Skip to content

Reflections on a talk gone wrong

The first talk I ever gave was at a conference in 1988. (This isn’t the one that went wrong.) I spoke on Constrained maximum entropy methods in an image reconstruction problem. The conference was in England, and I learned about it from a wall poster. They had travel funding for students. I sent in my abstract, and they liked it. All excited, I walked over to a travel agent office, reserved a round-trip to London, walked over to the bank, withdrew $300, and walked back to the travel agent to buy a ticket. Then I went to the Coop and spent another $100 or so on a suit. When I was preparing my talk, Hal Stern gave me some excellent advice. He said, don’t try to impress them or blow them away. Just explain what you did and say why you think it’s interesting. I flew to London and took the bus to Cambridge and went to the conference. I was the only statistician there so I ended up being the statistics expert. My talk was supposed to be 15 minutes long and was scheduled for the session before lunch on the last day. By the time my turn came up, the session was already running 20 minutes late. My talk went OK but I think that most of the audience was eager for lunch by then. (Recall this research finding.)

I’ve given a few zillion talks in the 30 years since, and they’ve mostly gone well. Usually when it doesn’t go well it’s because I don’t know the audience. Once I spoke at an applied math seminar. I was talking all about logistic regression but they didn’t know what logistic regression was! It wouldn’t have been hard for me to cover the background in 5 minutes, but I hadn’t thought to do that, so they didn’t see the point of what I was talking about. Another time I spoke to a group of people who worked in numerical analysis and, again, I didn’t motivate the statistical problems, so it wasn’t clear to the audience what were the difficult parts in our research. Another time I was giving a remote talk—this was a couple years ago, well before coronavirus—and there had been some miscommunication. I’d been hoping for an open conversation with lots of questions, but the organizers were expecting a 45-minute talk. What ended up happening was that I spoke for about 15 minutes and then asked for questions and discussion topics. Nobody had anything to say, so that was that!

Also, sometimes some people will like a talk and some won’t. It’s hard to satisfy everybody, and sometimes it’s hard to know how things went. People will tell you the positive feedback, but you don’t usually hear the negative reviews to your face. Once time I got a warning a couple days before my talk: the organizer contacted me and told me that there were certain topics I should not bring up because they might upset someone in the audience! The punch line is that person didn’t end up showing up.

The most recent time I gave a disaster of a presentation was a few months ago: it was in a session with many other speakers, I had a 20-minute slot and it was running 5 minutes late. 15 minutes should be enough time for a talk, but I somehow wasn’t in the right frame of mind for giving a short presentation. I spent the first 5 or 10 minutes getting warmed up—this can work fine in a long talk, see for example here—but in a short talk it just doesn’t work. I never really had a chance to get to the interesting part. My bad.

Reflections on a talk gone wrong

– Remote talk. No energy of the crowd. Limited opportunity for audience feedback via laughter and facial expressions. (You’d think that seeing faces on zoom would work. But, no, I’ve found that people on zoom have blank faces compared to a live audience.)

– I didn’t lead off with my target message. The audience didn’t know what my talk was about or where I was going. This can be a particular issue when your talk is one of several in a long session.

– Failing to engage the audience from the beginning. Again, a slow warmup can work well for an hourlong stand-alone talk, but not for something shorter.

– Not conveying the importance of the problem. This is related to not making the theme of the talk clearer.

– Poor flow. I didn’t have a plan for how the 15 minutes would go.


  1. Adede says:

    Applied mathematicians didn’t know what logistic regression is? That’s disturbing.

    • Andrew says:


      Maybe they knew it under a different name. Small changes in terminology can confuse people. This happens to me too, when I see a presentation given by someone in another field and I get hung up on details that the speaker has taken for granted.

    • Michael says:

      Eh. Logistic regression really is just completely irrelevant to what 99% of mathematicians are working on. I never even heard of it until I left mathematics. For some reason people expect mathematicians to spend time analyzing data sets like statisticians, but they really don’t do that. Sure, it’d be nice if everyone knew the basics of other fields, but realistically speaking you can’t expect people to remember things that at most they were briefly exposed to in an undergrad class 15 years ago and then never thought about since.

      • mr 80s says:

        Few people outside of statistics, economics and machine learning know what logistic regression is. I had particle physicists who did data analysis at accelerators asking what it is. Unless it’s already part of the culture of the field and students are made to learn it, people will just pick up statistical techniques and vocabulary as they need them.

      • Carlos Ungil says:

        > For some reason people expect mathematicians to spend time analyzing data sets like statisticians, but they really don’t do that.

        The person you respond to said “applied”! Are you suggesting that mathematics has any other applications outside of statistics?

  2. Jonathan (another one) says:

    “Limited opportunity for audience feedback via laughter and facial expressions. (You’d think that seeing faces on zoom would work. But, no, I’ve found that people on zoom have blank faces compared to a live audience.)”

    This. I’ve found that the two things that assure a good speech are (a) planning and (b) modification of the plan as necessary based on live audience reaction. That’s whether I’m giving one or listening to one. Way too many speakers ignore (b).

  3. It’s a terrible feeling (at least for me) to feel like the talk you just gave is a dud. (Like you, I think most of my talks have gone well.) One of my Zoom seminars at a virtual meeting early this summer was like this; the embarrassment still lingers.

