Principle 8 - Finish When You Are Done, Preferably Sooner

How long is a good science talk? I do, actually, have a number in my head. Actually, several numbers. Different numbers for scientists at different career stages. I'll reveal my numbers at the end of this post. First, some advice.

A good science talk is almost never longer than the time allotted for the talk. And, by almost never, I really do mean very close to never. In fact, going over time can magically turn a good talk into a bad talk. Back when I was a student at Cornell, I invited a scientist to speak at our natural history club about his work on whale communication. A day before the talk, the speaker asked me how much time we had allotted for the talk. This was the first person I had ever invited to give a talk and I made the fatal mistake of saying "As much time as you need", by which I meant, you know, "40-60 minutes." Well, after 2.5 hours I finally had to interrupt this prestigious scientist to call it quits. Of course, everybody hated this talk. The talk contained an enormous amount of fascinating information. My colleagues didn't hate the information, or even the presentation, they hated being held prisoner in that room for hours. Technically, the speaker didn't go over time, but he did speak for longer than everyone expected. This souring effect kicks into action remarkably quickly once you pass the expected quitting time. Going over by five minutes can leave a bad taste. Try to finish early.

Good talks tend to focus attention on only one or two major points. Perhaps one point for a short talk, maybe two points for a longer talk. Three points max! Your goal should be to communicate the arguments and evidence in support of these major points as clearly and concisely as possible. Everything in your talk should be in the service of these major points. Tangential work, peripheral points, and completely unrelated issues will only tend to confuse the audience. By limiting the number of topics, you should be able to keep your talk on schedule. 

As a general guide I would say aim to speak for about two thirds of your allotted time. If you succeed, your audience will be fully engaged and people will have lots of questions. The Q&A will likely eat up the remaining allotted time. By the way, the best evidence that you have given an unsuccessful (or even a boring) talk is that your audience has very few questions. Perhaps counter-intuitively, abundant questions are the best sign of a successful talk. Your audience was fully engaged and that is why they have questions. They don't have questions because they are thoroughly confused; they have questions either because they are a little confused or because they have thought of extensions to your work. Unfortunately, most people will not tell you if you have given a boring talk, or even a moderately unsuccessful talk. This is a shame because I have discovered that the vast majority of people want to give good talks, but they have not received honest feedback. So, in the absence of direct feedback, you can use the number and quality of questions as a rough gauge of the quality of your talk. 

Now for some numbers. I would say most junior folks (undergrad, grad student, early post doc) should be able to present a coherent, compelling talk on their work in 15 minutes. To put that in perspective, I will tell you a little secret. The talks at the HHMI Investigator meetings are fifteen minutes long with five minutes for Q&A. These are talks by some of the best scientists in the world, and they get only 15 minutes every two years to tell a compelling story to their colleagues. So, I don't think it is harsh to limit junior scientists to the same time span. As scientists become more senior, their science becomes more mature and it is appropriate in a regular seminar to listen to them for a bit longer. I think 40 minutes is probably optimal, with 20 minutes for Q&A. Longer than that can still work, but you better have explained everything very clearly and still have important compelling things to say toward the end. 

Unfortunately, we have all suffered through talks that eat up every minute of the allotted time, or even talks that extend beyond time. Sometimes this reflects poor organization by the speaker and sometimes it reflects excessive exuberance by the speaker to tell you everything they have ever done (the "take no prisoners" talk). In either case, they have missed the point of giving a talk. Tell a story well and well within your allotted time. The old saying is true: "Nobody complains if a talk finishes early."

Principle 7 - Tell and Show

The human voice is the best instrument for communicating complex information to other humans. So, your talk will go best if the humans in the audience actually hear what you say. I have so far given advice about how to help the audience focus on what you are saying, by removing words from slides and by using black slides to help the audience focus on your words. This new principle, "Tell and Show", provides some relatively subtle advice about how to guide the audience through your work. 

I commonly observe that speakers click on a slide and then proceed to walk the audience through the contents of the slide. For example, they might say something like "As you can see in this slide, the length of the wombat's nose is correlated with the the length of their toenails." It is usually more effective to do this in reverse. First, say "Because of everything I have just told you, we wondered if there might be a correlation between the wombat's nose and their toenails. So, we measured 50 wombat's and this is what we found." Hopefully you said this while there was no distracting slide in the background (use black slides). Click to the data slide, explain the axes and describe the result. "As you can see, we plotted wombat nose lengths on the X axis and wombat toenail lengths on the Y axis and we observed a remarkably strong correlation between these traits."

One obvious advantage of tell and show is that you end up telling the audience about each point twice. They hear the result once before they see the data and then again when they are looking at the data. This helps guide them toward the relevant observation in the data and increases the likelihood that they will remember the result after you move on to your next point. 

A more subtle advantage of tell and show is that it forces you to focus on one result at a time and devote one full slide to each result. This will help clean up the appearance of your slides. "Tell and show" also weakens the argument that you should put words on slides to remind the audience what the slide is about. With tell and show, you introduce the theme of the next data slide before you show the data, which is far more effective than putting words on slides, which end up distracting the audience from the data. 

Principle 6 - Practice, Practice, Practice

I have discovered that many scientists don’t practice their talks. Science may be the only profession where folks regularly stand up in front of a big group of people and discuss their work without having first practiced their presentation. This doesn’t make any sense to me. The members of your audience decided to take valuable time out of their day to come hear you speak. If there are a hundred people in the room, then they have invested a hundred people-hours to listen to you. Isn’t it worth a few hours of your time to practice your talk? Why would you waste their time with a talk that doesn’t get to the point or that does not clearly communicate the central message of your work? I have no idea where this culture of throwing together slides and “winging it” came from, but it is pretty obvious which science talks have been practiced and which have not.

