Your eyes jump three or four times a second, and you have never seen one

The movement you have never once caught

Look at your left eye in a mirror. Now look at your right eye. Now back again.

You will not catch the movement. Not once, not at any speed, no matter how hard you stare. Someone standing beside you sees your eyes flick across plainly. Vision researchers use exactly that failure as a plain description of the effect they study [1].

Your eyes do not glide across a page or a room. They jump. These jumps are called saccades, and during ordinary looking people make them about three to four times a second [2]. Between jumps the eye holds still for roughly 250 milliseconds, which works out to about four saccades a second [3]. Each jump itself lasts somewhere between 20 and 300 milliseconds. The longest of them reach speeds of up to 700 degrees a second [4]. Even common saccades exceed 400 degrees a second [5].

At those speeds the image sweeps across the back of your eye. It smears, the way a photograph smears when someone swings the camera during a long exposure [5]. Researchers call the result an intra-saccadic motion streak. The smear is real, it lands on your retina several times a second, and you have never seen it.

Nor do your eyes stop when you try to hold them still. Asked to fix on a point, the eye keeps making tiny jitters and slow drifts that no amount of effort removes [23]. Those tiny jitters carry their own brief dip in vision [7], and that suppression held across every condition the researchers tested [8].

Two accounts of how the smear is hidden, and the argument between them

There are two explanations, both first set out in 1950, and the field has not fully settled between them [1].

The first is active. When the brain sends the command to move the eye, it also sends a copy of that command to the visual areas, which then handle the self-made motion differently [1]. The strongest evidence is a matter of timing. Sensitivity to contrast starts falling 50 to 100 milliseconds before the eye begins to move. It reaches its lowest point as the movement starts, and recovers shortly after [1]. A drop that begins before the eye has moved cannot have been caused by the image.

The second account is passive, and needs no motor signal at all. The smear arrives sandwiched between a steady picture before it and a steady picture after it, and the later picture masks it [1][5]. Show a still eye the same sandwich and the brief motion also becomes hard to see [1].

Work published in the Journal of Vision in July 2026 compared the two directly, using patterns built to be visible only during a saccade [1]. Masking sharply cut reports of sudden-onset motion but barely touched smooth-onset motion. Repeated exposure did the opposite: it reduced detection of the smooth kind and left the sudden kind alone [1]. The authors conclude the two mechanisms are complementary rather than rival - masking handles abrupt transitions, and a kind of tuning-out handles the predictable motion your own eyes create [1].

There is a third layer underneath both. Recent work suggests sensitivity is already reduced in the retina itself, before anything reaches the brain [1]. Brain imaging points at the magnocellular pathway, the fast channel that carries motion [6]. The machinery that launches the movement also sits low in the brain. In monkeys, activity in the superior colliculus, a midbrain structure, predicts when the eyes will move. It does so far better than activity in primary visual cortex [24].

The honest position is that nobody can yet say which of these does most of the work.

What fills the gap

The brain does not hand you a blank frame and let you notice it. It fills the space with a prediction it has already prepared.

Before the eye moves, cells in the visual areas shift their sensitivity toward the place the eye is about to land. This is called predictive remapping. It has been recorded in early visual cortex, in the area V4, and in the frontal eye field, the region that plans eye movements [9].

A study published in eLife in January 2026 pushed this further. Using brain scanning, researchers made peripheral targets vanish before the eye could reach them. The target never landed on the centre of the retina. Even so, researchers could decode which target it had been from the part of primary visual cortex that serves the centre [10]. The brain had told the centre what to expect before the eye arrived.

The stitch is also fragile in a revealing way. What you glimpse in the corner of your eye speeds up your handling of the same object once you look straight at it. But inserting a brief blank period at the start of the new fixation abolished that benefit entirely [11]. The join only works if the next real picture arrives immediately.

Alongside all this, the signals tied to eye movement compress both space and time around the movement [12]. The record you are handed is not just missing a piece. It has been quietly edited around the missing piece.

The smear is not thrown away

Here is the turn. The streak you never see is not discarded - it is read and used.

Researchers recorded brain activity and eye position together while shifting a target in mid-flight [5]. When a continuous streak was available, the target’s new position could be read from the brain signal as early as 50 milliseconds after the movement ended. The small corrective jumps that follow a saccade also happened faster [5]. The smear helps your eye land properly. It just never reaches you.

The size of the jump matters too. Work from the University of Rochester and Weill Cornell found that saccade size filters what you can see afterwards. It sorts visual information by fineness of detail rather than by location. Even very small changes in how far the eye moved noticeably changed what was visible next [13]. A movement you cannot detect is setting the terms of the view that follows it.

In birds the same movement does something stranger still. Pigeon retinas have no blood vessels of the usual kind, and saccades were found to drive glucose levels inside the eye and shape the responses of visual neurons [14]. There the eye movement is not only aiming the camera. It is feeding it.

You do not experience any of this as something you did. Tested directly, the tiny fixational saccades produced no measurable sense of agency - no faint feeling of having caused the change [15]. The same holds for the stillness you think you see. Hold an image perfectly steady on the retina and it fades from view within seconds [25]. That is part of why the eye refuses to stop moving.

When the machinery slips

All of the above describes a system working. When it does not work, the concealment fails and the world visibly moves.

Nystagmus is a rhythmic, involuntary oscillation of one or both eyes, and it happens when the system that holds steady aim fails [16]. Patients describe it plainly as wobbly or dancing eyes, and how much sight is affected varies widely from person to person [17]. Oscillopsia is the related symptom in which still objects appear to jump, shake or shimmer, often worst while walking or driving [18].

One rare condition shows how little it takes. Superior oblique myokymia is a brief, small, high-frequency twitch of a single eye, usually caused by a blood vessel touching a nerve. Across 35 patients, 30 of them - 85.7 per cent - reported double vision or the world appearing to move [19]. The eye movement itself is tiny. The perceptual consequence is not.

Any of this belongs with a doctor. Jumping vision, new double vision or involuntary eye movement is something to have assessed by a qualified clinician, not something to work out from an article.

What is real here, and what is oversold

The mechanism is solid. Saccades happen, the smear is real, and you cannot see it. That much is not in dispute.

The commercial and diagnostic claims built on top deserve more care. Eye tracking as a clinical readout is a genuine research field, and differences do show up. Saccade and pursuit measures have been quantified against controls in multiple sclerosis [21]. Fixational eye movements differ after concussion [22]. But the effects are often modest, and honest studies say so. One study of early-stage Parkinson’s disease found only minimal involvement of the tiny fixational saccades in a working-memory task, with moderate effects on their size and nothing more [20].

The gap between “measurable in a laboratory” and “useful for you” is where most overselling lives. Treat any product promising to read your mind or your health from your eye movements as a claim to be checked, not a result already in.