Gender variation in writing : analyzing online dating ads

This dissertation presents a study of gendered language variation and linguistic indexicality in computer-mediated communication. A two-pronged approach combining the analysis of language production in a corpus of 103,000 English-language online dating ads with a language perception study (891 participants) is taken towards identifying the usage patterns and social meanings of nine features of e-grammar (Herring 2012). The indexicalities of features exhibiting gendered patterns in production as well as perception, emoticons (e.g. : ) ) and prosodic items (e.g. haha), are discussed in light of their linguistic and social context. Drawing on empirical research on American gender ideologies, the study argues that they index characteristics such as friendliness and emotional expressiveness, both stereotypically associated with women. In an instance of indirect indexicality (Ochs 1992), they are then linked to femininity in this type of computer-mediated communication. In production, the same features exhibit a strong audience effect (Bell 1984): women, for instance, use them more frequently in ads directed at other women. Throughout the analysis, the study makes use of and illustrates use cases for computational tools such as machine learning algorithms or automatic part-of-speech tagging in sociolinguistic research. At the same time, it attempts to strike a balance between a quantitative, data-driven approach and the nuanced analysis of gender identities and linguistic indexicality in the performance of gendered identities