Who’s the “Real” Victim? How Victim Framing Shapes Attitudes Toward Sexual Assault

Abstract

People accused of sexual assault are often described as the “real” victim by their defenders, but the impact of “victim framing” on public opinion is unknown. We investigated this issue across four experiments (N = 2,614). Online U.S. adult participants read a report about an alleged sexual assault that framed the female accuser as the victim (of assault), framed the male alleged perpetrator as the victim (of false accusations), or was neutral about victimhood (baseline). Relative to those in the baseline condition, participants in the assault- and allegation-victim conditions generally expressed more support for the victim-framed protagonist and less support for the other protagonist. The consistency of these effects varied with how often the victim frame was instantiated and whether the report described a fictionalized or real-world case. Across all contexts, however, participants who identified the victim-related language as influencing their evaluations exhibited strong framing effects. This suggests that social-pragmatic reasoning is a key mechanism by which victim framing shapes moral judgments.

Keywords

victim framing victim blaming communication language and thought moral evaluation social-pragmatic reasoning open data open materials preregistered

In a confidential July 2018 letter leaked to the media, Dr. Christine Blasey Ford alleged that U.S. Supreme Court nominee Brett Kavanaugh sexually assaulted her at a party in the 1980s. On September 16, 2018, Ford spoke publicly about her accusations, providing a detailed account of the event and its aftermath—an account she reaffirmed under oath before the Senate Judiciary Committee. On September 21, the conservative-leaning Washington Times published an article with the headline, “Christine Blasey Ford is not the victim here—Brett Kavanaugh is” (Chumley, 2018). Kavanaugh was confirmed on October 6.

This is hardly the first time an alleged perpetrator of sexual assault has been cast by their defenders as the “real” victim in an effort to mitigate blame. This rhetorical device, which we call victim framing, is distinct from the more frequently discussed act of victim blaming: holding the victim of a crime responsible for what happened to them, often by citing their behavior (e.g., choice of clothing) as a cause of the attack (Niemi, 2017). The apparent goal of victim framing, by contrast, is to directly foster empathy for the perpetrator. This is arguably justifiable in cases in which the alleged perpetrator is innocent. However, false allegations of sexual assault are rare (2%–10% of cases; Lisak et al., 2010), and victim framing can also be deployed by guilty parties as a bid for public support. Despite anecdotal evidence for its increasing prevalence in the #MeToo era (Hesse, 2019), victim framing has received little attention from researchers. Whether it has any measurable impact on how people evaluate sexual assault cases is an open question.

On the one hand, decades of research have shown that subtle linguistic cues shape reasoning (e.g., Holmes et al., 2021; Loftus & Palmer, 1974; Thibodeau & Boroditsky, 2011; Tversky & Kahneman, 1981). Casting perpetrators as victims might encourage people to view them through that lens, leading to increased support. On the other hand, victim framing could have the opposite effect if observers suspect that the alleged perpetrator is merely evading responsibility for their actions. It might also backfire if the victim label focuses observers’ attention on the perpetrator, which has been shown to make the accuser seem less blameworthy (Niemi & Young, 2016).

In light of recent high-profile instances of victim framing, the lack of targeted research on the topic, and diverging predictions about its efficacy, we set out to systematically assess the effects of victim framing and the mechanisms that drive them.

All data and materials are available on OSF (https://osf.io/ks5tr/ and https://osf.io/ydf2u/, respectively). In addition, the preregistered design and analysis plans for Experiments 2 to 4 can be viewed at https://osf.io/4jgsz/. All experiments were approved by the institutional review board at Colorado College. Participants in all experiments were located in the United States, were at least 18 years old, and had reliable performance (≥ 95% rating) on Amazon’s Mechanical Turk (MTurk; Buhrmester et al., 2011).

Experiment 1

Method

Participants

We recruited a convenience sample of 606 participants through MTurk. We aimed for 100 participants per condition, comparable with sample sizes in other linguistic-framing studies. Table 1 shows participant demographic data.

Table 1.

Demographic Data for All Experiments

Variable	Experiment 1	Experiment 2	Experiment 3	Experiment 4
Time of data collection	December 2018	July 2019	August 2019	August 2019
N
Sampled	606	847	615	626
Analyzed	606	808	599	601
Gender
Female	44%	46%	47%	47%
Male	56%	53%	53%	52%
Mean age (SD)		36.4 years (11.7)	35.6 years (10.6)	36.8 years (11.6)
Race/ethnicity
White	73%	71%	71%	66%
Black	9%	10%	12%	15%
Asian	6%	10%	6%	7%
Latinx	5%	5%	6%	6%
Multiracial	5%	4%	4%	5%
Political affiliation
Democrat	45%	48%	51%	45%
Republican	28%	23%	20%	26%

Design, materials, and procedure

The experiment had a 3 (frame: assault victim vs. allegation victim vs. baseline) × 2 (detail: sparse vs. rich) between-subjects design. Participants were told that the experiment was about news consumption, and each participant was randomly assigned to read one of six different news reports describing an alleged sexual assault on a college campus. Such incidents are common, receive media attention, and often feature competing claims of victimhood. In the court of public opinion, and sometimes in the literal courtroom, a broad sample of adults forms judgments of these cases.

Statement of Relevance

False allegations of sexual assault are rare, yet people accused of assault are often described as the “real” victim by their supporters. We explored whether this purposeful attempt at persuasion—which we call victim framing—actually works. In our experiments, people read a news report that described an alleged sexual assault, often in vivid detail. A quote from a friend, reflected in the headline, framed one character as the victim (of assault or false accusations). We found that this framing worked: People in the framing conditions supported the victim-framed character more than did people in a baseline condition. Interestingly, this was true only for people who told us that the victim label affected their thinking. This means that victim framing can shift public opinion but only when people reason that the victim label has been intentionally applied to someone who deserves their support. Our work shows that the way people talk about sexual assault can shape how others view it.

