What Are Saccades? How Your Eyes Actually Move While Reading

Stand in front of a mirror and look at your left eye. Now shift your gaze to your right eye. Back to the left. Repeat a few times and try to catch the moment your eyes actually move.

You won't catch it. You'll see your eye stopped in one position, then stopped in the other, never caught in between. Someone watching your face from the side sees the movement perfectly clearly.

That's the first lesson about saccades, and it matters more than it seems for any discussion of reading speed.

What a saccade is, and what a fixation is

When you read, your eye doesn't glide smoothly along the line, even though that's exactly what it feels like. It stops, jumps, stops again.

The stop is called a fixation, and it lasts, on average, between 200 and 250 milliseconds. That's when information gets picked up — the only moment new text actually reaches your brain.

The jump between two stops is the saccade. It lasts between 20 and 40 milliseconds and covers roughly seven to nine character spaces. These figures come from Keith Rayner's 1998 review, which pulls together two decades of eye-tracking measurements.

Two honesty notes before we go further. These are averages, and averages move: on difficult text, fixations get longer and saccades get shorter, sometimes considerably. And they come from research on English and other alphabetic languages — the order of magnitude carries over to other scripts, the exact decimal doesn't.

Not every saccade moves forward. Between 10 and 15% are regressions — jumps backward over text you've already covered. For an average reader, that's roughly one move in eight. When a regression is a legitimate correction and when it's a sign your attention wandered is a separate discussion, but it's worth knowing that regressions aren't a flaw: they're part of the normal architecture.

There's also a special case, the return sweep: the long jump from the end of one line to the start of the next. It has a charming quirk. It rarely lands exactly right. It usually touches down a few characters too far to the right, and the eye immediately fires a small corrective saccade back to the left.

The thirty milliseconds when you're blind

Back to the mirror. That little experiment is the textbook illustration of a phenomenon with its own name: saccadic suppression. During a saccade, visual sensitivity drops sharply, and Rayner treats the consequence as a basic premise of the entire field of reading research: information is extracted during fixations, not during jumps.

It makes mechanical sense. An image sweeping across the retina at saccade speed would just be a blur. The brain would rather not process it at all than try to untangle it.

So your eye alternates between still frames and brief moments of blindness, and yet you perceive a smooth, continuous page, with no jumps at all. That continuity is a reconstruction, not an observation.

So why does the eye stop at all? Because sharp vision is packed into a tiny area at the center of your gaze. Rayner and Bertera showed as far back as 1979 what happens when you mask out exactly that center while someone reads: speed collapses, even though the rest of the page stays perfectly visible. The rest of the visual field can't pick up the slack. How wide the field you actually pull nameable letters from really is turns out to be a measurement in its own right, and the result disappoints everyone.

A fixation isn't a simple snapshot either. Schotter, Angele, and Rayner synthesized dozens of studies in 2012 showing that, while you're fixating one word, you're already pulling partial information from the next one: its first letters, its length, sometimes enough to skip it entirely. The saccade that follows is programmed on the basis of that preview, before it even starts.

The uncomfortable arithmetic: your eye stands still, it doesn't run

Put the numbers together.

A 225-millisecond fixation plus a 30-millisecond saccade gives a 255-millisecond cycle. Movement takes up just over a tenth of it. Take the other ends of the ranges, 250 plus 20, and it drops below 8%.

However you combine them, the result is the same: out of one minute of reading, your eye sits still for about 54 seconds.

Now reread the central instruction of half the speed-reading courses out there. "Move your eyes faster."

Faster through what, exactly? Through the part that takes up a tenth of the time. Even if you could eliminate movement entirely, teleporting your gaze instantly, you'd gain under 10%. A reader at 250 words per minute would land around 275. That's not nothing, but it's not the promise on the cover, either.

The instruction isn't wrong. It's aimed at nothing. It pushes exactly on the one piece that was never going to give much back.

What's fixed by design, and what actually moves

The duration of a saccade isn't a variable you can tune. For a given amplitude, it's nearly constant and predictable — a relationship so stable that physiologists gave it a name: the main sequence. A longer jump takes more time, a shorter one takes less, and that's about all there is to say about it.

On top of that, a saccade is ballistic. Once it's launched, the trajectory can't be corrected mid-flight. The decision about where you'll land is made entirely before the eye moves. There's no steering wheel.

So what actually moves? Two things, and both live somewhere other than where you'd expect.

The number of stops, first. There's real room here, and how many stops you actually make on a line depends on both the text and the width of the column you're reading.

The duration of the stops, second. Except it doesn't shorten on command. Just and Carpenter proposed the hypothesis in 1980 that has organized the field ever since: the eye stays on a word for exactly as long as processing it takes. A rare word holds you longer. A tangled sentence holds you longer. Fixation duration isn't a setting, it's an outcome.

Which means the thing you can actually change isn't in your eye. It's in your head.

Dot drills train a movement you already have

The classic "saccade training" exercise puts two dots on a page, a few centimeters apart, and asks you to jump your gaze between them, faster and faster, for a few minutes a day. Sometimes it's columns of numbers instead, sometimes a metronome.

Look at what it's actually asking for. Jumps between two large, fixed, predictable targets, with nothing in them to extract.

Your eye already makes this kind of jump tens of thousands of times a day. From the steering wheel to the rearview mirror, from a screen to a cup, from one person's face to another's. And those jumps are far larger than the saccades on a line of text, which cover under a centimeter on the page.

The drill trains, at an easy load, a movement you already perform under a much heavier load, all day, without thinking about it.

The problem isn't that the exercise is hard or dangerous. The problem is transfer. Rayner and colleagues reviewed the speed-reading literature in 2016 and reached a sober conclusion: the gains that show up in programs like this come from skipping text, not from an eye that's become faster, and comprehension pays the difference.

Spending five minutes jumping between two dots won't hurt you. It's a harmless warm-up, and eye fatigue after hours of reading is a real, separate problem. But it's a comfort routine, not an engine for speed.

What's actually worth training instead

If fixation duration follows processing, then the lever is the material itself and what you already know about it. Rayner shows in the same review that fixations shorten on words that are frequent, familiar, predictable from context. A better vocabulary really does speed up your eye — just indirectly, and slowly, over months of reading.

Three concrete things, in the order worth trying them.

Read in narrow columns when you have the choice. Fewer return sweeps, fewer corrections, fewer chances to lose your place between lines.

Watch your regressions for one page without correcting them. You'll find that many of them don't come from your eye at all, but from the moment your mind wandered off and you snapped back mechanically.

And measure yourself with comprehension included, not just a stopwatch. A faster pace with half the understanding isn't actually faster.

Your eye isn't the brake. It already spends nine-tenths of its time standing still and working. What changes is how quickly that work gets done.

Measure your speed and comprehension