A shaky hand holding a video camera invariably turns a treasured moment into an annoying, jittery momento. More recent consumer cameras thoughtfully offer stabilization mechanisms to compensate for our unsteady grip. Our eyes face a similar challenge in that they are constantly making small movements even when we try to maintain a fixed gaze. What should be substantial, distracting jitter passes completely unseen. Position changes from large eye movements (saccades) seem to be corrected on the basis of extraretinal signals such as the motor commands sent to the eye muscle, and the resulting motion responses seem to be simply switched off. But this approach is impracticable for incessant, small displacements, and here we describe a novel visual illusion that reveals a compensation mechanism based on visual motion signals. Observers were adapted to a patch of dynamic random noise and then viewed a larger pattern of static random noise. The static noise in the unadapted regions then appeared to 'jitter' coherently in random directions. Several observations indicate that this visual jitter directly reflects fixational eye movements. We propose a model that accounts for this illusion as well as the stability of the visual world during small and/or slow eye movements such as fixational drift, smooth pursuit and low-amplitude mechanical vibrations of the eyes.