January 20, 2010
Simple schematic for a good talk
My colleague John King sent me the diagram below. It is an easy-to-remember schematic for how to organize a talk. Start with some context (we often say "the motivation"), drilling down until you get to your specific problem, then describe in detail your contribution in solving this problem. Finish by heading back up to the surface: return to the context, reminding your audience of the connection between your problem and the broader context, this time adding the implications of what you have discovered for the broader context. This would naturally lead into a description of your ongoing research programme on this topic, if you have one. (I advocate *against* a "future work" slide unless you are actually committed too, and probably already engaged in, the future work.)
I like this, but, at least for a seminar talk (50-80 minutes rather than the 12-18 typical at many conferences), I would precede the big donut with a smaller semi-donut, which is the left half of the big donut in miniature. That is, spend the first minute (and maybe one slide) quickly setting the scene (the context), then briefly but clearly introduce the problem you tackle, state your results (without demonstrating or supporting them, then sum up your contribution. After that brief intro, get with the more detailed context - contribution - context program.
This is very similar to the classic storytelling exposition I've elsewhere advocated. As in a typical (not post-modern) novel or movie, immediately set the scene, identify the protagonist and the conflict he or she faces. In stories the resolution (results) often aren't broadcast up front (though they may be foreshadowed), to build suspense. Sometimes that kind of suspense-building works in a scholarly talk, but often giving a sneak peak at the results is helpful / necessary in motivating the audience to be interested in what comes.
Posted by jmm at January 20, 2010 06:13 PM