What reading costs: your eyes, the page, and the speed you cannot buy
40 min
Two hosts talk the lesson through. The voices are synthetic; the script was written from this lesson and checked against it, and asserts nothing the lesson does not.
- State a realistic reading rate for yourself and calculate how long a given text will take
- Explain why reading rates far above the measured band cost comprehension, and what the eye is doing that makes that so
- Compare skimming and speed reading, and say what each one actually delivers
Lesson 3 ended with a reader going to find an easier book. This lesson is about the other resource you were spending: time. And it is where the course closes a door, because there is a shortcut on sale and it doesn't work.
Start with your own number
Before any of the research, measure yourself, because a rate you have measured is worth more than a rate you have read about.
Pick any five hundred words of ordinary non-fiction. A page of this lesson will do. Read it at your normal reading speed, the one you would use if you meant to understand it, and time yourself in seconds.
Then divide 500 by the seconds and multiply by 60. That is your words per minute on this text.
Write the number down. You will need it twice more in this lesson.
What the measurements say
The figures come from Marc Brysbaert's 2019 meta-analysis in the Journal of Memory and Language, which pooled 190 studies of reading rate covering 18,573 participants.1
Put these three in order from slowest to fastest, and guess the gap between them. Reading aloud. Reading fiction silently. Reading non-fiction silently.
Show the answer
Aloud is slowest, at 183 words a minute. Then silent non-fiction at 238, then silent fiction at 260.
Most people get the order right and the size wrong. The whole spread, slowest of the three to fastest, is 77 words a minute. Reading rate is not a dial with much travel on it, and that is the first thing the measurements tell you.
For adults reading English silently, the averages are 238 words a minute for non-fiction and 260 for fiction. Reading aloud, from 77 studies and 5,965 participants, is 183.1
The spread matters as much as the average. In the paper's own words, "For silent reading of English non-fiction most adults fall in the range of 175 to 300 wpm; for fiction the range is 200 to 320 wpm."1
Now put your own number next to those. Most readers who do this find they're slower than they expected, and a good many discover they are not slow at all and have simply been told that 300 is normal.
That number has a history, and the paper traces it: through Huey in 1901, Tinker's studies, S. E. Taylor's sample of 12,143 readers, Carver, and Rayner's later reviews.1 Three hundred is at the top of the measured range rather than in the middle of it, and it is the figure this meta-analysis exists to correct.
Tinker's own verdict, quoted in the same paper, gives you the far end of what careful reading looks like: "even with easy material, 500 wpm is very fast reading".1
Four numbers, drawn to one scale, are the whole argument of the rest of this lesson. The first is what was measured. The last is what is for sale.
Measured rates from Brysbaert 2019, and the 500 wpm judgement is Tinker's, quoted there. The 1,000 wpm ceiling is from Rayner and colleagues 2016 as the APS summary reports it. Reading aloud, at 183, and silent fiction, at 260, sit too close to 238 to draw apart at this scale. The 1,500 mark is not a measurement of anything: it is an illustrative marketing claim, and this lesson takes it apart below.
You surveyed a book in lesson 2. Work out how long reading it will actually take, using your own measured rate, and say what you will do about the answer.
Show the answer
Books vary from about 250 to 400 words a page, so count the words on one full page of your own book and use that figure. If you cannot get at the book, 300 a page is a fair guess for non-fiction. Multiply by the page count and divide by your rate. A 320-page book at 300 words a page is about 96,000 words, which at 250 words a minute is six hours and twenty minutes of reading, and more if the book is hard.
What most people do with that number is the interesting part. Six hours is three two-hour evenings, which is manageable and which nobody plans. The alternative, and this is what the survey was for, is deciding on evidence that this is not the book, and spending the six hours on one that is.
Why the ceiling is where it is
A note on how this course read its main source here, before it is used. Most of this section rests on a 2016 monograph by Rayner and colleagues for Psychological Science in the Public Interest, and we did not open it. Everything attributed to it comes from the Association for Psychological Science's public summary and from Brysbaert's account of it, and the lesson claims nothing beyond what those two support. You are entitled to know which of a course's sources were read in full and which were not, and this one was not.
Reading is not a smooth sweep across a line. The eye jumps, stops, and jumps again, and it takes in text only during the stops. The jumps are called saccades and the stops fixations.
Two things follow, and they are why the numbers above do not move much.
