Rf Readit Fast Add to Chrome
← Blog

Does speed reading actually work? The honest answer.

Most speed-reading claims don't survive controlled testing. The version that does is more modest — and more useful — than the marketing tells you. Here's what 30+ years of eye-movement research actually shows.

Readit Fast · ·9 min read

The short answer is: mostly no — and the version that does is what we built. But “mostly no” is a flippant framing for a question with real research behind it, so let’s actually answer it.

If you’ve spent any time in this category, you’ve seen the claims: “Read 3x faster!” “1,000 WPM with full comprehension!” “Become a speed reader in 21 days!” The peer-reviewed evidence on every single one of those claims is cleanly negative. But underneath the marketing, there’s a real, modest, replicable speed-reading effect — somewhere in the 30–50% range above your baseline. The trick is being honest about which version you’re selling.

This post walks through what the research actually says, why most speed-reading products lie about it, and what an honest implementation looks like.

The lie at the center of the category

Most speed-reading marketing collapses into a single claim shaped like this: “Read at [400, 600, 1,000, 1,500] words per minute with full comprehension.”

The problem is the full comprehension part. There is a peer-reviewed scientific finding, replicated multiple times since the 1980s, that says: above roughly 350–500 WPM, comprehension drops measurably and predictably, and above 600 WPM it collapses. The 2016 review So Much to Read, So Little Time, published in Psychological Science in the Public Interest by Keith Rayner and colleagues, surveys this entire literature and concludes: “there is no evidence to support that speed-reading techniques can be used to dramatically increase reading speed without substantial loss of comprehension.”

That’s not us editorializing. That’s the entire field’s review-level conclusion in a flagship journal.

So why does the marketing keep working? Because most users never run the comprehension test that would catch the lie. The “I can read 800 WPM!” feeling is actually the feeling of skimming successfully — your eyes are moving fast, you’re catching some words, and you have a sense that the article was about something. That sense is real. It’s also not reading.

What the research actually shows

The cleanest single paper in this literature is Schotter, Tran & Rayner (2014), titled — and you can’t make this up — “Don’t Believe What You Read (Only Once): Comprehension Is Supported by Regressions During Reading.”

Here’s the core finding: when you read normal text, your eyes do three things. They fixate (pause on a word for ~200 milliseconds), saccade (jump to the next word in ~30 ms), and occasionally regress (jump back to a word you already read). Skilled readers regress on roughly 10–15% of words. The Schotter paper used a clever experimental setup — a “trailing mask” — that made it physically impossible for participants to regress, and then measured comprehension. The result: comprehension dropped meaningfully, on both literal and inferential questions.

The implication is that regression isn’t wasted time. It’s how your brain checks its understanding mid-sentence. When you encounter an ambiguous word (“bank” — financial or river?) or a complex parse, you sometimes need to flick your eyes back to confirm. Strip that out and the brain’s error-correction loop has nowhere to run.

This matters for the speed-reading category because the dominant speed-reading mechanism — RSVP, or Rapid Serial Visual Presentation — physically eliminates regressions. RSVP shows you one word at a time at a fixed point on the screen. There’s no “back” to flick to. Spritz, the most famous RSVP product, was published in 2014 with the marketing claim that it preserved comprehension at 500+ WPM. The Schotter paper was published the same year. They were directly responsive: the cognitive scientists were calling out the product category at exactly the moment it went viral.

A follow-up paper, Benedetto, Carbone, Pedrotti, Le Fevre, Yahia Bey & Baccino (2015), tested Spritz directly across 250, 350, 500, 650, and 950 WPM. Comprehension was intact at 250 WPM, mostly intact at 350 WPM, and dropped sharply from 500 WPM upward — first on inferential questions (the kind that test whether you understood the argument), then on literal questions (the kind that test whether you saw the words).

So: at 250 WPM, RSVP is roughly as good as natural reading. At 350 WPM, you’re paying a small comprehension tax. At 500+ WPM, you’re skimming with extra steps.

The 400–500 WPM cliff is real, and audiences feel it

If you spend an hour on Reddit, Hacker News, or Quora threads about speed reading, you will see the same comprehension cliff described in user-generated language across completely independent threads:

“I think that is ‘reading with a headache.’ I don’t think my comprehension would ever catch up.” — physicist forum user

“Anything above 400-500 WPM becomes incomprehensible noise.” — Reddit user reviewing Speechify’s 5x mode

“The speed was there but I understood almost nothing.” — Bartosz Czekala, Universe of Memory

Three independent sources converging on the same number — 400–500 WPM — using the same physical metaphors (“headache,” “noise,” “blur”). That’s not coincidence. That’s a real ceiling that real readers hit and recognize, even if they can’t cite the Schotter paper that explains it.

Bionic Reading: the public failure

The category’s most recent credibility collapse is worth highlighting because it shows what happens when a speed-reading claim meets a real test.

Bionic Reading went viral in 2022 with the claim that bolding the first 1–3 letters of each word (“fixation prompts”) improves reading speed and comprehension. Tweets racked up. The original developer’s validation study had n=12 participants.

In 2023, Readwise — a respected read-later and highlights tool — ran a public replication with 2,074 participants. The result: readers using Bionic Reading were 2.6 words per minute slower than reading plain text. Comprehension was identical. The effect was, statistically, a wash.

A 2024 paper in Acta Psychologica — yes, peer-reviewed — was titled, plainly, “No, Bionic Reading does not work.”

The category is still rebuilding from the resulting credibility hit. The Bionic Reading iOS app currently sits at 2.4 stars (147 ratings).

The lesson isn’t that Bionic Reading was uniquely bad. It’s that most claims in this category have never been tested at scale, and the few that have been tested have a tendency to collapse.

So what version of speed reading actually works?

Here’s the honest band, drawn from the research above:

How Readit Fast handles this

We built the product around the research, not against it.

The default WPM is 200, free forever, no signup — the speed where everyone is comfortable. Email sign-in unlocks 350 WPM, which is roughly the comprehension-preserving ceiling for most people on most material. Pro unlocks the full 1,500 WPM ceiling — but with three things attached:

Regression as a first-class control. Tap to pause, swipe back, scrub by paragraph. RSVP without rewind contradicts its own foundational research. We keep the rewind.

AI comprehension quizzes after every article. A 3-question quiz testing inferential comprehension (the kind that degrades first when you read too fast). Below 70% and we tell you. The comprehension claim is not asked to be taken on faith — it’s tested every session.

Adaptive pacing. Hard sentences slow down. Easy transitions speed up. You set the target WPM; the engine handles the variance based on Flesch-Kincaid difficulty and semantic density.

That’s the version of speed reading that survives the research. It’s slower than what the rest of the category promises. It’s also actually true.

Bottom line

Does speed reading work? Yes — modestly, predictably, around 30–50% above your baseline, on most material, with comprehension intact, if the implementation respects regression and measures retention. Anything else is selling you a feeling.

If you want to test this on your own reading, the Readit Fast Chrome extension is free at 200 WPM with no signup. Read your next article through it. Take the comprehension quiz at the end. The number on screen is the only honest answer to “did I actually get it?”

References cited:

Try the reader you just read about.

200 WPM free, no signup. 350 with email. Pro at $2.99/week unlocks AI quizzes, semantic chunking, and the 1,500 WPM ceiling.

Add to Chrome — free Or get launch updates