The goal of a perfect job talk is to convince the audience that what you have done is an important piece of work, and that you will continue to produce important math in the future.

Spend the first 20 minutes or so explaining what your field is about -- what are the main questions? what motivates these questions?

In general, classification questions (eg. finite simple groups) or long standing conjectures (eg. Fermat's last theorem) are considered to be the best kind of questions. A conjecture by someone like me will not cut it. A conjecture by your advisor is OK if he/she is famous enough, but it is not great.

You want to relate your work to as broad an area of mathematics as you can.

Explain what you personally have done.

In this sense joint work may fare a bit worse than solo work. On one hand, if a recognized leader in the field has joint work with you, it is interpreted to mean that you are worth that person's time. On the other hand, the department is not hiring your collaborators. So a mixture of joint and solo work is good.

Blow everyone away in the last 5 minutes (but not more) of the talk.

This rule is not universally believed in, and, when believed, is less than readily admitted.

However, I do believe that you don't want the audience to understand everything you've talked about. You want to prevent someone from deciding: "I understand this well. It must be pretty trivial."

You want to convey the impression that you have your own research program.

One more bit of information that may be relevant. In the present crowded job market, the departments are risk averse. They want to hire someone who they are pretty sure is going to get tenure.

Mention contributions made by people in your audience.

This may be a little sycophantic but they're going to be wondering if what you do is of any interest to any of the present faculty.

You may try and keep your talk fairly light and amusing.

A joke here or there can help (though if overdone could backfire). A lot of people in the audience will be wondering whether you'd be a pleasant colleague. Here's your chance.

Do not neglect the basics of a good talk:

clear, large blackboard writing, clear speech (not too fast or slow, directed to your audience, good eye-contact, a degree of humility (pomposity always gets marked down, it seems) but not so humble that you don't sell yourself), not erasing material as soon as it's written, ...

Also, the way your handle questions could be very revealing.

Dress neatly, act confident, don't rely too heavily on your notes.

Be honest.

Here is a story from Nigel Boston: ``I was recently at a talk where the speaker tried to make out like he'd got a better result than he really had. An expert in the audience stopped him in mid-talk to point out that his result meant little without some proof of non-degeneracy. The speaker wriggled a little but he was caught out and his talk left a very poor impression of his work AND him. People will now be wary of his claims, unless they make the effort to check them.''

back to Eugene Lerman's home page

last modified November 28, 1997