Figure 1

Box and whiskers plots of accuracy for each submission. The line within each box signifies the median, the bottom and top hinges of the box correspond to the first and third quartiles, the bottom and top lines extending from the box correspond to 1.5 * IQR (inter-quartile range), and the points to outliers beyond this range. Top plot shows accuracy in within-category testing. The two dashed horizontal lines indicate chance accuracy, with color matched to testing round. In within-category testing, chance was 8.3% for the 12 nouns, 12.5% for the 8 actions, and 10% for the 10 properties. Chance for between-category testing was 10% for the 10 alternatives.