Experiment 1 used a modified spatial cueing paradigm that was introduced by Egly et al. [J. Exp. Psychol. Gen. 123 (1994) 161] to investigate the cost incurred in shifting attention within an object as opposed to shifting attention between objects. Subjects were presented with two outline rectangles and had to detect a target (a luminance increment) that could appear in the cued location (valid trials), in an uncued location inside the cued rectangle (inside-invalid trials), or in an uncued location inside the uncued rectangle (outside-invalid trials). Valid trials were faster than invalid trials, and inside-invalid trials were faster than outside-invalid trials. In Experiment 2, the two rectangles were joined to form a unitary object. Here, no difference was found between outside-invalid trials and inside-invalid trials. Experiment 3 showed that the delayed response on outside-invalid trials in Experiment 1 was not due to attention needing to cross the figural borders in order to re-orient to the uncued rectangle. The results were interpreted as showing that an extra cost is incurred for shifting attention between different objects.