People with different levels of experience should use practice time for different reasons. For young scientists, it is useful beyond measure to practice your talk in front of other people. Choose the audience for your practice talks carefully. Don’t invite too many people. A large audience can result in lots of advice, perhaps too much, and it can be difficult to parse out the useful suggestions from the gratuitous comments. Usually it is best to have just a few people who work in your area and a few from outside your area. I have found it useful to have an audience with a mix of experience, like one or two senior colleagues together with a few junior colleagues. In general, junior colleagues often help to identify points that aren’t clear in the talk, while senior colleagues often help to reorganize the talk or rephrase parts that are not clear, though some people are great at both!

There are many good reasons to practice in front of an audience. Obviously, this practice audience will give you feedback on how to improve the talk. Less obviously, you can tell during the practice talk what is working and what is not. Make sure you actually look at people in the audience while you practice. For one thing, this helps you learn how to talk to the audience like you are talking to people. In addition, you can often gauge from people’s facial expressions whether they are getting what you are saying or not and whether they are bored or engaged. Many people recommend video-recording your talk so that you can watch yourself give the talk, to learn what works and what doesn’t. I am sure this can be helpful, but I should emphasize that I am not interested in turning scientists into actors. I just want them to communicate their science clearly. Talks need not be super-smooth, elegant productions. I just want scientists to explain what they have discovered in clear, plain language. 

I think the importance of practicing is already appreciated by the younger generation and it is now common practice for young scientists to practice their talks. I think what is less well appreciated is the fact that it is most useful to practice early and practice often. Practicing early gives you the opportunity to seriously rework a talk, should that be necessary. Often, it is not entirely clear what is wrong with a talk and it can take a few days or a few weeks to find a better way to tell your story. If you practice only a few days before your presentation, then you have foresworn the opportunity to seriously revise your talk. 

Practicing often provides many benefits. I practice my talks. Then I practice again. Then I practice some more. For new talks, I usually practice at least three times. One reason I practice so much is that I try very hard to avoid jargon in my talks. I don’t always succeed at killing all the jargon, but that is my goal. Explaining complex science in plain language is just plain hard! It takes practice to find plain language that is both accurate and concise to explain your science. I find that this effort is paid back tenfold by the fact that most of the audience seems to understand my talks, which I know because many people ask sophisticated questions about the work.

As you become more experienced in giving talks, you can reduce the ratio of practice talks in front of people to practice talks on your own. Sometimes I practice in front of the folks in my lab, but usually I just practice in an empty room. Practice talks in front of people, and in front of cameras, are great ways to learn to avoid the worst habits (like talking to the floor). I learned to avoid these bad habits long ago and I no longer practice to improve my performance or to appear “smooth” in my presentation. Instead, I practice to hear the flow of the story. 

What do I mean by this? For one thing, words spoken out loud can sound different from words written on a page. So, sometimes, the logic of your scientific paper will not convert nicely into a story for a talk. Second, papers often contain more details than are required to communicate the core ideas of your work. I always strongly recommend that folks cut most of the supporting information from their talks and focus on presenting the critical data and experiments that support their conclusions. Remember that your audience will absorb and remember far less information from your talk than you carry in your head. If you can communicate one or two really critical ideas, then you will have succeeded. Everything in your talk should be in the service of communicating those critical ideas. Scientists routinely over-estimate what members of the audience know and understand, causing many presenters to take shortcuts through the material. Presenting complex work to a live audience should not be the first time that you have tried to explain this complexity with words. If you practice, and practice again, you are more likely to find a flow of words that communicates the big messages of your talk without getting stuck in the weeds. 

A Talk Is Not A Paper: The Structure Of An Effective Talk

A verbal presentation of science is a fundamentally different beast than a scientific paper. Many of the things that make talks boring are elements that are transferred directly from the written papers. It is extraordinarily tempting, I know, to give a talk in which you recapitulate every experiment that you have done, with all of the controls, to convince your audience that the exciting results you are presenting represent a fundamental advance in the field. Certainly, the aficionados in the audience will appreciate your rigor. But, of course, they probably won't entirely believe you until they read the paper anyway. The big problem is that dragging the audience through the thicket of your experiments will very likely put most of your audience to sleep. To engage the widest possible audience, you must make a tradeoff between breadth and depth of presentation. Usually, breadth is better than depth. The aficionados will understand that you are trying to reach a broad audience and, because you have obviously energized the room, they will look forward even more to reading your paper. The non-aficionados will be thrilled to learn about something so interesting that they had not heard before. You will have begun their engagement in a conversation on the topic that will continue in the question-and-answer session and, hopefully, far into the future.

So, how, exactly, do you convert your great scientific paper into a talk? I will discuss several specific issues in future posts. For now, I will focus on the overall structure of the talk. Below, I provide an outline to one approach to an effective structure. There are probably an infinite number of effective structures. I recently heard a brilliant talk by a student who seemed to break all the rules by showing us data and then saying, "Now, let's just ignore that finding for a while. We'll come back to it." She had us on the edges of our seats for the whole talk and the denouement was in fact very satisfying. So, my advice is meant only to get your creative juices flowing so that you can find a talk structure that works well for you. The overall goal is to keep the audience intrigued and, dare I say, entertained.

First, you will keep or lose your audience in the first few minutes. Remember, they are on your side, at least at first. So, take their goodwill and run with it. Make it clear from the very beginning what the problem is and how cool you think it is. This may require backing up a bit from the specific topic of your paper. Not everyone is going to immediately think that studying a few cells in the cockroach mushroom body is the coolest project ever, so give them something to grab on to. In plain language, talk about the general nature of the problem. Talk about how, for example, animals learn by associating different sensory experiences and that it is important to identify and understand the anatomical basis for this learning. But, make it specific, make it palpable. Talk about how cockroaches learn to associate your footfall with danger and to skitter into the wallboards when they hear you coming.

