Reading Page Estimator

238 words a minute, from 190 studies, and it is not a studying rate

Marc Brysbaert’s 2019 meta-analysis in the Journal of Memory and Language is the reference for how fast adults actually read, and its first sentence is the figure: “Based on the analysis of 190 studies (18,573 participants), we estimate that the average silent reading rate for adults in English is 238 words per minute (wpm) for non-fiction and 260 wpm for fiction.”

What the meta-analysis found

The paper is explicit that these are narrower claims than they look: “The values of 238 and 260 wpm are valid for adults between 18 and 60 years without reading problems. They are lower for younger children and older adults. In addition, they are only valid for native speakers. Second language readers have lower reading rates.”

And then there is the detail that matters most for a page about textbook assignments, which sits in the study-selection criteria rather than in the abstract. To enter the meta-analysis, a study had to meet this condition: “The task was reading for comprehension or fun. This excluded studies with proofreading or the type of task Tinker used. It also excluded studies that tested studying for fact retrieval.”

So the best-supported number for adult reading speed was arrived at by leaving out the activity you are probably here to estimate. If you are reading forty pages of a textbook because there is a quiz on Thursday, 238 wpm does not describe you, and neither does any tool that quietly assumes it.

Reading so that you can recall it runs at about 100 words a minute

Brysbaert addresses studying directly, while taking apart an older theory that predicted it. Ronald Carver proposed that readers shift between five “reading gears” depending on their goal. The paper summarises them as “reading for recall (140 wpm), reading for recognition (200 wpm), reading out of interest (300 wpm), skimming (450 wpm), and scanning (650 wpm)”.

The meta-analysis did not support that as a set of distinct modes. “We failed to find evidence for a distinction between reading for recognition and reading out of interest”, and no bimodal distribution turned up that could have rescued the idea. But the time cost of reading for recall was real, and the paper puts a number on it:

“We further failed to find a different type of reading for recall than for recognition/interest. It is true that reading for recall takes more time (100 wpm seems to be a good estimate, although in several studies it was even lower), but this was not due to a different type of reading. What seems to happen is an increased structuring and organization of text information, as can be concluded from the many regressions and rereadings observed.”

Two things follow, and both change how you should read the output above. Reading to remember costs roughly twice the time of reading to understand. And the extra time is not spent moving through the text more slowly; it is spent going back. A model that multiplies a word count by a forward reading rate is estimating a different activity from the one you are about to do.

The same 40 pages at 300 words a page, which is 12,000 words

One assignment, a spread of 50 minutes to two hours, decided entirely by a box most people leave alone.

What to put in the words-per-minute box

The box starts at 200. That is a perfectly plausible rate for an individual, since most adults fall between 175 and 300 on non-fiction, but it is not the average: the 95% confidence interval for the mean is 230 to 246, and 200 sits below it. So the default has an awkward property. If you are reading to understand, 200 understates your speed and the page will tell you the job is longer than it is. If you are reading to be tested, 200 overstates your speed by about double and the page will tell you the job is shorter than it is. It is wrong in opposite directions for the two things students use it for.

Two honest ways to fill it in. Either measure yourself, which takes five minutes and is described below, or use the published adjustment for how hard the text is, which is the next section.

Longer words, slower reading, and the one adjustment the paper hands you

Brysbaert anticipates exactly the case where a single average fails, and supplies a formula for it: “To capture word length in a reading rate calculator for individual texts, the following equation can be used for English: Expected reading rate text = 238 * 4.6/average word length text.”

The 4.6 is the average word length the 238 figure was calibrated against. The worked example in the paper is the paper itself: “For instance, the average length of the words in the present article is 5.1 letters, meaning that the expected reading rate is 238 * 4.6/5.1 = 215 wpm (translating to roughly two hours of solid reading for the present article without the references).”

The formula, run out

The first two rows are the paper’s. The rest is the paper’s formula applied to other word lengths, rounded to whole words per minute.

One caution about that last row, because it is the sort of thing that gets misquoted. The paper reports in a footnote that the “mean word length of content words in English is 5.9 letters for fiction and 6.7 letters for non-fiction text”. Those figures are for content words only, not for all words including the, of and a, so 6.7 is not the number to feed into a formula calibrated on every word in a text. It tells you the direction academic prose pushes, not the magnitude.

If you want the magnitude for your own book, it is two minutes of work: count the letters in 100 consecutive words of the chapter you have been set, divide by 100, and run the formula. A chemistry chapter and a paperback are not the same text at the same speed, and this is the published way of saying how much they differ.

Words per page is the input this page cannot know

In page mode, every minute reported above is directly proportional to the words-per-page box. It is the largest source of error in the estimate and the box almost nobody touches.

40 pages at 238 wpm, varying only the words per page

A 40% swing in the answer, from an assumption rather than a measurement.

Here is the honest position on that assumption. I looked for a published figure for the number of words on a printed page and did not find one I would put on this page. The numbers in circulation, 250 and 275 and 300, come from writing and self-publishing sites that do not say what they counted or on which books. The closest thing to a measurement in the peer-reviewed literature is a side result of Brysbaert’s own book-reading study, in which 48 regular readers each timed themselves through a whole book: “Mean number of pages per book was 385 (SD = 166), mean number of words was 107,000 (SD = 65,800).” Dividing one mean by the other gives about 278 words a page.

