Fig 1.
The 27 different smiling faces.
The 27 smiles represent all possible combinations of the three spatial factors (mouth angle, smile extent, dental show) at three different levels (low, medium, high). The numbers 1–27 have been included post hoc for labelling purposes and were not present in the animations.
Fig 2.
Definitions of the spatial parameters used in the study.
Mouth angle is the angle between the green and blue lines. Smile extent is the length of the green line. Dental show is the distance between the lower and upper lips.
Fig 3.
Smile 22 with various amounts of timing (delay) asymmetry.
All animations started with the same (symmetric) neutral expression and ended with the same (symmetric) smiling expression, so the asymmetries were only visible for a few frames of the 250 ms animation.
Fig 4.
Histograms of demographic variables.
Age distributions for female (left) and male (middle) participants, as well as the alcohol consumption numbers (right).
Fig 5.
Screenshots from smile study iPad app.
The welcome screen, consent screen, and instructions screen that were shown to participants at the onset of our study.
Table 1.
Models for smile effect function ηS.
Table 2.
Information criteria for the fit models.
Table 3.
Model fit information for symmetric and asymmetric smiles.
Fig 6.
Predictions for SSANOVA model of symmetric smiles.
The top row plots the estimated main effect functions for the three covariates: age, gender, and drinking. The bottom row plots the estimated smile effect function predictions for each of the 27 smile animations depicted in Fig 1. Within each subplot, the shaded regions or bars denote 90% Bayesian confidence intervals.
Fig 7.
Visualization of the smile effect.
A heat-map plotting the three-way interaction between the smile parameters. The three vertical bars behind each face denote the predicted score for the three response variables: effective, genuine, and pleasant (respectively). Greener colors correspond to better (i.e., higher rated) smiles, and redder colors correspond to worse (i.e., lower rated) smiles.
Fig 8.
Dental show effect at different levels of angle-extent.
(a) Two smiles with smaller angle-extent combinations. (b) Two smiles with larger angle-extent combinations. Increasing dental show makes the smile worse (i.e., less successful) for (a) and better (i.e., more successful) for (b).
Fig 9.
Perceived emotions for the 27 smiling faces.
The percentage of subjects who selected the given emotion (rows) for each smile (columns). The top subplot depicts the results for all seven emotions, whereas the bottom subplot provides a more detailed look at the non-happy emotions that were perceived from each expression.
Fig 10.
Predictions for SSANOVA model of asymmetric smiles.
The top row plots the estimated main effect functions for the three covariates: age, gender, and drinking. The bottom row plots the estimated timing asymmetry effect function. Within each subplot, the shaded regions or bars denote 90% Bayesian confidence intervals.