POINTLESS TESTING: I'm spending the weekend (and I'm going to spend the next few weeks) studying intensively for my Written Qualifying Exams, and it's hard to be enthusiastic. The WQE is a new test format that's only been in use for a couple of years and it replaces the old Major Field Exam that used to be required in order to proceed in the graduate degree program. The WQE is very different from the MFE, however.
The MFE was written specfically for each student by his or her professors, and was based on the classes that the student had taken and the research that the student was doing. A PhD is supposed to be a narrowly focused degree, and the MFE was designed to ensure that the student was proficient in the field (within computer science) that he or she was going into. The tests explored the field in generous depth (so I'm told) and could be quite hard. However, it was generally understood that a student's chance of passing was mostly related to how much his advisor liked him and how confident his advisor and committee were that he would be a successful PhD candidate.
The WQE, on the other hand, was designed to be more objective and broader in scope. In theory, it doesn't go into as great a depth as the MFE did but covers a much wider range of topics -- most of which, by necessity, do not fall within the student's research focus. The main goal was to reduce the subjectivity of the grading, and so the scoring is done using a double-blind methodology wherein the students do not know which professor wrote each question, and the professors do not know which answer set belongs to which student. Questions are written by ten different professors, each an expert in their field, and these same professors then score the answers to the questions that they wrote.
It's an interesting idea, but I don't like the way it works in practice. The professors that I've talked to all admit that they would not be able to pass the WQE if they were to take it, because it's simply too broad. Despite the fact that the test is not supposed to be particularly deep or advanced, if a student hasn't recently taken a class with the professor who writes a question it is likely that the student won't know whatever nuance of the field the professor decides to incorporate into his question. Additionally, no notes or books are allowed to be used during the exam, which is unrealistic in my mind because in real life almost anyone would use a book (or a computer!) to help them calculate network flows or to design arithmatic logic units. Except experts in the various fields, of course.
The end result is that each time the exam is given there is an approximately 50% pass rate. This is a very scary thing, let me tell you. Even though we have three chances to pass, it's nerve-wracking to have to study and study and study and know that if you fail you'll have to spend another six months studying again. To top it off, when you get your test back there isn't much feedback on what you did wrong, so it's not very easy to correct yourself.
Since the exams have only been given in this format four times, no one has failed three times yet and been kicked out (as is supposed to happen, based on the information I have seen). Several people have taken it twice and failed and then just not taken it again because they're too scared. I've heard rumors that people won't actually be kicked out, but who knows? Who wants to be the first to test that hypothesis?
The grading of the MFE may have been subjective, but it seems to me that that's more in the spirit of what grad school should be about. Computer science is an incredibly broad field, and I think it's unreasonable to expect any one person to have expert knowledge of every area within it; likewise, I'm not sure what purpose a broad, shallow test format serves, since that's basically what being an undergrad is all about. Graduate school should be about research and innovation, not endlessly poring over undergrad textbooks and trying to guess what professor is going to get picked to write which question.