Data are drawn from an online survey about emerging adults’ sexual violence experiences administered in 2017 (procedures approved by the Institutional Review Board at the University of Michigan). We used Facebook and Instagram advertisements targeted to 18- to 25-year-olds nationally in the United States to recruit individuals to complete a 15 to 20-minute survey containing items described below. Following survey completion, participants could provide contact information and be entered to win one of four $75 Amazon.com gift cards. Specific details of recruitment are described elsewhere (see Bonar et al., 2021).

Of the 6,115 participants who initiated the survey, 4,010 completed the entire survey. Of the 4,010 participants, 2,848 (71.02%) indicated they had seen sexual content posted on social media by a friend and thus were asked questions related to online rape myths; of the 2,848, 91.6% (n = 2,609) had complete covariate data and were thus included in the present study.

We used descriptive statistics to review variable distribution. Then, we randomly split the sample into two groups (a seed was set for reproducibility): a model training group and a testing group. Exploratory factor analysis was conducted with the training group. The exploratory factor analysis employed a Geomin oblique rotation, as this type of rotation provides correlations similar to those of the confirmatory factor analysis without having to specify specific loading structures and is the common rotation type for exploratory factor analyses (Asparouhov & Muthén, 2009; Schmitt & Sass, 2011). The number of factors selected was based on eigenvalues and a scree plot “elbow,” fit statistics of the model (RMSEA, CFI, TLI, SRMR), review of item-factor loading and theoretical interpretation (Furr & Bacharach, 2013). Once the factor solution was selected, a confirmatory analysis was conducted with the testing group. The fit statistics of RMSEA, CFI, TLI, and SRMR were used to evaluate confirmatory model fit (Hu & Bentler, 1999); acceptable cutoffs are values < 0.06 for RMSEA; >0.95 for CFI and TLI; and < 0.08 for SRMR. In a final set of analyses, a mean score across the nine rape myth items was calculated (ranging from 0 to 1) to represent how many types of rape myths were seen in online spaces. Finally, we produced a correlation table to evaluate bivariate association between covariates and rape myths and used a multivariable OLS regression model to evaluate the association between the mean of rape myth exposure score and covariates.

Of the 4,010 individuals who completed the item assessing frequency with which they had seen content posted by friends on social media about sexual assault, 71.02% (n = 2,848) reported having seen such content. Of those 2,848, 91.6% (n = 2,609) provided complete data for the online rape myth items and covariates; 237 did not answer any further questions about the rape myths they saw online, and 2 did not answer information about cannabis, opioids, and/or illicit drugs. Of the 2,609 participants, mean age was 20.9 years (SD = 2.2 years), just under half were biologically male (n = 1,202, 46.1%), and most were White (71.6%, n = 1,868; Table 1). Most participants identified their gender as either man (n = 1,169, 44.8%) or woman (n = 1,284, 49.2%); 27 (1.0%) of participants identified as a trans man, 4 (0.2%) identified as a trans woman, 116 (2.2%) identified as genderqueer, and 75 (1.4%) identified as a gender not listed, such as “agender,” “non-binary,” or “gender fluid.”

On average, participants reported having between four and five social media accounts (M = 4.16, SD = 1.25). The rape myth item with the greatest endorsement was seeing content reflecting that “rape accusations are often used as a way of getting back at someone,” which 31.9% (n = 832) of the sample endorsed. More than a quarter of participants (25.5%, n = 666) reported exposure to content reflecting that “people don’t usually tend to force sex on others, but sometimes they get too sexually carried away.” The fewest participants reported that friends posted content supporting the idea that “people from nice middle-class homes almost never rape” (n = 114, 4.4%). When evaluated as a summary score, on average, participants reported seeing one or two rape myths posted to social media by friends (M = 0.15, SD = 0.20).

More than one-third of study participants reported rape victimization (n = 902, 34.6%), though fewer reported attempted rape perpetration (n = 171, 6.6%) or perpetrating a sexual assault in a situation where the individual “misread” the situation (n = 141, 5.4%). On average, most participants reported less than monthly risky drinking (defined as five or more drinks in 1 day for males, and four or more drinks in 1 day for females; M = 1.20, SD = 1.08). On average, cannabis was used monthly or less than monthly (M = 1.38, SD = 1.52), and illicit drug use (e.g., cocaine, heroin, ecstasy, mushrooms; M = 0.28, SD = 0.61), and prescription drug misuse (M = 0.26, SD = 0.65), were less common. All descriptive statistics are located in Table 2.

