Planning and Giving a talk

Ph.D. Professional Seminar. 2 April 1997.

Ken Birman

• Know your audience. Decide what you want them to learn from your talk.
• Go slowly even if you need to omit detail.
• Include that hard technical core, even if you only convey the flavor or thrust of the argument.
• Say less, but say it clearly and forcefully.
• Adapt as you go. Don't read the slides.

• Only plan on conveying one "idea" per 20 minutes allocated. More information than this cannot be effectively included. However, an idea isn't just a single sentence, it is a developed region of your talk. Using the focus of one or two "ideas" which are developed in the course of your talk is a useful structuring technique as well.
• Engage your audience. If you don't, they will lose interest. Tips on this are given below.
• Each talk should "peak" with a central 10 minutes or so of hard, core material. If your presentation is technical, you may lose some of your audience here. This is the one time in your talk that that will be acceptable, so long as you are teaching the rest of your audience.
• As an engineer, you must relate your work to the real world. Be sure to make this connection in your talk.
• Remember to reiterate your key themes throughout the talk.

• 20 minute talk to specialists:
  • 2 minutes warmup
  • 5 minutes to state model and problem
  • 5 minutes for core results
  • 5-8 minutes for prior work and recap/summary
• 20 minute talk to a general audience:
  • 5 minutes warmup
  • 5 minutes to state model and problem
  • 5 minutes for core results
  • 3-2 minutes for prior work and recap/summary
• 50 minute talk to specialists:
  • 5 minutes warmup
  • 10 minutes to state model and problem
  • 25 minutes for core results
  • 5-3 minutes for prior work and recap/summary
• 50 minute talk to a general audience:
  • 10 minutes warmup
  • 10 minutes to state model and problem
  • 15 minutes for core results
  • 10-15 minutes for prior work and recap/summary

Try to find out in advance who your audience will be. If this is not possible, assume a general audience. However, you should also assume that your audience is full of very smart people who lack background on your topic. The audience is doing you the favor of listening to you. Your job is to convince them that the topic is interesting and important and that they should read the paper. The role of the talk is not to educate but to advertise.

• Do not get too detailed. A talk should convey the style and impact of your work but usually can't cover the real details. Don't include details that are too small, say too much about things that are not central to your talk or let your slides get too packed. More on this below.
• Losing your sense of timing is a big problem. Many talks drift. However, it takes years of practice to get timing right without a clock visible. It is probably a good idea to have a clock with you, in case there isn't one that you can see in the lecture room. Watches are not a great time source because they are too hard to check. Don't trust time keepers; they tend to wake up one minute before your slot ends and are only good for cutting you off. Having your timing be off is particularly bad if you allow yourself to get disoriented or panic. Try to recover gracefully. Some ideas on dealing with timing are given below.
• Never, ever let yourself run over the allocated time for the talk. Even if you start late. Even if you are interrupted with questions. Even if you are told you can run over, unless they insist, and then you should at least pause at the scheduled ending time to let any who needs to leave do so.
• Don't go out on a limb. If you don't know something, don't fake it.
• Outlines aren't necessarily a good thing. Try to use an alternative method of guiding your audience through your talk. Ideas for this are given below.
• Covering slides is insulting and annoying. Don't do it.
• Abuse of pointers is confusing. It also makes you talk at the screen rather than the audience. Point only when absolutely necessary.
• Never read your slides to the audience.
• Avoid lists; they are very boring. It is okay to have a list on a slide, but if you do, don't read the whole thing. Use it to indicate that there are many things which fall under the category of this list, but then focus on only one or two items.
• Don't let questions throw you off your timing. Try to answer briefly and move on. More suggestions on this are given below.
• Don't apologize; believe in your talk.

Most speakers wish they could present their hardest technical results. But most audiences aren't looking for that. Only a few people in your audience will appreciate what you actually did. Instead, focus on defining the problem and why it is interesting and the overall tone of your work. Don't rush to reach the "good stuff"; if you go too fast at the beginning the audience won't understand why the problem matters. This means that you will have at most 10 minutes for the real meat of the talk. When you get there, make it clear that it is interesting and why. It is okay to lose them here, but make sure you pick them up again in your recap.

While you are being careful to give the big picture, make sure that you don't omit the core of your work. The audience expects the talk to have a core insight and they want to learn from you. Build up to your core, then hit them with your basic results, place the results with respect to prior work and end with a recap which places what you did in context again, including pointers for future work and a visionary closing sentence.

