A short history of RSVP

Before rapid serial visual presentation was a product, it was a piece of laboratory apparatus — a way of showing words one at a time, at a fixed point, for exactly as long as the experimenter chose. It was built to answer questions about how fast the mind can take language in. Decades later it was sold as a way to make that speed unlimited, which is a strange fate for an instrument whose founding results said it was not.

An instrument, not a technique

The method emerged in psycholinguistics in the 1960s and 1970s. Presenting words serially at one location gave experimenters something the ordinary page never could: complete control over exposure timing. On a page, a reader decides how long to look at each word and when to look back; with RSVP the experimenter decides, and the reader’s comprehension at each rate becomes a measurement. Mary Potter’s laboratory at MIT made the method famous, using it to probe how quickly words — and, strikingly, pictures — could be understood at all.

The early findings were double-edged, and both edges matter to everything that came later. People could recognise words presented remarkably fast — far above ordinary reading speeds — which suggested the eyes were not the limiting factor. But holding on to what was recognised was another matter: push the rate, and comprehension of connected sentences fell away even while individual words were still being caught. The instrument had located the bottleneck, and it was not vision. It was the language machinery downstream, assembling words into meaning at its own unhurried rate.

For decades that was roughly where matters rested. RSVP appeared in hundreds of experiments — on attention, on the brief blindness that follows spotting one target in a fast stream, on how much of a sentence survives at speed — and in almost none of them was it proposed as a way anyone should actually read. It was the treadmill in the laboratory, not a way of walking around. A handful of researchers did ask whether it might serve readers with vision loss, since it moves the text instead of the eyes. The idea stayed academic, waiting for a reason to leave the building.

2014, the year of 1,000 words per minute

Then came Spritz. In early 2014 a small company announced a licensing deal that would put its RSVP display on a major phone maker’s new devices, and for a season the technique was everywhere. The pitch was clean: your eyes waste time moving; fix the word at one point, mark an optimal letter in red, and read at 500, 600, 1,000 words per minute. Headlines about finishing a novel in ninety minutes wrote themselves, and the demonstration applet was genuinely startling the first time anyone tried it.

The number did the selling. A thousand words per minute is four or five times the adult baseline of 200–300, and the demonstrations felt real because words genuinely can be recognised at that rate — the laboratory had established exactly that, decades earlier. What the demonstrations did not show was the other half of the laboratory record: that recognising a stream of words and understanding a text are different achievements, and that the second one degrades as the rate climbs, whatever the first one is doing.

The ecosystem the moment left behind is worth listing, because most of it is still here. Spritz licensed its display into other people’s apps. Reedy carried the idea to Android; Outread to the iPhone; open-source spritzers appeared for every browser; smartwatch apps discovered that the technique fitted their screens better than lines of text ever had. The products vary in polish and in honesty of framing, but they are recognisably one family — and the family descends from apparatus, not from the speed-reading classrooms of the previous century.

Why the hype receded

The criticism, when it consolidated, was not that RSVP was fake. It was that the promise had misidentified the bottleneck. Reading speed is limited by language processing, not by eye movement; remove the eye movements entirely and reading gets somewhat faster — not multiples faster. A well-known 2016 review of the speed-reading evidence delivered the general verdict plainly: no known method multiplies reading speed without cost, and beyond roughly 400–500 words per minute on unfamiliar prose, what is happening is skimming, whatever the app calls it.

RSVP also carries a cost of its own, one the page never had: the words vanish. A reader on a page looks back — a regression — on roughly one saccade in seven, and many regressions are comprehension repairs, not errors. A stream permits none of them. For simple prose at moderate rates that may matter little; for dense argument it matters a great deal. Nor was the arc of overpromise and correction new. The same story had already played out with the classes and eye-training regimes of an earlier generation, told in what happened to Evelyn Wood.

The instrument that found the bottleneck was sold as the cure for it.

The 2014 problem, in one line

What survived scrutiny

Strip the hype away and a residue of real findings remains. Removing eye movement does remove a real cost: some 10–15% of reading time is spent with the eyes in motion and functionally blind, and RSVP genuinely recovers much of that, along with the return sweep, the lost lines and the crowded periphery that come with a page. Modest gains at moderate rates, with comprehension held and checked, are consistent with the evidence in a way the headline numbers never were.

The second survivor is geometric. A fixed-point display needs almost no space — one word’s worth — which makes RSVP nearly the only way to read continuous prose on a watch face or any screen too small for lines. That use case was barely mentioned in 2014 and has quietly become one of the strongest, as reading on the wrist sets out. The third survivor is the original one: laboratories still use RSVP daily, for exactly the reason it was built. Control over timing never stopped being valuable.

There is a candour dividend too, which nobody designed. Because a stream sets the rate explicitly, it makes the speed–comprehension trade visible in a way ordinary reading hides. A reader on a page who slows down for a hard paragraph rarely notices doing it; a reader in a stream who has to drop from 450 to 300 words per minute has been told something concrete — about the text, or about themselves. The instrument keeps measuring, even inside a consumer app. That was always its nature.

What a post-hype RSVP is for

The honest answer is that RSVP reading is a pace instrument. It cannot multiply your speed, but it can hold a speed — steadily, adjustably, without the dawdling and drifting that ordinary reading invites — and it can tell you the truth about what happens to your comprehension as the rate changes, which a page never quite can. Used that way, at rates the language system can actually sustain, it is less a speed-reading machine than a metronome with a memory.

That is the tradition this app works in, with one addition the laboratory never tested: every word arrives with a sense-matched icon, on the bet that restoring imagery protects comprehension as rates rise. We state it as a bet because that is what it is — nobody has run the study, including us. And the stream slows for long words, commas and sentence boundaries rather than ticking uniformly, because the founding experiments were clear that uniform timing is the enemy of connected sense. Sixty years on, the instrument is still teaching. The lesson was always about the reader, not the eyes.