Abstract
Real-world objects typically consist of multiple parts. When searching for a multi-part object, are all of its parts used, or just a select few? In experiment 1, participants were cued with one, three, or all four parts of a photorealistic teddy bear target, then searched for the whole target among three other images of whole teddy bears. We found that reaction times (p < .001), accuracy (p < .001), guidance (as measured by time to target fixation; p < .05), and verification time (as measured by time between target fixation and response; p < .001) all improved with the number of target parts added to the cue, suggesting that all of these parts were loaded into visual working memory and used to facilitate search. However, some parts were more useful than others. Guidance (p < .05) was best when the torso of the target bear was cued; accuracy (p < .05) and verification time (p < .01) were best when the bear's head was cued. In experiment 2, participants were cued with the whole bear, but now the search display depicted 4-part (whole bear), 3-part, or 1-part target and distractor bears. Accuracy (p < .05) and verification time (p < .05) were again best when the target's head appeared in the search display, but no benefit was found for search guidance (p > .05). We conclude that the head/face is preferentially weighted in the target representation when the target contains a head/face, in conjunction with other parts. However, when considered separately, the torsos of our targets allowed for the most effective guiding template, perhaps because they were more homogenous in texture and color. Our results suggest that search guidance increases with the number of target parts, but that all of these parts are not equally effective in guiding search.