Abstract
These three related studies created a set of ecologically valid scenarios for assessing relative associations of both attraction and sexual coercion risk-recognition in college women’s heterosocial situational drinking decisions. The first study constructed nine scenarios using input from heterosexual drinking women in the age cohort (18-30) most likely to experience alcohol-related sexual coercion. In the second study, 50 female undergraduates (ages 18-25) assessed the salience of three important dimensions (attraction, risk, and realism) in these scenarios. The third study was a factor analysis (and a follow-up confirmatory factor analysis) of the elements of coercion-risk as perceived by the target group with two female samples recruited 1 year apart (Sample 1: N = 157, ages 18-29); Sample 2: N = 157, ages 18-30). Results confirmed that the scenarios could be a useful vehicle for assessing how women balance out risk and attraction to make in-the moment heterosocial drinking decisions. The factor analysis showed participants perceived two types of situations, based on whether the male character was “Familiar” or “Just Met” and perceived themselves as happier and more excited with Familiar males. However, in contrast to HIV risk studies, Familiar males were perceived as higher risk for unwanted sex. Future research will use the six scenarios that emerged from the factor analysis to study how attraction and risk perception differentially affect young adult women’s social drinking decisions.
Despite increased attention both in research and the public media, the risk of coerced sex or rape remains a disturbing risk for young adult women (ages 18-25). In their seminal study 30 years ago, Koss, Gidycz, and Wisniewski (1987) reported 27.5% of U.S. college women had experienced at least one incident of rape or attempted rape after the age of 14. A review by Gidycz, McNamara, and Edwards (2006) summarized several subsequent studies showing a consistent rate of 16% to 18% of U.S. college women reporting a rape or attempted rape just in the 3 to 6 months prior to being surveyed. These incidents often occur in the context of a social situation (i.e., “acquaintance rape” vs. “stranger rape”; Abbey, 2002; Testa & Livingston, 2009; Ullman, 2003), and the most robust finding in this area is that sexual assault is associated with alcohol consumption by both men and women. Alcohol use was present in one half to two thirds of reported sexual assault cases (Ullman, 2003).
Alcohol administration studies suggest that although women are aware of sexual assault risks when they are sober, a moderate to high (.08% or above) alcohol dose is associated with less awareness of risk cues (Testa & Livingston, 2009) or possibly less discomfort with risk cues (Davis, Stoner, Norris, George, & Masters, 2009). Furthermore, in a retrospective study, Parks and Fals-Stewart (2004) found that sexual aggression incidents were 9 times as likely to occur on women’s heavy drinking days versus non-drinking days. In a prospective study, women’s daily reporting for 8 weeks using Interactive Voice Response showed that sexual aggression occurred significantly more often on their heavy drinking days (Parks, Hsieh, Collins, & Levonyan-Radloff, 2011). Although the offender is completely culpable, such research highlights the potentially important association between sexual assault and women’s alcohol use. As elucidated in the following review, the mechanisms underlying this connection are currently poorly understood, but better knowledge about women’s drinking decisions may be helpful in reducing their risk level (see, for example, Norris, 2011).
Young adult women engage in alcohol-related heterosocial interactions (parties and going out to bars, for example) usually because they are attracted by the possibility of a rewarding relationship (Abbey, 2002) and alcohol consumption is strongly associated with the good times and fun of dating (Abbey, 2002; Cowley, 2014; Crowe & George, 1989). However, they may fail to recognize or respond when warning signs for coercion are present (Norris, 2011; Norris, Nurius, & Dimeff, 1996). The cognitive mediation model of sexual decision making (Norris, Masters, & Zawacki, 2004) suggests that especially when a woman is attracted to a man, she may have difficulty distinguishing between the man’s reciprocal expression of sexual interest versus his statements and behavior warning of sexual assault risk. In addition, she may recognize risk but fail to heed coercion cues because of a “positive process” (also known as “unrealistic optimism”) effect (Taylor & Brown, 1988; Weinstein, 1980) commonly seen when people believe themselves to be better able than the “typical” person to control risky situations (Abbey, 2002).
