Abstract
When a drawing of two circles is rotated in the image plane, a cylinder is vividly perceived (Musatti 1924). We hypothesize that this percept is a compromise between two preferences of the visual system: minimal deformation and slow motion. We quantified this percept and asked: Is the perceived shape identical across observers?
Observers, viewing monocularly, adjusted one circle's size relative to the other such that the perceived cylinder is uniformly thick, and indicated which circle was in front. The resultant circular size ratios differed reliably between observers.
To determine whether this difference was due to perceived viewing distance or shape or both, the cylinder was replaced by an imperfect cube. The same observers psychometrically determined whether the front square was larger than the back square, and the perceived distance was calculated from the square size ratio at PSE. The distances differed reliably between observers.
The perceived viewing distance was alternatively measured whereby observers touched a stimulus with an LED attached to a finger. The stimulus was the same two circles, except only one of which was visible. (A control with binocular viewing was also conducted.) All stimuli in our studies were presented using a two-way mirror at 45°. The LED's location was tracked by four cameras. The average viewing distance from multiple sessions was remarkably consistent with that from the cube psychometric measurement, within-subjects.
Combining the viewing distances and circular size ratios, we found that the perceived cylindrical shapes differed reliably between observers, indicating their differently weighted preferences.
We deliberately used an impoverished stimulus to study the way in which prior preferences compromise to reach a (bi-stable) percept, and how this compromise differed between observers. The principle of this compromise is hypothesized to be stimulus-independent, and this study is one step toward testing it.