Changing features do not guide attention in change detection: evidence from a spatial cuing paradigm.

Smilek, Eastwood, and Merikle (2000) demonstrated that the detection of change was facilitated when the target character changed in many rather than few of its features. Specifically, the function relating search response time to display set size was shallower when more features changed than when fewer features changed. The researchers interpreted these results as indicating that large feature changes provide preattentive guidance of focal attention to the location of the change. We tested this preattentive guidance hypothesis by examining change detection performance in the context of a spatial cuing paradigm. The hypothesis predicts that (1) the cost on invalidly cued trials should be less when more features change than when fewer features change, and (2) the features manipulation should have no effect on validly cued trials. In contrast to these predictions, our results show that cuing effects are equivalent across all levels of feature change and that a robust effect of the features manipulation is observable for both validly and invalidly cued trials. We argue that large feature changes do not provide preattentive guidance and in fact can be detected more readily after attention is already in place at the target location.