the imperative is what indexes motherhood: mothers order their children to do things for the children’s own good, especially putting on coats. This argument only points out just how specific indexicality can be, however, as a single phrase can call up the stereotypical class of people known in the United States as ‘mothers.’ In addition, we might observe that imperatives used by women are more often than not used with such ‘motherly’ content. In any case, note how even such a brief, unremarkable use of language leads to an intricate web of identity indexes. Such are the minute, commonplace, but complex indexical meanings that rely on and recreate identity discourses and ideologies. These indexicalities are thus both sensitive to and indexical of social context and the other possible relevant identities being performed in a particular instance of talk-in-interaction. Bonnie McElhinny’s (1994) study of police officers in Pittsburgh, Pennsylvania, is particularly relevant in this light. She showed that when the identity of being a police officer is salient in an interaction, such a speaker tends to use less affect in her or his speech. This lack of affect indexes a distant, emotionless stance, a stance that is not indexical of femininity in Pittsburgh. However, the women officers told McElhinny that they did not feel less feminine in these situations, because gender is less relevant as an interactional identity; in such interactions they think of themselves primarily as police officers. As women have joined men on the police force, the affectless and emotionless stance has thus come to index not so much masculinity, but ‘police officer.’ Further complicating the connections between linguistic form and identity is the fact that indexes are always potentially creative as well as presupposing (Silverstein, 1976). That is, an index can be used to change what is in play in the context (creative) rather than simply respond to the already mutually understood context (presupposing). The classic example of such context-changing uses is that observed by Blom and Gumperz (1972) in Hemnesberget, Norway. In Hemnesberget at that time there was a diglossic situation in which the linguistic variety of Norwegian known as Bokma˚l was the official language of government and commerce, while Ranama˚l was the local variety (dialect) more typically used for intimate settings. Blom and Gumperz described an instance in which an inhabitant goes to the community office for official business and interacts with another local inhabitant with whom he is friends. The official business is conducted in Bokma˚l (indexing in a presupposing manner the situation), but when the official business is finished, the customer switches to Ranama˚l to indicate he wishes to have a less formal conversation about local matters. This switch from Bokma˚l to Ranama˚l is a creative index, because it changes the context from formal to informal, rather than responding to an external cue such as a change of setting or the arrival of a new participant. While this distinction between creative and presupposing indexicality is important and useful to begin with, LISA researchers have more recently emphasized the fact that indexes always have both a creative and a presupposing aspect. Sidnell (2003), for example, described how language is used to create a male space in a Guyanese rum shop. While this setting is understood by the Guyanese to be a male space, it is nonetheless often populated by women for various reasons. Sidnell analyzed a stretch of talk-in-interaction in which one of the men drinking in the rum shop designs ‘‘his talk . . . to preserve the observable gender exclusivity of the activity’’ (2003: 338). That is, the way his talk is designed makes the space a male one even when he is addressing a woman in it, and thereby helps to recreate the ‘male-only’ context, in effect (creatively) peripheralizing women who are present. A final complication comes from the ‘level’ of knowledge that speakers use in interpreting indexical meaning. While most indexical meanings are not overtly discussed by speakers, some linguistic features nevertheless become the focus of social discussion and overt knowledge. For example, Americans have become aware of, and can articulate, the masculine indexicality of dude (they are aware of its use on this societal level), but have much more trouble explaining what it means in terms of local stances. Morford (1997) analyzed the use of second-person pronouns in French in this light. She showed that on one level of indexicality, the use of tu and vous in French indexes a relationship between two speakers based on a reciprocal vs. asymmetric use of the forms. However, Morford found that French speakers have a knowledge about (for example) what kinds of families have asymmetric use between parents and children, and that some speakers make conscious choices about whether to use tu or vous with other speakers. Most importantly, these choices have to do with the interplay between this metapragmatic awareness and the kinds of identities that speakers wish to create. Morford cites one particularly vivid example of a teacher who explains that he constructs status as a teacher in part through his asymmetric use of second person address with the support staff of the school. Performance of Identity in Mediated Social Practice While the distinctions made above are important, individual indexical forms such as ‘pronoun’ or ‘accent’ are idealizations of meanings that are tangled in a messy, changing, and locally contestable web of meaning. That is, there are complex meaning relationships not just between a linguistic form and a specific social identity, but that there are also connections of meaning among different kinds of identities, and different social entities and categories such as institutions and census groups. LISA researchers do not search for psychological representations of such links, but rather they investigate the manner in which such connections are created and used in interaction. In fact, these connections need not be present in any individual’s psyche before a speech event; it is by participating in the event itself that speakers bring indexicalities and identities about, because they are needed for speakers to ‘make sense’ of the talk. In this way, LISA is similar to critical discourse analysis (see Critical Discourse Analysis) in arguing that the interpretations that speakers make of the ongoing talk based on ‘background knowledge’ or conversational principles (see Pragmatics: Overview) are not used in a value-free way but are a resource for doing identity in interaction. One of the most significant bodies of work along these lines is Jane Hill’s (1993, 1995, 1998) work on what she calls ‘‘mock Spanish.’’ In this work, the process of indirect indexicality and the ideological structuring of background knowledge come together to reproduce racist discourses that are covert even for those using them. Hill took as her starting point the nonce borrowing of Spanish (or forms that resemble Spanish) by Anglos, such as Hasta la vista, baby, popularized by actor and politician Arnold Schwarzenegger in the film Terminator. Hill showed that these forms of Spanish are used strategically by speakers to create an ironic speaker stance with respect to stereotypically Hispanic (Latino) identity, stances that are often demeaning to this identity category (e.g., lazy, disorderly, given to rudeness), in effect distancing the self from such identities by gently parodying their principal index, Spanish language forms. A similar process is at work in a conversation I analyzed between two white American fraternity men (Kiesling, 2001). The two men in that conversation discuss the roster of the fraternity’s intramural basketball team. Embedded in this discussion is a short exchange in which each boasts about his basketball ability. At strategic points in both parts of this speech event, the interactants use linguistic features of African American Vernacular English (AAVE) (see English, African American Vernacular) to index their boasting stance, as well as a stance of basketball expertise. The crucial point is that the men are not trying to ‘be black,’ but they are drawing on an association of ‘blackness’ with basketball and boasting. In other words, when performing ‘basketball player’ it helps to perform ‘black’ as well. A similar point was made by Bucholtz (1999), who showed that AAVE features are used in a fight narrative told by a white male high school student in California to represent a tough stance. Other articles appearing in the same journal issue as Bucholtz’s made similar points and represent an important body of work for LISA understandings of language and identity. Issues of race and racism in language, and gender and sexism in language have therefore been central in LISA research. Over the past decade, Susan Gal and Judith Irvine (Gal and Irvine, 1995; Irvine, 2001) have developed an influential set of concepts for understanding how speakers’ ideologies of language are involved in the social production of indexicals of identity categories. Their view recognized three semiotic effects of ideology as a factor in indexicality: iconicization, recursivity, and erasure. Iconicization is the ideological endowing of a linguistic feature with and characteristic that is taken to be like some essence of a recognizable identity or type of person. In the example from Hill above, mock Spanish is understood to be iconic of some inherent characteristic, laziness or disorderliness, of Spanish speakers. Recursivity is the manner in which ideologies are represented in parallel recurring form with respect to multiple levels of social organization. In the example above, stereotyped differences between black and white men in the United States (the stereotyped black characteristics of athletic prowess and fighting ability) become the basis for negotiating differences between fraternity brothers, both of whom are white. In effect this use recursively reproduces the black-white identity distinctions in interactions with only white male speakers. Finally, erasure is the effective simplification of both formal and sociological complexities of experience, and can render all but a single term of a schematic dichotomization invisible to speakers. In all of the above examples, the Anglo white language style is invisible as a specifically alternative register, while the Spanish and AAVE are marked and highly salient; the white Anglo vernacular register is erased and re-enforced as normative by such uses, rendering its social peripheralizing power less visible. Studying Identity through Language and Contextualized Language Variation Up to this point, we have generally focused on language and identity as performed when speakers use languages or varieties they recognize as different. But the processes of indexicality, and the tendencies toward iconization, recursivity, and erasure are at work even when a single variety is being used, not only when varieties (or languages) are mixed. As sociolinguistic variationists continually remind us, language is inherently variable in both grammar and use (Guy, 1996); such variability is a valuable resource for speakers in identity work. Several researchers have been incorporating insights from LISA studies into the search for how this variability is used by speakers to create social meaning. Penelope Eckert (2000, 2002) is the most prominent of these researchers and has done the most to move this new perspective of meaning in variation forward. She advocates understanding variation as a resource for creating personal styles and ‘‘stylistic practice as a process of bricolage, in which ways of being are transformed through the strategic reuse of meaningful resources.’’ (2002: 5) In her ethnographic analysis of sociolinguistic variation in a Detroit-area high school, she showed how vowel variants vary alternately with other aspects of social practice and style, such as ‘cruising’ (driving around a particular route with friends without a destination) and wearing wide pants cuffs. She showed that while the variables she analyzed have general meanings throughout the school (e.g., whether they are associated with the urban rather than suburban areas of Detroit), they are used in specific clustered ways to create personal styles by individuals. Students of Eckert’s are moving this view of variation and identity forward considerably. Sarah Benor (2001) investigated how the language of Orthodox Jews differs depending on their background and orientation to integrate into the Orthodox community. Her work is significant in two respects. First, it showed that identity is not just about the outward practices and styles of speakers, but that the internal orientations and ideologies of speakers can influence their use of language, and that the relationship between these orientations and language use are not necessarily linear. Second, she showed that adults, as they become integrated socially and ideologically into a new community with a different way of speaking (even if that new way is based on the rate of use of a linguistic variant), can change their speech to the new way of speaking. This finding suggests that speakers have much more control over how they speak