Debbie Hawhee, who I think you know, has put up a nice little post about exams (you calling my post little?) over on her station. It reminds me that I once thought up what I imagined to be the ideal exam structure. I would go like this:
→ read extensively in the field(s) for a year or so
→ write somewhere between one to three articles for publication based on and using that reading
→ have committee member read said article(s), providing revision suggestions
→ meet for oral examination
→ move on to diss, integrating written materials in some way
I should say that the PhD exam process I went through in grad school was not completely dissimilar from this, the main difference being that the writing got condensed into 72 mad, wild, crazed, fret-filled hours. My MA exams were neolithic, involving six hours, small desks, and several blue books. As I had to take the exams twice (I wrote about Katherine Hayles' Choas Bound the first time around; not a popular topic with my reader), that meant twelve hours total and double-the-blue-books.
Anyway. What I like about the write-an-article/have-a-conversation model is that it approximates the work we do, and that seems to make sense to me in terms of assessment.