250 Words Per Minute Is the Number Reading Time Calculators Use. A 2019 Meta-Analysis of 17,000 People Found the Real Answer.
The number 250 words per minute appears everywhere as the average adult reading speed. It shows up in reading time calculators, content strategy guidelines, UX writing recommendations, and school-level assessments. Medium used it to calculate article reading times for years. It is probably slightly wrong, and the interesting question is why the inaccurate number persists and what a better one looks like.
The persistence of 250 words per minute traces back to studies conducted in the 1950s and 1960s, primarily on paper, with self-selected groups of college-educated American adults. Those studies found averages in the 250-300 words per minute range, and the round number became a rule of thumb that content platforms adopted without much scrutiny. The figure spread through citation chains where each subsequent source referenced the previous one rather than the original research.
Whether you use 238 or 250 as the baseline, what matters more for reading time estimates is understanding what factors move real readers away from any average, and what reading time labels actually do to reader behavior when they appear on published content.
The Brysbaert Meta-Analysis: A Better Number
A 2019 meta-analysis published in Psychological Bulletin by Marc Brysbaert, a cognitive psychologist at Ghent University in Belgium, analyzed 190 independent studies involving 17,887 participants from 18 countries reading in their native language. The study, titled "How many words do we read per minute? A review and meta-analysis of reading rate," is the most comprehensive synthesis of reading speed research conducted to date.
The mean reading rate Brysbaert's team found for adults reading silently in their native language on paper was 238 words per minute. This is close to 250 but not identical, and the difference compounds across long content: a 2,000-word article takes approximately 8.4 minutes at 238 wpm versus 8.0 minutes at 250 wpm. For a high-traffic publication calculating reading times for thousands of articles, the systematic underestimate is meaningful.
On screen, the mean rate was lower. Brysbaert's analysis found that screen reading is approximately 10 to 30 words per minute slower than paper reading, depending on the display technology, text size, line length, and contrast. E-ink displays like early Kindles tested closer to paper reading speeds. Standard computer monitors, particularly older lower-resolution displays, showed larger speed penalties. Modern high-DPI smartphone screens have not been as extensively studied, though more recent research suggests the gap between screen and paper has narrowed as display quality has improved.
Skilled readers tested significantly faster than the mean: approximately 300 words per minute for the upper quartile. Readers with dyslexia or other reading differences tested substantially slower. The variability within any group of readers is wide enough that a single number can mislead. A reading time estimate calculated for an average reader will underestimate time for a third of the audience and overestimate it for another third.
Why Reading Speed Is Not Uniform Within an Article
Reading time is also not constant across a text. People process headlines and subheadings faster because they are shorter, simpler, and serve orientation rather than argument-following. Introductory sentences that summarize familiar concepts are processed faster than paragraphs introducing new material. Dense explanations, statistical tables, code samples, and arguments that require pausing to evaluate slow readers down substantially below their average rate.
An estimated reading time that treats the entire article as uniform ignores this texture. A 1,000-word article that is 800 words of familiar background and 200 words of new technical content will take longer to read than a different 1,000-word article that covers familiar material throughout. The word count is identical; the cognitive load is not.
This is why reading time estimates should be understood as order-of-magnitude guides rather than precise predictions. A "5 minute read" label means the article is roughly in the 1,000-word range and will take most readers between four and eight minutes. It does not mean the article takes exactly five minutes for any specific reader.
Why Publishers Add Reading Time Labels
Despite this imprecision, adding a reading time estimate to content measurably changes reader behavior. Parse.ly, an analytics platform used by publishers including The Atlantic, Reuters, and hundreds of digital media organizations, has observed that articles with reading time labels have lower bounce rates and higher scroll depth than comparable articles without them. The mechanism appears to be that a reader who sees a specific time commitment makes a deliberate decision before starting, which creates a follow-through effect.
Medium introduced reading time estimates in 2013 as part of their pitch to readers that they were a platform for longer, more substantive reading. The explicit time labels communicated that the content was worth a real investment of attention, differentiating Medium posts from the short social media updates and thin blog posts that were common at the time. The format was later adopted by publications including Longreads, Substack newsletters, and editorial sites at major media organizations.
For content creators, the reading time label functions as a retention mechanism. A ten-minute article without a reading time label surprises readers who expect shorter content, and surprise is a common trigger for abandonment. The same article labeled as a ten-minute read self-selects for readers who are willing to make that commitment, improving the match between content and audience.
Newsrooms and content marketing teams have used this principle to design their content inventories intentionally. Short pieces under 3 minutes for social referral traffic. Medium pieces of 5 to 7 minutes for newsletter readers. Long-form pieces over 10 minutes for loyal subscribers who engage with depth. The reading time estimate makes the content categorization visible to readers without requiring the publication to explain their content strategy.
Speaking Rate and Audio Content
Average speaking rate for audio narration, podcast delivery, and presentation scripts runs approximately 130 to 150 words per minute, roughly half the average silent reading rate. Fast conversational speech reaches 180 to 200 words per minute; deliberate, instructional speech may run as slow as 100 to 120 words per minute. Radio and podcast hosts often target 150 to 160 words per minute as a pace that feels natural but remains easy to follow.
A script intended for audio needs a different estimate than a written article because the relationship between word count and time is fundamentally different. A 500-word podcast intro runs approximately 3 to 4 minutes when spoken aloud at a measured pace, not 2 minutes as a reading time calculator would estimate. Presenters writing conference talks, podcast hosts scripting episodes, and video producers planning narration length all need the speaking rate calculation rather than the reading rate calculation.
The practical consequence is that reading time estimates and speaking time estimates should be calculated separately and not assumed to be proportionally related beyond the rough 2-to-1 ratio.
What the Number Actually Tells You
Conclusion
A reading time estimate is most useful as a content production and editorial planning tool rather than as a precise user-facing claim. For an author, knowing that a draft is currently a 12-minute read signals that cutting it to 8 minutes requires approximately 800 words of reduction. For an editor, knowing that the four articles scheduled for this week total 45 minutes of reading time helps calibrate the publication's weekly output against audience engagement data.
For readers, the reading time label is a filter. It lets people decide whether a piece matches the time they have available before investing any attention. A reader with three minutes at lunch chooses differently when the label says 10 minutes than when it says 3 minutes. The label reduces mismatched expectations on both sides.
Frequently Asked Questions
What is the average adult reading speed?
A 2019 meta-analysis in Psychological Bulletin by Marc Brysbaert, covering 190 studies and 17,887 participants, found a mean of 238 words per minute for native-language reading on paper, and slightly slower on screen.
Does reading on screen take longer than reading on paper?
Yes. Research consistently finds screen reading is about 10 to 30 words per minute slower than paper reading, depending on display technology and text formatting. The difference is meaningful for reading time estimates.
How many words per minute do people speak?
Average speaking rate for presentations and narration is roughly 130 to 150 words per minute, about half the average reading rate. Fast conversational speech reaches 180 to 200 words per minute.
Why do articles with reading time labels have lower bounce rates?
A reading time label lets readers make a conscious commitment before starting. Research from Parse.ly shows that readers who know the time investment tend to follow through, reducing bounces compared to articles without the label.