Language in the world: properties and powers
Linguistic autonomy
Interaction between linguistic levels
However, further intralinguistic properties can be identified when the interactions between the different levels of language are taken into account, and this begins to throw doubt on any notion of absolutely autonomous - i.e. separate - levels of the linguistic system.
Poets, humorists and many other creative users of language take advantage of coincidences between the sounds of words and their meanings. Such resources of language play are affordances of the linguistic system itself, as is demonstrated by the fact that both poems and jokes often lose a lot in the process of translation. Written languages afford different kinds of potential for play and creativity in the interaction between sound, visual symbol and meaning. Lewis Carroll's quip in Alice in Wonderland, Take care of the sense and the sounds will take care of themselves', and the visual poem The Mouse's Tail' (ibid,) are both examples of interplay between the different levels of language, whose meaning is emergent from this interaction.
Another example of intralinguistic interaction is where speakers of a variety with rich case marking, in the process of many uses over a long time, lose a phonological distinction which also serves as a morphological marker. Nettle (1999) suggests that this small phonological change could in time lead to a 'shift in state space' for this variety, from a highly inflected one to one with increasingly fixed word order. Simulations, using probabilistic modelling, can replicate patterns of change in a language, but it is crucial to maintain the distinction between human beings, who can make choices about language and how to use it, and linguistic systems, whose independence and autonomy from speakers can only ever be limited. Nevertheless, we propose that examples such as these illustrate properties of languages which are, as we would expect of a cultural emergent property, partially autonomous, in that components of the linguistic system interact with each other leading to effects on, and changes in, the system itself.
Let us now turn our attention to the interplay between two further levels of the linguistic system, namely grammar and lexis. We may start by observing that many scholars of child language development now doubt the existence of an autonomous grammar module, arguing instead for 'a unified lexicalist approach to the processes by which grammar is acquired, used, and represented in the brain' (Bates and Goodman 1999: 37). These authors point out that *... the historical trend in modern linguistics has been to place in the lexicon more and more of the work that was previously carried out in a separate grammatical component' (p.53). Relations internal to the linguistic system which act as constraints on what speakers can do with the linguistic resources available to them may be explored with reference to the empirical evidence of linguistic corpora. Here we are sympathetic to Beaugrande's (1997: 187) proposal that '[o]ur safest recourse is to suspend the project of separating grammar from lexicon and to use a large corpus of authentic data to determine how far the two sides share the "expressive work" in a given language or discourse domain' (cf. Halliday 1978). Corpus research is revealing ever more robust patterns of collocation, colligation, semantic preference and prosody which cannot be accounted for by descriptions of language based on slot-and-filler categories (Clear 2000), As Sinclair (1991: 108) puts it:
[t]he model of a highly generalized formal syntax, with slots into which fall neat lists of words, is suitable only in rare uses and specialized texts. By far the majority of text is made of the occurrence of common words in common patterns, or in slight variations of those common patterns.
Sinclair discusses, for instance, the word set, which frequently occurs in combination with other common words to make phrasal verbs such as set about, set in, set off and so on. Corpus evidence reveals that the meaning of the combined 'chunk' is not reducible to either of its constituents (see Carter and Sealey in press). Although neither set nor in has any particular connotations, either positive or negative, the collocates of set in often denote something unpleasant. The kinds of thing which 'set in' are 'drought', 'rain', 'depression', 'rot' and so on. Thus as Sinclair says, '[t]he co-occurrence of two quite common little words can unexpectedly create a fairly subtle new meaning that does not seem to be systematically related to either or both of the original words' (1991: 68) - another example of intralinguistic emergence.
Corpus linguists report their findings in terms which, once again, connote the partial independence of aspects of language. For example, Beaugrande (1997: 209) identifies core items (in this case, the dispositive 'giving to') as 'strong attractors'. Examples from Hunston (2002) include (with agential metaphors highlighted): '... concordances can be used to give very general ideas about the ways that words behave and the meanings that can be associated with patterns' (pp.50-1); 'Collocation is the tendency of words to be biased in the way they co-occur' (p.68). We would obviously reject any literal implication that words and phrases are choosing how to 'behave', and we do not assume that the authors intend us to draw any such inference, but this attribution of a limited degree of agency to items within the linguistic system, based on quantitative evidence, may be read as a further indication of the emergent properties of language itself.
