Abstract
We blink frequently, yet most of our eye blinks go unnoticed because of suppressed visual input during blinks. To maintain visual continuity in a dynamic world, we need to predict the behavior of objects during visual suppression. For example, while tracking a moving object, we would ideally minimize the pursuit error at the moment our eyes open after a blink, by predicting the motion of the object.
In this experiment, we used a 2AFC task to investigate if participants (N=23) performing a horizontal smooth-pursuit could determine perceptually if the target had jumped forward or backward during an eye blink. The behavioral data indicates that participants consistently perceived backward jumps as more continuous.
We also analyzed participants’ gaze data during the task to compare the smooth pursuit gain at the instant just before and after the blink. We found that the pursuit gain was reduced just after the blink, i.e., gaze was lagging behind the target.
During the period of eyelid closure, there is no retinal slippage. Hence, we need to rely on our internal prediction model to accurately estimate the target motion. Many studies have shown that we underestimate the duration of eye blinks. Assuming that pursuit accuracy before the blink is an indication of the robustness of our internal model, a possible explanation for the decrease in pursuit gain might be that our prediction model uses an underestimated duration of eye blinks to predict target locations after blinks. Thus, the temporal underestimation of a blink may later manifest as a spatial underestimation in the form of a pursuit error. To test this, we modeled the subjects’ smooth pursuit across blinks using kinematics. Our model shows that the duration of a blink is compressed by about 61%, consistent with values reported in recent studies.