Ideologies and practices of identity

Asif Agha

University of Pennsylvania


[Comments on Bucholz, ‘Word up: Social meanings of slang

in California youth culture]


            Mary Bucholz’s elegant paper raises several interesting issues regarding the formulation and deployment of signs of social identity. An unusual feature of the paper is the comparative treatment of two very different schemes of identity reckoning in a single ethnographic context: the sign-values attached to place (i.e. the locale where one ‘hangs out’ in the schoolyard) and to the uses of slang (i.e. register variation in speech). Since both cases involve an ideological dimension, the comparative frame of the paper raises some general issues:

·      In what sense do these ideologies formulate signs of social identity?

·      How are these ideologies identified in actual data?

·      What is the evidence that the ideologies are socially shared?

·      What is the semiotic basis for distinguishing ‘ideology’ from ‘practice’?

I’d like to comment on each of these issues in turn, starting with the first.


Social identity

In what sense do these ideologies formulate signs of social identity? In each case the ideology treats particular forms of behavior—habits of hanging out, habits of utterance—as conveying information about the actor who performs the behavior in the instance. Both ideologies thus evaluate forms of behavior as indexicals of actor type. In the case of slang the criterial signs are linguistic forms; we are thus inclined to call this case a ‘language ideology’. In the other case the criterial signs are facts of habitual presence (‘hanging out’) in certain schoolyard spaces; in this case we are dealing with —what we might as well call—a ‘geosocial ideology’ of personhood.

In both cases the ideology links performable behaviors to social personae. Some of the personae discussed in the paper involve matters of group-relative status (cf. “social position,” p. 26); these are analyzable in terms of demographic variables, e.g., age-set, ethnicity, gender, regional affiliation. Others involve matters of interactional stance, e.g., speaker’s coolness, or emergent alignments between speaker/addressee; these have a more indirect relationship to social-demographic categories of personhood.

Now, whether the “identity” effect is construed as a matter of group-membership or of interactional stance the effect is always marked by patterns of actor-focal indexicals in interaction. Ideologies of identity do often essentialize some among the behaviors performed by social actors as indexicals of group membership; but this is not necessary. The ideology of ‘coolness’, for example, centers on matters of individual essence, not group-affiliation.

Yet even when a behavioral sign is ideologically associated with fixed social categories of personhood, the persona indexed by the sign is only ‘readable’ under conditions of co-occurrence with other signs; any such effect is therefore limitlessly defeasible or modifiable by co-textual accompaniment by other signs (see pp. 8-9 below). Signs of identity appear much more bounded and discrete in the data of decontextualized reportage than they do under conditions of performance.

We are now talking about two entirely different perspectives on signs of identity: an ideological perspective and a performative perspective. The discussion of slang in the paper allows us to see the difference between the two rather sharply. But before I turn to slang, let me take up the next two questions about ideology mentioned above—the question of identifiability and of sharedness (see above)—for the geosocial ideology first.


Schoolyard space

            What is the empirical evidence by which we identify the geosocial ideology? The evidence consists largely of maps drawn by students. The maps constitute a metapragmatic discourse linking facts of habitual presence in specific schoolyard spaces to facts of group affiliation. The link is formulated very explicitly by writing ethnonyms (Asian Gastas, AfAm, Asst White People, etc.) next to place-descriptors (steps, hill, blocks) within the maps themselves. So the metapragmatic discourse formulates geosocial equivalences by linking ethnonyms to toponyms.

 What is the evidence that the ideology thus identified is socially shared? The main evidence is that maps drawn by two different social categories of respondents—John Doe (African American) and Mark (EuroAmerican)—formulate comparable geosocial equivalences (viz., steps=Euro Americans, blocks=Asian Americans, hill=African Americans). The ideologies can be shown to be shared insofar as the content of these metapragmatic discourses—especially maps (2a) and (3)—have similarities, one with another. The residue—the differences left over— are emblematic of “different subject positions” (p. 8) within the ideological order.

In other words, just as the evidence for the existence of the ideology is the data of metapragmatic discourse, the evidence for the sharedness of the ideology is the degree to which the metapragmatic discourses produced by different informants have commonalties of content. Thus the analysis of several aspects of the ideology (its existence, its locatability, its degree of sharedness, its variation by subject position) depends upon the analysis of metapragmatic discourses (their production, their producers, their content, replicability of content across a population).

Let us now consider the case of slang.



