Saturday, February 13, 2010

A Ph.D without knowledge?

Sometimes I struggle with defining the minimum that someone should be able to do and know before they are allowed to graduate. Here is a story (possibly an urban legend) which might put things in perspective.
At Princeton after a student submits their Physics Ph.D thesis they are given an oral exam by a faculty sub-committee. After a short presentation on the thesis by the student they are asked questions about it for half an hour or so. Then the faculty can ask the student ANY question about ANY area of physics (e.g., why is the sky blue? why is diamond called "ice" by criminals? explain the twin paradox in special relativity.) I believe the purpose of this is to show students how much they do NOT know, particularly about basic physics.
Many years ago there was a very bright young student who completed a thesis on general relativity and was about to go and do a postdoc with Stephen Hawking. He was asked, "How big is an atom?" He did not know!
But, he got his Ph.D.
So, who was the student? Someone once told me it was Nathan Myhrvold, who for many years was Chief Technology Officer at Microsoft.


  1. The viva (oral exam) in the UK also serves a similar purpose. I was examined by Stephen Hayden & Andy Schofield. I learnt a lot from spending 3 hours talking with these two about my thesis. But one of the things I learnt a most clearly was how little I new.

    At one point they asked me if I could draw the phase diagram of the Hubbard model in 2D. I was quite proud when I gave, what I thought, was the "correct" answer - no, it's an unsolved problem! But Andy asked me what I _thought_ the answer was. I just stood there for so long "thinking" that Andy had to suggest that started with some axes! He then helped me work through the problem and taught me about some effects I didn't know about - e.g. Nagaoka ferromagnetism.

    It was tough and scary at the time. But, despite that it was a great experience. I learnt lots and it's helpful to have two such smart and knowledgeable people think about your work and interact with on in such detail.

    Thanks Steve & Andy!

    I really think the lack of a compulsory oral exam in Australia is to the detriment of the students.

    PS I passed also despite not being able to answer all the questions.

  2. I had two postdoc interviews recently, and was delighted to get to talk at length with other theorists about my phd work. I learnt heaps about my own work, and was exposed to challenges that just never came up while actually doing the work. This sort of experience makes me sad that we don't have a thesis defense in Australia. Brief end of year reviews are not really an adequate replacement for a robust defense (I would think).

    On the question of how much one should be expected to know, I was pleased to find that none of these two potential employers expected me to know even vaguely what their field was, nor even too much about the broader scope of my own field (within reason... whatever that means). However all seemed quite concerned with the motivations behind different calculations - perhaps a good measure of one's immersion in the field, and how conversant they are with the literature?

    This question of phd expectations is quite a common thread in this blog, and I always look forward to reading the next installment. For what it's worth I think someone who comes through the Australian system is at a disadvantage. The lack of graduate coursework (and robust undergrad coursework), and the possibility of getting away with doing extremely little variety of research is dangerous. Going from a relatively unknown to a dauntingly prestigious institution I'm feeling the pressure from my lack of education.

    Perhaps it's in my head though.

    While I've got the floor, thanks Ross for this blog. I enjoy it heaps. My approach to my own research has changed dramatically since reading it as well. In particular I appreciated your call to doing slow science, and was convicted of your challenge to learn the tools you need to solve a problem rather than using the tools you know exhaustively on all problems you can find (problem directed research I suppose). Both very simple, helpful tips that are vastly different to my recent approach :).