Then, with your audience attuned to the fact that you are discussing an interesting problem, feel free to proceed quickly to something specific about your project. Why the cockroach? Why the mushroom body? Why these cells? Use photos or illustrations or movies to rapidly orient your audience to the nature of the problem and the specific aspects you have studied. Don't put words on these slides, other than labels. I discuss this in further detail in Principle 1: Don't Put Words on Slides. Guide your audience through the images. Point to specific places in the illustration when you discuss this thing or that thing. Don't rush through these introductory slides (or any slides). Allow the audience to absorb the complexity of the images you are presenting and to accept your guided tour of the most important aspects for this presentation. Do not overwhelm them, but introduce them to the richness of the problem you have studied. 

Now, start revealing your methods and experiments. This is the most important point where you must strike a balance between breadth and depth. You want to provide sufficient information to demonstrate that you have performed your work carefully and that your experiments are sensible, but you don't want to drown your audience in details of your experimental methods. You can say "We generated transgenic Drosophila" without providing details of the injection protocol, the transgene structure, or the landing sites used (unless, of course, your talk is actually about new methodology related to transgenesis). This will take some practice, but when you are using widely-used methodology you can usually assume that your audience will believe that you used appropriate methods with appropriate controls. They can read your paper for the details if they are interested. 

It is rarely helpful to present all of your experiments. The best papers include multiple experiments, probing the problem from orthogonal directions, to demonstrate that a result is true. In a talk, that is not always necessary. If you are presenting a mind-blowing, non-intuitive result, then it is probably critical for you to present each of the experiments that supports the view that this result is correct. But, most scientific results are not mind-blowing and non-intuitive. Mostly, they are incremental. (In a good way! Science is mostly incremental.) The results often make a lot of sense given what we know already. If you are giving one of these talks, then you probably don't need to present every experiment and every control that you performed that led you to your conclusions. Instead, focus on the most compelling results. Tell the audience in some detail and with care the killer experiment that really proved the point. You can also tell the audience "There are several other experiments that support this view, but given the limited time, I will not discuss them today. If you are interested, I am happy to discuss them with you later." Then, they know that there are other results that support your conclusion and that you are not trying to hide anything. This allows the audience to focus on the key messages in your talk.

It is commonly suggested that speakers should "Tell the audience what you plan to tell them, then tell them, then tell them what you told them." I agree, but only in the broadest terms, not in the specifics. I do not think it is useful to provide the audience with a laundry list of the what you plan to tell them. (And never use a bullet point slide for this!) Instead, get them excited about what you are going to tell them by igniting their curiosity about the problem, just as your curiosity drove you to do the project. Then, when you are actually telling them about experiments, feel free to tell them the same thing in multiple ways. Maybe once in plain language, and then once in more technical language. Don't repeat yourself verbatim, but find multiple ways of explaining each result. Finally, as you wrap up the talk you have two choices. If you have performed a suitable job of engaging your audience, it is unlikely that a formal conclusion will be necessary. Everyone will understand precisely what you did and why you did it. In this case, you won't need to "Tell them what you told them." However, if you feel uncertain about this, then you should feel free to remind the audience of what you presented. Remind them of the problem and the approach you took. And point out how your results have contributed to our understanding or helped to solve the problem.

Now, a few words about slide content. Please, I really beg of you, do not present the figures from your paper directly in the talk. First, delete the figure panel labels (a, b, c, etc.). These panel labels do not communicate any useful information to the audience. More importantly, figure panels from papers are usually relatively small and the full figure can be complex, with many experiments presented in a single figure. You are likely simply going to confuse the audience by presenting many panels from a single figure in a single slide. In a talk, however, you are not constrained to showing small images because you can show multiple panels consecutively, rather than simultaneously. Show your beautiful experimental results in all their glory in images as large as possible. Allow the audience to really absorb these data. Allow them to, effectively, "interact" with the data and to come to their own conclusions about the data as you guide them through the results.

Try this exercise. Cut all slides except those that illustrate data, preferably the results from key experiments. Then, working from that skeleton, be extremely critical about whether a new slide really adds value that is better communicated with the slide than with your voice. Hopefully you will find that you are left with only images or movies that communicate information more efficiently than your voice. 

Then, for the talk itself, just stand, face your audience, and speak. Introduce your topic. Show slides when they communicate information more efficiently than your words can. And engage with the audience. Talk to them like you really want them to understand what you are saying. DO NOT talk to the slides, or even half-way to the slides. Face and talk to the people who came to hear you. Practice this. Practice it again. Keep practicing until it feels comfortable. I routinely practice talks, even talks that I have given previously, two or three times in the days leading up to a presentation. There is no such thing as too much practice. I was going to reserve a whole blog entry for the Principle "Practice, Practice, Practice." But maybe you get the point.

I am often asked how much detail should be provided in a talk. Younger speakers are often understandably nervous about not presenting their experiments in full, lest the audience think that they are not rigorous scientists. My usual advice is that you should aim to lose most of the audience on occasion, as you delve into the details of an experiment, but then rapidly pivot back to a more general conclusion that can be understood by everyone. If you drag your talk into the morass of details surrounding your experiments and you don't return, then you can be assured that most people in the room will not follow you. If they are generous, they will think you are a great scientist. More likely, they will think the talk is boring. Picture your talk like the silhouette of a very peaked mountain range rising from and then descending into the plains. Start very general, then ratchet up the details until you hit a peak of an elegant experiment. Very rapidly summarize this result for a more general audience. Then, you are free for an ascent up another peak. Of course, it is ideal if you can always use plain language, with minimal jargon, even when you are explaining your most sophisticated experiments. (Most talks abound in jargon. So you must examine your language very carefully to minimize the jargon.) The audience will not think you are dumb for using plain language. They will think you are brilliant for explaining such complicated work with such straightforward language. The ability to explain the most complex problem in the plainest of language is a sign of true understanding.

