Figure 8: Results of Experiment 6.

(a) Crowd lifelikeness of large groups is well predicted by the average of individual item ratings from independent observers. This graph depicts the ratings (black circles) of one representative participant judging 102 unique temporal groups. There is a high Pearson correlation, r=0.910 between the ratings of average crowd lifelikeness (y axis) and predicted ratings of the groups generated from single-item ratings (x axis). (b) Averaged data for five observers. The x axis represents the number of items displayed in the set. The y axis represents the magnitude of the Fisher z score. The Fisher z score in the whole set condition indicates that participants are able to perceive ensemble lifelikeness even in large groups of items (in this case 12 items). The Fisher z scores increase as the number of items displayed increases. This pattern rules out the possibility that participants engaged in random subsampling techniques to accomplish the task. Error bars represent s.e.m. Finally, the dashed line represents the memory capacity limit in this experiment. Participants remembered on average five items in each display. However, their ensemble coding performance continued to significantly increase between 8 and 12 items, indicating that ensemble coding performance is not solely dependent on explicitly remembered items.