---
title: "How to Speak: Patrick Winston on Giving Talks That Matter"
author: "Patrick Winston · MIT Lecture (shared by @ronycoder)"
description: "The Uniform Code of Military Justice specifies court‑martial for any officer who sends a soldier into battle without a weapon. There ought to be a similar protection for students, because students..."
source: "https://adhenawer.net/posts/original/how-speak-patrick-winston-giving-talks-matter.html"
lang: en
---

TL;DR · key points

1. The Uniform Code of Military Justice specifies court‑martial for any officer who sends a soldier into battle without a weapon.
 2. The quality of your communication — your speaking and your writing — is largely determined by a simple formula: how much knowledge you have, how much you practice with that knowledge, and your inherent talent.
 3. This point came to me decades ago while skiing at Sun Valley during Celebrity Weekend.

## THE FORMULA: KNOWLEDGE, PRACTICE, AND TALENT 0:00

The Uniform Code of Military Justice specifies court‑martial for any officer who sends a soldier into battle without a weapon. There ought to be a similar protection for students, because students shouldn't go out into life without an ability to communicate. Your success in life will be determined largely by your ability to speak, your ability to write, and the quality of your ideas — in that order.

The quality of your communication — your speaking and your writing — is largely determined by a simple formula: how much knowledge you have, how much you practice with that knowledge, and your inherent talent. And the “T” in that formula is very small. What really matters is what you know.

This point came to me decades ago while skiing at Sun Valley during Celebrity Weekend. I spotted Mary Lou Retton — Olympic gymnast, perfect tens on the vault — struggling on a novice slope. I realized I was a much better skier than an outstanding Olympic athlete, simply because I had the K and the P. All she had was the T. With the right amount of knowledge, you can get a lot better than people who may have inherent talents.

Today you will see examples of what you can put in your armamentarium of speaking techniques. Some heuristic, some technique — maybe only one — will be the one that gets you the job. It's a very nonlinear process; you never know when it's going to happen. But by the end, you'll have been exposed to ideas that will ensure you get the maximum opportunity to have your ideas valued and accepted.

## RULES OF ENGAGEMENT: NO LAPTOPS, NO CELL PHONES 2:57

Before we start, we need a rule of engagement: no laptops, no cell phones. We humans only have one language processor. If your language processor is engaged browsing the web or reading your email, you are distracted — and worse yet, you distract the people around you. Studies have shown that. And if I see an open laptop, it drives me nuts, and I do a worse job — which means your friends who came to pay attention don't get the performance they came for.

## HOW TO START: THE EMPOWERMENT PROMISE 4:08

Some people think the right thing to do is start a talk with a joke. I don't recommend it. In the beginning of a talk, people are still putting their laptops away, becoming adjusted to your vocal parameters. They're not ready for a joke. It usually falls flat.

What you want to do instead is start with an empowerment promise. Tell people what they're going to know at the end of the hour that they didn't know at the beginning. That's the reason for being there. For example: “At the end of this 60 minutes, you will know things about speaking you don't know now, and something among those things will make a difference in your life.”

## FOUR HEURISTICS: CYCLE, FENCE, PUNCTUATE, QUESTION 5:22

The first heuristic is to cycle on the subject. Go around it, go around it again, and go around it a third time. People sometimes say “tell them what you're going to tell them, tell them, then tell them again” — as if people weren't intelligent. But the real reason is this: at any given moment, about 20% of the audience will be fogged out, no matter what the lecture is. If you want to ensure a high probability that everybody gets it, you need to say it three times.

The second heuristic is to build a fence around your idea so it's not confused with somebody else's idea. If I were teaching a Martian what an arch is, I'd say “this is an arch — and that is not an arch, and that is not an arch either.” In a technical setting: “My algorithm might seem similar to Jones's, except his is exponential and mine is linear.” That's a fence that keeps your idea distinguishable.

The third heuristic is verbal punctuation. Because people will occasionally fog out and need to get back on the bus, you need landmarks where you announce a good place to rejoin. Enumerate: “The first idea was cycling. The second idea was building a fence. The third idea — verbal punctuation.” Numbering creates seams in the talk where listeners can climb back on.