Principle 5 - Don't Take A Data Dump On Your Audience

Have you ever noticed that after some talks many people are eager to ask questions, but after other talks the audience is silent and everyone seems itching to get out of the room? There are at least two reasons for the itchy pants phenomenon. One is that the talk was boring and the the speaker has failed to engage the audience in the exciting elements of the science. I have discussed some remedies to the boring talk in previous posts. A second kind of problematic talk is the polar opposite: the take-no-prisoners talk, in which the speaker tells us everything they have done in a tsunami of slides. This is really a data dump talk. But, data dumps can take many incarnations. The take-no-prisoners talk is one extreme. I also see data dumps on a smaller scale. Sometimes a single slide can be a data dump!

I have never seen a data dump that improved a talk and most data dumps thwart conversation. Data dumps present the material in such a shallow manner that the audience has no idea how to begin thinking about the conceptual issues raised by the work. This is my explanation for why, most of the time, audience members have few questions at the end of a data dump talk. The speaker has provided no window into the science, just a facade of sciency things. Sometimes, when I am feeling really cynical, I think that speakers use data dumps to intentionally thwart conversation, by throwing up bulwarks to hide chinks in the armor. But, even if this is not the case, it is hard to know, because the data and ideas fly by so fast!

The worst style of data dump involves presentation of an undigestible quantity of data in a short time. I often observe extremely successful scientists giving data-dump talks—the take-no-prisoners talk—where they ever so briefly mention each of the projects going on in their lab, together with a historical review of everything they have ever done. When presented with excellent rhetoric and flair, these talks can seem very impressive. But, they are essentially advertising. While advertising can be entertaining and enjoyable, it can also be misleading and corrosive to understanding. Fundamentally, these are not science talks. They are not talks that invite understanding and a conversation about the science.

Data dumps appear commonly in genomics and other 'omics talks. 'Omics might be defined as the process of implementing a technological innovation to produce a ridiculous quantity of data. Because nobody can digest a vast amount of data quickly, especially when presented in a "novel" graphic form, 'omicists therefore bear a particular burden to explain their work slowly, carefully, and clearly. Instead, I often observe 'omics talks with slides flying by, each slide containing a different graphical representation of the 'omics data, and the speaker glibly rattling off one conclusion after another. The audience has no chance to evaluate the data and, therefore, is in no position to assess the conclusions. Of course, I have seen some 'omics talks that were beautifully presented. In such cases, the 'omics data were presented after a clear explanation of the biological question being addressed and the manner in which the 'omics data could be used to answer the question. And then, the 'omics data were presented in a deliberate and cautious manner, together with a discussion of the limitations of the methodology. In the few cases were I have seen this done well, I felt like I really learned something, and I had a ton of questions after the talk.

I have observed that labs that build tools for other scientists often present data dump talks. Often, the best tool builders are very prolific and their many tools really are very useful to a large community of scientists. Therefore, they often use a talk as an opportunity to advertise the existence of all their great tools. These talks are more infomercial than science talk. The problem with this is that a laundry list of great tools provides little insight into the process of how tools are built or the creativity and scientific ingenuity that underlies the work. These scientists keep the science veiled in mystery, rather than bringing other scientists behind the scenes and engaging everyone in a deeper understanding of the science of tool building. I recently saw a young acolyte of a great tool-building lab give a mini version of a data dump talk. Predictably, the audience was stumped for questions. Everyone assumed that building the tool involved a rather boring sequence of protein engineering steps. I was convinced, however, that there must be interesting science buried in the story. It was only under cross examination that the speaker told us that, in fact, the project was stalled for years because they couldn't figure out how to make the complex protein fusion stable. It was only when they thought to use the homologous protein isolated from a thermophilic bacterium that the protein was stabilized and the project started to work, leading to the great tool we were shown. A deep understanding of biology, a leap of insight, and voilá, we have a tool that nobody thought could be built. That is an exciting story that raises more questions than it answers. That's a talk I want to hear.

For these most egregious styles of data dumps, the remedy is very simple, just throw out most of the topics. Speakers should limit themselves to just one or two major topics for each seminar and take care to present the data and ideas underlying each topic in a cohesive story. 

Data dumps also occur on smaller scales; more like little rabbit data dumps, or data dumplings, rather than the big bear data dumps I discussed above. A data dumpling can be something as simple as showing a figure of data without explaining the axes and without taking the audience through the data. That is, anytime you show data and draw a conclusion, without walking the audience through the data, you have taken a little data dump on the audience. Data dumplings are extremely audience dependent. For example, a phylogeny of species is a very data rich image that is easy for many evolutionary biologists to absorb quickly. However, scientists in other fields may need a slower introduction to the meaning of branch points, branch lengths, and other features of a tree. Likewise, an audience of neurobioloogist will have no problem interpreting a rich dataset of spike trains, but most people outside the field will need a more gentle introduction to the data. Molecular biologists will scan a Western blot quickly, just as ecologists will quickly appreciate a species area curve; but you will need a little time to explain each to the opposite audience. 

These observations have led many people to offer the advice that speakers should "know your audience" before preparing a talk. This is terrible advice that has contributed to the balkanization of science. It is much better to assume that you don't know anything about your audience (except that they are scientists) and that they don't know anything about the topic you will present. If you start with that premise, you will quickly realize that you must explain your goals and approaches in plain language, that you should limit how many topics you will discuss (avoiding the big bear data dumps), and that you should carefully prepare how you will present and discuss the key data to communicate your most important messages (thus avoiding the little rabbit data dumps). Maybe someday I will write the Principle "Don't know your audience," but for now this short explanation will have to suffice.


Principle 4 - Don't Tell Jokes, But Don't Avoid Humor

A duck walks into a bar, orders a drink and says "Put it on my bill."

Does that joke inspire you to read the rest of this post? I didn't think so. That is usually how people feel when speakers tell a joke or present "funny" slides during a talk. For some reason, many people have been told that they should try to win over their audience at the beginning of a presentation by telling a joke or funny story. You can hear politicians doing this all the time. How funny are they? 

