Abstract
We all know that finding a target is easy if the distractors are dissimilar. However, why are target-absent searches systematic? More specifically, what property of the distractor makes some target-absent searches easy? Surprisingly there have not been any attempts to answer this question. Here we set out to investigate this question using a combination of behavioral and fMRI experiments in humans. In Experiment 1, subjects had to view a search array and indicate whether any oddball target is present or absent. Target absent times were highly systematic, as evidenced by a strong split-half correlation across subjects (r = 0.75, p < 0.0005). We hypothesized that the target-absent search time might depend on how distinctive an object is compared to other objects. We measured the pairwise dissimilarity between all pairs of images in Experiment 2. For each object, we measured its distinctiveness as its average distance from all other objects in the experiment. This quantity, derived from target-present search times, was strongly predictive of the target-absent search time (r = -0.77, p < 0.0005). In Experiment 3, we measured brain activations using fMRI while subjects performed a visual search task to investigate the neural correlates of distinctiveness. We hypothesized that if a brain region is computing distinctiveness of an object then its activations during target-absent search should predict the distractor's distinctiveness. On each trial, subjects had to indicate whether an oddball target was present or not using a key press. Consistent with previous work, neural representations in the lateral occipital (LO) region matched best with visual search behavior (r = 0.67, p<0.00005). Importantly, neural activations in a region anterior to LO predicted distinctiveness during target-absent searches (r = -0.53, p < 0.005). Taken together our results show that distinctiveness computations in higher visual areas predict systematic variations in target-absent search times.