Much of what Weir (2004) has to say about a priori context validity is relevant (pp 17-21). In the item "Write an Email to your friend/teacher/wtf..." what are we testing, if not the ability to write an email? Is handwriting, speed of handwriting, part of the construct? Is the ability to edit with a keyboard part of it? If we're to be in the domain of every day email writing, then is it reasonable to exclude automated spell and grammar checks?
I'm sort of reaching for my Critical Pedagogy hat again here, because of the possibility of established conservative educational opinion that tests should in someway hurt and be difficult, and that spell checks are somehow cheating... Which is interesting of itself, of course. [Is there scope for a CP angle on Testing Theory? Or am I looking around for windmills to tilt at?]