Development and Validation of a Video Measure for Assessing Women’s Risk Perception for Alcohol-Related Sexual Assault

Abstract

Objective

Using an iterative process, a series of three video scenarios were developed for use as a standardized measure for assessing women’s perception of risks for alcohol-related sexual assault (SA). The videos included ambiguous and clear behavioral and environmental risk cues.

Method

Focus group discussions with young, female heavy drinkers (N = 42) were used to develop three videos at different risk levels (low, moderate, and high) in Study 1. Realism, reliability, and validity of the videos were assessed using multiple methods in Studies 2 and 3. One hundred-four women were used to compare differences in risk perception across the video risk level in Study 2. In Study 3 (N = 60), we assessed women’s perceptions of the low and high risk videos under conditions of no alcohol and alcohol.

Results

The realism and reliability of the videos were good. Women who viewed the low risk video compared to women who viewed the moderate and high risk videos perceived less risk for SA. We found an interaction between alcohol and risk perception such that, women in the alcohol condition were less likely to perceive risk when watching the high risk video.

Conclusions

As the video risk level increased, women’s perception of risk increased. These findings provide convergent evidence for the validity of the video measure. Given the limited number of standardized scenarios for assessing risk perception for sexual assault, our findings suggest that these videos may provide a needed standardized measure.

Both prior victimization and alcohol intoxication have been associated with sexual assault (SA, see Testa & Parks, 1996 for a review). One of the mechanisms through which prior victimization and intoxication have been hypothesized to place women at risk for additional SA, is through reduced ability to perceive SA risk cues (Cloitre, 1998; Norris, Nurius, & Dimeff, 1996). Previous research has used written, audio, and video scenarios of heterosexual interactions embedded with SA risk cues (e.g., alcohol use or non-use by the victim and/or perpetrator, sexual precedence, resistance by the victim) to assess differences in women’s ability to perceive risk for SA under conditions of alcohol intoxication and as a function of prior victimization (e.g., Davis, Stoner, Norris, George, & Masters, 2009; Messman-Moore & Brown, 2006; Soler-Baillo, Marx, & Sloan, 2005; Testa, Livingston, & Collins, 2000). At the present time a limited number of measures exist that have been systematically developed or validated, and used in multiple studies to assess women’s perception of risk for SA (e.g., Hanson & Gidycz, 1993; Marx & Gross, 1995; Messman-Moore & Brown, 2006). The current video measure was developed using an iterative process and tested for evidence of validity in a series of three studies in an effort to provide a new standardized measure for assessing women’s perception of risk for alcohol-related SA.

Researchers have suggested that the validity of measures designed to assess women’s ability to perceive risk for SA lies in the development of extremely realistic scenarios with salient and meaningful cues for the population being studied (Gidycz, McNamara, & Edwards, 2006; Noel et al., 2008). Such suggestions are consistent with a need to develop scenarios that provide a range of cues (i.e., verbal and nonverbal behavioral, environmental), as well as clear and more nuanced risk cues (e.g., subtle or ambiguous; Davis et al., 2009; Norris et al., 1996) that often precede SAs. The majority of scenarios that have been used in SA perception research to date have been written and audiotaped scenarios. These modes of scenario administration may not convey the facial expressions, location of eye-gaze, or touch that occurs; as well as the isolated feeling of the environment, or the number and gender of the other individuals present in the immediate surroundings during a heterosexual social drinking interaction that can precede a SA. In an effort to include a broader range of cues (verbal, nonverbal behavioral and environmental) and potentially allow participants to more fully immerse themselves in the story, we developed our measure using a video scenario recorded from the female character’s first-person perspective. While it remains an empirical question whether the mode of scenario presentation, written, audio, or video, increases validity, it is clear that standardized measures for assessing risk perception need to be developed and assessed for reliability and validity.

To date, the majority of scenarios used to assess women’s perceptions of risk for SA have not been tested for reliability and validity. We are aware of two studies that provide some psychometric properties for their scenarios. Marx and Gross (1995) reported on the reliability of their audio scenario, and Messman-Moore and Brown (2006) indicated that a group of experts provided feedback on the validity of their written scenario. In addition, Messman-Moore and Brown found that college women’s responses to the written scenario (i.e., decreased perception of risk) were associated with an increased likelihood of SA over an 8-month follow-up period. Although some data are available for these two measures, psychometric data are unavailable for most measures used to assess risk perception for SA.

The Current Study

In an effort to develop a standardized measure for assessing young women’s perceptions of risk for alcohol-related SA, we used an iterative development process, as well as assessments of reliability and validity across multiple studies. Compared to other measures that have been used in the past, our measure is unique by incorporating the following: 1) three videos that increase in level of risk cue presentation (low, moderate, and high); 2) subtle and clear verbal, nonverbal, and environmental risk cues for SA in each video; 3) and all videos are recorded from the female character’s first-person perspective. The videos depict a social interaction between a young female and male meeting for the first time at a mutual friend’s party (i.e., a social drinking event).

The videos were developed based on findings from focus group discussions with young, female, heavy drinkers (Study 1), a population at high risk for alcohol-related SA. Focus groups are a technique that is often used to develop new measures and ideas for measures from target populations (e.g., McMahon & Farmer, 2011; Noel et al., 2008). Evidence of construct validity was assessed by randomly assigning women to watch one of the three videos and testing for differences between groups (Study 2a) on impressions of the video (e.g., realism) and perception of risk. We obtained feedback about the believability and consistency of the video measure for assessing risk perception for sexual assault from an expert panel as additional evidence of validity. We revised the videos based on findings from Study 2a and feedback from the Expert Panel. We then reassessed the videos with a smaller sample to determine whether the realism and desired areas of the videos targeted for improvement had changed (Study 2b). Researchers have theorized that alcohol places a woman at increased risk for SA through its pharmacological effects on her cognitive abilities, specifically her ability to perceive risk (Abbey, Buck, Zawacki, & Saenz, 2003; Testa et al., 2000). Thus, we would expect that women, after consuming alcohol, would be less likely to perceive risk for SA. We used the revised low and high risk videos to conduct an experimental alcohol administration study to assess differences in women’s perception of risk for SA after consuming alcohol (.08 BAC) compared to a no-alcohol control group (Study 3) as further evidence of validity.