than is often assumed. Benor’s work showed further the Eckertian view that most variables provide general meaning resources that acquire specific meanings in local contexts, since each analyzes essentially the same linguistic form, but show how its indexicality changes depending on the community and the individual speaker’s style. I have been working with a more atomistic view than Eckert’s, but one that is consistent with her perspective. This view focuses on the performance Identity in Sociocultural Anthropology and Language 499of stances in interaction rather than personal styles. We can understand Eckert’s personal styles to be repertoires of stances; the variables used habitually by speakers to create these styles are primarily concerned with creating stances – relationships of the moment. In an analysis of the -ing variable used in a college all-male fraternity in the United States (Kiesling, 1998), I showed that the men who speak with a high rate of the ‘standard’ velar variant [-N] in meetings are more likely than those whose rate is lower to be taking a stance at that point highlighting their structural power in the fraternity (through age or office-holding). These men also use grammatical forms that help them take an epistemic stance of certainty. Men who use the nonstandard coronal articulation [-n] in meetings tend to take a stance that highlights their connections to the aggregate of men (solidarity) or their ability to work hard for the fraternity. They also tend to use more profanity and other nonstandard features, such as double negatives and ain’t. The point is that these differing rates of the variable shows the fraternity men to be taking particular stances at various points in specific interactions; it is not just that each man is from a certain class or ethnic background and uniformly reflects that in speech production. Moreover, one can explain why – in the context of the situation, the man’s personality, and cultural discourses and ideologies of masculinity, for example – he would take such a stance. This focus on stance moves us further toward an explanation of why speakers use the forms they do, and how these forms mean in interaction. Such an interactionally local understanding of the indexicality of linguistic forms does not entail that wider cultural discourses, in this case norms of gender, are not also being indexed. In fact, we see here how the meanings of variation also exhibit the property of recursivity: the kinds of stances that the men take, sorting themselves out relationally, are all stances that index masculinity in their culture as well. In other words, they are able to ‘be men’ of one or another sort – not more or less male – while they are using either variant, because the dominant cultural discourses of masculinity include approved types that are both structurally or intellectually powerful (the standard users) and that are friendly or physically powerful (the nonstandard users). These cultural discourses are being recursively played out in the momentary stances the men take in the meetings. In terms of method, in all of these directions of research, we find a curious marriage between the quantitative, objective, paradigm of variationist linguistics and the qualitative, subjectively-focused view of culture and interaction from sociocultural anthropology. One of the crucial ways that anthropology has informed the study of language and identity is in fact a focus on the speaker’s interpretive frameworks, to which we turn now. Ethnographic Investigation of People’s Interpretive Frameworks One of the strengths of LISA studies is the in-depth investigation of the speakers’ interpretive frameworks – how speakers themselves respond to language as reflective or constitutive of identity. This metapragmatic awareness, and its relationship to other indexicalities, is an important area of research in LISA. Silverstein (2001) proposed a classification that specifies under what kinds of conditions, and how, a linguistic form is likely to be accessible to metapragmatic awareness. While there is not space to explain these principles here, Silverstein’s work suggested that certain types of linguistic forms are more accessible resources for speakers to use to connect with identities. Holland and Skinner (1987) took another approach, and argued that identities be understood as organized in the cultural models that speakers hold for the world. Such models are more specific than ideologies; they are in essence normative narratives of life trajectories or domains to which actual people are compared, especially in language. Holland and Skinner used the cultural model concept to understand how women at a southern U.S. college semantically organize the language they use to describe men. Such interpretations of the relationship between language and identity are beginning to be articulated with traditional dialectological questions about how language relates to place. Johnstone (2004a, 2004b) suggested that a speaker’s emplacement – that is, being culturally connected to place – is not simply another census identity, but one that is discursively negotiated like others. Speakers who live in the same geographical space take different orientations to it as a culturally meaningful place, and local dialect forms may function in various ways in everyday talk, in self-conscious performances, and in metalinguistic talk about the dialect, to index these orientations. The use of dialect forms to index orientation to place may help determine which forms are preserved and which are lost over time. This work indicates that even such taken-for-granted categories as place are ideologically organized by speakers, and that this ideological organization can affect the way speakers use language, and its meaning in terms of identities. Language and Identity over Time Future research in LISA is likely to focus on further understanding the ways in which indexicality works to create identities in interaction, from a number of perspectives. For one, the question of awareness is still a very open one, in which even what counts as awareness is not entirely clear. One can imagine a typology of awareness that can then be matched with Silverstein’s accessibility hierarchy. Awareness may also be promoted by the media as well as faceto-face communication, and it is not clear the ways in which these modes of mediation differ for the creation of identity categories and indexicalities. Finally, the ways in which the use of these indexicalities construct and re-create identity categories, and especially relations of dominance among them, will continue to be a puzzle for some time.