The guide
A cue card works because it leaves out almost everything you plan to say
A script pulls your eyes down and a crowded slide makes the room read. A few numbered cards with big type and a time on each keep the talk in your head and the clock in your hand.
- 3 by 5 inches the most common index card size in North America
- 0.2 to 2 degrees letter height, as seen by the eye, that is read at full speed
- 40 cm the standard reading distance in the print size study
- 281, 168, 112, 39 seconds for four cards sharing a ten minute talk by word count
- 160 words a minute natural conversation, as one research paper puts it
Read the full guide Fold the guide
A talk fails in one of two familiar ways. The speaker reads, head down, in sentences that were built for the page. Or the speaker turns to the screen and reads that instead, while the audience races ahead through the same words. Both are the same mistake. The full text of the talk has been put where someone can read it, and reading has replaced speaking. A cue card is the opposite bet. It holds a few words that mean nothing to anyone but you, and it trusts you to supply the rest. That trust is the whole point.
Why the script and the crowded slide both fail
Writing and speech are different instruments. The United States Air Force handbook on communication, The Tongue and Quill, is blunt about the word-for-word script: reading aloud often sounds dull, and a briefing read from a manuscript tends to lack spontaneity, eye contact and engagement with the audience. Why, it asks, should people attend a reading if they can read it themselves? Its advice for the spoken message is everyday language, contractions, short sentences, personal pronouns and repeated key words. A script is written language by construction, so reading one aloud delivers the wrong kind of sentence.
The listener’s position makes this worse. A reader can see where a paragraph or a section ends, and can go back. A listener cannot. The Wikibooks chapter on oral presentations makes the point directly: a speaker has to recreate those breaks out loud, because the audience cannot see the progression of the argument. A script tempts you to write for the reader you are, not the listener they are.
A slide full of sentences fails for a related reason, and the same handbook names it from the speaker’s side. Do not read slides, it says, because it is insulting. Keep one thought on each, which gives the audience time to turn back to you, and avoid the table that looks busy on a screen. If the words are on the wall, the room reads them and stops listening, and the speaker has become a slow narrator for a document.
So the notes have to live somewhere the audience cannot see, and they have to be too short to read aloud. That is a card.
One idea, a few words, a number
The alternative has a name. The Wikibooks chapter calls it extemporaneous speaking: a talk delivered from a prepared outline or from note cards that hold the main ideas and arguments, with only quotations copied word for word. The Air Force handbook describes the same thing as prepared speaking, with notes or an outline but not a script, and with words that are found at the moment of delivery and so sound natural. The usual card is the index card, most often 3 by 5 inches in North America, and presentation notes are one of the uses Wikipedia lists for it.
An open textbook from College of the Canyons is specific about the cards themselves. Do not write or print on both sides, it says, because flipping a card distracts the audience; slide the finished card aside instead. Keep the words few enough that a glance is enough. The exception is an extended exact quotation or a group of statistics that must be delivered precisely. A Wikibooks chapter on planning a piece of writing describes the same habit from the other end, one topic to a card and the next topic on the next card, which is what lets a stack of cards be put in order at all.
Then comes the detail that sounds trivial until the day it matters: number the cards. The handbook asks for bold page numbers even on a full manuscript, the College of the Canyons textbook wants the number bold in the upper right corner, and the sample card in the Wikibooks chapter is headed Note card #1. A deck gets dropped, or shuffled in a bag, or two cards stick together under hot lights. A number in the corner turns a disaster into a few seconds of sorting. A count beside it, card 7 of 12, also tells you at a glance how much of the talk is left without any arithmetic.
How much goes on a card is a matter of nerve. The Wikibooks chapter observes that inexperienced speakers worry they will forget something important unless it is written out, and its remedy is practice with the cards, not more words on them. Most points need two or three words. A title, a figure or a quotation may need the exact wording.
