Figure 7.

Group results for Experiment 3. (A) Sensitivity and accuracy calculated relative to the entire set of 18 faces. The negative slope clearly shows the integration of information into the ensemble percept. As more information (i.e., faces) became available (x-axis), subjects got closer to the mean of the 18 faces. Improvement in performance continued beyond four faces, indicating that at least four faces were integrated into the ensemble percept. (B) Subjects are significantly better at judging the average identity of a group of faces than they are at judging the identity of a single face. Notably, we replicated the result from Experiment 1 even though the group face judgment requires the integration of multiple viewpoints. Error bars represent the standard deviation of 1,000 bootstrapped samples. The shaded regions represent the 95% confidence intervals of the bootstrapped distributions.