Abstract
Rates of sexual victimization have remained steady over several decades, and preventative interventions to reduce men’s sexually aggressive behavior have been largely ineffective. As such, research has endeavored to find novel approaches to identify women at increased risk for sexual victimization. Sexual assault scripts, or “cognitive models” that women adhere to that guide their beliefs about sexual assault are posited to influence their victimization risk. Prior studies on sexual assault scripts primarily have been qualitative in nature; however, recent work yielded a 27-item measure of putative risk for sexual victimization called the Sexual Assault Script Scale (SASS). The SASS has four subscales called Stereotypical Assault Scripts, Acquaintance Assault Scripts, Assault Resistance Scripts, and Date/Friend Assault Scripts, which were found in prior work to be internally consistent and associated with putative risk factors for sexual victimization. The focus of the current study was to test the measurement invariance of the SASS among Hispanic and non-Hispanic White college women who were recruited in the prior study. Four hundred sixty-nine (N = 469) Hispanic and 415 non-Hispanic White US undergraduate heterosexual or bisexual women from a Southwestern university in the United States completed the SASS. Confirmatory factor analysis (CFA) replicated the prior four-factor model with an acceptable fit to the data, and tests of measurement invariance revealed the SASS to be invariant across Hispanic and non-Hispanic White college women, suggesting that the SASS is measuring a similar construct in these groups.
In spite of the development of numerous preventative interventions aimed at decreasing the prevalence of sexual violence on college campuses in recent years, rates of victimization have remained stable. Moreover, these interventions, while reasonably effective at changing attitudes, have been relatively ineffective at changing behavior (Anderson & Whiston, 2005; Ellsberg et al., 2015). Additionally, very few programs have been successful in reducing rates of assault among college women (e.g., Marx et al., 2001; Rowe et al., 2012; Senn et al., 2015), and even less effective at reducing incidents among women who are at high risk for sexual victimization (i.e., previously victimized women). As such, research has explored a variety of approaches to identify and intervene with women who are at heightened victimization risk. The intention of these strategies is not to blame women for their assault; indeed, perpetrators, who are commonly men, are entirely responsible for their acts of sexual aggression. Yet due to the paucity of effective preventive interventions to reduce men’s sexual aggression, a focus on women’s risk factors remains necessary at present.
One promising avenue of work has been on examining women’s scripts of sexual interactions, including unwanted sexual experiences. Sexual scripts are “cognitive models” (Rose & Frieze, 1993, p. 499) that are posited to influence how people evaluate and guide their social and sexual interactions. These scripts also are thought to influence women’s risk of having negative sexual experiences, including sexual assault (Kahn & Mathie, 2000). Notably, women often report perceiving their sexual assaults as an event other than rape (Bondurant, 2001; Kahn et al., 1994) and exhibit also a tendency to think of rape as something involving a stranger who is physically violent (Kahn et al., 1994). These “blitz rapes” are quite infrequent in the United States; indeed, most rapes are perpetrated by an acquaintance or date (Bondurant, 2001; Kahn et al., 1994), who use low levels of physical force and no weapon (Littleton & Breitkopf, 2006). In addition, most rapes are known to occur inside and involve alcohol use (Littleton & Breitkopf, 2006). Consequently, researchers have posited that a “cognitive mismatch” between what typically occurs during a sexual assault and a stereotypical conceptualization of this event may place women at risk for sexual victimization by a partner, acquaintance, or friend (Kahn et al., 1994; Kahn & Mathie, 2000).
While most work in this area has been qualitative and cross-sectional in nature, several prospective studies have found that certain script features (e.g., outdoor location, nonforceful resistance, unfamiliar perpetrator, alcohol and drug use, acceptance of sexual aggression) predicted sexual assault during a follow-up period (Krahe et al., 2007; Turchik et al., 2009). Thus, scripts that are discordant with the common features of sexual assault (e.g., outdoor location), as well as scripts that “normalize” high-risk behaviors (e.g., alcohol and drug use) are thought to interfere with women’s capacity to identify and respond to high-risk situations. In the broader script theory literature, scripts also are posited to influence both attention to and memory for the contextual features of social situations (Baldwin, 1992; Schank & Abelson, 1995). In sum, this work suggests that women’s cognitive processing about what constitutes a sexual assault may be relevant in understanding their risk for experiencing sexual violence.