The point of this principle is that most people can't pull this off (like most politicians) and, more importantly, you don't need to win over your audience at the beginning with a joke. They are on your side from the very beginning. You don't have to win them over with humor. You won them over as soon as they walked in the door. They took time out of their busy day, or they traveled from a distant location, to hear your talk. Your job is simply to present your work in an engaging, honest, and concise manner.

The problem with forced attempts at humor is that you have lost track of why the audience came to hear you in the first place. They didn't come to a comedy club. They didn't come to a hear a TED talk. They came to hear about your science. 

A second problem with telling jokes during science talks, which are inherently international, is that understanding jokes often requires fluency in the native language and deep familiarity with cultural norms. Thus, even a joke well told will tend to leave the non-native listeners cold. Or, worse, they will think your joke was part of the science and they will puzzle over how to incorporate your unusual words into the flow of the science that they came to learn about.

At the beginning of my thesis defense, I made the mistake of telling a joke. It wasn't the end of the world. But the joke bought me nothing. The audience was there to hear bout my work, not to hear my lame attempt at humor. These days, I find that the audience often laughs at one or more points during my talks. We are having a "conversation" about my science and sometimes humor sneaks in. I never plan these humorous moments, they arise naturally in the moment. Some speakers do, obviously, plan humorous moments and this can work, especially if it is well practiced. But it is certainly not necessary and those new to giving science talks should not feel like they need to tell jokes or be funny to give a successful talk. In most cases, it will detract from the talk.

Leave the jokes to the masters. They find it hard enough to make an audience laugh. You have a much easier job. You have data and your audience is hungry for data. Feed them data and everyone will be happy.

Principle 3 - Show Your Data

When we talk with colleagues in anticipation of attending a talk, we say "I'm looking forward to hearing her talk. I hear that she has collected an impressive dataset on wombat nose lengths." We don't say "I'm looking forward to reading all the material in her slides. I am especially eager to trust implicitly her over-simplified data slides." 

A talk provides a special opportunity that is not available when reading a paper. We get to hear directly from the scientist their motivation for the work and to be guided through potentially complex datasets to the key discoveries. Remember, a talk is the beginning of a conversation. The conversation should continue in the question-and-answer period and, hopefully, long after. But, it begins with you and, especially, with your data. The audience wants to engage deeply with your data. Showing your data in some detail is one of the most effective ways of engaging a science audience in your work. 

Unfortunately, science figures usually show only the central tendency (often the mean or the median) of data plus some measure of the variability. There are two problems with this presentation of data. First, it communicates little information. It is often faster, and no less informative, simply to say "We found that treatment A generated a highly significant increase in wombat nose sizes compared with treatment B" than to show the mean and error bars with some cryptic indication of "significance" for the two treatments. Second, this reduced representation of the data is an opportunity lost. When you show the data that underly the summary statistics, you provide an opportunity to engage the audience. Plots that show all the data allow the audience to intuitively grasp the sample size, the extent of variability in the experiment and the distribution of this variation. In the plots below, for example, the summary statistics (mean ± SD) hide the fact that the data were sampled from populations with different distributions.

Showing the data allows the audience members to do what they like to do, which is to be scientists who evaluate data. This will both keep them engaged with your talk and bolster their confidence in your work.

There is an entire hornet's nest of problems with the visual display of results as simply the mean plus or minus some measure of error. These problems extend to papers as well as talks. First, let's deal with the error bars. Very often in talks, speakers provide no indication of what the error bars represent. In neurobiology, where I am a recent interloper, speakers are particularly egregious. Over the past four years, I can't remember hearing a single speaker indicating, either in speech or in text on the slide, what the error bars represented. I look at the bars and I think "Well, I've done behavior experiments and there is no way that they are showing me the standard deviation of the data, so those must be standard errors." Once I figure out that I am looking at standard errors of the mean, I scan the slide trying to find the sample size, so that I can try to back-calculate, on the fly, what the standard deviation should look like. (As you all know, the standard error is the standard deviation divided by the square root of the sample size. Therefore, the standard error, unlike the standard deviation, decreases systematically with increasing sample size. So, if I can just find the sample size, I have some hope of estimating the standard deviation, which I use to get a feel for the variability of the experimental treatment.) But, I almost never find the sample size on the slide and I am therefore stuck taking a huge leap of faith that what the speaker is telling me is true, because it is very challenging to divine from the slide whether the data supports their statements. It is at this point, if I haven't already been hopelessly confused by the speaker, that I stop listening and start thinking about something else. I once read on Twitter—from someone that I don't recall confiding in—the following tweet: "Rumor has it that David Stern stops listening if you use the standard error of the mean in a scientific talk." I don't think I've told many people this factoid about me, so I was shocked (and pleased) to learn that my peevishness about standard errors was gaining attention.

What, really, is the problem with the standard error? Well, fundamentally, of course, there is nothing wrong with the standard error. It is a perfectly valid way to estimate the confidence in your estimate of the mean. If you have come up with a new way to estimate the melting point of water, then you definitely want to report your estimate along with the standard error, so that we can all understand how much error there may be in your estimate. The problem is that the standard error is often used to represent the variability in an experiment (not in an estimate of the central tendency), so that a "simple" visual comparison can be made with controls and other experimental results. Use of the standard error of the mean in this context is really a bit of visual trickery and can be extremely misleading. Remember, to test whether two results are "significantly different," we usually compare the difference in the means of the data from each treatment with the variability in the experimental treatments, usually measured as the variance (which is the standard deviation squared). For most experiments, it is therefore far more intuitive to give the observer a sense of the variability in the treatments (standard deviation) than a sense of the variability in the means (standard error). Showing the standard deviation allows observers to quickly and properly interrogate the data and come to their own conclusions about the results. Using the standard deviation (and eschewing the standard error) is really only a half solution, however. It is far better to actually show all the data that underlie the estimates of the mean and standard deviation together with estimates of the mean and standard deviation. During a talk, this has the huge advantage of engaging the audience in a deep way with your data. 

