In the popular debate, the rising importance of masculinity for young men is often cited as a reason for the growing political polarization based on gender. Yet, recent scholarly research on femininity and masculinity (F&M) has generated inconsistent findings, for instance, regarding the effect of F&M on women’s and men’s mental health, political aggression, support for radical right-wing parties, and attitudes toward gender equality. We suggest that inadequate conceptualization of F&M may be a reason for these inconsistent findings. First, most research treats F&M as unidimensional concepts. Second, most existing measures rely on top-down, normative views of gender from the 1970s and 1980s, which may not reflect how people experience or express F&M today. As Thompson and Bennett (2015) conclude, “there is value for a 3rd generation of measures to capture the changing face of [women’s] and men’s gendered lives.”
In this paper, we use survey instruments, innovations in Natural Language Processing (e.g., large language models and topic modelling), and psychological experiments to identify and measure distinct types of F&M. To differentiate subjective feelings of F&M, we use open-ended survey items implemented in a representative survey of the Norwegian population (fielded in November 2025). In particular, we ask respondents of different genders to recall situations in which they felt feminine and masculine over the last couple of months. We then use state-of-the-art large language models to parse these responses into clusters (types) of F&M.
To assess the added value over unidimensional measures, we will also include the single-item F&M ratings from the European Social Survey. We test whether our generated types of F&M capture variation that a unidimensional measure does not. We also examine whether our types correlate more with socio-economic background conditions and are more predictive of respondents’ reported well-being and political orientation than unidimensional measures. Finally, we draw on the language-generated types of F&M to suggest an inventory of closed survey questions that can be validated and used in subsequent work to measure F&M. The paper thus furthers our understanding of how different forms of femininity and masculinity may affect key political outcomes.
