Candid Engineer is blogging about giving a 20 minute research presentation, and it reminded me that I posted a very detailed set of suggestions about this at the old DrugMonkey blog. Since I'm a lazy fuck, here is a repost.
I have recently been helping one of my trainees with a short seminar presentation, and last week I was at a subfield conference that featured many short seminars delivered by trainees and PIs. Here are some of my thoughts concerning the proper design and delivery of a short seminar (15-20 minutes). Some of these ideas are also applicable to long seminars.
The rest is inside the crack.
(1) It is much harder to deliver an effective 15-20 minute "short" seminar than it is a 45-60 minute "long" one. With a short seminar there is simply no room for error, no opportunity to recover from a mistake, no leeway for poor time management. In a typical long seminar, you can invite the audience to chime in with questions as they arise, and thus any confusion you sow can be dispelled in real time. In a typical short seminar, no questions are entertained until the end, and any confusion cannot be dispelled until after the seminar is over.
A long seminar can be constructed in independent modules, so that if you go slower than expected, you can simply cut out one or more modules of information, and the seminar is still a coherent whole. A short seminar delivered by a trainee is typically a single module, although PIs with advanced short seminar skillz can deliver two modules, if at least one of them is very schematic and light on data.
(2) However much data you think you should present in a short seminar, you are almost certainly mistaken. Inexperienced or poorly trained presenters always include way too much data. The question to ask yourself is, how much data can you expect a person not already familiar with to digest in 15-20 minutes? Here's a hint: It's not the amount of data in a Journal of Neuroscience paper with 10 figures and 5 supplemental figures.
Even someone with a quicker than average mind can only process in 15-20 minutes the quantity of data that would typically be presented in two or three typical Journal of Neuroscience multipanel figures. If you try to make ten different points, your audience will remember none of them. If you try to make two or three points, there is a chance your audience will remember one or two.
(3) There is no such thing as too much background and context. If you don't provide at the beginning of the talk a sufficiently rich theoretical context for the data, your audience will not be able to follow your explanation of your data because it will not fit into any conceptual framework in their minds. Give them this conceptual framework. And if you don't provide a sufficiently rich motivational context-i.e., why your audience should care about what you are saying-your audience is not going to emotionally engage and make an effort to listen to you.
Of course, you should be sensitive to what background and context your audience can be expected to already possess. If you are speaking at the Society for Nephrology meeting, you probably don't have to explain what a glomerulus is or convince them that its study is important for understanding kidney function. However, the downsides of providing insufficient background and context far outweigh the risk of insulting your audience's intelligence.
(4) The overall structure of a short seminar should be like two pyramids tip-to-tip: start very broad with background and context, narrow in gradually to the specific questions you address experimentally, and then broaden back out at the end to the implications of your experimental results to the big picture. In a short seminar, you can only enact this structure once; in a long seminar, you can repeat it two or three times.
(5) However, fast you naturally talk, it is almost certainly too fast for a seminar (short or long). It is counterproductive to try to cram more information into a talk by speaking quickly. Talk very slowly and enunciate clearly; your audience will thank you, especially those who are not native English speakers. Any idea you want the audience to remember needs to be repeated at least twice, and preferably more. Repeat conclusions and assumptions that flow from what you have already said as part of your transitions into what you are going to say next. If you do this smoothly, by the end of the talk you will have repeated your key points quite a few times, yet you will not sound repetitive.
(6) Each slide should have only as much data as would go into a single panel of a multi-panel figure. For bar and line graphs, use color to distinguish experimental groups, not hatching, shades of gray, symbol shapes, or line dot-dashes. Don't use colors that look like an infant's Playskool toys. There should be no extraneous information on a slide; if you don't explicitly refer to it, it shouldn't be there. Use animation only very rarely, and only if necessary to make a point. (I once attended a talk on circadian rhythms, and every slide had a little clock at the top left of each slide with animated hands whirling around. Do I need to explain how stupefyingly horrible and distracting that was? Let's put it this way, I don't even remember who gave the talk or what it was about, but I can still picture the infernal whirling clocks.)
State each slide's essential point in large text at the top. Use only a single sans serif typeface for all the slides in your talk, and make sure all text and labels are plenty big. Myriad and Syntax are good humanist sans serif typefaces, while Arial and Helvetica are too "cold"; no Times New Roman nor, for the sake of all that is good in the world, comic sans. Use a pure white background and black text for all slides, not a dark background and light text. Gradient background are just a distraction. Use all-text slides *very* sparingly, and only use bullet points; no full sentences or paragraphs.
(7) Use the laser pointer very sparingly, and only when necessary. A well-designed slide should not require much pointing, as the information should naturally flow from top-left to bottom-right. DO NOT WAVE THE LASER POINTER AROUND LIKE A MANIAC! Try to look at the audience, not the screen. If you are in a multi-talk session, try to insist that you use your own laptop rather than any integrated system; this greatly reduces the risk of problems with advancing slides.
(8) The question-and-answer period at the end is as much dominance-submission theater as it is an opportunity for clarifying and amplifying points made in your talk. You must make sure that you appear the dominant authority and the questioners appear submissive to your authority. This requires both deep knowledge and practice at expressing your knowledge with a demeanor of absolute confidence.
If a question is framed in a way that explicitly or implicitly calls into doubt any aspect of your interpretations, feel free to appropriately reframe the question before answering it. If you are not clear on what a question is really getting at, request clarification before answering. If you still only have a vague idea of the question, you should say, "I think what you're asking is blah, blah, blah", and then answer that question. If after requesting clarification you still have absolutely no clue what the question is, simply state, "I'm sorry but still don't understand your question, why don't you ask me afterwards", and move on. It is very annoying to an audience to sit there while a speaker answers a question that the audience knows wasn't asked.
(9) In terms of general verbal demeanor, DO NOT APOLOGIZE OR BE SELF-DEPRECATING UNDER ANY CIRCUMSTANCES, EVEN IN AN ATTEMPT AT HUMOR. Recite your experimental findings in the present tense as if they are simply facts about the world. "As you can see here, animals with knock-out of the bibbly-gibbly gene exhibit a severe deficit in gabbly-babbly behavior." Don't use phrases like, "we observe(d)", "we see/saw", "we were able to show", "we find/found" etc.
In the introduction of the the talk, don't say things like, "I am going to try to convince you of...", "I want to describe our efforts to...", or "I'm going to present evidence in support of our hypothesis that..." Say things like "We have tested the hypothesis that..." or "We have developed a new technique for..." In the conclusion of your talk, recite your main conclusions simply and directly. Don't say things like, "I hope I have convinced you..."
(10) Ultimately, your audience is going to remember a lot more about *how* you gave your talk than *what* you said. If they really become interested in your science, then they will look up your papers, e-mail you to discuss things further, etc. So the most important thing is to come across as very smooth, adept, and authoritative. *You* are the expert on what you are talking about, and *you* are in the naturally dominant position in the speaker/audience relationship. You should exploit that dominance to create a powerful impression in the minds of your audience. These are the people who review your papers and grants, decide whether to give you jobs, promotion, and tenure, invite seminar speakers to their departments and to present at conferences, and vote on officer and committee positions in scientific societies.