postUnsolicited Advice on Academic Talks

This is inspired, in part, by Eugenia Cheng's own guide.

  1. Don't start by reading the talk title. Rather, thank the people who made the talk possible: organizer, collaborators, funding bodies.
  2. Capture my attention. Your audience is easily bored and distracted, you need to actively captivate their curiosity.

    To see how it's done, open YouTube or TikTok.

    1. Show me something interesting immediately! An impactful statement, a question. This means also visually interesting: no wall of text (see also below). This applies to all slides but if it's hard for you, it's worth spending effort on these slides specifically: it's also easier because this part of the talk should not be overly technical.
    2. Give me an hook. Tell me what is going on, but leave something out: curiosity is a combination of not knowing and knowing enough to know you are not knowing. In blander terms, tell me why your talk is worth following further.

    In particular, I find the old adage "tell me what you're going to tell me, then tell me, than tell me what you told me", quite stale, unless is a very (1h+) long talk. Else it's just wasted time.

  3. Manage your audience's focus on your slides.

    If you transition to a wall of text, a big diagram, a long derivation, etc., you're overloading your audience attention and they will almost certainly stop listening to you as they try to understand what's going on there. It's a fraction of a second, but it is enough to lose the thread and eat into their attention budget.

    It's really not hard to get around this: at a minimum, break down your slides into smaller chunks. One slide per idea/concept. To keep context around from a slide to another, use progressive overlays, so that the same amount of content is revealed in digestible chunks, timed with your speech.

    I also really like to use overlay transitions that change how the content is presented, e.g. evolving an equation as it is manipulated. These overlays guide attention.

  4. Don't read your slides...

    If you really read every word from your slides, you'll kill your audience with boredom. If you read some of your slides and some not, then you need your slides to be walls of text that, as per above, overload people's attention into guaranteed confusion. In any case, it is a guaranteed bomb.

    The main point here, tied to the above, is that

    your slides are aides to your speech, not vice versa
    It's called a talk, not a read!

    Nevertheless, don't panic: you can use speaker notes! You can take all the words you want to say out of the slides and into the notes.

  5. ...but make your slides readable!

    In 2026, it is expected that your slide deck will be available shorly after your talk. This is both for your fellow conference attendees (parallel sessions means some of your audience couldn't make it to the live) and for future visitors of you or the conferences' website.

    Therefore, your slides should somehow stand on their own without your live commentary. I rarely came across slides that were so in service of the speech as to be unreadable in and on themselves, the problem is usually the opposite. However, remember in beamer and any decent slide-making software, you can always add 'speaker notes' to a slide, and those can help adding missing context (and help you prepare the talk: instead of writing in the slides, write in the speaker notes). A great example of this are some Ingo Blechschmidt's slides.

  6. Speak loud!

    Corollary of this: use the microphone. Unless you have a professional-grade command of your voice projection, people will not hear you in the back, and if you ask do you hear me without a microphone? they don't even hear you're asking.

    Corollary of the corollary: keep your microphone right next to your mouth. If it's handheld, move it with your head if you turn. If it's clipped, clip it as close as you can on your clothes, or keep it in your hand if you can't put it close enough (social media normalized that by now).

    Bonus reverse-advice: if you are audience, or chair, you should also speak loud and use the provided microphone when you ask questions. The question is for the speaker, but the conversation is for everyone, so you should make sure everyone hears you. Often there is a poor soul running around with a microphone, wait for (and thank) them. If you are the speaker, and someone asked a question not everyone could hear (because they did not use the microphone, or are in first row and did not speak to the room), repeat it. Think of the remote folks if the talk is hybrid!

  7. Do not go over time.

    This can be a real faux pas with some people, which will (somewhat rightly) think you are wasting their time. It will also eat into your question time, which is not ideal.

    For me the issue with overtime is it just signals the talk was not prepared with attention: it is a problem you solve by testing the talk just one time, you don't even need a captive audience.

    The way to solve overtime is to cut content and manage the focus of the talk. That is often why 20-30mins talks are more enjoyable that 60min ones: it is really easy to fall into the trap that more time means more detail, or more stuff, rather than better exposition. If you go over time, there is a good chance you are trying to say too much or give too much detail. Remember there is a limited amount of information that you can leave to your audience, so you will have to budget that carefully.