Fig. 2: Controlling for visual cues.

A A custom LED display developed specifically for this study. This display, an array of 110 × 8 LEDs, each covering 1.9\({}^{{\prime} }\) on the horizontal meridian, was designed to provide no persistence and no background luminance. The insert shows the time-course of activity of one of the LEDs with the brief exposures used in our experiments, measured with a high-speed photocell. B–F Performance measured with 5 ms exposures (N = 7 subjects). Both B proportions of correct responses and C the discriminability index improved as the Vernier gap increased (*p = 7.21 × 10⁻⁴ in B and 1.65 × 10⁻³ in C; paired two-tailed t-tests) and were significantly above chance (B: **p = 4.5 × 10⁻⁴, ***p = 5 × 10⁻⁵; C: **p = 1.17 × 10⁻³, ***p = 7.49 × 10⁻⁴, two-tailed t-tests). Graphic conventions are as in Fig. 1H, I, with diamonds representing mean values ± SEM across subjects. D Probability of “Right” responses as a function of both the eye displacement in a trial (X_E) and the small misalignment on the retina caused by the display resolution (\({X}_{R} \, < \, 1.{9}^{{\prime} }\); one LED, see panel A). Negative and positive X_R indicate that, on the retina, the bottom bar was shifted to the left or right, respectively. Each diagonal line represents a Vernier offset X on the display. E Marginal probability of “Right” responses as a function of the eye displacement in a trial for both X_R < 0 and X_R > 0. The shaded regions represent one SEM. Perceptual reports are influenced by X_R (the oscillations in both curves) but primarily driven by X_E (the overall trend). F Mean performance ± SEM in the trials in which X_E and X_R possessed opposite signs. Subjects successfully completed the task even when X_R predicted the wrong response (*p = 0.0297 and **p = 3.54 × 10⁻⁴ above chance; two-tailed t-test). Source data are provided as a Source data file.