language as a cultural emergent property
'Real language' in applied linguistics = 'real sentences' in the classroom?
Let us take one topic of debate as an example of how competing theories about what constitutes 'language' are implicated in an applied linguistics problem. This topic is chosen because it relates closely to the issue of 'real' language. It is the question of whether language teachers should make use of invented sentences or should restrict themselves to authentic examples only. Although at one level this is an argument about pedagogy, at another it rests on (possibly competing) beliefs about the 'real' properties of language.
Several applied linguists are explicit proponents of the use in the classroom of attested examples of the language-to-be-learned. One core element of their argument is that what learners are engaged in is the processing (both productively and receptively) of language behavior. In other words, what they usually want is engagement with those very aspects of language that Chomsky would classify as epiphenomena of 'real language'. As Cook (2001) points out, advocacy of authentic sentences in teaching has a long history, but the arguments have been given new momentum with the advent of linguistic corpora. Thanks to this particular technological development, those responsible for providing learners with examples of the target language (that is, mainly, teachers and textbook writers) now have access to large databanks of examples of language behavior, for example in the Bank of English (the result of a collaboration between a university, Birmingham, which has primarily scientific purposes, and a publisher, HarperCollins, with its commercial, and thus, indirectly, pedagogic priorities). One of the principal uses to which the Bank of English has been put is to contribute to textbooks for learners of English as a foreign language, and Sinclair has this to say about the approach (emphasis added to key words):
It is now generally accepted that it is extremely difficult to invent examples which sound realistic, and which have all the features of natural examples. I am convinced that it is essential for a learner of English to learn from actual examples, examples that can be trusted because they have been used in real communication. (COBUILD 1990: vii)
Another corpus with a partly pedogogic motivation is CANCODE (the result of a collaboration between Nottingham University and Cambridge University Press). This corpus comprises transcripts of spontaneous spoken English, and is thus a repository for examples of the kind of spoken language behavior that non-corpus-based text-books often omit. The best course of action to remedy this deficit, claim Carter and McCarthy (1995: 154), 'would seem to be to expose learners to natural spoken data wherever possible' (emphasis added).
Now the theories made most explicit in the debate about this topic (e.g. Carter 1998; Cook 1998) are primarily pedagogic, secondarily linguistic and only tangentially ontological: that is, the writers are not explicitly debating the nature of what language 'really is'. Our argument, though, is that ontological claims are entailed by the claims that are made in this debate. So, while not intending to impute to the protagonists positions they would not adopt themselves, let us try to elucidate some of the claims about the nature of 'real language' which are implicit in this debate.
Claim 1: The use of the terms 'natural', 'actual', 'attested' and 'authentic' as near synonyms for 'real' points to one ontological proposition: what's real about language is what can be empirically observed. If one wants 'real' language data, one consults a repository of examples collected by observers of language behavior (in the form of transcribed recordings or of communicative writing).
Claim 2: Language has some properties which are rarely observed empirically, and evidence of these (in this case for teaching purposes) has to be artificially brought into being. One of these properties is a distinction between form and meaning. Thus an example sentence which illustrates the formal, grammatical properties of a language may fail to illustrate its communicative, semantic properties. (Perhaps the most well-known invented sentence used to illustrate this is Chomsky's 'Colourless green ideas sleep furiously', which is grammatically acceptable but semantically obscure at best.) This 'real' property of language (Le. the potential to be grammatical but communicatively inauthentic) is not revealed in language users' routinely observed behavior.
Claim 3: Language has some properties which are not reducible to individual instantiations of it, such as utterances. One illustration of this, from a proponent of the use of invented examples, is the observation that, thanks to the human capacity for creativity, language users can, and often do, put even the most bizarre sentences into some kind of imagined context which will render them as discourse (Cook 2001). An illustration of this claim from the opposing position, that is from proponents of corpus-based teaching, is that it is only by looking at lots of examples of individual sentences (or, more accurately, of concordance lines) that some of the properties of language become evident.
We will discuss an ontological position which we hope will help to accommodate, or at least adjudicate between, those aspects of these claims that seem to be in conflict. But first let us examine where we have reached in respect to determining what might be 'real' about language for these various applied linguists.
If what is real about language could be exhausted by what can be empirically observed, then neither the attested sentence nor the invented one could be an 'example' of any of the properties which might underlie that sentence. The concept, for instance, of 'relative clause' is a theoretical description, and is not given, unmediated, to the observer of any sentence which contains one, invented or attested. Many, perhaps most, speakers of a language manage perfectly well as communicators with no concept of 'relative clause', though this concept may be useful to learners of an additional language, which is why applied linguists concern themselves with how best to teach it: it becomes, pedagogically, 'a target structure'. The idea of a target structure presupposes that language is 'structured'. This claim is largely uncontroversial, although, there are those who are skeptical that 'the resources of the descriptive apparatus available' can capture any supposed 'underlying "system" , of linguistic universals (Harris 2001: 126). However, if there are such structures within language, then it is often a teacher's goal to illustrate one such aspect of it, to make it more apparent, more central and uncluttered by features which are irrelevant and potentially misleading.
Both putative example sentences share the property of being instantiations of some more generic characteristic of the linguistic system- perhaps of the fact that the system includes structures such as relative clauses. (Indeed, as Cook (2001) points out, the terms 'invented' and 'attested' need not be mutually exclusive: all attested sentences had to be 'invented' by their original speakers; and sentences invented by teachers or textbook writers become attested, even if only in the pedagogic context for which they were produced.) In pedagogic terms, the 'ludicrous' example is said by Cook to have the advantage of being memorable. For Carter and McCarthy, and others, the authentic example scores on another dimension, by illustrating what is typical, and, when contextualized among large numbers of similar examples, by revealing to learners the structure (again) clothed in so many variants of incidental features that it is the common property which becomes apparent. Two routes to the same end: to make apparent to the learner what the analyst/teacher knows is there, despite the fact that it is not empirically observable. So protagonists on both sides, it would seem, share (at least implicitly) the realist's belief in two levels of reality: the empirical (i.e. that which can be perceived by the senses), and something distinct from the empirical (i.e. phenomena, such as those labelled by linguists 'relative clauses', which cannot be seen directly but which nevertheless exist as features of language).