Qualitative work on sexual assault scripts has produced a plethora of rich and informative data, yet analyzing these data remains a tedious and time-consuming approach for identifying script features that may increase women’s victimization risk. In an effort to facilitate a more pragmatic and efficient measurement process, Yeater et al. (2019a) developed a quantitative measure of sexual assault scripts called the Sexual Assault Script Scale (SASS). To develop the SASS, Yeater et al. (2009a) conducted three separate studies using mixed-methods methodology with 1,031 undergraduate women recruited from the psychology subject pool at a medium-sized Southwestern university (see Yeater et al., 2019a for an in-depth presentation and discussion of the development of the SASS).
In Study 1, a group of undergraduate women was interviewed and asked to imagine a situation in which a man was verbally or physically coercing them into a sexual experience. Women were instructed to describe a hypothetical situation—not an incident that they had experienced in the past—and were asked to imagine what was happening in the situation. The interviewer probed for additional contextual information as needed (e.g., Who is the man? How long have you known him? What kind of place and situation will you be in? What types of sexual activity, if any, will occur?). Once saturation was obtained (i.e., novel information was no longer being obtained from interviews), no new participants were recruited for the study. All interviews were coded and analyzed using a qualitative data analysis software program (i.e., NVivo). The coded data and qualitative data analysis results were used to create phrase codes that then were used to develop 74 Likert-type items for the SASS. Items were written such that women were asked “how likely” they believed various contextual and interpersonal features would be present during a hypothetical sexual assault (e.g., How likely is it that you have never seen him before? How likely is it that both you and the man have been drinking?).
In Study 2, 500 undergraduate women were recruited and asked to complete the SASS (among other measures). On the SASS, participants were asked: “When considering the following statements, please imagine that you are in a situation in which you are being coerced by a man into a sexual experience. Please indicate the extent to which you believe each of the following would be happening during this experience.” For each item, participants indicated on a 5-point Likert-type scale, from 1 (not at all likely) to 5 (completely likely), the likelihood of each statement. Exploratory factor analysis was used to examine the data, and four factors emerged that were labelled: Stereotypical Assault Scripts (17 items; e.g., “How likely is it that you are walking somewhere outside when the man approaches you?”), Acquaintance Assault Scripts (10 items; e.g., “How likely is it that both you and the man have been drinking?”), Assault Resistance Scripts (6 items; e.g., “How likely is it that you fight back physically?”), and Date/Friend Assault Scripts (7 items; e.g., “How likely is it that you two are just spending time together?”; Yeater et al., 2019a).
In Study 3, a separate sample of 500 undergraduate women was recruited, and the procedure was the same as that in Study 2. Exploratory structural equation modeling (ESEM) was used to replicate the model found previously through the EFA. Poor loading items were dropped from the model, and the model re-estimated using confirmatory factor analysis (CFA). These steps resulted in a final measure with 27 items that mapped onto the original factors—Stereotypical Assault Scripts (13 items). Acquaintance Assault Scripts (4 items), Assault Resistance Scripts (6 items), and Date/Friend Assault Scripts (4 items).
The SASS subscales were found to be internally consistent across Studies 2 and 3 (Yeater et al., 2019a). Notably, these studies also found support for the criterion validity of the SASS, as the subscales were associated significantly with measures of putative risk for sexual victimization, including a prior victimization history, alcohol risk, number of sexual partners, and sociosexuality (Yeater et al., 2019a).
The initial psychometric findings for the SASS were encouraging, yet additional work is needed to further establish the validity of the measure. Thus, the focus of the current study was to test the measurement invariance of the SASS in a sample of Hispanic and non-Hispanic White college women. Examining the measurement invariance of the SASS is important, as if the measure is found to be invariant across groups of college women, users can assume that the same psychological construct is being measured in these groups. Sexual scripts are posited to be culturally specific (Frith & Kitzinger, 2001); thus, work that focuses on the influence of demographic variables on these scripts is relevant. We chose to test measurement invariance across Hispanic and non-Hispanic White college women, as the Hispanic population in the United States is large and growing rapidly (Passel et al., 2011), and our university is located in a region of the US that has a large, self-identified Hispanic population of undergraduate students. In addition, college enrollment of racial/ethnic minorities continues to increase (KewalRamani et al., 2007), which highlights the importance of ensuring that our measures are equally applicable across populations. Given many college students will identify as Hispanic, it is imperative to examine whether the SASS is tapping the same construct across Hispanic and non-Hispanic White college women.