We varied which individual was described as the victim in the headline and main text of the report and how much detail was provided about the alleged assault and the timing of the accusations (see Table 2). We included the detail manipulation to mirror variability in media coverage about sexual assault and to see whether victim framing would be less effective when more details were available.

Table 2.

News-Report Stimuli Used in Experiments 1 and 2

Report section	Frame
Report section	Assault victim	Allegation victim	Baseline
Headline	Victim of Sexual Assault Faces Long Road Ahead (May 4, 2018)	Victim of Sexual Assault Allegations Faces Long Road Ahead (May 4, 2018)	Students Embroiled in Sexual Assault Investigation Face Long Road Ahead (May 4, 2018)
Framing	Reactions on campus are polarized, but friends have rallied around Sawyer. “Emma is a victim of sexual assault who deserves to be believed,” said a close friend of both students. “She has been traumatized by this experience. The investigation ahead is going to be long and grueling, but I know she will survive it.”	Reactions on campus are polarized, but friends have rallied around Bradley. “David is a victim of false allegations who deserves to be believed,” said a close friend of both students. “He has been traumatized by this experience. The investigation ahead is going to be long and grueling, but I know he will survive it.”	Reactions on campus are polarized, but friends have rallied around both Sawyer and Bradley. “Emma and David should each be allowed to tell their side of the story,” said a close friend of both students. “This experience has been challenging for both of them. The investigation ahead is going to be long and grueling, but I know the truth will come out.”

Note: Participants read one of the three headlines, followed by either the sparse-detail or rich-detail paragraph (see text). The end of the paragraph was framed in a way that was consistent with the headline, for a total of six unique reports.

Participants in the sparse-detail condition read the following passage:

Emma Sawyer, a student at Livingston University, filed a complaint against fellow student David Bradley, alleging that Bradley sexually assaulted her at a campus party last fall. University authorities are investigating Sawyer’s allegations, which Bradley has firmly denied.

Participants in the rich-detail condition read the following passage:

Emma Sawyer, a student at Livingston University, filed a complaint against fellow student David Bradley, alleging that Bradley sexually assaulted her at a Halloween party hosted by the Sigma Chi fraternity last fall. In the complaint, Sawyer claimed that Bradley lured her to a private bedroom in the fraternity house and proceeded to remove her clothing, pin her to a bed, and force himself onto her despite her protestations. University authorities are investigating Sawyer’s allegations, which Bradley has firmly denied. Bradley also questioned why the allegations had not surfaced until six months after the purported incident and noted that he and Sawyer were both intoxicated on the night in question.

In both conditions, an additional section was included at the end of the paragraph. This section framed the report in a way that was consistent with the headline by quoting from a supporter of Emma (assault-victim condition), David (allegation-victim condition), or both (baseline). This resulted in a total of six unique reports (see Table 2). Participants were able to advance to the next screen after 15 s (sparse-detail condition) or 25 s (rich-detail condition).

Support

After reading the report, participants answered eight questions assessing their level of support for the two protagonists. Using scales from 0 (none/not at all) to 6 (a lot/very), participants indicated how much empathy they had for Emma and David, how believable each of them was, how much harm each had experienced as a result of the incident, and how responsible each was for the incident (reverse scored). Questions about Emma and David were interleaved throughout this and the following sections, and the order of the protagonists was randomized across participants. For each participant, we computed a mean support score for each protagonist by averaging their responses to these four questions (α = .81).¹

Additional measures

Participants also rated how apt the words victim and survivor were for describing Emma and David and provided their general opinion of each protagonist. These data were collected for a different project assessing the connotations of “victim” and “survivor” labels (Schultz et al., 2020) and will not be discussed further. Next, participants indicated whether the report reminded them of any recent real-life news stories and completed a measure of personal experience with sexual assault, a measure of acceptance of interpersonal violence (Burt, 1980), and the modified Illinois Rape Myth Acceptance Scale (McMahon & Farmer, 2011). Finally, participants answered basic demographic questions. Further details and analyses of these measures are provided in the Supplemental Material available online.

Results

Participants in the baseline condition expressed more support for Emma (M = 4.16, SE = 0.08) than David (M = 2.78, SE = 0.09), t(196) = 9.12, p < .001, d = 0.65. This suggests that there was relatively little victim blaming in the absence of victim framing.

To assess the effects of victim framing, we conducted separate 3 (frame) × 2 (detail) analyses of variance (ANOVAs) on support scores for each protagonist. In both analyses, there was a main effect of frame—support for Emma: F(2, 600) = 13.95, p < .001, η_p² = .04; support for David: F(2, 600) = 12.01, p < .001, η_p² = .04. Planned contrasts showed that support for Emma was higher than baseline (M = 4.16, SE = 0.08) in the assault-victim condition (M = 4.44, SE = 0.08), t(400) = 2.34, p = .02, d = 0.23, and lower than baseline in the allegation-victim condition (M = 3.79, SE = 0.10), t(399) = −2.93, p = .004, d = 0.29. Conversely, support for David was higher than baseline (M = 2.78, SE = 0.09) in the allegation-victim condition (M = 3.07, SE = 0.09), t(399) = 2.35, p = .02, d = 0.23, and lower than baseline in the assault-victim condition (M = 2.45, SE = 0.10), t(400) = −2.53, p = .01, d = 0.25 (see Fig. 1).

