Switch branches/tags
Nothing to show
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
68 lines (40 sloc) 8.79 KB

The fans of PowerPoint are presenters, rarely audience members.

Edward Tufte, the Cognitive Style of PowerPoint

thoughts on giving talks

This is a running list of thoughts and best practices for giving talks. They are a mix of general guiding priciples and specific practices to make talks more engaging, clearer, and more applicable to your audience. But first:

why you should care

Say you're giving a department seminar. Maybe you anticipate 20-30 people to show up to your 50 minute-long seminar. The people coming to your talk are busy but they're putting everything on hold to go listen to you talk for 50 minutes. 20 - 30 man-hours of work are being wasted on you and your talk--that's roughly that amount of work most people get done from Monday - Thursday. Arguably, you have an obligation to give a smart, engaging, interesting talk.

best practices:

have a conversation with your audience. don't lecture

The best talks I've been to are conversations, where the presenter is leading you through their thought processes and telling a story. Most presenters give more of a lecture than a conversation when giving a talk, but lectures are (honestly) often boring. By giving a lecture we also tend to emulate other lecturers and teachers we've seen, most of which don't give good or compelling presentations. Think of a talk as a conversation between you and your audience, not a lecture.

assume people know less than you think they do

Presenters, especially grad students, often feel like imposters that have to prove themselves and how much they know. One symptom of this is that in a presentation they will assume that all these smart people in the room already know all the background to their work. Never assume this.

Most of the time, people know less than you think they do. Even if they know what you're saying, if you're presenting in an interesting way on an interesting topic, it isn't the end of the world to hear some things you already know.

Consider two possibilities. The audience knows what you're talking about or they don't. If they know everything and you go over things that people already, the worst case scenario is that people get bored. When you go on to talk about your research, hopefully people will perk back up. If they don't know everything, they will be lost and confused, and you will never get them back to understand your work and how it fits in to a larger picture.

get excited

When you're excited about what you're talking about, people like your talk. Enthusiam is infectous. Even if a subject matter isn't all that interesting to me, if the speaker is really excited about the subject, I get drawn in. Presumably you're giving a talk about something you've done or something you're interested in. Show the audience that you're excited about it! Walk around the room, gesticualte, vary your voice.

the focus is on you, not your slides

No one likes a talk because the slides were good. Bizarrely, when people are flipped out about giving a talk, they sink all their time into slides and hardly any into practicing actually giving the presentation. A presentation should be about you, the speaker, and not the slides. (On his website David Sterns reinforces this idea by encouraging people to use black slides so the people can't focus on what's on your slides while giving a talk.) Slides are a great tool but they should not be a crutch; they should be used sparingly in a minimalistic way that supports your story and what you say; the slides should not be the story. If the projector were to go out you should still be able to give an engaging talk.

Only include information on a slide that is absolutely necessary and you want your audience to absorb. Don't add tangential information or anything that will distract an audience member from the idea you want to communicate with your audience. Don't add visual 'chartjunk' to your slides. If it's on your slide, it should have a clear purpose. Many people have very complex backgrounds and weird lines and other elements that apparently imbue the slide with some aesthetic beauty. That only people that think this are presenters. Strip down your slides to the bare minimum needed to support what you're saying.

explain your graphs

You are used to looking at graphs you produce. Chances are you've made them fifty different ways and their interpretation is insanely obvious to you. It won't be to your audience. This is the first time most of your audience will see a particular graph you show them, so take some time and make sure they understand it:

  • Tell us what the x axis is.
  • Tell us what the y axis is.
  • Tell us the pattern or lack of a pattern you see, and the stats that back you up. If the null hypothesis is not obvious, tell us what your null hypothesis is.

For the love of God, please do not copy and paste figures from papers. Remake them whenever possible. Make sure someone in the back row can make out the x axis title and all the annotations. Tell us where the statistically significant differences are, and what test you used to determine whether the differences are statistically significant. Never combine green and red in a plot. It's good practice not to include more than two graphs on a single slide. Tell us the sample sizes used and preferably actually show the data (instead of a boxplot or barplot). If you include error bars on your figure, you must explain what they are, otherwise they are totally meaningless (they could be x% confidence intervals, standard error of the mean, standard deviation, credible intervals, etc.).

give credit where credit it due, esp. if they're in the room

If someone made an important discovery or insight to your work, something that you're building off of in your own research, let people know, especially if these are people that will be in the room when you give your talk. Most people like to hear that they're doing good work. You'd want the same thing. Acknowledging others also lets people know that you've done your homework and that you aren't so stuck up that you can't recognize others' work.

don't go over time

This seems really obvious but I find that presenters rarely finish when they're supposed to. Going over time is especially egregious because people have places to be and it's rude to assume they want to hear you for even longer than they initially budgeted.


This is related to the above thought. I find that I have to practice my talk out loud at least five times before giving it in front of people. You want to make sure your transitions are smooth and that you know exactly what you're going to say.

outline your talk first

Talks, papers, any sort of presentation needs a basic organizing scheme. Spend time outlining what you want to say in your talk, then figure out how to say it. Often I find students focus on the latter instead of the former. The result is usually criminally complicated slides that do not play nice with each other: it's difficult to follow the presenter's logical thought process. Outlining your talk before you start making slides protects you from making this mistake.

bonus: use a consistant color theme and use color to highlight important information

a note on graphics

You are used to looking at the graphs in your talks. Chances are you've spent hours looking at dozens of variations of whatever the graph is you're presenting. Your audience will not come armed to your talk with the same advantages.

It’s common to confuse ‘showing data’ with ‘understanding,’ and it’s very easy to show lots of data poorly. Is the point of your talk to impress people with graphics so opaque that they can’t be understood, or to inspire your audience with an interesting finding or surprise them with something they haven’t thought of before?

You want to show data so that people can grasp a simple idea and ultimately guide them in forming an intuition about a process or relationship. If you think, ‘I can’t think of a better way to show this, I know it might be confusing at first glance,’ chances are you don’t actually know what you want to communicate. First, figure that out. Then strip away the figure piece by piece until that point and nothing else comes across. It’s preferable for a figure to do one thing well than several things very poorly.

There is also a tendency to use figures from papers in talks. This is lazy. Sometimes figures from talks are appropriate, but we generally like to pack in as much as possible into figure papers, where the reader can take as much time as she needs to absorb and understand. A talk is different. Multi-panel figures are cumbersome and bit-heavy. Axis labels are hard to read from far away. Nuanced points we make in papers are assumed or glossed over in a talk. Remake your figures and make them simpler and more digestible for your viewer.

This website ( is also a great resource.