With respect to specific predictions, we expected that the SASS would be invariant across Hispanic and non-Hispanic White groups. Our prediction was supported from prior work at our university that found that Hispanic and non-Hispanic White women reported similar acculturation experiences, such as those pertaining to mainstream comfort and social affiliation (Crawford et al., 2017). Given the extent of acculturation reported by our Hispanic college women, we expected that their scripts pertaining to sexual assault would be similar to those of their non-Hispanic White counterparts. With respect to what that meant for our outcomes, we expected that (a) the underlying factor model of the SASS would fit the data well in both groups (i.e., configural invariance); (b) the factor loadings in the model would be equivalent across groups (i.e., metric invariance); and (c) the thresholds in the model would be equivalent across groups (i.e., scalar invariance). That is, we expected that there would be no differences between the configural invariance model, the metric invariance model, and the scalar invariance model in our analyses.
Methods
Participants
As part of a larger study (Yeater et al., 2019a), 1,000 undergraduate women attending a medium-sized Southwest university were recruited through a massive email recruitment effort. Eligible participants were between the ages of 18 and 24, since women in this age range are at highest risk for sexual victimization (Bureau of Justice Statistics, 2013).
While the original sample was racially/ethnically diverse, with 423 (42.3%) identifying as non-Hispanic White, we had adequate power only for testing measurement invariance among Hispanic and non-Hispanic White participants. Thus, the datasets of women who identified as a racial group other than non-Hispanic White and who did not endorse a Hispanic ethnicity were eliminated from the analysis. Additionally, women who identified as lesbian (n = 16) were not included in the analyses, because of the small sample size of women who self-identified as such, and because we were interested first in examining sexual scripts among women who were interested in dating men. Of the 1,000 participants, data from 884 participants were used for the current analyses. Four hundred sixty-nine (53.1%) self-identified as Hispanic and 415 (46.9%) self-identified as non-Hispanic White. Most Hispanic participants indicated their race was White (n = 306, 65.2%) or Other (n = 124, 26.4%); however some identified as American Indian/Alaskan Native (n = 14, 3.0%), African American (n = 13, 2.8%), or Asian (n = 12, 2.6%). The mean age for this sample was 20.7 years (SD = 1.62). With respect to sexual orientation, 784 (88.7%) participants were identified as heterosexual and 100 (11.3%) were identified as bisexual. A majority of participants were single (n = 749; 84.7%) and had a mean of 2.83 (SD = 1.30) years of completed college.
Measures
Demographic questionnaire.
This self-report measure gathered participants’ age, marital status, race and ethnicity, and academic status.
Sexual Assault Script Scale (SASS; Yeater et al., 2019a).
The SASS is a putative measure of sexual assault risk that includes 27 items and assesses the extent to which participants believe certain contextual and interpersonal features are likely to be present during a hypothetical sexual assault (see description above and Yeater et al. (2019a) for additional information on the development of the SASS). For each item, participants indicate on a 5-point Likert-type scale (not at all likely to completely likely) the likelihood of each statement. Examples include “How likely is it that you try to get away but fail to do so?” and “How likely is it that you scream?” The subscales were internally consistent across two separate studies (i.e., Stereotypical Assault Scripts subscale, .89 and .89, Acquaintance Assault Scripts subscale, .82 and .80, the Assault Resistance Scripts subscale, .82 and .79, and Date/Friend Assault Scripts subscale, .71 and .66). Scores for all four subscales are summed with higher scores representing greater endorsement of those scripts. The subscales also were found to be associated significantly with putative risk factors for sexual victimization; specifically, prior sexual victimization, alcohol problems, number of sexual partners, and sexual attitudes, supporting the criterion validity of the SASS (Yeater et al., 2019a).