Fig. 1.

Support for (a) the accuser and (b) the alleged perpetrator in Experiment 1, by frame and level of detail. Boxes denote the interquartile range, the middle line indicates the median, whiskers extend from the 10th to the 90th percentiles, and the plus sign denotes the mean.

There was also a main effect of detail on support for David, F(1, 600) = 5.86, p = .02, η_p² = .01, with more support when the report had sparse detail (M = 2.89, SE = 0.08) than when it had rich detail (M = 2.64, SE = 0.07). There was no main effect of detail on support for Emma, F(1, 600) = 0.02, p = .89, η_p² < .01, and no interaction between frame and detail for either protagonist—Emma: F(2, 600) = 1.86, p = .16, η_p² = .01; David: F(2, 600) = 0.38, p = .69, η_p² < .01. Thus, victim framing was effective regardless of the level of detail about the incident, but additional graphic details decreased support for the alleged assailant.²

Discussion

Experiment 1 demonstrated that victim framing works: Relative to the baseline condition, participants in both the assault-victim and allegation-victim conditions expressed more support for whoever was labeled the victim—and less support for the other individual—regardless of the level of detail provided about the incident. We sought to replicate these results in Experiment 2 using a larger sample and preregistered design.

Experiment 2: Replication

Method

Participants

Following our preregistered target sample size and exclusion criteria, we recruited 847 new participants through MTurk. Participants who failed an initial attention check were prevented from completing the experiment (n = 39). Our final sample (N = 808) provided greater than 99% power to detect a main effect of frame based on the Experiment 1 effect sizes, as determined using G*Power (Version 3.1; Faul et al., 2007; see Table 1 for participant demographics).

Materials and procedure

Experiment 2 was identical to Experiment 1, with one exception: Before completing the personal experience measure, participants were shown the report again and asked to provide a rationale for their evaluation of the protagonists by (a) copying and pasting the part of the report they found most influential into a text box and (b) indicating any other information that contributed to their evaluation. This enabled us to investigate participants’ reliance on the victim frame for their evaluations.

Results

Participants in the baseline condition expressed more support for Emma (M = 4.37, SE = 0.07) than David (M = 2.69, SE = 0.08), t(269) = 12.34, p < .001, d = 0.75, again indicating low levels of victim blaming in the absence of victim framing.

As shown in Figure 2, the results partially replicated those of Experiment 1. In 3 (frame) × 2 (detail) ANOVAs, there was a main effect of frame on support for each protagonist—Emma: F(2, 802) = 3.07, p = .047, η_p² = .01; David: F(2, 802) = 6.87, p = .001, η_p² = .02. Support for Emma was lower than baseline (M = 4.37, SE = 0.08), albeit not significantly, in the allegation-victim condition (M = 4.17, SE = 0.08), t(537) = −1.89, p = .059, d = 0.16, but was similar to baseline in the assault-victim condition (M = 4.41, SE = 0.07), t(537) = 0.41, p = .68, d = 0.04. Conversely, support for David was lower than baseline in the assault-victim condition (M = 2.38, SE = 0.08), t(537) = 2.69, p = .007, d = 0.23, but similar to baseline in the allegation-victim condition (M = 2.79, SE = 0.08), t(537) = 0.88, p = .38, d = 0.08.

Fig. 2.

Support for (a) the accuser and (b) the alleged perpetrator in Experiment 2, by frame and level of detail. Boxes denote the interquartile range, the middle line indicates the median, whiskers extend from the 10th to the 90th percentiles, and the plus sign denotes the mean.

Unlike in Experiment 1, there was a main effect of detail on support for Emma, F(1, 802) = 6.61, p = .01, η_p² = .01, with more support when the report had sparse detail (M = 4.43, SE = 0.06) than rich detail (M = 4.21, SE = 0.06), but no main effect of detail on support for David, F(1, 802) = 0.51, p = .47, η_p² < .01. Mirroring Experiment 1, however, results showed no interaction between frame and detail for either protagonist—Emma: F(2, 802) = 0.44, p = .65, η_p² < .01; David: F(2, 802) = 0.21, p = .81, η_p² < .01—again showing that the level of detail about the incident did not moderate the framing effects.

We used automatic text coding to determine whether participants’ rationales in the two framing conditions included the victim-related language from the report (i.e., the words victim, traumatized, or survive or cognates with the same root). Across both conditions, 43% of participants (“citers”) included one or more of these victim-related words in their rationales. In 2 (assault victim vs. allegation victim) × 2 (citers vs. nonciters) ANOVAs on support for each protagonist, the interaction was significant—Emma: F(1, 534) = 14.71, p < .001, η_p² = .03; David: F(1, 534) = 20.58, p < .001, η_p² = .04. As shown in Figure 3, the overall framing effect was driven by citers: For them, support for Emma was higher in the assault-victim than allegation-victim condition, t(227) = 4.69, p < .001, d = 0.62, whereas support for David was higher in the allegation-victim than assault-victim condition, t(227) = −5.79, p < .001, d = 0.77. In contrast, for nonciters, there was no significant difference between conditions for either protagonist (ps > .46, ds < 0.09). We also conducted preregistered analyses on support difference scores (support for Emma – support for David) that corroborated the main findings (see the Supplemental Material).³

Fig. 3.

Support for (a) the accuser and (b) the alleged perpetrator in the assault-victim and allegation-victim conditions of Experiment 2, separately for participants who cited victim-related language as influential (“citers”) and those who did not (“nonciters”). Boxes denote the interquartile range, the middle line indicates the median, whiskers extend from the 10th to the 90th percentiles, and the plus sign denotes the mean. The dashed line indicates the mean in the baseline condition.