We had five specific hypotheses: 1) Women assigned to the moderate and high risk video groups would perceive more risk than women assigned to the low risk video group (Study 2a); 2) Women would find the videos realistic, and the realism ratings of the videos would not differ based on video risk level (Study 2a); 3) Additional evidence for validity of the videos would be obtained if our four expert panelists from the field provided consistent responses (i.e., no [unexpected] significant differences among experts or across videos) on measures of realism; and recognition of cues within the videos; 4) After revision women would rate the videos as more realistic (Study 2b); and 5) There would be an interaction between alcohol and video risk level such that, women assigned to the alcohol condition would be less likely to perceive risk cues for SA when watching the high risk video only compared with women in the no alcohol condition (Study 3).

Study 1: Development of the Video Measure

The purpose of Study 1 was to obtain information about women’s sexual experiences (positive and negative) when interacting with men in social drinking situations, in order to facilitate the development of the new video measure for assessing risk perception for SA. Women were asked to describe situations and behaviors that ranged from harmless, but uncomfortable to concerning or even frightening to their sexual safety.

Method

Participants

A total of 42 young (M = 24.9; SD = 2.7) women participated in one of five focus group discussions. All of the procedures for the research described in all three studies were approved by the University’s Institutional Review Board. Participants for all three studies were recruited through advertisements placed in the primary entertainment newspaper (used to advertise local events, music, shows, etc.) and on Facebook. These ads asked for “female, social drinkers, between the ages of 21-30, who date or have sex with men, to call the project if interested in learning more about a paid research study on social interactions.” Women were screened by telephone to determine eligibility for the study. The following general eligibility criteria were used for all three studies. Any additional criteria will be described under Studies 2 and 3. Eligibility requirements included being: 1) between the ages of 21 and 30; 2) reporting drinking in social settings (e.g., at a bar, party, sporting event) 2 or more times each month; 3) not permanently abstaining from drinking alcohol; 4) self-reporting good physical and mental health; and specific to Study 1, 5) being willing to participate in one of the scheduled focus groups. Of the 62 women who called the study and were screened, 87.1% (n = 54) were eligible to participate and 77.8% (n = 42) of eligible women participated.

Two-thirds (66%) of the women were White, 24% Black, 5% of mixed race, 5% of another race, and 10% reported being Hispanic. Most (83.3%) had some college education, with 19% indicating current student status. In addition, 69% worked either full- or part-time over the past year, and reported a median household income within $ 20,000-$30,000.

Procedure

For all three studies, the research was conducted at the Research Institute and all participants provided informed consent prior to beginning participation. The focus groups were facilitated by the Principal Investigator and one of the Co-Investigators on the study. Prior to the discussion, participants were asked to complete an anonymous, brief questionnaire about their history of victimization and their usual drinking patterns. Total participation time for the study was 1.5 hours and women were remunerated $40 for their time. We conducted focus group discussions until data saturation was reached (i.e., no new information was being collected). At the beginning of each discussion, women were asked to maintain the confidentiality of all individuals in the group, advised that they did not have to disclose any experiences as their own, and fully informed of the purpose of the study. All discussions were digitally audio-recorded and later transcribed for qualitative data analysis.

Brief Questionnaire Measures

The questionnaire was designed to take no more than 15 minutes to complete and did not ask for any identifying information. The following measures were included in the questionnaire, as well as in the questionnaires for Studies 2 and 3, and thus, will not be described again.

Demographics

Information on race/ethnicity, marital status, education, work status, and income were collected.

The Daily Drinking Questionnaire

(DDQ, Collins, Parks, & Marlatt, 1985) was used to assess average weekly drinks, as well as frequency of heavy episodic drinking (HED; consumed 4 or more drinks in a 2 hour period) during a typical month. The DDQ has good convergent validity with the longer Drinking Practices Questionnaire (Cahalan, Cisin, & Crossley, 1969; Collins et al., 1985) and has been used extensively with young adult populations (e.g., college students; Corbin, Mcnair, & Carter, 1996).

History of victimization

Childhood sexual abuse (CSA) was assessed using 6 items that asked women whether they had ever experienced a variety of specific sexual behaviors, ranging from exposure to penetrative sex, prior to age 14, from an individual who was not a boyfriend (Miller, Downs, & Testa, 1993). Internal consistency was strong in the current sample (Cronbach’s α = .84). SA history since the age of 14 was assessed using Testa et al’s (2010) 20-item modified version of the Sexual Experiences Survey (SES). This version of the SES assesses five forms of sexual aggression (unwanted contact; attempted intercourse; completed vaginal, anal, or oral intercourse) across four tactics that perpetrators use (verbal coercion, threats of harm, physical force, incapacitation by alcohol or drugs). Validity of this measure has been established (Testa, VanZile-Tamsen, Livingston, & Koss, 2004), and internal consistency was strong for the current sample (α = .84).

Focus Groups

Each focus group began with a clear description of the type of information we were hoping to gain during the discussion (i.e., descriptions of uncomfortable/negative drinking social interactions they knew of other women experiencing or had themselves experienced with men, behaviors that men engaged in socially that made women feel flattered, wary, uncomfortable, concerned, or scared). We also reminded women that regardless of their own behavior or their alcohol consumption that they are never responsible for aggression or violence they experience from a man. We then asked women to describe situations they were “aware of” in which other women had become uncomfortable or frightened during social drinking situations with men. We probed for details on where these situations occurred, how they began and ended, as well as what made these situations and “the male’s” behavior more or less concerning/frightening. We asked women to be specific about which behaviors were clearly frightening and which were concerning or not frightening, but only annoying.

Data Analyses

The questionnaire data were analyzed using descriptive and basic inferential statistics (e.g., ANOVA, Chi-square) to provide a description of the sample. The digital recordings from the five focus groups were transcribed and subjected to qualitative thematic analysis (Krueger, 1994; Parks, Miller, Collins, & Zetes-Zanatta, 1998). During thematic analysis specific questions are assessed for answers that emerge across multiple focus groups. These answers are then organized into “themes.” For example, we were interested in the locations in which these types of interactions with men most often occurred. Thus, locations that were mentioned were recorded by three data coders (i.e., Principal Investigator, Project Director, and a Co-Investigator). These locations were then assessed for commonality across focus groups and coders. A theme was found for a location that occurred in all focus groups and consistently among coders. This procedure also was used for coding of behavioral cues. Lists of behaviors that women found frightening, concerning, and annoying were created for each of the focus groups. Types of behaviors (e.g., shoulder rub, knee touch, breast rub) that were common to most groups (i.e., 4 of 5) were further consolidated into categories of behavior (e.g., touching) where appropriate. Lists of qualifiers that women used to describe the intensity of behaviors or how behaviors made them feel (e.g., concerned, fearful, flattered) were also developed. Behaviors and qualifiers were compared across the three coders. Percent agreement was high (greater than 90%) and all discrepancies were discussed and resolved. The behaviors and qualifiers were used to develop the scripts for the videos.

