Abstract
Purpose: We recently showed that humans extract precise information of angular velocity from rotational motions. The angular velocity at a given point is the ratio between the speed at that point and its distance from the center of rotation. Here, we investigate how precisely humans locate the center of rotation. Methods: Stimuli consisted of random-dot patterns undergoing rotations, viewed through an annular window. In a first experiment, we measured with a 4AFC task, how precisely a subject could estimate the position of the focus of rotation, parametric on the inner radius of the window. Next, we tested angular-velocity discrimination, with a single-stimulus paradigm. In this protocol, the distance between the center of rotation and the geometric center varied randomly between trials. Results: Subjects could not tell precisely in which of four quadrants of an 8° disk the center of the rotation was. We estimated that the localization error was about 1.5°. Interestingly, this error did not increase substantially when we substituted an annulus for the disk. We then wondered whether, in our original angular-velocity-estimation experiments, subjects overcame this large error by using the geometric center as a surrogate for the center of rotation. The second experiment tested whether such a substitution of centers occurred. Subjects had no problems discriminating between angular velocities despite random displacements of the center of rotation. Conclusions: Humans cannot locate the center of rotation precisely, but can discriminate well between angular velocities. Computational analysis suggests that the reason is that the center of rotation and the angular velocity are stable parameters of rotation models. In other words, an error in the estimate of one of these parameters does not affect the estimate of the other. Finally, we also conclude that the vanishing point of the rotational flow and its geometric center are not important estimators of the center of rotation.
Supported by NEI Grants EY08921 and EY11170.