Eye Movements in Reading: Fixations, Saccades and Span

When you read, your eyes do not glide along the line. They stop on a word for about a quarter of a second, jump seven to nine letters forward in a few hundredths of a second, and stop again. The stops are called fixations, the jumps are saccades, and the occasional jump backwards is a regression. All of the reading happens during the stops.

These movements have been measured for decades with eye trackers, and the figures are stable enough to test the claims that speed-reading courses make about them. The short version: your eyes are already efficient, and they are not what limits your reading speed.

What are fixations, saccades and regressions?

The figures below are for skilled adults reading English silently, taken from Rayner (2009).

These are averages with a lot of spread. A single fixation can be as short as 50 to 75 milliseconds or last more than half a second, and a saccade can cover one letter or more than fifteen. The spread is not noise. It follows the text.

Why do your eyes stop longer on some words and skip others?

Fixation time tracks mental work. Your eyes rest longer on words that are rare, long or unexpected in their sentence, and move on quickly from common, predictable ones. When a text gets harder, fixations get longer, saccades get shorter and regressions become more frequent.

Not every word is looked at directly. Content words such as nouns and verbs are fixated about 85 percent of the time, while short function words such as the and of are fixated only about 35 percent of the time. Words of two or three letters receive a fixation roughly a quarter of the time; words of eight letters or more are almost always fixated, and long words often get two fixations. Skipped words are still processed, on the fixation just before or just after the skip.

So the pattern you would see on an eye tracker is uneven: quick hops through easy stretches, a longer stop on a technical term, an occasional step back. That unevenness is the reading system matching its effort to the text, moment by moment.

What is the perceptual span, and can you widen it?

Sharp vision is confined to the fovea, the central two degrees or so of the visual field. Around it lies the parafovea, out to about five degrees on either side, where letters are progressively blurrier. This is the reason the eyes have to move at all.

The perceptual span was measured with a clever method introduced by McConkie and Rayner in 1975. An eye tracker controls the screen so that only a window of text around the point of gaze is readable and everything outside it is replaced with other letters. Researchers then shrink or grow the window to find the size at which reading becomes normal. For English, that window runs from 3–4 letters left of fixation to 14–15 letters to the right. Three further findings matter for speed claims:

• Words are identified in a much smaller zone. The word-identification span reaches only about seven letters to the right of fixation: the word you are on and perhaps the next one or two. Further out, you pick up word lengths and spaces, which helps aim the next saccade.
• Nothing useful comes from other lines. Readers do not take in information from the lines above or below the one they are reading.
• The span is set by language processing, not by eyesight. It is asymmetric in the direction of reading (larger to the left for Hebrew), far smaller in Chinese (about one character left and two to three right, because each character carries more), and narrower in beginning readers than in skilled ones.

That last point explains why the span resists training. It shrinks when the fixated word is hard to process and grows with reading skill, which suggests it is a result of fluent reading and not a separate visual capacity you can stretch. Attempts to train it directly, by flashing words in the periphery, have not led to faster reading with good comprehension, according to the review by Rayner et al. (2016). Our guide to peripheral vision training looks at what such drills are still good for.

In one fixation you can identify the word you are looking at and, at best, one or two short words after it. Claims about reading a whole line or a whole page in a single glance contradict what eye-tracking measures.

How much of reading time is spent moving the eyes?

Very little. With saccades of about 20 to 35 milliseconds and fixations of about 250 milliseconds, Rayner et al. (2016) estimate that only around 10 percent of reading time is spent with the eyes in motion. The brain also keeps processing the previous word during each jump, so that time is not wasted either.

This single figure undoes the most common sales pitch for speed-reading technology, which is that eye movements eat most of your reading time and removing them will set you free. Even a display that removed every saccade could save roughly a tenth. The other nine tenths go to recognising words and building meaning, which is the same work whatever the display.

What does this mean for speed-reading claims?

• "Stop regressing." Regressions are mostly short hops to the previous word, with longer ones when a sentence has been misread. In an experiment by Schotter, Tran and Rayner (2014), words were masked as soon as the eyes had passed them, so looking back was useless. On trials where readers tried to look back, comprehension accuracy was about 75 percent with normal text and around 50 percent, which is chance, with the mask. See regression in reading for which re-reads are worth reducing.
• "Take in more words per fixation." The seven-letter identification zone sets the ceiling. Reading in short phrases is realistic; reading in whole lines is not.
• "Eliminate eye movements with RSVP." RSVP shows words one at a time at a fixed spot. It saves the small saccade cost but also removes word skipping, the preview of the next word and the option to look back. It can be a comfortable way to hold a pace; it does not change how fast you understand.
• "Zigzag down the page." Recordings of trained speed readers show them fixating only some lines and skipping most. That is sampling the text, and in a separate study speed readers could not answer detail questions about parts they had not fixated.

The broader conclusion, set out in does speed reading work?, is that eye movements are controlled by comprehension and not the other way round. Change the understanding (more vocabulary, more knowledge of the subject) and the eye movements change by themselves: shorter fixations, longer saccades, fewer regressions.

Can you train your eye movements? How to practise sensibly

You cannot usefully drill fixations and saccades into a faster pattern, and there is no good evidence that eye exercises raise reading speed on ordinary text. What you can practise is steady attention and a deliberate pace, and you can use what you now know about your eyes to practise honestly.

1. Warm up for one minute. A 30-second gaze-following drill and one or two Schulte grids are enough to settle your attention. Treat them as a focus routine, not as eye strengthening.
2. Read with a pacer for five to ten minutes. Use a mode that keeps the full text visible, start at a speed where you follow nearly everything, and go up by 25 WPM at a time only while that stays true.
3. Allow yourself to look back. When a sentence does not make sense, pause and step back. That is the regression doing its job. What you want to reduce is re-reading because your mind wandered.
4. Try one-word display on easy text. Two or three minutes of RSVP at your normal pace shows you what reading without eye movements feels like, and what you lose with it.
5. Check understanding. After each session, write two or three sentences about what you read without looking.

One caution. This article describes typical reading in people with healthy vision. Eye exercises do not correct eyesight or treat eye conditions. If you regularly lose your place, see double, or get headaches or sore eyes when you read, an optometrist is the right person to ask. For a full training plan, start with how to read faster.