Heard or Halted? Gender, Interruptions, and Emotional Tone in U.S. Supreme Court Oral Arguments

This study examines how interruptions during U.S. Supreme Court oral arguments shape both the semantic content and emotional tone of advocates’ speech, with a focus on gendered dynamics in judicial discourse. Using the ConvoKit Supreme Court Corpus (2010-2019), we analyze 12,663 speech chunks from advocate-justice interactions to assess whether interruptions alter the meaning of an advocate’s argument and whether interruptions toward female advocates exhibit more negative emotional valence. Semantic shifts are quantified using GloVe-based sentence embeddings, while sentiment is measured through lexicon-based analysis. We find that semantic similarity between pre- and post-interruption speech remains consistently high, suggesting that interruptions do not substantially alter argumentative content. However, interruptions directed at female advocates contain significantly higher levels of negative sentiment. These results deepen empirical understanding of gendered communication in elite institutional settings and demonstrate the value of computational linguistic methods for studying power, discourse, and equity in judicial proceedings.

💡 Research Summary

This paper presents a computational linguistic analysis of interruptions during U.S. Supreme Court oral arguments from 2010 to 2019, focusing on their impact on semantic content and emotional tone, with particular attention to gendered dynamics. The study addresses two primary research questions: 1) Do interruptions alter the core semantic meaning of an advocate’s argument? 2) Do interruptions directed at female advocates carry a more negative emotional sentiment compared to those directed at male advocates?

The research utilizes the ConvoKit Supreme Court Corpus, analyzing 12,663 “speech chunks” – continuous segments of an advocate’s speech between interruptions or pauses during an exchange with a single justice. Advocate gender was inferred using a name-gender reference dataset. The methodology employs two key techniques. First, to measure semantic shifts, the study uses pre-trained GloVe word embeddings. The text of each speech chunk is converted into an average sentence vector. For each advocate, a composite embedding is created for all their interrupted chunks and another for all uninterrupted chunks. The cosine similarity between these two vectors is then calculated to gauge how much the meaning changes after an interruption.

Second, to assess emotional tone, the study applies lexicon-based sentiment analysis using the NRC Emotion Lexicon. For every interrupted chunk, the proportion of words associated with negative sentiment (including anger, fear, sadness, and disgust) relative to all emotion-bearing words is computed, creating a “negative sentiment ratio.” This metric is then compared across gender groups.

The findings reveal a clear dichotomy. On the question of semantic content, the results show an exceptionally high degree of semantic similarity between interrupted and uninterrupted speech. The mean cosine similarity score was 0.997, indicating that interruptions do not substantially derail or alter the fundamental meaning of an advocate’s argument. Advocates consistently return to their original argumentative trajectory post-interruption.

However, the analysis of emotional tone uncovers a significant gender difference. Interruptions directed at female advocates contain a statistically significant higher level of negative sentiment compared to those directed at male advocates. A Welch two-sample t-test confirmed this difference (p = 0.0078). This finding persists even after controlling for potential confounding factors such as advocate experience, case year, and ideological alignment between the justice and the advocate in a regression model. The coefficient for male advocates remained negative and significant, suggesting that the higher negativity in interruptions towards women is not an artifact of other variables.

In conclusion, the study demonstrates that while interruptions in this high-stakes institutional setting may not succeed in changing what advocates say (the semantic content), they are associated with a gendered difference in how they are spoken to (the emotional tone). The interruptions directed at women are imbued with more negative emotion. This insight deepens our understanding of the nuanced ways in which power and gender bias can manifest in elite discursive practices, moving beyond simple frequency counts to a more qualitative assessment of communicative dynamics. The research also underscores the value of computational text analysis methods for uncovering subtle patterns in legal and political discourse.