That figure deserves its caveats stated rather than stripped. It is a ratio of two averages over 48 books in six languages, in whatever editions the readers happened to own, and the word counts were themselves estimated from a word-count website and from sampling pages. It is an order-of-magnitude check, not a constant. The standard deviations say the rest: a book in that sample was about as likely to run 220 pages as 550.

Words per page is decided by trim size, type size and leading, which are set per edition. The practical consequences are simple. If your syllabus gives a word count, use word mode and skip the problem entirely. If it gives pages, count the words on three pages of your own copy, average them, and type that in. That single step removes the biggest error in the estimate.

How to measure your own rate, and why a one-minute test will not do it

Brysbaert makes a methodological recommendation that applies as much to you timing yourself as to a laboratory: “The second recommendation is that a good assessment of reading rate requires more than a one-minute test. This is particularly true for silent reading. Although the average of short tests is representative, there is much more variability between studies for short tests than for long tests ... On the basis of the present review, a minimum of 5 min seems required. Preferentially, this involves one long text rather than five short ones, as the latter seem more susceptible to demand characteristics.”

So the procedure, in the book you are actually going to read:

1. Open the chapter you have been assigned, at a page of ordinary prose rather than equations or figures.
2. Read for a timed five minutes, one continuous stretch, at the speed you would really use.
3. Count the words you got through. Count one full page properly, then multiply by the pages covered rather than estimating the whole run.
4. Divide by five. That is your rate in words per minute, for that book, at that purpose.
5. Type it into the box above, and type your measured words-per-page figure in as well.

Do it twice if you have ten minutes: once reading to follow the argument, once reading the way you read before a test, pausing and going back. You will get two numbers that differ by something like a factor of two, and you will know which one to use for which week.

If English is not your first language

The headline figures do not apply to you, and the paper says so twice. In the conclusions: the 238 and 260 values “are only valid for native speakers. Second language readers have lower reading rates.” And in the section on it: “Reading speed in second-language (L2) speakers is considerably slower than in first-language (L1) speakers. Indeed, reading rates below 100 wpm are not uncommon.” The example given is Hirai (1999), who measured English reading rate in Japanese university students who had had six years of formal English instruction at school plus university courses on top. The paper also cites Dirix and colleagues (2019), who “observed 10% slower processing rates when participants read or studied texts in a second language”.

Across one plausible lecture hall, then, the right value for the box runs from under 100 to over 300 words a minute. That range is the reason the box is editable, and it is why leaving it at 200 is a decision rather than a default.

What this page does not know

Your reading rate. It starts at 200 wpm, which is below the measured average for reading to understand and about double the estimate for reading to recall. Until you measure yourself, that number is a guess with a factor-of-two error bar on it.

What you are reading for. There is one speed box, and the published figures for understanding a text and for being able to reproduce it differ by roughly 2.4 times. The page cannot tell which you mean, so it applies whatever is in the box to both.

How many words are on your pages. No published constant exists, and the estimate is directly proportional to the value you enter.

Note-taking, highlighting and rereading. None of it is in the arithmetic. And it is worth knowing before you budget that time that John Dunlosky, writing about his group’s review of ten study techniques, says of highlighting, rereading and cramming that “my colleagues and I found that these strategies are not that effective, especially if students want to retain their learning and understanding of content well after the exam is over”, while “We rated two strategies—practice testing and distributed practice—as the most effective of those we reviewed”. On rereading in particular: “Despite its popularity, rereading has inconsistent effects on student learning”, and “A survey conducted at an elite university revealed that 84 percent of the students studied by rereading their notes or textbooks.” So the hours this page does not count are, for most people, hours spent on the method with the weakest support.

Problem sets, proofs, worked examples, figures and tables. A page of equations is not a page of prose and a words-per-minute model does not describe it at all. The reading-rate literature above is about continuous text; Brysbaert’s inclusion criteria required that “the stimulus materials were normal text”.

Your week. The sessions box divides the total into equal parts. It does not know your timetable, your other four courses, or that one of those sessions is at 11pm.

Whether the number helps. An estimate is useful for deciding whether forty pages is a Tuesday evening or a weekend. It is not a prediction, and nothing above should be read as advice about how you ought to study.

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: the 238 and 260 wpm silent reading estimates with their sample of 190 studies and 18,573 participants, the SD of 51.2 and the 230 to 246 confidence interval, the 183 wpm oral reading figure, the 175 to 300 and 200 to 320 individual ranges, the study-selection criterion that excluded studying for fact retrieval, the 100 wpm estimate for reading for recall and the finding that its cost comes from regressions and rereadings, Carver’s five reading gears, the word-length formula and its 215 wpm worked example, the content-word lengths of 5.9 and 6.7 letters, the 48-book study’s 385 pages and 107,000 words, the second-language figures including Hirai (1999) and Dirix et al. (2019), and the recommendation of a five-minute minimum test.
• John Dunlosky, “Strengthening the Student Toolbox: Study Strategies to Boost Learning”, American Educator, Fall 2013: that highlighting, rereading and cramming “are not that effective” for retaining learning past the exam, that practice testing and distributed practice were rated “the most effective of those we reviewed”, that rereading “has inconsistent effects on student learning”, and the survey finding that 84 percent of students at one university studied by rereading their notes or textbooks.

Every figure on this page was read against the sources above on 1 October 2026. Nothing you type leaves your browser. If your syllabus gives a word count rather than pages, the essay word counter is the other end of the same arithmetic. More: the rest of the MyCampusKit tools.