There is a limit on how much can be taken in per stop, and the eye needs time to move between stops. So a rate is roughly a count of fixations multiplied by what each one delivers, and you cannot raise either term far without giving something up.
And the movements are not random. Rayner and colleagues describe saccades as movements that "naturally move to the parts of text in most need of processing and to move back to material that has not been adequately understood".2 That second half is the important one. The backward movements that speed-reading courses describe as wasted motion are the mechanism by which comprehension repairs itself.
What the summary and Brysbaert's account establish is enough for the argument. Skilled readers already run at roughly 200 to 400 words a minute. Rates above about 1,000 aren't possible without severe loss of comprehension. And, stated flatly: "an increase in reading speed is accompanied by a reduction in comprehension of the material being read".2
A reading app shows you one word at a time, in a fixed place on the screen, so your eyes never move. It can run at 800 words a minute. What does the evidence predict about what you will remember?
Show the answer
Less than you would from reading the same text normally, and the gap grows as the speed rises. The technique is called RSVP, and because it removes eye movements entirely, it removes the backward movements too. The summary's finding is that eliminating eye movements "generally impairs the recall accuracy of material presented at faster speeds".2
Notice what the problem is not. It isn't that the words go by too fast to see. It's that you've lost the ability to go back, and going back is how you fix a sentence you half-understood, which is most sentences in anything hard.
The other common promise is training your peripheral vision to take in more per fixation. That runs into a fact about eyes rather than about training: peripheral information "appears fuzzy and unfocused", which is why you move your eyes in the first place.2
The claim, taken apart with your own arithmetic
Take a marketed figure of 1,500 words a minute with full comprehension.
That is twenty-five words every second. Put your own measured rate beside it. For most readers it is five to eight times what they just recorded on a passage they were actually trying to understand.
Now ask what would have to be true. Either the eye is taking in five to eight times more per fixation, which runs into the fuzziness of peripheral vision, or it is making five to eight times as many fixations per second, which runs into how long a fixation takes and how far the eye can move. Those are the only two terms in the product.
So the arithmetic and the evidence point the same way, and the sources say so plainly: comprehension falls as speed rises, and above roughly 1,000 words a minute the fall is severe.2
Now the fair question, which this course would rather ask than skip. Is there any version of the claim that survives?
There is, and it is worth having. Speed-reading programmes do appear to train skimming, meaning getting the gist of running text quickly.2 That is a real skill with a real use. It is a cousin of the survey you learned in lesson 2 rather than the same thing: the survey reads the compressed copies of a book, its contents page and index and blurb, while skimming in this sense reads the text itself fast. Systematic skimming, to give the survey its proper name, is one of the two sublevels of inspectional reading rather than a level of its own. What the programmes don't deliver is the thing being sold, which is analytical reading at five times the speed.
And there is a second thing that genuinely raises your rate, which nobody can sell you in a weekend: vocabulary and practice.2 A word you know well is processed faster than a word you half-know. That is lesson 3 arriving from the measurement side, or part of it is. Knowing a text's words already is one of the things lesson 3 called background, and it is the part that shows up in the timing.
The five gears, and why this course won't hand them to you
You may meet a model of reading with five gears and speeds attached: reading for recall at the slow end, then reading for recognition, reading out of interest, skimming, and scanning at the fast end. It is Carver's, and it is a proposal rather than a finding.1
Brysbaert tested it against the pooled data, and the section of his paper reporting the result is headed with the conclusion: there is no evidence for reading gears except for the distinction between reading and scanning.1 He found no difference between reading for recognition and reading out of interest, and no bimodal pattern that would have rescued the model.
One finding from that test survives and is worth more than the ladder was. Reading for recall is slower, and the reason is not that it is a different gear. It is, in the paper's words, "an increased structuring and organization of text information", visible as "many regressions and rereadings".1
Read that twice, because it is this course's central claim arriving from the measurement side. The extra time does not go into moving the eyes more slowly. It goes into what the reader is building.
Screens
One more cost, and this one is smaller and more conditional than the headlines.
Delgado and colleagues pooled 54 studies covering 171,055 participants and found that reading on screen came out worse than reading on paper, with an overall effect of about a fifth of a standard deviation.3
The conditions are the finding. The effect is carried entirely by expository text and is absent for narrative. And it is significant only where people read under time pressure.3
Later work fits that shape rather than overturning it. A 2022 meta-analysis in Media Psychology found no negative effect of screen reading for narrative texts, which is what the 2018 result already predicts for stories.3
So the honest statement is not "screens are ruining reading". It is that the medium matters most exactly where this course operates, which is hard expository text read against a clock, and hardly at all when you are reading a novel in an armchair.
