9312224 Braun In prior work, we have studied concurrent visual task situations in which the observer directs visual attention at one part of a visual scene, thus leaving other parts unattended, but attempts to judge attributes of both the attended and unattended parts of the scene. These experiments have demonstrated that a significant amount of information, especially about salient objects, can be discriminated in unattended regions of the field of view. Here, we formulate a hypothesis which could explain these observations, and propose to test it psychophysically. This hypothesis is based on our earlier proposal that a "saliency map" guides visual attention. Our new hypothesis it states that the saliency map provides a rudimentary but functional alternative to visual attention, and that it can sometimes mediate nonattentive stimulus selection, nonattentive figure-ground segregation, and nonattentive access to awareness. To test this hypothesis, we propose to conduct experiments on unattended but salient objects, investigating feature discrimination, as well as feature conjunctions and other "binding" phenomena. Ongoing research in our laboratory, supported by the NSF grant "An integrated, neurally based model of selective, visual attention," (BIR 92-14238), aims to build a model of visual attention and the saliency map at the systems level, on the basis of neurally plausible elements. The psychophysical experiments proposed here will help to guide this effort and will help decide what perceptual phenomena (grouping, feature conjoining, apparent motion, tracking) should be our benchmarks in evaluating the model.***
