In Theory, it makes no difference what you SELECT in an EXISTS sub-query as it is just syntactical sugar. (ie Nothing is actually selected.)
I seem to remember someone doing a test in SQL2000 which showed that SELECTing a constant was marginally quicker than SELECT * as SQL2000 seemed to look up the column names. I have no idea if this is still true with SQL2008.
A lot of development shops specify that you should use SELECT 1, in EXISTS sub-queries, so they can easily check for lazy developers putting SELECT * in the rest of their code.