Results

Quantitative findings from the questionnaires indicated that this sample of women drank heavily and had high rates of SA. Women reported consuming an average of 16.3 drinks (SD = 12.9) weekly, and engaged in HED an average of 7.2 times (SD = 7.6) in a typical month. Not surprisingly given their heavy drinking, rates of prior victimization were very high among the sample, with 61.9% (n = 26) having a history of CSA and 71.2% (n = 30) having experienced adolescent/adult SA as a result of physical force or incapacitation.

Focus Groups

Findings from the thematic analysis of the focus group transcripts indicated that women felt SAs were likely to occur in the following locations: bars when “hanging out;” public places (e.g., subways, events) or sporting events with a lot of people around; and at house parties that involved alcohol, crowding, or people “crashing” overnight. We asked women to describe behaviors that men engage in that cause them to feel “concerned or frightened” for their sexual safety. Five specific behaviors were described in all of the focus groups. These included: 1) unsolicited touching, specifically a lingering, rubbing, stroking touch; 2) men who become aggressive when told “no”; 3) men’s unexpected sexual comments/jokes, or compliments; 4) men encroaching on one’s personal space (following women, cornering them at the bathroom, restricting their movement with his body); and 5) men not responding to cues of disinterest. Women qualified their level of concern or fear with most of these behaviors. The most common qualifier was the degree of isolation the women felt in the situation (i.e., due to crowding, lack of friends, ratio of men to women, etc.), followed by the degree of persistence by the man. When isolation and persistence were high, all of these behaviors, except compliments, were seen as high risk for SA.

Video Development and Description

Based on the thematic analysis findings, three written scripts were developed and three videos were professionally produced, using professional actors. The primary male and female characters were White, while the other party goers were from varied racial and ethnic backgrounds. The videos were recorded from the female’s first-person perspective (i.e., her hand was seen and her voice heard) and depicted a male and a female meeting for the first time at a mutual friend’s party. The videos ranged in length from 149 to 169 seconds and included behavioral (e.g., compliments, touch, eye contact) and environmental cues (e.g., number of friends present, noise level, gender ratio, alcohol cues) at three risk levels (low, moderate, and high). The videos incorporated the behaviors, as well as the qualifiers (e.g., isolation, persistence) that women had specified in the focus groups to establish the three levels of risk and the specific cues. The three scripts remained consistent across risk levels with the exception of the changes in the specific behavioral and environmental risk cues. For example, eye contact changed from the male maintaining eye contact with the woman throughout most of the interaction during the low risk video to the male primarily focusing on the female’s chest and engaging in full scans of her body (i.e., “looking her up and down”) during the high risk video. Environmental cues involved escalation from a quieter, mixed gender party where the female clearly had friends with her in the low risk video to a loud, nearly all male party where the female knew only one female and none of the males in the high risk video (i.e., a louder, more isolated environment, with fewer friends available for support).

In addition, specific alcohol cues were present in the videos. These also escalated in prominence across the three risk levels of the videos. In the low risk video males and females were drinking from red plastic cups and playing drinking games. In the high risk video only males were present and drinking, playing drinking games and drinking from liquor bottles. In addition, the interaction between the male and female around alcohol differed by risk level. For example, in all three videos the male asks the female if he can get her another drink, to which she replies “no thanks, I’m good.” In the moderate and high risk videos the male grabs the female’s cup out of her hand in an effort to get her another drink anyway. His response escalates from a shrug in the low risk video, to “fine” in the moderate video, and an attempt to convince her to get drunk and live it up in the high risk video.

Study 2: Validation of the Video Measure

Study 2 was designed to assess the initial construct validity and realism of the newly developed video measure, using three dependent measures of risk perception and a realism scale with women from the target population. We also obtained evidence for the validity of the video measure from four experts in the field. We then modified and reproduced the videos based on the data collected from the target population and expert feedback and conducted a small follow-up validation with the target population to reassess realism and evidence for validity of the measure.

Method

Participants

Seventy-five young (M = 24.1, SD = 2.6) women participated in the initial validation (2a), while 30 young (M = 23.8, SD = 2.3) women participated in the follow-up validation (2b) phases of Study 2. Participants for both the initial and follow-up validation were recruited and screened for eligibility to participate in the research as in Study 1. For the initial validation, the majority of women who called and were screened for the study were eligible to participate (73.5%, n = 86) and completed the study (86%, n = 74). For the follow-up validation, the majority of women who called and were screened for the study also were eligible to participate (76.2%, n = 32) and completed the study (93.8%, n = 30).

The racial make-up of the initial and follow-up validation samples was similar to the metropolitan area in which the study was conducted. Slightly more than half (51.9%) of the women were White, 34.6% were Black, 8.7% of mixed race, 4.8% of another race. Less than 9% of women reported being of Hispanic ethnicity. The majority of women (84.6%) had some college education, with 23.1% indicating current student status. In addition, 65.4% worked either full- or part-time over the past year, with an average household income within $20,000-$30,000. There were no significant differences in racial or ethnic make-up, marital status, education, work status or income between women who participated in the initial validation and follow-up (n = 30) validation. In addition, there were no significant differences in the demographic characteristics between women who participated in Study 1 and those who participated in Study 2.

Procedures

All procedures for both the initial and follow-up validation were the same. Total participation time for each was 1.5 hours and women were remunerated $40 for their time. Women were randomly assigned to watch one of the three videos (low, moderate, or high risk). Prior to watching the video, women were instructed to imagine themselves as the female character in the video interaction. They were then told to use the space bar on the computer keyboard to indicate any point at which they felt uncomfortable or concerned during the interaction. They were also told that if at any point during the video they wanted to leave the interaction, to indicate this by pressing the enter key. However, they were informed that the video would continue to play after they pressed the enter key and that they should continue to indicate through pressing the space bar any additional points in the video that made them feel uncomfortable.