It is within the context of this inherent conflict of heterosocial situations that women make in-the-moment drinking decisions. Decision Theory (see Janis & Mann, 1977; Kuther, 2002) posits that response selection (in this case, how much alcohol to drink) is based on balancing danger signals (e.g., assault risk cues) versus impelling cues (e.g., attraction, signaling possible rewards). Enhancing recognition of risk cues may be an important method of reducing drinking, and much research has been directed to studying how to enhance women’s risk recognition (see Gidycz et al., 2006), emphasizing the negative side of the balance. However, attraction (the positive side of the balance) could be an important determinant of women’s drinking decisions as well, as it is for other risky decisions. For example, in a recent study (Lennon & Kenny, 2012), women projected themselves as more likely to have unprotected sex with a man they rated as “attractive” even though they also rated him as more likely than others to have a sexually transmitted disease. A parallel process in making drinking decisions, in which attraction outweighs risk, might begin to explain the strong connection between drinking and sexual assault. However, to date, the relative influence of attraction cues on women’s drinking decisions in heterosocial situations is under-researched. One method for addressing this gap is through systematic laboratory studies allowing female participants to balance the relative influences of attraction versus risk recognition in making in-the-moment drinking decisions.
Laboratory studies have the advantage of allowing a researcher to isolate and manipulate a number of factors that might otherwise vary unpredictably in an in vivo study (Noel et al., 2008). However, for a laboratory study to be useful, it needs to have as much external validity as possible, without putting the participant in an embarrassing, unsafe, or possibly harmful condition. Using scenarios has been a common method for circumventing ethical concerns in studying sensitive topics, such as sexual coercion or risky sex (see Maisto, Carey, Carey, & Gordon, 2002; Maisto, Carey, Carey, Gordon, & Schum, 2004; Maisto, Carey, Carey, Gordon, Schum, & Lynch, 2004). In a scenario study, participants typically are asked to project themselves into a vignette and then to make decisions about how they will behave. Such judgments are considered to be an approximate assessment of the behavior being studied (see Bieneck, 2009, for a review of scenario methods).
For example, Livingston and Testa (2000) used a vignette in which young adult women were to see themselves as interested in a character named Mark. To summarize the story, Mark calls from a bar to ask the woman to join him. When she puts him off because she is tired, he comes over to her home anyway with pizza and beer. Participants were asked what they would do and researchers assessed whether participants recognized the sexual coercion risk, whether they would let Mark come in (and why), and what precautions they would take (if any).
Thus, to study the relative influence of attractiveness cues versus risk recognition on women’s drinking decisions, this article details our development of a series of ecologically valid vignettes about heterosocial situations incorporating risk, pressure to drink, and, most importantly, a strong element of attraction. In these scenarios, participants can project themselves into several situations that can systematically vary the intensity of attractiveness in potentially risky situations; thus, in future studies, researchers using the scenarios can assess the relative contributions of these variables to drinking decisions. This article details the steps we followed to develop this research instrument (a series of short ecologically valid scenarios) so that the scenarios can be used across several studies, thus advancing the field through increasing comparability of laboratory research in this area. Generally, we followed the framework described by Noel et al. (2008) for developing sexual assault-related scenarios with men, but adapted it for the female-specific needs of this study.
Much of the original scenario generation was done by female undergraduate research assistants (RA) drawn from the same population we wanted to assess, using material elicited from a previous focus group study of young adult women (Noel, Ogle, Maisto, & Jackson, 2016). An important aspect of scenario studies is that potential participants representing the target group must perceive the specific cues as reflections of critical factors (in this case, drinking pressure, sexual coercion risk, and attraction) and the vignette must seem realistic to them. The scenario may portray objective reality perfectly, but if a participant does not perceive it as such, she may not react characteristically. Moreover, developing several scenarios that all incorporate these elements allows some flexibility so that if a participant has difficulty visualizing herself in one, another is available.
After we developed a well-validated set of risk scenarios, we winnowed the large number of vignettes by factor analyzing women’s judgments of sexual coercion risk in each one. The final goal was to provide researchers with a set of clear and ecologically relevant scenarios that could be useful across several studies for comparison purposes.
To this end, the current article comprises three related studies. The first was a qualitative project aimed at constructing the scenarios using input from women in the age cohort (18-30 years) most likely to experience alcohol-related sexual coercion. The second was a quantitative study that assessed the salience of the three important variables (attraction, risk, and realism) with the intended target group. The third study was a factor analysis (and a follow-up CFA) of the elements of risk for coercion as perceived by young adult female participants in two large samples recruited 1 year apart.