Discussion

Experiment 2 partially replicated Experiment 1: Participants expressed less support for the non-victim-framed protagonist relative to baseline, regardless of how much detail was included about the incident. As in many replication studies, the effect sizes were smaller (Open Science Collaboration, 2015). Notably, we observed strong victim-framing effects among citers—participants who indicated that the victim-related language influenced their decision-making. This suggests that social-pragmatic reasoning may be a key driver of these framing effects. That is, citers may have reasoned that the victim-related language was chosen by the writer to be informative and adjusted their evaluations accordingly (Grice, 1975; Sher & McKenzie, 2006; Sperber & Wilson, 1986).

Our goals in Experiment 3 were to replicate the moderating effect of citing victim-related language and to examine whether a minimal instantiation of the victim frame would still yield reliable framing effects. Whereas more elaborate linguistic framing often yields stronger effects (Flusberg et al., 2020; Thibodeau, 2016), a single instantiation of a frame can be sufficient (Thibodeau & Boroditsky, 2011).

Experiment 3: Minimal Framing

Method

Participants

We recruited 615 new participants through MTurk, using the CloudResearch platform (Litman et al., 2017). Sixteen failed an initial attention check. Our final sample (N = 599) provided greater than 88% power to detect the main effect of frame based on the effect sizes across Experiments 1 and 2 and greater than 85% power to detect the interaction of frame and citing victim-related language based on the Experiment 2 effect sizes (see Table 1 for participant demographics).

Materials and procedure

Experiment 3 was identical to Experiment 2, except for one critical difference: The reports were modified to include only minimal victim framing—one instance in the headline and one near the end (see Table 3).

Table 3.

News-Report Stimuli Used in Experiment 3

Report section	Frame
Report section	Assault victim	Allegation victim	Baseline
Headline	Victim of Sexual Assault Braces for Investigation (May 4, 2018)	Victim of Sexual Assault Allegations Braces for Investigation (May 4, 2018)	Students Brace for Sexual Assault Investigation (May 4, 2018)
Framing	Reactions on campus are polarized, but one student said, “Emma is the real victim here.” A university representative declined to comment.	Reactions on campus are polarized, but one student said, “David is the real victim here.” A university representative declined to comment.	Reactions on campus are polarized. A university representative declined to comment.

Note: Participants read one of the three headlines, followed by either the sparse-detail or rich-detail paragraph (see text), which remained unchanged from Experiments 1 and 2. The end of the paragraph was framed in a way that was consistent with the headline, for a total of six unique reports.

Results

In the baseline condition, participants again expressed more support for Emma (M = 4.36, SE = 0.09) than David (M = 2.32, SE = 0.09), t(201) = 12.59, p < .001, d = 0.89. In 3 (frame) × 2 (detail) ANOVAs on support scores, the main effect of frame was in the expected direction, although not significant—Emma: F(2, 593) = 1.59, p = .21, η_p² = .01; David: F(2, 593) = 2.31, p = .10, η_p² = .01 (see Fig. 4). As in Experiment 1, there was a main effect of detail on support for David, F(1, 593) = 4.47, p = .04, η_p² = .01, with more support when the report had sparse detail (M = 2.48, SE = 0.08) than when it had rich detail (M = 2.25, SE = 0.08), but no main effect of detail on support for Emma, F(1, 593) = 0.21, p = .65, η_p² < .01. There was no interaction between frame and detail for either protagonist—Emma: F(2, 593) = 0.28, p = .75, η_p² < .01; David: F(2, 593) = 0.50, p = .61, η_p² < .01.

Fig. 4.

Support for (a) the accuser and (b) the alleged perpetrator in Experiment 3, by frame and level of detail. Boxes denote the interquartile range, the middle line indicates the median, whiskers extend from the 10th to the 90th percentiles, and the plus sign denotes the mean.

Although the victim-framing effects were not statistically significant in Experiment 3 alone, a cross-experiment comparison showed that the magnitude of these effects was comparable with the more elaborate framing of Experiments 1 and 2. In 3 (frame) × 2 (experiment: 1 and 2 vs. 3) ANOVAs, there was a main effect of frame on support for each protagonist—Emma: F(2, 2007) = 9.60, p < .001, η_p² = .01; David: F(2, 2007) = 12.15, p < .001, η_p² = .01—as well as a main effect of experiment on support for David, F(1, 2007) = 24.10, p < .001, η_p² = .01, with more support with elaborate than minimal framing. There was no main effect of experiment on support for Emma, F(1, 2007) = 2.43, p = .12, η_p² < .01, and no interaction between frame and experiment for either protagonist—Emma: F(2, 2007) = 0.92, p = .40, η_p² < .01; David: F(2, 2007) = 1.50, p = .22, η_p² < .01—suggesting that minimal and elaborate victim framing did not differ qualitatively in efficacy.

As in Experiment 2, the framing effects were moderated by whether participants cited the victim-related language (in the minimally framed report, just “victim”) as influential. In 2 (assault victim vs. allegation victim) × 2 (citers vs. nonciters) ANOVAs on support for each protagonist, the interaction was significant—Emma: F(1, 393) = 6.65, p = .01, η_p² = .02; David: F(1, 393) = 13.05, p < .001, η_p² = .03. As shown in Figure 5, citers (30% of participants in the framing conditions) showed significant framing effects—Emma: t(118) = 3.52, p < .001, d = 0.65; David: t(118) = −4.34, p < .001, d = 0.81—whereas nonciters (70%) did not—Emma: t(275) = 0.09, p = .93, d = 0.01; David: t(275) = 0.34, p = .74, d = 0.04. As in Experiment 2, we also conducted preregistered analyses on support difference scores (support for Emma – support for David) that corroborated the main findings (see the Supplemental Material).

