How to give a good scientific talk

Recently I got drawn into a short discussion on how best to give scientific talks. As a reaction to that discussion, the following is in part a note to myself and in part a piece of advice in case an early career scientist happens on this blog via a search engine and finds this useful. As always, while the following is based on my own experience as a career scientist speaking at various conferences from ca. 25 to more than 2,000 participants, please note that I am not a professional instructor for public speaking courses. In addition, the culture of speaking will vary from one area of research to the next, and what is appropriate in mine may or may not be appropriate in yours.

What we are talking about here are the typical self-submitted, non-keynote talks at scientific conferences. They generally last 12 min plus 3 min for questions or 15 min plus 5 min for questions, and the resources at your disposal are generally PowerPoint slides projected at the wall behind you, a laser pointer and a microphone. You may have a remote to advance the slides or you stand behind a lectern with a keyboard and press the arrow keys to change slides. So here is my advice.

Design of the presentation

• Prepare your presentation so that it will fit into the allotted time with enough time left for questions. There is nothing more obnoxious than a speaker who continues speaking although the session chair is urging them to finish. It is extremely inconsiderate towards all later speakers, and you will loose a tremendous amount of goodwill that way. While it is much more considerate to terminate your talk when the time runs out half way through, that will of course only make a slightly less bad impression. So again: Prepare your presentation so that it will fit into the allotted time. Period.
• The first step is to make sure that you are not trying to squeeze too much into it. Yes, you have done all these fancy analyses with your data and find them all very interesting, but perhaps it is not a good idea to attempt to explain four different methods and show four different results in the course of only twelve minutes. Ideally, take one simple question or hypothesis, one clear way to answer or test it, and visualize the results with one or two clear graphs. Perhaps mention that you have done more than just that, that this is part of a larger study, but that due to time constraints you will focus on just one aspect of your work.
• A good rule of thumb that works for me is that there should be approximately one slide per minute, but you will have to learn how that ratio is for you. Obviously it depends on the slides in question - some will be up only for a couple of seconds, such as a picture of your study organism, while results or methods slides may be up for significantly more than a minute - but for sufficiently large numbers of slides it evens out and such rules of thumb seem to work well.
• The less text on your slides, the better. Walls of text are overwhelming; if you are lucky the audience will read them instead of paying attention to you, and if not the audience will simply be annoyed. Best are ca. three bullet points per slide spelling out key concepts, technical terms or key results that are too important to only be mentioned once in your speech. Examples would be the questions or hypotheses, formal names of the methods, tests, markers or data sources, and P values, Fst values or other similarly crucial results. Always be as concise as possible, e.g. phrase a research question as "bias against spiny plants?" instead of "do field botanists collect spiny species less often than unarmed species?"
• As a side remark, there are those who argue that the ideal presentation has zero slides, and that the ideal slide has zero text. I consider that to be nonsense. Don't know in what area you are working, but it is literally inconceivable to give an intelligible presentation in my field of research without slides because there is always a graph or phylogenetic tree involved that people quite simply have to see to make sense of what you have done. Showing a picture of your study group at the beginning also increases the identification of the audience with the topic. And as for the text, having the aforementioned key concepts and terms written out can be very helpful especially to those not familiar with a particular method, and writing out the results of statistical tests makes sure that they don't get lost when somebody gets distracted by their neighbor at the second you mention them.
• Font size should be at least 18 points, better 20 points. Use simple sans-serif fonts like Arial or Calibri. Use different font colours and sizes sparingly, if at all. Space the text out.
• Use a simple slide template. Black font on white background is best, not least because that is what people are used to from reading books or screens anyway. Because many of your figures will be black on white, they will also look better in a PowerPoint file with a white background. Definitely avoid obnoxious templates like neon text on a black background or your audience will have a headache after the first five slides. And yes, I have seen scientific presentations like that.
• Use a consistent slide layout, e.g. bullet points always on the left half, graphs and photos always on the right half.
• A well-designed presentation does not need a laser pointer because you can use circles or arrows in PowerPoint and then simply say "as indicated by the arrow" or similar. Thus you avoid speaking to the screen and disruptive arm movements, and you will be unconcerned if the laser pointer at the venue has run out of power.
• On the other hand, have as few animations as possible and preferably none. They are in most cases superfluous, silly and annoying, and they take ages to click through later when you have to go back to a slide in the middle to answer a question after the talk. What is more, they are the first thing to go wrong when there are compatibility issues between your computer and the one at the conference.
• Related to the previous observation, it is best to save the presentation not only as a PowerPoint file but also as a PDF, just in case there are serious compatibility issues, and, well, PDFs don't have animations.

