Gender stereotypes are reflected in the distributional structure of 25 languages

Cultural stereotypes such as the idea that men are more suited for paid work and women are more suited for taking care of the home and family, may contribute to gender imbalances in science, technology, engineering and mathematics (STEM) fields, among other undesirable gender disparities. Might these stereotypes be learned from language? Here we examine whether gender stereotypes are reflected in the large-scale distributional structure of natural language semantics. We measure gender associations embedded in the statistics of 25 languages and relate these to data on an international dataset of psychological gender associations (N = 656,636). People's implicit gender associations are strongly predicted by gender associations encoded in the statistics of the language they speak. These associations are further related to the extent that languages mark gender in occupation terms (for example, 'waiter'/'waitress'). Our pattern of findings is consistent with the possibility that linguistic associations shape people's implicit judgements.