The fourth heuristic is to ask a question. The standard amount of time you can wait for an answer is about seven seconds — it feels like an eternity. The question must be carefully chosen: not so obvious that people are embarrassed to answer, and not so hard that nobody has anything to say.

These are samples of heuristics for your armamentarium. If this persuades you that there is knowledge to be had, then I've already succeeded. Watch the speakers you admire, ask yourself why they're effective, and build your own personal repertoire and style.

## TIME AND PLACE 10:37

What's a good time for a lecture? Eleven a.m. Most people are awake by then, hardly anyone has gone back to sleep, it's not right after a meal, and people aren't fatigued.

As for the place — the most important thing is that it be well lit. Whenever the room is dimly lighted, it signals to humans that we should go to sleep. Whenever I speak somewhere, the first thing I tell the audiovisual people is: keep the lights full up. They often reply, “But people will see the slides better if we turn the lights off.” And my reply is: it's extremely hard to see slides through closed eyelids.

The place should also be cased — in the colloquial sense of how you'd case a bank before robbing it. Whenever I speak somewhere, the first thing I ask my host to do is take me to the room so I can deal with any weirdness. When I arrive, I imagine all the seats filled with disinterested farm animals; that way I know no matter how bad reality is, it won't be as bad as that.

Finally, the room should be reasonably populated. If only ten people show up in a big hall, everyone wonders what's going on elsewhere that's so much more interesting. You want a right‑sized place — doesn't have to be packed, but it has to be more than half full.

## TOOLS OF THE TRADE: BOARDS 13:52

The blackboard is the right tool when your purpose is informing — teaching, lecturing. Slides are for exposing ideas, not teaching them. I use the board for several reasons.

First, graphic quality. A board lets you easily exploit graphics in your presentation. Second, speed. The speed at which you write on a blackboard is approximately the speed at which people can absorb ideas. If you go flipping through a bunch of slides, nobody can go that fast. Third, a board gives you a target — something for your hands to do.

Novice speakers often become suddenly aware of their hands, as if they were private parts that shouldn't be exposed in public. They shove them into their pockets — considered insulting in some parts of the world — or clasp them behind the back. I was once at a convent in Serbia where a nun pulled my hands off my back because in that culture it was extraordinarily insulting. The supposed reason: hidden hands look like they might conceal a weapon. With a board, your hands have something legitimate to do — you can point.

I once watched Seymour Papert give a lecture I thought was terrific. I went back a second time — first time for content, second time for style. I noticed Papert was constantly pointing at the board. And then I noticed that none of the stuff he was pointing to had anything to do with what he was saying. Nevertheless, it was an effective technique.

## TOOLS OF THE TRADE: PROPS 16:35

The custodians of knowledge about props are the playwrights. Decades ago I saw Ibsen's “Hedda Gabler.” It's about a woman in an unhappy marriage, whose husband is about to lose an academic job to a competitor who has just written a magnificent book. As the play opens, a pot‑bellied stove sits on stage with only faintly glowing embers. Throughout the play, the stove is always there, the fire grows bigger and hotter, and you just know that manuscript is going to go into that fire. It's what I remember about the play, decades later.

Playwrights have props figured out, but they're not the only ones who can use them. Seymour Papert once used a bicycle wheel to teach people to look at a problem the right way. Spin the wheel, blow on the edge — which way does it tip? Mechanical engineers immediately invoke the right‑hand screw rule, align their fingers in some configuration, and get it right about 50% of the time — their fancy education reduces them to a coin flip. But put a piece of duct tape on one spot on the wheel. Forget the whole wheel and think only about the piece under the tape. When you blow on it, it goes down. The next piece, same thing. The next piece, same thing. So the wheel must tip that way. You'll never wonder again — because a prop demonstrated the right way of thinking.

Another example from my undergraduate days: Alan Lazarus was teaching conservation of energy in 26‑100. A long wire hung from the ceiling with a massive steel ball. Lazarus pulled the ball up against the wall, pressed his head flat against the wall to steady himself, and let go. The pendulum takes many seconds to swing over and back, and then gently kisses Lazarus on the nose. You have all those seconds to think: this guy really believes in the conservation of energy. Do not try this at home — the first time, there's a natural human tendency to push rather than let go.

