A Linear Grammar of Speech 101
This comment is broadly in line with Sinclair’s (1991: 110) observation that,
at times, speakers assemble speech from lexical elements coterminous
with orthographic words (the open-choice principle), but in the majority
of cases they assemble speech from items coterminous with more than
one orthographic word (the idiom principle), which may appear to be
analysable into smaller segments. Stubbs (2002: 14) agrees and argues
that combinations of words in phrases are a strong candidate for the core
semantic unit of language. The implication appears to be that breaking
down chunks into smaller segments does not help to construct a transpar-
ent and descriptively accurate grammar.
However, the diffi culty is that as of yet no-one has successfully identifi ed
and coded these core semantic units. Accordingly, caution is in order when
coding grammatical chains. Brazil (1995: 44) remarked that his use of
traditional terms such as N and V elements was:
no more than a convenience, and one which we must be prepared to
abandon if and when the need arises.Thus, if, or perhaps when, corpus linguists identify the core semantic units
of the language a grammar of used language should be represented as
a chain comprised of such units. Such a recoding would have the dual
advantage of rendering the coding more descriptively transparent as well as
more psychologically refl ective of how speech is produced. Until that day,
however, there is no choice but to code using traditional conventions.
To summarize, the discussion in this section leads us to propose that a
PHR (phrase) element be added to Brazil’s descriptive coding in three
instances. First where orthographic words coalesce into a larger element,
such as an idiom, which cannot be decomposed without loss of meaning;^18
second where V elements are in phase; and fi nally where the V–ing pattern
refers to a future activity.
4.4 Two Features of Spoken Language
Brazil claims that his chaining rules have the potential to describe all pos-
sible instantiations of telling and asking exchanges. A moment’s refl ection,
however, is enough to show that the rules are too restrictive to account for
all instances of English speech. This section looks at two features of spoken
language: ellipsis and dysfl uencies which have been chosen because, while
unmentioned in Brazil (1995), their existence is incontrovertible and they