A card can also carry a cue for the voice. The Air Force handbook treats the pause as the spoken form of punctuation: short ones divide the points inside a sentence, longer ones mark the move from one main point to the next, and a speech should never be hurried. A word such as pause on the card, marked so that it is not read aloud, puts that advice where the speaker will see it.
That is the working definition of one idea per card. A card is a unit you could move, cut or skip without rewriting its neighbors. If a card carries two ideas, you cannot drop one when time runs short, and you will lose your place between them. The cards themselves are ordinary enough that printers know them by name. The printing industry’s list of standard media names includes index cards of 3 by 5 and 4 by 6 inches, beside the international A6 at 105 by 148 mm and A7 at 74 by 105 mm.
How big the type has to be
Type on a card is read under worse conditions than type in a book. It sits at the end of your arm or on a lectern, the light is whatever the room gives you, and you have a fraction of a second before looking up again. Guessing at a comfortable size is risky. There is better evidence than a guess.
Gordon Legge and Charles Bigelow reviewed decades of reading research and described a fluent range of print size, the span over which text can be read at maximum speed. Measured as the angle the lowercase x-height makes at the eye, it runs from about 0.2 to 2 degrees. At the 40 cm they take as a standard reading distance, those angles correspond to x-heights of 1.4 mm and 14 mm.
The useful part is that the measure is an angle, so it travels. Hold the card at 60 cm instead of 40 and the same angles need letters one and a half times taller: about 2.1 mm of x-height at the small end. Set a card on a lectern a meter away and the floor is 3.5 mm. The x-height is only part of a type size, so the point size you ask for is always larger than the x-height you need, by an amount that depends on the typeface.
Two cautions belong here. The bottom of that range is where full reading speed begins for continuous text in good conditions, not a target, and a glance in stage light deserves a wide margin above it. And the study is about size, not about contrast, so it does not rescue pale ink on a tinted card. The web accessibility guidelines give a usable yardstick for that second problem: a contrast ratio of at least 7 to 1 for their stricter level. Black on white, black on yellow and white on a dark card all clear it easily.
Fitting type to a card is therefore a constraint problem. The largest size that fits is the right one, and a minimum you refuse to go below tells you when the card holds too much. The honest response to a card that will not fit is to split it, not to shrink the letters.
A time on every card
Every talk has a time limit, and the advice about it is consistent. Know your time constraints, the Air Force handbook says, and know where you can shorten the speech if you must. The Wikibooks chapter tells a speaker who runs far over the allotted time to go back and narrow the scope. Neither says how to notice the problem while there is still time to act on it. A time written on each card does. If the card says you should be done with it by the four minute mark and your watch says six, you know now, not at the end.
A budget needs a rule for dividing the total. Equal shares are the simplest. Shares in proportion to the words in each card’s notes are usually closer, since a card with more prompts tends to take longer. Whole seconds make this slightly awkward, because fair shares rarely come out whole. The method the Census Bureau describes as the Vinton or Hamilton method, used for apportioning seats in the House from 1850 to 1900, solves exactly that: give each party the whole number in its quotient, then hand out what is left, one at a time, to the largest fractional remainders.
Here is a case you can reproduce. A talk of ten minutes is 600 seconds. Four cards carry notes of 50, 30, 20 and 7 words, 107 in all. The exact shares are 280.37, 168.22, 112.15 and 39.25 seconds. The whole parts add to 599, so one second is left, and it goes to the first card, which has the largest fraction. The result is 281, 168, 112, 39. As running times, the cards should end by 4:41, 7:29, 9:21 and 10:00.
Those last four numbers are what belong on the cards. A duration tells you how long a section should take, which is hard to judge while speaking. A clock time tells you where you should be right now, which takes one look at a watch.
