Meta Talk:
How to Give a Talk So Good There Will Be Pizza Left for You

David Evans
University of Virginia Theory Lunch
25 January 2007

Talk Slides: [PPT] [PDF]

Tell a Story

All talks should tell a story (rabbit story example). Don't fall into the trap of giving a list. A good story builds and releases tension to keep the audience interested. All good stories have a beginning, middle and end.

The Beginning (30 seconds to 30% of your time)

• Introduce characters. Characters may be familiar creatures or abstract things. If your characters are not cute and furry, ou need to give your audience a reason to care about them.
• Put them in a predicament: motivate an interesting and important problem.
• Suspense is good for tales, but not recommended for technical talks (unless you are Andrew Wiles).

The Middle (50-85% of your time) — build up to a resolution of the problem. A technical talk should include at least one technical nugget including a clear example. For a conference talk, you probably only have time for one; for a seminar-length talk, you should have a few small technical nuggets and one large one.

The End (1 minute to 20% of your time) — good endings have a moral!

Respect your Audience

Have a Goal. Make sure you know what the goal of your talk is, and design your talk about achieving that goal.

Prepare. Think about the structure of your talk and the most important things you want your audience to remember before constructing your talk. Practice your talk. You should practice in your head, out loud with an imaginary audience, to a friendly audience, and to critical audiences.

Use Time Wisely. Advertisers pay over $2.5M for 30 seconds during the SuperBowl (how much is your audiences' attention worth?), so don't complain if you only have 5 minutes. Plan your talks to use the available time as usefully as possible (but for seminars, don't be afraid to deviate from your plan if you get interesting questions). You should plan your talk to fit in the available time leaving plenty of time for questions. No one ever complains that a talk is too short.

Don't Apologize. If you feel the need to apologize for not preparing a good talk, prepare a good talk.

Don't Be Arrogant. Don't dismiss questions rudely (it is fine to defer questions until later), don't belittle previous work (it is fine and important to explain how your work is better though).

Don't Be Boring. If you are bored, you are being boring.

Use Props. When you can, find a way to use props in your talk; ideally, use live, creepy animals. People are unlikely to forget a talk that includes unleashing a swarm of mosquitoes or revealing a live snake.

Slide Tips

It is often better to give a talk without using slides, but in most research settings difficult to speak without some visual aids. If you do use slides, use them as a supplement to your talk, not as the primary focus.

Fight the Power. PowerPoint is designed for making list talks not story talks. If you use it, be conscious of this and fight it.

Variety. Don't spend 2 minutes per slide. You should have some long, deep slides (at least 5 minutes), some medium slides (2-3 minutes), and some short slides (30 seconds).

Simple Design. Use simple, sans serif fonts, and plain colors. If you want people to take you seriously, don't use "Comic Sans MS". If you are speaking in a dark room, it is best to use bright (white, yellow, gray) text on a dark (black, dark blue, dark green) background. If you are speaking in a light room, use a white background with black text. You can have a small footer with your slide number, name, and title, but don't waste screen space and audience attention on fancy backgrounds, sidebars, headers, or footers.

Other Advice

Patrick Henry Winston's, How to Speak (summarized by Cal Newport). I especially like this advice:

Have one meaningful picture per slide. If it's found in Microsoft's clip art gallery, it's not meaningful.

Oliver Danvy, Issues in making an oral presentation (//www.brics.dk/~danvy/issues.html)

Michael Ernst, Giving a technical talk

Simon Peyton Jones, How to give a good research talk (//research.microsoft.com/Users/simonpj/papers/giving-a-talk/giving-a-talk-slides.pdf) — Lots of good advice, but I don't agree with his advice about not preparing, or with his font choices or slide designs.

Guy Kawasaki's blog on How to Change the World (//blog.guykawasaki.com/pitching_presenting_and_speaking/)

Norman Ramsey, Giving Talks (//www.eecs.harvard.edu/~nr/students/talks.html)

TED Commandments. The first three:

• Dream big. Strive to create the best talk you have ever given. Reveal something never seen before. Do something the audience will remember forever. Share an idea that could change the world.
• Show us the real you. Share your passions, your dreams ... and also your fears. Be vulnerable. Speak of failure as well as success.
• Make the complex plain. Don't try to dazzle intellectually. Don't speak in abstractions. Explain! Give examples. Tell stories. Be specific.

The real entertainment gimmick is the excitement, drama and mystery of the subject matter. People love to learn something, they are "entertained" enormously by being allowed to understand a little bit of something they never understood before. One must have faith in the subject and in people's interest in it. Otherwise just use a Western to sell telephones! The faith in the value of the subject matter must be sincere and show through clearly. All gimmicks, etc. should be subservient to this. They should help in explaining and describing the subject, and not in entertaining. Entertaininment will be an automatic byproduct.

Richard Feynman, Letter to Mr. Ralph Brown, Advisory Board in Connection with Programs on Science (in Perfectly Reasonable Deviations from the Beaten Track)