Abstract
This study seeks to discover which spatial relations among image elements (e.g., line segments or edges) tend to induce grouping, that is, cause image elements to bind together into perceptual “objects.” Many candidates have been proposed: non-accidental properties such as collinearity, cotermination and parallelism; relatability; more global factors such as symmetry and skew symmetry; and many others. Empirical investigation of these factors has, however, been hampered by the lack of “objective” methods for determining the perceived organization of an image. The current research uses a new method for studying perceptual grouping, based on the finding that perceptual comparisons between two spatially separated probes are faster if the two probes lie on the same perceptual object (as in “object-based attention”). We use this effect to determine the degree to which two image elements are, in fact, “bound” or grouped together by the visual system. We conducted a large omnibus experiment testing the grouping strength of pairs of line segments in many spatial configurations, testing a wide variety of potential grouping factors, including collinearity, relatability, cotermination, parallelism, symmetry, skew symmetry, contact, geometric codimension, and others. We also varied the time (SOA) the configuration was presented prior to the appearance of the probes, allowing us to determine the time-course of the influence of each potential grouping factor. These data paint a detailed picture of the chronology of object formation, as the grouping percept gradually develops over the first few hundred milliseconds of processing. Generally, the data reveal that grouping begins with more local spatial relations (e.g. collinearity and relatability) and gradually spreads to more global or higher-order relations e.g. symmetry). Generally, grouping seems to be induced by the degree of “regularity” exhibited by the configuration,rather than by any one essential property.