Your top priority is to be understood by the audience, so always aim for clarity.  Make sure the logic of your talk is simple, and that you explain your methods and results in appropriate detail.  Think carefully for each topic you discuss what level of detail, high or low, is appropriate.  Organize your presentation logically, not chronologically.

Give less detail on topics that are well-known to the audience already, such as general background and motivation, or unsurprising conclusions.  Spend most of your time on new ideas, new methods, new results, and new "lessons learned."  Give more detail on

• new algorithms, step by step
• the specific intuition behind individual algorithm features
• exactly why a new algorithm is expected to be better than a previous method
• algorithm time and space complexity, and scalability
• the explanation of equations
• the design and logic of experiments
• a quantitative description of sample datasets and experimental runs
• the results of related experiments by other authors.

Do not just skip over mathematics.  Do present mathematical equations and explain proofs, but avoid getting lost in formalism.  Do not automatically reuse mathematical notation from the paper you are presenting; if possible, use simpler notation.  For every equation presented, give an informal but careful precise explanation of what it says.  Plan your explanation of each equation in advance and make notes for yourself.  In particular, make sure that each variable used in the equation is defined carefully and explicitly.  In your notes, have an explanation of what each variable represents.

To keep the audience interested, more than half of your slides should have text and a figure, table, diagram, chart, or picture.  Do not have many consecutive slides that are text only.

Keep all graphics as simple as possible, with no distractions.  Charts should have clear legends and labels on axes, etc.  In a diagram, each different type of box and arrow should have a clear semantics.  Be sure to use colors or shades of gray that are very easy to tell apart.

Regardless of what display technology you use, check in advance whether your slides are readable when projected.  They can look very different from what you expect.  If you use a laptop, make sure that your presentation is "ready to go" without fumbling.  If you use printed transparencies, make sure that they are in the right order and numbered.

Every slide should have a title that is less than one line long.  Except for the title, written text should consist of entire sentences as opposed to "telegraphese."  Each sentence should have very simple grammar, with a clear subject and main verb.  Avoid sentences in the passive, which have a subject grammatically but not conceptually.   For similar reasons, avoid pronouns.

Eliminate words that are not absolutely necessary.  If a sentence or slide is too complex or contains too much information, split it into two.  Avoid sentences of three or more lines.

Use blank space between sentences to increase clarity, and use indentation to indicate nesting of ideas.  Do not waste space on bullets for top-level sentences, but do use bullets for nested sentences if appropriate.  When a sentence must occupy two or more lines, think about where to place each line-break so that the sentence is as easy to read and understand as possible.

Face the audience as you speak, with both your body and your head.  Do not face the screen or projector, and do not stand in front of the screen.  When you display a slide, check that no text is projected off the screen.   Then point to items on the screen, not on the projector.

Don't read sentences word for word from slides.  Speak at a normal speed, not too slowly or too quickly.  Try not to speak hesitantly or softly.

For more suggestions and guidelines on giving an academic talk in computer science, see Ian Parberry's speaker's guide.
 

Most recently updated on June 21, 2001 by Charles Elkan, elkan@cs.ucsd.edu