Procedure
The study was approved by our university’s Institutional Review Board. Eligible participants were sent an email that described the study as one focused on women’s expectations of dating situations and sexual experiences. Interested students who wished to participate read the consent form online, indicated their consent to participate by entering the study website, and, upon entry, completed the SASS and the other measures given in the Yeater et al. (2019a) study. Completion of the measures took approximately 20 minutes. Participants were provided with a link to a secondary survey that collected contact information for a random drawing of Amazon gift cards ($20), debriefed as to the purpose of the study, and provided with a list of counseling resources in the event they were interested in using them.
Data Analytic Strategy
Measurement invariance testing involves fitting successively more restrictive models to the data (Chen et al., 2005). As applied to our data, the first step, which tested the configural model, evaluated whether the factor structure of the SASS was equivalent across the two ethnic groups. The next step, the metric invariance model, determined whether the factor loadings of the SASS were equivalent across Hispanic and non-Hispanic White participants. Finally, the last step, the scalar invariance model, tested whether the thresholds were equivalent across our groups. A Wald χ2 difference test (using the DIFFTEST feature in Mplus) was utilized to compare the fit of the configural, metric, and scalar models, with a non-significant χ2 indicating the fit of the more constrained model (with parameters constrained to equivalence across groups) does not fit significantly worse than a less constrained model. Degrees of freedom for difference testing was based on the difference in the number of degrees of freedom of the models.
To test whether the theoretical four-factor model of the SASS was invariant across non-Hispanic White and Hispanic college women, we utilized a multiple-groups CFA framework. Sample size recommendations for CFAs vary, however a critical sample size of 200 participants per group has been identified (Kline, 2011), and the current sample surpassed these recommendations. Statistical analyses were conducted using robust weighted least squares estimates (WLSMV) via the Mplus software (Version 8.0; Muthén & Muthén, 2015). Given that the items on the SASS were highly non-normal, they were treated as ordered categorical for the analyses. WLSMV estimators have been found to be robust when using non-normal and categorical data (Li, 2016). Acceptable model fit was determined by a non-significant χ2 (p > .05), root mean square error of approximation (RMSEA < 0.08), and Comparative Fit Index (CFI > 0.90). We also examined correlation residuals greater than |.10| by groups (Kline, 2011).
Results
Preliminary Analyses
Independent-samples t-tests revealed no significant differences between the groups in age [non-Hispanic White (M = 20.68, SD = 1.58) vs. Hispanic (M = 20.73, SD = 1.66) women] or in completed years of college [non-Hispanic White (M = 2.86, SD = 1.30) vs. Hispanic (M = 2.80, SD = 1.30)]. Chi-square analyses also revealed no significant differences between the groups for marital status, χ2 (4) = 3.33, p = .51 or sexual orientation, χ2 (1) = 1.66, p = .20. Table 1 presents the descriptive statistics of the SASS subscales for the Hispanic and non-Hispanic White women. The subscales had similar internal consistencies across groups, with the Date/Friend subscale having the lowest internal consistency (all α > .68). Table 2 presents the correlations between the SASS subscales among the Hispanic and non-Hispanic White women.
Descriptive Statistics of the Sexual Assault Script Scale (SASS) Subscales for Non-Hispanic White and Hispanic Women.
Correlations Between the Sexual Assault Script Scale (SASS) Subscales in Non-Hispanic White and Hispanic Women.
Note. Correlations for non-Hispanic White women are above the diagonal; correlations for Hispanic women are below the diagonal. *p < .05, **p < .01.
Results of Confirmatory Factor Analyses on the 27-Item SASS for Non-Hispanic White Women/Hispanic Women.
Note. Entries are factor loadings for Non-Hispanic White Women/ Hispanic Women. SASS = Sexual Assault Script Scale. In our studies, we presented the items to each participant in a random order.