Students always say: more chalk, less PowerPoint. Why? My lunatic fringe view is that it has to do with empathetic mirroring. When you watch me write on the board, the mirror neurons in your head become actuated — you can feel yourself writing. When I gesture about the steel ball going this way and that, you can feel the ball as if you were me. You can't do that with a slide or a picture. You need to see it in the physical world.

## TOOLS OF THE TRADE: SLIDES AND THEIR CRIMES 24:49

I was once at Logan Airport after a miserable conference and a horrible flight. I was having a cup of coffee when a stranger approached: “Are you Professor Winston? I'm on my way to Europe to give a job talk — would you mind critiquing my slides?” “Not at all,” I said. “You have too many and they have too many words.” “How did you know?” he asked, thinking I'd seen his talk. I hadn't. It's always true: always too many slides, always too many words.

Let me walk through the crimes of slide preparation.

Crime one: reading your slides. People in your audience know how to read, and reading will annoy them. Have only a few words per slide, in an easy‑to‑read font. If you're standing way over here and the slide is way over there, the audience falls into a tennis match, swinging their attention back and forth. Slides should be condiments to what you're saying, not the main event.

How do you fix this? Get rid of the background junk — distraction. Get rid of the words — when the slide has only a sketch, everything I said about a previous version I'm no longer licensed to say, because it's not on the slide. Get rid of the logos — you don't need them. Get rid of the title — I can tell you this is the rules‑for‑slides section without it being on the slide.

By reducing words on the slide, you're allowing people to pay more attention to you. We have only one language processor — we can either use it to read or to listen. A student of mine once ran an experiment: teaching web‑based programming ideas where half the information was on slides and half was spoken, with a control group where the halves were reversed. What did subjects remember best? What they read on the slides. When slides have a lot of material, audiences stop paying attention to the speaker. One subject in the after‑action report said: “I wish you hadn't talked so much. It was distracting.”

Eliminate clutter. No reason for bullets on most slides. The “too many words” problem is a consequence of a crime Microsoft committed by letting you use fonts that are too small. Have a sample slide you can use to determine the minimum readable font — probably 40 or 50 point, and 35 is about as small as you can go, not because it's unreadable, but because you're probably using it to cram too many words.

Crime: the laser pointer. In the old days, wooden pointers turned presentations into baton‑swirling contests. Now we have laser pointers — a wonder more people aren't driven into epileptic fits. When you use one, you can't avoid turning your head toward the screen. You lose eye contact and engagement. A student once said to me while we watched a talk: “We could all leave and he wouldn't know.” If you need to identify something in your image, put a little arrow on the slide itself and say, “Look at the guy at the end of arrow number one.” No laser needed.

Crime: too heavy. When people ask me to review a presentation, I ask them to print it out and lay it on a table. It's easy to see whether the talk is too heavy — too much text, not enough air, not enough white space, not enough imagery. A good talk by contrast has air in it, mostly pictures, only three or four text slides, and when you reach those, you give the audience time to read them.

The “hapax legomenon” slide: the kind you can get away with exactly once per presentation, per paper, per book. The famous diagram of the complexity of governing Afghanistan is one example — impossibly complex, and that impossibility is the point. But you can only have one.

Do these crimes actually occur? Yes. Hands in the pockets. Crimes of time and place — the Bartos Theater at MIT, reached by descending into a dark corridor that makes you wonder if there are torture implements around the corner, and then opening into a gloomy space perfect for watching a movie but not for giving a talk. I once attended a talk in Stata where the speaker stood far from the slides, used a laser pointer, and displayed the 80th slide — the first of ten conclusion slides. The sponsor was reading his email. The co‑sponsor was examining the lunch menu. Another attendee looked like he was paying attention — but that was only because it was a still picture.

## INFORMING AND INSPIRING 37:00

In an informing talk, start with a promise, as I did today. Then comes the question of how to inspire. A few years ago our department chairman asked me to give this talk to new faculty and emphasize what it takes to inspire students. I hadn't thought about it, so I ran a little survey. Incoming freshmen told me they'd been inspired by a high school teacher who told them they could do it. Senior faculty said they'd been inspired by someone who helped them see a problem in a new way. But everyone, without exception, said they'd been inspired when someone exhibited passion about what they were doing.