Study 1: Developing Scenarios for the Target Group That Are Salient for Attractiveness, Pressure to Drink, and Pressure to Have Unwanted Sex
Method
Participants
Because the aim was to make the scenarios ecologically relevant, all of the participants in the studies described below were female university students in the target age group of 18 to 25 recruited from Introductory Psychology classes at a 15,000-student southeastern U.S. public university. All received course credit for their participation. The university institutional review board approved all procedures.
Step 1: Scenario development and writing
In the previous study with several focus groups, 44 young adult women (age 18-30; see Noel et al., 2016) provided original content describing drinking-related social situations they perceived as risky for sexual coercion (e.g., they interacted with a man who was attractive to them but who could potentially pressure them to have sex, even if they did not want to). Using this content, a team of five female undergraduate RAs (ages 18-25, four Caucasian, one African American) developed 11 scenarios, carefully wording them so that the scenarios were “realistic” (i.e., similar to what women in the focus groups had experienced or had heard about from friends), used language that was peer-appropriate, and, most importantly, were attractive, despite also being risky for alcohol-related coerced sex. Scenarios were worded in the second person (“you”) and the woman in the story interacted with a particular man. The man’s behavior was described, but not his physical appearance beyond a positive description (e.g., “a good-looking guy”) so that his appearance could be imagined by the participant reading the vignette. Each scenario was about a paragraph long and ended with an opportunity for the woman to drink alcohol. In each case, we were also careful to avoid drawing specific attention to the possibility of sexual assault, because, as in real life, the scenario ending was uncertain—it could be the beginning of a good relationship, or a bad incident, and the participant had to make a drinking decision based on only the available knowledge. Because it is often cited, we added, the “Mark” scenario (Livingston & Testa, 2000), so that altogether 12 scenarios described potentially dangerous, but attractive, sexual and drinking situations. To avoid a confounding order effect, each scenario was named with two randomly selected letters and a double digit number, and in subsequent studies, they were presented in random order. In addition, the vignettes were titled “Romance Scenarios” to highlight the attractive aspects and not the sexual coercion potential.
Step 2: Qualitative feedback from the target group
The scenarios were further refined with a qualitative study. Individually, 21 heterosexual female (ages 20-25, 19 Caucasian, two African American) undergraduate psychology majors read the 12 scenarios. They gave written and oral feedback to one of the female RAs about the descriptive language, and discussed how to maximize the plausibility and realism of each scenario. They also rated the “realism” of the scenarios on a scale of 0 to 7, 7 being “as close as possible to real life.”
On the basis of this feedback, some wording was slightly changed, including re-wording of the Livingston and Testa (2000) scenario (named “CG56” in the appendix). Furthermore, three scenarios were dropped through this process because they were deemed “unrealistic.” More than half of the participants rated these three as below the midpoint (3.5) of the scale for “realistic.” This left a total of nine scenarios (appendix) for potential use, with a mean “realistic” rating across the nine scenarios of 5.2 (SD = 1.2).
The next step was to assess the salience of the cues with a new sample of participants from the target group, young adult women.
Study 2: Assessment of Scenarios for Salience of Attraction, Pressure to Drink, and Pressure for Sex
Participants
Fifty female undergraduates (ages 18-25, M = 18.92, SD = 1.31; 44 Caucasian, four African American, two Hispanic) responded to a psychology department research website asking for women who drank alcohol to participate in a questionnaire study. Originally, 52 participants volunteered, but because all of the scenarios related to attraction to men, data were dropped from two women who said they were exclusively lesbian. All participants reported drinking alcohol at least once per month in the past 90 days, and the average amount they drank per occasion was three to four standard drinks, as assessed by a retrospective drinking measure based on the Quantity-Frequency Index (QFI) questionnaire (Cahalan, Cisin, & Crossley, 1969). This questionnaire has been shown to be both reliable and valid with college students (Hagman, Clifford, Noel, Davis, & Cramond, 2007).
Procedure
Each of the nine scenarios was presented individually and in random order by a female RA who read the scenario out loud and provided a copy for the participant to read along. Each scenario was read aloud to provide extra assurance that the participant was attending to the scenario and not rushing through them to finish the experiment quickly. To preserve anonymity, the participant recorded her ratings on a computer screen that faced away from the RA. Participants were instructed to imagine themselves as the “typical” female undergraduate and, using a scale of 0 (not at all) to 7 (very much), rate (a) how much she would be attracted to the man, (b) how much she would want to continue a relationship with the man, (c) the happiness and excitement she would feel in the situation, (d) the pressure that she would feel to drink, and (e) how likely it was that the man would pressure her to have sex, even if she did not want to. Finally, participants rated how “realistic” the situation was (how much is it like situations you have seen or heard about?).