First, we examined eigenvalues for possible solutions (1-factor: 5.509; 2-factor: 0.803; 3-factor: 0.704; 4-factor: 0.482) and examined the scree plot. Though the “elbow” of the graph was around 2, the values suggested a 1-factor solution. As these were slightly conflicting, we compared the fit statistics and interpretability of a 1-factor and 2-factor model before choosing the final solution. The fit statistics for both the 1-factor (RMSEA: 0.03; CFI: 0.99; TLI: 0.99; SRMR: 0.057) and the 2-factor solution (RMSEA: 0.014; CFI: 0.99; TLI: 0.99; SRMR: 0.036) were both acceptable given the cutoff scores suggested by Hu and Bentler (1999), though the 2-factor solution was slightly better fitting. The selection of the 1-factor solution was supported by the loading scores of each item onto the 1-factor solution, the acceptability of fit, and the ease of interpretation of a parsimonious model given existing theory. Thus, we selected a 1-factor solution (Table 3).

We next tested the 1-factor solution using a confirmatory factor analysis in the other half of participants (testing group). The fit statistics of the 1-factor solution in the testing group suggested a good fitting model (RMSEA: 0.021; CFI: 0.99; TLI: 0.99; SRMR: 0.038). Cronbach alpha of the nine items within the testing group was .77.

Our final set of results is to evaluate the associations between covariates and electronic rape myth exposure. First, we used a correlation table to examine bivariate correlation between covariates and electronic myth exposure score (M = 0.15, SD = 0.20, range 0–1; Table 4). Then, we used multivariable linear regression to evaluate the association between the covariates and electronic rape myth exposure (R² = .029; Table 5). Greater exposure to rape myths posted by friends on social media sites was associated with greater attempted rape perpetration (β = .052, SE = .016, p< .005), rape victimization (β = .045, SE = .009, p< .005), and illicit drug use (β = .021, SE = .008, p < .05). Those with greater exposure were more likely to be male (β = .017, SE = .008, p < .05) and younger in age (β = −0.008, SE = .002, p < .005).

Sexual assault is a widespread problem in the United States which disproportionately occurs during adolescence and early adulthood (Basile et al., 2016), yet little work has been done to examine rape myths within the online space. In the present study, more than two-thirds of the broader study sample (n = 4,010) of emerging adults surveyed reported seeing discussion of content promoting rape myths/sexual assault on social media. This proportion is particularly concerning given that greater endorsement of rape myths often decreases individual perception of the responsibility of actions (e.g., one or more parties were intoxicated, someone was wearing tight clothing that shows certain body parts), which has harmful implications, including an increased risk of sexual assault perpetration (Fansher & Zedaker, 2022).

Beyond examining the prevalence of the broad topic of online rape myths, it is important to gather nuanced information on the specific forms and classifications of violence expressions to inform future interventions (Lewis et al., 2017). Our data suggest that among emerging adults who see sexual assault content posted by friends, the most common expressions of rape myths are those which suggest that sexualized violence acts are used as a method of revenge (reported by one in three); that rape is the result of being sexually carried away (reported by one in four); that rape is the consequence of wearing revealing clothes (reported by one in five); that rape often occurs because the person being violated was not sufficiently clear with their “no” (reported by one in six); and that the person who is violated has some responsibility for a rape if they were intoxicated (reported by one in six). Rape myths are prevalent, but the several specific myths that are most common could serve as focal points for future education campaigns. Addressing exposure to rape myths in the online space is important for three reasons. The first is given the propensity for victim-blaming-type posts to spread quickly on social media platforms; the second is that the vast majority of emerging adults use social media to socialize, for entertainment, and to gain important news/health-related information; and the third is that this age group appears to be particularly swayed by the information they see online (Auxier & Anderson, 2021; Bialik, 2018; Mitchell et al, 2021; Perrin, 2020; Shearer, 2021).

One of the major challenges in the field of online sexual violence and intimate partner violence research is a lack of coherence across measures (Caridade et al., 2019). Having multiple measures reflecting different conceptualizations or definitions of a given phenomenon creates confusion and a lack of consensus on the prevalence of a problem and can be particularly problematic when evaluating changes over time. In an effort to not replicate this challenge in the measurement of online rape myths, we examined the psychometric properties of our adapted rape myth scale specific to social media postings by friends. Using EFA, we found preliminary support that the nine items were psychometrically related and can be combined to form a scale. We then tested the structure using CFA and found additional support for a 1-factor solution. By splitting the sample in half to build independent training and testing sub samples, we were able to increase validity of our findings and potential reproducibility. We also acknowledge the potential limitations of this new measure which could benefit from refinement and/or extension. Specifically, given the constraints of our parent study with respect to participant burden, we were limited in the number of items from the IRMA we could adapt and include, and it may be that other items better or more completely measure this construct. Nonetheless, we believe this measure provides a useful starting point for future work to accurately characterize online rape myth exposure. For example, researchers could extend this area by using a content analysis of the available social media data to help articulate the various rape myths that may be missing from the current measure. Researchers could also use focus groups and/or think aloud procedures with emerging adults to better understand how they interpret the items presented.

