The Leek Group Guide to Giving Talks
This guide is part of the series of guides that have so far covered data sharing, reviewing papers, and writing R packages. The guide is designed primarily for students in statistics and computational biology although some of the ideas will be useful in other disciplines as well.
This guide is a synthesis of some ideas that I got from the following resources:
Uh, cause you got invited?
When you are first starting out you should accept pretty much every opportunity to speak about your research you get. In approximate order of importance, the value of talks early in your career are:
Later the reasons evolve, altough not be as much as you'd think. The main change is that 4 evolves more into "to show people you are a good speaker".
The importance of point 1 can't be overstated. The primary reason you are giving talks is for people to get to know you. Being well regarded is absolutely not the goal of academia. However, being well known and well regarded can make a huge range of parts of your job easier. So first and foremost make sure you don't forget to talk to people before, after, and during your talk.
Point 2 is more important than point 3. As a scientist, it is hard to accept that the primary purpose of a talk is advertising, not science. See for example Hilary Mason's great presentation Entertain, don't teach. Here are reasons why entertainment is more important:
That being said, be very careful to avoid giving a TED talk. If you are giving a scientific presentation the goal is to communicate scientific ideas. So while you are entertaining, don't forget why you are entertaining.
It depends on the event and the goals of the event. Here is a non-comprehensive list:
The biggest trap in giving a talk is assuming that other people will follow you because you follow the talk. In general it is always better to assume your audience knows less than you think they do. People like to feel smart. I have rarely heard complaints about people who went too basic in their explanations, but frequently hear complaints about people being lost. That being said, here are some structural tips. Your mileage may vary.
The last point is particularly important. Usually by the results section people are getting a little antsy. So a completely disparate set of results with little story behind them is going to drive people bonkers. Make sure you explain up front where the results are going (e.g. "Results will show our method is the fastest/most accurate/best ever"), then make sure that your results are divided into sections by what point they are making and organized just like they would be if you were telling a story.
There are only a few hard and fast rules here. I really like Zach Holman's style guide speaking.io and his actual talks. My suggestion is pick one template/style and go with it for a few consecutive talks to avoid costly overhead. See what you like and what you don't, then edit. Here are the only hard and fast rules.
If you have a figure in your talk you should present it in the following way.
If you are giving a stats talk you will probably have some equations. That is ok. But the way you present them is critically important to giving an understandable talk. Here are a few important points:
When giving a job talk you have an additional job on top of the four main goals we talked about above. The goal is to present your complete professional persona to a bunch of people who (mostly) won't know who you are. You should tailor this a little bit to the place you are applying to a job, but don't try to pretend to be something you are not. You can show a little more theory at a theory place and a few more results at an applied place, but don't claim you are proving theorems all the time if you aren't.
You should include both at the beginning and the end of the talk brief summaries of all the stuff you have worked on and plan to work on so they get an idea of who you are in a complete sense. But only talk about one specific project when giving the talk. There is nothing more detrimental to your chances of getting a job than going way over time or not getting to give your whole talk.
Two things that I think you want to convey when giving a job talk are:
Traditionally, people demonstrated #2 by showing that they could prove really hard theorems. At some places, that is still a really good thing to do. Other things that might make you unique are the ability to write amazing R packages that get used, the ability to analyze massive data sets other people can't, the ability to teach courses that students will want to take but the department doesn't have, or ideas about a brand new research area (with some data to back them up).
Before giving a job talk ask around and get a feel for the sorts of things that people have heard about the place you are applying so you can be smart about your choices of how/what to present.
Inevitably you will get hard questions during your talk. The most important point is not to panic and not to get defensive. It is way better to just say I don't know, then to get upset. When you get asked a really hard question you should:
The key is to distinguish what kind of response you are giving. Are you giving a response where you know the answer because you actually looked into that? Are you giving a complete guess where you have no idea? Or, what is more likely, are you somewhere in between?
Most importantly, don't feel embarrassed! Hard questions happen to everyone and if you are polite, explain how sure you are about your answer, and try your best it won't be a problem.
Almost everywhere you give a talk there will be a person who is upset/intense/aggressive. Do not fall into the temptation to be aggressive in return. Answer their first questions politely just like everyone else. If they keep asking really aggressive/lots of questions, you are within your rightst to say: "You are bringing up a lot of good issues, I'd be happy to discuss with you after the presentation more in depth, but in the interest of time I'm going to move on to the next part of my talk". It is ok to keep things moving and to finish on time.
Do it. People will love you for it. If you are the last speaker in a session and others have gone long, adapt and go shorter. The value you gain by making your audience happy >>>> the extra 5 minutes of details you could have explained.
If you have weblinks in your talk you need to post it online. There is no way people will write down any of the links in your talk. Two good places to put your talks are https://speakerdeck.com/ or http://www.slideshare.net/. But then link to all the talks from one, single, memorable site you can show people at the very end of your talk. All my talks are at http://jtleek.com/talks/. You are welcome to send a pull request with your talk there if it is Leek group related, or you can put them on your own short and sweet web link.
I personally like Slideshare a little better these days because the slides are much easier to view on mobile phones and iPads, which is the most likely place someone will read your talk.
Use at your own risk, but these are talks I like or have given (or both :-).