‘But this one was so . . . male.’ A Corpus-Based and LLM-Augmented Analysis of Language and Gender Bias in Barbie

The 2023 film Barbie has sparked discussions on women’s empowerment and patriarchal norms, weaving feminist themes with critiques of gender roles while subtly reflecting patriarchal undertones that marginalize women (Myisha et al., 2023). We conduct a corpus-based analysis to investigate gender bias and differences in utterances’ distribution of part-of-speech (POS), affective values, and gender-linked classifications across three distinct scenes: 1) Barbie Land, 2) Real World, and 3) Post-Barbie Land, each representing the matriarchal, conventional and patriarchal theme, respectively. Leveraging large language models (LLMs), we extend Bradley and Lang’s (1999) affective norms to assess the emotional properties of film scripts, marking a novel application of LLMs in film script analysis.

Findings reveal gender-based differences and reflect power dynamics in and across scenarios, supporting prior research regarding gendered lexical preferences (e.g., Argamon et al., 2003; Jasmani et al., 2011). The results of affective analysis indicate that females consistently demonstrate higher valence and arousal scores across scenes, aligning with feminine communication styles emphasizing emotional rapport and expressivity, whereas males prioritize dominance (Tannen, 1990; Holmes and Stubbe, 2003). These variations underscore how affective language use is co-constructed by gender and socio-narrative context. Notably, gender-linked classifications (Barbie/Ken tones) align more closely with theoretically grounded gendered linguistic features than the broad stereotypes. This study highlights the linguistic construction of gender roles and power dynamics in Barbie, offering insights into the interplay between subversive feminism and traditional patriarchal norms and demonstrating how language reflects and challenges gender stereotypes in the film.