Contrary to what you and I might experience when we explore the world, our eyes do not provide us with a continuous and stable view of it. They jump several times each second in rapid movements called saccades. Because the eye projects the world onto the retina, we should see the world shift abruptly each time the eyes move—the visual scene should feel unstable, yet the brain uses sophisticated mechanisms that ensure it does not.
A recent study, titled “High-fidelity but hypometric spatial localization of afterimages across saccades”, published in Science Advances, found that common afterimages, like the faint shape we all see after looking at a bright light, provide insights into how the brain achieves that stability. The work was conducted by Richard Schweitzer, Thomas Seel, Jörg Raisch, and Martin Rolfs, researchers from the Cluster of Excellence Science of Intelligence in Berlin. Using afterimages as an experimental tool, the team set out to measure how accurately the brain predicts the visual consequences of its own eye movements. The study reveals that these predictions are very accurate, but are subject to systematic errors.
The phenomenon that afterimages follow wherever we direct our gaze was already documented by Aristotle and reveals a striking dissociation: While the visual world appears stable when eye movements constantly shift the world across the retina, afterimages seem to drift across the scene despite remaining fixed on the retina. Visual stability and the apparent motion of afterimages may therefore be two sides of the same coin–the brain’s attempt to account for its own eye movements.
To examine these mechanisms, the experiments had to be conducted in complete darkness—the opposite to normal everyday vision where the richness of the visual scene provides constant feedback that helps the brain estimate each eye movement. Sitting in the dark, participants first fixated a bright flash that created an afterimage and then looked over to a second, briefly illuminated light source. Then, once the afterimage became clearly visible, brief probe lights appeared at specific positions, and participants reported whether the afterimage seemed to lie to the left of the probe light, to the right, or directly aligned with it.
From these responses the researchers could estimate where the afterimage was perceived. Eye-tracking measurements monitored where participants really looked—allowing the researchers to determine how closely perception tracked the actual movement of the eye.
Afterimages closely followed the eyes: The larger the eye movement, the farther the afterimage appeared to move in space. Yet this match was not perfect. “On average, the perceived shift of the afterimage reached about 94 percent of the actual eye movement,” says Richard Schweitzer, lead author of the study. “In practical terms, perception follows eye movements very closely, but not perfectly.”
This small undershoot, called hypometria, held across individuals and remained consistent across different directions and sizes of eye movements. This suggests a systematic inaccuracy in the brain’s prediction rather than a random error. Even though the difference is subtle enough that most people never consciously notice it, understanding it requires looking at how the brain updates space after each eye movement.
Now what actually determines where the afterimage appears? One possibility is that its perceived location is determined based on visual feedback that becomes available after each eye movement. The researchers tested this directly. In some trials, the saccade target (i.e., the light that participants were told to follow) remained briefly visible after the eye landed; in others it was shifted slightly to create deliberately misleading feedback.
Neither manipulation changed where participants perceived the afterimage. Indeed, there is good evidence that the brain uses an internal copy of the command sent to the eye muscles, called an efference copy, to predict how the visual scene should shift. That signal effectively tells the brain: “the eyes just moved this far”, allowing perception to anticipate the consequences of the movement instead of waiting for new visual input to correct perception afterward. Movements of afterimages now reveal that visual predictions derived from the efference copy fall short of the eye movement’s true consequences.
That raises a natural follow-up question: If perception depends on the brain’s efference copy, what happens when those movements themselves change? Eye movements are not fixed. When the eyes consistently miss their targets—say, due to fatigue of the eye muscles—people gradually adjust how far their eyes move. This process, known as saccadic adaptation, can be introduced in the lab by shifting the target of an eye movement with each saccade. This trick provided another insight into the brain’s prediction of the visual consequences of eye movements: As participants’ saccades became shorter through adaptation, the perceived shift of the afterimage shortened with them. Yet, the small systematic undershoot remained, whether saccades were adapted or not.
That remaining mismatch may not be a flaw. Natural eye movements often fall slightly short of their targets, so it makes sense that the brain’s internal estimate reflects this tendency. Assuming a stable visual environment–where objects do not suddenly change their positions during saccades–observers can use visual cues in everyday life to learn how much the visual scene typically changes after a given eye movement. If saccades tend to fall slightly short, it would only be reasonable to expect a slightly smaller visual shift as well. What may matter more than perfect accuracy of the movement is that perception stays reliably aligned with it.
If afterimages remain fixed on the retina, then why do they appear to move with our gaze? One possible explanation is that the brain uses its knowledge about the consequences of an upcoming eye movement to predict where an object should appear on the retina after the saccade –a process known as predictive remapping. If this prediction is accurate and matches the object’s actual position, as confirmed by visual feedback, the object is perceived as stable.
In normal visual environments this works well. But an afterimage inevitably violates this prediction: because it stays fixed on the retina while the eyes move, the brain can only conclude that it moved in the same direction. In this case, the size of the prediction error corresponds to the size of predicted visual change.
“Afterimages become a useful tool for studying how the brain keeps the visual world stable by predicting the sensory consequences of its own movements,” says Schweitzer. Understanding these predictive mechanisms may provide insights beyond basic vision science, for example in robotics, virtual reality, and clinical studies of eye-movement disorders, where linking movement with sensory consequences reliably is essential.
• Human eyes move several times per second in rapid jumps called saccades, yet the visual world appears stable.
• Afterimages allow researchers to isolate the brain’s internal signals that track these eye movements.
• The brain predicts the visual consequences of eye movements with striking accuracy.
• However, perceived afterimage movement slightly undershoots the true eye movement, reaching about 94% of the actual shift.
• This consistent undershoot suggests a small but expectable bias in the brain’s internal estimate of eye movement-induced change.
• The findings help explain how the brain keeps the visual world stable despite constant motion of the eyes.
Science Advances
Experimental study
People
High-fidelity but hypometric spatial localization of afterimages across saccades
13-Mar-2026