The redundant-signals paradigm and preattentive visual processing

Physiological and cognitive models of vision agree that the pre-attentive processing of visual stimuli is organized in a parallel and segregated fashion. However, several incompatible models have been proposed for the subsequent processing stages. They differ in their assumptions about architecture (serial, parallel, or coactive/integrative), stopping-rules (self-terminating, or exhaustive), spatial specificity of saliency signal coding (signal pooling across locations, or spatially distinct processing), and dependency of target detection on the prior allocation of attention (pre-attentive, or post-selective). We review how studies employing the redundant-signals paradigm in visual pop-out search contribute to discerning between the different assumptions. We find strong support for the notion of a saliency map, into which feature contrast signals are pooled, and especially the dimension weighting account (1) receives further support: Instead of a priming mechanism that could increase weights for several dimensions independently, evidence favors a weighting mechanism that effectively limits the total weight available for allocation to the various dimensions through competitive interactions, whereby increasing the weight for one dimension goes along with decreasing the weights for other dimensions.