Abstract
Small eye movements produce coherent and random image motions. As they are normally kept totally unnoticed, it is unclear to what degree they might limit motion perception. I measured motion detection threshold in the presence of artificial jitter simulating small eye movements, to estimate the equivalent noise (i.e., the artificial jitter that deteriorates performance just the way the visual system's internal noise would), and to compare its value to actual eye-movement statistics. On a gray background, two dense random-dot fields were presented concentrically for 853 ms. The disk pattern was moving coherently, whereas the annulus pattern was stationary. The annulus was shown constantly (condition R), invisible constantly (condition A), or flickered coherently at 9.4 Hz (condition F). The disk motion was the linear sum of Brownian random jitter and translation in one of eight directions. The observer was asked to identify the translation direction. For each condition, the translation speed at motion detection threshold was determined at several levels of jitter amplitudes, and the resulting threshold-versus-jitter function was fitted with the noisy linear amplifier model (e.g., Pelli & Farell, 1999, JOSA A). The estimated equivalent noise for condition R, where relative motion existed, was almost negligible. In contrast, that for condition A was significantly greater and was correlated with variability of small eye movements. Moreover, in condition F, which would evoke illusory jitter in the central disk even if it were static (Murakami, 2002, VSS), an even greater equivalent noise was found. These results suggest: (1) relative motion is a strong cue to reduce the visual system's internal noise; (2) small eye movements are the major source of it when no relative-motion cue is available; (3) the flickering annulus does not so much reduce noise as create a wrong relative-motion cue to eye-movement-originated image jitter, degrading performance of the current task.