Abstract
Recent categorical search work demonstrated that searching for an exemplar among items of the same category is faster for heterogenous categories relative to homogeneous categories (Hout et al., 2017). However, precisely quantifying heterogeneity is difficult. Multidimensional Scaling (MDS) is a method used to model similarity (Hout et al, 2015), quantified as the distance between items in a hypothetical “psychological space.” The more dissimilar two items are perceived to be, the larger the distance between them in space. We used MDS distances to quantify category heterogeneity. We selected stimuli from a large MDS image database (Hout et al., in prep) which provides similarity estimates for 1200 images across 20 distinct categories. To derive our measure of category heterogeneity from the database, we averaged the distances between each exemplar within each category (e.g., average distance between all cars). Therefore, categories with larger average distances could be described as more heterogeneous in appearance. We then correlated these values with performance in a visual search task in which participants searched for a single target item among non-targets from the same category. We expect RT’s to decrease with increased heterogeneity because target-distractor similarity of any given search display should be lower given the within category nature of each display. Consistent with prior work, average RTs to target categories were negatively correlated with the average MDS distance. These results indicate that as average distance between exemplars within a category increases (i.e., category heterogeneity increases), search RTs for those exemplars decrease. We conclude that using MDS-derived similarity ratings to quantify category heterogeneity can be a useful and valid method to select stimuli for vision research.