How to Make Sure Your Talk Doesnt Suck

David Tong

This is an annotated version of a talk I gave at a summer school for first year graduate students. Anything sitting in a box, like this, summarizes what I said about each slide. Or, at least, what I meant to say.

Its Good to Talk

The talks that you give will determine the jobs that you get

Its important to give seminars. Its almost certainly more important than you realise. Of course, you have to do good work, and you have to write papers that youre proud of. But this isnt enough to guarantee that you will succeed. To be blunt: no one will read your papers. And, if someone does, chances are that the paper includes your supervisor, or other collaborators, as co-authors and your contribution may not be appreciated. Going around and giving talks is the main method that you have to advertise your work, and more importantly, to advertise yourself. I can put this in perspective. I did four postdocs before coming to cambridge. And for three of these I can pinpoint the talk that got me the job --- meaning that, if I hadnt given that talk, I wouldnt have been offered that job. This means that its important to give as many talks as you can, and to practice giving as many talks as you can.

Obvious Corollary

Sign up now!!!

Ways in which you can suck

Assume the Audience is Omniscient

Introduction
As you all know, the remarkable Kontsevich-Soibelman wall crossing formula is given by

d du

Z (u)V

( ;u) K =0

The number one pitfall that students fall into is simply pitching the talk way too high. And its understandable. I remember as a grad student that there seemed to be a chasm between what I knew, and what everyone else knew. And this meant that, obviously, everyone else already knew everything that I knew. Of course, this isnt the case. If youre one year into a PhD it means that youve spent one year studying a fairly specialized topic that isnt common knowledge. And you need to be aware of this when giving talks

Theres no hard and fast rule which determines the amount of knowledge you can assume of your audience. It depends very much on who youre speaking to and you should try to get a good idea of this when preparing the talk. But a good rule of thumb is that anything taught at an advanced Masters level --- say a the level of Part III in Cambridge --- can be assumed to be common knowledge. However, its extremely hard to make a talk too simple. (It can be done. I once saw a famous physicist give a talk at the TASI summer school which Id previously seen at a Borders bookstore)

Acknowledge the expert in the room

by saying But, of course, I dont need to explain that to this audience

Another extremely common pitfall, is to pitch the talk at a single expert in the audience, ignoring the many graduate students and postdocs that are present. Perhaps theres one person in the room whose work plays a fundamental role in your talk. But you still need to explain this for the benefit of everyone else. Just because Einstein is sitting in the front row, doesnt mean that poor Peter Debye in the corner knows what the black ring solution in AdS5 looks like.

Show the Audience Your Back

(and mumble)

Introduction
The problem of bound states in strongly coupled quantum field theory is difficult.
Why are there mesons and baryons, but no pentaquarks, in QCD? If we varied some of the quark masses, how would the spectrum change?

A recent mathematical development due to Kontsevich and Soibelman solves this problem in supersymmetric theories

People relate to the human face much better than the human arse. Look at your audience. Make eye contact. It doesnt matter how clear your explanations are, if you cant engage your audience, people will turn off.
[There followed a miserable performance in which I faced the board and tried to mumble, but completely failed to make my voice inaudible]

Get as Much on That Page as Possible

(and speak really fast)

This ones fairly self-explanatory. Keep the slides clean and simple. Just because you display more and speak faster doesnt mean that people will learn more. The concept of baud rate is important.

Say I know you cant read this, but

Make sure all your graphs are tiny and illegible. Never label the axis. Throw in equations that are just cut and pasted from a paper, with lots of indices and redundant notation that wont actually be relevant for the talk

Use stupid colours. Green on white is always a good idea

Just Read to the Audience

Another common technique is the following. You write every single thing that you want to say on the slide, turn to the screen, and you simply read. Now, this isnt a completely terrible technique. If you read it well, and pause in the right places, it can be quite effective. In fact, Ashoke Sen uses this technique and gives some of the clearest talks Ive seen. But hes Ashoke Sen. And were not. On the whole I wouldnt recommend it.

