Guest post: How to give a science talk
August 19, 2011 § Leave a comment
Andrew Murray recently wrote this absolutely brilliant piece on how to give a science talk for the benefit of local graduate students, and has kindly allowed me to reproduce it here.
1. You can never give too much introduction.
2. The introduction must include a defining question.
3. It’s very hard to show too little data.
4. Writing a talk out is the only way to be sure you’ve faced and solved all its difficult moments.
5. Tell them what you’ll tell them, tell it to them, and then told them what you told them
6. Never, ever exceed the time you’ve been allotted.
7. You must rigorously distinguish conclusion from inference from speculation.
8. Most people speculate too little rather than too much.
9. No representation without explanation.
10. Crappy, pilfered-from-the-web, diagrams produce crappy talks.
11. Be modest and generous.
12. Anticipate, encourage, understand, and answer questions.
13. There is one supreme edict. You get back what you put in: crap in, crap out.
Your success as a scientist will depend on how well you can present your work and the work of others. With practice, thought, and coaching, everyone can learn to communicate effectively. Above all, remember that your listeners have many things on their minds, including the fate of the paper they sent in a month ago, getting to daycare to pick up their kids, and whether the Sox will beat the Yankees. You need to keep their attention so that your talk triumphs over these other vital topics and you need to make sure that you can get their attention and comprehension back if they fall off the wagon for a couple of minutes. Failing at either means that much of your audience will lose the thread of your talk and drift off into their own affairs.
Like a good meal, there are four parts to a good talk. The appetizer is the introduction, which explains the problem that you want to study, why it matters, how it is related to the big questions that scientists want to answer, and the progress you have made in answering it. The main course explains how you constructed your experiments, computations, or simulations, what results you produced, and what you have concluded from them. Dessert is what you infer from your data and conclusions, and coffee is your speculations on how your work reveals the structure of your section of the scientific universe and where you think you and others should go next.
In his introduction to the reworking of Strunk’s Elements of Style, E. B. White said “Will [Strunk] felt that the reader was in serious trouble most of the time, a man floundering in a swamp, and that it was the duty of anyone attempting to write English to drain this swamp quickly and get this man up on dry ground, or at least throw him a rope.” But even without a rope, a reader can go back and reread opaque sentences. A listener cannot rewind you. Lose them for more than a minute and the quicksand will pull their attention irretrievably away from you and the science you are explaining.
The tips below should help you avoid the most common disasters. If you prefer watching to reading, Uri Alon can be seen talkin’ bout talkin’ (http://www.youtube.com/watch?v=5OFAhBw0OXs and parts 2 thru 6!)
Introduce, introduce, introduce
Remember that talk by the guy who thought that any sane scientist knew about the exciting challenge of distinguishing between nucleophilic attack and nuclear explosion in the mechanism of bubblyomuctase, the enzyme that dominates his every hour? Sadly, he forgot to introduce the enzyme, the competing mechanisms, and any reason why you should care, leading to one of the nicest fifty minute naps that you’ve had in years.
Don’t be like this! Remember how little you once knew or cared about your current project and help your audience up the gangway of the SS My Supremely Interesting Research. Tell them the overall question that motivates you, the history of the problem, and how your approach is designed to produce new and fascinating insights. Don’t be afraid to spend twenty minutes of a fifty minute talk on the introduction, as long as you are giving information that your listeners will need to understand and appreciate your science.
No data flood
A paper is an impersonal and permanent object. Without being able to respond to their questions, you need to convince skeptical reviewers and readers that you’ve done all the controls, considered every possible interpretation, and referenced every soul who has ever considered the problem. A talk is interactive. You need to show enough data to convince your audience that you know how to do an experiment, that your data are clean, and that you’ve dealt with one or two major caveats to your preferred interpretations, but no more.
This means that you probably can’t use the figures from your soon-to-be-famous paper. Left untouched, the figures contain too much information, like controls that will take too long to explain, and are too compressed and too poorly labeled to be absorbed on the fly. If they are presented as snippets that still contain the tell-tale A, B, C, and D that denoted their individual segments in the paper, many of the same problems still exist, and you are advertising that you think so little of your audience that you couldn’t even be bothered to modify the figure for the talk.
