Mark comments in a post earlier today: "I don't know of any memorable linguistic April Fool's hoaxes." Well, I think I know of at least two.
About a decade ago, Chris Barker released by email the abstract of a new paper he had written. It sounded fascinating: formalizing principles and parameters theory arithmetically to get a number-theoretic model of it, and using techniques from modern cryptography, he had managed to prove that the problem of discovering the right parameter settings from a set of data depended on a trapdoor function: it was as hard as breaking modern prime-number encryption. The paper had been made available at an ftp site. I downloaded it straight away (back in those dark days, it was the first scholarly paper I ever obtained by downloading). When I unzipped it, I found all I had was a polite apology to the effect that although the idea was neat he had not in truth been able to achieve this result. Appended was a recommendation to check the date of release. It was, of course, April 1.
The other was when David Pesetsky, writing from Mark's
doctoral alma mater, put out what I suspect was a faked
ASCII screen dump from a fictive newswire story about how
the weight of the snow on the roof of MIT's Building 20 had
collapsed it and buried Noam Chomsky, Morris Halle, Jay Keyser,
the editorial office of Linguistic Iniquity Inquiry,
and the entire MIT Linguistics department in a welter of roofing
material and slush. He had quotes from Chomsky and everything.
Sympathy notes flooded into the department, though I suspect
slush did not. Bitten once by Barker's cruel hoax, I did not make
any linguistic inquiries. The April 1 dateline had only a 1/365.25
= 0.0027 probability, after all, and April was a little late for major snow
buildup even in the ghastly climate of New England. I never asked
around, but I suspect that Noam
passed that April 1 warm and dry, and as comfortable as that
(now demolished) building ever got.