After watching the video, women were asked to fill out a series of questions about their impressions of the video, followed by standardized measures about their substance use, victimization history, and individual and demographic characteristics. They were then debriefed about the purpose of the study and asked to provide feedback about the procedure and their emotional response to the videos and victimization questionnaires. Women were provided with comprehensive information on support services located in the immediate area.

Measures

Risk Perception

Women’s ability to perceive risk in the videos was assessed using three measures. The first measure was the number of times women indicated, by pressing the space bar on the computer keyboard while watching the video, that they were uncomfortable or concerned during the interaction with the male. The number of key presses was corrected for the length of the video, and a higher number of key presses indicated greater discomfort or greater perception of risk. The second measure was the percentage of women who indicated wanting to leave the interaction in the video, based on the risk level of the video watched. Women could indicate wanting to leave the interaction by pressing the enter key. A higher percentage of women wanting to leave the interaction indicated greater risk perception. The third measure assessed women’s impression of the primary male character in the video using a 22-item semantic differential scale, originally developed by George et al. (1988). Participants were asked “What kind of person do you think the male is? Please rate him on the following characteristics.” Pairs of adjectives were then presented on opposing ends of a 7-point scale. For example: “flirtatious/unflirtatious;” “aggressive/non-aggressive;” “attractive/ unattractive.” Items were reverse coded as necessary and summed to create an overall impression score. Higher scores indicated a more negative impression of the male character. Reliability was excellent for this measure for the initial (α = .90) and follow-up (α = .93) validation phases. These and similar dependent measures have been used to assess risk perception in other studies (e.g.,George, Gournic, & McAfee, 1988; Marx, Calhoun, Wilson, & Meyerson, 2001; Messman-Moore & Brown, 2006).

Impressions of the Video

The realism of the video was assessed using 4 items. These items were measured on a 7-point Likert scale (1 “No/Not at all” to 7 “Yes/Absolutely”). Items asked participants about the realism of the male and female characters, interaction, and their ability to imagine themselves as the female character in the video. For example, “Were you able to imagine yourself as the main female character in the video?”; “How realistic did the male’s behavior/words feel?” These 4 items were summed to create a realism scale. Internal consistency for this scale was good for the initial (α = .68) and follow-up (α = .78) validation.

An additional 8 items were used to assess participants’ own impression of the male character’s attractiveness and perception of his attractiveness to other women, his similarity to their male friends and dating partners, and their comfort and enjoyment when viewing the video and the party situation. For example, “Was the male in the video similar to men with whom you are friends?”; “Was the interaction with the male comfortable?” These items were expected to differ across risk level. These 8 items also were rated on a 7-point Likert scale (1 “No/Not at all” to 7 “Yes/Absolutely”). Because these items were expected to differ based on risk and were not designed to measure one factor (e.g., realism), internal consistency was not calculated. Demographics, Drinking Patterns, and Victimization Histories. These were measured as in Study 1. Internal consistency was good for the victimization measures for both the initial and follow-up validation (CSA: .88, .82; SES: .82, .77).

Expert Evaluations

Four individual experts (two female, two male), who are well-published in the area of SA research and scenario methodology, were asked to compare the three video scenarios quantitatively and to provide qualitative impressions of the videos. They were asked to rate the video using 17 items, on a 7-point scale from 1 ‘Not at All’ to 7 ‘Absolutely.’ Three of the items were not expected to vary across the three video risk levels (e.g., “The interaction was believable”), while the remaining 14 items were expected to vary based on risk level of the video (e.g., “The male character touched the female character”; “The male was aggressive”). We assessed reliability for the 14 items expected to vary and found that Cronbach’s α was .68 for the low risk, .43 for the moderate risk, and .80 for the high risk video. The expert panelists were also asked to describe anything that they “felt was a glaring or critical flaw in the video tapes” as well as anything they “liked or found positive about the video tapes.” We assessed their impressions of the videos to: 1) determine whether they recognized the cues that we had imbedded within the videos, 2) determine how appropriate they felt the cues and social interactions depicted in the videos were for assessing SA risk perception, and 3) to gain insight into any problems with the videos that we might have inadvertently created that would undermine their use as a measure of risk perception.

Data Analyses

Initial Validation

Our aim was to develop three video scenarios that varied in level of risk cues (environmental and behavioral) for SA, from low to high, while maintaining the equivalence of non-risk cues across the three videos. The initial validation (Study 2a) was designed to test this aim. To test hypothesis 1 (H1), we used hierarchical linear regression analyses for the continuous risk perception outcomes (i.e., measures: impression of the male character, number of times the women indicated concern while watching the video [i.e., key presses]) and logistic regression for the dichotomous risk perception outcome (i.e., percentage of women who wanted to leave the scenario [yes/no]). Given evidence from the literature that prior victimization history and alcohol use can influence risk perception, we included CSA and adolescent/adult SA, as well as weekly drinking as covariates. Race also was included as an independent variable (White/Minority). Video risk groups were dummy coded and the low risk group was used as the referent category. To test hypothesis 2 (H2), we conducted Analysis of Variance (ANOVAs) with difference contrasts to compare the women assigned to each of the video risk levels (low, moderate and high) on their impressions of the video, including realism.

Finally, at the end of the study, we asked women to describe anything that felt unrealistic or needed improvement in the videos. These comments were assessed for common qualitative themes that might suggest that the videos were not acceptable to the participants or were not ecologically valid.

Expert Panel

To test hypothesis 3 (H3), the quantitative items evaluated by the expert panelists were individually subjected to ANOVA. Qualitative comments were summarized and assessed for common concerns and suggestions for improvements to the videos.

Follow-up Validation

Using the revised videos, ANOVAs with difference contrasts were used to assess whether we had successfully varied the SA risk cues and held consistent the non-risk cues, across the three video scenarios. We again compared the women assigned to each of the video risk levels (low, moderate and high) on their impressions of the video, as well as the three measures of risk perception and included CSA and adolescent/adult SA as covariates. To test hypothesis 4 (H4), we assessed the change in realism from the initial to the validation study using ANOVA.

