For many decades, Patrick Winston, professor in the CSAIL at MIT, gave a one-hour lecture on “How to speak”. You can watch it yourself on YouTube as well as on the official OCW course page where you can also find captions and transcript. There are many summaries of it already written; here is my own take on it.

Why care about how to speak?

Because success in life depends on your ability to speak, your ability to write, and the quality of your ideas - in that order!

What you will get out of this talk

Quality (of communication, and everything else) = f(K, P, T), a function of:

  • Knowledge (most important)
  • Practice
  • Talent (least important)

With the right amount of knowledge, you can be better than others with inherent talents (but less knowledge).

This talk gives you the knowledge for how to speak, providing a toolbox of speaking techniques/heuristics.

Rule of engagement

Winston has a “rule of engagement” of no laptops and no mobile phones during his talk:

We only have one language processor. We will be distracted. We’ll distract others around us. We’ll annoy the presenter who’ll do a worse job.

This should be a rule in all meetings. It’s less obvious in online meetings, but if you’re not going to be paying attention, please at least turn off your video!

How to start

Don’t start with a joke (people are still adjusting to you and won’t be ready for it).

Start with a promise: what knowledge will the audience have gained by the end of the talk?

4 Samples of heuristics when giving talks

  1. Cycle on a subject ➰➰➰

    Tell everything 3x, for it to stick.

    Ideally, in various different ways!

  2. Build a fence around your own idea.

    Avoid confusion with others’ ideas. Make clear what is different about what you’re talking about.

  3. Use verbal punctuation.

    Refer back to the outline, enumerate, point out seams in the talk. Provide landmarks so those who got lost can get “back on the bus” again.

  4. Ask a question.

    It is normal to wait up to 7 seconds for a response (even if it may feel like an eternity to you)!

    Choose the question carefully: too obvious, and people will be too embarassed to answer; too hard, no-one will have anything to say.

When listening to talks, look out for why speakers are effective.

Build your own personal repertoire and style.

The tools

Time & place

Best time: 11 am. Most people awake, not yet asleep, and it’s not after meal-induced food coma.

Place should be…

  • well-lit! Otherwise audience will fall asleep.
  • “cased” (as in bank robbery): make yourself familiar with venue in advance.
  • populated: of the right size, at least half full.


Board (and chalk) for informing. (Slides for exposing, see below.)

Advantages of a board:

  • graphic quality (literally draw connections)
  • writing speed: appropriate to how fast audience is able to follow!
  • target for pointing at with your hands


They make a talk memorable.

Winston’s reason for why board & props are more effective than powerpoint slides when informing/teaching is that they allow for “empathetic mirroring”, where audience can feel themselves writing, experiencing - and this requires the physical world.


Use slides for exposing ideas (as in a job talk or conference presentation).

Critique any set of slides

There are too many, and they have too many words. This is always true!

Slides should be the condiments to what you’re saying, not the main event.

Avoid the basic crimes

  • Do not read out transparencies. Only use a few words that are easy to read.

    Minimum font size: 40 to 50 points (so you aren’t tempted to put too much on the slide - then people read instead of listen)!

  • Be in the image. Standing far away from projection divides audience’s attention.

  • Use images, keep them simple. They should be handles for your ideas. Don’t use goofy clipart.

  • Eliminate clutter. No background pattern. No need for logos. Can even avoid slide titles and the bullet points themselves!

  • Don’t use a laser pointer: you lose eye contact with the audience. Just put arrows directly on the slide instead. This is also true of online presentations.

Review slides

Lay them all out on a table, see if there’s enough “air” in them.

Several programs also support some kind of “overview mode” to look at the whole slide deck at once.

Something the audience doesn’t understand (and that is the point): you can have at most one of this (“hapax legomenon”) per work (whether presentation, paper, or book).

For one of my favourite examples of a talk that uses minimal slides to great effect, watch Edward Felden’s amazing invited talk at NeurIPS 2018!

Special cases

Informing / teaching

Start with a promise.

How to inspire? Show someone they can do it; let them see a problem in a new way; exhibit passion.

How do we teach how to think? Humans are storytelling animals. So provide stories, the questions that need to be asked about stories, how to analyse stories and compose multiple stories, how to evaluate their reliability.



Winston gives two reasons why people fail in oral exams, but this also applies more widely:

  • Failure to situate: you need to put your research in context, what is the history, what is the impact, why is this interesting?
  • Failure to practice: you need to practice with people who don’t know what you’re doing (not your office mates, not your supervisor). Ask them to make you cry.

Technical talks

Job talks (not just in academia): need to show vision and that you have done something - in the first 5 minutes!

How to show your vision: 1) a problem someone cares about and 2) your approach (what is new about it?).

How to show you have done something: enumerate the steps that are required to solve the problem (even if you haven’t done all of them).

Conclude by enumerating your contributions.

Getting famous: getting recognition


We get used to being famous. We don’t get used to being ignored. Ideas are like children, we don’t want to send them into the world in rags. They should be recognized for their value! So it is legitimate to spend time on their presentation.


“Winston’s star” (at the example of his thesis):

  • Symbol (e.g. arch)
  • Slogan (e.g. one-shot learning)
  • Surprise (e.g. only one example is needed, not millions)
  • Salient idea: one that sticks out, not necessarily the most important (e.g. near-miss)
  • Story: how is it done, how does it work, why is it important

How to stop

Final slide

  • Don’t list collaborators (dillutes your contributions); list them on first slide instead.
  • Don’t use “Questions?” / “The End” / URL… just wastes space and your opportunity to leave people with who you are.
  • Don’t list “Conclusions”; audience cares about what you have done.
  • Contributions: sandwich like with job talks, what does the audience get out of it?

Final words

  • Tell a joke. (That way people think they had fun the whole time!)
  • Don’t say “thank you”. (You don’t want to imply people only listened out of politeness!)
  • Salute the audience. Show that you valued the time & place.

Winston finishes the talk by reminding us that how to present and package ideas is important.

In the talk he makes use of his suggestions himself, which makes for great re-watching.

One tool he used without talking about it is the use of small stories to add flavour, without distracting from the content.

If you want to learn more about how to speak, check out Winston’s book Make It Clear. Its table of contents is already a great overview/summary.