31 July 2008 I am currently finishing off my PhD write-up, and hopefully will be submitting this academic year, and defiantly this calendar year. At this stage is when you look back and think what you did right and wrong, and here are my thoughts on advice I would give to someone starting a PhD tomorrow. For now I'll stick to stuff relevant to actually doing the PhD itself.. :)
Use a bibliography database
If there is one thing that I did right from day one of my PhD, it was to maintain a bibliography database. As an undergrad I did not bother maintaining bibliographies, as it was easier to keep a printout of everything and look back at it as needed. More importantly it seemed easier to maintain citation lists manually than to learn how to maintain them automatically. This is OK for the 1-2 months you might spend on an essay (or project report) that in the end only has a dozen or so citations, but a PhD thesis that has 50-100 citations (from a collection of perhaps 200+) and is written over a year or so is something quite different. Although bibliography software is intended for automatic citation list generation, the important thing is really keeping track of what you read and where.
EndNote was recommended to me and I used it for a while, in the end I settled with JabRef (with periodic exporting back to EndNote, as I tended to use Word for drafting stuff). One nice feature JabRef has is that it allows you to classify papers into arbitrary categories, in addition to storing comments. Of course, like a lot of people, I write comments on papers while reading them. Upon finishing the paper, these comments need to be written into a mini-review and added to your bibliography comments. Assuming the copy of the paper with the comments on has not been misplaced, it is a lot easier to glance over a self-written summary than to dig out the paper.
The value of a good bibliography system really shows up when writing up the thesis. Quite often I would vaguely remember reading something of relevance, but cannot remember where or when ('when' often being over a year ago). Selecting bibliography categories I think the paper was under and/or doing a search for keywords within my comments would often narrow down the list of publications, and in the worst case I would end up re-reading 2 or 3.
Use a notebook
Although using loose paper for notes allows for sorting at another time, I found that in practice it results in paper all over the place. This is a problem made worse as you will likely also have printouts everywhere, and in my case I often did work at locations other than at my research lab desk (flat, union, study rooms, etc). Unlike a taught course where you can at least file things according to subject, with a PhD you are unlikely to have a filing system from day one that will still be operating a few years later.
Write up ideas/notes into reports
Although it was originally because my industrial sponsor was supposed to receive reports at least every 3 months, writing stuff up as you go along means you constantly think about how your research is presented. The process of making reports helps filter out ambiguities that are present in notes, because the understanding of notes is highly dependent on what is going through your mind at the time.
Although most of the stuff you write in reports will never see the light of day (and looking back, some of the stuff will make you cringe), a lot of writing up will be rewriting of said reports rather than getting down information for the first time.
Follow interesting diversions
There will be times when the main thrust of your project seems to have stalled, and you find yourself branching off somewhere. Sometimes this may be due to boredom with some core aspect of the project, or because you are considering a "preresuquite" problem that has a solution that may make one of your original objectives irrelevant. See where it leads, as in the worst case it is more productive than having a 2-week lie-in.