Purchase this article with an account.
George Sperling, Joetta Gobell, Chia-huei Tseng; Trapezoidal illusions: Windsurfers versus runways. Journal of Vision 2009;9(8):691. doi: https://doi.org/10.1167/9.8.691.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We demonstrate here a windsurfer illusion, a naturally occurring trapezoidal illusion in which the small end of a sail viewed at a distance appears to be pointed away from the observer even when it is pointed toward them. This naturally occurring illusion is so compelling that observers are both unaware of their gross perceptual misinterpretation and are unable to perceive the scene veridically. The windsurfer illusion is largely context-free; we demonstrate that context can be critical: the same 2D shape on the different walls, ceiling, or floor of a room produces vastly different shape interpretations. Four experiments further investigate the joint effects of object shape, retinal orientation, head position, relative motion, and the direction of gravity on automatic depth perception of white outline trapezoids rotating back-and-forth on a black background. The trapezoids are perceived as rotating in 3D; observers report the side that appears to be nearest. Left- and right- pointed trapezoids are perceived mirror-equivalently; the longer edge of the trapezoid is reported in front on 74% of trials (traditional trapezoidal illusion, “windsurfer effect”). When the same display is rotated 90 deg to produce a runway retinal configuration, there is striking perceptual asymmetry: the long edge is perceived in front 97% of trials when it is on the bottom but only 43% when it is on top (“runway effect”). The runway effect also occurs when the head is tilted 90 deg or when displays on the ceiling are viewed from the floor. The strikingly different 3D perceptions produced by similar 2D screen trapezoids are quantitatively explained by a model that assumes just three bias factors contribute additively to perception: (1) Implicit linear perspective, (2) assumed viewing from above in head-centered coordinates, and (3) relative height in the retinal field. The model accounts for 93% of the variance of the data.
This PDF is available to Subscribers Only