PMC

(a) Original drawings (top row) were altered so that either the least distinctive part was swapped with the least distinctive part of a different category (second row) or the most distinctive part was swapped with the most distinctive part of another category (third row). (b) Comparison of the percentage of correct categorizations for the original shapes (from Experiment 2), swapped indistinctive parts and swapped distinctive parts (error bars indicate standard errors). (c) Bar plots showing how often the category of the swapped-in part determined the categorization choice. The indistinctive (left) and distinctive (right) parts are shown in red on the shapes’ silhouette. Bars significantly different from chance (p < 0.01) are marked with an asterisk.

(a) Individual responses of all participants for an example Variation, showing which parts were marked as most distinctive (red), second most distinctive (orange), and third most distinctive (yellow). Aggregating these responses results in the shape on the right. The redder each point in the contour, the more distinctive it was evaluated across all participants. (b) Comparing aggregated responses between Exemplars and corresponding Variations (a subset shown here) suggests that in most categories similar parts (e.g., the indentation, spike or twirl) were identified as distinctive across most Variations.

All possible combinations of common and yeast and human-specific components in a model with five components are tried as a target for the DISCO rotation. The deviation of the DISCO-SCA solution from this target is displayed against the number of distinctive components. The targets are labelled by the numbers between brackets (the first number indicates the number of distinctive components for GC, the second for LC). The line connects solutions with the lowest deviation for a fixed number of distinctive components.

All possible combinations of common and yeast and human-specific components in a model with five components are tried as a target for the DISCO rotation. The deviation of the DISCO-SCA solution from this target is displayed against the number of distinctive components. The targets are labelled by the numbers between brackets (the first number indicates the number of distinctive components for yeast, the second for human). The line connects solutions with the lowest deviation for a fixed number of distinctive components.