Abstract
The grouping of image parts into a perceptual whole can help to deduce an objects' shape when it is only partially visible. Here we studied the temporal characteristics of the process of grouping parts into wholes by measuring eye movements during visual search. Visual search for an occlusion-defined surface among non-surface stimuli can proceed efficiently across the visual field, that is search times show little dependence on the number of items (Davis & Driver, 1994). However, search for a target surface among other surfaces is inefficient (Gurnsey, Poirier & Gascon, 1996). We asked whether the grouping of parts into wholes might speed visual search compared with a condition that contained identical local parts that could not be grouped into wholes.
We measured reaction times and eye movements while participants (n=12) were searching for the odd-one-out target (singleton) among a field of uniform distracters (4, 8, 12 items). Search items were Varin inducers of illusory contours that were either properly aligned (whole condition) or outward pointing (part condition). The target-defining feature was a difference in the inducers' openings (+/− 10° from 90°) which corresponded to a difference in the perceived shape enclosed by aligned inducers.
Search was inefficient for both, a whole among wholes, or a part among parts, but the time required to find the singleton increased at only half the rate in the former condition. However, the dwell times per item - as estimated by the slopes of linear fits to individual reaction times - slightly overestimated the difference in fixation duration between the whole and part condition (∼140 vs. 260ms). Fixation durations that were calculated directly from the eyes' scan path yielded a difference of 60ms between fixations of wholes (140ms) and parts (200ms). Therefore processing of the perceptual whole did indeed need less time than processing of the individual parts.
This work was supported by a TransCoop fellowship from the Alexander von Humboldt Foundation to M. Maertens and R. Shapley.