Fig. 5.

Support for (a) the accuser and (b) the alleged perpetrator in the assault-victim and allegation-victim conditions of Experiment 3, separately for citers and nonciters of victim-related language. Boxes denote the interquartile range, the middle line indicates the median, whiskers extend from the 10th to the 90th percentiles, and the plus sign denotes the mean. The dashed line indicates the mean in the baseline condition.

Discussion

The main effect of frame, although not significant in Experiment 3, did not differ reliably from the effect found in the previous experiments, which used a more maximal instantiation of the framing language. Critically, we found confirmatory evidence for the moderating effect of citing this language: Only participants who cited the word victim as informing their evaluations exhibited significant victim-framing effects.

In Experiments 1 to 3, participants evaluated a fictionalized scenario, which allowed us to control for familiarity but limited ecological validity. The report in Experiment 4 described a real event: Dr. Ford’s allegations against now-Justice Kavanaugh. Data were collected approximately 10 months after Kavanaugh’s Supreme Court confirmation, providing a natural cover story for participants accustomed to expressing attitudes toward salient sociopolitical events in online studies. This minimized any demand characteristics associated with our design.

Experiment 4: Real Event

Method

Participants

We recruited 626 new participants through MTurk using CloudResearch. Twenty-five failed an initial attention check. Our final sample (N = 601) provided greater than 88% power to detect the main effect of frame based on the effect sizes across Experiments 1 and 2 and greater than 85% power to detect the interaction of frame and citing victim-related language based on the Experiment 2 effect sizes (see Table 1 for participant demographics).

Materials and procedure

The report described Ford’s allegations and closely matched the reports used in the previous experiments (see Table 4).

Participants in the sparse-detail condition read the following:

Participants in the rich-detail condition read the following:

Christine Blasey Ford, a psychology professor at Palo Alto University, testified before the Senate Judiciary Committee today, alleging that U.S. Supreme Court nominee Brett Kavanaugh attempted to sexually assault her at a house party in the 1980s when both were in high school. In her testimony, Ford claimed that Kavanaugh lured her to a private bedroom, attempted to remove her clothing, pin her to a bed, and force himself onto her despite her protestations, but that she eventually escaped. The Senate Judiciary Committee is investigating Ford’s allegations, which Kavanaugh has firmly denied. Kavanaugh also questioned why the allegations had not surfaced until 36 years after the purported incident, and only just before his Supreme Court confirmation hearing.

As in the previous experiments, an additional sentence at the end of the paragraph framed the report in a way that was consistent with the headline (see Table 4), for a total of six unique reports.

Table 4.

News-Report Stimuli Used in Experiment 4

Report section	Frame
Report section	Assault victim (Ford)	Allegation victim (Kavanaugh)	Baseline
Headline	Victim of Sexual Assault Braces for Investigation (September 27, 2018)	Victim of Sexual Assault Allegations Braces for Investigation (September 27, 2018)	Nation Braces for Sexual Assault Investigation (September 27, 2018)
Framing	Reactions across the nation are polarized, but friends have rallied around Ford. “Christine is a victim of sexual assault who deserves to be believed,” said a high school friend of both. “She has been traumatized by this experience. The investigation ahead is going to be grueling, but I know she will survive it.”	Reactions across the nation are polarized, but friends have rallied around Kavanaugh. “Brett is a victim of false allegations who deserves to be believed,” said a high school friend of both. “He has been traumatized by this experience. The investigation ahead is going to be grueling, but I know he will survive it.”	Reactions across the nation are polarized, but friends have rallied around both Ford and Kavanaugh. “Christine and Brett should each be allowed to tell their side of the story,” said a high school friend of both. “This experience has been challenging for them. The investigation ahead is going to be grueling, but I know the truth will come out.”

Note: As in the previous experiments, participants read one of the three headlines, followed by either the sparse-detail or rich-detail paragraph (see text). The end of the paragraph was framed in a way that was consistent with the headline, for a total of six unique reports.

The support questions concerned the individuals in the news reports (Ford: α = .85; Kavanaugh: α = .89), and the question about whether the report reminded participants of real-life news stories was replaced with one assessing their familiarity with the case described (see the Supplemental Material). The method was otherwise identical to that used in the previous experiments.

Results

In the baseline condition, participants expressed more support for Ford (M = 3.81, SE = 0.12) than Kavanaugh (M = 2.61, SE = 0.13), t(203) = 5.04, p < .001, d = 0.35. As shown in Figure 6, the effects of victim framing were somewhat inconsistent overall. In a 3 (frame) × 2 (detail) ANOVA on support for Ford, there was a main effect of frame, F(2, 595) = 3.12, p = .045, η_p² = .01, and an interaction, F(2, 595) = 3.30, p = .04, η_p² = .01. When the report had sparse detail, support for Ford was higher in both framing conditions compared with baseline—assault victim: t(196) = 2.91, p = .004, d = 0.41; allegation victim: t(201) = 3.18, p = .002, d = 0.45. When the report had rich detail, support for Ford did not differ significantly from baseline for either framing condition (ps > .86, ds < 0.03). In the analogous ANOVA on support for Kavanaugh, there were no significant effects (ps > .16, η_p²s < .01).

Fig. 6.