First, some initial remarks about slang as a register. I have argued elsewhere (Agha 1996, 1999) that a lexical register is comprised of much more than a linguistic repertoire: that the traditional view which seeks to reduce a register to a linguistic repertoire fails to explain how the repertoire is distinguishable from the rest of the language in the first place; that the identifiability of a lexical register requires appeal to native metapragmatic discourse as a criterion; and that once we consider such data we see that a register is a discursive formation involving not only lexical repertoires but metapragmatic stereotypes of personhood, activity, and social relations which, once linked ideologically to lexemes, become performable through the use of such forms.

In the case of slang the importance of metapragmatic criteria is particularly plain— indeed, inescapable—since the repertories of slang are among the most rapidly changing of any linguistic register’s (see, e.g., Eble 1996). In fact, in the case of age-graded slangs both the register’s repertoires and its users change at an extremely rapid pace: the repertoires change through lexical innovation, the users simply by growing out of the age-set. Hence the continuity over time of slang as a cultural formation depends much more on the persistence of ideologies of language than on a fixed repertoire or a fixed group of users. 

Such ideologies—which typify slang as a sub-standard variety of the language, associated with informal, casual interactions, and with particular types of speaker—are themselves replicated across a population through normative metadiscursive institutions such as prescriptive schooling, lexicography and the like. Given such replication by institutions the ideology can acquire a greater historical continuity and demographic spread than do the lexical repertoires that count as slang forms for a given generation. The ironic result is that different generations of adult speakers may readily share a common attitude about youth slang (and its users) without ever having known (or used) the same slang words  (let alone the words currently employed in the slang).

 Bucholz’s paper shows that even within youth communities variations in slang repertoire are linked to social classifications of speaker. Once speech is personified in this way the deployment of particular slang repertoires makes possible both the performance of normative identities and the tropic manipulation of speaker persona through various types of displaced usages, such as the performance of role fragments, voicing phenomena, and tropes of irony and sarcasm. The account employs three main kinds of metadiscursive data.


(1) Explicitly metadiscursive data

The first type involves speech-speaker correlations evidenced by overtly metadiscursive conversations, i.e., conversations which take slang lexemes, glosses, usage, etc., as overt topics of discussion. Interestingly, although slang is an overt topic in these cases, much of the actual discussion focuses thematically not on lexical forms but on the social personae associated with these forms.

In some of the cases a particular persona is indexically inhabited by a speaker through a particular usage, then commented upon by interlocutors.  In (1), for example, where Priscilla has been talking about the use of marijuana, her use of ‘stoned’ marks a shift from standard English to slang. Her interlocutors comment ironically on her usage and present contemporary slang alternatives to ‘stoned’; the interaction has turned now to the negotiation of speaker persona and group-relative (including generational) identity. This case involves naturally occurring metadiscursive activity. In the cases discussed on pages 12-13, in contrast, the metadiscursive data is elicited by the analyst; here, Bucholz is able to control for demographic category of respondent, a technique that reveals asymmetries of competence over slang repertoires across a population of speakers, such variation itself comprising a system of second-order indexicals of speaker identity (Silverstein 1996).

Yet though the two cases differ in whether the data is naturally occurring or elicited, they are similar in that the association between speech difference and speaker difference is established analytically by appeal to explicit metadiscourses about slang.


(2) Implicit correlations

The second main type of data involves implicit correlations of speech and speaker type. In these cases the skewing of usage within naturally occurring text corpora itself implies something about the type of speaker who uses these forms. Thus the fact that ‘hella’/’hecka’ is used by a range of speaker types [examples (2)-(3)] constitutes the main evidence that the form is not linked to a specific ethnic/gender group, but is a Bay Area “regional marker” (p. 16). In contrast, the skewing of the uses of ‘patna’ (p. 17) suggests that the term is strongly associated with African American speakers.

Here, slang is not an overt topic of discussion. The utterances in question are only implicitly metapragmatic in import: they perform rather than describe links between speech variants and speaker types. The patterns of skewing within text corpora suggest that these correlations may be regular. Yet by themselves such data are merely suggestive. They require some sort of emic confirmation. A partial confirmation is presented by the data of tropes.

(3) Interactional tropes

The third type of data involves interactional tropes of various kinds. The construal of these tropes (as described by Bucholz) presupposes the stereotypic values linked to lexemes such as ‘patna’. Yet the construal of these usages as tropes is based on textualized effects, i.e., on the co-occurrence of criterial lexemes with other signs whose effects are not wholly congruent with the personae linked to lexical forms. Thus when Jay uses the term ‘patna’ in (7c) the co(n)textually readable fact that he is not African American calls attention to the possibility that he may be likening himself to an associated persona such as that of a hip-hopper.

