🚀 go-pugleaf

"Attested" data is, for me, anything which has been widely circulated
and which can be reproduced by some speakers (at least > 1). My 
preference is to use unelicidated data, which I collect during
casual encounters with native speakers, prefereably monolinguals.
Unfortunately, I can only do this in a few languages -- Russian, Khmer
and some varieties of English, with a bit of German as well. When I
need data from other languages I try to use data from linguists who
speak the language and have collected data from native speakers.

I don't badger informants. If I did, I would wind up with tons of
data just like the stuff that appears in LI and in much generative
work on parasitic gaps, resumptive pronouns and the like. Until I
hear an English construction spontaneously produced, I don't trust it
much. Much of the data I read strikes me as far too artificial. On
the other hand, what would we do if we had to create a set of rules
for George Bush's attempts at sentence formation? <grin>

Eric Schiller