Texture-density aftereffects to filled-in and suppressed portions of textures

((F.H.Durgin and R.C.Cole)) Swarthmore College.
(1997) Investigative Ophthalmology & Visual Science, 38, S636.

Purpose: Interocular transfer of texture density aftereffects was investigated in the filled-in blind spot region and in normal peripheral retinal regions to clarify the relationship between interpolative processes, perceptual experience, and perceptual adaptation. Partial "occlusion" of a 10° diam. monocular texture in the periphery may be accomplished by a single bold circular contour 5° in diameter in other eye producing a "donut" of texture. Effects of such suppression on aftereffect strength were compared to those of blind spot interpolation and of actual peripheral texture donuts.

Methods: Twenty-four observers were monocularly adapted using a dual monitor stereoscope to dense textures in one of three experimental conditions. Gaze was monitored throughout. Adapted regions were 10° diam. centered on blind spot and on a comparably peripheral region. Test regions were either 10° or 5° diam. (concentric) so that transfer to central portion in the region corresponding to blind spot or to the hole of a texture donut could be assessed. Aftereffect strength, expressed as log ratio at PSE between target and unadapted regions, was measured with a staircase method.

Results: In the first condition, aftereffects to textures presented across the blind spot were found in the corresponding interior region of the contralateral eye (aftereffect size = -0.36 -- ln(ratio at PSE)), and these did not differ from those found with actually complete textures (-0.41). In the second condition, aftereffects to texture annuli (-0.16) in normal retina were found to be significantly weaker than those to textures filled in across the blind spot (-0.45), t(7) = 3.37, p < .05. In the third condition the central portion of the adapting textures were perceptually suppressed by a rivalrous stimulus presented to the contralateral eye. In this condition, aftereffects to blind-spot-presented (-0.27) and to actually complete textures (-0.31) did not differ when assessed with textures presented in the central portion.

Conclusions: Interocular aftereffects were found to filled-in textures at the blind spot which were significantly stronger than to annular textures. Strong aftereffects were also found even when binocular rivalry suppressed the perception of filling-in. Texture can act "filled-in" with respect to adaptation without being experienced as seen.

Supported by the Swarthmore College Research Fund and the Howard Hughes Medical Institute.

The Rubber Pencil Illusion: Beyond Persistence

((A.C. Huk1, F.H. Durgin2, T.A. Banton3, D. Aks4, D.A. Lewis2, S.Gold2, and R. Jain2)) Stanford University1, Swarthmore College2, University of Virginia3, University of Wisconsin--Whitewater4
(1997)Investigative Ophthalmology & Visual Science, 38, S378.

Purpose: Pomerantz (1983) argued that the rubber pencil illusion was due to the visual persistence of a curved spatial envelope, however, Ishiguchi (1988) found that the motion of two dots in complex motion (endpoints of a virtual bar) was sufficient to produce the perception of a bending rod. Moreover, the rubber pencil illusion may be produced when much of the middle of the pencil is occluded. Because the curved spatial envelope discovered by Pomerantz is sufficient, but not necessary to produce the rubber pencil illusion we sought to investigate parameters and conditions necessary to produce the perception of bending behind an occluder.

Methods: Apparent motion sequences were constructed in which a black rod was shown behind a gray occluding rectangle. The rod protruded equally on each side when horizontal, but rotated 24, 18, or 162 deg clockwise and counterclockwise about an (occluded) point 1/3 along its length. A rubber pencil illusion is defined as an appearance of "flexing" such that the right end seems to rotate more than the left end. The three canonical frames of the sequence are shown above. Such frames produce an impression of flexing (SOAs 100, 150, and 200 msec). Observers fixated the center of the gray rectangle and adjusted the rotation of the right end of the bar until it seemed rigidly linked to the left. Two frame sequences (no horizontal bar), which fail to produce subjective "flexing impressions" were also tested, as a control for the Poggendorf Effect.

Results: The point of subjective rigidity was systematically distorted by this manipulation.


Conclusions: Partly occluded, kinematically simple stimuli (off-center rigid rotations in the plane) are sufficient to produce a kind of rubber pencil illusion. Differences between 2-frame and 4-frame sequences suggest that more than Poggendorf Effects are relevant to the apparent misalignment of the left and right bars. Video analysis of both the rubber pencil illusion and of true rubber rods, suggests that persistence, even in video frames, supports the perception of curvature. However, differential motion amplitudes may also contribute to temporal and spatial mismatches -- especially the more complicated wave patterns that may be perceived in the standard rubber pencil illusion.

