Registres et dialectes selon John Gumperz
John J. Gumperz (1968). The Speech Community.
International Encyclopedia of the Social Sciences (pp. 381-6).
New York: Macmillan.
Repris dans Alessandro Duranti (Edited by),
Linguistic Anthropology. A Reader,
Oxford: Blackwell, 2001, pp. 43-52.
Ce texte est antérieur à Language in the Inner City de William Labov (1972), sur le vernacular (le sociolecte) des ghettos noirs. Le mot sociolecte ou «dialecte social» ne s'est pas encore introduit dans la langue de travail des sociolinguistes, non plus que le mot registre inventé dans les années 1950 pour désigner une variété linguistique appropriée à une situation d'emploi particulière. Un registre est une variété linguistique (un sous-ensemble d'une langue) caractérisée par un certain nombre d'écarts (par rapport à la langue standard) ou variations spécifiques — qu'elles soient lexicales, syntaxiques, phonologiques, morphologiques ou pragmatiques — et réservé à des situations circonscrites. Mais c'est bien l'étagement des différents registres de la langue ou «niveaux de langue» que désigne l'expression superposed variation et chez Gumperz un registre est approprié à une situation sociale particulière.
1 / La communauté de parole comme champ d'action
«the speech community as a field of action»
Most groups of any permanence, be they small bands bounded by face-to-face contact, modern nations divisible into smaller subregions, or even occupational associations or neighborhood gangs, may be treated as speech communities, provided they show linguistic peculiarities that warrant special study. The verbal behavior of such groups always constitutes a system. It must be based on finite sets of grammatical rules that underlie the production of well-formed sentences, or else messages will not be intelligible. The description of such rules is a precondition for the study of all types of linguistic phenomena. But it is only the starting point in the sociolinguistic analysis of language behavior…
Just as intelligibility presupposes underlying grammatical rules, the communication of social information presupposes the existence of regular relationships between language usage and social structure. Before we can judge a speaker's social intent, we must know something about the norms defining the appropriateness of linguistically acceptable alternates for particular types of speakers; these norms may vary among subgroups and among social settings. Wherever the relationships between language choice and rules of social appropriateness can be formalized, they allow us to group relevant linguistic forms into distinct dialects, styles and occupational or other special parlances. The sociolinguistic study of speech communities deals with the linguistic similarities and differences among these speech varieties.
In linguistic homogeneous societies the verbal markers of social distinctions tend to be confined to structurally marginal features of phonology, syntax, and lexicon. Elsewhere they may include both standard literary languages, and grammatically divergent local dialects. In many multilingual societies the choice of one language over another has the same signification as the selection among lexical alternates in linguistically homogeneous societies. In such cases, two or more grammars may be required to cover the entire scope of linguistically acceptable expressions that serve to convey social meanings.
Regardless of the linguistic differences among them, the speech varieties employed within a speech community form a system because they are related to a shared set of social norms. Hence, they can be classified according to their usage, their origins, and the relationship between speech and social action that they reflect. They become indices of social patterns of interaction in the speech community.
2 / Deux dimensions de la variation linguistique, les dialectes et les registres
«two dimensions: the dialectal and the superposed»
Although the bulk of research on speech communities that was conducted prior to 1940 is historically oriented, students of speech communities  differ markedly from their colleagues who concentrate upon textual analysis . The latter tend to treat languages as independent wholes that branch off from uniform protolanguages in accordance with regular sound laws. The former, on the other hand, regard themselves primarily as students of behavior, interested in linguistic phenomena for their broader sociohistorical significance. By relating dialect boundaries to settlement history, to political and administrative boundaries, and to culture areas and by charting the itineraries of loanwords in relation to technical innovations or cultural movements, they established the primacy of social factors in language change, disproving earlier theories of environmental or biological determinism…
The shift of emphasis from historical to synchronic problems during the last three decades has brought about some fundamental changes in our theories of language, resulting in the creation of a body of entirely new analytical techniques. Viewed in the light of these fresh insights, the earlier speech-community studies are subject to serious criticism on grounds of both linguistic and sociological methodology. For some time, therefore, linguists oriented toward formal analysis showed very little interest. More recent structural studies, however, show that this criticism does not affect the basic concept of the speech community as a field of action where the distribution of linguistic variants is a reflection of social facts. The relationship between such variants when they are classified in terms of usage rather than of their purely linguistic characteristics can be examined along two dimensions: the dialectal and the superposed.
Dialectal relationships are those in which differences set off the vernaculars of local groups (for example, the language of home and family) from those of other groups within the same, broader culture. Since this classification refers to usage rather than to inherent linguistic traits, relationships between minority languages and majority speech (e.g., between Welsh and English in Britain or French and English in Canada) and between distinct languages found in zones of intensive intertribal contact (e.g., in modern Africa) can also be considered dialectal, because they show characteristics similar to the relationship existing between dialects of the same language.
Whereas dialect variation relates to distinctions in geographical origin and social background, superposed variation refers to distinctions between different types of activities carried on within the same group. The special parlances described above  form a linguistic extreme, but similar distinctions in usage are found in all speech communities. The language of formal speechmaking, religious ritual, or technical discussion, for example, is never the same as that employed in informal talk among friends, because each is a style fulfilling particular communicative needs. To some extent the linguistic markers of such activities are directly related to their different technical requirements. Scientific discussion, for instance, requires precisely defined terms and strict limitation on their usage. But in other cases, as in greetings, forms of address, or choosing between "isn't" and "ain't," the primary determinant is the social relationship between speakers rather than communicative necessity. Language choice in these cases is limited by social barriers; the existence of such barriers lends significance to the sociolinguistic study of superposed variation.
