Daniela Witten writes:

Devoted readers of this column will know that I am a parent to three young humans. And while I love them beyond measure, the fact remains: keeping three humans alive (or four, depending on your take on husbands) is basically my limit. I am up to my ears in mammals. Now, please don’t @ me with comments about how a pet bunny is critical for a child’s emotional development, or how guinea pigs are underrated. I meant what I said: I’m tapped out, mammal-wise.

My kids are perpetually disappointed that I won’t get them a dog, but they find solace in their very close relationship with Callie*, our friends’ miniature goldendoodle.

Callie is a princess among poodle mixes, a queen among canines, and a wonder among wolf descendants. When she enters the dog park, the other pooches stop and stare. She is the belle of her bougie neighborhood. She makes the other designer dogs look like run-of-the-mill rescues. Simply put, Callie is a sight to behold, and she knows it.

Now, enter Ralph and Bess, our dear friends and Callie’s owners—er, I mean, dog parents. Though Ralph and Bess are equally besotted with Callie, they interact with her in different ways.

When Ralph feels cold, he puts Callie across his lap like a blanket or under his feet like slippers. When he feels bored, he teaches her to walk on her hind legs or to roll over for treats. When he feels tired, he takes a nap with Callie as his pillow.

Bess interacts with Callie differently. When Callie feels cold, Bess snuggles Callie on her lap. When Callie feels bored, Bess has her do tricks for treats. And when Callie feels tired, Bess finds a place for Callie to take a cuddly nap.

Basically, Ralph interacts with Callie to make Ralph feel good, whereas Bess interacts with Callie to make Callie feel good. Who does Callie prefer? Not surprisingly, the answer is Bess.

This is actually not a column about dog ownership (a topic that I, as a non-dog owner, am in no way qualified to discuss). Instead, it’s a column about how to give a research talk. In this metaphor, the audience is Callie, and you get to decide whether to be Ralph or Bess. Will you be a Ralph, and make yourself feel good, or a Bess, and make your audience feel good?

To answer this question, it’s worth considering the reason you’re giving a talk. Is it:

1) To make yourself feel smart?

2) To make the audience think you’re smart?

3) To teach the audience about your research?

I’m only going to say this once: never let the answer be 1). If you want to feel smart, then please find another way to scratch that itch, like completing the Sunday crossword or studying Ancient Greek or engaging unsuspecting and disinterested strangers in arguments about politics while on public transit. But, I beg of you, never a research talk.

What about reason 2)? Also no. Your audience does not care to watch your performative show of intelligence.

So, via process of elimination, the reason to give a talk should be 3): to teach your audience about your research. We will now consider how Ralph and Bess might approach this goal, and which strategy will be most effective.

Ralph copy-pastes from a paper.

In most cases, our talk is based on a paper, and the fastest way to prepare slides is to copy-paste chunks of text (perhaps transformed into bullet points) and technical results.

However, while long stretches of text and detailed technical results are a good way to share information in a paper, they are problematic in a talk. It is hard for the audience to keep track of more than a minimal set of notation in a talk (an audience member can’t flip back to a previous slide if they forget notation) — and talks with a lot of text and details are boring! Yes, it’s easy to create a talk by copy-pasting… but it rarely leads to a great result.

Bess creates a visual to explain her main ideas.

When I teach courses on presentation skills for grad students, I assign an exercise that is simple in concept but challenging in execution. I ask students to take a statistical idea, and create a 10-minute presentation to explain it with visuals and without any words or equations.

For instance, consider a two-sample t-test. I’m not asking for students to, say, present simulation results showing the power of a two-sample t-test: rather, I’m asking them to explain the concept and logic underlying a two-sample t-test, without any words or math. This requires really thinking about what the key point of a two-sample t-test is, and communicating that—and only that! —through a series of visuals. For instance, I might make a series of visuals to represent (i) a sample from a N(0,1), (ii) a sample from a N(c,1), (iii) the two sample means, (iv) the difference between the sample means. Then I might create an animation iterating through (i)–(iv) a whole bunch of times, to illustrate (v) convergence to the normal distribution. Remember, no words and no equations: only visuals.

I really do recommend that you try out this exercise for your own research project. Yes, your research is undoubtedly more complicated than a t-test, and I fully understand that this won’t be easy. But if you think hard enough, then you can surely distill the key point—the creative spark, the interesting idea, the thing you really want your audience to take home from your talk—down to a series of figures, which you can then use both to introduce this point in your talk, and as a “running” visual aid throughout your talk.

Ralph tries to explain it all to the audience.

You cannot squeeze an entire 30-page paper into a 45-minute talk. I promise that if you try, then your talk will be neither understandable nor interesting. Instead, think about the most interesting couple of points of your paper, and explain them as clearly as possible. 

I personally think that there’s no place in a talk for detailed proofs, or for more than one carefully chosen simulation result. Instead, focus on the big-picture main idea: the scientific or statistical problem that you’re solving, and the creative insight that went into solving it.

Bess thinks about what the audience knows.

There’s a big difference between the set of people who read my papers and those who attend my talks. For instance, if I’m giving a department seminar, then members of the department who are working in completely unrelated areas may attend out of general interest. By contrast, typically only people in my research area will read my papers.** So I need to target a department seminar towards a much broader audience than my papers.

To achieve this, I focus my talks on the big-picture main idea: I spend a substantial portion of the talk explaining what problem I’m trying to solve, why it’s important, what solutions were available before my work, and so on. Establishing this framing will substantially decrease the time available to present my own research contributions (and means that I certainly will not have time to present every single detail!). Furthermore, it’s hard to do, since this requires zooming farther out on the problem than I do in my papers. But it’s worth it, since giving a talk that almost nobody understands is a waste of everybody’s time: the audience’s, and my own.

Before I close, I’d like to revisit reason 2). I’ve said that the goal of a talk is not to show the audience how smart you are. But there is perhaps one exception: if you’re very junior and are giving a job talk. Then the primary goal of your talk should still be 3), but you could reserve the last five minutes for technical material to “wow” your audience with your raw intellect, in case there are curmudgeons in the department who judge a talk by how technical the content is. Five minutes is plenty of time for this, and it should come at the end of the talk: if you put it earlier then you’ll lose the interest of the rest of your audience.

There is no one way to give a good talk, but there are many paths towards a bad talk. At the end of the day the question is: did you consider your audience’s needs, or your own? Be a Bess and not a Ralph. Callie (and your audience) will appreciate it.

Daniela Witten lives in Seattle with four human mammals, and—coming soon—a hermit crab. Her children can’t wait.

* All names changed to deter would-be dognappers.

** To make this more precise: of course, I hope that some people outside of my immediate research area read my papers! But if they find that they are missing background knowledge required to understand my papers, then they can pause and fill in the gap before continuing with my paper. This is not possible during a talk.