In the final step of our analysis, we examined the associations between greater exposure to online rape myths posted by friends on social media platforms (using the online rape myth scale) with sexual violence experiences. Our findings that online rape myth exposure is associated with greater attempted rape perpetration and victimization builds on past research regarding rape myths and sexual behaviors which were not specific to the online space (Fansher & Zedaker, 2022). Further, results showing that exposure to friends’ posts endorsing rape myths is significantly associated with one’s own actions extends research on the clustering of beliefs within social networks (Banyard et al., 2022; Brechwald & Prinstein, 2011). While our results support and extend prior work examining rape myth beliefs outside of online spaces, we are unable to draw conclusions regarding directionality; an alternative interpretation of our findings could be that those with exposure to in-person sexual violence are more aware of rape-related material in online spaces, and thus have more precise recall of this type of content. This alternative explanation highlights the need for future prospective work to evaluate temporal directionality and causality. This would hold particularly important implications for ongoing and future social media reform to better manage exposure to certain content/misinformation and actively work toward dispelling rape myths (Hedrick, 2021).

Beyond the correlation of exposure to rape myths and experiences of sexual violence, we also considered correlations with sociodemographic variables and risk behaviors. Younger age and male sex were associated with greater online exposure to rape myths, which are consistent with prior research on rape myths that are not specific to online spaces (Beshers & DiVita, 2021). Importantly, the present study psychometrically tested items which did not use gender or sex-specific language to create a scale on rape myths within the online space. This focus is critical given that the majority of the literature on rape and sexual assault is focused on heterosexual relationships, and on male perpetrators and female victims (Boyle & Rogers, 2020). While biologically female bodies are at high risk of sexual assault (Basile et al., 2016), implying perfect correlation of sex and gender excludes individuals on the basis of biological differences (e.g., intersex individuals) and social identities (e.g., trans, non-binary gender), which do not fit into hegemonic binary sex and gender categories.

Finally, our findings regarding substance use and rape myths were somewhat surprising. Prior work has found a positive association between alcohol use and rape myths (Dworkin et al., 2017; Navarro & Tewksbury, 2017), lower perceived capability of perpetrators and greater responsibility of victims when intoxicated (Grubb & Turner, 2012), and the co-occurrence of alcohol use and sexual violence (Anderson et al., 2017). Yet, we did not find a significant association between alcohol use and online exposure to rape myths. Additionally, cannabis use was not associated with rape myths. These findings may be due in part to differences in the normativity of alcohol and cannabis use relative to rape myths. Although alcohol and cannabis use are relatively prevalent behaviors in emerging adulthood (Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration, 2021), in line with other researcher findings, it could be that increased societal knowledge and awareness around rape and culpability have lowered the social normativity of these beliefs (Beshers & DiVita, 2021), pushing rape myths from a social norm into a set of antisocial beliefs. This could help explain our finding of association between illicit drug use and greater rape myth exposure. Youth who use illicit drugs are at higher risk of engaging in a broader range of antisocial behaviors (Fitzsimons & Villadsen, 2021) and are more likely to be part of friendship networks with individuals who also display a broader range of antisocial behaviors (Sijtsema & Lindenberg, 2018), which could include rape myths.

Several limitations require mentioning. Although we adapted items from a standard measure, reviewed the items for face validity, and psychometrically tested the items to understand their interrelatedness, replication, and potential expansion, other samples, is required. As noted earlier, due to the cross-sectional design, we are unable to draw conclusions about temporal associations. The study design of the current project did not purposively collect data from non-binary identifying individuals and thus does not allow the current piece to explicitly examine rape myths relative to gender identities; however, we strove to not remove individuals who do not identify within the gender binary from our analyses and thus chose to use biological sex. Future work on online rape myths which uses a purposive sampling framework to represent multiple forms of gender (beyond the binary man/woman) is an important next step. Finally, our measure of exposure relied on individual recall for the prior 6 months; although prior research shows a positive correlation between recall and observed measures of social media use (e.g., what is seen on social media; Parry et al., 2021), future research should include objective variables (e.g., scraping social media data) of exposure to rape myths and explore validity of reported recall relative to duration of time. Despite limitations, our findings using a large data set of emerging adults set the stage for future investigations and begin to quantify the ways in which exposure to rape myths occurs in the online world. These findings could inform future peer network interventions as well as social media messaging campaigns to address and disrupt these harmful belief systems.