Dissociating preattentive vision and preattentive attentional guidance

Abstract

In Feature Integration Theory (FIT; Treisman & Sato, 1990), efficient visual search performance can either be driven by preattentive vision, or by focused attention that is effectively guided by preattentive information. Whereas independent dimensional modules handle basic feature detection, a salience map of locations serves to guide attention in other searches. Recent theoretical models have largely abandoned this distinction; for example, the guided search theory (Wolfe, 1994) assumes that all searches are guided by a salience map. Based on an assumption that preattentive vision does not signal location information, the present study investigated and provided evidence for a distinction between searches that rely on preattentive vision and searches that rely on attentional guidance. Participants detected features or judged target locations (left or right side of the display); because location information is needed for the location judgement task, the salience map should be implicated in responses. Results showed a dissociation between tasks in terms of both dimension-switching costs and cross-dimension attentional capture, in which dimension-switching occurred only in the target detection task and attentional capture occurred only in the location judgement task. These results reflect the use of two different mechanisms, one dimension-specific and the other dimension-general, which map onto existing proposals for dimensional modules and the salience map. In a feature discrimination task, results precluded an explanation based on response mode (detection versus discrimination). We conclude that the FIT architecture should be adopted to explain the current results, and that a variety of visual attention findings can be addressed within this framework.