Support for (a) the accuser (Ford) and (b) the alleged perpetrator (Kavanaugh) in Experiment 4, by frame and level of detail. Boxes denote the interquartile range, the middle line indicates the median, whiskers extend from the 10th to the 90th percentiles, and the plus sign denotes the mean.

A cross-experiment comparison showed that victim framing was somewhat less effective for the real event in Experiment 4 than the fictionalized vignette in Experiments 1 and 2, which also used a maximal instantiation of the frame. In a 3 (frame) × 2 (experiment: 1 and 2 vs. 4) ANOVA on support for the accuser, the interaction was significant, F(2, 2009) = 8.54, p < .001, η_p² = .01, and victim framing had a stronger effect on support for the fictionalized accuser than on support for Ford. In the analogous ANOVA on support for the alleged perpetrator, the interaction was not significant, F(2, 2009) = 1.99, p = .14, η_p² < .01. In both ANOVAs, there was also a main effect of experiment—accuser: F(1, 2009) = 9.86, p = .002, η_p² = .01; alleged perpetrator: F(1, 2009) = 8.75, p = .003, η_p² = .004—with more support overall for the fictionalized protagonists than for Ford and Kavanaugh.

Critically, however, the effects of victim framing were again moderated by whether participants cited the victim-related language as influential. In 2 (assault victim vs. allegation victim) × 2 (citers vs. nonciters) ANOVAs on support for each protagonist, the interaction was significant—Ford: F(1, 393) = 14.82, p < .001, η_p² = .04; Kavanaugh: F(1, 393) = 27.95, p < .001, η_p² = .07. As shown in Figure 7, citers (37% of participants in the framing conditions) showed the same framing effects as in Experiments 2 and 3—Ford: t(144) = 3.08, p = .002, d = 0.51; Kavanaugh: t(118) = −5.26, p < .001, d = 0.87—whereas nonciters showed framing effects in the opposite direction—Ford: t(249) = −2.57, p = .01, d = 0.33; Kavanaugh: t(249) = 2.20, p = .03, d = 0.28. As in Experiments 2 and 3, we also conducted preregistered analyses on support difference scores (support for Emma – support for David) that corroborated the main findings (see the Supplemental Material).

Fig. 7.

Support for (a) the accuser and (b) the alleged perpetrator in the assault-victim and allegation-victim conditions of Experiment 4, separately for citers and nonciters of victim-related language. Boxes denote the interquartile range, the middle line indicates the median, whiskers extend from the 10th to the 90th percentiles, and the plus sign denotes the mean. The dashed line indicates the mean in the baseline condition.

Discussion

We again observed significant victim-framing effects for participants who cited the victim-related language as influential. This is noteworthy because the report concerned a high-profile, familiar case, suggesting that victim framing has ecological validity. We also observed a reverse framing effect for nonciters, who expressed more support for Ford when Kavanaugh was labeled the victim, and vice versa. One possibility is that these participants reacted negatively to the victim frame on the basis of their prior beliefs and pointed to other aspects of the report to justify their elevated support for the non-victim-framed individual. An exploratory analysis detailed in the Supplemental Material supports this explanation: Nonciters in the allegation-victim condition were more ideologically liberal and less accepting of rape myths than those in the assault-victim condition. That is, participants inclined to back Ford from the outset may have been put off by the portrayal of Kavanaugh as the victim, leading them to redouble their support for Ford. According to this interpretation, people use social-pragmatic reasoning to infer what the frame is communicating; when this inference runs counter to strongly held convictions, they reject it and victim framing backfires.

General Discussion

We investigated the consequences of victim framing across four experiments. Participants read about an alleged sexual assault that framed the accuser as the victim (of assault), framed the alleged perpetrator as the victim (of false allegations), or was neutral about victimhood (baseline). Relative to the baseline condition, participants in the two framing conditions generally expressed more support for the victim-framed protagonist and less support for the other protagonist. The strength and consistency of these effects varied with how often the victim frame was instantiated in the report and whether the case was fictionalized or real. Across all contexts, however, participants who explicitly cited the victim-related language from the report as influencing their evaluations exhibited strong victim-framing effects.

These results provide the first empirical demonstration that victim framing can shape public opinion. Whereas previous work suggests that drawing attention to alleged perpetrators increases support for their victims (Niemi & Young, 2016), labeling an alleged perpetrator a victim appears to have the opposite effect, garnering support for the perpetrator and reducing support for the accuser. That said, we also replicated a range of findings from the victim-blaming literature, which suggests that our experiments elicited participants’ genuine attitudes toward the protagonists. An exploratory cross-experiment analysis, detailed in the Supplemental Material, revealed that participants who reported more personal experience with sexual assault and less acceptance of rape myths, as well as female and more liberal participants, expressed more support for the accuser and less support for the alleged perpetrator. Similar observer characteristics have previously been linked to support for assault victims and perpetrators (Grubb & Turner, 2012; Hayes et al., 2013; Nagel et al., 2005; Niemi & Young, 2016; Suarez & Gadalla, 2010). Notably, the effects of victim framing remained significant when we controlled for demographic factors and were not moderated by those factors. This suggests that victim framing and observer characteristics contribute independently to people’s evaluations of alleged perpetrators and their accusers.

The effects of victim framing are broadly consistent with a dyadic account of moral reasoning. This account posits that people evaluate moral violations by comparing the situation at hand with the prototypical schema of an immoral act: an intentional agent inflicting harm on a vulnerable patient (Gray & Wegner, 2011; Schein & Gray, 2018). After someone has been judged a moral agent or patient, they become typecast in that role, and patients are seen as less capable of committing immoral acts such as assault (Gray & Wegner, 2009). Victim framing may nudge people to slot whomever is labeled a victim into the patient role, resulting in increased empathy and reduced blame attributions.

