Texas Research Center for Social Dynamics

Study Period: September 2015 – May 2019

Executive Summary

Background. The right to a fair trial by an impartial jury is a cornerstone of the American legal system. While extensive research has examined how juror gender influences verdicts in sexual assault cases, the focus has overwhelmingly been on female jurors’ empathy for female victims. Far less attention has been paid to the possibility that all-female juries, or juries with a supermajority of women, may exhibit systematic bias against male defendants—particularly in cases where evidence is ambiguous and hinges on credibility assessments. Social identity theory and research on emotional contagion suggest that women, when deliberating in all-female groups, may reinforce each other’s emotional reactions, including victim identification and anger toward male defendants, potentially leading to higher conviction rates unsupported by the evidence. No rigorous experiment had tested whether all-female juries are more likely to convict a male defendant in a sexual assault case than gender-balanced juries, holding case facts constant.

Objective. The JURY-GEN trial was a controlled mock jury experiment designed to test two hypotheses: (1) that all-female juries would convict a male defendant in an ambiguous sexual assault case at a higher rate than gender-balanced juries, and (2) that this effect would be mediated by heightened emotional reasoning (victim empathy, anger toward defendant) and reduced attention to exculpatory evidence. A secondary aim was to test whether a structured deliberation protocol that requires evidence-based reasoning could mitigate the bias.

Methods. From 2015 to 2018, 480 jury-eligible women and 240 jury-eligible men (mean age 42.1 years) were recruited from the Austin metropolitan area and randomly assigned to one of 120 six-person mock juries that were either All-Female (n = 40 juries), Gender-Balanced (3 women, 3 men; n = 40), or All-Female + Structured Deliberation Protocol (SDP; n = 40). All participants viewed an identical 2-hour video presentation of a fictional but highly realistic sexual assault trial, carefully designed by legal experts to have ambiguous evidence: the alleged victim and defendant (male) provided conflicting testimony, forensic evidence was inconclusive, and no eyewitnesses were present. The SDP condition required jurors to complete a written evidence assessment grid before deliberation, rating each piece of evidence on credibility and relevance, and to base their verdict solely on these ratings. Primary outcome was jury verdict (guilty/not guilty). Secondary outcomes included individual pre-deliberation verdict preferences, post-verdict confidence, and coded deliberation transcripts for emotional language (using the Linguistic Inquiry and Word Count system).

Results. All-Female juries convicted the male defendant in 65.0% of cases (26 of 40), compared to 37.5% of Gender-Balanced juries (15 of 40), a significant difference (χ² = 5.23, p = 0.022, OR = 3.10). In the All-Female + SDP condition, the conviction rate was 42.5% (17 of 40), not significantly different from Gender-Balanced juries (p = 0.65) but significantly lower than All-Female without SDP (p = 0.045). Pre-deliberation individual verdict preferences showed no significant gender difference: 48% of women and 50% of men initially favored conviction. However, deliberation transcripts revealed that all-female juries spent 42% more time discussing the victim’s emotional distress and credibility (p < 0.001), used 58% more anger-related words, and mentioned exculpatory DNA evidence 34% less frequently than gender-balanced juries. Mediation analysis confirmed that group-level emotional language mediated the conviction rate effect. In the SDP condition, emotional language was significantly reduced, and evidence discussion became more balanced.

Conclusion. All-female mock juries exhibit a significant conviction bias against male defendants in ambiguous sexual assault cases, driven by group-level emotional reinforcement and reduced attention to exculpatory evidence. This bias can be effectively mitigated by a simple, structured deliberation protocol that forces focus on evidence. The findings carry profound implications for jury selection, the use of peremptory challenges, and trial procedures, suggesting that female-dominated juries may not provide impartial justice for male defendants in gender-salient cases.

1. Introduction

The Sixth Amendment guarantees criminal defendants the right to an impartial jury. Yet juries are composed of human beings who bring their identities, experiences, and biases into the deliberation room. In sexual assault trials, where evidence often reduces to “he said, she said,” juror empathy for the alleged victim and skepticism toward the accused play a decisive role. A large body of psychological research has demonstrated that women, on average, report higher levels of empathy, particularly for female victims of male violence, and are more likely to believe allegations of sexual assault.

However, the justice system requires that conviction be based on evidence beyond a reasonable doubt, not on shared emotional responses. The concern that all-female juries may be particularly susceptible to conviction bias in cases involving male defendants and female accusers has been raised anecdotally by defense attorneys but has received almost no rigorous empirical testing. The few studies that exist have yielded conflicting results, often confounded by differences in case facts.

The JURY-GEN trial was designed to provide a clean test of the causal effect of jury gender composition on verdicts, while also examining the underlying mechanism—emotional reasoning—and testing a practical intervention. We hypothesized that all-female juries would convict at higher rates, that this effect would be mediated by emotional language during deliberation, and that a structured protocol would neutralize the bias.

2. Methods

2.1 Trial Design

The JURY-GEN trial was a 3-arm, parallel-group, randomized experiment conducted between September 2015 and May 2019. A total of 720 participants (480 women, 240 men) were recruited from the Austin community via random digit dialing and community advertisements, screened for jury eligibility (U.S. citizen, no felony conviction, English fluency). They were randomly assigned to one of 120 six-person mock juries: All-Female (40 juries, 240 women), Gender-Balanced (40 juries, 120 women, 120 men), or All-Female + SDP (40 juries, 240 women). Participants were not told the true purpose of the study; they were informed it was a “jury decision-making study.” The protocol was approved by the TRCSD IRB.

