A Linear Grammar of Speech 87
view but instead argues that a linear grammar also provides a feasible
description of used language. It does this by showing that the objections
raised against the feasibility of linear grammars are not applicable to the
proposed grammar of used language.
An objection raised against linear grammars is that sentences must be
parsed to be understood (Singer 1990: 57). To exemplify the point, he
produces the following examples:
(1) Wild beasts frighten little children
(2) *Beasts children frighten wild little^2It is clearly correct to argue that because ‘sentence’ (2) fails to comply with
the formal rules of grammar it is grammatically unacceptable. However, an
alternative view, that ‘sentence’ (2) is judged ungrammatical because it fails
to fulfi l any conceivable communicative purpose, appears equally feasible.
According to this view ‘sentences’ are judged to be grammatical only if
they are capable of fulfi lling a conceivable communicative purpose in the
context in which they were produced. A further argument in favour of the
belief that sentences must be parsed in order to understand their meaning,
is Chomsky’s view that people can recognize some nonsense sentences as
grammatical, e.g. (3) or ungrammatical, e.g. (2).
(3) Colorless green ideas sleep furiously. (Chomsky 1957)But again, as (3) fails to satisfy any obvious communicative need it is unlikely
to be an increment except in marked communicative situations. Accordingly,
a grammar of used language does not have to concern itself with explaining
how (3) can be identifi ed as an abstract unit which grammarians label a
sentence. But if one were forced to explain the apparent grammaticalness
of (3) a tentative explanation could go as follows: (3), like the nonsense
poetry of Edward Lear,^3 could conceivably satisfy a communicative purpose
in a particular genre of language such as a children’s story. Because of this
potential to fulfi l an imagined though marginal communicative need, people
may judge that (3), unlike (2), has the potential to be grammatical, and
therefore under experimental conditions judge it grammatical.
While it is certainly true that a message can be interpreted by parsing its
constituent parts this does not necessarily entail that the message cannot be
described linearly. Chomsky argues against the feasibility of linear grammars
by demonstrating that a fi nite state grammar^4 is incapable of generating all
the possible sentences of a natural language. He (1975: 30–1) speculates