Observers with experience looking at data (i.e. scientists) can usually intuit whether two samples look like they have come from samples with different means, which is usually what statistics are used to infer. Thus, in talks, it often becomes unnecessary to report the details of the statistics. You can say, if you like, whether results are "statistically significant," but this quickly becomes pointless when it is obvious that two samples with reasonable sample sizes have different means.

Finally, it really is just as easy to plot all the data as it is to generate any other kind plot. Often, I plot the raw data together with the mean and standard deviation. It is easy, then, for the observer to see both the raw data that underly the conclusion and the central tendency of the data. Many plotting programs, including the powerful plotting routines available in the free statistics package R, provide sophisticated methods for plotting all the data. For many experiments, the routines available on the BoxPlotR web server will suffice.

Here is an example from a great recent paper. Tenaillon and colleagues plotted one summary of their results as bar graphs with standard errors, which you can see on the left (or on top, if you are reading on your phone). In a review I wrote that highlighted their work, I re-plotted their data, which you can see on the right (or bottom on your phone). 

I plotted all the data with random horizontal jitter for four of the categories, along with the mean and standard deviation beside each cloud of points. The re-plotted data provide a more intuitive presentation of their experimental results (with the exception of the background shading, which imparts no information and was added by the journal).  

In some cases, when you have a lot of data, showing all the data may obscure interesting results. This can often be solved by altering the way individual points are represented. In this example, in an interesting paper by Cooper et al. the raw data were plotted (yay!), as shown on the left (top), but each point was small and uniformly black, which obscured the differences in data density in different regions of the plot. When I re-plotted these data for my book, seen on the right (bottom), I made each point larger and transparent, which allows the reader to see more easily that a huge fraction of the data have a relative fitness of 1, producing a trend that does not agree entirely with the regression.

Plotting raw data together with summary statistics also provides a more intuitive understanding of how the data generated the summary statistics. In this second figure from the Tenaillon paper, I fused their plots of the raw data and summary stats (on the top) into a single figure (on the bottom). To clarify the distribution of the data, I reduced the size of each data point and the thickness of the checkerboard lines, which attract more attention than they deserve in the original plots. I now feel like I have a more intuitive grasp of the support for the summary statistics, which are shown as shades of color.

For talks, when I present the quantitative data from multiple related experiments, I often include the data on the same "slide," but I rarely flash up all the data at one time. First, I introduce each experiment, then I reveal the relevant data. This allows the audience to follow the logic of each experiment and to engage in the data. Here is an example of five consecutive views of a slide from one of my current talks. First, I tell the audience what they see in the panels along the top, then I walk them through groups of relevant experimental results. If I had simply flashed up all the data and said "Here you can see that multiple binding sites confer robustness against temperature variation" and then moved quickly to the next slide, probably nobody would have time to digest the mass of data to convince themselves that this is in fact true. By breaking up the data, I force the audience to focus on each subset of the data as I discuss it. They "analyze" the data as they see it and they "discover" the result for themselves.

When you show all the data you are illustrating your results much in the way that most scientists engage with their own data. They are a smart crowd and they will not be overwhelmed by all the data. On the contrary, they will be flattered that you have entrusted them with a genuine view of your science. 

Postscript: While I was preparing this post, Weissgerber et al. published a useful paper that promotes plotting all the data and they discuss the many reasons that this is advantageous. I wholeheartedly endorse their conclusions and believe that showing all the data is likely to reduce publication of questionable results, as I discussed for a special blog post I wrote for BMC  to accompany a paper I published on a not-unrelated topic in BMC Biology.

Principle 2 - Use Black Slides

I once knew a famous developmental biologist who would sit in the front row at talks, bow his head low, and shield his eyes from the slides with his hands. He would listen intently to the description of the problem and the experimental approach and then cautiously peek up from above his hands to glance at the data when he thought the results were being shown. He would then quickly nestle his head back into his hands. He always asked the best questions after the talks. This biologists came up in the era before Powerpoint and I believe he was deeply distracted by all the graphics that are now displayed on slides. He wanted to understand everything the speaker said, so he did what he had to do to focus his attention on the speaker's words. 

The best way to convince the audience to listen to what you are saying is to eliminate any distractions. Remove extraneous visual stimuli and everyone in the audience will look at you and listen to what you are saying. Remember, they came to hear what you are saying and they are on your side, so don't be afraid to capture their full attention with your words. I admit that it can be somewhat unnerving to look at the audience and to realize that everyone is looking directly at you. A normal response is to want to look down or to look away. Resist this temptation! Look back at the audience. You will get used to this and you will soon learn that seeing everyone looking at you is the best sign that you are truly communicating with your audience. If you have lost your audience, you will see many heads not looking at you and not looking at your slides. Your "listeners" will be looking instead at whatever electronic device they brought with them. They are bored and they are not learning anything from your talk.

In the first entry, Principle 1, I provided the simplest way to focus the audience on your words and the relevant data, by removing words from slides. In later posts I will discuss other ideas to improve these data slides. But now I want to introduce you to my favorite "trick" for keeping the audience's attention. Insert black slides whenever you can explain things more clearly with words than you can with images. For example, I almost always start my talks with a black slide. When I start talking, everyone is looking at me and I know I have their attention from the beginning. I start by describing the problem. That is, I start, immediately, by talking about the science. Often I start with a very big picture view of the problem. Sometimes I start by discussing a commonplace observation that is related to the topic. Only rarely is an image required to introduce the broad view, but then, I quickly move to images that help me to explain the problem under study.

