The Professor Jim Yorke (University of Maryland)
Checklist for talks with overhead projectors:

A live test audience

All mathematicians and scientists hear many unintelligible talks by professional speakers (i.e., professors) who think they are being clear. I believe these occur because the speaker did not get adequate feedback while preparing. All talks should be tested on live audiences. In our seminar students must Pre-test talks with an audience of students before the announced version is given, even if that is also a practice for a talk at meeting. Speakers are frequently amazed to find out that their basic material is not known by the audience.

The test talk is primarily for intelligibility, not timing, and is to see what kind of difficulty an audience would have with your explanations. Hence the test audience should speak up when they do not understand what is being said, not hold questions until the end.

Feedback: When you sit through a lecture that nobody understands, as often happens, it is almost always because the speaker does not know that he/she is being unintelligible. When you give a talk, do NOT ask if it was OK. That will usually yield a positive response. Instead ask what was wrong with the talk. Tell people you are going to give it again and need feedback. It is good if you can tell them this before you give the talk so they take notes.

If you give a practice talk that is judged a failure, don't worry about it; just fix it. It is the final talk that counts. I once tested a talk on test audiences before the real event.

The Main Point

Have a clear idea of what the main point is in your talk and tell the audience early in the talk. You may get lots of questions because some of the audience is lost, and you may run out of time before the talk is over, so be sure that your main point is early enough in the talk that it does not get lost if the talk ends early.

Outline? Pare down to the Essence

Don't waste time in an early talk outlining your talk. People who present outlines in short talks rarely get through the material outlined. More generally a major art in giving a talk is seeing how much material can be omitted. For each transparency, ask if there is a way not to present the material or explain it more briefly. You should be aiming at explaining one central idea or achievement. Talks sometimes begin with unessential material. When the audience interrupts with numerous questions about what the speaker has said, the talk never gets to the essence and is a disaster.

Practice with Overhead Projectors

Practice using two overheads if you can have two in the final talk.

There is a strange resistance among many speakers to use more than one overhead projector, as if the audience could perfectly remember the previous material. Yet if they were lecturing on a black board and had only one meter-square panel, they would find it constricting to have to erase the board every couple of minutes. The audience will appreciate the use of two even if your initial reaction as speaker is that only one is needed.

The simplest approach is to alternate between projectors, putting the first transparency on one and the second on the other. When one projector is brighter or better than the other, just put the new transparency on the better projector and the previous one on the backup. When a particularly central transparency is encountered it can be left for a longer time on the backup projector.

Your test audience should criticize you when you if you stand in the way of some of the audience, blocking either screen. Point to material on the screen if possible and not to the transparency on the overhead projector since then you will be blocking your audience's view. Practice with an audience distributed to the far left and right and in front and do not block any of the audience.

Credits

Make sure the audience knows what YOU have done and what part of your material has been done by others; in particular distinguish background material from your material. Don't be shy about claiming credit.

Give credit to people who did the background work by name at least; (it is usually cumbersome to give a full reference, tho it is good to have available.) If you don't give credit, someone will think you did that work and are claiming credit it for it.

Voice

Speak uniformly loudly. Some speakers drop their voices at parts of sentences. It does not suffice to say 90% of a sentence loudly. When answering a question of someone sitting in the front, keep you voice loud so that they whole audience can hear you, and remember that many have probably not heard the question.

Eye contact

Make frequent eye contact with the audience. See if you can interact with the audience.

Transparencies

Transparency titles

Each transparency should have a title. It should tell the audience what they are looking at, what the point is of the transparency, and it should be underlined so that it is clear it is a title.

Double Size type

ALL type on a transparency should be visible from the back of the room. Never use unenlarged typescript in a transparency. It might work in some small rooms with some overheads but usually it fails. Type should be enlarged from regular print by at least a linear factor of 2, so your font should permit at most 40 characters for the width of the page.

It is important to keep the total amount of text on a transparency small; but also assume that the audience is much less likely to understand a point that you say than if you write and talk about it.

If a figure you copy has some small type (such as scales), wipe it out before making the transparency and re-write the material by hand in large visible letters.

Erase material that you do not want to draw to the attention of the audience to.

Color

Use some color. If your transparencies are black and white, then underline in color, use color boxes and write over some letters in color.

Number your transparencies

Number your transparencies so that when they get mixed up, they are easy to sort for your next presentation.

Many speakers seem more concerned about keeping transparencies in order than in keeping them in front of the audience; they remove a transparency the second they finish talking about it, giving the audience no time to really see and understand what was written; then they take several seconds finishing some comments and putting down the transparency, and picking up a new one. Instead you should pick up your next transparency before you take the previous one off, so you can keep your material in front of people for a maximum time.

Apologies?

Do not apologize in your talk for anything, for example, for having forgotten to say something earlier. If you forgot, just say "Now is a good time to tell you ...". They won't know that you think you screwed up. You are supposed to give an appearance of mastery. Do not apologize even for having the sniffles, since you just draw the audiences attention to it. If you have failed despite the above paragraph above to get to your main point, don't tell the audience that you didn't get to your main point. That is just telling them that they wasted their time listening to you. Hopefully they enjoyed what they did hear.

Repeat key ideas

Find the places where you present ideas, equations, or definitions at the bottom of a transparency; after you introduce it in your talk, the audience might have 5 seconds after you go over it to understand before you grab the sheet from the screen; so figure out how to extend the time they have to understand it. For example discuss some illuminating aspect of it.

If you plan to cover up the bottom of a transparency and show only the top, make an extra copy with just the material on the top; hence the next slide will duplicate the top and continue on. One slide can thus be broken into several stages. Many audience members dislike seeing large parts of the screen blacked out; it creates a dark room with only a slice of transparency showing.

Graphs and Figures

Your graphs should be intelligible. If you are using a logarithmic scale, don't write the values of the logs. write for example 1 10 100 1000 Etc. When people instead write logs base e instead of actual values, they are presenting material they cannot expect the audience to understand. In such a talk would you know what exp(6) is for example? (Ans.: exp(3) is about 20 so exp(6) is 400)