A Ph.D. defense differs from a typical academic talk. A good academic talk primarily conveys technical insight. To be clear, a good Ph.D. defense also conveys technical insight, but a Ph.D. defense must also convey another critical point:
"The defender deserves a Ph.D."
A Ph.D. committee is looking for the answers to three questions:
Has the candidate conducted original, first-class scientific research?
Could the candidate conduct an independent research program?
Can the candidate persuasively communicate results of her work?
A good thesis proposal
A proper defense begins at the thesis proposal. A good thesis proposal makes a subsequent defense significantly easier. A poor thesis proposal delays graduation and reduces the odds of success. A good proposal forms a contract between student and committee: If I do X, Y and Z, you agree that this constitutes a Ph.D.
Check in with the committee
Students should check in with their committee members at least once a semester between proposal and defense. The committee should receive a copy of any new publications. If conditions arose during the thesis proposal, students should follow up with committee members to ensure that those conditions are being satisfied. The goal is to make the committee feel like the candidate is addressing their concerns, and to simultaneously convey a sense of activity and momentum to the candidate's research.
Structure the talk
In addition to teaching the technical aspects of the candidate's dissertation, the presentation for a defense should contain the following ingredients:
What is the problem I'm trying to solve? Why is this problem interesting or important? Why is my solution interesting?
A thesis is an active, declarative statement briefly but accurately summarizing the candidate's dent in human knowledge.
Exhibiting the contract established at the thesis proposal establishes the boundaries for success and establishes a natural flow to the defense.
Which of my peer-reviewed publications support my thesis? What theorems have I proved? What systems have I built? What experiments have I run? What evaluations have I performed? What case studies have I conducted?
Who else has attacked this problem or similar problems? In what ways is my solution better? In what ways is it limited?
While explaining the technical content of a dissertation, the candidate must impart a sense to the committee that she can convey the "intuition" behind complex ideas. Good illustrations, examples and animations impart this sense.
What will I or would I do next in this line of research? How can I build on the work I've done?
A practice run with the advisor is a must before the talk itself. A minimum of three more practice runs with lab mates or in the mirror will bring the talk together. The primary goal of practice runs is to debug the "flow" of the talk. A secondary goal is to sniff out likely questions from the committee. Put the answers to critical questions in the talk itself. Prepare back-up slides for the answers to other questions.
Most questions in a defense should be easy to answer. Answer them fully, honestly and politely. In an academic talk, one can often deflect a bruising or harassing question with a plea to "take the question offline" for later discussion. In a defense, there is no offline: almost all questions must be answered. The lone exception is if a question is off-topic because it is outside the scope of the contract established by the proposal. In this case, the defendant may insist that it would be a good basis for future investigation. If the proposal failed to establish a solid contract, there is no escape. If a committee member seems to be agitated while asking a question, it is critical to remain composed. Lashing out at a committee member is a proven way to tack on onerous conditions to graduation. If a committee member feels more work needs to be done, do not let them suggest what needs to be done. Ask if specific actions would be sufficient, and escalate until the committee member accepts.
A Ph.D. defense is an unusual trial: the goal is to prove that it has in fact been a trial by a jury of one's peers. As with any human endeavor, understanding psychology grants an edge to the defendant.
Dress like a professor
Dressing like a professor drops subconscious hints to the committee that one is no longer a mere graduate student. [If you're at a loss for what "dressing like a professor" means, hang out in humanities building for an afternoon to observe the faculty.]
Feed your committee
Psychologists studying will power have found relationships between the ability to make good decisions and blood glucose. Judges making parole decisions were found to be more likely to grant parole after eating. Even an assortment of small snacks could help at the margins--particularly when a defense drags into its second hour and the committee fatigues.
Watch your posture
Good posture conveys confidence. Poor posture conveys uncertainty.