Preparations for the talk

• Before the conference, give a practice talk in front of a small number of friends and colleagues, preferably including your supervisor if applicable to your situation, and ask for honest feedback. Do not blindly follow all advice, of course, but comments like "this is hard to read" or "maybe you should mention this before that" are really valuable.
• If you find during such a practice talk that you needed more than the available time, throw things out. Make your talk shorter. Seriously, rushing through it at higher speed is not an option.
• If, on the other hand, you find that you have two or three minutes more than you thought, don't change anything; be happy that you can talk a bit slower and that you don't have to be nervous about running out of time.
• Practice your talk so that you know what slides/topics come after each other. The point is not to memorize all your talk by heart because that would come across as too artificial and stilted. It is better to only know what things to mention on each slide and then to present them freely, so what you should memorize are a few keywords for each slide. That being said, especially as a beginner I always had one or two slides per talk where I tended to fumble the transition. In that case, it can be useful to memorize the first sentence for that slide; once you have delivered it, you are back in the flow and can speak freely again.
• At the conference itself, and unless you are the first speaker on the first day, you will have the opportunity to observe other speakers. Use it. Take note of the arrangement of the lectern relative to projection wall and audience, of the availability of screen and keyboard on the lectern, whether remote controls and/or laser pointers are supplied, how far away from the microphone a speaker needs to be for good sound quality, and any problems previous speakers may encounter with those arrangements and pieces of equipment. I have often wondered about colleagues who speak at the third day of a conference and are still tripped up by one of those details, as if they had been asleep during all previous talks.
• Among scientists, wearing a good shirt and clean trousers or equivalent dress is usually adequate, so don't over-think your appearance. Unless you give the talk as part of a job application, of course.

Giving the talk

• During the first two sentences, note whether you are too close to or too far away from the microphone. If you hear a loud blasting noise coming out of the speakers every time you exhale, it is the former; if you cannot really hear your voice coming out of the speakers at all, it is the latter.
• Speak loud and clear, enunciate and articulate well. That means that you should speak reasonably slow; force yourself to take breaths and make pauses between sentences instead of breathlessly rushing through your talk.
• Speak to the audience, not to the wall. A good idea is to look somewhat unfocusedly into the upper third of the audience while you are speaking. This has the advantage of giving most of the audience the impression that you are looking at them, but by not focusing your gaze you avoid actually reading somebody's face and becoming stressed or distracted because they happen to frown at that point. I find it more useful and less distracting to "read" the audience by paying attention to the sounds emanating from it than by watching the mimics of its individual members.
• If the last slide consists of acknowledgements, I would strongly suggest not to read it all out. There are few things more annoying than having a speaker waste half the time for questions by reading out the names of fifteen people and three funding bodies and mentioning all their contributions in loving detail. Better focus only on very few really important contributors and leave the less important ones for reading, e.g. "finally, I would like to thank all the people listed here who have made this work possible, especially my PhD student Jane Doe and our collaborator Joe Average who provided many indispensable samples, and of course the XSF for funding. And I thank all of you for your attention."
• When answering a question after the talk, make sure that you do not just have a dialog with the person who asked it. If you have reason to suspect that the rest of the audience was unable to hear the question, repeat it into the microphone before answering. Do not be ashamed to admit that you cannot answer a question - a reply on the lines of "this was beyond the scope of our study" is much better than inventing some evasive nonsense that patently does not address the question at all.