[Avery Andrews (920511.2048)]
(Bill Powers 920502.1200)
Observations of grammatical errors and subsequent corrections are
useful (people do this sort of thing, and I'm pretty sure that both
Bruce Nevin and Penni Sibun know more about it than I do), but I don't
think they are a substitute for ordinary grammatical observations, for
the reason that you can't apply the Test, and systematically create
disturbances for them to neutralize. All you see is them changing
their mind about what to say and backtracking a bit, and there's
really only an inference that control is what is involved, and you
don't have much of an angle on what they're controlling for either
(sounding cultured? like a real person? or what?)
Extended work on an individual's language is also problematic from the
point of view of both practicality and ethics--to record really large
samples of someone's ordinary use of language would pretty well
trash their privacy, and those of the people they were talking to.
I suspect that it wouldn't be worth it, either, on the basis
that an individuals `internalized grammar' is a selection from what's
floating around in the speech community, & there wouldn't be much
significance to which bits happened to fetch up together and any
particular person's brain.
Of course the standard generative grammarians `can you say this: ...'
method also has pretty strong limitations.