It is increasingly common for primary linguistic fieldwork to be conducted with “last speakers,” as swingeing language extinction brings a belated attention to the need to document endangered languages. Data from “last speakers” must, however, be treated with caution, given that the variety they speak may have been simplified through various processes of language death (see Schmidt 1985: 41) – though this is by no means always the case – and/or heavily influenced by interference from whatever other language(s) they use in day-to-day communication. Nonetheless, many detailed and subtle grammars of Australian languages, for example, have been written on the basis of data from a single last speaker; recent examples are Dench's (1995) grammar of Martuthunira, and Harvey's (1992) grammar of Gaagudju.

Such works clearly validate the possibility of carrying out linguistic fieldwork with last speakers, but it is imperative that researchers be aware that the definition and identification of “last speakers” is highly problematic, and prone to constant redefinition from both the speech community's and the researcher's point of view. There are of course rather obvious cases, where the death of one “last speaker” is followed by the fortuitous discovery of another speaker, equally or more fluent, in some other location, or where the community's definition of total linguistic competence adjusts to the erosion of stylistic, grammatical, or lexical complexity, so that in a succession of what the community considers “last speakers” each knows less, in some objective sense (see section 1).