Saccades to spatially extended objects: The roles of observer goals and part-structure cues in determining within-object landing position

How is the perceptual representation of a spatially extended object translated into a discrete saccade vector? A wealth of evidence suggests that the object center of area (COA) is computed as a default index. However, saccades are also sensitive to the location of task-relevant information, which does not always appear at the COA. In addition, task-relevant object regions often correspond to discrete object parts. Here, we manipulated part-structure cues and participant goals to examine their potential interaction in controlling within-object landing position. In Experiment 1, participants executed saccades to spatially extended objects in a visual search task. Within-object landing position was systematically influenced by both factors: (1) a bias toward locations known to contain task-relevant information, and (2) a bias toward regions of part segmentation. A rapid orienting task to a single, abruptly appearing object (Experiment 2) showed only the latter bias, suggesting that unlike the strategic bias, the segmentation bias was driven primarily by physical stimulus properties. Finally, the two biases combined, such that the largest effect of strategy was observed when the relevant location coincided with a discrete object part. Together, the results indicate that saccade target selection was controlled by the combined effect of goal-directed processes, segmentation, and a default bias toward the COA.