A Linear Grammar of Speech 99
which coincides with the time reference of the fi nite verb. Examples (27) to
(29), from Quirk and Greenbaum (1973: 48) demonstrate the problem
with this analysis.
(27) They are washing the dishes {now}/{later}
(28) He’s moving to London
(29) The president is coming to the UN this week
Examples (28) and (29) show that the V-ing pattern has the potential to
signal an anticipated future happening. Example (27), without the addition
of the time adverbial, has ambiguous time reference. It seems impossible
to equate Brazil’s analysis with the potential futurity of utterances which
contain V–ing verb patterns. Perhaps the most likely explanation, which
accounts for the potential futurity patterns, is that the fi nite and non-fi nite
verb have merged, through the process of grammaticalization^12 into a single
lexical element which is used to label an utterance with a future time
reference. Accordingly it is suggested that example (27) be coded in two
different ways depending on the open-choice and idiom principles.
(30) They are washing the dishes [now]
N V V' d N
(31) They are washing the dishes [later]
N PHR-V d N
The coding in example (30) indicates that no grammaticalization has
occurred. The fi nite V element indicates a speaker selection of undifferen-
tiated time; the following V’ element indicates that the washing occurs
as the same time as the speaking. The coding in (31) on the other hand
indicates that grammaticalization has occurred and that the washing has
not yet occurred at the time of speaking and that are washing represents
one slot and not two in a grammatical chain.
The fi nal strand of evidence reviewed here is from research into speech
errors which provides some further evidence for the existence of pre-
assembled phrase-like elements. However, caution must be exercised
in generalizing from pathology as the breakdown of a system may not
necessarily refl ect its normal workings. Jackendoff (2002) observes that
little is known about how speech, rather than the abstract system of
language, is produced. Perhaps the most complete account to date is that
of Levelt (1989) who proposed a four-stage model of language production:
conceptualizing (translating or encoding thought into language); formulating