Home Notes Experiments Articles Papers DPhil Talks Conferences
Misc Stats Teaching Future Profiles Contacts Media Search

How to Give a Talk

My favorite tips on how to give a talk are by Gordon Bower (click here to see the original). Below are a set of tips compiled from Gordon, other excellent talk-givers, and my personal experiences at conferences and on the job market.
First, the bad news: the best thing for a talk is an interesting topic and some compelling data. The good news is that improving every other aspect of your talk is really easy. Here are some tried and true do's and don'ts.

I. Preparation and Practice:

Find out as much about your audience and your room as possible. Who will be there? Only social psychologists? All kinds of psychologists? Undergrads? Physicists? Art history professors?

Always bring a bottle of water with you.

Dress sharp, but never wear anything uncomfortable. I pity women who try to give talks in heels, too-tight skirts, or improbable hair arrangements, or guys who seem strangled by their ties. You will give a much better talk if you're comfortable. If your tie strangles you, forget the tie. Give at least one practice talk wearing what you plan to wear for the big event. Make sure you can raise your arm to point at things (without splitting seams or revealing mid-riff).

When travelling, always bring a disk copy of your talk, and always put a copy of your talk somewhere you can easily download it. If something goes wrong, or you need to make a last-minute change, this will give you the best chance. Always bring a bunch of empty transparencies and a few overhead pens.

If you plan to give a powerpoint presentation, always bring back-up transparencies. This is good for two reasons: (1) you will not set off to the conference or the job talk without having a fully prepared and printed talk, (2) you'll have a back-up when something goes wrong. I have yet to go to a conference where at least 3 talks were not ruined by faulty LCD projectors. Don't take the chance. Always bring a back up.

Never check your talk at the airport. Keep it with you. In fact, if you're on the job circuit, don't check anything. You're in a stressful enough situation already - no need to make it worse.

You must practice giving your talk, several times. First, do it alone. It's weird talking to an empty room, but you'll get used to it. What this practice buys you is the transformation of your talk from an abstract set of points into a verbalized story. It gives you the chance to find and fix any verbal hiccups you may have in the talk, or any places where you might be missing transitions or explanations.

Get as wide a practice audience as you can. And listen to the practice audience's advice! When listeners tell you that something is confusing, they are always, by definition, correct.

If you're preparing for a job talk, do two practice runs for a selected audience of close student and faculty friends and then do one more with a broader local audience before taking it on the road.

II. Style
Everybody gets nervous before talks. Don't let nervousness get in your way - take advantage of it instead. A slightly nervous edge can add zing to a talk. Feeling nervous is very similar to feeling excited, so tell yourself you're just excited and act that way.

Exude self-confidence. Steer clear of hedges such as "kinda", "sorta", "basically" and the like - these often convey very little other than that you're unsure of yourself.

Be excited to be there. Be excited about your work.

Right before your talk, chat with an audience member or the organizer of the talk - this takes up the time you would otherwise spend sitting alone and getting nervous.

Don't give your talk sitting down. Unless you're giving a very informal talk to only a few people, stand. If you've got more than 10 people, they'll expect a performance.

Talk to the audience, not the screen.

Talk simply, like you're telling a story to a friend. Unless you're the next Martin Luther King, don't orate. It makes you seem really full of yourself.

A talk is not a written paper. Talks have an informal narrative style and are dramatic rather than detailed or completely informative.

The model for the short talk is the campfire story -- teller of a mystery. Talk informally as though you were telling your grandmother what you did and why. Complexity of expression is uncorrelated with wisdom, intelligence, and originality; it's perfectly correlated with audience puzzlement and boredom.

Be with the people. Walk toward and away from the audience as well as left and right to help break down the implied barrier.

Do not read your slides to the audience. The slides should be mostly pictures, plus a very sparse outline of the talk to help the audience follow what you're saying. Ultimately, reading a talk is better than giving a terrible, incoherent talk --but only a little.

Ask real and rhetorical questions to keep people actively engaged. Get people to raise their hands to make predictions. (who thinks it will work this way?)

Don't be self-deprecating in job talks. It's fine in lab talks and small lectures, but not job talks.

Humor can be great, but there are several cautions. (1) it has to be topical - don't put up dilbert cartoons that only sort of relate to your topic - that's lame. (2) it damn well better be funny. if you're not a good joke teller, don't do it. a failed joke can be really difficult to recover from

III. Figures & Examples
Use lots of examples.

Use lots of figures. A picture is worth way more than a thousand words. Try to develop a talk that is entirely in pictures. Then go back and add one or two words per slide.

Use props. Talks are about show and tell and keeping your audience amused, so you can inform them painlessly about what you are doing. Whenever possible, bring and use props: videotapes, examples of stimuli, etc.

Make sure all your demo's work. Cue the videos and check the projectors - make sure everything works. Practice turning things on and off so it goes smoothly.

Use color. Audiences these days expect color. But don't go overboard. Making your talk visually attractive is one thing, but don't turn your slides into a circus. Different projectors will make your colors look different - the more colors you use, the better the chances they'll look really gak.

