Victim Language Project

Participants will be randomly assigned to read a vignette that refers to the complainant either as the “victim” or as the “complaining witness.” Participants will also be randomly assigned the gender of the complainant in the vignette, so that the complainant is either male or female. After reading the vignette, participants will answer a series of questions about their beliefs regarding the defendant, the complainant, and the alleged incident.

Participants will read one of four versions of the vignette below. Bracketed language indicates the text that varies by treatment condition:

Imagine you are serving on a jury for a case involving a man who allegedly punched [another man/a woman]. At trial, the defendant claimed this never happened and that the [victim/complaining witness], [John/Jane Doe], completely fabricated the entire story to get back at him for a heated argument they had earlier that night. The [victim/complaining witness] says the defendant punched [him/her] after the heated argument. Phone records show that [the victim/complaining witness] texted [his/her] friend that night that the defendant had just punched [him/her] and asked what to do. The friend told [him/her] to call the police. Immediately after receiving this advice, the [victim/complaining witness] called the police, who took [his/her] statement. There were no eyewitnesses or video evidence of the alleged assault. The police report noted no visible injury on the [victim/complaining witness].

After reading the vignette, participants will answer the primary and secondary outcome questions described below using 0–100 slider scales.

2026-07-26

2026-08-07

The two primary outcomes are (1) the extent to which participants believe that the defendant punched the victim/complaining witness and (2) the extent to which participants agree that the alleged victim is indeed a victim. Both outcomes are measured on a 0–100 slider scale, where 0 means “absolutely not,” 50 means “too close to say,” and 100 means “absolutely yes.”

Secondary outcomes include participants’ ratings of:
• How violent or aggressive they believe the defendant is.
• Whether they believe the defendant will commit crimes in the future.
• What share of fault they believe the victim/complaining witness bears for the incident.
• What share of fault they believe the defendant bears for the incident.
• How severely they believe the defendant should be punished.
• How believable they find the victim/complaining witness’s account.
• How believable they find the defendant’s account.
• How much sympathy they feel for the victim/complaining witness.
Each secondary outcome is also measured using the same 0–100 slider scale. Higher values indicate more of the construct described in the corresponding survey question. These outcomes are intended to shed light on potential mechanisms through which victim terminology may influence mock jurors’ judgments.

The experiment uses a 2x2 factorial design. The experiment randomly varies: (1) the term used to refer to the complainant, either “victim” or “complaining witness”; and (2) the gender of the complainant, either male or female.

Participants will not be informed of the study hypotheses or the treatment assignment during data collection. There is no direct interaction between researchers and participants during data collection. Researchers analyzing the data will not be blinded to treatment assignment because treatment assignment is recorded in the survey data and is necessary for the planned analysis. No human coding of open-ended participant responses is planned. These procedures are intended to reduce unintended influence from participants, researchers, coders, or analysts having knowledge of the study design, treatments, measures, or outcomes during the research process.

Not available

Randomization will be conducted by Qualtrics using computerized random assignment. Participants will be randomly assigned to treatment conditions with equal probability.

The unit of randomization is the individual survey participant. There is no clustered randomization.

No

Not applicable. There are no clusters because randomization occurs at the individual-participant level.

400 survey participants.

100 participants per treatment arm, for a total planned sample of 400 allocated equally across the four cells of the 2x2 factorial design:
• 100 participants: male “victim”
• 100 participants: female “victim”
• 100 participants: male “complaining witness”
• 100 participants: female “complaining witness”
Equivalently, 200 participants will be assigned to the “victim” label condition and 200 participants will be assigned to the “complaining witness” label condition; 200 participants will see a male complainant and 200 participants will see a female complainant.

For the main treatment contrast on the primary outcomes, a sensitivity power analysis indicates that the study can detect a minimum effect size of approximately Cohen’s d = 0.281 with 80% power at α = 0.05. At 90% power, the minimum detectable effect is approximately Cohen’s d = 0.325.

Victim Language Project Pre-Analysis Plan