What happened to Evelyn Wood

For about twenty years, a large number of intelligent people believed that reading could be made four or five times faster with comprehension intact, and they believed it because they had watched someone do it. The demonstrations were real. The graduates were not frauds. What went wrong is more interesting than a hoax, and the lesson it leaves is the one every reading product since has had to answer for, ours included.

Evelyn Wood was an American schoolteacher who, in the late 1950s, put together a method she called Reading Dynamics. The origin story, as it is usually told, involves noticing that a handful of people read extraordinarily fast without apparent effort, and that sweeping a hand down the page seemed to help. She built a course around the observation. Within a few years it had become an institution with branches across the country, and within a decade it was part of the furniture of American self-improvement.

The scale is easy to underestimate now. This was not a niche product. Reading Dynamics was taken up by professionals, by university students, and — most usefully for its reputation — by staff inside more than one American administration. Presidential interest in speed reading was reported widely and repeatedly, and that association did more for the method than any experiment could have. When the White House is said to be taking your course, nobody asks to see the comprehension scores.

The demonstration

It is worth being precise about what was being promised, because the modern versions of the promise are vaguer. Reading Dynamics did not offer a modest improvement. Rates in the thousands of words per minute were routinely described, and the marketing explicitly held that comprehension would rise rather than fall, on the argument that the trained reader was taking in complete thoughts instead of dismantling sentences one word at a time. That second half — comprehension improving with speed — is the claim that mattered, and it is the one nobody was checking.

The claimed mechanism had three parts. First, the eye was said to be capable of taking in whole blocks of text at a glance rather than a few characters at a time — so a trained reader would move down a page vertically, absorbing bands of print rather than tracking lines. Second, the hand was used as a pacer, sweeping ahead of the eye to set the rhythm and to stop the reader drifting backwards. Third, the inner voice was to be eliminated, on the argument that silently pronouncing every word chains you to speaking rate.

In the room, this looked extraordinary. A graduate would take an unseen book, move a hand down page after page at a pace that made a spectator laugh, close it, and then talk fluently about what was in it. The demonstration is the whole story of speed reading in miniature, because everything about it was true except the interpretation. The pages really were turned. The account really was coherent. The question nobody in the room was equipped to ask was where the coherence had come from.

What happened when it was measured

The laboratory answer arrived steadily rather than in one blow. Once reading rate and comprehension were measured together, on material the reader had not chosen, the pattern was consistent: rate went up substantially and comprehension came down to roughly what you would expect from skimming. Ask general questions about a familiar topic and scores stayed respectable. Ask about detail, about the structure of an argument, about anything that could not be guessed from a heading and a first sentence, and they fell away.

The reason turned out to be more flattering to the readers than to the method. Skilled adults are very good at reconstruction. Given a topic, a few sampled phrases, a sense of how this kind of text is usually organised and a stock of prior knowledge, a competent person can produce an account of a chapter that sounds exactly like the account of someone who read it. They are not lying. They genuinely feel they have read it. What they have done is build a plausible model of the text from fragments and expectations.

The eye-movement work closed the other door. The perceptual span in reading is a property of the visual system, not a muscle to be trained, and it is far narrower than whole-block perception requires. Subvocalisation turned out not to be a defect either: it is measurable in the speech musculature and it supports comprehension on difficult material, so suppressing it removes something that was working for you. A well-known 2016 review of the speed-reading evidence gathered this up and concluded what the individual studies had implied for decades — the trade-off between speed and comprehension is real, and it is not a training problem.

Everything in the demonstration was true except the interpretation.

On watching a page turn

What the training actually gave people

This is where most accounts stop, and stopping there is unfair. Reading Dynamics was wrong about the mechanism and still changed how a great many people read, because the drills smuggled in a set of genuinely useful habits under a false explanation. That happens more often in the history of self-improvement than anyone likes to admit.

• Permission to skim. Most educated adults are taught that skipping is cheating. The course licensed strategic skimming as a technique, which is exactly what it is when you choose it deliberately.
• Triage. Deciding in ninety seconds whether a document deserves an hour is a distinct skill from reading, and it is the one that most changes how much you get through in a week.
• Pacing. A hand moving down the page is a metronome. Readers who were drifting, re-reading out of habit and stopping to check their phone were given an external rhythm, and rhythm alone recovers a surprising amount.
• The discovery of the ceiling. Most people read considerably below their own comfortable maximum, out of habit rather than capacity. Finding that out is worth the price of the course even if every explanation attached to it was wrong.

There is also the confidence, which is easy to sneer at and shouldn't be. A reader who believes the pile is manageable opens it. A reader who has decided that reading is slow and effortful puts it off. If a course changed the second reader into the first, it did something, even though the something is not what appeared on the certificate.

Every generation gets its own version

The claim never dies; it changes vehicle. Mid-century it was a classroom course and a hand sweeping down a page. In the 1990s it was software that flashed text at you on a desktop computer. In the 2010s a word-stream engine went viral on the argument that eye movement was the thief, and the app stores filled with variants within a year. Now the pitch tends to be a language model that reads the thing on your behalf, which at least has the virtue of being clear about what it has replaced.

What stays constant is the shape of the offer: a large multiple, comprehension asserted rather than measured, and a demonstration that feels convincing while you are inside it. The families differ in mechanism, which is worth understanding properly and is the subject of our field guide. They do not differ much in how they talk about themselves.

Which brings us to us

We are selling a reading app. We are in this lineage, not observing it from outside. Our own family removes eye movement entirely, which buys the motion cost and then runs straight into the finding that the bottleneck is language rather than vision. Our distinguishing idea — pairing every word with an image chosen by sense, so the imagery channel is carrying something while the words move — is our own reasoning built on findings we did not generate. It has not been tested as a combination by anyone, and a person reading this in 2050 may well file it alongside whole-block perception.

The one structural difference we can point to is not a claim about reading. It is that the app contains an instrument that can prove us wrong and we ship it switched on: a recall test after each passage, mixing real targets with plausible distractors, with the score printed next to your speed and exportable. Evelyn Wood's demonstrations had no such instrument in the room, which is the entire reason they persuaded people for twenty years. If you push the app to 700 words per minute and your recall collapses, the app will tell you so, and you should believe the app over us.