Abstract
It has long been established that the difficulty of determining whether a target is present or not among a set of distracters depends on the heterogeneity of said distracters. If they are homogeneous then the target will be detected effortlessly via peripheral vision; if they are heterogeneous then a series of saccades will be required. Previous research has uncovered large, stable individual differences in how observers search arrays of line segments when the distracters have low variance on one side, and high variance on the other [Nowakowska, Clarke & Hunt, 2017]. While a small number of observers follow the optimal strategy (saccade to high variance side on target absent trials), many others perform at chance, or even ‘counter-optimal’. Here, we present a new series of experiments (N = 60 in total) in which we replace the line segments with different classes of items (icons, mosaics, polygons). An Eyelink 1000 was used, and each experiment involved around 200 trials, half of which contained the target. The suitability of the each class of stimuli was validated in a pilot study to confirm that the target, when present, is clearly visible when on the homogeneous half of the display, and only visible after careful inspection on the heterogeneous side. The results show that search behaviour can change dramatically when we vary surface-level details of the task. When observers are faced with an array of computer icons (i.e., applications and folders), they all search optimally, ignoring the homogeneous side and directing their attention to the heterogeneous side of the display. Oriented line segments are visual primitives, assumed to generalize to more complex objects. Yet they produce highly variable and sub-optimal search, while complex objects are searched with near-perfect efficiency.