D'Zmura, and Bauer, Jolicoeur, and Cowan demonstrated that a target whose chromaticity was linearly separable from distractor chromaticities was relatively easy to detect in a search display, whereas a target that was not linearly separable from the distractor chromaticities resulted in steep search slopes. This linear separability effect suggests that efficient colour visual search is mediated by a chromatically linear mechanism. Failure of this mechanism leads to search performance strongly influenced by number of search items (set size). In their studies, linear separability was confounded with distractor heterogeneity and thus the results attributed to linear separability were also consistent with the model of visual search proposed by Duncan and Humphreys in which search performance is determined in part by distractor heterogeneity. We contrasted the predictions based on linear separability and on the Duncan and Humphreys model by varying the ratios of the quantities of the two distractors and demonstrated the potent effects of linear separability in a design that deconfounded linear separability and distractor heterogeneity.
Convex hull test of the linear separability hypothesis in visual search