Abstract
The classical rod-and-frame experiment (Witkin and Asch, 1948) employs a roll-tilted square frame that generates a substantial influence on the frontal plane orientation visually perceived as vertical (VPV). Since the original discovery of the frame's influence, the frame has been treated as a unitary perceptual object — a gestalt. We ask whether the frame's influence is any more than the additive combination of separate influences produced by the lines which constitute the sides of the frame as would be expected from a gestalt. VPVs were measured for 4 erect subjects monocularly viewing either a 1-line, 2-line, or 3-line section of the frame, or the full 4-line frame on a frontoparallel plane in otherwise total darkness. VPV was measured by S's setting of a 5 deg. long line to appear vertical. The visual stimulus was presented at each of 7 roll-tilts [−22.5 deg. (ccw) to +22.5 deg. (cw)]. Two frame sizes were used: For the large frame each side was 48 deg. long (eccentr=24 deg.); for the small frame each side was 16 deg. long (eccentr= 8 deg.). VPV changed linearly with roll-tilt for each of the 16 conditions. For the large frame, the average slopes of the VPV-vs-roll-tilt function (measuring sensitivity to induction) for the 1-line, 2-line, 3-line, and 4-line cases were 0.24, 0.27, 0.28, and 0.29, respectively; for the small frame, the average slopes were 0.13, 0.19, 0.23, and 0.24, respectively. Thus, the influence of the large frame was very near to the average of the influences of the individual lines; for the small frame the influence manifested a great deal of (linear) summation between the influences of the individual lines. However, for both cases, the full frame was considerably less influential than the simple additive combination of the influences of the individual lines; no higher level gestalt property is involved. Both small and large frame results fell along the same negatively accelerated function of the total length of the lines in the inducing field. We note that the individual lines of the roll-tilted frame also produce systematic influences on visually perceived eye level (VPEL), another dimension of egocentric space perception. Recently we described a 3-stage computational neural model (Vis Res, 2001) that accounts quantitatively for VPEL data with inducing lines of different orientations, lengths, and different numbers of lines. A similar model would account for the VPV results.
Supported by NIH grant EY 10534.