Fig 1.
Distribution of all 1021 landmarks and semi-landmarks on the facial surface.
Semi-landmarks were initially positioned to depict various areas of the face but were allowed to slide along curves (curve semi-landmarks) or the surface (surface semi-landmarks) during TPS (Thin Plate Spline) transformation. In order to avoid clutter, surface semi-landmarks are only depicted on one side of the face.
Fig 2.
3D PCA graph displaying facial variation according to sex (in SD units), in the entire sample, as explained by PC1 (20.9%), PC2 (15.1%) and PC3 (9%).
The corresponding 3D facial morphings represent the average female (red) and male (blue) facial shapes. A best-fit superimposition (upper right) reveals the surface differences in facial shape between males and females and the color-map (lower right) displays the magnitude of those differences, as distance of the female from the male face (positive: forward).
Fig 3.
a. PCA graph displaying facial shape variation in females (in SD units), as explained by PC1 (21%) and PC2 (15.1%). The corresponding 3D facial morphings show the shape extremes from -3 to +3 standard deviations of PC scores within each axis. b. Best fit superimpositions of shape transformations created from the Procrustes coordinates corresponding to -3SD and +3SD of PC scores within each axis from the average shape configuration in the female population. The superimpositions display the direction of shape variation explained by each PC.
Fig 4.
a. PCA graph displaying facial shape variation in females (in SD units), as explained by PC3 (9.1%) and PC4 (7.2%). The corresponding 3D facial morphings show the shape extremes from -3 to +3 standard deviations of PC scores within each axis. b. Best fit superimpositions of shape transformations created from the Procrustes coordinates corresponding to -3SD and +3SD of PC scores within each axis from the average shape configuration in the female population. The superimpositions display the direction of shape variation explained by each PC.
Fig 5.
a. PCA graph displaying facial shape variation in males (in SD units), as explained by PC1 (20%) and PC2 (16.8%). The corresponding 3D facial morphings show the shape extremes from -3 to +3 standard deviations of PC scores within each axis. b. Best fit superimpositions of shape transformations created from the Procrustes coordinates corresponding to -3SD and +3SD of PC scores within each axis from the average shape configuration in the male population. The superimpositions display the direction of shape variation explained by each PC.
Fig 6.
a. PCA graph displaying facial shape variation in males (in SD units), as explained by PC3 (8.1%) and PC4 (6.3%). The corresponding 3D facial morphings show the shape extremes from -3 to +3 standard deviations of PC scores within each axis. b. Best fit superimpositions of shape transformations created from the Procrustes coordinates corresponding to -3SD and +3SD of PC scores within each axis from the average shape configuration in the male population. The superimpositions display the direction of shape variation explained by each PC.
Table 1.
Comparisons (independent t-tests) of self-perceived facial attractiveness scores within the entire sample and within subgroups.
Fig 7.
Best fit superimpositions showing the surface differences between the least and most attractive variation of a female (upper row, left) and male (upper row, right) face.
The magnitude of difference between the two surface images can be visualized on the color maps (lower row), where distances between the most attractive and the least attractive face are shown (positive: forward).
Fig 8.
Best fit superimposition showing the surface differences between the least and most attractive variation of a white female (upper row) face.
The magnitude of difference between the two surface images can be visualized on the color maps (lower row), where distances between the most attractive and the least attractive face are shown (positive: forward).