Later during the talk, I use black slides periodically as a visual pause. In these interludes, I review what I have presented so far. I try to use plain language to make sure that everyone in the room is following the major implications of each set of experiments or observations. Then, before I click to the next slide—which might show data, or a model, or some complex image which I will explain in detail—I say something along the lines of "Now, to test this idea we did experiment X" or "This raises an interesting issue, which is illustrated clearly in the following image." Then, I click to the next slide and we get back into the experiments.

These black slides also give me an opportunity to check on the state of the audience. Of course, when I present data slides, I usually point at specific items in the slides. This means I am not looking at the audience. With a black slide up, I can turn to the audience and discover if people are still interested. Are people looking at me or are most people looking at their iPhones? If they are looking at their iPhones, then I need to do something to recapture their attention. (I hope, though, that by the time I finish this blog you will have all the skills and tricks required to keep everyone's attention through the entire talk.) One simple trick to recapture the audience is, simply, to pause. Just stop talking for a few seconds. Everyone will look up to see what happened. Then, when everyone is looking at you, bring everyone back by summarizing what you have shown so far. 

It takes a little practice, but don't overstay your "visit" to the black slides. Don't launch into a huge monologue. The black slides provide a break from the visuals, and the visuals provide a break from the black slides. Everything in moderation.

Note, I said black slides, not blank slides. I have argued with some folks who feel that a blank white slide is just as effective as a blank black slide. They may be right, but I am not so sure. My impulse when I see a white slide is to look at the slide expecting something to appear on the slide soon. In contrast, my impulse when I see a black slide is to think "Oh, there is nothing to see here" and I automatically look at the speaker and listen to what they are saying. I would love to see a controlled study of the effect of white slides versus black slides on audience attention! For now, I encourage you to use black slides.

One technical note. If you do decide to start with a black slide, which I strongly encourage, be sure to alert the audio-visual experts who are helping you with your slides that your first slide is black. Otherwise, they may automatically start clicking through your talk looking for the "real" first slide. Needless to say, this will distract the audience from what you are saying!

Principle 1 - Don't Put Words On Slides

It may not seem intuitive, but including words on slides has a bigger detrimental effect on the quality of a talk than any other issue I will address. 

The slides in most science talks contain many words. This simple fact has many consequences, none positive. Most obviously, words on slides impel listeners to read the words on the slide. If you are talking while they are reading, then you generate cognitive dissonance that makes it difficult for the audience to understand either your spoken words or your written words. So, if you want your audience to read the words on a slide, then, presumably, you should shut up. But, I don’t recommend that. Instead, just cut all the words form your slides. 

This recommendation applies also to the dreaded “Title” slide. Title slides are a symptom of treating a talk like a paper (which I discuss further here). What, exactly, are you trying to communicate with a title slide? Who you are? They know that. What you will talk about? They very likely know that. Where you are from? Maybe that is unique information, but it is definitely not information that is pertinent to the science you are going to present. There is no information on any title slide that I have ever seen that improved communication. Instead, title slides often contain gratuitous images and information that distracts the audience from your critical first few sentences. So, with a title slide, you have very quickly ceded the first battle. You have let the audience not focus on your words and let their gaze wander over words (and sometimes images) that communicate very little information. Delete the title slide.

Now, onto the words on the remaining slides. These fall into two main categories. First, many talks include slides with bullet lists, or multiple full sentences. Sometimes these slides summarize the talk structure or sum up the conclusions. I have heard all the arguments about how these slides help guide the audience through the structure of your talk and clearly present the conclusions. There are at least two problems with these slides, however. The first is that speakers often read the lists or sentences. I believe this causes everyone’s comprehension of the subject to slow down. We read quickly, but we speak slowly. You therefore generate a conflict in the minds of the audience members by asking them to do both. Instead of asking them to think about an important issue, you are forcing them to artificially slow down their reading so that your words match their perception of the written words. So, not only have you probably not communicated your point to the majority of the audience, but you have insulted their intelligence to boot (they know how to read!). The second problem is that a well-organized talk really doesn’t require written signposts. These summaries are a crutch that contributes to making science talks clunky and boring.

In the second category of problematic-words-on-slides are titles for each slide. Many instructional materials on "how to give a good science talk" promote the use of titles on slides. The vast majority of science talks currently include slide titles. It seems that slide titles are currently the "industry standard" and accepted as an excellent way to communicate relevant information about the real information (the data) in the slide. We are all victims of this training—I was too, until recently—and I strongly encourage you to break the habit. Slide titles are a poor replacement for clearly stating the point with words. Often, titles purport to “describe” the content of the slide. About half the time these titles are cryptic, however, and simply describe the topic of the material presented in the slide. Since you are presumably talking about this topic, the title is redundant and the audience may have read the title and ignored you while you were speaking. The other half of the time, titles state the conclusion of the experiments presented in the slide. Some will read this conclusion and ignore what you say about the experiments, or worse, they will allow the title to color their interpretation of the data you present in the slide. Never write what you think the conclusion is. It is far better to guide the audience through the data—engage them with the material—and lead them to the conclusion. I explore this topic in further detail in Principle 3: Show Your Data. For now, just delete the titles.

Some words on slides are useful. Citations to previously published work are efficient ways to point the audience to some of the relevant literature. Keep it short, though. Usually the first author, date and journal title are sufficient to allow the audience to track down the reference. Other useful words include labels on figure axes and labels of graphical items. Avoid abbreviations. Use fully spelled out words whenever possible to simplify audience comprehension.

I commonly hear two reasons in support of putting words on slides, including titles and bullet lists. First, I have heard that many people recommend putting a title on the slide because it helps orient audience members who may have lost track of the thread of the talk. Hmm. Really? You are going to start with the assumption that the audience became distracted and lost your meaning? So, a title is really your confession that you are giving a boring talk and people are losing track? We can’t accept this assumption. Since we are going to help you give a non-boring talk, let’s start with the opposite assumption. People will be riveted by what you are saying. Nobody will lose track. And titles are therefore both redundant and distracting.

