Figure 5.

Global and local shape integration in hierarchical stimuli. (A) We investigated how global and local shape combine in visual search using the multiscale part sum model. According to the model, the dissimilarity between two hierarchical stimuli can be explained as a weighted sum of shape differences at the global level, local level, and cross-scale differences across and within objects (see text). (B) Observed dissimilarity plotted against predicted dissimilarity for all 1176 object pairs in the experiment. (C) Local and cross-scale model terms plotted against global terms. Colored lines indicate the corresponding best fitting line. Asterisks indicate statistical significance: *** is p < 0.0005, **** is p < 0.00005. (D) Visualization of global shape relations recovered by the multiscale model, as obtained using multidimensional scaling analysis.