Results and Discussion
Means and standard deviations of the ratings for each scenario are shown in Table 1. To summarize, all nine scenarios were rated as acceptably realistic. The lowest mean rating for realism was 4.66 (SD = 1.81), well above the midpoint of the scale (3.5).
Participants’ Ratings of the Nine Scenarios (Study 2).
Note. The ratings are on a scale of 0 (not at all) to 7 (very much). The questions were (a) how likely is it that the typical undergraduate woman would be attracted to the man; (b) how much she would want to continue a relationship with this man? (c) In this situation, how happy and excited would she feel? (d) How much pressure would she feel to drink alcohol? 5) How likely is it that the man would pressure her to have sex, even if she did not want to; (e) How realistic is this scenario? In all cases, N = 50.
In addition, the men in eight of the scenarios were rated above the midpoint on attractiveness and on the likelihood that the woman would want to continue a relationship with him. Happiness ratings were high for the same eight scenarios. Thus, it appeared that with these eight scenarios, the aim of making the situations and men attractive to the target population was achieved. The ninth scenario, the one (CG56) adapted from Livingston and Testa (2000), was not rated as positively as the others. However, because it was close to the midpoint on the positive scales, we decided to retain it for purposes of comparison with the literature in this area.
Finally, participants appeared to recognize pressure to drink in each of the nine situations as well as a high likelihood of pressure to have sex, even if unwanted, with mean scores again well above the midpoints on the scales. Overall, then, the target group, young adult women, appeared to perceive the nine scenarios as realistic, attractive (with exception of CG56), but risky for drinking and unwanted sex.
Study 3: Demonstration of Use: Deriving the Factor Structure of Young Adult Women’s Perception of Risk for Alcohol-Related Sexual Coercion
The next step in refining and strengthening the set of scenarios was to assess the factor structure of the “risk” concept. Although the risk ratings were well above the midpoint for each of the nine scenarios, there was also some variability that suggested our participants did not perceive a unitary concept of “risk” when making ratings. The utility of having several valid scenarios available was the ability to test empirically if “risk” is a unitary concept. Thus, in this portion of the study, we used two samples taken at different times, one to test the factor structure of risk recognition in young adult women and the other to cross validate it. The two samples were recruited and data were collected 1 year apart.
Method
Participants
Sample 1 participants were 157 female undergraduates ages 18 to 29 (M = 19.22, SD = 1.8, 72% freshmen or sophomores) recruited in the fall semester, 2012, for a study of “women ages 18 to 30 who date men or who want to date men.” Originally, 207 volunteered for the study but because all of the scenarios related to drinking and attraction to men, data were excluded from 50 who described themselves as exclusively homosexual/lesbian, or reported no drinking days in the last 90, or both. The majority (88%) described themselves as White or Caucasian. The rest described themselves as African American, Biracial, and Hispanic or Latina. Of the total sample, 90 (57.3%) said they were single, 60 (38.2%) said they were in a “steady dating” relationship, and the rest (4.5%) were divorced, widowed, or did not answer the relationship question. All received course credit for their participation in the study.
Sample 2 participants were a similar cohort recruited 1 year later, in the fall semester of 2013. Of the original sample (N = 220), again 157 met criteria (heterosexual alcohol consumers between 18 and 30; M age = 18.68, SD = 1.69; 79% freshmen or sophomores); 81% self-described as White or Caucasian, the rest were African American, Asian American, and Hispanic or Latina; 53% were single, with 45% in a steady dating relationship. The second sample was significantly younger than the first, t(313) = 3.49, p < .001, but in real terms, the difference was only about 6 months of age. Otherwise, t tests showed no differences between the two samples on any of the demographic measures or question responses (see “Measures” section). Thus, they appeared to be similar to the participants in Sample 1.
Measures
Measures for each sample included a demographics questionnaire (age, ethnicity, education level, sexual orientation, and dating status), the set of nine “Romance Scenarios” and a response sheet with a series of questions for each scenario (see below). We collected drinking and drug use data via a retrospective drinking and drug use measure based on the QFI questionnaire (Cahalan et al., 1969). Each participant estimated her frequency of drinking hard liquor, wine, and beer in the past 90 days and the quantity of each she had consumed per drinking day within the past 90 days (from none up to six or more standard drinks).
