Average Reading Speed, Explained (With Real Sourced Numbers)

The most rigorous recent source on this is a 2019 meta-analysis by researcher Marc Brysbaert, appearing in the academic Journal of Memory and Language, which pooled reading-speed data from a large number of individual studies spanning several decades rather than relying on any single smaller sample. Pooling that much data pointed to an average of roughly 238 words per minute for adults reading English non-fiction silently — a figure notably lower than the 250-300 WPM range still commonly cited in older, less rigorously aggregated sources.

The gap between 238 WPM and the older, higher estimates isn't a sign that people got slower readers over time — it more likely reflects that older individual studies used smaller sample sizes, different text difficulty levels, and less standardized measurement methods than a large meta-analysis pooling many studies together can achieve. When in doubt about which figure to trust, a large, recent meta-analysis pooling many prior studies is generally a more reliable estimate than any single older study, which is why this site uses the Brysbaert figure as its default.

238 WPM is an average across general-audience non-fiction prose — it is not a fixed constant true of all text. Brysbaert's own analysis and related reading research consistently show that fiction tends to be read somewhat faster than non-fiction on average (simpler sentence structures and more familiar vocabulary), while technical, academic, or unfamiliar-jargon-heavy text is read meaningfully slower, sometimes 30-50% slower than the general-prose average, since readers spend more time parsing unfamiliar terms and complex sentence structures.

This is exactly why /reading-time-calculator/'s estimate should be treated as a reasonable planning baseline for general-audience writing rather than a precise prediction for every kind of text — a technical whitepaper and a casual blog post of the same word count will not actually take the same amount of time to read for most readers, even though the calculator reports the same estimate for both unless you manually account for the difference.

Reading research consistently shows reading speed increasing through childhood and adolescence as decoding skills and vocabulary develop, typically plateauing in early adulthood, and then gradually declining somewhat in older adulthood, though with wide individual variation at every age. Brysbaert's meta-analysis specifically focused on adult reading speed, which is why 238 WPM is presented as an adult baseline rather than a number meant to apply to children's reading speed, which is meaningfully slower on average, particularly for early readers still developing decoding fluency.

Second-language readers reading in a non-native language also typically read more slowly than native speakers of that language, for reasons similar to the technical-text slowdown — more cognitive effort spent on word recognition and grammar processing leaves less capacity for fast, fluent reading. If you're estimating reading time for an audience that includes a meaningful share of non-native readers of the text's language, a somewhat longer estimate than the general 238 WPM baseline is more realistic.

Reading silently and reading aloud are genuinely different cognitive and physical tasks, and speaking speed is reliably slower — commonly cited in the 130-150 WPM range for natural, unhurried spoken delivery, versus the 238 WPM silent-reading baseline. This is why /reading-time-calculator/ reports two separate figures rather than one: reading time for someone silently reading your text, and a separately-calculated, slower speaking time for estimating how long a script will take to read aloud, whether that's a speech, a video voiceover, or a podcast script.

The gap between the two exists because speaking involves actual articulation time (physically producing each sound), natural pauses for breath and emphasis, and typically some deliberate pacing for audience comprehension — none of which apply to a reader's eyes scanning a page. Confusing the two is a common practical mistake: estimating a video script's runtime using the faster silent-reading WPM will produce an estimate that's meaningfully too short compared to the actual recorded runtime.

/reading-time-calculator/ divides your live word count (computed the same way /word-counter/ counts it — whitespace-split tokens) by 238 for the silent-reading estimate and by a slower rate for the speaking-time estimate, then reports both. If you know your specific content is unusually technical, unusually simple, or aimed at a specific age group or reading-proficiency audience, treat the tool's estimate as a reasonable starting point to mentally adjust rather than an exact figure — the underlying research this site's default is based on is itself an average across a very wide range of real reading conditions.