Quoting John Cowan <[log in to unmask]>:

> Andreas Johansson scripsit:
> > Quoting Ray Brown <[log in to unmask]>:
> >
> > > In view of the richness of our language and the wide use of metaphor
> > > in 'ordinary' English (one doesn't have o resort to 'poetry'),
> > > I wonder if it is, in fact, possible to produce a sentence using
> > > standard English words in acceptable grammatically formed sentence
> > > that is completely stupid from a semantic point of view. Now there's
> > > a challenge  :)
> >
> > The easy solution would seem to be oxymorons: He's got a five-wheeled
> > trike.
> I think it would be worthwhile at this point to recall Chomsky's original
> motive for constructing the CGISF sentence.  He was not directly concerned
> with semantics at all: rather, he was refuting a false theory of syntax.
> A behaviorist (I don't remember who) had proposed that syntax was
> associational: in particular, that a sentence was grammatical if and only
> if each pair of consecutive words had a high collocational frequency in
> the known corpus of the language.  But by devising the CGISF sentence,
> Chomsky showed that even though "colorless green", "green ideas",
> "ideas sleep", and "sleep furiously" were all extremely rare if not
> nonexistent collocations, the sentence as a whole was still grammatical.
> Consequently, syntax had to be (broadly) generative: it had to have
> the capability of constructing utterly novel sentences.

I might be missing something, but it seems to me that a such hypothesis would
be refuted more easily and convincingly by presenting a sentence in which each
consecutive pair has a high collocational frequency, but which taken as a
whole is flatly agrammatical?

(English, with its general lack of agreement, isn't the easiest language to
construct such sentences in, but try for instance "I killed themselves".)