Procedure
The protocol used in both samples was approved by the university’s institutional review board as an anonymous questionnaire project. We administered questionnaires in a classroom with groups of 16 or fewer participants at a time. Participants sat in desks with an empty desk on either side to preserve privacy and folders with the demographics questionnaire, the QFI, and the response sheets were placed on designated desks. To begin each administration, a RA (all were female) read aloud a statement affirming the participant’s right to voluntary and anonymous participation. Next, she read the directions for the demographics questionnaire and the QFI. Once the participant had completed this portion of the study, she placed her questionnaire in the folder on her desk.
Each RA then presented the scenarios using Microsoft PowerPoint, reading them aloud while the participants read each scenario on the computer screen. The order of scenario presentation was determined randomly. Following each scenario, the RAs asked the participants to complete the items on the response sheet accompanying the particular vignette. Then, the RA presented the next scenario, and so on, until all of the scenarios had been presented and the items accompanying each had been completed. Response sheets were put in a folder as each was completed.
The response sheet at the end of each scenario included the following questions:
After all nine scenarios had been presented in random order, the participant re-read each scenario and answered two further questions about each:
Questions 3 and 4 were asked in random order. Both ratings were obtained to avoid the possible effects of “unrealistic optimism” (noted in the Introduction). We were concerned that participants might overall rate themselves as better able to handle the risk, thus decreasing the risk ratings on an absolute scale. We hypothesized that the “typical” rating would illustrate an objective risk rating, while the “self” rating could assess the participant’s perception of her own personal risk.
Once the last form was completed, the participant put all her forms in her folder and handed it to the RA. RAs collected the folders and deliberately mixed the order of the folders so the participants could feel safe in their anonymity. Each participant was given course research credit, asked if she had any questions or comments, and thanked for her assistance. RAs also routinely gave each participant a slip of paper with contact information for campus health services and the community rape crisis and counseling center.
Results
Participants’ Drinking and Dating Behavior
Sample 1
Participants’ self-reports of drinking and dating in the last 90 days showed a large amount of variability. The number of days drinking in the last 90 ranged from 1 to 90 with an average of 19.98 (SD = 17.56) days. On occasions when they drank, the mean amount consumed was reported as 5.66 (SD = 0.77) on the scale of 1 to 7 (corresponding to about 3.5 standard drinks per drinking occasion). In terms of dating, 25 (15.9%) said that they had not dated a man in the last 90 days, but that they did want to date men. On the average, the participants reported dating 16.21 days (SD = 18.91) of the last 90 days (ranging from 0 to 80).
Sample 2
Sample 2 participants reported drinking from 1 to 78 days in the past 90 (M = 16.25, SD = 14.54), and the amount consumed per drinking occasion ranged from 1 to 6 standard drinks (M amount score = 5.63, SD = 0.69—again, roughly equivalent to 3.5 standard drinks). A total of 22 women (14%) had not dated in the past 90 days (but said they wanted to). The mean number of dates in the past 90 days was 16.66 (SD = 19.42) ranging from 0 to 90.
Judgments of Risk for Sexual Coercion in Each Scenario
Sample 1
Ratings of sexual coercion risk could range from 0 “not at all likely” to 7 “very likely.” The most important “risk rating” for judging the scenario was the level of risk for a “typical female college student.” The means of these ratings ranged from 4.77 (SD = 1.83) to 2.82 (SD = 1.89), depending on the scenario (see Table 2). Mean ratings for risk to “self” were about two points lower than “typical” for each scenario. Paired-samples t tests showed that, for each scenario, participants rated their risk of sexual coercion as significantly lower than that of the “typical” female college student (ts ranged from 12.8 to 16.0; dfs: 154-156; all ts significant at the p < .001 level). However, the self-risk ratings also correlated significantly with the “typical” risk (i.e., showed the same pattern of riskiness ratings; Pearson’s rs ranged from .59-.71; all p < .001), thus suggesting that although absolute scores for the self-ratings were lower, the risks were detected and the situations were seen on the same continuum of risk for self and for the “typical” female college student.
Risk Ratings for “Typical College Female” and “Self” (Study 3, Samples 1 and 2).
