Visual suppression occurs during saccades
Aliases: saccadic omission · saccadic masking · trans-saccadic insensitivity
What it is
A saccade flings the fovea to a new location, usually in a few tens of milliseconds. During that motion, and for tens of milliseconds on either side, sensitivity to displacement, flicker, and low-spatial-frequency change drops. The phenomenon is saccadic suppression, sometimes called saccadic omission: people rarely report seeing the smeared world in flight. If a UI aligns a state change, cursor jump, or highlight with that window, the user may encode neither the old frame nor the new one.
Why it happens
Two pathways stack. One is physical smear: rapid rotation turns edges into streaks that are hard to read. The other is an active gain drop coupled to the motor command, especially for large-scale luminance transients and motion. Even when the image is artificially stabilized, detection of probes flashed during a saccade still worsens. Suppression begins about 50 ms before the eyes move and takes tens of milliseconds after they stop to recover, so the vulnerable window is wider than the movement itself. Fine detail was never going to be encoded in flight; what suppression mainly shuts down is the transient channel that would otherwise flag “something changed.” That is why a change timed to a saccade is harder to notice than the same change during a fixation.
Studying it
The classic method locks a probe—flash, displacement, contrast reversal—to different phases of a saccade and compares detection thresholds mid-saccade, pre-saccade, and during fixation. The tracker must sample fast enough to place the probe in the motion window; coils or high-speed video are far more trustworthy than a 30 Hz consumer camera. Comparisons typically vary probe spatial frequency, presence of a full-field transient, and saccade amplitude. Outcomes are detection rate, localization error, and reports of “did it flash.” In interface work, saccade-contingent refresh tests whether a status bar, CAPTCHA redraw, or cursor teleport lands exactly when the user is flinging the eyes. Laboratory probes are inserted by the experimenter; product jumps are produced by the layout, so ecological validity is not automatic. Suppression is stronger at low spatial frequencies; it hurts a panel-wide luminance switch more than high-contrast text that remains after the eyes land.
Where it stops holding
Not every change is eaten by suppression. A high-contrast difference that sits on the post-saccadic landing point and remains once fixation resumes can still be read on the next pause—what is missed is the moment of change, not the static result afterward. Slow compensatory movement from the vestibulo-ocular reflex (VOR) during head turns does not share the same suppression depth as a voluntary saccade; horizontal-saccade lab numbers do not transfer unchanged to compound eye-and-head motion in VR. Epilepsy, some drugs, and severe fatigue alter the time course. If a consumer tracker misses short saccades, an update that should have been suppressed will be scored as “the user was looking.”
Applying it
- Do not make the only status cue a brief flash or an untransitioned full-page replace and assume the user was staring at it; the critical result should still be readable on the next fixation.
- When cursor teleports, list reshuffles, or modal popups cannot avoid a jump, keep a static difference afterward (position, label, focus ring) so encoding can resume once suppression lifts.
- Verify by replaying saccade-locked recordings and listing every jump shorter than a fixation that leaves no discriminable after-mark.