You don't need to have an outline for your talk. It may not even be desirable. Of course, even if you omit showing your audience an outline, you should have one in your head and you should express the spirit of it to your audience. Let them know what to expect. More importantly, let them know why the talk is interesting and what your innovation concerns. This includes explicitly stating what you are going to show them and why this is interesting, emphasizing who would benefit from this work, positioning your work in the area and explaining the importance of the result. Then, when you reach your result, remind them of what you promised to show them. Remind them again that you have met this promise at the summary. For you as a speaker, have a sense relative to your mental outline of what your goals are and how long they take you to meet. This will help if you have to adjust your timing.

• Talk to them, not the slides or the blackboard.
• Pause after key points and let them soak in.
• Relax. Look at the audience. Enjoy yourself. Take your time.
• A little humor never hurts, but limit it.
• Look at the room when you speak and watch to see if they are following you.
• Pause often and look around.

If your audience is making you uncomfortable, instead of looking at individuals, look at gaps between them or at the rear walls. Make sure you continue looking out towards the audience; don't look at the slides or the blackboard. Concentrate on speaking in one minute "chunks" and then look at the audience during the pauses and focus on what you are saying in between. Slow down even if you need to skip details. Hit the important points and remind them why this matters and how it relates to the big picture.

Think ahead of time about what matters most in the talk and what points you really want to communicate. Keep these in mind at all times.

It isn't necessary to hit every point on every slide. In fact, many good speakers look at the slide they are showing but then "ignore it" and focus on the audience. It is normal to skip a point on a slide or even an entire slide. The audience doesn't want to be read to. If you start to fall behind, just skip ahead, making sure that you hit your main important points. During your wrap-up you can have a few words on what you didn't have time to cover.

Don't put too much on a slide. Trust powerpoint font sizes and margins. Then use as a rule of thumb that slow speakers need to count on 2 minutes per slide and faster speakers can perhaps get as little as 1.5 minutes per slide.

Your talk might start late. In fact, most seminar start about five minutes late. Introductions cut out of your time as well. Plan on this. Sometimes circumstances cause a long delay in the start of your talk. In this case, ask if you should try and end on time or if not when you should stop. When you are forced to trim back, think hard and be sure to hit your key points anyhow. It is a good idea to have an idea ahead of time about how you can cut your talk if it becomes necessary.

Timing must be balanced with getting your information across effectively. For the audience to understand your hard technical core, go slowly through the hard material. If you have formulas, ask for two projectors. This way you can leave key slides, such as your model definition, up on one and talk next to the other. Whenever you show a formula, you must give your audience time to read and understand it. Point out the key things, though not every detail. Speak slowly and to the point.

You can't show the audience everything that you did. Instead, show them a key theorem or two, instead of all eight or ten of them. Explain how your proof works for these sample theorems. Then state the other theorems but don't give the proofs. This works especially well if you can pick a theorem to present that gives the flavor of the proofs of the other ones as well. Remember that even brilliant people can be lost in details. Pick the point that you want to make and focus on it.

Think ahead about what the board should look like and where the information will end up with respect to each other. As always, you must design the material to assist the audience. Think about how it will look to them. When writing, you can read your writing but don't speak on the subject until you are done. You can also write and then turn to the audience and read them what you wrote.

Questions during your talk can throw your timing off. Answer questions briefly. If you do not understand the question, rather than asking for a clarification, guess at what the question should have been and restate it in your own words. Only use a sentence or two to answer. If the question is on a topic that you will cover soon, say that you have a slide on the question in a moment and when you reach that slide point it out.

Sometimes questioners will get more aggressive. They may not agree with you and will start to give a seminar on their reasons. In this case, gently thank him or her and say that you really need to move on now. You should answer a question once, and briefly, but avoid getting drawn into a debate. However, be sure not to seem to be evading the questioner if he or she as a real point. If the question gets to the heart of your talk, it may be best to spend a little more time dealing with the question.

At the end of your talk, if you have done a good job, you should expect 2 or 3 questions. Answer them politely and briefly. If the questioner is aggressive, complement the question but then explain your perspective briefly. Do not get defensive; always be polite. If you do not get any questions, smile and thank them for inviting you. On the other hand, if you do receive questions, make sure to watch the time. If the question period starts to stretch beyond 5 minutes or so either stop the questioning period but offer to take more questions up front or let the rest of the audience know that they are welcome to leave while you continue to take questions from those who are still interested.

Notes by hollandm@cs.cornell.edu