Results

Initial Validation

Drinking Patterns and Victimization History

Women reported consuming an average of 10.1 drinks (SD = 7.3) weekly, and engaging in HED an average of 6.8 times (SD = 10.6) in a typical month. Rates of prior victimization were high, with 59.5% (n =44) having a history of CSA and 40.5% (n =30) having experienced adolescent/adult SA. Weekly drinking and adolescent/adult SA history did not differ based on the video risk level to which women were assigned. However, more women with a history of CSA were assigned to the low (47.7%) and high (36.4%) risk compared to the moderate risk (15.9%) video, X²(2) = 15.60, p < .001.

Impressions of Video by Risk Level

The findings for women’s impressions of the videos are presented in Table 1. Women rated all three videos as realistic (M > 4). As hypothesized (H2), there were no differences in women’s ratings of the realism of the video. Women who watched the low risk video were more likely to agree that the male was more like their male friends than women who watched the moderate and high risk videos. Women who watched the low risk video also were more likely to indicate that the interaction felt good, and rated the video as more comfortable and enjoyable to watch compared to women who watched the moderate and high risk videos.

Table 1.

Descriptive Statistics for Impressions of the Videos by Risk Level for the Initial and Follow-up

	Risk Level
Variable	Low (M, SD)	Moderate (M, SD)	High (M, SD)	Eta2
1. Realism of Video
a. Initial	5.54 (1.16)	5.86 (0.92)	5.54 (0.98)	0.02
b. Follow-up	6.23 (0.69)	6.15 (1.00)	5.88 (0.73)	0.04
2. Party like your experiences
a. Initial	6.52 (0.65)	6.17 (1.40)	5.88 (1.74)	0.03
b. Follow-up	5.60 (2.07)	6.00 (1.49)	5.70 (1.49)	0.02
3. Male Attractive to You
a. Initial	2.92 (1.73)	2.67 (1.66)	2.28 (1.40)	0.03
b. Follow-up	3.40 (1.51)	2.50 (1.43)	2.60 (1.58)	0.05
4. Male Attractive to Other Women
a. Initial	4.64 (1.47)	4.04 (1.52)	4.24 (1.45)	0.02
b. Follow-up	5.00 (1.16)	4.60 (1.27)	4.40 (0.70)	0.04
5. Male similar to your friends
a. Initial	4.48 (1.58)a	3.17 (1.99)a	2.72 (1.93)b	0.11
b. Follow-up	4.70 (1.64)	3.60 (1.96)	3.10 (1.66)	0.06
6. Male similar to men you date
a. Initial	3.12 (2.13)	3.17 (2.14)	2.60 (2.04)	0.03
b. Follow-up	3.00 (1.63)	3.10 (2.28)	2.90 (1.45)	<0.01
7. Interaction w/male comfortable
a. Initial	4.04 (1.72)a	2.54 (1.82)b	2.44 (1.98)b	0.13
b. Follow-up	5.10 (1.29)a	3.20 (1.32)b	2.70 (2.41)b	0.19
8. You felt good about Interaction
a. Initial	4.04 (1.34)a	1.71 (1.08)b	1.60 (1.00)c	0.40
b. Follow-up	4.80 (1.69)a	3.10 (1.29)b	1.60 (1.27)c	0.28
9. Video enjoyable to watch
a. Initial	5.88 (1.09)a	4.17 (2.35)b	4.24 (1.76)b	0.12
b. Follow-up	5.10 (1.66)	4.90 (1.10)	4.60 (1.51)	<0.01

Note: Means with different superscripts are significantly different at p < .05.

Values: 1= No/Not at All, 4= Unsure/Maybe, 7 = Yes/Absolutely

Realism Scale: Mean score of 4 items is provided to facilitate interpretation on 1-7 scale. Midpoint is 4.

Risk Perception

The regression analyses for the three risk perception measures are provided in Table 2a. Minority status was associated with reduced risk perception (decreased key pressing), while watching the moderate or high risk video compared to the low risk video was associated with increased risk perception (increased key pressing). This regression accounted for 30% of the variance in the key press measure of risk perception. Watching the moderate or high risk video compared to the low risk video was associated with increased risk perception (increased negative impression of the male). This regression accounted for 38% of the variance in the impression of the male character measure of risk perception. Watching the high risk video increased the likelihood of choosing to leave the scenario by slightly more than 12 times (OR = 12.12, 95% CIs = 2.90, 50.66) compared with the low risk video. Women who watched the moderate risk videos perceived greater risk compared to women who watched the low risk video on two of the three risk perception measures, while women who watched the high risk video perceived greater risk compared to women who watched the low risk video on all three of the risk perception measures. These findings provide support for H1.

Table 2a.

Hierarchical Linear and Logistic Regression Analyses Predicting Risk Perception for the Initial Validation Sample

	Outcome
	Key Presses			Impression of Male			Leave Interaction
Step/Variable(s)	β	R2	Δ R2	β	R2	Δ R2	β	SE	OR
1. Race	−0.28*	0.08	0.08*	−0.19	0.04	0.04	−0.77	0.48	0.46
2. CSA	−0.21	0.15	0.07	−0.28*	0.12	0.08*	−0.45	0.54	0.64
Adol/Adult SA	−0.12			−0.07			0.01	0.53	1.01
3. Weekly drink	−0.12	0.16	0.01	−0.02	0.12	0.00	−0.02	0.04	0.98
4. Moderate  Risk Group	0.52***	0.35	0.20***	0.55***	0.43	0.31***	1.07	0.81	2.90
High Risk  Group	0.45***			0.63***			2.50	0.73	12.12***
	Total Adj. R2 = 0.30			Total Adj. R2 = 0.38

^*Note: p < .05,

^**p < .01,

^***p < .001. Race is dichotomous (0 = White, 1 = Minority). The video risk groups were dummy coded and the referent group for the analyses shown in the table was the low risk group. Additional analyses with the moderate group coded as the referent group indicated that the moderate and high risk groups were not significantly different for key presses (β = −0.08, p = ns) or impression of the male (β = 0.07, p = ns); while the moderate and high risk groups were significantly different (β = 1.43, s.e. = 0.69, OR = 4.18, p < .05) for leaving the interaction.

We reran the regressions to determine whether perception of risk for the moderate video differed from perception of risk for the high risk video. We found that there were no differences between the moderate and high risk videos for key presses and women’s impression of the male character. However, the likelihood of leaving the scenario was slightly more than 4 times higher for women who watched the high risk video (OR = 4.18, 95% CIs = 1.08, 16.09) compared to women who watched the moderate risk video. This finding provides partial support for H1, indicating that for one of the measures women indicated greater perception of risk when watching the high risk compared with the moderate risk video.

