Tuesday, July 19, 2011

"Defending" a PhD

I had the pleasure today of attending a stellar PhD defense by a student who did a rotation in my lab back in spring, 2006. That rotation involved several months of really intense work, and resulted in authorship on 3 really good papers. I loved watching that student present the outstanding product of his dissertation research over the past several years, even though he chose a different lab than mine in which to do that dissertation research.

But it got me thinking- why and how do we "defend" PhDs? Could we make the process better? Where did the word "defend" even come from-- who is "attacking"? Presumably, a main point of the defense is to convince an audience and a committee that the work was worthwhile-- both tackling important questions and coming up with compelling answers. Defenses typically have the "seminar" followed by the "closed session" with a dissertation committee, and it is the latter that actually decides if the student passes.

In cases like today's defense, the "closed session" that followed the public talk was not especially helpful. As in many good defenses, the committee felt obliged to ask questions and sound critical of tiny nuances even though they all thought the product was excellent. The student, meanwhile, is exhausted after the public defense and has to suffer through this "hoop". Perhaps a few pearls of wisdom or insights come up from the committee, but many such pearls will be forgotten amidst the ensuing celebration.

Worse yet, what happens when a defense is less-than-stellar? What happens if, at the closed session, major gaps are identified in the work or in the student's understanding? I think the answer is "usually nothing." In principle, such gaps should have been found in the "last committee meeting before the defense", but it's VERY common for many facets of dissertation projects to be incomplete at that stage, or holes to not be totally apparent because the committee doesn't have a fully elaborated document. Similarly, we expect advisors to prevent such awkward situations, but this doesn't always happen. This is bad for the student (who has an incomplete product and may be less competitive in applications), it's bad for the advisor (who either accepts a less-than-ideal product or assigns someone else to complete it), it's bad for the university (who's 'seal-of-approval' is stamped on an incomplete product), etc.

Here again, like so many things in graduate education, we have procedural inertia that fails us.

What are the alternatives? Let's start by going over the problems. One problem we face is that we don't want to humiliate students by having them give a public "ending" talk and then not allow them to leave afterwards. Another is that there's little or no "check" on changes to the dissertation (or associated research) after a defense. Finally, we create a situation where the student has to simultaneously prepare for a major public seminar as well as an evaluatory defense from their committee.

What if we instead had the closed session first, long before the public seminar has been announced publicly? The student presents the committee with a full draft dissertation and presents the work orally in much-abbreviated form (instead of the typical "closed session"). If there are holes in the experiments or analysis, the student will have some time to correct them (see below). Based on the closed session, the student is given a "pass" or "provisional pass" or "fail." Assuming a "pass" or "provisional pass", at a date no less than 2-4 weeks after the closed session, the student presents the public talk. Just before the public talk, the student also gives the committee the revised dissertation, noting what was done in response to their concerns (which may also be covered in the public talk). The committee attends this public talk already knowing that the work is solid and at least provisionally approved, and they just ensure that their concerns are addressed either in the talk or in the revised document. At that point, the committee signs off on the dissertation.

This splits the defense into two temporally separated pieces, swapping the traditional order, and it avoids the awkwardness of public shame if a student actually doesn't produce a satisfactory product in their first attempt. It gives the committee the chance to approve the final product (which bears their signatures of approval). And it gives the student more time to work on the public talk vs. the written dissertation separately.

There are problems with this approach, too, but I'd love to hear people's thoughts and/ or other ideas.


  1. In my department, there is a "data meeting" before the actual defense, which in many ways covers much of what you're mentioning. If the committee isn't happy with the result of the data meeting, the defense isn't scheduled until the recommended changes/ additions are made.

    I've also heard of the role of an opponent in some European schools, so you are actually *defending* your thesis against an attacker. It is a point of pride among professors to be asked to be an opponent.

  2. There are departments (I think UCDavis EEB does this) where the defense is replaced by an exit seminar. This is similar to your plan, but I'm not sure whether they do a B-exam/closed-session before the public talk.