1. LINGUISTIC TRADITIONS

For de Saussure, language structure (langue), rather than language in use (parole), was the proper object of linguistic enquiry, a position also adopted by Chomsky, who labelled this dualism competence and performance. Language was a biologically determined mental process and social factors should therefore be excluded from linguistic analysis.

The principles underlying a corpus-based approach to the analysis of language, however, are concerned with language in use as opposed to language as a mental construct. This view of linguistics can be traced back to Firth - who posited that language be studied as part of the social process - and is a stance also adopted by Halliday, whose “grammar is semantic (concerned with meaning) and functional (concerned with how language is used)” (Bloor & Bloor 1995:2). John Sinclair, initiator of the first corpus-based dictionary of general language (Cobuild 1987) at Birmingham University, also advocates the notion of meaning and use as central to understanding language.

One particularly important de Saussurean concept refers to the syntagmatic and paradigmatic dimensions of language. Syntagmatic (horizontal) relations represent a potential for the combination of items, whereas paradigmatic (vertical) relations represent a potential for substitution among items. Conventional linguistics has tended to focus on paradigmatic relations, a fact reflected in traditional dictionaries, which are constrained by orthographic considerations that would help a user locate items. Nonetheless, “paradigms are by definition things which do not go together” (Halliday 1994:xxxii), whereas “a substantial proportion of speakers´ associations to word stimuli are syntagmatic” (Fellbaum 1998:9).

A syntagmatic view of language takes account of the contribution of sense and syntax to meaning. The argument that “sense and syntax” (Sinclair 1991), or “meaning and pattern” (Hunston & Francis 2000), are associated is based on two pieces of evidence. Firstly, meanings tend to be distinguished by differing patterns, and secondly, words with the same pattern sometimes share aspects of meaning. According to Hunston & Francis (ibid:255-259), the fact that there is no perfect one-to-one correspondence between meaning and pattern is attributable to a lexicographic rather than a semantic worldview.


2. THE NOTION OF CO-SELECTION

Co-selection describes the general phenomenon of words that habitually keep company, to paraphrase Firth. (Note: Given time and space constraints, two particular cases of very fixed syntagmatic relationships - multiword and idiomatic expressions – will be excluded from this discussion; for a general overview see, however, Partington 1998; Nelson 2000). Sinclair’s definiens of what he refers to as collocation is “the occurrence of two or more words within a short space of each other in a text” (1991:170); this could logically refer to co-selection between lexical or grammatical items. Other authors draw a distinction between collocation and colligation (Firth 1957; Bahns 1993; Hoey 1997), using the former to refer to the co-occurrence of lexical items and the latter to the co-occurrence of grammatical items; Sinclair himself refers to colligation within a collocation context, yet has stated that “collocation in its purest sense recognises only the lexical co-occurrence of words” (1991:170). The debate is complicated further by the fact that the derived term collocate is generally used to refer to lexical items (cf. the Louw definition of semantic prosody below). For the purposes of this article, therefore, co-selection will henceforth be used to refer to the general phenomenon from a translator’s perspective, collocation to the co-occurrence of lexical items as per Sinclair, and colligation to co-occurrence with grammatical items. The preferred usage of individual authors will, naturally, be respected.

Sinclair describes what he terms as “collocational principles” (1991:109) as follows:

· The open-choice principle - on which virtually all grammars are based - refers to the possibility of selecting words to ‘slot and fill’ a unit, with the only restrictions being grammatical ones.
· The idiom principle, on the other hand, refers to “the large number of semi-preconstructed phrases that constitute single choices, even though they might appear to be analysable into segments”.


The open choice principle, therefore, would broadly correspond to paradigmatic choice (restricted by grammaticality), whereas the idiom principle corresponds to syntagmatic choice (restrictions may be lexical, grammatical or semantic). Sinclair identified colligation and semantic prosody as particular features of the idiom principle.

Colligation (defined above) has also been referred to in terms of collocational frameworks (Renouf & Sinclair 1991:128-144), which are units based on a grammatical, as opposed to lexical, core (e.g., ‘the/an...of’, ‘too…to’). Hunston & Francis’s (2000) analysis of ‘words and their patterns’ (from the perspective of concordance lines) and ‘patterns and their words’ (from the perspective of word-class labels) describe a dual approach to patterning from the lexical and grammatical perspectives; likewise, Benson et al. (1997) organise their dictionary in terms of both grammatical collocations and lexical collocations.

Semantic prosody, defined as “the consistent aura of meaning with which a form is imbued by its collocates” (Louw 1993:157), refers to an additional layer of perceived meaning, over and above that accorded by lexical and grammatical patterning alone. The notion of semantic prosody posits an initial selection of word or phrase in relation to which choices are realised at the lexical, grammatical and semantic levels. To use Sinclair’s example (as cited in Tognini-Bonelli 2001:104), the collocation ‘barely visible to the naked eye’ reflects an expression of semantic prosody (difficulty experienced implied by barely), a lexical choice (the notion of seeing) and the requisite colligation (to the).

Although Sinclair has largely focused on positive and negative connotations, other authors (Nelson 2000; Stubbs 1996) have broadened the meaning of semantic prosody somewhat. Nelson’s (2000) use of the term to include people and places associated with ‘business’ is possibly over-extended, since if semantic prosody is an ‘aura’, then it would typically occur at the textual or extra-textual level. Typical co-selection at the co-textual level is more correctly referred to as semantic preference.


3. CO-SELECTION AND TRANSLATION


Two issues relating to co-selection are of particular concern to the translator:


· The co-selection of items from memory may be affected by SL interference (what Toury has referred to as the ‘law of interference’ (1995) or Baker (1992:54) as “the engrossing effect of source text patterning”), with the result that a translation seems unnatural. Bahns (1993:61), for example, attributes the majority of collocational errors to SL interference.
· Given that appropriate co-selection within a particular register is outside the range of everyday language, linguistic intuition may be inadequate in terms of producing appropriate language and style.


Traditional translation resources (e.g. dictionaries) are primarily word-focused and therefore of little help in resolving either problem. Indeed, a ‘dictionary answer’ to a translation difficulty is the mark of the inexpert translator, who typically tends to translate word-for-word. A semantic rather than lexicographic worldview, however, is provided by text; to quote Lindquist (1999:179):

With the emergence of computer tools for translators, texts have become increasingly useful as a rich source of lexical data that enable translators not only to identify appropriate collocations but also to interpret lexical items in their pragmatic and linguistic contexts.

Moreover, as Heid (2001:793) points out in relation to the recording of terminological data:

The…structuring of collocations according to semantic criteria is most useful for tasks like text production because it not only helps organise the material but also to access collocations by the types of abstract meanings they potentially express.

For the translator, therefore, successful problem solution depends on a semantic representation of the word in context.