90 A Grammar of Spoken English Discourse
a new intermediate state. The speaker remains obliged to produce the
V element fell, which completes the telling. According to this view, if the
hearer proceeds up a garden path, it is not because the hearer has incor-
rectly parsed the sentence, but rather because the speaker has misjudged
the state of speaker/hearer convergence and assumed incorrectly that the
hearer recognizes which horse is under discussion.
The Chomskyan view of language, which argues that the language sys-
tem is mentally represented innate de-contextualized knowledge which a
speaker accesses prior to use, has been criticized by Hopper (1987, 1998)
as being incapable of explaining how language is both ontogenetically
and phylogenetically acquired and used. Hopper argues that grammar
emerges from the discourse and is itself shaped by the discourse as much as
it shapes the discourse. Structure is not, he says, the result of an overarch-
ing set of principles but is instead the spreading of regularities in discourse.
Hopper (1998: 159) argues that each individual’s speech is ‘a vast collection
of hand-me-downs that reaches back to the beginnings of the language’.
The language each individual uses is infl uenced by the speaker’s unique
and individual experiences with the language. Everyday language is not a
collection of freely constructed novel sentences but is instead built up out
of combinations of ready-made regularities previously experienced by the
speaker and pre-existing in the discourse.
Some empirical support for Hopper’s theory is found in Elman (1990:
195–203) who describes a ‘sentence generator program’ which he used
to construct a set of two- and three word ‘sentences’. After a number of
training sets the program developed internal representations which allowed
it to predict which kind of words followed other words. Despite not being
trained to recognize the categorical distinction between nouns and verbs
the programme learned to recognize that certain words (verbs) typically
followed other words (nouns) and that certain verbs prospected a direct
object while others did not; it learnt how to distinguish transitive from
intransitive verbs (ibid. 199). Weber (1997) provides some further support
for Hopper’s theory. He claims that because linguistic meaning is inherently
emergent it can only be explained by a grammar such as Hopper’s which is
‘dynamic, individual and indeterminate’. More support for Hopper’s theory
is found in Pierrehumbert (2001: 143) who states that a usage based theory
is a more accurate predictor of the realities of speech than a rule based
approach. She (ibid. 143) argues that a usage based theory is better able
to predict and explain regularities and differences in the lenition of
phonemes across languages and dialects than can a rule based context
free theory such as Chomsky and Halle (1968).