Notice what the word count cannot tell you. Those 107 words are notes, not the talk. Natural conversation, in the phrase of one research paper on speech, runs at 160 words a minute, and the Air Force handbook puts the average speaking rate at 180, so ten minutes of speaking is on the order of sixteen to eighteen hundred words. The notes are a small fraction of that by design, and a pace computed from them describes your cards, not your voice.
A plan is only a guess until it meets a stopwatch. Practice, practice, practice, the Air Force handbook says more than once, and its checklist for briefings with slides includes gauging the time they need. The Wikibooks chapter adds that the practice should be done with the outline or the cards that will be used on the day, and the College of the Canyons textbook says the same and adds a clock: time the speech to be sure it fits the time assigned.
A useful rehearsal records two things per card: what you planned and what you took. The comparison is rarely flattering and always informative. An opening that was budgeted a minute and took two is the most common finding, and it is one you cannot discover by reading the notes silently.
The timing has to be real time. A browser offers a clock that only moves forward, which is the right tool for measuring elapsed time, and a timer built on it keeps counting correctly however unevenly the screen redraws. It also has to pause honestly. A break to answer the door should add nothing. Keep the table when you finish: saved as comma-separated values, a format simple enough that a short memo defines it, it opens in any spreadsheet for comparison with the next run.
When a screen is the better card
Paper has limits. It cannot show you a clock, it cannot tell you that you are behind, and the advice to write on one side only means a long talk becomes a thick deck. A phone or tablet on the lectern can do all three, and it never gets dropped out of order.
A screen brings its own failure modes, and they are worth knowing before you depend on one. Screens dim and lock. The web’s wake lock feature exists to prevent that, but the specification releases the lock whenever the page stops being visible, so switching apps for a moment ends it. Full screen display on the web must be started by a deliberate press, and on some phones a page cannot take over the whole screen at all. And a screen at arm’s length is subject to the same arithmetic about letter height as a card.
There are things neither medium can do. A card cannot make a weak structure strong, and a time budget divided by word count is a starting estimate, not a measurement of how you speak. Printing on both sides only works if the printer’s flip matches the layout, which is why print standards distinguish turning the sheet on its long edge from turning it on its short edge. Print one sheet and check it before printing twenty. The through-line is restraint. Leave the sentences in your head, put one idea and one time on each card, make the letters bigger than seems necessary, and find out how long it really takes before anyone else does.
Sources
Every source consulted for this guide. Links checked October 6, 2026.
- The Tongue and Quill (Air Force Handbook 33-337, 27 May 2015, incorporating Change 1) United States Air Force (archived copy at the Internet Archive)
- Rhetoric and Composition/Oral Presentations Wikibooks
- Index card Wikipedia
- 8.5: Preparation, Practice, and Delivery Tammera Stokes Rice, Fundamentals of Public Speaking, College of the Canyons (LibreTexts page shared under CC BY)
- Basic Writing/Invention Wikibooks
- Does print size matter for reading? A review of findings from vision science and typography Gordon E. Legge and Charles A. Bigelow, Journal of Vision 11(5):8, 2011 (PubMed record with abstract, archived copy at the Internet Archive)
- A high-performance speech neuroprosthesis Francis R. Willett and others, Nature 620, 2023
- A4 paper format / International standard paper sizes Markus Kuhn, University of Cambridge Computer Laboratory
- PWG Media Standardized Names 2.0 (MSN2) The Printer Working Group, Candidate Standard 5101.1-2013
- Internet Printing Protocol/1.1: Model and Semantics (RFC 8011) Internet Engineering Task Force
- Historical Perspective United States Census Bureau, Congressional Apportionment
- Web Content Accessibility Guidelines (WCAG) 2.2 World Wide Web Consortium (W3C)
- Screen Wake Lock API World Wide Web Consortium (W3C)
- Element: requestFullscreen() method MDN Web Docs, Mozilla
- Performance: now() method MDN Web Docs, Mozilla
- Common Format and MIME Type for Comma-Separated Values (CSV) Files (RFC 4180) Internet Engineering Task Force