Visual search efficiency is modulated by symmetry type and texture regularity

Abstract

More than a century of vision research has identified symmetry as a fundamental cue, which aids the visual system in making inferences about objects and surfaces in natural scenes. Most studies have focused on one type of symmetry, reflection, presented at a single image location. However, the visual system responds strongly to other types of symmetries and to symmetries that are repeated across the image plane to form textures. Here we use a visual search paradigm with arrays of repeating unit cells that contained either reflection or rotation symmetries but were otherwise matched. Participants were asked to report the presence of a target tile without symmetry. When unit cells tile the plane without gaps, they form regular textures. We manipulated texture regularity by introducing jittered gaps between unit cells. This paradigm lets us investigate the effect of symmetry type and texture regularity on visual search efficiency. Based on previous findings suggesting an advantage for reflection in visual processing, we hypothesized that search would be more efficient for reflection than rotation. We further hypothesized that regular textures would be processed more efficiently. We found independent effects of symmetry type and regularity on search efficiency that confirmed both hypotheses: Visual search was more efficient for textures with reflection symmetry and more efficient for regular textures. This provides additional support for the perceptual advantage of reflection in the context of visual search and provides important new evidence in favor of visual mechanisms specialized for processing symmetries in regular textures.

Figure 1.

The 17 wallpapers rendered with a comma-like symbol as the repeating element. Illustration based on Wade (1993).

Figure 2.

Examples of 4 × 4 “no jitter” search arrays based on wallpaper group P4 (A) and PMM (B). An example of a PMM in the “jitter” condition is shown in C. Across all stimuli the individual unit cells were the same size. In the jitter stimuli, the overall array was larger, but the unit cell size was the same. For all jitter conditions, the unit cells were presented on a 50% gray background, large enough to contain all possible unit cell positions, to avoid discontinuities along the search array edge.

Figure 3.

An example of a fundamental region and corresponding PMM and P4 unit cells, with the fundamental region highlighted with a black dotted line. Here, you can see that the random dot pattern from the fundamental region is repeated four times to create a unit cell but using different transformations. This allows for consistency between the two symmetry types regarding the amount of white, black, and shades of gray in each stimulus.

Figure 4.

Example images from Experiment 1; PMM target absent versus present example. The “random unit cell” target tile is in the second row from the top, the third column from the left.

Figure 5.

(A). Reaction time data across the four experiments. Error bars reflect the standard error of the mean. (B) Slopes of the corresponding visual search function, averaged across participants, for each of the four experiments. Error bars reflect the standard error of the mean. It is evident that slope values are smaller (more parallel) for PMM than for P4, and for non-jittered compared to jittered conditions.

Figure 6.

d′ plotted in the same way as reaction time, with d′ data across the four experiments in (A), and slopes of the corresponding visual search function in (B). Error bars reflect the standard error of the mean. Slopes across array sizes are relatively flat and similar across conditions. The only exception is un-jittered PMM, where the slope is flatter than the others. This implies that this condition was the easiest overall, which is consistent with the reaction time results.