Abstract
Feature integration effects are prevalent in many visual cognition tasks that require stimulus identification. In many cases, these feature integration effects are "nuisance" effects in that they obscure the effects under study, possibly leading to invalid inferences. Eliminating these effects has proven difficult. Here, we manipulated a paradigm in which feature integration effects have been shown to be surprisingly absent in order to learn about the limits of such effects limits. In the prime-search paradigm, individuals sometimes make a response to a centrally presented prime stimulus (a colored circle). This is followed by a visual search where the target stimulus may or may not appear in a placeholder matching the prime stimulus. Typically, response are faster when the target's placeholder matches the prime than when it does not, but only when individuals respond to the prime. Across four experiments we gradually increased the overlap in stimulus location and response demands between the two tasks to determine what is necessary to cause feature integration effects to appear. We found that stimulus location overlap alone was insufficient to generate feature integration effects. Similarly insufficient was having location overlap and response overlap (i.e., both responses with the spacebar). With stimulus location overlap and a lateralized response set, however, feature integration effects appeared. This remained the case when we removed the distractor item from the second task. Interestingly, the feature integration effects were consistent with those predicted when a task switch occurs, rather than those found more commonly. This study indicates that the lack of integration effects in the prime-search task result from participants representing the two parts as separate tasks along with the lack of stimulus location and response set overlap. Furthermore, the data has implications for understanding the limits of feature integration and the interplay between task switching and feature integration.
Meeting abstract presented at VSS 2018