orgtheory.net

Leslie Lamport, author of LaTeX, amongst many other things, provides a guide to How To Present A Talk. It was written in 1979, but modulo a couple of changes its advice applies equally well today. For instance:

WHAT TO SAY
1. Describe simple examples rather than general results. Try to make the examples much too simple — you will not succeed.

2. Don’t use formalism. If your results cannot be described simply and informally, then there is no reason why anyone should be interested in them.

3. It is better to be inaccurate than incomprehensible. The place for accuracy is in the paper. (However, false advertising is unethical.)

HOW TO SAY IT
1. Don’t put too much on a slide — a picture of a thousand words is worthless. …

3. A rapid sequence of slides has a hypnotic effect. Unless you are a licensed hypnotist, don’t use more than one slide per minute.

- Time your talk. Running over your allotted time is a mark of incompetence, and displaying your incompetence is a poor way to get someone to read your paper. Remember that talking to an audience takes longer than talking to a mirror.

He also provides the best, most direct advice ever given to people responsible for chairing a talk:

Be utterly ruthless about enforcing time limits. Warn the speaker when he has 10 minutes left and when he has 5 minutes left, and stop him in midsentence when his time is up. The audience will be grateful. (A loud alarm clock works quite well if you don’t turn it off until the speaker has finished talking.)

I can think of worse uses of my ASA dues than the purchase of large manual alarm clocks whose employment at conference sessions is mandated on threat of expulsion from the organization.