I find the conclusion unlikely to have the effect size they describe and there are two similar explanations that fit the data better. Take the name "Maria" in

I find the conclusion unlikely to have the effect size they describe and there are two similar explanations that fit the data better. Take the name "Maria" in the US. In the 1930's a girl with that name would have strong odds of being from a Catholic background, probably from a large family, maybe Hispanic/Italian/Porto Rican. Then the film the "Sound of Music" came out, and the number girls of being given the name Maria tripled in the 1960's and 1970's. And these new Maria's were born to parents who were trying to evoke a white, blond, energetic, musical Austrian image. The "stereotypes" associated names change each generation. One big of tell of this, from the linked study, is that adults don't revert to chance on children's names, but are in fact actively worse than chance. In fact, adults are more "worse than chance" on matching children than adults are "better than chance" a matching adult faces. It's not that adults aren't getting data from looking at children's faces, it's that they are mapping what they are getting to an incorrect mapping!