What to Say in a Good Research Talk

John Farrell
Department of Computer Science
James Cook University
Seminar May 1994

Abstract

Giving a good research talk is not easy. Hints on mechanical aspects of the speaking process, such as format of accompanying overheads, behaviour during the talk, and so on, abound, but few authors focus on the content of the talk; i.e. on the process of designing a good talk in the first place. I make some suggestions to assist novice and experienced researchers alike to construct a good research talk. The structure of this paper (though not the content) is heavily based on the paper by Simon Peyton Jones, John Hughes and John Launchbury. This is a useful paper, and I recommend that it be read as a companion to this one.

What This Paper is About

By a research talk, I mean a presentation of some research work you have done. This work is original to you, and hopefully original to your audience as well. Thus my basic premise is that you know something that the audience doesn't, and the purpose of the talk is to tell them about it. My hypothesis is that if you start by saying the right thing, finish by saying the right thing, and say all the right things in the middle, you have given a good talk. Consequently, my guidance is based on choosing how to start, how to finish, and what to do in the middle.

Where to Start

You must start your talk with something the audience can understand. In a research talk, this is usually a statement of the problem. Exactly what you choose as your problem statement depends on your audience.

Generally speaking, your audience will consist of the following types of people:

Graduate Students
Scientists who do not share your research interests
Scientists who do share your research interests

Graduate students, depending on the advancement of their studies, probably don't know much about the forefront of research in your chosen area, nor even a lot of the background, so to speak to them it would be necessary to identify a well-known problem (we need to cure cancer, we need to write software that works). Scientists who do not share your research interests are likely to be familiar with the important problems in the area, so you are able to choose a less general problem to start with (we need to discover the structure of a virus, we need to make programs more readable). Finally, scientists who do share your research interests not only understand the problem you are trying to solve, but will be familiar with previous attempts, will have ideas of their own, and may be conducting similar research. In this case, you may choose a very specialised statement of your problem (we need to determine when eager evaluation may be substituted for lazy evaluation without introducing non-termination).

The composition of your audience will depend on the forum - a departmental seminar is likely to consist mostly of types (1) and (2), an international conference will be mostly types (2) and (3), and a specialised workshop will be mostly type (3). In each case, you should choose a different problem statement, but in all cases it should be a problem that the audience already appreciates to be a problem. If they agree with your choice of problem, they will be more willing to listen to your method of solution. If you start by saying something the audience doesn't believe, they won't believe the rest of what you say either.

Holtzman [1970] calls the starting point on which everyone agrees the Common Premise. The more specific your Common Premise is, the less ground you have to cover to your solution, and the more time you will have to cover your work in between. Your Common Premise should understood by the majority of the audience, and believable with a quick explanation to the others. This gives us the following rule to choose where to start the talk:

Rule 1: Choose the most specific Common Premise suitable to the audience.

Where to Finish

The preponderant tension in construction of a seminar is between the small amount of time available and the large amount that you could say about your topic. No matter how fast you speak or how many overheads you prepare, you will only be able to communicate a limited number of ideas in the time period, so you should aim to present only that many ideas. If you have written a paper to go with the talk, put everything you want to say in the paper; if your talk is successful people will read it anyway. So, as you can't say everything you want to, you must carefully choose the things you do say.

My technique to achieve this is to choose the single most important thing I want the audience to learn, and construct the rest of the talk so as to achieve this single goal. I shall call this goal the Ultimate Idea. The Ultimate Idea should be characteristic of your work, with the result that comprehension of the Ultimate Idea gives a listener an understanding of your research.

The Ultimate Idea is not necessarily, and in most cases hopefully not, a summary of the entire field. Any field which deserves devoted research is too complex to be summarised in a short talk, and you should not aim to do so. However, to keep the attention of the audience, the Ultimate Idea should be novel, unusual or unexpected, and it should be a solution to, or related to a solution to the Common Premise. It should be an idea which demonstrates the benefit of research in your field, and your aptitude at carrying it out. We have the following rule to choose where to finish the talk:

Rule 2: Choose an Ultimate Idea which characterises your work, and demonstrates progress towards the solution of the Common Premise.

Where to go in Between

Having chosen how to start the talk, and how to finish it, the structure of the body of the talk becomes obvious - you present ideas in a sequence which leads the audience naturally and comprehensibly from the Common Premise to the Ultimate Idea. The most important restriction on this progress is:

Rule 3: Proceed directly from the Common Premise to the Ultimate Idea. Do not get sidetracked.

Irrelevant sidetracks will confuse your audience and lose their attention. If you can't bear to leave a sidetrack out, it is worth presenting in a separate seminar, but don't clutter up this one!

The key to presenting successful research talks is to keep it simple, even painfully simple. The people you are speaking to do not have experience with your ideas and notation, and will not comprehend your ideas instantly. However if you can present a series of clear steps which lead from a problem to a solution, your audience will understand and you will have achieved your goal in presenting the seminar.

We now present a series of ideas which help you achieve comprehension and credibility.

