Disclaimer: I'm not familiar with the format of departmental exams given in other curricular groups, at other departments, at other universities, or even in different years, so these thoughts really only apply to that of the BO departmental given at SIO in 2009. However, I suspect, given the strong "tradition-esque" property of departmentals, that it is conducted in a similar manner to single-session oral exams elsewhere (and elsewhen).
My main issue with the process of departmental exams is the opaque process in which they are conducted. First, the issue of which members comprise the departmental committee varies from year to year. I believe it typically includes five members, one physical, one chemical, and three biological. At least one of each is one of the professors who taught the corresponding introductory course that year. However, replacements occasionally occur when certain members are unavailable. The remaining two biological members are drawn from the faculty, though it is unclear to me how they are chosen.
Once the committee members and time for the exam are decided, first-years are simply told to show up. No information is given about what will be asked, what the committee is looking for, etc. Yes, individuals can contact the committee members and ask questions about what they are likely to be asked about, and the committee members will likely respond truthfully, but it's still a very mysterious process.
Aside from the fact that the committee expects you to know *something* about oceanography, it's unclear what the expectations are. It seems that they take into account things like what classes you took, how large your courseload was, what research you did, your background, etc. I suspect the departmental exam is thought of as a test of your ability/qualifications to be a Ph.D student. My personal opinion is that this is an extremely subjective decision and that each committee member probably has different views on what qualities are most important.
I also think that a departmental exam isn't necessarily the best (i.e. accurate and/or fast) way to determine success of a Ph.D. student. It's not like you can track the students that don't pass the departmental exam to see how well they do in graduate school. There's probably an overly strong emphasis on academia success, which doesn't seem like a good idea, either. Given current numbers of graduating students and the number of positions available, it's clear that not all students will end up in (or even want to be in) academia. Success in academia depends on somewhat different factors than success in other fields for which a degree would be useful.
I think what really irks me is that the subjective/opaqueness of the committee's decision means that it's really hard to practice for. Practice committees with students probably does a close job of approximating the situation, but it's hard to say how useful any feedback will be. The feedback you get from passing (or not passing) the exam is usually along the lines of 1. take class x, 2. study more on subject y, which focuses just on the knowledge portion of the exam.
Finally, there's a huge element of luck. 90 minutes seems like a long time, but there are five committee members, who usually take 15-18 minutes each to ask questions, with some slack for follow-up questions and for the committee to take a break between examinees. Because the committee can ask virtually anything, it's likely that you won't know the answer to every question. (One aspect of the exam is probably to see how you respond when you don't know the answer and try to bootstrap from what you do know, make connections from related knowledge, etc.) However, it's also possible that the committee stumbles upon your area(s) of expertise. I'm not sure what they do in such a situation, but obviously you're at an advantage for passing if you know the answers to all the questions.
