The short version
Reading regressions are backward eye movements, the jumps you make to re-read something you have already passed. Even skilled readers send 10 to 15% of their eye movements backwards. They happen because comprehension failed, not because your eyes misbehaved. They are a repair, and suppressing them does not fix the fault.
How your eyes actually move across a page
Your eyes do not glide along a line of text. They jump.
Watch someone read and you will see it. The eye lands, holds still, then flicks forward and lands again. Reading is a sequence of stops, and the stops are where everything happens.
Get this picture straight and most speed reading advice collapses on contact. Our speed reading guide covers the full mechanism. This page is about the jumps that go the wrong way.
Fixations: where the reading happens
A fixation is a pause. It lasts about 250 milliseconds, and virtually all of your visual input happens inside it.
Fixations consume most of your reading time. This is not a small point. It means reading speed is set by what you do while your eyes are still, not by how fast you can move them.
The window of text you can use during a fixation is small. The eye resolves letters across about 20 characters, and you can identify words only 7 to 8 characters to the right of where you are looking. That is the word you are on, plus the next one.
Saccades: only 10% of the time
A saccade is the flick from one fixation to the next. It takes 25 to 30 milliseconds. It is fast, and it is cheap.
Saccades take up only about 10% of your reading time. Ninety per cent goes on the pauses.
So if you could make your eye movements twice as fast, and you cannot, you would save five per cent of your reading time. That is the entire prize on offer from eye-speed training, and it is why every product built around moving your eyes faster is aimed at the wrong target.
During a saccade you are effectively blind. The image is smeared and the brain discards it. You take in nothing while the eye is in flight, which is another reason the movement itself is not where your speed lives.
What a regression is
Most saccades go forward, left to right, in the direction of the text.
Some do not.
The back-saccade
A regression is a backward saccade. Your eye jumps back to a word or a phrase it has already fixated, sometimes a few words, sometimes to the start of the line, sometimes to the paragraph above.
You are rarely aware of the short ones. They happen in a fraction of a second and you do not experience them as a decision. The long ones you notice. Those are the moments when you realise you have lost the thread and go back to find it.
Ten to fifteen per cent of eye movements
Even skilled readers send 10 to 15% of their saccades backwards. This is not a beginner’s problem that good readers have trained away. It is a standing feature of competent reading.
Roughly one movement in eight goes back over ground you have already covered. Every one of them costs you a fixation, and fixations are where your reading time lives.
That is the honest size of it. It is not the whole of your reading speed and it is not nothing. Cut the wasted ones and you get a real gain, measured in tens of words per minute, not thousands. Anyone telling you that killing regressions will take you to 1,000 words per minute is selling something. We deal with that number on how to read 1,000 words per minute.
What a regression costs you
Work it through. A fixation is where your reading time goes. A regression buys you an extra fixation, plus the saccade to get there and the saccade to get back.
If one movement in eight goes backwards, and each one drags a fixation behind it, a real slice of your reading time is spent covering old ground. That is time, and it is attention, and attention is the scarcer of the two.
There is a second cost that nobody counts. Every time you go back, you drop the sentence you were building. You have to reassemble it from the start. On a long sentence, that is worse than the eye movement itself.
Why regressions happen, and it is not your eyes
This is the part the speed reading industry gets wrong, and it gets it wrong in a way that sells hardware.
Regressions come from weak comprehension
Your eye goes back because the meaning did not land.
That is the whole explanation. The word arrived, the brain failed to fit it into the sentence being built, and the system did the obvious thing. It went and got the word again.
Regressions are a symptom. They are the visible trace of a comprehension process that has stalled. The eye is not misbehaving. The eye is doing exactly what a well-designed system should do when the meaning breaks down: go back and check.
Which means training your eyes not to go back is like disconnecting a warning light. The light stops flashing. The fault stays.
The real causes, in order
Comprehension breaks down for a handful of reasons, and each has its own fix.
The text is too hard. Unfamiliar vocabulary, dense syntax, an argument that assumes knowledge you do not have. Your eye goes back because it should. No drill fixes this. Reading more in the field does.
Attention drifted. You read four lines while thinking about something else, then surfaced and had to go back. This is the largest single cause for most adults, and it has nothing to do with reading skill at all.
The preview failed. You get a partial look at the next word before your eye lands on it, and that look is what settles an ambiguous one. English is full of words that only make sense once you see what comes after. Lose the preview, on a bad screen or in a flashing-word app, and the sentence stalls and the eye goes back for it.
The sentence is genuinely ambiguous. Sometimes the writer built a garden path and you walked down it. Going back is the correct response, and we come to that below.
