Changes in Facial Shape throughout Pregnancy—A Computational Exploratory Approach

Abstract

Facial cognition serves an important role in human daily interactions. It has been suggested that facial shape can serve as a signal for underlining biological condition, and that it is correlated with, among others, health, fertility, and attractiveness. In this study, 14 women were photographed during three consecutive trimesters of pregnancy, and the levels of their facial sexual dimorphism, asymmetry, and averageness were computed. Facial sexual dimorphism in first trimester was higher than in the second trimester (F(2, 22) = 5.77; p = 0.01; ηp2 = 0.34, post-hoc Tukey HSD test p = 0.007). Similar pattern was visible for asymmetry (F(2, 22) = 3.67; p = 0.04; ηp2 = 0.25, post-hoc Tukey HSD test p = 0.05). No statistically significant changes in measurement of averageness were observed. Results from Bayesian complementary analyses confirmed the observed effects for sexual dimorphism. The evidence for trimester differences in asymmetry and averageness was inconsequential. Based on the preliminary results of this exploratory study, we suggest that previously found decrease in observed facial attractiveness during pregnancy can be related to the decrease in computed facial femininity (possibly mediated by the changes in facial adiposity).

A replication that requires replication: Commentary on Caton et al. (2022)

“…Unfortunately, there are some instances of recent questionable usage of these GM methods by some researchers in evolutionary and social psychology (e.g., Antar & Stephen, 2021;Marcinkowska et al, 2021;Marcinkowska & Holzleitner, 2020). These studies tend to use landmark configurations originally proposed for stimuli manipulation and generation of 'morphs' in software such as PsychoMorph, i.e., they rely on tools not primarily designed for biometric purposes.…”

“…Another problematic practice that appears in some recent works (Antar & Stephen, 2021;Marcinkowska et al, 2021;Marcinkowska & Holzleitner, 2020) is that they do not take into account the distinction between landmarks and semilandmarks. Landmarks and semilandmarks are treated as 14 having the same morphological reliability and semilandmarks are then not allowed to slide along the curves in the generalised Procrustes analysis.…”