Expert Panel Ratings

Quantitative

The ratings of the expert panelists on the 17 items are provided in Table 3. We expected differences based on risk level of the video for 14 of the 17 items. The expert panel did not differ in their ratings for the three items we did not expect to differ: Item 1, the video was shot from the first person perspective; Item 14, the female’s behavior was neutral; and Item 15, the interaction was believable. Among the 14 items we expected to differ, 10 were significantly different, four were not significantly different. The four that were not different suggested specific areas within the videos that needed revision to delineate differences in the three levels of risk. These areas included: the execution of the touch cues, the portrayal of isolation, and the male asking for personal information. Expert panel findings indicated that the videos were consistent in the presentation of non-risk cues across the three levels of risk, thus supporting H3.

Table 3.

Expert Panel Impressions of the Videos

	Risk Level
Variable	Low (M, SD)	Moderate (M, SD)	High (M, SD)	Eta2
1. Video in 1st person perspective	7.00 (0.00)	7.00 (0.00)	7.00 (0.00)	-
2. Male encroached on female	3.25 (1.50)a	5.25 (0.96)	6.75 (0.50)b	0.71
3. Male was offensive	1.75 (0.50)a	4.25 (0.96)b	6.75 (0.50)ab	0.92
4. Male touched female	2.33 (1.53)	6.00 (1.73)	6.00 (1.73)	0.62
5. Female does not know anyone	1.25 (0.50)a	3.25 (0.96)b	6.50 (0.58)c	0.93
6. Female rejects male touch	6.50 (0.58)	5.50 (1.29)	5.25 (2.06)	0.16
7. Male respects female’s space	5.25 (0.96)a	3.25 (0.50)b	1.50 (0.56)c	0.86
8. Male is nervous	2.57 (1.71)a	2.00 (0.82)b	1.00 (0.00)b	0.36
9. Male is aggressive	2.50 (1.00)a	4.50 (1.00)b	6.75 (0.50)c	0.84
10. Male spoke politely	6.00 (0.82)a	4.00 (0.82)b	1.50 (0.58)b	0.89
11. Male made appropriate  compliments	4.50 (1.29)a	3.00 (1.63)b	1.00 (0.00)b	0.65
12. Male looked in female’s eyes	6.25 (0.96)a	3.75 (0.96)b	2.50 (0.58)b	0.82
13. Male asked for personal information	4.75 (1.71)	5.50 (1.73)	6.00 (1.41)	0.12
14. Female’s behavior was neutral	4.75 (1.71)	5.00 (1.41)	4.75 (1.50)	0.01
15. Interaction was believable	5.25 (0.96)	6.00 (0.82)	5.75 (0.50)	0.18
16. Situation felt isolated	2.25 (1.50)	2.75 (1.71)	4.00 (2.45)	0.16
17. Potential for situation to end badly	2.25 (0.50)a	4.25 (1.89)b	5.50 (1.29)b	0.57

Note: Means with different superscripts are significantly different at p < .05.

Values: 1= No/Not at All, 4= Unsure/Maybe, 7 = Yes/Absolutely

Shaded Items: These were items we expected to differ based on video risk cue level and thus, suggested potential areas needing modification during revision of the videos.

Qualitative

The qualitative comments from the experts corroborated their quantitative ratings and provided feedback on areas of the scenarios that they felt needed to be modified to improve the ecological validity of several specific risk cues and the videos overall. For example, they felt the atmosphere in the videos was not effective at portraying a true house party (e.g., not dark enough, not loud enough, not a large enough or ethnically diverse enough crowd). They felt that the conversational flow between the two main characters needed to be more natural, and that the acting of the touch cues in each scenario needed to be done in such a way as to be more clear or obvious. Most of the cues were recognized by all of the experts, with the exception of precedence for the main female character knowing the male character through a mutual friend. In addition, they indicated that the later part of the conversation, in which personal information was exchanged, was hard to understand, as the female appeared to be mumbling.

Qualitative feedback from the female participants that was relevant to modifying the video scenarios was minimal and indicated that they were not concerned about the issues pointed out by the expert panelists. Specifically, a number of women commented on the male character being a “friend of a friend,” and how that made her trust him more. Women did not mention the ethnicity of the male or female characters, typically only commenting on their negative feelings or opinions of the male’s behavior (e.g., “he was creepy” or “a jerk”).

Modification of the Videos

Based on the initial validation phase data and the expert evaluations, it was determined that there were several areas in which the videos could be improved to differentiate risk level, between the moderate and high risk videos and improve the overall ecological validity of the three videos. Therefore, a second set of videos was produced. The verbal scripts for the second set of videos were not altered. Several of the changes to the videos were aesthetic, these were done to increase the generalizability and ecological validity of the videos and included moving to an apartment that was larger, could accommodate more people, and had more flexibility in lighting. We further increased the ethnic diversity of the party goers, and improved the conversational flow and articulation of the scripted words between the primary male and female characters. In an effort to increase the differentiation between the moderate and high risk videos, we modified several environmental and behavioral risk cues. We increased the sense of isolation by changing the male to female ratio and noise level from low to high risk (escalation to more men than women, and louder music). In addition, we improved the execution of the touch and approach cues, so that the touch by the male was more realistic and the female’s withdrawal was more obvious. We then conducted the follow-up validation to determine whether the changes had improved the differentiation between the moderate and high risk videos, while maintaining the realism and evidence for validity of the measure found in the initial validation.

Follow-up Validation

Drinking Patterns and Victimization History

Women reported consuming an average of 10.2 drinks (SD = 7.3) weekly, and engaging in HED an average of 5.1 times (SD = 9.0) in a typical month. Rates of victimization were high among the sample, with 50.0% (n =15) having a history of CSA and 40.0% (n =12) having experienced adolescent/adult SA. Weekly drinking, HED, CSA, and rates of adolescent/adult SA did not differ from women who participated in the initial validation. Women also did not differ in their weekly drinking rates or histories of CSA and adolescent/adult SA based on video risk level assignment.