Write and practice
When distinguished faculty give talks at meetings, some are clear and some are muddy, but almost all end in overtime. Why? Because they didn’t write their talks out and they didn’t practice them. When you are old and gray you can be just as shoddy, but until then you must be clear, compelling, and sure that you will finish before people start shuffling and staring at their watches.
Writing it out word for word is the only way that you will know that you have solved every awkward transition, explained opaque nomenclature, and made the logic of your experiments and your analysis clear. When you read and consider your own words, you will see that some of the problems aren’t solved and you’ll have to go back and try again. Counting words tells you how long your talk will be. Anyone can make a reliable connection between the number of words they’ve written and how long the talk will take. Once you’ve made this calibration, you will see that your talk is too long and you will have to decide what to throw out and what to keep and simplify.
Practice helps. However cleverly you think you’ve introduced, simplified, and charmed, a single practice in front of a mix of experts and innocents will reveal that there’s still work to do. You can ignore some of the advice that you get, but you have to get advice, listen to it carefully, and have an explanation for any resistance you’re putting up. Practicing in front of people I know and admire is one of the most painful things that I’ve ever done, but the talks that were practiced were good, and the off-the-cuff ones weren’t.
Writing a talk doesn’t mean that you should read it from the podium or memorize it word for word. But it does mean that you know what you plan to say, how to connect the different parts of the talk, and how to avoid the “Oh, I thought the next slide was a Western blot, not this crazy micrograph!” moment. When I coach student presentations and the talk is horrid, the answer to “Did you write it out?” is almost always “I meant to, but I ran out of time.” I don’t buy that excuse and your audience won’t either.
Repetition is OK
Modest repetition helps your audience remember and understand your message. Giving them an outline of the talk at the beginning helps them anticipate what is to come; so do transitional slides between each distinct section of your talk. You can overdo this and come off sounding like a school teacher instructing a set of particularly dense and wayward pupils, but it’s a commoner and a worse sin to give the audience no road map at all.
Conclusion ≠ Inference ≠ Speculation
The action of your mind on your evidence produces conclusions, inferences, and speculations. A conclusion is something that your evidence argues cannot be denied, such as “After I delete this gene, my cells are dead.” An inference is the most likely explanation of the facts based on history, your new evidence, and Ockham’s five-cent razor. A speculation is your opinion about how the universe would work if you’d been its grand designer. I strongly believe there isn’t nearly enough speculation, and that when there is, the speculators often don’t distinguish between conclusion, inference, and speculation. If you don’t make the distinction, your audience will assume you don’t know it. Avoid both errors.
Pictures, pictures, more pictures, and words
Long before we were verbal, both in ontogeny and phylogeny, we were visual. Listen to yourself and you will hear the metaphors and similes that produce pictures in the minds of your listeners. Great diagrams make an inherently complicated subject seem easy, and awful ones can make the simplest subject incomprehensible. As you listen to talks, critique each slide you see. Good slides are elegant, simple, titled, modest, and produce a linear sequence of connected ideas. Bad slides are ugly, crowded, untitled, boastful, and shown in an apparently random order.
Keep the number of slides under control. A simple, well-designed slide takes two minutes to explain, which means that you can only use 7 slides if you have a 15 minute talk.
Your words and images have to fit together. This may seem so obvious as to be insulting, but watch talks carefully, and you’ll see that it isn’t. Unless you practice in front of your slides, you won’t know how to integrate the flow of words and images so that they reinforce rather than clash with each other.
A call to revolt: no representation without explanation
Watch carefully, and you will find that fewer than one speaker in ten explains everything that appears in every image they show. Everything means every symbol, every axis, every line on a graph, every lane in every gel, outlining at least one cell in every micrograph, and on and on and on. You and I may think we’re not guilty of this sin, but indulge in the infinite penance of watching yourself giving a videotaped talk and you’ll know that you’re guilty too. I have a colleague who is a recent and extremely successful immigrant from another science. He perennially and politely asks “What’s that circle” and “What does that line mean?” and thus serves as a fantastic example of how to educate others by being unafraid to ask questions. I tell students that they have to imagine Professor X listening to their talk and make sure that their combination of slides and speech are completely free of representation without explanation.
