Abstract
Continuous psychophysics is a new methodological approach gaining traction in vision science, because it enables measuring perceptual uncertainty significantly faster compared to classical psychophysics. Instead of performing forced choice trials, subjects track a target on a screen, promising broad applicability with untrained subjects. However, due to the naturalistic task, several perceptual, cognitive, and motor parameters contribute to behavior in nontrivial ways, which are not yet well understood. For example, because subjects track the target with a cursor, there is not only visual uncertainty about the target, but also about the cursor. Here, we address the question of how visual uncertainty about the cursor affects tracking behavior. Subjects tracked a target, which moved on a random walk, using a mouse cursor. In two conditions, we manipulated visual uncertainty about either the target's or the cursor's position. In the target uncertain condition, the target was displayed as a dot cloud pattern, while the cursor was displayed as a single dot and vice versa in the cursor uncertain condition. The dots in the pattern were drawn from a Gaussian distribution, and we manipulated the perceptual uncertainty by changing its standard deviation. For different levels of target uncertainty, we observed differences in both root mean squared error (RMSE) and temporal lag, in accordance with previous results. For different levels of cursor uncertainty, we observed a similar effect on the RMSE, but a smaller effect on temporal lag. We conclude by showing why visual uncertainty about the cursor and the target affect tracking behavior in different ways with our recently developed optimal control model of continuous psychophysics. In particular, because the subjects control the cursor, but not the target, the uncertain visual information is weighted in different ways against internal model predictions, resulting in the observed behavioral effects.