It is perhaps no more than trite to suggest that multilingualism has been a fact of social life in Africa for a very long time, but facts lack significance until they are brought within an ordered frame of reference, and for many years both linguist and social anthropologist were more concerned with uniformity than with diversity, with a holistic rather than an atomistic view. In social anthropology this was due in part to a legacy of Malinowskian functionalism with an emphasis on units like that of the tribe, worked out in conditions very different from the African setting. 1 It also derived in part, I think, from the development of techniques for the intensive study of small communities over long periods: if such study were to form a viable basis for extrapolation to the tribe or tribal group, 2 then there must be some presumption of uniformity. Characteristically, concern for language in such studies was pragmatic: one learned the language of one’s people and believed intuitively in an entity labelled ‘their language 3 . In some cases language featured specifically in the definition of tribe, 4 but more usually it constituted simply one of a number of distinctive features. Occasionally this was questioned, thus, ‘If the Nupe tribe is thus not a local group in the strict sense of the word, it is not a linguistic 2unit either’, 1 but Nadel was certainly one of the few people at the time to appreciate the social significance of language and language variation, though in his Saussurean distinction between the operational and structural aspects of languages he misses completely the significance of the operational aspects for linguists. 2 Linguists, for their part, eschewed the situational view of language stressed by Malinowski, and concentrated on making extremely detailed descriptions of a limited corpus of material obtained from one or more informants. This is not to question the value of the single-informant technique as a procedural imperative in the face of linguistic complexity, but to ignore the fact that ‘even a single individual comprises a range of styles and a span of history’, 3 and to use the material as a sample from which to extrapolate to the point where one could say that it constituted a grammar of language X seems to me much more questionable.