A Linear Grammar of Speech 111
kind of deserted is not notated as a separate increment. Both examples appear
to fulfi l a communicative need but an increment is only possible if it con-
tains an end-falling tone; the point is simply that increment boundaries
cannot be determined without reference to intonation.
The fi nal point to be considered discusses elements which Brazil’s gram-
mar does not encode. These fall into two categories, the fi rst of which is
linking elements such as and, but and so. Brazil (1995: 218) argues that the
absence of such linking elements does not alter the communicative value of
the utterance. However, a difference in linking element seems to alter the
communicative value of (58),
(58) She went to the local school and/but got into OxfordTherefore in the interests of producing a grammar that codes as many
meaningful elements as possible, they are coded here using the Cobuild
convention as C. Brazil’s coding of linking elements using an ampersand
appears to suggest that linking elements are always additive. The second
category consists of ‘miscellaneous elements like well, anyway, and I mean in
circumstances where they cannot be said to represent sense selections or
enter into the organization of chains’ (Brazil 1995: 214). Nevertheless, such
elements express interpersonal meaning and hence they are again coded
using the Cobuild conventions. Appendix 1 reprints the fi rst 25 lines of the
analysis in Brazil (1995) to illustrate the suggested changes.
4.6 Conclusion
This chapter has demonstrated that objections to the idea that a grammar
of used language is feasible, based upon arguments that fi nite state gram-
mars cannot generate all the possible sentences of a language, are not
applicable. Discussion of the work of Hunston and Francis (2000) showed
that lexical items contract syntagmatic relations with other lexical items. An
individual lexical element not only prospects a following lexical element
based upon its class membership but also prospects a following lexical ele-
ment based upon its own pattern. Sinclair’s open-choice and idiom principles
were explained and discussed. Some evidence based on the discussion of
verbs in phase, the analysis of the V–ing pattern, and slips of tongue sug-
gests that the idiom principle is the default. However, because no foolproof
way of identifying such idioms presently exists, the descriptive coding used
in this book will employ Brazil’s conventions and not attempt to encode