Motivation. Recent evidence suggests that extra-retinal signals play an important role in the perception of 3D structure from motion (SfM). According to the stationarity assumption (SA, Wexler, Lamouret, & Droulez, 2001), a correct solution to the SfM problem can be found for a moving observer viewing a stationary object, by assuming a veridical estimate of the observer's translation. According to SA, perception of surface slant should be: (1) more accurate for active than for passive vision; (2) more accurate for natural than for virtual objects (because of the cue-conflict inherent to virtual stimuli). Method. We performed three experiments involving both active and passive observers. The task was to estimate the slant of a static random-dot planar surface. We manipulated the surface slant and the translation speed the observer's head. The translational displacements and orientation of the participant's head were recorded on-time by an Optotrack Certus system and the virtual stimuli were generated in real time on a high-definition CRT (passive observers received a replay of the same optic flow). Natural simuli were dotted planar surfaces. Results. Perceived surface orientation increased with both increasing slant and translation velocity. These systematic biases were found for both virtual and natural stimuli, and for both active and passive observers. Conclusion. Extra-retinal information available to active vision is not sufficient for a veridical solution to the SfM problem. Also for active vision, the first-order properties of the optic flow are the main determinant of perceived surface slant. If the first-order properties of the optic flow are kept constant, the surface is perceived as having a constant orientation, regardless of actual orientation; if the first-order properties of the optic flow are varied (e.g., by manipulating the translation speed of the observer's head), surface slant is perceived as varying, regardless of whether distal slant is constant.
展开▼