Don't switch color schemes from slide to slide. If the "verb recall" column is yellow in slide 1, then it damn well better be yellow in slide 2.

Be mindful that about 20% of your audience has some kind of vision impairment. Avoid red/green contrasts.

In visuals, make it simple, clear and obvious. Don't clutter slides with irrelevancies. No more than 7 words on a visual. No more than 7 numbers on a visual (round them to one or two significant digits).

One word can abbreviate whole phrases. If you have lots of results you must show, use many slides, not one cluttered slide. Idealize graphs, no lightning-bolt data. Ask: are the exact values all that terribly important for my point?

In PowerPoint, NO fancy fade-inds. No slides swooshing in from the left, no dissolves. Just don't.

IV. Clarity, Clarity, and more Clarity
The three most important things in a talk are: clarity, clarity, and clarity. Nothing matters if the audience doesn't understand what you did and why.

Explain the task in terms of what the subjects were doing, not in terms of abstract theoretical manipulations.

Be redundant. Say the same thing several times in different ways. It's all new to your audience, so give them the best chance of understanding you.

If some manipulation is particularly hairy, make a picture or diagram explaining it. Before you go on, it's ok to say "Does anyone have a clarification question about how this worked before I go on?"

Present data kindly. If you must present lots of data, present each piece separately on a different slide. Present the most important data first! (Present manipulation checks first when it is necessary for your argument, but not otherwise). What the audience wants to know is "Did your experiment support your primary hypothesis?" so answer this question first.

Speak slowly, loudly, and clearly. Make sure the people in the back can hear you.

Use large fonts. Anything smaller than 24 point is too small. If you photocopy a paper from a book and project that, you deserve severe punishment.

Text is clearest when it's black on a white background. Yellow text on blue background is not so good.

V. Mechanics
Point to the projection (screen), not the source. Here's why:

The projection (the screen) is a shared artifact - both you and your audience can see it.

You want the audience to keep their attention on you. When you go up to the screen and point, they're with you. When you point at the projector, they have to look past you toward the screen and you lose their attention.

Most people pointing at the projector will end up getting their shoulder in the way and blocking the projection. Very annoying.

Pointing at the projector will often jiggle the slide. Also annoying. In general, don't touch the slides after you've put them up. Step away from the projector.

There are occasions when you cannot reach the projection to point at it directly. Put your hand into the light and make shadow pictures: use the shadow of your hand to point at the part you want to deal with.

Do not use a pointer. Here's why:

If you're nervous, the pointer dramatically magnifies the shaking of your hand. That leaves a bad impression.

People cannot find where a laser points very quickly. You probably zip it around and circle things. You're making your audience dizzy. Or you say "like this here" and they don't see where you point because the laser is already somewhere else.

Very few speakers are capable of speaking without playing with the thing that's in their hands. It's distracting. You shouldn't have things in your hands when giving a talk.

Do not adjust the slide unless it's falling off. It makes you look really nervous. Get away from the projector and point at the screen. You won't be blocking the view of your audience and you won't look as nervous.

Be sure the projection is on the screen. Whenever you put a new slide on, take a look back to see that it's displayed properly on the screen.

Do not cover up parts of the slide. The "overhead striptease" act can be very distracting. If you'd like to keep something in suspense or build up information gradually on a slide, use an overlay transparency.

Powerpoint makes it really easy to do "multiple overlay" slides. These can be quite effective.

Put up a slide only a moment before you want to refer to it. Give the audience time to read it or describe it to them. Remove the slide when you want the audience to attend fully to you again.

Do not let anyone darken the room. The darker it gets, the less alert people will be. If you must talk in a dark room, bring a small flashlight so you can see your notes.

VI. From beginning to end
In a job talk, start by saying something like "I'm honored to be here today. Thank you very much for inviting me. I'm very excited to have this opportunity to tell you about my research."

Prepare your first two sentences like they were a Madison-Avenue advertisement for you and your talk. Grab the audience in these first sentences.

Example weak start: ``The research I will tell you about stems from earlier work by Johnson published in Cognitive Psychology which led to a lot of follow ups; and I want to thank my collaborators, Jim and Dorothy Smith''.

A better start: ``How do we understand language'? How can I figure out the meaning of what you say? Some people believe we have a mental dictionary with fixed entries and we assemble the meanings out of this fixed dictionary. Another theory is that we only have flexible procedures which decompose compound phonetic strings into basic morphemes from which we compute a meaning for the utterance . . .''

Get interest and attention first, with a rhetorical question, anecdote, or startling statement or paradox. Assume your audience is an Introductory Psych class of undergraduates.

Before you can say what you did, you must say why you did it. What's the big picture?

You must be very selective of what you can say in a short time. Most short speeches can barely carry one main idea plus its support. Resist the temptation to tell everything you know or every thought you had about it: only the most interesting and important thing can be said.

Ask yourself "What is the take-home point here?" Say the take-home message early and often.