Do the Striptease

Do the Striptease
Reveal information one tiny tiny piece at a time

Do the Striptease
Reveal information one tiny tiny piece at a time Its because you dont want your audience reading ahead before you say something.

Do the Striptease
Reveal information one tiny tiny piece at a time Its because you dont want your audience reading ahead before you say something. You want them listening to you.

Do the Striptease
Reveal information one tiny tiny piece at a time Its because you dont want your audience reading ahead before you say something. You want them listening to you. Dont you?

Do the Striptease
Reveal information one tiny tiny piece at a time Its because you dont want your audience reading ahead before you say something. You want them listening to you. Dont you?
But thats not the way people learn in seminars. Its not a linear process. People scan backwards and forwards, trying to build the big picture and fill in the gaps in their understanding. Its one reason why a blackboard is better than powerpoint.

Ive noticed that the striptease technique is much less common now than it was ten years ago, and its mostly older scientists who still use it. I suspect that, in the 1970s, there was a guy going around summer schools giving lectures like this one, saying Ive got this great new way to teach people: you cover up the transparency with a piece of paper, and you slowly pull it down to reveal your calculations. Its an educational breakthrough.

How not to suck

Simple, Obvious, Advice

Put a lot of effort into writing a talk. (At least a day). Practice. Be aware. Learn from other peoples talks.

I have a lot less advice to give on how to make your talk good, than on how to avoid making it bad. Some things are obvious. Work hard at it. (It took me one day to prepare this, and a further two hours of practice). Watch other peoples seminars and figure out for yourself what makes a good speaker, and a good talk.

The Structure of the Talk

In this talk, I focussed mainly on presentational issues. With more time, I would have also described the structure of a good talk. However, much of my advice would have been lifted straight out of an excellent article by Bob Geroch, which can be found at gr-qc/9703019.

Forms of Presentation
Decide which method suits you best. (And suits the talk best). My opinion: Blackboard > Powerpoint > Transparencies
Use tex4ppt to write equations in powerpoint Use latexit to write equations on a Mac

Blackboard talks are guaranteed to proceed at a reasonable pace, and much of the information remains on the board for the audience to review. Also, I find that I can deliver blackboard talks with much more energy than powerpoint presentations. But sometimes, like now, powerpoint is simply more appropriate.

Forms of Pointers
Stick > Laser >> Pen or Mouse

Its useful to use a pointer to guide people through calculations and graphs. A solid stick is best. Laser pointers tend to move too fast. Also, if youre nervous, the tiny shakes of your hand are amplified to enormous oscillations of the screen. Never never never point at transparenices on the projector with a pen. Never use the mouse pointer. You will give your audience an epileptic fit.

Coping with Nerves

Take a bottle of water. Or a glass of wine. Learn the first 3-4 minutes by heart, so that you can get into your stride. Invite questions early.
This may sound paradoxical, but its much easier to relax when having a one-on-one dialogue with someone than when standing on stage attempting a monologue.

This is one of the hardest parts of giving a talk. And I dont have any magical advice. Other than to say that nerves arent necessarily a bad thing. They get the adrenaline flowing. Talks I give where Im not nervous typically bomb.

Coping with Questions

Be completely honest. Dont bullshit. Its ok to say I dont know.

The way you answer questions reveals a lot about you as a scientist. Its important to do a good job. Ultimately, there are few tricks that you can use. You simply need to know your subject well and answer clearly. There is an important caveat to the second point above. If you have to say I dont know three times in the same talk, then the game is basically over. From then on, lie through your teeth.

The Golden Rule

Never never never go over time. Never. You might think that, given another five minutes, you can get more across. But youre wrong. No one is listening at that point. You are merely pissing people off.

The End

Dont leave that awkward silence hanging, where the audience dont know if youve finished or are just having a deep thought.

Thats all I have to say. Thankyou for listening.