Don’t pilfer the Internet
The second commandment states “Thou shalt not make unto thee any graven image” and was carved into stone tablets by God’s fingers to warn Moses’ followers of the dangers of lifting images from the Internet and slapping them into their talks. Occasionally, days of diligent search will reveal that someone else has made just the introductory slide you need, but 99% of the ones I see aren’t the right slide for the talk I’m listening to: they contain extraneous detail, they lack critical features, and all too often they show both defects at once. To avoid these problems, you’ll have to learn how to scan in your own drawings, use a simple drawing program (like the ones in Powerpoint, Keynote, or Open Office), or sell your soul to master the power, the glory, and the impenetrable “help” offered by Illustrator and Photoshop.
If you do use other people’s images, make sure you acknowledge them and the source of the image.
Be modest and generous
You don’t want to overhear people saying “I can’t believe I heard him give another of those ‘How I invented enzymology’ talks.” But if you don’t mention the names of the shoulders you stand on and the contributions of your competitors, you’ll earn the label. If you have collaborators, they deserve both name and picture. If you enormously enjoy watching others try to hog the limelight, admire but don’t emulate. In science, unlike People, there is bad publicity.
Part of modesty is generosity. The rest is not overselling your science or yourself. Your goal is to open a window that lets your audience see a problem in a different light. If you sound like an infomercial, the sensible conclusion is that you’re trying to sell a point of view and listeners will infer that everything you say should be subjected to independent verification. If you end up being the subject of the talk because of your costume, your bombast, or your jokes, people will remember you, not your science. This doesn’t mean that you can’t be charming, make jokes, or reveal that you have an iconoclastic soul, but remember that you are educating, not speed-dating.
Obey two unwritten rules. Never talk about anything you won’t answer questions on, even if they are from your stiffest competitor. If the answer to a question is “I can’t say, we’re applying for a patent,” or “My collaborator won’t let me discuss her results,” I get up and leave the room. Don’t show quotes from reviews of your papers where history reveals that your claims were reasonable and the reviewers erred. In your career, you too will misjudge papers that you review.
Encourage and answer questions
You can ruin a good talk by handling questions poorly and you can rescue a mediocre one by answering them well. The worst you can do is demean questioners by cutting them off in mid-sentence, and giving long, aggressively pompous, defensive answers to a question that is different from the one they were trying to ask. Perfection means that you begin your talk by encouraging your audience to interrupt you, let every questioner finish their question, rephrase it if you have the slightest doubt that you understood it, give an answer that is as clear and as short as completeness will allow, and gracefully acknowledge any logical flaws, missing controls, or other defects that your audience has found. Many of my talks have been rescued by a perceptive question that revealed I had utterly failed to communicate a key concept and gave me a chance to try again. You can never anticipate every question, but trying will reveal correctable flaws in your talk and give you a sporting chance of looking thoughtful and open-minded when your audience shows that they care enough to inquire about and challenge your presentation.
GIGO (Garbage in, garbage out)
In public speaking, you get back what you put in. The person who gave the last stunning, effortless talk you saw probably spent days thinking of how to tell their story, wrote it out at least once, and practiced it many times. And the last person who made you wonder whether prayer could make their talk or your life end now most likely did none of these things. There are a very few people who were born with silver tongues and the ability to invent clear and compelling explanations on the fly and a very few who find it almost impossible to utter a single clear declarative sentence. The rest of us lie somewhere in the cozy middle, meaning that we can give hellaciously poor talks if we choose not to bother and fantastic ones when we do. Complain all you want that, unlike subject X, biology has been mesmerized by form and forgotten substance, but people will and should judge you by your presentations. Don’t set foot on the podium until you’re sure that your audience will give the right responses to these questions:
Did you tell them the question you are asking?
Did you explain what others have done to set the stage?
Have you explained how you got your data and what you can conclude from them?
Have you convinced them that you can think critically about your own work?
Do you successfully distinguish between conclusion, inference, and speculation?
Did you point out the new questions your work raises?
Did they prefer the whole experience to gouging their own eyes out with a fork?
A note on journal clubs
Everything that I’ve said applies to journal clubs with a single exception and several amplifications. The exception is that you’re talking about someone else’s work rather than your own. This means that giving a good introduction is even more important than it is in your own talks. The amplifications are
i) even less is even more, meaning reduce the detail and complexity of the paper as much as you can;
ii) explaining the design of the experiments is even more critical;
iii) you will have to redraw and simplify the figures that describe the outcome of experiments;
iv) you should opine on which of the authors’s conclusions, inferences, and speculations you buy; and
v) you should say what you would do next.