That's easy for me — I do artificial intelligence. How can you not be interested in AI? If you're not interested in artificial intelligence, you're probably not interested in interesting things. When I lecture in my AI class, it's natural to talk about what I think is cool and how exciting some new idea is.

During the promise phase you can express how cool something is. Here's an example from a lecture on resource allocation — the kind of math you'd need to schedule airplanes or factories. I open with a demonstration of coloring the states on a U.S. map so no two neighbors share a color. The naive program runs slowly — we're not going to wait until it finishes, because the sun will have exploded and consumed the Earth before it's done. But with a slight adjustment, which you'll understand in the next 50 minutes, here's what you get — instant solution. Isn't that cool? You have to be amazed by something that goes from longer than the lifetime of the solar system to a few seconds.

## TEACHING PEOPLE TO THINK 39:55

Ask faculty what the most important thing they do is, and they'll say, “Teach people to think.” Ask them how, and you get a blank stare. My belief: we are storytelling animals. We begin developing our story understanding and story manipulation skills with fairy tales in childhood and continue through professional schools — law, business, medicine, everything — and throughout life. If that is what thinking is, then teaching people to think means giving them the stories they need to know, the questions to ask about those stories, mechanisms for analyzing them, ways of putting them together, and ways of evaluating how reliable a story is.

## ORAL EXAMS: SITUATE AND PRACTICE 41:36

Oral exams used to be a lot scarier; in the old days, reading literature in a foreign language was part of them, and the failure rate was high. Looking back, the most common reasons for failure were failure to situate and failure to practice. To situate means talking about your research in context: “This is a problem being pursued all over the world. There's been no real progress in 30 years. Everyone is looking for a solution because it would impact many other things.” That's situating in time, place, and field.

Practice matters — but don't practice by showing your slides to the people who share your office. If they already know what you're doing, they'll hallucinate material in your presentation that isn't actually there. Same with your faculty supervisor — they know your work, so they hallucinate too. Gather some friends who don't know what you're doing, and begin the session by saying: “If you can't make me cry, I won't value you as a friend anymore.” Then the actual faculty oral will feel easy.

Also: the amount of flack you get from someone is proportional to age — inversely. Older people understand where they are in the world; young people are trying to show old people how smart they are, so they are viciously picky. Whenever you can stack your examining committee with gray hair, do it.

## JOB TALKS: VISION AND ACCOMPLISHMENT IN FIVE MINUTES 44:36

Years ago, I was sitting at a bar in San Diego as a member of the Navy Science Board, with Dolores Etter from Colorado (who'd written 21 books to my 17 — enough to make me spit) and Bill Weldon from Texas (who knew how to use rail guns to drive steel rods through tank armor). I asked: “What do you look for in a faculty candidate?” Within one microsecond Dolores said: “They have to show us they've got some kind of vision.” Quickly followed by Bill: “They have to show us they've done something.” How long does a candidate have to establish these two things? Five minutes. If you haven't expressed your vision and shown you've done something within five minutes, you've already lost.

Vision consists of a problem somebody cares about and something new in your approach. My stump‑speech version: “The problem is understanding the nature of human intelligence. The approach is asking what makes us different from chimpanzees and Neanderthals. Is it merely a matter of quantity, or do we have something fundamentally different? The answer is yes — we do. We are symbolic creatures. Because we are symbolic, we can build symbolic descriptions of relations and events, string them together into stories — and because we can make stories, that's what makes us different.” That's how I start most talks about my own research.

To show you've done something, enumerate the steps needed to solve the problem. You don't have to have done all of them — you can simply show what needs to be done: specify the behavior, enumerate the constraints, implement a system (because engineers don't believe we've understood something until we can build it), and demonstrate. Then conclude by enumerating your contributions — a mirror of the steps. That's a general‑purpose framework for a technical talk.

## WHY YOU SHOULD CARE ABOUT GETTING FAMOUS 48:05

I thought about this at a fundraising event for Save Venice. I was seated next to the late Julia Child. All evening, people came up asking her to autograph things or telling her she'd changed their lives. Eventually I asked: “Ms. Child, is it fun to be famous?” She thought about it and said, “You get used to it.” But then it occurred to me: you never get used to being ignored.