Note. Ratings are from Sample 1 (N = 157) and Sample 2 (N = 157) collected 1 year apart. Ratings were based on a scale of 0 (not at all) to 7 (very much) in response to the question, “How likely is it that the man would pressure her to have sex, even if she did not want to,” and “How likely is it that the man would pressure you to have sex, even if you did not want to.” The differences between “typical” and “self” were all significant (all ps < .001).
Sample 2
The mean risk ratings for the “typical female college student” ranged from 4.69 (SD = 1.81) to 2.46 (SD = 1.97), depending on the scenario (see Table 2). Again, Table 2 shows that the mean ratings for risk to “self” were about two points lower than “typical” for each scenario. Paired-samples t tests showed that, again for each scenario, participants rated their personal risk of sexual coercion as significantly lower than “typical” (ts ranged from 12.9 to 17.60; dfs: 154-156; all ts significant at the p < .001 level). As with Sample 1, the self-risk ratings also correlated significantly with the “typical” risk (i.e., showed the same pattern of riskiness ratings; Pearson’s rs ranged from .59-.67; all p < .001). Thus, to analyze the factor structure of the risk ratings, we were confident in using just the one (“typical”) risk rating per scenario in each of the two samples because they were similar to the self-rating but provided a great range.
Analysis Strategy
Analysis of the nine scenarios for an underlying factor structure for “risk” proceeded in two steps. First, we conducted an exploratory factor analysis (EFA) using SPSS 19.0 for Windows to derive a preliminary factor structure for the “typical” risk ratings. This analysis was conducted using Sample 1 (n = 157). Following the data analytic approach of Briere, Malamuth, and Check (1985) and Walker, Rowe, and Quinsey (1993), we used principal components analysis (PCA) to extract the factors, and a Varimax rotation was used to increase interpretability. Once we derived a factor structure, we conducted a CFAs with Sample 2 (n = 157) to test the fit of the derived factor structure for a models in which the latent factors are allowed to correlate and in which the latent factors were constrained to be uncorrelated, respectively. The CFA was conducted using AMOS 6.0. To test the fit of the derived model, the comparative fit index (CFI), normed fit index (NFI), and root mean square error of approximation (RMSEA) were used.
PCA
The PCA of the nine ratings from Sample 1 (n = 157) resulted in a two-factor solution. The first factor accounted for 52.8% of the variance and had an eigenvalue of 4.76. The second factor accounted for 12.6% of the variance and had an eigenvalue of 1.13. Only two factors had eigenvalues greater than 1.0, and the scree plot was consistent with a two-factor solution. Overall, the two factors derived accounted for a total of 65.5% of the variance.
Extracted factors were rotated using a varimax rotation to improve interpretability. Table 3 presents the item loadings on the two derived factors in Sample 1. Item loadings of .40 or higher were considered significant (Stevens, 2002). Six items loaded exclusively on a factor (three on each), and three items loaded on both factors. The first factor consisted of three items relating to scenarios in which the male was someone whom the individual had just met or had recently met and was labeled Just Met Scenarios. As seen in Table 3, these scenarios were PN38, ZL49, and AT01. The second factor consisted of three scenarios in which the male was casually or well-known to the individual and was labeled Familiar Scenarios. These scenarios were RL22, JK96, and BD50 (Table 3). Calculation of Pearson’s r for the two factors showed that they were highly correlated with each other (.481, p < .001) but also significantly different in that the Just Met scenarios were judged as significantly lower risk than the Familiar scenarios, t(155) = 10.11, p < .01. As noted, three of the scenarios (AG25, WF63, and CG56) did not load differentially on either factor so these were dropped for clarity of interpretation of the factors.
Item Loadings for the Two Derived Factors in Sample 1 (Study 3).
Note. Based on a principal components analysis with Varimax rotation of 157 participants’ ratings of sexual coercion risk for a “typical” undergraduate female. The first three loaded on Factor 1, the second three loaded on Factor 2, and the last three did not load differentially.
CFAs
Two CFAs were conducted using the Sample 2 (n = 157) to test the fit of two variations of the derived two-factor model to the data. We hypothesized that, given that the two factors from the PCA were correlated, a model in which the latent factors were allowed to correlate would be a better fit to the data than a model in which the latent factors were constrained to be uncorrelated. As noted earlier, the six scenarios that loaded significantly onto one of the two factors derived in Sample 1 were included in the CFAs.
