Abstract
Feature singleton search tasks have been characterized as being independent of the number of non-target elements in the display (Treisman and Gelade, 1980; Wolfe, 1994). Previous work from our lab has shown that reaction times on feature singleton search tasks in fact increase logarithmically with the number of non-target (lure) elements in the display and that the steepness of the logarithmic function is modulated by lure-target similarity (Madison, Buetti & Lleras, 2014). Duncan and Humphreys (1989) proposed an effect of distractor heterogeneity on search performance such that distractor-distractor similarity modulates search efficiency, and that distractor rejection is facilitated by the repetition of identical distractors (a mechanism called spreading suppression). Here we challenge those claims and demonstrate that in parallel search, we find no evidence for distractor-distractor effects, nor for spreading suppression (grouping of distractors). We tested search performance in displays containing multiple types of lures simultaneously (e.g., a red target amongst yellow, blue and orange lures). To evaluate performance on these mixed displays, we first computed estimates of logarithmic processing costs for each type of lure (obtained from a different set of subjects) when the target was presented amongst uniform fields of each type of lure (e.g., a red target amongst only orange distractors). We then used those estimates to predict what reaction times in mixed displays ought to be if (a) each item in the display was processed independently from one another and (b) the processing of all items was performed in unlimited capacity parallel fashion. In two separate experiments, with different combinations of lures, this model (a simple linear addition of logarithmic processing costs), was able to account for 95.67% and 94.41% of the variance of reaction times in mixed displays. These results argue strongly against both distractor-distractor interactions as well as spreading suppression effects in parallel search.
Meeting abstract presented at VSS 2015