Illusory motion in Enigma. (A) Enigma (1, 3). As one views the image the concentric purple annuli fill with rapid illusory motion. Upon careful fixation, the perception of illusory motion decreases or even stops completely. Subsequent eye movements bring back the illusory percept. (B) Simplified version of Enigma used in the present experiments (spatial characteristics slightly modified from those used in the actual experiments, for clarity). Illusory motion is seen to travel lengthwise along the gray stripe. (C) An epoch from a trial during the experiment. (Top) Schematic representation of the stimulus, which does not change over time. (Middle) Perception of the stimulus: the gray target stripe intermittently fills with fast or slow illusory motion. (Bottom) Subjects' report of their perception.

Microsaccades can drive the perception of illusory motion in Enigma. (A and B) Perceptual transitions are correlated with previous microsaccades. (A) Average microsaccade rates around reported transitions toward faster (red) vs. slower/no (blue) illusory motion. (B) Average probability of transitions after microsaccade onset. The results are expressed as the percentage increase over a correlation where microsaccades happened at random times. Color code as in A. (C–E) Changes in microsaccade rates are not triggered by the perceptual transitions. Color code as in A. (C) Distribution of latencies from correctly identified physical transitions to reported transitions (control experiment). Dashed vertical line and surrounding gray band: average latency and standard deviation (SD) for all subjects (520 ± 190 ms). (Inset) Latency and performance parameters for individual subjects. (D) Average microsaccade rates around physical transitions in motion speed. Microsaccade rates decreased after both increases and decreases in physical motion speed. The vertical gray band indicates the estimated moment of the reported transitions with respect to the physical transitions (mean latency and SD from C). (E) Average microsaccade rates around reported transitions in nonillusory motion speed. The same data as in D are now realigned to the time of the subjects' report. The vertical gray band indicates the estimated moment of the physical transitions with respect to the reported transitions (from C). (F) Microsaccade rates around reported transitions in Enigma. Shown is the same graph as in A, now indicating the estimated moment of the perceived (illusory) transitions (vertical gray band; extrapolated from C). Microsaccade rates depart from baseline ≈740 ms before the time of the perceived transitions. (G and H) ROC analysis. The ideal observer can predict the type of illusory transition in Enigma (faster vs. slower/no motion) based on microsaccade rate. (G) The ideal observer's prediction (green line) becomes significantly better than chance ≈1,060 ms before the reported transitions (i.e., ≈540 ms before the perceptual transitions; estimated time of perceptual transitions indicated by vertical gray band). Horizontal dashed line indicates chance level (probability = 0.5). Thin gray line indicates the significance level, the level at which the ideal observer performs above chance with P < 0.01 (determined by permutation analysis, see SI Methods for details). Significance is reached whenever the green line is above the gray line. (H) The ideal observer cannot predict the type of transition in motion speed (faster vs. slower/no motion) for nonillusory moving objects. Other details are as in G. All of the panels show the average result for all subjects with the shaded red, blue and green areas indicating the SEM across subjects.