Here's how to think about it: your ideas are like your children. You don't want them to go into the world in rags. So you owe it to them to have the techniques, mechanisms, and thoughts about how to present ideas so they're recognized for the value that's in them. That's why packaging is a legitimate concern.

## THE WINSTON STAR: FIVE S'S FOR BEING REMEMBERED 51:22

Every element starts with S.

Symbol. Associate a visible symbol with your work. The arch from my PhD thesis is an example — and the arch was also the symbol.

Slogan. A phrase that gives people a handle on the work. For the arch research, the slogan was “one‑shot learning” — the program learned something definite from every example presented.

Surprise. An unexpected claim. The surprise: you don't need a million examples; you can learn from one, if you're smart enough to use that one example appropriately.

Salient idea. Not “important” — sticking out. Some theses have too many good ideas and nobody can tell what the thesis is about. Pick one idea that sticks out. In my case: the “near miss” — something that isn't an arch but doesn't miss by much.

Story. Tell the story of how you did it, how it works, and why it matters.

Accidentally, my arch work had all five. I didn't know why it got mildly famous until years later.

## HOW TO STOP: FINAL SLIDE 53:58

The worst final slide is a giant list of collaborators. Divide by a thousand, and your nice piece of work becomes tiny. Recognize your collaborators on the first slide, not the last.

A “Questions?” slide is the worst possible way to end — it can sit up there for 20 minutes, squandering prime real estate and an opportunity to remind people who you are.

“The End” — I often see this, though I've never seen it written down. It wastes the opportunity. “Any Questions?” with a thinker silhouette — even worse.

A “Conclusions” slide seems reasonable but falls short. Nobody cares about your conclusions; they care about what you have done. Your final slide should be labeled “Contributions.” It mirrors the way a job talk ought to be a sandwich — open with what you'll do, close with what you did. That final slide stays up there while people ask questions and file out, so make it the one that shows your contributions.

## HOW TO STOP: FINAL WORDS 56:55

You could finish with a joke. It works. By the end, people are adjusted to your voice and ready for a joke. I once asked my colleague Doug Lenat, a fantastic speaker, for his secret. He said: “I always finish with a joke. That way people think they've had fun the whole time.”

Do not end with “thank you.” It won't send you to hell, but it's a weak move. It suggests everyone stayed out of politeness, longing to be somewhere else, but enduring out of good manners — and that's what you're thanking them for. Once wild applause has started, you can mouth a silent thank you. But the last thing you say should not be “thank you.”

“But doesn't everybody do it?” What everybody does isn't necessarily the right thing. Political speeches show how to end without thanking: Governor Christie closed a Republican keynote with “…we will stand up once again for American greatness, for our children and grandchildren. God bless you and God bless America.” A benediction ending. To keep things nonpartisan — Bill Clinton's Democratic keynote in the same cycle: “If that is what you want, if that is what you believe, you must vote, and you must re‑elect President Barack Obama. God bless you and God bless America.” Watch him closely: at one point he seems to press his lips together, forcing himself not to say thank you, and he throws in a little salute.

I could conclude by saying, “God bless you and God bless the Massachusetts Institute of Technology,” but it might not work quite as well. The point is: you don't have to say thank you. There are other things you can do.

People figure this out over time. In the good old Latin Mass, the priest closed with “Ite, missa est” — approximately: “Okay, the Mass is over, you can go home now.” At musical concerts, you know it's time to clap not at the end of the song but when the conductor walks over and shakes hands with the concertmaster. Those are conventions that tell you the event is over.

One more possibility: salute the audience. Say something about how much you've valued your time with them. “It's been great fun being here. It's been fascinating to see what you folks are doing at MIT. I've been stimulated and provoked by your questions. It's been really great, and I look forward to coming back on many occasions in the future.”

## CLOSING SALUTE 1:02:24

There it is. I'm glad you're here — because by being here, you've demonstrated an understanding that how you present and how you package your ideas is an important thing. I salute you for that. And I suggest that you come back again, and bring your friends.