A narrative style is preferable in talks. Research is done to tell a story, going from problem, goal, plan through actions (observations) to outcomes, resolution, and a moral (conclusion). Avoid a written journal-style organization.

In longer talks, tell the audience your plan. You should also come back to it to let them know where they are in the talk as you go along. This can really help people put it all together.

In your plan, focus on the questions you're trying to answer. This will get your audience inetersted and will also help them understand what you're doing.

example useless plan:

previous studies
experiment 1
experiment 2
experiment 3
further questions

much better plan:

The history of cats and dogs
Do dogs really chase cats?
Why do dogs chase cats?
How do cats feel about this?
Will cats and dogs ever get along?
Implications for the Arab-Israeli conflict.

This way you can come back throughout your talk and answer the questions one by one.

Desribing your experiments. You are not duty-bound to describe every condition of your experiment, not every result, not every analysis. In particular, suppress complications and unresolved loose-ends or incomprehensible pieces of results -- don't lay your confusions on the poor listener. Your goal is to tell a simple coherent story, to interest and to entertain, not to tell the complete unvarnished messy truth. Your first rule is: tell a simple mystery story that has a neat wrap-up and don't confuse or bore your audience. Not telling the whole truth is not the same as telling a falsehood. Speeches are for conviction, written papers for corrections!

Describing your data. In narrative talks, descriptive and inferential statistics should be suppressed. Speak "eyeball-effects" rather than F-values. Say "These words were remembered very much better than those", NOT "The mean recall for the two categories was 8.76 and 4.37, and difference gave an F of 13.8 which with 1 and 14 degrees of freedom was statistically significant at the .01 level." A better attitude towards description is "Holy baloney, look at that!"

The first thing to do when you put up a graph is to explain what the axes are and what the colors mean. "On the vertical axis I've plotted reaction time, and here we have females on the left, and males on the right. The yellow bars represent how quickly people solved spatial problems, and the red bars how quickly they solved verbal problems." Only now are you ready to say what you found.

Bring up alternative explanations or potential problems when you think people in your audience will think of them. Don't wait till the end of the talk. If you wait till the end, you'll have people in the audience who've been sitting and stewing on this criticism ever since they thought of it, and they probably haven't heard a word you've said afterward. You don't need to address the criticism right then. You just need to acknowledge it, and say that you would be glad to discuss later how to address it or how to reconcile the two view points. It puts your audience at ease, and they will be more accepting of what you say knowing that you're being thoughtful and forthcoming with them.

If you want to say something controversial or speculative, mark it as such. The audience will be much more accepting if they know that you know that what you're saying is speculative. It makes you look careful and thoughtful, but at the same time interesting.

Summarize in two steps: First summarize your findings. Second, show the meaning of your findings for the "Big Picture". Finally, point out what other provocative questions your findings suggest.

VI. Ending and dealing with questions
Do not go over your given time. Even if you start late, it's a courtesy to the audience to end as close to on time as possible. Talking overtime can easily lose you all the points you've previously gained.

If you have more material that you desperately want to cover, make it easy for the audience to ask you a question about it afterward.

Don't worry about "tough" questions: they almost never come. You know more about the research than anybody, so you have a great advantage. Don't be intimidated by "big shots" in the audience (if there are any): most are struggling to comprehend, and ask only simple questions.

If a question comes you don't know about, it's okay to say "I don't know". Or to say "That's a tough one I haven't thought about -- or I'll need more time to think about that" -- or "Fine idea -- would be worth trying in an experiment". You don't have to have instant answers for everything. If you don't understand a questioner, ask him to rephrase it so you can understand. If he asks three questions, answer any one of them and move on.

Plant at least one pithy question with a friend so he/she can direct it to you in case no one else pops up with a quick question. Often the audience needs time to think of some question to ask about -- so give the audience a long time to come up with a question.

Learn how to say "shush": If you feel that questions are leading you off your track, inform your questioners of this fact, and tell them you will return to the issue later on.

This is your talk. Don't let someone else take control of it by forcing you to deviate from your organizational plan. If someone requires clarification, then answer them briefly and continue. If someone wants to argue philosophy (e.g., "But don't you think that psychology errs when it thinks of people as real?") don't take the bait. A good standby is something like "That's an interesting question and I've given it some thought. In fact, I'll be addressing that issue in a few minutes, but if I don't answer that particular question, please ask it again at the end of my talk. "

Don't agree to criticisms you don't understand.

Don't get defensive.

Act interested in the questions. You are not defending a fortress, you are talking openly about scientific ideas with interested colleagues. If someone does try to attack you, turn them to your side by saying something like "That's exactly the kind of thing I think we need to spend more time thinking about. So, let's think together about what kind of evidence we would find convincing." Then they're thinking with you, not against you.

Prepare slides that address common questions. This is where practice comes in handy. If you get some question more than once, prepare a slide to address it. Your audience will be very impressed with your foresight.

Written by Lera Boroditsky

mnemon.net 2002. all rights reserved.