Gender, language, and representation in the United States Senate

We explore how gendered language in Senate floor debates evolves between the 101st and 109th sessions (N=229,526 speeches). We hypothesize that female Senators speak like women in the general population, that their speeches focus on traditionally designated women's issues, and that they use female linguistic strategies found in the general population when discussing low politics or women's issues. We also expect women to speak like legislators, adopting more male linguistic approaches for high politics issues or in election year speeches and for female senators to use more male linguistics as time served in the Senate increases. Using a suite of computational linguistics approaches such as topic modeling (Latent Dirichlet Allocation), syntax and semantic analysis (Coh-Metrix), and sentiment analysis (LIWC), our analyses highlight the distinct roles of women speaking for women (e.g. promoting issues like education or healthcare), women speaking like women (e.g. using personal pronouns), and women speaking as Senators.