For the CFA of a model in which the two factors were constrained to be uncorrelated, the results, based on typical fit statistics, showed that the derived two-factor model significantly deviated from the data, χ2(9) = 69.10, p = .026, CMIN/df = 7.68 (this expression is the chi square divided by the degrees of freedom). Fit statistics for the model were as follows: CFI = .796, NFI = .781, RMSEA = .172 (90% CI [confidence interval] = [.135, .210]). Based on these statistics, this model was considered to be a poor fit.
The CFA of the correlated factor model revealed that the derived two-factor model significantly deviated from the data, χ2(8) = 17.46, p = .026, CMIN/df = 2.18; however, the df adjusted χ2 of 2.18 is considered acceptable, as it is less than 3. Fit statistics for the model were adequate: CFI = .957, NFI = .927, RMSEA = .086 (90% CI = [.03, .14]). The model was judged to be an acceptable fit overall (Kline, 2011) and was a significantly better fit to the data than the uncorrelated model, χ2(1) = 51.64, p < .001. As with Sample 1, the Just Met scenarios were rated as significantly less risky than the Familiar scenarios, t(156) = 8.88, p < .01.
Association of Current Drinking With Scenario Risk Perception
As a final step in our analyses, we combined both samples (N = 314) to assess any associations between participants’ current drinking (past 90 days) and their judgments of risk in each of the six final scenarios. In three scenarios (one “Familiar” and two “Just Met”), there were small positive correlations between drinking frequency and judgments of risk (r = .13, p < .02; r = .14, p < .01; r = .13, p < .02, respectively). No other correlations were found with drinking frequency and risk judgment or usual drinking amount and risk judgments.
General Discussion
The main purpose of the current project was to develop a sound scenario method for laboratory assessment that can test the relative influences of attraction and of sexual coercion risk on young adult women’s drinking decisions in heterosocial situations. We constructed and refined a set of ecologically relevant scenarios for this target group in an organic method that used content and feedback derived directly from the group to be studied. Validating the scenarios with the target population was more important than reproducing what the researchers might see as “objective reality.” Because we plan to use these scenarios in subsequent studies with young adult women (and we hope other researchers will, as well), we wanted a set of scenarios with which we could be sure that the target participants felt attraction and were aware of risk.
We systematically constructed the scenarios, using content elicited from women in an earlier study, but then tested the scenarios to ensure that they met the three criteria (attraction, risk, and realism). Furthermore, using two large samples of young adult women, collected 1 year apart, we were able to determine the factor structure of the women’s risk perception in these scenarios. A principal components factor analysis on the first sample was validated with a CFA of the second sample, suggesting that the two-factor structure was robust.
The two risk components seen in this study appear to differ primarily in regard to the degree to which the male character is known to the female. Although it is well-documented that most sexual assaults are perpetrated by someone known to the victim (Ullman, 2003), there is little research on the degree to which familiarity and attraction influence perceived sexual coercion risk and drive drinking decisions. However, risky sex HIV-prevention research may offer a useful parallel to the sexual coercion risk literature. Zawacki and colleagues (2009) suggested that women tend to divide “attractive” men (men with whom they would like to pursue a relationship) into those whom they know well or at least “pretty well,” and those whom they have just met, for example, at a large gathering or through social media, such as Facebook. This finding is consistent with the present study. However, in the HIV-prevention study (Zawacki et al., 2009), women associated better familiarity with less risk of STD’s (because they thought the men were more likely to be “clean”) and were more willing to have unprotected sex. In the current study, women associated familiarity with higher risk for a different possible negative consequence: unwanted sex demands.
The contrast between the two studies suggests that the women in our study may have been responding to the thought, “Yes, I know this guy, so he is going to feel freer to push me to have sex even if I don’t want to.” Perhaps they are recognizing that as “risk,” because it IS risk for unwanted sex. Women often have trouble communicating that they do not want to have sex unless and until they feel comfortable (Muehlenhard & Peterson, 2005). With someone she has had more interaction and more familiarity and to whom she is attracted, a woman might feel like she has “run out of excuses” and cannot say no to him without “being rude.” Several women in the focus group study (Noel et al., 2016) said, when answering questions about why they might have unwanted sex with minimal protest, that “being rude” to an attractive man was considered very negative for the woman (see also Peterson & Muehlenhard, 2007; Rinehart & Yeater, 2012).