Such a construal is a text-level effect since it depends on accompanying signs co-textually conveying information about Jay’s ethnic identity. In this case several facts about both interlocutors are known to us: that both are EuroAmericans, that Charlie is a ‘stoner’ and Jay a ‘hip-hopper’; we also know that both have just agreed (7c, lines 1-7) that the term ‘blood’ sometimes slips out in their speech, even though, as Bucholz observes, the term may be “off-limit to white speakers” (p. 21). Under these textual conditions the fact that Charlie goes on to offer racially neutral slang terms for ‘person’ (i.e., ‘man’, ‘dude’), while Jay offers ‘patna’, may suggest that each is staking out a different “identity claim” for himself. Hence the semiotic basis of this hypothesis is a pattern of text played out across several interactional turns; none of the individual turn contributions in (7c) suffice to make clear any identity claims as such.

At the same time the fact that Jay switches to creaky voice when uttering ‘patna’ (7c, lines 19-20) brings into play other indexical partials as well, suggesting for example that he is voicing the utterance from a position somewhat distinct from his own. Hence Jay’s usage is somewhat layered in terms of competing performed effects. But this inference is, again, motivated not by the slang lexeme alone but by its place in a pattern of discursive text: it is the suprasegmental contrast that implicitly frames the use of ‘patna’ as a potential instance of voicing (in the Bakhtinian, not the phonetic sense!).

Similarly, Norman’s use of the term ‘patna’ for Al in (8) appears to perform an ironic interactional stance; but the stance is readable only by appeal to co-occurring semiotic material (cf. Bucholz’s discussion of the marked use of ‘got’ and ‘be’, p. 22). Here, the text pattern calls into question the denotational appropriateness of ‘patna’ for the person in question, namely Al, who is not African American but aspires to hip-hop culture.

Note that the construal of these tropes (at least as Bucholz presents them) depends on the stereotypic association of ‘patna’ with African American speakers and their activities. In this sense the interactional tropes in (7c) and (8) confirm the hypothesis based on the data of implicit correlations in (4), even though the effects actually performed in the two cases are not the same.


The ideology-practice continuum

In the case of slang, all three types of data—the explicitly metadiscursive discussions, the implicit correlations, the interactional tropes—contribute to our understanding of both ideology and practice, though in different ways, and to different degrees.

The explicit discussions are much more transparently ideological since they overtly articulate links between speech and speaker persona (though from different subject positions). In contrast, the implicit correlations evidenced in text corpora call into play (and the interactional tropes appear to manipulate) ideological stances which they do not overtly describe. These are cases of interactional practices which play upon values articulated within the ideological order.

The terms ‘ideology’ and ‘practice’ are sometimes viewed as competing theoretical constructs in the linguistic anthropology literature today, even as the boundary between them grows increasingly blurred. It is unclear that any hard and fast  line can now be drawn. Here however is a proposal that may account for some of the more common uses of these terms.

At its most distinctive the term ‘ideology’ is used for the explicit content of metapragmatic discourses, and the term ‘(meaningful) practice’ for the implicit aspects of (e.g., presuppositions of) patterns of semiotic activity. (When we speak of tacit ideologies, on the other hand, we are often interested in the distorting presuppositions of particular practices; here however the boundary is less clear). Similarly, claims about the sociocentric aspects of these constructs rest on the replicability of semiotic data: Claims for the ‘sharedness of ideology’ are often based on the comparability of content of metapragmatic discourses produced by different members of a  population (cf. pp. 3-4 above); and the claim that some activity is a ‘social practice’ often rests on whether the activity at issue is genrefied, habitual, routinized, etc., over a population of individuals. Since many types of data can be evaluated from either perspective, claims about ‘ideology’ and ‘practice’ often involve complementary perspectives on the significance of social behavior.

In any event, this is a set of issues that Bucholz’s paper opens up in a very interesting and provocative way. I hope these cursory remarks will open up the possibility of a fuller discussion of the various issues involved.




Agha, Asif. 1996. Stereotypes and registers of honorific language. Language in Society, vol. 27, no. 2, pp. 151-194.

Agha, Asif. 1999. Register. Journal of Linguistic Anthropology, vol. 9, nos. 1-2, Special Issue: Language matters in anthropology: A lexicon for the millennium. Reprinted in Key Terms in Language and Culture, A. Duranti, ed., 2000, Oxford: Blackwell.

Eble, Connie. 1996. Slang and sociability: In-group language among college students. Chapel Hill: University of North Carolina Press.

Silverstein, Michael. 1996. Indexical order and the dialectics of sociolinguistic life. In R. Ide et al., eds., SALSA 3.266-95