Figure 5

(A) Average expression intensity (top) and genuineness (bottom) scores as a function of the three databases (WFD, ED, and US). Error bars are 95% CI. Lines with asterisks indicate significant pairwise differences (p < 0.001). (B) Average intensity scores were highly correlated with genuineness scores for images in the WFD (blue markers), ED (red markers), and US (green markers) databases. The corresponding solid lines are the result of the best fitting linear regression of the form y = mx + b. (C) The relationship between genuineness and intensity for the WFD redrawn to examine the influence of modal categorization. Color reflects the emotion most often used by human participants to categorize the image. The solid lines are the result of the best fitting linear regression, as in (B), and their length marks the data range. Thus, any clustering of colored dots would indicate that particular emotions are perceived as more genuine and/or intense than others. However, this is not the case, all colors are dispersed, and we observe positive relationships between genuineness and intensity for every emotional category. That said, this plot does not take into account categorical spread and specificity score.