Supported by the Swarthmore College Research Fund and the Howard Hughes Medical Institute.

Stereoscopic 3-D Shape Constancy with real objects in the real world: Effects of Contextual Disparity Contrast

((R.A.Jain, L.Tsina, and F.H.Durgin)) Swarthmore College.
(1997) Investigative Ophthalmology & Visual Science, 38, S904.

Purpose: Durgin et al. (1995) reported surprisingly good stereoscopic depth constancy using real objects in a well-structured environment. We conducted two similar studies to examine the influence of contextual disparity contrast in the oft-reported failures of constancy. In one experiment, disparity contrast was introduced by varying viewing distance. In another, disparity contrast was introduced by varying the absolute size (another ostensibly-irrelevant variable for shape judgments) of otherwise identically shaped objects.

Methods: Objects were wooden cones with height-to-base-radius ratios of 1, 2, 3, and 4. Shapes were viewed binocularly, tip-on, while a chin and forehead rest immobilized the head. Observers adjusted a triangular icon on a simultaneously viewable computer screen to indicate the perceived aspect ratio of each cone. In Experiment 1, all cones had base radii of 2.5 cm, and were viewed, by 36 undergraduates, at distances of 60, 100, and 140 cm. Trials were blocked by viewing distance and order varied to isolate contextual order effects. In Experiment 2, 36 undergraduates made judgments of the same shapes, but the within-subject manipulation involved blocked sets of judgments about cone sets with different base radii. (5 cm and 2.5 cm). Viewing distance (46, 92, or 138 cm) was varied between subjects.

Results: In both experiments, impressive shape constancy was found with between-subject comparisons, both of viewing distance (Exp. 1 and 2) and of base radius (absolute size -- Exp. 2). However, failures of shape constancy consistent with effects of contextual contrast (of absolute disparity) were found for both changes of viewing distance (Exp. 1) and changes in absolute size (Exp. 2).

Conclusions: Failures of stereoscopic shape constancy might not always reflect inherent failures to take viewing distance into account. In enriched viewing environments, which seem to minimize such failures, contextual contrast effects arise when the same-shaped objects are judged successively at different distances or simply at different sizes.

Supported by the Swarthmore College Research Fund and the Howard Hughes Medical Institute.

Comparisons of numerosity perception across stimuli type

((D.A.Lewis and F.H.Durgin)) Swarthmore College.
Investigative Ophthalmology & Visual Science, 38, S638.

Purpose: Regular dot patterns appear more numerous than do random ones. One model to account for this proposes that each dot "occupies" a circular area of space and that the combined occupied area is diminished when dots are clustered because their areas overlap (Allik & Tuulmets, 1991). This might occur because visual images are undersampled (blurred) or analyzed at large spatial scales. The present experiments were designed to test the blurring interpretation of the occupancy model by using dots filtered to minimize low spatial frequencies.

Method: Ten subjects made forced choice numerosity comparisons of simultaneously presented texture fields which varied in number and in the minimum spacing of their elements (larger minimum spacings produce more regular textures). A staircase method was used to assess points of subjective equality between random and regular texture numerosity perceptions. Numerosities of 40, 80 and 160 dots were used in the random field. Measurements were made with white dots against a gray backgrounds as well as balanced dots (Carlson et al, 1984), which minimize low spatial-frequency information. Lower contrast gray dots were also used to assess affects of simple visibility. If the blurring hypothesis were correct, then balanced dots ought not to demonstrate the classic regularity/numerosity effect.

Results: Although the effects of regularity varied as a function of dot type, F(2,8) = 3.72, p <.05, the direction of these effects was not that predicted by the blurring explanation of the occupancy model. Contrary to the blurring hypothesis, regularity effects were stronger for balanced dots than for gray dots, and certainly no weaker than for white dots. Lower contrast (gray) dots showed somewhat reduced effects. Expressed as the proportion of dots in the regular compared to the random arrays at the PSE, gray dots had an average distortion factor of 0.97, white dots 0.92, and balanced dots, 0.87.

Conclusions: The simple blurring hypothesis of the occupancy model was not supported by this data. To the extent that area-based models (e.g., CODE -- Van Oeffelen & Vos, 1983) are correct, they must be assumed to operate on contrast information, rather than luminance energy.