Confirmatory Factor Analyses
Prior work on the SASS found that a four-factor model demonstrated the best model fit for the data (Yeater et al., 2019a). Thus, in the current study, a CFA of the four-factor SASS was conducted using the Hispanic and non-Hispanic White groups. The CFI and RMSEA model fit statistics revealed that the four-factor model adequately fit the data (CFI = 0.931; RMSEA = 0.062, 90% confidence interval = 0.059 – 0.066, p < .001), however the χ2 test was significant (χ2 (702) = 19095.025, p < .001) indicating some residual misfit of the model to the data. An examination of correlation residuals indicated several correlation residuals greater than |.10|, however the largest correlation residuals were similar in both the Hispanic and non-Hispanic White samples. Overall, the four-factor model of the SASS was an acceptable fit of the data using the overall sample, and as such, we used the four-factor model in tests of measurement invariance across the two ethnic groups. Table 3 presents the results of the CFA for the Hispanic and non-Hispanic White groups.
Measurement Invariance
A test of configural invariance indicated that the model fit the data reasonably well in both groups (Non-Hispanic White: χ2 (318) = 900.609, p < .001; CFI = 0.938, RMSEA = .066 (90% CI = 0.061 – 0.072); p < .001; Hispanic: χ2 (318) = 1096.312, p < .001; CFI = 0.913, RMSEA = .072 (90% CI = 0.068 – 0.077); p < .001). We next tested a metric invariance model with all loadings constrained to be equal across groups and results revealed a non-significant difference between the two models, χ2 (23) = 15.297, p = .883, suggesting that the metric invariance model did not fit significantly worse than the configural model. Next we estimated the scalar invariance model with both loadings and thresholds constrained to be equal across groups and found a non-significant difference between the two models, χ2 (77) = 79.802, p = .391, suggesting that the scalar model did not fit significantly worse than the metric model and the scalar model provided a reasonable fit to the data (χ2 (736) = 1997.993, p < .001; CFI = 0.931; RMSEA = .062 (90% CI = 0.059 – 0.066); p < .001). Overall, the tests of invariance indicated that the factor structure of the SASS was statistically equivalent across the Hispanic and non-Hispanic White groups.
Discussion
The primary aim of the current study was to evaluate the measurement invariance of the Sexual Assault Script Scale (SASS) among Hispanic and non-Hispanic White college women. The SASS was developed in prior work to measure putative risk for sexual assault among college women (Yeater et al., 2019a). The measure has four subscales that were shown in this past work to be internally consistent and criterion valid (i.e., they were associated significantly with measures that tap risk for sexual victimization). Nonetheless, additional work was needed to examine the validity of the SASS; hence, our current efforts to test whether the measure was invariant across two common racial/ethnic groups of college women.
We first conducted a CFA of the four-factor SASS derived in our prior work using the Hispanic and non-Hispanic White groups. Overall, the results suggested that the model fit was adequate with similar residual misfit across groups. Importantly, our tests of measurement invariance suggested that the SASS was invariant across the configural, metric, and scalar models, suggesting that the SASS was measuring a similar construct in Hispanic and non-Hispanic White college women. In addition, the subscales were internally consistent across both groups.
Limitations and Future Directions
The data for our analyses came from a cross-sectional study; thus, we cannot conclude whether certain aspects of sexual assault scripts increase risk for sexual violence; future prospective studies should endeavor to examine the predictive validity of the SASS. Nonetheless, related work in the sexual assault script area suggests that particular elements in women’s scripts (e.g., outdoor location, alcohol and drug use) predict sexual victimization experiences (Krahe et al., 2007; Turchik et al., 2009); thus, upon further testing, the SASS may prove to be a useful measure for identifying at-risk college women. Additional work examining the convergent and discriminant validity of the SASS is also needed, as the original study examined only convergent validity (i.e., associations with measures of putative risk for victimization).
The reliability of the Date/Friend Assault Script subscale was low as compared to the other subscales. In addition, the significant χ2 and correlation residuals indicate some residual misfit and additional work may be required to improve the factor structure of the SASS.