What people get wrong
"Speed reading is a skill I can buy." What is on sale is skimming, meaning going through running text fast for the gist. That's a real skill and a useful one, and it isn't what the advertisement says you are buying.
"Subvocalisation is slowing me down." The inner voice is a common target of these programmes. The rate limits above are measured on ordinary silent reading, inner voice and all, and nothing this course found supports suppressing it as a route to comprehension.
"Regressions are wasted motion." They are the repair mechanism. A reader who never looks back understands less, not more.
"I read at 600 words a minute." Usually true, and usually measured without a comprehension test. Add the test and the number comes down, which is the entire finding.
"Screens are ruining reading." The effect is small, conditional on the kind of text, and conditional on time pressure. Report it that way.
Practice
Measure yourself on three passages of five hundred words each: something easy, something ordinary, and something from your own field or study.
Write the three rates down side by side. The spread between them is the point of the exercise, and it is usually larger than people expect. That spread is lesson 3's claim in numbers: the same reader, the same eyes, different amounts of background.
Then take a text you would normally read on a phone under time pressure, read it on paper with no clock, and write one line about what you noticed. You are not running an experiment on yourself, and one comparison proves nothing. You are finding out what the conditions in the meta-analysis feel like from the inside.
Connections
Lesson 1 gave you the levels, and this lesson attaches numbers to the choice between them: inspectional reading is fast because it reads less, not because it reads the same amount faster.
Lesson 3 explained why your rate changes with the text, and this lesson measures the change.
Lessons 5 to 10 teach the slow methods: outlining an argument, coming to terms with an author, reading a paper, reading a primary source. They are worth learning precisely because the shortcut is closed, and now you have the evidence that it is.
Go deeper
- Brysbaert, "How many words do we read per minute?", Journal of Memory and Language 109 (2019). Free, and the history of the 300 wpm figure in the early sections is worth reading on its own.
- The APS summary of Rayner and colleagues on speed reading. Short, free, and the source of most of what this lesson says about eye movements.
- Delgado and colleagues, "Don't throw away your printed books", Educational Research Review 25 (2018). Paywalled at the publisher, but the abstract carries the three conditions that make the finding usable.
Sources
- Marc Brysbaert, "How many words do we read per minute? A review and meta-analysis of reading rate", Journal of Memory and Language 109 (2019), 104047, full text. Read in substantial part: the abstract, the introduction, the historical sections on the 300 wpm figure, and the general discussion. The averages of 238, 260 and 183 wpm, the participant and study counts, the individual ranges quoted verbatim, the history of the 300 wpm norm, Tinker's judgement that "even with easy material, 500 wpm is very fast reading", and the test and rejection of Carver's gears including the quoted explanation of why reading for recall is slower.
- Keith Rayner, Elizabeth Schotter, Michael Masson, Mary Potter and Rebecca Treiman, "So Much to Read, So Little Time: How Do We Read, and Can Speed Reading Help?", Psychological Science in the Public Interest 17(1), 2016. Read at summary level only, via the APS public summary and via Brysbaert's account of it; the monograph itself was not opened, and nothing here goes beyond those two. The 200 to 400 wpm figure for skilled readers, the 1,000 wpm ceiling, the quoted sentence on speed and comprehension, the account of saccades and of RSVP, the quoted finding on recall accuracy, the fuzziness of peripheral vision, the constructive finding about vocabulary, and the concession that these programmes train skimming.
- Pablo Delgado, Cristina Vargas, Rakefet Ackerman and Ladislao Salmerón, "Don't throw away your printed books: A meta-analysis on the effects of reading media on reading comprehension", Educational Research Review 25 (2018), 23 to 38, doi:10.1016/j.edurev.2018.09.003. Read at abstract level. 54 studies, 171,055 participants, overall Hedges' g = -0.21, carried by expository text (-0.27), absent for narrative (0.01), significant only under time pressure.
Check your understanding
This lesson has a 5-question quiz. Pass it and the questions come back on a schedule in Review, so what you learned stays learned. Your progress is saved in your browser; no account needed.