Abstract
Wolfe and Myers (2010) tested whether materials guide our search in visual scenes. They used the stimuli from the Flickr Material Database (FMD) and found no evidence that our search is efficiently guided by materials. For example, searching a fur patch among stone patches turn out to be highly inefficient. FMD was developed to capture a range of real-world materials, in which the surface appearance per material class may vary largely. Here we present results from a standard visual search experiment using the images of canonical material modes (velvety and specular) as stimuli. Participants searched for a velvety image among specular images (distractors) vice versa. There were four set sizes, 4, 9, 16, and 25. Images of canonical materials provide key image features that trigger corresponding material perception, namely the bright contours for velvety and the highlights for specular material (e.g., Zhang et al., 2019). To ensure that participants only use these material-related perceptual features instead of sole perceptual features, the lighting direction was randomly varied throughout the experiment. In addition, the 3D-shape of the materials was either a sphere or a blob. Overall, there was a significant set size effect with a search slope of around 3ms/item in target-present trials and 4 ms/item in target-absent trials, indicating an efficient search. However, the efficiency significantly varied with the type of material. The slopes of velvety target were 4.5ms/item and 9ms/item in target-present and target-absent trials, respectively, while that of specular target was flat for both conditions. Hence, specular clearly constitutes a basic feature in the sense of Treisman’s FIT. For velvety, evidence is less strong. In other words, our study presents a first evidence that material perception may extract basic features, i.e., contradicting Wolfe and Myers’ (2010) study. To strengthen our claim further studies will test more canonical materials.