In HIV risk research, partner familiarity also appears to interact with alcohol consumption in determining sexually related behavior. Findings from the HIV risk literature suggest that greater alcohol consumption is related to an increased likelihood of sex with a casual partner (Leigh & Schafer, 1993; Parks et al., 2011; Temple, Leigh, & Schafer, 1993) and is related to both the level of sexual activity with unfamiliar partners and the degree to which the experience was later regretted by the person (LaBrie, Hummer, Ghaidarov, Lac, & Kenney, 2014). How alcohol might interact with partner familiarity and attraction in detecting and reacting to sexual coercion risk is not clear. In the current study, participants were not actually drinking alcohol when they made risk judgments. However, to answer a related question, we did assess for associations between participants’ previous 90-days drinking patterns and scenario risk judgments by correlating both frequency (days drinking) and usual amount with risk perceptions in these scenarios. Although technically we did find significant correlations (the more days drinking, the higher the risk perception) in three of the six scenarios, these correlations were quite small. Perhaps frequent drinkers are slightly more likely to detect alcohol-related risk when being offered a drink in a risky situation, but given the relative lack of strength in these findings, further study is needed to determine if this is a consistent and meaningful pattern.
The final set of six scenarios developed in the present study will be of assistance in assessing at least two major research questions. First, as discussed previously, do partner familiarity and attraction influence young adult women’s perception of sexual coercion risk in heterosocial drinking situations? The Cognitive Mediation Model (Norris et al., 2004) suggests that strong attraction will balance out or even counteract caution induced by risk perception. The different degrees of attraction in these scenarios could allow some within-subjects’ assessments of this question. Second, we can study the question most central to our line of research, how do familiarity, attraction, and sexual coercion risk perception combine to determine a young adult woman’s drinking decisions in heterosocial situations? Taken as a whole, it appears that familiarity is an important factor in sexual coercion risk perception, and in our subsequent studies, we will use this factor structure to assess differential drinking decisions women make with men based on the women’s differential coercion risk perception.
While we hope these carefully developed scenarios can provide a useful methodological advance in the laboratory, we also have to acknowledge some limitations. First, no non-drinkers were included in the study. Because our aim was determining factors that may be associated with potentially risky drinking decisions, we focused on women who drink currently. It may be that non-drinkers would have very different risk perceptions than drinkers and perhaps that can be determined in later studies. Second, 88% of Sample 1 and 81% of Sample 2 participants described themselves as Caucasian or White, meaning that only a relatively small number were women of color. While we did include women of color in each of our development steps, both as RAs and participants, their percentages were small, so it may be that the scenarios are most appropriate to Caucasian women. In the future, we would like to assess the experiences and risk judgments of a more diverse group of women and perhaps even develop scenarios specific to particular groups. As a third limitation, we were unable to assess any differential risk perceptions of women based on their past experience with sexual assault, because this factor was not assessed. We used an overabundance of caution in developing the scenarios because of the possibility of demand characteristics influencing the content and the judgment of risk. Thus, we did not reveal to the participants until the end of the session that the series of studies were concerning sexual coercion risk. That was a trade-off and in the future we would like to include women’s sexual experiences (both positive and negative) as a factor influencing judgment.
Over the long run, the development of these specific scenarios gives us the ability to test the relative influences of attraction, familiarity, and risk perception on young adult women’s drinking decisions and could have important implications for modifying those drinking decisions in heterosocial situations. For example, a prevention program that addresses women’s differential reactions to partner familiarity and type of risks could be more effective than one that assumes a unitary concept of “risk,” and the influence of attraction could be addressed in a decisional balance type of exercise.
Footnotes
Appendix
Acknowledgements
The authors thank their research assistants: Cynthia Bennett, Ashley Cottle, Mackenzie Earl, Allison Eddy, Sarah Gaweda, Rebecca Holliman, Kaitlyn Huff, Morgan Metzger, and Melissa Meglin.
Authors’ Note
Portions of this research have been presented at the Research Society on Alcoholism and the Association for Cognitive and Behavioral Therapy conventions in 2012 and 2013. Randi Loomis is currently attending New York University in New York, NY. Jennifer Heaton is currently at Embry-Riddle Aeronautical University in Daytona Beach, FL.
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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: All of this research was conducted at and supported by the University of North Carolina Wilmington. Dr. Maisto was partially supported by a grant from the National Institute on Alcohol Abuse and Alcoholism (2K05 AA16928) during this research.
