How to give a quals practice talk
Here's a collection of random tips for giving a successful practice
talk, based on my experience giving and attending quals practice talks
at the University of Washington. Following these tips won't guarantee
that you have a good presentation, but they address some of the more
common issues that come up in practice talks, especially for students
just learning presentation skills.
Although some of this is specific to the UW CSE qualifying exam
process, I hope that much of the advice is generally useful. I give
thanks to the UW CSE department for teaching its students excellent
presentation skills and having a student culture that encourages
friends to help each other learn how to give great talks.
Organization and Content
- Think carefully about what information you want to get across.
What are the most important two or three points in your talk? How can you
explain them most clearly? A 30 minute talk isn't long enough to go into
very many details. How does each slide support your main idea?
- What did you do and why should we care? Your talk should
focus on these two questions. Tell us what you did in the first few slides
(even if it's only at a high level; the details will follow) and why it's
important. Be sure to distinguish between your contribution and
- Aim for 20-25 slides for a 30-40 minute talk. Of course, this
number will vary by speaker, but it's a good rule of thumb. Anything more
than that, and you are almost guaranteed to run overtime.
- You don't have to tell us all the details. You can always
refer the audience to your paper for the full details. If you're worried
about it, have detailed slides on hand in case someone asks a question.
- Don't show code on your slides unless you have a very good
reason. "This is how I implemented the Foo algorithm" is a bad reason.
Instead consider demonstrating how the algorithm works on a specific
example. "These code snippets show how you can do in 3 lines what used to
require 100 lines of code" is a borderline acceptable reason, but only if
that code reduction is one of the main contributions of your work. Flash
the code up on the screen for a minute, then move on.
- Don't put too much text on your slides. Also, don't just read
off your slides. If you can do this, you have too much text on your
slides. A good rule of thumb is to use at least a 24pt font for
the body text. I use 32pt.
- Talk through your slides at least once before your
practice talk. Stand in front of a mirror and talk to yourself. Or
borrow a tape recorder and record yourself talking. You'll get a better
feel for how you sound, and you'll notice interesting things such as the
fact that you talk too fast or rely on hand gestures to get your point
across. In addition, you'll find out how long your talk will run before
inflicting it on the nice people attending your practice talk.
- Consider giving two practice talks. If this is your first time
giving a talk about your research, expect to make a lot of revisions to
your slides, especially after your first practice talk. Scheduling two
practice talks also means having two time slots available for people who
can't make one or the other. This may be overkill for some people; I gave
two practice talks and was greatly helped by both, but good speakers won't
find this necessary.
- Attend other practice quals talks, even if they're not directly
in your field. Find out what other people do right and wrong, and think
about how you could do better yourself. Think about how they structure
their talks and question whether it would work for yours.
- Don't use two projectors unless you have a very good reason.
It's worthwhile to examine your talk carefully and see if you can
restructure it to only use one projector. A lot of times when you think
you need both, you don't. Two projectors cause unnecessary overhead in
switching between the two and causing people to focus on each one in turn.
- Make copies of your slides as handouts. It's much easier to
make notes on the handouts than to try to scribble notes on pieces of
paper. If you use PowerPoint, print them 6-up (6 slides per page) and
preferably duplex. Bring enough copies to go around.
- Number your slides. They're easier to reference by number as
we start picking apart your presentation.
- Select a mix of people who are and aren't familiar with your
field. The people who are familiar with the grotty details of the
algorithms/systems/techniques/theorems you're discussing will likely have
detailed comments on the correctness of your work and will ask you probing
questions. The people who aren't will make sure your talk is accessible to
- Find out which of your fellow students give helpful comments
at other people's practice talks. Invite them to yours. Bribes help.
Tessa Lau |
Last modified on 02/21/04