Assert Your Authority

It is important to assert your authority to talk on the topic, as listeners not familiar with your work will be wondering whether you know anything or not. One way to demonstrate your qualifications is to give a concise and accurate summary of work done on this particular problem. This may be combined with the Common Premise, by giving a history of work leading from the general problem to the specific problem that you are addressing. You may demonstrate that you are familiar with the field by outlining major developments, and accredit them to known authors in the field. This ``namedropping'' insinuates that you are so familiar with the field that you almost personally know other researchers.

Of course, if you are a well-known researcher in the field, either your audience will know you already or you can start your talk with a history of your own work in the area (``since I proved the Continuum Hypothesis, I have been investigating...'').

Tell The Truth

In a court of law, witnesses are required to tell ``the truth, the whole truth, and nothing but the truth''. This is not the case in a technical seminar, as your listeners are not being paid to hear you speak. The whole truth is not necessary.

However it is often the case when presenting technical research that the reasons behind assumptions, proofs, etc, are very complex and require a great deal of explanation. This explanation should go into an accompanying paper, not into your talk. If you have successfully established your credibility with the audience, you can use up some of it by asking them to believe you. However this is bad practice and should be avoided if at all practical.

It can be sufficient simply to give an intuitive grasp of the reasons behind your dubious statement (this is called ``hand-waving''). This can be done by mentioning a related proof which is well-known, or by demonstrating a crucial part of the proof.

If you can't convincingly avoid this proof, it may be necessary to reconsider the design of your talk. If the proof is so important, it should probably have a talk of its own.

Returning to the court room, the requirements that you tell the truth and nothing but the truth remain. You are giving evidence to an audience which probably contains experts, and they will be less interested if you do not tell the truth. Furthermore, they could expose you as a charlatan during question time.

Rule 4: Tell the truth, but not the whole truth.

A Picture's Worth a Thousand Words

A more fancy and effective method of hand-waving is by drawing a picture. If an audience can see a graphical demonstration of why some statement is true, they are likely to believe you quickly. Furthermore, they are likely to understand the picture far more easily than they are a proof.

Rule 5: A picture's worth a thousand words, and billions and billions of Greek letters.

After the Talk

Any good talk deserves questions, and the lack of them is a sure sign of failure. If you do know what you are talking about, questions are to be welcomed rather than feared. As Lethbridge [1991] points out, most questions will be of a curious and sympathetic nature, and demonstrate interest in your talk. You should answer them with enthusiasm; the questioner may turn out to be a valuable colleague.

We now examine some common types of questions, and techniques for responding to them.

Questions about Material You Covered

Some questions will be about material you did cover in the seminar - the questioner may have misunderstood or misinterpreted what you were saying. Maybe you covered it too fast, maybe the questioner has it confused with some similar work.

In any case, this is the sort of question you should be able to answer - you chose to put the material in your talk, you should be able to explain it. If only one listener has such a problem, you shouldn't worry. If the whole audience needs clarification, you should rethink that part of the talk before you give it again.

Questions about Material You Didn't Cover

Many questions, and in some ways the most flattering, will be requests for additional detail on areas you didn't cover due to time constraints. However do not take this sort of curious question as an invitation to present another five slides!

If the question probes an area of which you do have a wide knowledge, say enough to satisfy the questioner but not enough to bore the audience as a whole. You might like to talk to the questioner later.

If the question probes an area which you are unfamiliar with, it is your job as a reputable researcher to have a reason for not knowing. A standard line is ``I haven't got around to looking at that yet.''

Doubts About Your Common Premise

Some members of the audience may have doubts about or serious disagreement with your common premise. However relevant their doubts are, your seminar is not the place to start a debate on the matter - in your seminar, the Common Premise is taken to be true! With experience in the field, you will learn the ways in which your Common Premises are attacked, and you will have to develop ways to answer these questions. While such questions are not welcome in your seminar, it is your duty as a researcher to consider their merit, so that you remain convinced that your research is worthwhile.

Irrelevant Rambling

Occasionally a questioner will take the opportunity of your question time to sermonise on a marginally relevant topic. In this case, it is often difficult to discern the question, let alone answer it. Such a speaker must be diplomatically shut up - the audience is here to listen to you, not them! Depending on your composure when faced with invasion of your fifteen minutes of fame, you may promise to ``discuss it later'', or give a short sermon of your own to assert your occupation of the pulpit.

Conclusion

In summary, you must design the talk for your audience. Start the talk with something they know, finish it with something they didn't know. In the middle, say things they understand. At all times, remember that you are, and act as if you are, an expert in your field.

References

[Holtzman 70]: Paul D. Holtzman, ``The Psychology of Speakers' Audiences'', Scott, Foresman's College Speech Series, Scott, Foresman and Company, 1970.
[Lethbridge 91]: Roger Lethbridge, ``Techniques for Successful Seminars and Poster Presentations'', Longman Cheshire 1991, ISBN 0582869838.
[Peyton Jones, Hughes & Launchbury 93]: Simon L. Peyton Jones, John Hughes & John Launchbury, ``How to Give a Good Research Talk'', ACM SIGPLAN Notices, Volume 28, No. 11, November 1993.