The second reason I hear is that words on slides help the speaker to remember what they intended to say. This excuse relates to many principles that I will discuss in the future. But let’s start with the basic issue of remembering what you want to say. All Powerpoint-like systems provide Presenter Notes that you can use to provide notes to yourself about what you want to say. This way, you can read what you want to say, but the audience has the impression that you are talking directly to them and not reading from your slides. This is an extremely helpful crutch as you learn to give better talks. Also, it is perfectly fine to write key words or sentences on index cards or a sheet of paper and look at your notes to remind yourself what you intended to say. Eventually, you will find you need these notes less and less. Later, we will discuss some of the other principles that are related to this same issue, like Practice, Practice, Practice and Don't Tell The Audience Everything. We will get to these principles in good time. For now, delete the words from your slides and put them in the Presenter Notes or write them down. You will discover that your audience will immediately start paying more attention to you, since they won’t spend any time reading your slides, which is exactly what we want them to do.

I can think of only one exception to the Don’t Put Words on Slides principle. The only time I have seen this done with any success was when an extremely engaging speaker read a quote from a book that was printed on the screen. We read along with him (silently) and it felt like a game. Even though we could have read ahead, we all read together as if we were a silent chorus. Like all successful exceptions, however, this works only when it is an exception! It breaks the rhythm of the presentation and invites the audience to participate as a community. 

So, if you can’t put words on slides, what are the slides for? Now we are getting somewhere! We will revisit the content of slides later. But, for now, just put information that is communicated better with images than with the spoken word. This is really the underlying reason for the Don't Put Words On Slides principle. In other words, use your speech to explain things that can be explained most clearly and efficiently with speech and use your slides to communicate things that cannot be explained efficiently with words. You could try to explain the Mona Lisa to someone with words, or you could just show them a picture. In contrast, when first introducing what you do, just say it. And when you say it, make absolutely sure that there is nothing visible that will distract the audience from your words. Then, they will understand you perfectly.

Some readers may have noticed that the Don't Put Words On Slides principle hews closely to principles espoused by Edward Tufte. See, for example, his take on the problems with PowerPoint. Also see Peter Norvig's brilliant parody of Abraham Lincoln's Gettysburg Address employing PowerPoint.

That’s it for Principle 1. As I have intimated, we will return to the words on slides issue later. All of these principles are related, so I will often refer back to earlier principles and forward to principles that we have not yet discussed.

How To Give a Talk

If you have ever had to stand up and present ideas or data in front of an audience, then this blog is for you. 

In my opinion, the vast majority of science talks are terrible. I don’t mean that the science is bad or that the speakers are not trying to present their work clearly. I mean that information is often presented in a muddled fashion, with visual distractions that prevent the audience from understanding the point and—all too often—the talks are simply boring. They are, therefore, a poor mechanism for communicating information to the audience. The result is that during most scientific meetings, I observe that the majority of the audience has tuned out of many talks. They are checking their e-mail, surfing the web, or just staring blankly at the slides, not really absorbing the information. Often, these audience members blame themselves. They think they are not smart enough or sufficiently advanced in the field to understand the complex material presented in the talk. This represents a huge missed opportunity to communicate the excitement of science to a broad audience.

Over the past twenty years of giving scientific presentations I have learned, developed and simply stumbled upon a number of very simple principles that anybody can use to present a good talk. These principles did not come naturally to me. I was not born with a gift for giving a good talk. I worked hard at it over many years and now I find that audiences are usually deeply engaged with my talks. Some of the principles I employ are adopted from the brilliant insights of Edward Tufte, others from the approaches espoused by Alan Alda, and some I have developed myself. In this blog, I will share these principles with you. 

Let's start, though, with an overview of the fundamental problems with most science talks.

First, most speakers manage to drain their talks of all of their enthusiasm for their work.

Second, most speakers rely on jargon and fail to communicate in plain language.

Third, most speakers employ their audio-visual aides in ways that detract from their message and make it hard for observers to understand the essential points.

These are the most obvious problems with the vast majority of talks. There is another category of talk that I call the “take-no-prisoners” talk, where an obviously successful scientist presents everything they have done or everything they are doing in whatever time slot is available. Often, slides flash across the screen at—literally—an incomprehensible rate. Do these speakers really think they are teaching us anything? Are they trying to impress with volume (rather than quality)? Are they insecure that we won’t think they are productive? I have no understanding of the psychology that underlies this talk format, but in its own way, it is as useless as the boring talk.

So, you may ask, now that I have trashed most science talks, what makes a good talk?

First, it should  be engaging. Just as engaging as a fascinating discussion with a good friend. How do you make it fascinating? Well, here’s the big secret. Just treat the talk like you are having a fascinating discussion with a good friend! In essence, that is all there is to giving a good talk. In practice, there are many tricks you can use to reach this goal. With some fanfare, I will call these tricks "principles."

In this blog, over the coming months (and perhaps years) I will describe in some detail the mechanisms I use to, essentially, transform a talk into a conversation. 

Here's the plan. Periodically I will state one of my principles in this blog and provide an argument for why it seems to work and how you can implement it in your talks. Most of the principles I plan to discuss are intuitive, at least in retrospect, like: Speak Plainly and Slowly , and Be Yourself. Some of the unintuitive principles include: Don't Confuse a Talk with a Paper and Don't Tell the Audience Everything and Stop When You Are Done, Preferably Sooner. I very much doubt I will end up presenting the principles in a coherent order. I am writing this blog in what remains of my “free” time, so I will not expend extra effort at the moment to organize the content. 

Eventually, all of my principles will be enumerated on this blog, everyone will learn them, nobody will give a bad talk, and we will all live in peace and harmony. Well, at least I will have a much more pleasant time attending scientific conferences.

So, let's begin. We start with my most important principle: Don't Put Words On Slides