Impressions of Video by Risk Level

We assessed the impressions of the revised videos for the follow-up sample to determine how the three videos were similar and different across the three risk levels, and to determine how their ratings were similar or different from the initial validation sample. The findings for the follow-up sample are presented in Table 1. Some of the findings (i.e., realism, party similar to previous experience, interaction was comfortable, watching video felt good) were similar to those from the initial validation, and most of the means trend in a similar direction as those from the initial validation. A comparison of realism ratings from the original videos (Initial validation) to the revised videos (Follow-up validation) indicated that the realism rating increased significantly (F(1, 102) = 4.40, p < .05). The increase in realism from the initial videos to the revised videos supports hypothesis 4 (H4).

Risk Perception

The regression analyses for the follow-up validation are provided in Table 2b. Minority status was associated with reduced key pressing (i.e., decreased risk perception). Watching the high risk video compared to the low risk video was associated with increased key pressing (i.e., increased risk perception). This regression accounted for 40% of the variance in the key press measure of risk perception. Watching the moderate or high risk video compared to the low risk video was associated with a more negative impression of the male (increased risk perception). This regression accounted for 53% of the variance in the impression of the male measure of risk perception. We reran the regressions to determine whether perception of risk for the moderate video differed from perception of risk for the high video. We found that perception of risk for women who watched the high risk video was associated with increased risk perception compared to women who watched the moderate risk video (β = 0.43, p < 0.05) for the key press outcome measure. However, there was no difference in risk perception between women who watched the high compared to the moderate risk videos for the impression of the male outcome measure. A chi-square indicated that there was a significant difference between the percentage of women who chose to leave the interaction when watching the moderate (30%) compared to the high (80%) risk (X²(2) = 14.07, p < .001) video, and presumably the low (0%) risk video.

Table 2b.

Hierarchical Linear Regression Analyses Predicting Risk Perception for the Follow-up Validation Sample

	Outcome
	Key Presses			Impression of Male
Step/Variable(s)	β	R2	Δ R2	β	R2	Δ R2
1. Race	−0.44*	0.19	0.19*	−0.30	0.09	0.09
2. CSA	−0.37	0.28	0.09	−0.27	0.41	0.33**
Adol/Adult SA	0.17			0.60**
3. Weekly drink	0.40*	0.41	0.12*	0.17	0.44	0.02
4. Moderate  Risk Group	0.03	0.53	0.12	0.49**	0.63	0.19**
High Risk  Group	0.45*			0.52**
	Total Adj. R2 = 0.40			Total Adj. R2 = 0.53

^*Note: p < .05,

^**p < .01,

^***p < .001. Race is dichotomous (0 = White, 1 = Minority). The video risk groups were dummy coded and the referent group for the analyses shown in the table was the low risk group. Additional analyses with the moderate group coded as the referent group indicated that the high risk group was different than the moderate for key presses (β = 0.43, p < 0.05); however the high risk group was not different from the moderate (β = 0.03, p = ns) for impression of the male.

Overall, these findings suggest that the second version of the videos (viewed by the follow-up sample) maintained the consistency and validity of the original videos as a measure for assessing women’s ability to perceive differences in risk for SA. In addition, there was greater distinction between the moderate and high risk videos, suggesting a general improvement in the specific risk cues in the videos. However, given the small sample size we did not have the power to detect small to moderate effects.

Study 3: Alcohol Administration Study

The purpose of Study 3 was to assess changes in women’s ability to perceive SA risk cues using the new video measure under conditions of acute alcohol intoxication. In order to assess these changes, we conducted an alcohol administration study using a mixed within-subjects design (2 video risk cue levels [low/high] × 2 alcohol conditions [no alcohol/alcohol]). This design allowed us to use the new video measure as a standardized presentation of risk cues for SA, as well as to standardize the dose of alcohol women consumed. All participants viewed both the low and high risk cue videos (within-subjects variable) and were randomly assigned to one of the two alcohol condition (between-subjects variable; no alcohol/alcohol).

Method

Participants

Sixty young (M = 24.4, SD = 2.7), women participated in the alcohol administration study. Participants for Study 3 were recruited, as for Studies 1 and 2, through advertisements in an entertainment newspaper and on Facebook. The wording for the advertisements remained the same as the other studies, however, the screening procedure and eligibility criteria changed given that women might be consuming a dose of alcohol equivalent to four standard drinks. Women were told that they might be consuming alcohol, the duration of time they could be required to be at the Institute if they consumed alcohol (5-6 hours), and payment for the study ($15 per hour). They were eligible to participate in the study if they met the original criteria for Studies 1 and 2, as well as the following additional criteria: 1) they could not be pregnant or trying to become pregnant; 2) they had to be willing to take a urine-based pregnancy test immediately prior to participating in the study; 3) they had to have consumed a minimum of 4 drinks in one drinking episode at least once in the past month; 4) they could not have a current drinking problem as determined using the short form of the Michigan Alcoholism Screening Test (sMAST; Selzer, Vinokur, & Rooijen, 1975; Taylor, James, Bobadilla, & Reeves, 2008); 5) they could not currently have a drug problem as determined using the short form of the Drug Abuse Screening Test (DAST-10; Skinner, 1982); 6) they could not report any negative physical side effects as a result of consuming alcohol; 7) they had to weigh less than 200 pounds, and 8) they could not score at a clinical level for anxiety or depression on the subscales of the Brief Symptom Inventory (BSI-18; Derogatis, 1993). These additional criteria were important for minimizing the potential negative side effects that could occur from ingesting the equivalent of four standard drinks if a woman was randomly assigned to the alcohol condition.

Given the more stringent eligibility criteria for Study 3, only 37.2% (n = 89) of the 239 women who called the project were eligible to participate in the study. In addition, 9.2% (n = 22) of women decided they were not interested in participating, 4.2% (n = 10) failed to show up for their appointment, one dropped out, and three were removed because of data irregularities (i.e., missing, inconsistent, one arrived intoxicated). Thus, a total of 60 women participated.

The racial composition of the sample for Study 3 was not significantly different than the racial composition of the sample for Study 2; 62.7% of women were White, 22% Black, 8.5% Asian-Pacific Islander or Native-American, and 6.8% were of mixed race. Less than 4% of women reported being Hispanic. Marital status (93.2% single), education (81.7% some college), work status (58.3%), as well as average household income (within $20,000-$30,000) did not differ significantly from Study 2 participants.