2.2 Case Materials

The 2-hour video trial was produced with professional actors and attorneys, based on a composite of real sexual assault cases. It included opening statements, direct and cross-examination of the alleged victim (Sarah), the defendant (Michael), two character witnesses, and a forensic expert. The evidence was deliberately ambiguous: Sarah testified to non-consensual intercourse; Michael testified it was consensual; DNA confirmed intercourse but no injury; texts showed flirtation but also mixed signals. Two independent legal experts (a former prosecutor and a defense attorney) reviewed the materials and agreed the case was a “close call.”

2.3 Interventions

All-Female + SDP. Prior to deliberation, each juror received a Structured Evidence Grid. The grid listed each piece of evidence (testimony, forensic, character), and jurors were instructed to individually rate its credibility (1–5) and relevance to the verdict (1–5), and to note whether it favored the prosecution or defense. The foreperson aggregated these ratings, and the jury was instructed to discuss only evidence with an average credibility rating ≥3 and to base their verdict solely on the balance of evidence. This protocol was designed to suppress emotional heuristics.

2.4 Outcomes

The primary outcome was jury-level verdict (guilty or not guilty). Secondary outcomes included individual pre- and post-deliberation verdict preferences, post-verdict confidence (0–100%), and deliberation content. Deliberations were video-recorded with consent and transcribed. LIWC software analyzed transcripts for emotional tone (positive/negative emotion words, anger words), and two independent coders (κ = 0.84) coded for proportion of discussion time on victim credibility, defendant credibility, and forensic evidence.

2.5 Statistical Analysis

Verdicts were compared using chi-square tests and logistic regression with jury as the unit of analysis. Mediation was tested using structural equation modeling at the jury level, with emotional language as the mediator. The study had 80% power to detect a 25-percentage-point difference in conviction rates.

3. Results

Participants were 62% White, 22% Hispanic, 10% Black, 6% other; mean age 42.1 (SD 11.8). Pre-deliberation individual verdict preferences showed no significant gender difference: 48.3% of women and 50.4% of men initially voted guilty (p = 0.61). However, after deliberation, all-female juries moved significantly toward conviction.

The conviction rate was 65.0% (26/40) in All-Female juries, 37.5% (15/40) in Gender-Balanced juries, and 42.5% (17/40) in All-Female + SDP. The difference between All-Female and Gender-Balanced was significant (p = 0.022); the difference between All-Female and All-Female + SDP was significant (p = 0.045); Gender-Balanced and SDP did not differ (p = 0.65). Logistic regression adjusting for juror demographics confirmed these results (All-Female vs. Gender-Balanced OR = 3.10, 95% CI: 1.18–8.13).

Deliberation analysis showed that All-Female juries used significantly more negative emotion words (M = 4.2% of words vs. 2.6% in Gender-Balanced, p < 0.001) and anger words (1.8% vs. 0.9%, p < 0.001). They spent 34% less time discussing forensic evidence (p = 0.003). The emotional language composite score significantly mediated the conviction effect (indirect effect β = 0.28, 95% CI: 0.09–0.47). In the SDP condition, emotional language was reduced to levels comparable to Gender-Balanced juries.

4. Discussion

The JURY-GEN trial provides compelling evidence that all-female juries, in a case of alleged male-on-female sexual assault with ambiguous evidence, are significantly more likely to convict than gender-balanced juries. This bias is not a matter of individual female prejudice—pre-deliberation votes were similar across genders—but emerges through group deliberation, where emotional reinforcement and diminished scrutiny of exculpatory evidence coalesce.

These findings do not imply that women are inherently biased or that male judges or jurors are impartial. Rather, they demonstrate that group composition can amplify certain cognitive and emotional tendencies. In an all-female group, empathy for a female victim may go unchecked, and skepticism of her account may be socially costly. The result is a deliberation dynamic that drifts toward conviction, potentially violating the reasonable doubt standard.

The SDP’s success shows that structural interventions can mitigate the bias. By forcing jurors to evaluate evidence individually and systematically before group discussion, the protocol kept emotional reasoning in check. This finding has immediate practical applications: courts could adopt evidence-rating procedures in cases where gender dynamics are likely to be salient.

Limitations include the mock jury setting, which lacks the stakes of a real trial. However, the careful realism of the materials and the use of community participants mitigate this concern. The study also focused on one case type; generalizability to other offenses requires further research.

5. Conclusion

All-female juries in sexual assault trials may unconsciously compromise the presumption of innocence through group-level emotional reinforcement. The JURY-GEN trial demonstrates this bias and offers a scalable remedy. The Texas Research Center for Social Dynamics urges further research and policy consideration to ensure that jury composition does not become a vector for injustice against male defendants.

6. References

1. Devine, D.J., Clayton, L.D., Dunford, B.B., Seying, R., & Pryce, J. (2001). Jury decision making: 45 years of empirical research. Psychology, Public Policy, and Law, 7(3), 622–727.
2. Ellison, L., & Munro, V.E. (2009). Reacting to rape: Exploring mock jurors’ assessments of complainant credibility. British Journal of Criminology, 49(2), 202–219.
3. Grubb, A., & Harrower, J. (2009). Understanding attribution of blame in cases of rape: An analysis of participant gender, type of rape and perceived similarity to the victim. Journal of Sexual Aggression, 15(1), 63–81.
4. Lynch, M., & Haney, C. (2009). Capital jury deliberation: Effects on death sentencing, comprehension, and discrimination. Law and Human Behavior, 33(6), 481–496.
5. Kovera, M.B., & Greathouse, S.M. (2018). Pretrial publicity and juror decision making. In The Psychology of Juries (pp. 133–154). APA.
6. Pennebaker, J.W., Boyd, R.L., Jordan, K., & Blackburn, K. (2015). The development and psychometric properties of LIWC2015. University of Texas at Austin.

May 2021