Swarthmore College Research Fund.

Adaptation to running depends on runner's frame of reference

((A.Pelah1, F.H.Durgin2, C.M.Miller2, T.A.Washington2, and M.Nelson2)) University of Cambridge1, Swarthmore College2.
Investigative Ophthalmology & Visual Science, 38, S1007.

Purpose: Running on an exercise treadmill strongly adapts peripheral motor pathways, leading to a compelling aftereffect (Anstis, 1995): after treadmill running subjects will show a marked inadvertent forward advance when asked to run blindfolded in place (RBIP). We investigated further the conditions that lead to this motor adaptation. As there are only small RBIP distances after natural running, it cannot depend only on the relative velocity between runner and surface. We reasoned therefore that expression of the aftereffect must depend either on 1) oddities specific to the movement of a treadmill belt, or 2) the runner's requirement to remain anchored to an external reference frame (i.e. the treadmill) whilst still advancing relative to the surface beneath him.

Methods: Ten subjects ran for 6 min along a quiet campus road while blindfolded and holding onto a bar attached to the back of a golf-cart driven at 10 km/h, and RBIP distances were measured before and after the run for 20 s. The golf cart task mimics treadmill running in that it also constitutes an external reference frame to which the runner remains anchored, while at the same time maintains the relative velocity of the runner and surface. However, it differs from a treadmill in that running is done on a normal surface. Thus, factor 1 alone would be implicated if there was no additional RBIP distance compared with natural running, whereas equivalent RBIP distances to treadmill running would support instead the importance of factor 2. To additionally consider subjects' use of localization cues, half of the Ss wore earplugs throughout the experiment. For comparison, RBIP distances were also measured before and after treadmill running indoors.

Results: All golf-cart subjects consistently demonstrated an approximate doubling of RBIP distance after the run. Earplugs and indoor treadmill Ss generally advanced longer distances but maintained the same factor change.

Conclusions: Expression of the motor aftereffect that can occur after treadmill running does not require a treadmill, but requires both the relative movement of running on a surface, and a reference frame (e.g. a stationary treadmill or moving golf cart) to which the runner is anchored. Hearing is useful in helping to maintain a constant spatial position during blindfolded running in place.

Oblique effects in visual search

((S.E.Wolfe and F.H.Durgin)) Swarthmore College.
(1997) Investigative Ophthalmology & Visual Science, 38, S367.

Purpose: Apparent global precedence in visual search (Saarinen, 1994; see Figure 1) is reversed to local precedence when stimulus orientations are rotated 45° (Durgin & S. Wolfe, 1997). Because these results can be explained by either the oblique effect or by the vertical symmetry between targets and distractors, we rotated these search stimuli by 22.5° to eliminate possible effects of symmetry (J. Wolfe & Friedman-Hill, 1992).

Methods: Targets and distractors were like those in Figure 1, but rotated 22.5°. Targets, defined by local or global orientation (blocked), were presented among 9 or 17 distractors (blocked). Sixteen subjects made 960 judgments each (50% target present). Median RTs were computed and submitted to a 2 x 2 (Target Scale x Distractor Number) repeated measures ANOVA. RT's were also compared to those found by Durgin and S. Wolfe (1997).

Results: RT data indicated that global search was just as (in)efficient as local search when all orientations were oblique. RT's in all conditions (M = 627 msec) showed no effect of spatial scale, F(x, yy) < 1, and were more similar to those found by Durgin and S. Wolfe for oblique targets (~600 ms) than for horizontal targets (~500 msec). Search was more efficient with 17 distractors (xx msec) than with 9 (yy msec), F(g, q) = tt.u, p < .05.

Figure1: Targets and distractors from Saarinen, 1994. Top vs. bottom differ in global orientation, left vs. right differ in local orientation. Search speed is predicted by absolute orientation (Durgin & S. Wolfe, 1997).

Conclusions: Delayed search for oblique targets was found for stimuli in which vertical symmetries had been removed. The nature of the oblique effect reported by Durgin and S. Wolfe cannot be explained by vertical symmetry between targets and distractors. It remains possible that mutually symmetric distractors formed a homogenous background, and that using non-symmetric oblique orientations produces a less homogenous background for search.

Supported by the Swarthmore College Research Fund and the Howard Hughes Medical Institute.

Created by Richa Jain, rjain1@swarthmore.edu