The Face of Emotion

Published: 08-10-2020| Version 1 | DOI: 10.17632/pdhrg3kkxr.1
Hosina alk,
Annukka Lindell


This research compared the two models of emotional lateralisation: the right hemisphere hypothesis and the valence hypothesis to ascertain whether the left cheek bias presents consistently for emotions of different valences. By using the left cheek bias paradigm we were able to determine which model was most supported. We assessed people’s perceptions of the five core emotions (happiness, sadness, anger, surprise, fear) expressed by left cheek and right cheek poses and aimed to determine whether perceptions of the emotions were influenced by the cheek shown (left or right cheek poses). It was hypothesised that if the valence hypothesis is correct, there would be higher ratings of positive emotions for right cheek poses and higher ratings of negative emotions for left cheek poses. If the right hemisphere hypothesis is correct, all ratings would be higher for left cheek poses across all emotional valences. We recruited 135 participants to complete an online survey depicting left and right cheek images. We found that the left cheek bias presents consistently for positive and negative emotions and thus, offered support for the right hemisphere hypothesis rather than the valence hypothesis.


Steps to reproduce

80 images of female and male models expressing the five core emotions using left and right cheek poses were collected from Pinterest. Images were edited to have a grey scale filter, a white background, and the dimensions of 7cm x 7cm. There were 16 images for each emotion; half were left cheek poses, half were right cheek poses; half were female models, half were male models; half were in their original orientation, and half were mirror-reversed to control for perceptual biases. Each image was accompanied by a single-item Likert scale stating 'this person appears (emotion)'. The survey was conducted by Qualtrics which automatically randomised and counterbalanced the order of the trials for each participant. Variables in the spreadsheet were coded as the following: Participant Gender (Categorical): 1 = Male 2 = Female Model Gender (Categorical): 1 = Male 2 = Female Emotion (Categorical): 1 = Happiness 2 = Anger 3 = Fear 4 = Sadness 5 = Surprise Orientation (Categorical): 1 = Original 2 = Reversed Cheek (Categorical): 1 = Left 2 = Right Emotional Valence (Categorical): 1 = Positive 2 = Negative Perception of Emotion (Categorical): 1 = Not at all 2 = A little 3 = Somewhat 4 = Moderately 5 = Very 6 = Extremely Two participants did not include their ages and one participant did not include their gender. Results were analysed using Generalised Estimating Equations ordinal logistic regression with perception of emotion as our dependent variable and cheek pose (left, right), emotional valence (positive, negative), gender of the model (female, male) and the gender of participants (female, male) as our independent variable.