We only tested the SASS on college women in the Southwestern United States; thus, our findings may not generalize to other groups of individuals in disparate locations and cultures. All people are posited to adhere to scripts, including sexual assault scripts (Clark & Carroll, 2008), and although we found the SASS to be invariant across groups of Hispanic and non-Hispanic White women, additional work needs to be done to test the invariance of the SASS in other groups, such as the LGBTQ community, other racial/ethnic groups, and men. Future work would profit from obtaining a large enough sample of lesbian and bisexual women to examine their sexual assault scripts, as these women are at heightened risk for sexual violence relative to heterosexual women (e.g., Drabble et al., 2013). Moreover, the Hispanic population is not homogenous—there is inherent diversity among individuals who self-identify as such; hence, future work might endeavor to test measurement invariance of the SASS among these groups. The SASS may not be invariant across Hispanic and Non-Hispanic White women who are college-aged yet not attending college. Given that we were able to show invariance across Hispanic and Non-Hispanic White women in our study, we predict that we would find similar results among less diverse college students using the SASS. However, this is an empirical question, and future work might endeavor to examine whether the SASS is invariant among “majority” college students in different areas of the United States (e.g., West vs. East coast college students). Finally, we also would not expect the SASS to be invariant across women and men, since related work suggests that sexual assault scripts differ between men and women (Clark et al., 2009).
Practice Implications
To date, the SASS remains the only existing quantitative measure of college women’s sexual assault scripts. Both this study and prior work (Yeater et al., 2019a) suggest that the measure, if shown in future studies to be psychometrically sound, may be an easy, cost-effective tool for measuring college women’s scripts and providing targeted interventions for at-risk women. The SASS might be used to provide college women with feedback about their scripts, that is, whether their scripts reflect what we know to occur, on average, during an assault, as well as whether their scripts appear to “normalize” behaviors that the literature shows increases women’s risk (e.g., alcohol and drug use).
Conclusions
A paucity of effective interventions exists to reduce men’s sexually aggressive behavior (Anderson & Whiston, 2005; Ellsberg et al., 2015); thus, research focused on augmenting women’s ability to reduce their risk for victimization remains a necessary adjunct to work on men’s sexual aggression. One promising avenue for understanding women’s risk for sexual victimization may be the “cognitive models” or scripts that women use when considering the contextual and interpersonal features that comprise a sexual assault. Certainly, related work has shown that women’s cognitive processing of sexual assault-related information is linked to heightened risk for sexual victimization (e.g., Yeater et al., 2010, 2019b). Again, our focus is not meant to intimate that women are responsible for sexual assault; that responsibility remains with the perpetrator. Yet it is our hope that work in this, as well as related areas, may help us to identify novel and effective methods by which we can reduce all women’s risk of experiencing sexual violence.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Author Biographies
Elizabeth A. Yeater, PhD, is an associate professor and director of Clinical Training in the Department of Psychology at the University of New Mexico. Her research focuses on identifying contextual and interpersonal factors associated with women’s risk of sexual victimization by using a social information processing model and methods borrowed from cognitive science.
Kristen N. Vitek, MS, is a clinical psychology graduate student at the University of New Mexico. She received a bachelor’s degree in psychology and sociology from the University of Illinois at Chicago and a master’s degree in clinical psychology from the University of New Mexico. Her research interests focus on trauma, particularly sexual victimization as it relates to interpersonal functioning.
Ryan S. Ross, MS, is currently a graduate student at the University of New Mexico in the Clinical Psychology Ph.D. program. He received his B.S. in Psychology, B.A. in Sociology, and M.S. in Clinical Psychology from the University of New Mexico. His research interests include gender differences in sexual trauma and risk factors for sexual victimization.
Meredith Blackwell, MS, is a clinical psychology doctoral student at the University of New Mexico. She received a bachelor’s degree in psychology and Arabic from the University of Mississippi, and a master’s degree in psychology from the University of New Mexico. Her research interests center on improving effectiveness of and access to culturally competent trauma treatments, especially for displaced and war-exposed groups.
Katie Witkiewitz, PhD, is a Regents’ Professor of Psychology at the University of New Mexico. The underlying theme of her research is the development of empirically-based models of addiction, with an emphasis on applying advanced quantitative research methods to better understand changes in alcohol and drug use behavior over time.
Kari Leiting, PhD, received her doctorate in clinical psychology from the University of New Mexico. Her research interests include sexual assault scripts, variables that may influence them, and what behavior scripts may predict. She is a staff psychologist at the Sioux Falls VA Health Care System.
