Author ORCID Identifier

https://orcid.org/0009-0002-2886-1163

Date of Award

Spring 6-14-2026

Document Type

Thesis (Undergraduate)

Department

Quantitative Social Science

First Advisor

Luke Chang

Second Advisor

Julie Kalish

Abstract

Sexual assault is a profound legal and social crisis. However, it is also fundamentally a linguistic one. The words used, or conspicuously not used, to describe victims, perpetrators, and their actions are not neutral arbiters of fact. They are powerful mechanisms that shape perceptions of harm, attributions of blame, and assignments of credibility. The central battleground for survivors is credibility, and while a “credibility discount” is often applied to female victims, the male victim is rendered nearly invisible. This research is therefore guided by one central, overarching question: how does a sexual assault victim’s gender influence the judicial language used, and do these patterns reinforce gender bias? To answer it, I built a stratified corpus of 97 contemporary U.S. judicial opinions (47 male-victim, 50 female-victim) and processed them through a custom NLP pipeline measuring 22 sexual-assault-specific framing detectors alongside broader linguistic features, with all comparisons stratified by criminal versus civil case type and validated under an opinion-only specification that excludes editorial headnote text. Four converging signatures of male-victim erasure emerged. Within the criminal stratum, female-victim opinions name the violence directly at substantially higher rates than male-victim opinions (Cliff’s δ = −.62), while male-victim opinions name the perpetrator at substantially higher rates (δ = +.63), both clearing Bonferroni correction. Within the civil stratum, female-victim opinions deploy criminal-conduct vocabulary at nearly twice the rate of male- victim opinions (δ = −.76). Direct victim-blame is near null at the within-criminal level, diagnostic of erasure operating through omission rather than active blame. These rhetorical findings sit atop a structural access asymmetry (34% of male-victim cases are criminal prosecutions versus 68% of female-victim cases, compounded by an 18× upstream filter-pool imbalance) produced by a multiplicative four-stage filtering process spanning policing, forensic infrastructure, civil channeling, and judicial language. Together the findings suggest that female and male victims face different forms of epistemic injustice—testimonial and hermeneutical, respectively—produced by the same patriarchal cultural matrix, and that the rhetorical erasure of male victimization requires its own conceptual and reform projects.

Download

Share

COinS