Preparing a talk

• Getting the timing right for a talk can be tricky; here are a few suggestions.
  • As a very rough guide, aim for no more than one slide for every two minutes in the talk. This may vary considerably if you have many pictures or other slides that go by very quickly.
  • The best way to determine length is by giving practice talks and timing yourself, accounting for the time to change slides. Choose a few slides and note down how much time you should have left when you start that slide. As it is hard to predict how nervousness and audience questions may alter speaking time, this will allow you to make small adjustments on the fly. Have in mind places where you can speed up (e.g. sections that can be omitted) or slow down, as needed.
  • In a conference talk, the session chair will typically indicate how much time you have left. In any setting, make sure you have your own timepiece at hand so that you can monitor your own progress.
• If your talk is based on a paper, be prepared to rethink the entire presentation. You have the advantage of being able to make use of colour and pictures to convey your ideas (as well as speech!), but the disadvantage of having to make the material understandable in real time. Consider the following suggestions with the limitations of the medium in mind.
  • Your presentation does not have to be comprehensive or complete; it is better to cover a little well rather than make the whole paper incomprehensible. In a case analysis you can give one representative case. Proofs do not require low-level details, and algorithms do not require implementation details.
  • Limit the amount of notation and terminology used in the talk. Members of the audience have to keep all this information in mind while processing the rest of your talk.
  • Simplify notation to make it easily understandable when read. Subscripts and superscripts can be difficult to process quickly. Make use of mnemonic symbols whereever possible as an aid to the audience. A three-letter abbreviation is more easily understandable than a random Greek symbol.
  • Where possible, use pictures to help illustrate your points. These can be given in conjunction with detailed information, resulting in both intuition and precision.
  • Repeat material from one slide to the next so that a point can be seen in its entirety or so that an illustration can be seen with accompanying text. A slide should be as self-contained as possible.
  • Provide your audience with "footholds" so that they can regain the ability to follow the talk should they get lost at an earlier point. These might be summaries of previous points, especially those that pertain to the remainder of the talk.
  • Consider the visual impact of each slide. Line breaks can make a difference in readability; make sure all words in a key phrase are on the same line. Unbroken blocks of text are monotonous; use structure and pictures to keep interest high. Choose a font that is big enough to be easily read in the room where you are presenting your talk.
  • Make judicious use of colour. Using a different colour to highlight new terms or important parts of pictures can make a slide more readable; gratuitous changes of colour may have the opposite effect. Not all colours are easily readable as text nor are all pairs of colours easily distinguishable (especially if lines are thin); be careful in your choices.
• The following points cover the main parts of a typical talk.
  • Title slide. For a conference talk, your title slide should include the title of your talk, your name, your department, your university, and your country.
  • Background. Start by putting your results in context. Make sure the audience understands why the subject of your talk is important and interesting, and give ample credit to those who have done earlier related work. It is adequate to give a citation as [Name, date], and it is not necessary to read out all the names when giving the talk.
  • Your results. Make sure you indicate clearly which work is yours and which work is due to others. Using charts to summarize results can make for a succinct way to convey what is known and what is not.
  • Summary and open problems.

Naomi Nishimura's home page