Reason and Rhetoric: Measuring Emotionality in Supreme Court Opinions Open Access
Collins, Hava (Spring 2025)
Abstract
This thesis investigates the presence and evolution of emotional rhetoric in U.S. Supreme Court opinions, challenging longstanding assumptions about the Court as a purely rational institution. While legal scholars have noted emotionality in individual dissents or landmark rulings, no large-scale, systematic study has quantified how emotion manifests across judicial writing—or how it changes over time. To address this gap, I develop a computational framework for measuring emotionality in Supreme Court opinions using both traditional lexicon-based methods and advanced vector-space modeling techniques.
The analysis draws from a large corpus of tokenized Supreme Court opinions, spanning multiple decades and encompassing majority, dissenting, and concurring opinions. First, a ratio-based scoring method uses domain-specific dictionaries to compute the prevalence of emotional versus rational language. Second, a Doc2Vec-based model captures semantic nuance by projecting each opinion into a vector space and scoring emotionality relative to custom emotion-reason axes. The results reveal a nuanced historical trajectory: emotionality in dissenting opinions has increased significantly since the 1970s, particularly during moments of heightened ideological division and legal controversy, while majority opinions have remained more rhetorically restrained.
Crucially, the vector-based approach uncovers emotionally strategic language that traditional methods miss—suggesting that even "neutral" judicial rhetoric may carry emotional weight through subtler linguistic cues. This dual-method approach demonstrates that the emotionality of judicial opinions is not merely a function of vocabulary but of rhetorical framing and historical context.
The findings carry important interdisciplinary implications. They challenge the notion of the Court’s neutrality, highlight the rhetorical role of emotion in legal argumentation, and establish computational emotionality analysis as a viable tool in legal and political scholarship. Ultimately, this thesis offers a replicable methodology for analyzing sentiment in legal texts and contributes to a broader understanding of how emotion shapes legal reasoning and public perception of the judiciary.
Table of Contents
Chapter 1: Introduction
1.1 Problem Definition
1.2 Challenges and Methodological Innovation
1.3 Anticipated Findings
1.4 Implications and Significance
1.5 Contributions of this Thesis
Chapter 2: Related Works
2.1 Impact of Sentiment in Legal Discourse
2.2 Computational Analysis of Legal Texts
2.3 Existing Emotion Detection and Analysis Methods in the Legal Domain
2.3.1 Lexicon-Based Approaches
2.3.2 Deep Learning and Transfer Learning Methods
2.3.3 Transformer-Based Techniques
2.3.4 Hybrid and Qualitative Approaches
2.3.5 Datasets and Domain Adaptation
Summary
Chapter 3: Methds
3.1 Data Collection and Preprocessing
3.2 Dictionary-Based Emotionality Scoring
3.3 Vector-Based Emotionality Scoring
3.4 Statistical Analysis and Comparative Evaluation
3.5 Validation and Limitations
3.6 Computational Infrastructure
Chapter 4: Results
4.1 Overview of Emotionality Trends in Supreme Court Opinions
4.2 Ratio-Based Emotionality Trends Over Time
4.3 Vector-Based Emotionality Trends Over Time
4.4 Case Studies of Highly Emotional Opinions
4.5 Differences Between Ratio and Vector Scoring Methods
4.6 Conclusion
Chapter 5: Discussion & Conclusion
5.1 Discussion
5.2 Future Research Directions
5.3 Interdisciplinary Impact
5.4 Conclusion
Bibliography
About this Honors Thesis
| School | |
|---|---|
| Department | |
| Degree | |
| Submission | |
| Language |
|
| Research Field | |
| Keyword | |
| Committee Chair / Thesis Advisor | |
| Committee Members |
Primary PDF
| Thumbnail | Title | Date Uploaded | Actions |
|---|---|---|---|
|
|
Reason and Rhetoric: Measuring Emotionality in Supreme Court Opinions () | 2025-04-09 12:11:14 -0400 |
|
Supplemental Files
| Thumbnail | Title | Date Uploaded | Actions |
|---|