    Two things I’ve learned about giving seminars on Zoom:

    (1) scribble on your slides a lot. Seeing motion, since people can’t see you except as a tiny head, seems to help the audience focus. I underline and circle things on the screen I’m sharing. I have a touchscreen, which makes it easier.

    (2) Find someone in the virtual audience who seems expressive, and pin that person’s video. I’m currently teaching a 60 person class, and this helps a *lot*. (In a class, one has some time to find out who these people are — harder in a seminar.)

  4. gec says:

    Part of the “knowing your audience” problem is knowing their expectations for how certain kinds of talk should go. Some places expect a breezy straight-through presentation, some places like lots of technical details, some places like lots of discussion. Especially when talking at departments where you only know one or two people, it’s hard to know which of these applies. And, of course, within an audience there will be people who prefer different structures.

    As an example, I often have to sit through 5-10 minutes of a researcher telling me some cute anecdote or personal story about how their topic affected them before they start talking about their work. I hate this. I’d rather learn about what they actually did. It’s like when you look up a recipe online and you have to scroll for a full minute past some personal story of the food before finding the actual ingredients. Or those clickbait links that say, “you’ll mind will be blown!” Just tell me what you have to say, and I’ll blow my own mind, thank you!

    The problem is compounded by the fact that the work is usually something technical that is at best vaguely related to the original story, so it doesn’t even really tie in.

    Anyway, I know this is a scientific communication strategy that is drilled into many students for some reason, so I don’t blame them for adopting it. And obviously it works for somebody, since it is pretty popular. What I don’t like is that it trains audiences who are used to that style that they can tune out at the beginning and pop in whenever they think they hear something interesting. So when someone gives a talk that actually has content from minute 1, those audiences get lost because they thought they could ignore it.

    I suppose the trick to get around this expectation is to do something shocking or unusual at the beginning of the talk or to ask a question of the audience that forces them to engage. It’s still kind of a waste of time, but at least it is shorter than a rambling anecdote.

    • Andrew says:


      My pet peeve on this was around 25 years ago when it seemed like half the statistics talks were on genetics, and every talk began with a set of cartoons of DNA unfolding and a statement of “the central dogma” of biology. It was so irritating! Just get to the damn statistics problem and give us background as needed. When I gave statistics talks on voting and elections, I didn’t start off with slides on the principles of democracy and the three branches of government.

      • gec says:

        At least now we have the technology to let speakers play the “Mr. DNA” sequence from Jurassic Park or, more apropos to your case, one of the Schoolhouse Rock songs about democracy.

  5. Survivor of many talks says:

    I have given hundreds of talks and people often tell me they are useful. But I remember my very first conference talk at a major conference. I was giving a contributed talk right after an invited talk by a moderately big name in the subfield. The room was packed with a couple of hundred people. When big shot finished his talk and I stepped up the room emptied like someone had activated the fire alarm.

  6. Christian Hennig says:

    The biggest issue I have with myself these days is that occasionally I put too much stuff in my talk. I’m keen on sticking to the time frame, so I’ll hurry, and that’s not good. People will miss stuff but I’m also better in a more relaxed mood. People pick up that I’m in a hurry and that will not motivate them. In a remote talk there are good chances that I won’t find out that it didn’t go well… not too bad for how I’m feeling afterwards but not a true advantage of an online talk really…

  7. Phil says:

    I only recall giving one talk that was an utter disaster, and It wasn’t for any of the reasons Andrew mentioned (or anyone else): I didn’t understand the material myself. It was a subject that was new to me — a sort of value-of-information problem in which I was trying to decide what measurements would help most in constraining a system with many unknown parameter values (something involving airflow in buildings with many rooms) — and I had some preliminary results but nothing really interesting. That would have been OK if I really had a strong grasp of the material, but I didn’t. I’d given talks previously for which I had similar worries, but they always worked out alright, so I assumed this one would too. No, it was terrible. My ignorance of my own material was plain to see, and the whole thing was just really embarrassing.

    Of course I didn’t know any of that when I agreed to give the talk, which was only something like two weeks before. I knew I didn’t have much to say at that moment, but figured I’d learn a lot in two weeks and would be able to do a good job. But other things came up, and, well, I just didn’t make as much progress as I expected.

    One potential lesson from this experience is to not agree to give a talk unless you’re sure you’ll have good results and will know what you’re talking about, but I think that would be the wrong lesson, at least for me: I find that once I agree to give a talk, I’m motivated to work hard and come up with something good, and I wouldn’t want to lose that even though it didn’t work out in this case. I think the lesson that applies to me is that it was OK to agree to give the talk, but when I was putting the slides together a couple of days before the date I should have realized I didn’t have enough and didn’t understand enough, and made some changes at that point. I had plenty of material I _did_ understand, on different topics that were somewhat related, so maybe I could have given one of those talks instead. Or I could have structured the talk so that the first half was stuff that I understood well, on a related but different subject, and then connected that to the new material and talked about the small amount I did understand…and perhaps brought up some of the stuff I didn’t understand but I knew was already known to others; there’s nothing wrong with admitting you’re still learning.

    At any rate, trying to give a talk about a subject I didn’t understand myself turned out to be a big mistake. Who would have guessed?

Leave a Reply