Why eye exercises will not fix this
You can find drills that train you to hold your eyes forward. Cover the line behind you with a card. Sweep a finger and never let yourself follow it backwards. Some courses build entire modules on it.
The drills do reduce regressions. That much is true.
What happens next is the problem. The comprehension failure that caused the regression is still there. You have simply removed the repair mechanism. So you read on, the meaning does not land, and you finish the page with a clean eye-movement trace and no idea what you have read.
Suppressing a regression does not create the understanding that the regression was trying to fetch.
This is the same error that runs through the whole field, and the apps make it worse. Speed reading apps flash single words at a fixed point, which removes regressions by making them physically impossible.
Give them their due. Removing eye movements does speed up word identification, and that cost is real. But identifying words is not reading. Comprehension needs a preview of what is coming next, and it needs the ability to go back when something does not fit. The apps remove both. Comprehension suffers, and it suffers more the longer the text runs.
They optimise the one part of reading that was never the constraint on understanding, and they break the two parts that were. Removing your ability to regress is not a feature. It is the fault.
Good readers regress less
The data is clear on this. Good readers show fewer back-saccades than average readers.
The temptation is to read that backwards and conclude that fewer regressions make you a good reader. It runs the other way. They regress less because they comprehend better. The comprehension comes first. The clean eye trace follows.
Think of what a strong reader has. Deep vocabulary, so few words stall the sentence. Prior knowledge of the subject, so the argument is half-predicted before it is read. Steady attention, so the thread is never dropped. None of it is in the eyes.
Every one of those reduces the number of times comprehension breaks down. So the eye has less repair work to do, and it goes back less often.
Chase the eye movements and you get nothing. Chase the comprehension and the eye movements sort themselves out.
Good readers also regress on purpose
Here is the finding that separates a serious account of reading from a sales pitch.
Good readers do not just regress less. They use regressions efficiently, to remove ambiguities.
A skilled reader hits a sentence that could mean two things, recognises it, and goes back to the word that decides it. One movement. Targeted. Then forward again.
A poor reader’s regressions are panic. A good reader’s regressions are surgery.
The average reader goes back because they have lost the thread and do not know where. So they re-read the line, or the paragraph, hoping the meaning will turn up. Sometimes it does. It is expensive either way.
The strong reader knows exactly which word failed. They fetch that word, resolve the ambiguity, and carry on. The regression is not a failure. It is a tool, and it is being used well.
So the goal is not zero regressions. Zero regressions on hard text would mean you were not checking anything, and a reader who never checks is a reader who does not know when they are lost. The goal is fewer accidental regressions and better deliberate ones.
Regressions on a screen
Most people now do most of their reading on glass, and it changes the numbers.
Long lines are the first problem. A wide browser window can run 120 characters across. At the end of one, the eye makes a big return sweep to find the start of the next line, and it lands wrong more often than you would like. Land on the wrong line and you regress, because nothing makes sense.
Scrolling is the second. Paper holds still. A page you scroll moves the text under your eyes, and every unplanned movement is a chance to lose your place.
The fixes are dull and they work. Narrow the column. Increase the line spacing. Scroll in blocks rather than continuously, so the text is still while you read it. Most reading modes in a browser do all three for you.
Newspapers worked this out a century before anyone measured a saccade. Narrow columns produce fewer return sweeps and fewer misses. More on this on reading on screen versus paper.
Why speed and comprehension travel together
Most people assume reading is a trade. Go faster, understand less. It is the intuition behind every worry about speed reading, and for skimming it is correct.
For real reading it is often wrong, and regressions are part of the reason why.
Regressions explain part of it. The reader who understands the sentence first time does not go back, so they are faster. The reader who is faster holds the sentence in working memory as a whole rather than a string of parts, so they understand it better. The loop runs both ways.
The honest route to fewer regressions therefore runs through comprehension, not eye control. Build the vocabulary. Build the background knowledge. Learn to hold the shape of an argument in your head while you read the parts of it. Do that and the backward jumps fall away on their own, because there is nothing left for them to repair. The methods are on how to improve reading comprehension.
It also sets a limit. On material you know nothing about, your regression rate will climb no matter how well you read, and it should. A physicist reads a physics paper in long clean sweeps. Hand them a legal contract and watch the eye start going backwards.
Children and struggling readers
Children regress far more than adults, and the reason is not their eyes either.
A child at around 150 words per minute is decoding. Each word is being worked out, sounded, checked. The sentence is being assembled from parts, and parts get dropped, so the eye goes back to fetch them. That is not a fault. That is what learning to read looks like from the outside. The numbers by age are on reading speed by age.
The trap is treating it as an eye problem. A child who regresses on every line does not need eye training. They need easier text, more of it, and a reason to want it. Volume builds fluency, and volume comes from books they will finish. See reading fluency and getting kids interested in reading.
