An angular expansion hypothesis quantitatively accounts for several well-documented biases in space perception

摘要

??In the past few years, evidence has accumulated which suggests that many well-known spatial distortions might be related to systematic biases in two perceived angular variables a?? the perceived direction of gaze (e.g. gaze declination from horizontal) and the perceived optical slant of surfaces (Durgin and Li, 2011; Durgin and Li, in press; Durgin, Li and Hajnal, 2010; Li and Durgin, 2009, 2010; Li, Phillips and Durgin, 2011). It has been shown that in the range between 0?° and 50?°, both the perceived declination of gaze and the perceived optical slant of surfaces are expanded with a linear gain of about 1.5; perceived optical slant also increases as a function of viewing distance. An angular expansion hypothesis has been proposed which assumes that (1) there are perceptual biases in these two angular variables and (2) that the relationships between these and other perceived spatial variables otherwise still adhere to a Euclidean geometry. Specifically, it has been shown that such an angular expansion hypothesis can quantitatively account for known exaggerations in perceived uphill and downhill geographical slants, as well as underestimation in perceived egocentric distance, and also the foreshortening of perceived egocentric distance relative to perceived vertical distance (height). In the present study, we provide evidence that the perceived azimuthal (left-right) gaze direction is also systematically biased. In contrast to the bias in vertical gaze direction (i.e. gaze declination), bias in perceived azimuthal gaze direction is relatively small. By including a quantitative estimate of the perceptual distortion in azimuthal gaze direction, the angular expansion hypothesis can provide partial accounts of a broader range of phenomena, including exocentric distance anisotropy (i.e. foreshortening in sagittal distance relative to frontal distance), the vertical-horizontal illusion, and even size constancy.