Preventing Local Clustering in Random-Dot Patterns: A New Method to Avoid Perceptual Artifacts

Patterns consisting of random dots are frequently used in visual science. One disadvantage of using random-dot patterns is the possible clustering of dots. These clusters have a lower spatial frequency than the one derived from dot size. The cluster formation is not present in patterns in which dots are homogeneously distributed, but with these patterns other problems, eg ambiguity in stereograms, occur. A good balance between homogeneous and random distribution has to be found. This problem has often been addressed in half-toning techniques. One of these techniques is force-field random dithering. In this technique each dot has a force field that repels newly placed dots. For generating random patterns, we used a modification of this dithering technique.

This technique is empirically compared with the traditional random-dot patterns. Subjects viewed for 70 ms a 6 deg × 6 deg square consisting of 100 × 100 dots placed randomly or with a force field. Each dot subtended 0.6 min arc. The task was to detect a vertical band of 30 × 100 dots with increased luminance, which could appear either left or right of the midline. The density of the dots was varied between 5% and 25%. The results indicate a significantly larger error rate when using the force-field generated pattern.

We conclude that subjects are using clusters as local cues. These results should warn investigators using random-dot patterns that local clusters could act as serious artifacts.