The same holds for an adult who has stalled. If you regress constantly on work documents, the answer is rarely a drill. It is usually that the document is above your current knowledge of the subject, and the fix is to learn the subject.
How to cut the wasted regressions
You do not attack the eye movement. You attack the thing that caused it.
Read with a purpose
Attention is the lever, and it is free.
Before you open a document, write down what you want from it. One question is enough. A reader who is hunting for something does not drift, and a reader who does not drift does not have to go back and pick up the thread they dropped.
No drills. No hardware. Just a reason to concentrate.
Get more out of the span you have
You cannot make your span bigger. Learn what it is and stop paying for the promise.
The eye resolves letters across about 20 characters, and words come out of only 7 or 8 of them, to the right of where you are looking. One line, not ten. One word plus the next, not five. Keith Rayner’s verdict on training peripheral vision to take in a whole line at a glance was blunt: “simply biologically impossible.”
What the span does give you is a preview. The next word is half-processed before your eye arrives on it, and that preview is what stops an ambiguous word stalling the sentence. You get more out of it by knowing more words, not by stretching your eyes. The real numbers are on visual span in reading.
Preview before you read
Two minutes on the headings, the first line of each section and the conclusion. Now you know the shape of the argument.
A brain that knows where a text is going predicts the next sentence, and prediction is what stops comprehension breaking down mid-line. Fewer breakdowns, fewer regressions. This is why every serious method starts with a preview, and it is why every page on this site opens with one.
Use a finger or a pen, but use it honestly
A pointer keeps the eye on the line and it does reduce the aimless drift-and-recover cycle. That much works.
What it must not become is a ban on going back. Use the pointer to set a pace and hold the line. If you hit a sentence you genuinely did not understand, go back and get it. Then carry on. A reader who refuses to go back on principle is not reading, they are just looking.
What a long fixation is telling you
One more finding, because it reframes everything above.
A long fixation is not a visual bad habit. It is a stall. Your eye holds on a word because the brain has hit something it cannot process yet. An unfamiliar word. A clause that will not fit the sentence you were building. An argument that assumes knowledge you do not have.
The pause is the processing. The regression that follows is the repair.
Reading rate is limited by how fast your brain extracts meaning, not by how fast your eyes take in text.
That sentence, from the people who spent their careers measuring it, ends the argument. Skilled readers do fixate for less time. They also fixate less often, jump further and regress less. Every one of those is downstream of the same thing. They understand the sentence faster.
So the eye trace is a readout, not a cause. Drilling the readout changes the readout. It does not change the reading.
Who says so
In 2016 five psychologists published a review in Psychological Science in the Public Interest asking how we read and whether speed reading can help. It is the paper the industry avoids citing.
Keith Rayner led it. He spent forty years measuring what eyes do when they read, and with George McConkie he built the moving window paradigm, the instrument that measures how much text a reader takes in per fixation. He had no course and no app to sell.
Mary Potter co-signed it. In 1969 she invented Rapid Serial Visual Presentation, the flashing-word method every speed reading app is built on. The paper that sets the limits of RSVP carries the name of the woman who created it.
They are the reason the 10 to 15% figure on this page is a number and not a guess.
What to expect if you do the work
Cutting wasted regressions is worth real speed, and it will not transform you on its own.
The average adult reads about 250 words per minute. Normal reading with full comprehension caps out around 300, and the fastest competent adults reach about 550. Those are the honest numbers. Cut the wasted regressions and you move inside that range. You do not leave it.
It takes weeks. Speed reading courses do sometimes produce real gains, and the reason is not the drills. It is that the student read a great deal of text over eight to twelve weeks. There is no shortcut. Reading a lot is the shortcut.
Be careful what you measure. If you cut your regressions to nothing and your comprehension drops, you have not learned to read faster. You have learned to skim, which is a different tool with a different use. The difference is set out on skimming versus speed reading.
If you want the drills sequenced rather than scattered, the speed reading course puts them in order, and the average reading speed page tells you what a normal starting point looks like so you can judge your own.
The other half of the job is the inner voice. Regressions and subvocalization feed each other, because a word that has to be pronounced before it means anything is a word that is easy to lose. If you want the practical side of that, start with how to stop subvocalizing.
Measure before you train, or you will not know whether any of it worked. Take the speed reading test, get a real number, then come back and pick one thing to fix. Purpose first. Span second. Leave the eye exercises alone.
Sources
Rayner, K., Schotter, E. R., Masson, M. E. J., Potter, M. C., & Treiman, R. (2016). So Much to Read, So Little Time: How Do We Read, and Can Speed Reading Help? Psychological Science in the Public Interest, 17(1), 4–34.