Our findings also implicate social-pragmatic reasoning, as we found reliable framing effects only when participants indicated that the victim-related language influenced their evaluations. Observers draw inferences about speakers’ communicative intentions in context, trusting that specific words and phrases were chosen to be informative (Goodman & Frank, 2016; Grice, 1975; Sperber & Wilson, 1986). Even “logically equivalent” phrases—such as saying that a basketball player makes 40% versus misses 60% of their shots—convey different information about the speaker’s perspective that observers readily infer (Sher & McKenzie, 2006). Such inferences seem to underlie associated framing effects—for example, judging that a player who misses 60% of their shots is worse than a player who makes 40% (Leong et al., 2017). In this case, observers infer that the miss frame was chosen to communicate a negative evaluation of the player, and their own judgments therefore shift in that direction.

Victim framing may work in a similar way. Participants who cited the victim-related language as influential inferred that the author of the report chose to describe one individual as a victim for good reason—to signal that they deserve support—and would have used similar language to describe the other individual if warranted. Participants who failed to make this pragmatic inference, weighed other information more highly, or were motivated to counteract the frame cited other aspects of the report and exhibited no (or reverse) framing effects. Importantly, the contrast between citers and nonciters is not merely a product of attentional differences. The two groups spent roughly the same amount of time reading the reports, and the rationales of nonciters typically included relevant factual information about the case (see the Supplemental Material). Whereas most explanations of framing effects invoke basic cognitive processes such as heuristics or analogy (e.g., Thibodeau & Boroditsky, 2011; Tversky & Kahneman, 1981), our findings highlight the critical role of social-pragmatic reasoning.

There is a curious connection between the social-pragmatic account and the concept of experimenter demand in psychological research. Both explanations propose that participants use the victim frame to infer how much support each protagonist deserves. The demand account goes a step further, positing that participants also infer the researcher’s hypothesis from the choice of frame and suppress their genuine attitudes to accede to it. Disentangling these accounts is difficult in framing studies, and most research does not directly address the issue. Our experiments contained relatively few demand characteristics, however. We used a between-subjects design and provided a plausible cover story—that we were interested in how people consume news reports. This cover story was especially strong in Experiment 4 because online studies often invite participants to express attitudes toward salient sociopolitical events. That said, sometimes experimenter demand can reflect real-world processes of attitude change. Much of what people think and do is a result of social influences, and our experimental context captured how people adjust to the intentions, beliefs, and norms communicated by others—especially in ambiguous situations. Victim framing may work in part by eliciting a kind of conformity to what is communicated, although further research is needed to fully tease out the underlying mechanisms.

In sum, we have shown that victim framing in the context of sexual assault is an effective rhetorical strategy, leading to increased support for whomever is labeled a victim, whether they are the alleged perpetrator or the accuser. The power of victim framing comes from using social-pragmatic reasoning to draw inferences from the framing language, but this can backfire when prior convictions motivate people to resist what they infer. An important direction for future research is whether these findings generalize beyond gender-stereotypical assault scenarios in an American context to cases involving individuals with marginalized identities and domains other than sexual assault. Nevertheless, we have shown that the language of victimhood—or its strategic deployment to cast alleged perpetrators in a more favorable light—can shape attitudes toward sexual assault in line with the ideological interests of the communicator, despite the polarized nature of current public discourse. These findings serve as a cautionary warning to journalists, lawyers, victim advocates, jurors, and the general public about the consequences of how we communicate about sexual assault.

Supplemental Material

sj-docx-1-pss-10.1177_09567976211045935 – Supplemental material for Who’s the “Real” Victim? How Victim Framing Shapes Attitudes Toward Sexual Assault

Supplemental material, sj-docx-1-pss-10.1177_09567976211045935 for Who’s the “Real” Victim? How Victim Framing Shapes Attitudes Toward Sexual Assault by Stephen J. Flusberg, James van der Vord, Sarah Q. Husney and Kevin J. Holmes in Psychological Science

Footnotes

Acknowledgements

We thank Paul Thibodeau and Nan Elpers for helpful feedback on earlier drafts of this article. We also thank Casey Pollard and Olivia Schultz for their help in developing some of the initial ideas explored in this article. Experiment 1 was presented at the 31st Annual Convention of the Association for Psychological Science and the 41st Annual Conference of the Cognitive Science Society, and the abstracts are archived in the conference proceedings.

Transparency

Action Editor: Mark Brandt

Editor: Patricia J. Bauer

Author Contributions

All the authors developed the study concept and designs. K. J. Holmes, J. van der Vord, and S. Q. Husney collected the data. K. J. Holmes analyzed and interpreted the data in consultation with S. J. Flusberg. S. J. Flusberg and K. J. Holmes drafted the manuscript, and J. van der Vord and S. Q. Husney provided critical feedback. All the authors approved the final manuscript for submission.

Open Practices

Data and materials for all four experiments have been made publicly available via OSF and can be accessed at https://osf.io/ks5tr/ and https://osf.io/ydf2u/, respectively. The design and analysis plans for Experiments 2 to 4 were preregistered on AsPredicted; copies are available at https://osf.io/4jgsz/. All preregistered analyses are described in the Supplemental Material available online. This article has received the badges for Open Data, Open Materials, and Preregistration. More information about the Open Practices badges can be found at .

