Skip to content

How to Speak

Channel: MIT OpenCourseWare
Video: https://www.youtube.com/watch?v=Unzc731iCUY

MIT professor Patrick Winston's classic lecture on the mechanics of effective speaking: a toolbox of concrete techniques for talks, job interviews, and getting your ideas recognized. Core claim: speaking is a learnable skill governed mostly by knowledge and practice, not inborn talent.

Key points

  • Success formula: your success in life is determined largely by your ability to speak, your ability to write, and the quality of your ideas - in that order. Communication quality = f(Knowledge, Practice, Talent), where talent (T) matters least. A skier with knowledge and practice beat an Olympic gymnast (pure talent) on the slopes - knowledge beats talent.

  • Rule of engagement: no laptops or phones. Humans have one language processor; a screen splits it, distracting the user and everyone around them.

  • How to start: do NOT open with a joke - the audience is still settling and adjusting to your voice, so jokes fall flat. Open with an empowerment promise: tell people what they will know by the end that they do not know now.

  • Four samples (techniques always on his mind while speaking):

  • Cycle on the subject - go around it three times, because about 20% of any audience is fogged out at any moment.
  • Build a fence around your idea so it is not confused with someone else's (e.g. "my algorithm resembles Jones's, except his is exponential and mine is linear").
  • Verbal punctuation - enumerate and mark seams in the talk so people who drifted off can "get back on the bus."

  • Ask a question - another re-engagement device. Choose it carefully (not too obvious, not too hard) and wait; roughly 7 seconds of dead air is acceptable.

  • Time and place: best lecture time is ~11am (awake, not post-meal, no one going back to sleep). The room should be well lit (dim light signals sleep - "hard to see slides through closed eyelids"), cased in advance for surprises, and more than half full.

  • Tools - boards, props, slides:

  • Board is best for informing/teaching: graphic quality, a writing speed that matches how fast people absorb ideas, and a target for your hands (novices hide hands in pockets/behind back, read as insulting or weapon-concealing). Seymour Papert pointed constantly at the board even at things unrelated to his point - still effective.
  • Props are memorable: cites Ibsen's Hedda Gabler stove/manuscript; Papert's spinning bicycle-wheel + duct-tape trick to reason about gyroscopic torque; the classroom pendulum-to-the-nose demo of conservation of energy ("do not try this at home" - the urge is to push).

  • Why chalk/props beat PowerPoint: his "Lunatic Fringe" theory of empathetic mirroring - mirror neurons let the audience feel themselves writing or feel the ball move.

  • Slides (for exposing, not teaching): there are always too many slides and too many words. Do not read your slides - the audience can read, and too much text forces them to read instead of listen (a student's experiment showed subjects remembered what they read on slides over what the speaker said). Fixes: strip background junk, cut words, drop logos and even the title, eliminate clutter/bullets. Minimum font ~40-50pt (small fonts just tempt you to cram). Do not stand far from the screen (creates a "tennis match"). Avoid the laser pointer - it forces you to turn away and lose audience contact; use an on-slide arrow instead. Print slides and lay them on a table to spot a "too heavy" talk. The exception slide ("epsilon-x-lambda" / apex): a deliberately incomprehensible slide (e.g. the Afghanistan-complexity chart) usable exactly once per talk/paper/book.

  • Informing and inspiring: surveying who inspired people - freshmen were inspired by a teacher who said "you can do it"; senior faculty by someone who helped them see a problem a new way; everyone by someone who exhibited passion for the work. Express how cool the material is (his map-coloring / resource-allocation demo: from longer than the lifetime of the solar system down to seconds).

  • Teaching thinking: we are storytelling animals; teaching people to think means giving them the stories, the questions to ask of stories, and ways to combine and evaluate them.

  • Oral exams and job talks: the usual reasons people fail an oral exam are failure to situate (place your work in context) and failure to practice. Practice with friends who do NOT know your work - people who know it hallucinate content that is not there (your advisor is a poor debugger for this reason). A job talk must show vision (a problem someone cares about + something new in your approach) and that you have done something (list the steps to the solution) within the first ~5 minutes. Structure it like a sandwich, closing by enumerating your contributions.

  • Getting famous / getting recognized (Winston Star, all S's): package your ideas so they are recognized - "your ideas are like your children; you don't want them to go into the world in rags." Five elements, illustrated by his own arch-learning PhD work: Symbol (the arch), Slogan ("one-shot learning"), Surprise (you can learn from a single example, not a million), Salient idea (an idea that sticks out, not necessarily the most important - here the "near miss"), and Story (how it works and why it matters).

  • How to stop:

  • Final slide should NOT be a giant collaborator list (recognize collaborators on the FIRST slide instead), a "Questions?" slide, a "The End," or even "Conclusions" (nobody cares about conclusions; they care what you did). It should be labeled Contributions and stay up while people ask questions and file out.
  • Final words: a joke works now (Doug Lenat's secret - "people think they've had fun the whole time"). Do NOT end on "thank you" / "thank you for listening" - it implies people stayed only out of politeness; it is a weak move (you can mouth thanks once applause starts). Better endings: a benediction (Christie's / Clinton's "God bless you and God bless America" - he shows Clinton visibly pursing his lips to avoid saying thank you), or salute the audience by expressing how much you valued the time and place.

Note: throughout, Winston demonstrates his own techniques as he names them (opening with a promise, asking a question whose answer is "ask a question," ending by saluting the audience).

Reception

  • Overwhelmingly reverent; widely regarded as one of the best talks on speaking ever recorded. Recurring gratitude that "we live in an age where we can watch something like this for free," and that the internet gives access to the world's best professors.
  • Many memorial comments: Professor Winston died in 2019, so commenters note "Rest in Peace" and share stories - a top comment recalls him arriving at 6am to rehearse a class he had taught for 30+ years, calling it his commitment to the students.
  • Commenters most admire that he practices what he preaches in real time - opening with confidence-building, using the ask-a-question-about-asking-questions move, applying his own techniques as he teaches them ("next level brilliant").
  • The most-liked practical comment is a full written summary of the whole talk (Start / Time and Place / Tools / Job Talk / Getting Famous / How to End) that many call the definitive save-and-reread digest. Another top comment quotes the success formula verbatim.
  • Jokes about the laptop rule are popular ("I turned off my laptop when he said to and missed 90% of the talk"), as is the military-briefing gloss on his cycling point ("tell them what you'll tell them, tell them, tell them what you told them").
  • Mild pushback on the "never say thank you" advice: several argue a sincere, sparing thank-you has real value, one teacher relating how an unexpected genuine thanks moved a class precisely because it was rare.