This distinction between dialectal and superposed varieties obviates  the usual linguistic distinction between geographically and socially distributed varieties, since the evidence indicates that actual residence patterns are less important as determinants of distribution than social interaction patterns and usage. Thus, there seems to be little need to draw conceptual distinctions upon this basis .
3 / Les langues standard sont des langues-relais
«Wherever standard languages are well-established,
they act as the ultimate referent that determines the association
of a given local dialect with one language or another.»
Superposed and dialectal varieties  rarely coincide in their geographical extent. We find the greatest amount of linguistic diversity at the level of local, tribal, peasant, or lower-class urban populations. Tribal areas typically constitute a patchwork of distinct languages, while local speech distribution in many modern nations takes the form of a dialect chain in which the speech of each locality is similar to that of adjoining settlements and in which speech differences increase in proportion to geographical distance. Variety at the local level is bridged by the considerably broader spread of superposed varieties, serving as media of supralocal communication. The latin of medieval Europe and the Arabic of the Near East form extreme examples of supralocal spread. Uniformity at the superposed level in their case, however, is achieved at the expense of large gaps in internal communication channels. Standard languages tend to be somewhat more restricted in geographical spread than classical languages, because of their relationship to local dialects. In contrast to a society in which classical languages are used as superposed varieties, however, a standard-language society possesses better developed channels of internal communication, partly because of its greater linguistic homogeneity and partly because of the internal language loyalty that it evokes.
In fact, wherever standard languages are well-established they act as the ultimate referent that determines the association of a given local dialect with one language or another. This may result in the anomalous situation in which two linguistically similar dialects spoken on different sides of a political boundary are regarded as belonging to different languages, not because of any inherent linguistic differences but because their speakers pay language loyalty to different standards. Language boundaries in such cases are defined partly by social and partly by linguistic criteria.
4 / Répertoire d'une communauté de parole
The totality of dialectal and superposed variants regularly employed within a community make up the verbal repertoire of that community. Whereas the bounds of a language, as this term is ordinarily understood, may or may not coincide with that of a social group, verbal repertoires are always specific to particular populations. As an analytical concept the verbal repertoire allows us to establish direct relationships between its constituents and the socioeconomic complexity of the community.
We measure this relationship in terms of two concepts: linguistic range and degree of compartmentalization. Linguistic range refers to internal language distance between constituent varieties, that is, the total amount of purely linguistic differentiation that exists in a community, thus distinguishing among multilingual, multidialectal, and homogeneous communities. Compartmentalization refers to the sharpness with which varieties are set off from each other, either along the superposed or the dialectal dimension. We speak of compartmentalized repertoires, therefore, when several languages are spoken without their mixing, when dialects are set off from each other by sharp isogloss bundles, or when special parlances are sharply distinct from other forms of speech. We speak of fluid repertoires, on the other hand, when transitions between adjoining vernaculars are gradual or when one speech style merges into another in such a way that it is difficult to draw clear borderlines.
 A savoir: les spécialistes de la dialectologie ou des pidgins et des créoles.
 A savoir: les indo-européanistes, les romanistes ou les slavisants, les spécialistes des unions de langues au sens de Troubetzkoy.
 Les deux principales critiques portées contre la dialectologie, qui conduisirent Chomsky à formuler la distinction devenue classique entre compétence et performance, étaient 1°) de procéder par collecte empirique de mots et de sons sans tenir compte des situations d'emploi, et 2°) de comparer les matériaux ainsi collectés dans une perspective diffusionniste.
 Ces parlers spéciaux, mentionnés dans un paragraphe antérieur de cet article, étaient essentiellement the languages of occupationally specialized minority groups, c'est-à-dire le jargon d'une profession, ou encore l'argot et les langues parlées entre initiés. Aujourd'hui, dans la région parisienne, on en donnerait comme exemple le «neuf-trois».
 This distinction… obviates the usual…: Cette distinction entre variation dialectale et variation d'un registre à un autre «pare aux [difficultés de] la distinction habituelle» entre variation géographique et variation sociale.
 Cette base, ce serait la géographie (actual residence patterns), mais la géographie ne joue qu'un rôle mineur dans «la distribution des variétés» (varietal distribution).
 Autrement dit: les registres (superposed) et les dialectes.
 Ce sont les «langues classiques» en position de langues dominantes dans une diglossie. L'invention de la diglossie par Ferguson date de 1959.
 Une «grande» langue, standardisée, sert de relais aux dialectes et aux langues locales, qui sont dites «petites» langues, dans leur association au sein d'une union de langues à base pragmatique. Le concept forgé par Gumperz d'une langue servant de «référent qui détermine l'association d'une [langue] locale» (referent that determines the association of a given local [language]) avec toutes les autres au sein d'une même aire sociolinguistique, c'est tout simplement le concept d'une langue que j'ai nommée langue-relais.
 Un répertoire plus ou moins riche et un plus ou moins cloisonné.
 Par un découpage (bundles) suivant des lignes de démarcation isoglosses bien tranchées.