Supplemental Material

Additional supporting information can be found at

Notes

References

Buhrmester

Kwang

Gosling

S. D.

(2011). Amazon’s Mechanical Turk: A new source of inexpensive, yet high-quality, data? Perspectives on Psychological Science, 6(1), 3–5.

Burt

M. R.

(1980). Cultural myths and support for rape. Journal of Personality and Social Psychology, 38(2), 217–230.

Chumley

C. K.

(2018, September 21). Christine Blasey Ford is not the victim here—Brett Kavanaugh is. The Washington Times. https://www.washingtontimes.com/news/2018/sep/21/christine-blasey-ford-not-victim-here-brett-kavana/

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191.

Flusberg

S. J.

Lauria

Balko

Thibodeau

P. H.

(2020). Effects of communication modality and speaker identity on metaphor framing. Metaphor and Symbol, 35(2), 136–152.

Goodman

N. D.

Frank

M. C.

(2016). Pragmatic language interpretation as probabilistic inference. Trends in Cognitive Sciences, 20(11), 818–829.

Gray

Wegner

D. M.

(2009). Moral typecasting: Divergent perceptions of moral agents and moral patients. Journal of Personality and Social Psychology, 96(3), 505–520.

Gray

Wegner

D. M.

(2011). Morality takes two: Dyadic morality and mind perception. In Mikulincer

Shaver

P. R.

(Eds.), The social psychology of morality: Exploring the causes of good and evil (pp. 109–127). American Psychological Association.

Grice

H. P.

(1975). Logic and conversation. In Cole

Morgan

(Eds.), Syntax and semantics (Vol. 3, pp. 41–58). Academic Press.

10.

Grubb

Turner

(2012). Attribution of blame in rape cases: A review of the impact of rape myth acceptance, gender role conformity and substance use on victim blaming. Aggression and Violent Behavior, 17(5), 443–452.

11.

Hayes

R. M.

Lorenz

Bell

K. A.

(2013). Victim blaming others: Rape myth acceptance and the just world belief. Feminist Criminology, 8(3), 202–220.

12.

Hesse

(2019, July 16). For Robert Foster’s defenders, men are victims and #MeToo is the real villain. The Washington Post. https://www.washingtonpost.com/lifestyle/style/for-robert-fosters-defenders-men-are-victims-and-metoo-is-the-real-villain/2019/07/15/e17e7836-a4d4-11e9-b8c8-75dae2607e60_story.html

13.

Holmes

K. J.

Doherty

E. M.

Flusberg

S. J.

(2021). How and when does syntax perpetuate stereotypes? Probing the framing effects of subject-complement statements of equality. Thinking & Reasoning. Advance online publication. https://doi.org/10.1080/13546783.2021.1963841

14.

Leong

L. M.

McKenzie

C. R.

Sher

Müller-Trede

(2017). The role of inference in attribute framing effects. Journal of Behavioral Decision Making, 30(5), 1147–1156.

15.

Lisak

Gardinier

Nicksa

S. C.

Cote

A. M.

(2010). False allegations of sexual assault: An analysis of ten years of reported cases. Violence Against Women, 16(12), 1318–1334.

16.

Litman

Robinson

Abberbock

(2017). TurkPrime.com: A versatile crowdsourcing data acquisition platform for the behavioral sciences. Behavior Research Methods, 49(2), 433–442.

17.

Loftus

E. F.

Palmer

J. C.

(1974). Reconstruction of automobile destruction: An example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behavior, 13(5), 585–589.

18.

McMahon

Farmer

G. L.

(2011). An updated measure for assessing subtle rape myths. Social Work Research, 35(2), 71–81.

19.

Nagel

Matsuo

McIntyre

K. P.

Morrison

(2005). Attitudes toward victims of rape: Effects of gender, race, religion, and social class. Journal of Interpersonal Violence, 20(6), 725–737.

20.

Niemi

(2017). Victim blaming. In Nadal

K. L.

(Ed.), The SAGE encyclopedia of psychology and gender (pp. 1756–1757). SAGE.

21.

Niemi

Young

(2016). When and why we see victims as responsible: The impact of ideology on attitudes toward victims. Personality and Social Psychology Bulletin, 42(9), 1227–1242.

22.

Open Science Collaboration. (2015). Estimating the reproducibility of psychological science. Science, 349(6251), Article aac4716. https://doi.org/10.1126/science.aac4716

23.

Schein

Gray

(2018). The theory of dyadic morality: Reinventing moral judgment by redefining harm. Personality and Social Psychology Review, 22(1), 32–70.

24.

Schultz

Flusberg

S. J.

Holmes

K. J.

(2020, June). Victim or survivor? How everyday language frames sexual assault [Poster presentation]. Society for the Psychological Study of Social Issues Summer Conference, Denver, CO, United States. (Conference canceled)

25.

Sher

McKenzie

(2006). Information leakage from logically equivalent frames. Cognition, 101(3), 467–494.

26.

Sperber

Wilson

(1986). Relevance: Communication and cognition (Vol. 142). Harvard University Press.

27.

Suarez

Gadalla

T. M.

(2010). Stop blaming the victim: A meta-analysis on rape myths. Journal of Interpersonal Violence, 25(11), 2010–2035.

28.

Thibodeau

P. H.

(2016). Extended metaphors are the home runs of persuasion: Don’t fumble the phrase. Metaphor and Symbol, 31(2), 53–72.

29.

Thibodeau

P. H.

Boroditsky

(2011). Metaphors we think with: The role of metaphor in reasoning. PLOS ONE, 6(2), Article e16782. https://doi.org/10.1371/journal.pone.0016782

30.

Tversky

Kahneman

(1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453–458.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.05 MB