As for why this bias exists, the authors are still unable to say, although they do think that this study, by demonstrating that the bias is already exhibited by children as young as four, “significantly constrains the range of possible explanations.”
Human faces are sexually dimorphic, and adults are able to rapidly and accurately classify the sex of faces even in the absence of typical sex-defining features such as hairstyle, clothing, facial hair, or makeup. Children show an increase in the ability to accurately classify both children and adult’s faces by sex in the absence of external cues such as hair from ages 7 to 9 years of age, although their performance remains below that of adults.
Children under the age of seven can distinguish male from female faces, but they are less good at it when the faces are stripped of gendered markers. This makes it even less likely that the bias towards male faces has anything to do with the faces themselves, as opposed to our socially constructed default male bias.
We found the male bias in our youngest group of children from 4 to 8 years of age, evidence that the male bias for face pareidolia emerges before the ability to determine gender from facial cues alone is fully developed. This suggests that the bias to perceive illusory faces as male is unlikely to be driven by a sophisticated ability to discriminate the distinguishing visual features of faces associated with each sex. Consistent with this, in our previous study with adults a computational analysis of the visual features in the same examples of face pareidolia we used here did not show an association with their perceived gender, and control experiments ruled out explanations based on pre-existing semantic or visual gender associations with the objects that contain the illusory face.