The method

What one word at a time does, and does not do.

Flash Reader has shown text one word at a time since 2010. Here is what the research says about reading that way, and what we built because of it.

What it is.

Rapid serial visual presentation, RSVP for short, shows a text one word after another at a single fixed point, so your eyes never have to move. Psychologists have used it as a lab method since 1970. Spritz made it famous in 2014 with a red letter and two small ticks that mark where to look (Benedetto and colleagues 2015 describe how it works).

Normal reading is not a smooth glide. Your eyes rest on a word for a moment, a fixation, then jump to the next one, a saccade, and now and then jump back to read something again, a regression. While they rest on one word they already take a preview of the next, so it is half read before they land on it.

What the evidence says.

Start with normal reading. Skilled readers average 200 to 400 words a minute, and eye movements take at most a tenth of reading time, so taking them away cannot make anyone several times faster. The limit is understanding the language, not moving the eyes (Rayner, Schotter, Masson, Potter and Treiman 2016).

In the stream, speed decides how much you keep. At 250 to 350 words a minute, understanding matches reading the same text on a page. It degrades at 400 to 450 and collapses above about 500, and inference goes before anything else (Di Nocera, Ricciardi and Juola 2018; Acklin and Papesh 2017).

Even at a gentle speed it has a cost. In a study of the Spritz reader at 250 words a minute nominal, readers recalled 60 percent of the literal detail against 72 percent for the same text on a page. They blinked half as often, and at matched rates they were no faster (Benedetto and colleagues 2015).

Going back is part of reading. Regressions are repairs: when readers were stopped from making them, understanding dropped, and not only for tricky sentences (Schotter, Tran and Rayner 2014).

There is a hopeful sign. A 2026 chapter found that a stream timed by the language itself kept more understanding at a higher nominal speed than a stream at one fixed pace. It is preliminary: how many people took part, and how, is not yet clear from what is published.

Speeds in words a minute.
SpeedWhat happensSource
200 to 400Skilled readers on a page.Rayner and colleagues 2016
250 nominalLiteral recall 60 percent against 72 on a page. Blinks halved.Benedetto and colleagues 2015
250 to 350Understanding matches a page.Di Nocera, Ricciardi and Juola 2018
400 to 450Understanding degrades.Di Nocera, Ricciardi and Juola 2018; Acklin and Papesh 2017
Above about 500Understanding of connected text collapses, inference before anything else.Di Nocera, Ricciardi and Juola 2018

What we built because of it.

A fixed target. Each word is centred by its width on the same spot, and the letter under the two ticks is red, so your eyes always have one place to rest. Where the eye lands matters: a word is recognised fastest when the eye lands slightly left of its centre, and slowest on its last letter (Nazir and colleagues 1992). A target near the middle of every word keeps you away from the slow end.

Natural pacing. In the classic tuning study, giving words with punctuation twice the time was the best liked change of all, and showing the most common short words for half the time did no harm (Castelhano and Muter 2001). Natural pacing works the same way: words that close a clause or a sentence stay longer, the commonest little words pass quicker, and long or rare words get a little more time.

Play restarts the sentence. When readers in an earlier Muter study could jump back two, four or eight words, or to the start of the sentence, they chose the sentence start most often (reported in Castelhano and Muter 2001). So you pause with a tap anywhere, and play picks up at the start of the sentence you were in.

The zones, and 250 to start. Flash Reader starts you at 250 words a minute, and the speed ruler shows where understanding holds and where it starts to slip, from the thresholds above. You can go faster. The ruler tells you what it costs.

Effective speed. Benedetto and colleagues counted the time spent on pauses when they measured speed, so we do too. The effective speed in your stats includes every pause, which makes it the honest number for how fast you read.

Rest breaks. Readers in the stream blinked half as often, which dries the eyes. On long reads Flash Reader reminds you to blink and look away, and later suggests a proper stop.

Chunk, a mode of its own. Showing a sentence sized piece at a time was as well liked as a normal page in the studies Castelhano and Muter review. Chunk shows two to four words at once, centred, for slower reading and longer texts. It has no red letter, because a group of words has no single point to aim at.

Reader, for when you stop. Feeling lost is a complaint the same paper documents. Reader shows the full text at your place, so a pause never leaves you guessing where you are.

The warm up. Each session starts a little slower and eases up to your speed over the opening sentences. That one is for comfort. No study tested it, and we do not claim one did.

Where it genuinely shines.

One word at a time is not the best way to read everything. It does well with:

  • Short texts and small screens. On a smartwatch, people read short passages faster in the stream with the same understanding, though they still preferred normal text (Gannon, He, Gao and Chaparro 2016).
  • Material you already know. Review and rereading, where the cost to literal detail matters least.
  • Readers with ADHD. In a study of 76 students, readers with ADHD understood about seven percent more in the stream than when their eyes had to move, while the other readers understood about six percent less. The authors call it “an initial step”, and so do we (Moussaoui, Siddiqi, Cheung and Niemeier 2025).
  • Some kinds of low vision, with real limits. For readers with low vision the stream beat scrolling text only once the letters were at least eight times the smallest size they could make out, and about a quarter read faster with scrolling (Fine and Peli 1998).

What we refuse to claim.

Speed reading has a long record of big promises. These are the ones you will not hear from us.

  • A thousand words a minute with full understanding. The review by Rayner and colleagues 2016 takes that promise apart.
  • That silencing your inner voice speeds you up. Hearing the words in your head helps you understand them, and the limit is the language, not the voice (Rayner and colleagues 2016).
  • That eye movements waste most of reading time. They take at most a tenth of it (Rayner and colleagues 2016).
  • That bolding the start of each word helps. A test with 2,074 readers found no gain in speed or understanding (Readwise 2022), and a peer reviewed study came to the same answer (Acta Psychologica 2024).
  • That anyone learns it in five minutes. There is no evidence for that (Benedetto and colleagues 2015).
  • That dark mode reads better. Dark text on a light background read better in daylight in the studies by Piepenbrock and colleagues, 2013 and 2014. The Ink theme is there for dim rooms and for taste.
  • That a dyslexia font helps. OpenDyslexic gave no gain in reading rate or accuracy (Wery and Diliberto 2017). Flash Reader offers it because some readers like it, with no promise attached.

Flash Reader has no comprehension test in this version, so it reports your speed and its zone, never how much you remember.

Sources.