A Linear Grammar of Speech 91
To conclude, this subsection has argued that a grammar of used language
does not have to explain the grammaticality of nonsense sentences which
fulfi l no communicative need. Demonstrations that all the potential sen-
tences of the English language cannot be generated linearly do not entail
that a linear grammar is incapable of accurately describing used language.
The attested diffi culty that hearers have in understanding garden path
sentences can also be explained by a grammar such as Brazil’s. In recent
years, the theoretical underpinning of context free theories of language
has come under attack and the alternate hypothesis that grammar emerges
from regularities found in the discourse is not incompatible with a grammar
of used language.
4.2 The Prospection of Lexical Items
Brazil (1995) argues that grammar as an abstract system only exists in the
broadest terms. There are few rules as to what might be theoretically said,
though in practice many possible utterances are extremely unlikely. The
need to successfully achieve communicative ends ensures that only utter-
ances which match individual hearer’s previous expectations are likely to
be produced. He claims that speakers in pursuit of their communicative
goals produce lexical items which anticipate further lexical items. Stubbs
(2002: 20) agrees and argues that communicative competence involves
expectations of what is likely to occur in the discourse. Sinclair and
Coulthard (1975) introduced the notion of prospection, which they explain
as something occurring in discourse leading the hearer to expect some-
thing else to occur. Similarly Tadros (1985) speaks of prediction: the choice
of one element determining a following element (see also Slobin 1978: 17)
for an almost identical view, though Slobin argues that syntax is the
ultimate determiner of the position of an element in a sentence. Brazil’s
view is a little different; he argues that previously occurring lexical items
create expectancies which can be fulfi lled by a prospection from a limited
set of choices, i.e. a one-to-a-few relationship between the previous set of
choices and the prospected rather than a one-to-one relationship between
the previously uttered lexical element and the prospected element.
Hunston and Francis (2000) explore the issue of lexical prospection in
their careful study of the Bank of English and propose a pattern grammar; a
grammar based on lexical patterns rather than syntactic rules. They claim:
The patterns of a word can be defi ned as all the words and structures
which are regularly associated with the word and which contribute to its