Abstract
The depth interval of a test target line in the sagittal plane is underestimated compared to a matching target line in the frontal plane. To reveal how such foreshortening depends on the representation of the ground surface, Exp 1 placed the matching target on the ground at (i) 2.5m; between the observer and the test target (near condition), (ii) 9.5m; beyond the test target (far condition); thus biasing the ground surface range to be sampled. We found that with monocular viewing the foreshortening is less in the near condition. This is attributable to an increased ability to sample the near ground surface for constructing an accurate global ground surface representation that is vital for space vision. In contrast, the far condition emphasizes the sampling of the distant ground. Exp 2 tested how varying viewing criteria, which presumably adjust how various processing levels weigh visual information, affect foreshortening. Using an L-shaped target on the ground with a 44cm frontal limb and an adjustable sagittal limb, observers judged the Ratio of Aspect Ratio (RAR) of the L-shaped target in the light with (i) a physical dimension criterion, (ii) an angular size/retinal image criterion. It is predicted that the latter criterion leads to a larger foreshortening since it is ideally impervious to size constancy. Our results support this prediction, but only to a point, as the measured-RAR from the retinal image criterion is larger than the predicted-RAR based solely on the retinal image of the L-shaped target. Further, when we compared our data from the retinal image criterion to the measured-RAR data in the dark (with a physical dimension criterion), where performance is based on the visual system's intrinsic bias (an implicit slant plane), we found that the foreshortening is similar. Thus, it is likely that the intrinsic bias determines the upper limit of foreshortening when the extrinsic visual information on the ground is either ignored or unavailable.
Supported by NIH (R01 EY014821)