Unsolicited Advice XI: How to Write a 5 Minute Talk

In the American astronomical calendar, early January looms large. Tis the season of the winter meeting of the American Astronomical Society (the AAS). Thousands of astronomers are currently gathering up their rain gear, preparing to descend on Seattle next week. A significant fraction of these thousands are currently preparing posters and talks for the meeting. Unfortunately, AAS presentations are almost entirely 5 minute talks, in contrast to the more typical 50 minute colloquia on which we've all cut our teeth. This short format presents a real challenge, and requires a very different approach. So, in CV's long running Unsolicited Advice series, I now give my basic guidelines for writing a short talk. 1. In a 5 minute talk, you can usually only teach people about one new thing. The key to framing the talk is to figure out that One New Thing, and then build the rest of the talk around it. 2. So, the very first step is to pick the absolute best single visual to show the One New Thing. This is usually an awesome plot. 3. Make this plot the first slide you prepare. Don't start with history and motivation and data acquisition and analysis techniques -- if you do, you'll prepare way too much material, and find yourself at 4 minutes and 59 seconds without having actually gotten to your results. Instead, generate the one or two slides that show the One New Thing, and explain its meaning. (If it helps, think of this as "How would I explain my results, if someone walked into my office and asked what I was working on?" In all likelihood, you'd pull out the coolest plot, and start explaining it.) If there are multiple features to explain about the awesome plot, consider repeating the plot on multiple slides, and explaining one feature of the plot on each slide. 4. Now, work backwards. What are the essential facts needed to understand the plot and appreciate its awesomeness? Operationally, this usually means one or two slides of motivation (covering why this is an interesting topic, and why there are questions that still need to be answered) and one or two slides describing the data and/or analysis (covering just enough that people understand the outline of what you've done, and accept the validity of your approach). 5. When presenting these essential facts, you have to return to the point in Step 1. Since you can't teach people more than one new thing in a 5 minute talk, your motivation/data/analysis slides should function more as "reminders" of what people already know. Your audience has read a lot of papers, have done lots of observations, have written many analysis codes, and have seen tons of colloquia -- since most will have this background, all you really need to do is trigger the right memories. Note that this is different than assuming your audience are experts on your particular project -- you still want to walk them through the essential facts from Step 4. However, you don't need to explain what the VLA is, nor do you need to explain anything that would be routinely covered in beginning graduate coursework. Reminding people? Yes. Explaining details of routine astronomical tools? No. 6. Add a conclusion slide to the end of your talk, ideally with two or fewer take home messages. Do not recap what you did. If you can, reproduce the plot for your One New Thing. 7. Now, go back through your whole talk, and eliminate every single word you possibly can. Turn sentences into phrases (i.e. "We took 2000 DEIMOS spectra with 8 hour exposure times in Spring 2008" becomes "2000 DEIMOS spectra"). Eliminate any details that aren't essential facts. You simply don't have time to effectively convey nuance and complexity, so you need to strip such frippery. You are trying to keep your audience on track, and do not want them distracted by unnecessary detail. Rule of thumb: If you have to use smaller than 48 pt font, you've got too many words. 8. Take yet another pass through your talk, working on making the visuals as clear as possible. Use Powerpoint/Keynote to generate new >48 pt axis labels for all plots. Add arrows and text boxes to point out key features in plots. Make sure all point symbols are clearly explained, and can be easily distinguished from 50 feet away (i.e. open and closed symbols of contrasting colors are better than X's and crosses). Eliminate all yellow or orange symbols from white backgrounds, since no one can ever see them when projected. If there are additional lines on your plot that you will not have time to talk about (i.e. from fits, or models), remake the plot without the lines, or hide them behind white lines. Adjust the image contrast and/or line thickness to make all lines and symbols beefier, and easier to see from far away. 9. Finally, practice your talk, ideally with a helpful non-expert colleague roped into watching. Find out what parts they didn't understand, and fix them. So, best wishes, and see you in Seattle!