Beverage Administration

Women in the alcohol condition consumed vodka and tonic with lime. Women in the no alcohol condition consumed tonic with lime. The amount of alcohol women received was calculated based on the formula provided by Curtin and Fairchild (2003), using 40 proof vodka mixed at a ratio of 1 ml alcohol to 3.5 ml of tonic to achieve a target BAC of .08. Women in the no alcohol condition received the equivalent amount of tonic as women of similar weight received of alcohol and tonic in the alcohol condition.

Procedures

Women participated in the study, individually, in the bar laboratory at the Research Institute. Women were asked to abstain from drinking alcohol or consuming drugs for 24 hours before their participation. Prior to participating in the study, all women were asked to provide informed consent. They also were required to take a urine-based pregnancy test to insure that they were not pregnant prior to alcohol administration. Prior to all study procedures, women were breathalyzed to ensure a starting BAC of 0.0. Women were randomly assigned to one of the two alcohol conditions and told upon arrival of their alcohol assignment. They were weighed to determine the amount of alcohol and mixer required to achieve the desired BAC (.08) or the amount of liquid equivalent they needed to consume for the no alcohol condition. Women were seated at the bar to consume their beverages over a 20-minute period (i.e., for both conditions). The measures for Study 3 remained the same as for Study 2. Following beverage consumption, during the 20 minute absorption period women in both alcohol conditions were seated at a computer in the bar laboratory to complete a series of standardized questions about their demographics, health, and drinking and drug use. Following the 20 minute absorption period, women were breathalyzed. If women in the alcohol condition were at or near .08 BAC, they began watching the first video. If they were not near the desired BAC, they were breathalyzed again at 5-minute intervals twice more to determine whether they were near .08 BAC. After the additional 10 minutes of absorption, all women began the protocol.

The within-subject design of the study allowed us to control for participants’ individual differences in risk perception. Women were only presented with two of the three videos. We chose to present the low and high risk videos because we found clear and consistent differences between women’s perception of risk between the low and high risk videos during Study 2. The videos were presented to all women in the same order – first the low risk then the high risk – to reduce the likelihood that the cues depicted in the high risk video would influence women’s opinions of the male character in the video and potentially increase their risk perception overall. After each video, women responded to the same measure used in Study 2, and were breathalyzed. The instructions for viewing the videos and the use of the keyboard to indicate concern and desire to leave the scenario also remained the same as for Study 2.

As for Study 2, after watching both videos, women were asked to fill out a series of standardized measures about their own victimization history and provided with the opportunity to ask questions about the study. Women who participated in the alcohol condition were provided with food and a comfortable space to eat and relax until their BAC dropped below .02. Women in the control condition were debriefed and released. All participants were provided with a ride to and from the Research Institute in a cab paid for by the study.

Debriefing

Debriefing for Study 3 was similar to Study 2 and included an additional assessment of participants’ physical and emotional symptoms and current level of intoxication prior to leaving the bar laboratory. As in Study 2, participants were asked to provide feedback about the procedure and their emotional response to the videos and victimization questionnaires. Any women who indicated feeling upset, concerned or traumatized by the video or questionnaires were provided with an opportunity to speak with the PI or a licensed clinical psychologist at the Institute. None of the participants indicated that they were feeling upset or needed to talk about the study with study personnel. In addition, none of the women indicated greater than mild physical symptoms as a result of alcohol consumption.

Data Analyses

To test hypothesis 5 (H5), we conducted mixed repeated measures ANOVAs with risk cue level (low vs. high) of the video as the within-subjects measure and alcohol condition (no alcohol vs. alcohol) as the between-subjects measure. This analysis was conducted for each of the three dependent measures of risk perception (impression of the male character, number of times the women indicated concern while watching the video [i.e., key presses], and percentage of women who indicated wanting to leave the scenario). Simple effects were assessed with difference contrasts for all significant interactions.

Results

Women in Study 3 reported an average of 11.1 drinks (SD = 8.6) weekly, and engaged in HED an average of 6.1 times (SD = 7.6) per month. These rates are not significantly different than those reported for the women in Study 2. The rate of women reporting a history of CSA (38.3%) was significantly lower (X²(2) = 8.99, p < .01), while the rate of women reporting a history of adolescent/adult SA (53.3%) was not significantly different compared to women in Study 2. Weekly drinking, HED, CSA, and adolescent/adult SA rates did not differ by alcohol condition.

Blood Alcohol Concentration (BAC)

The average BAC for women in the alcohol condition prior to watching the low risk video was .071 (SD = .019), prior to watching the high risk video was .072 (SD = .016), and after watching the high risk video was .075 (SD = .016). These BACs are not significantly different (F (2, 58) = 2.78, p = ns) and indicate that the alcohol dosing protocol was successful in achieving a moderately high level of intoxication.

Risk Perception

The means and standard deviations are provided in Table 4a and repeated measures ANOVA results in Table 4b. Significant interactions between video risk level and alcohol condition were found for key presses and for women’s impression of the male character. This interaction was not found for the measure of women wanting to leave the interaction. As hypothesized (H5), women in the alcohol condition pressed the space key significantly less and had a significantly more positive impression of the male character (lower mean) than women in the no alcohol condition when viewing the high risk video only (see Table 4a). This difference was not observed during the viewing of the low risk video – resulting in the interaction.

Table 4a.

Alcohol Experiment Means and Standard Deviations for each Outcome

	Outcome
	Key Presses		% Leaving		Impression of Male
Risk Level of Video:	Low	High	Low	High	Low	High
Alcohol Condition
No Alcohol	0.5 (0.8)	8.3a (5.6)	6.7 (25.4)	66.7 (47.9)	3.7 (0.6)	5.9a (0.6)
Alcohol	0.4 (1.4)	3.8b (3.3)	6.7 (25.4)	46.7 (50.7)	3.6 (0.6)	5.4b (0.7)

Note: Significant simple effects within alcohol by risk interactions are indicated by different superscripts. Lower means for “Impression of the Male” indicate reduced risk perception; thus, higher value for all outcomes indicates greater risk perception.

Table 4b.

Alcohol Experiment Mixed Repeated Measures ANOVA Results for each Outcome

	Outcome
Source of Variation	Key Presses		% Leaving		Impression of Male
	F	Eta2	F	Eta2	F	Eta2
Between Subjects
Alcohol Condition	7.54**	0.14	0.23	<0.01	3.28	0.06
Within Subjects
Risk Level	36.14***	0.66	14.15***	0.26	162.37***	0.75
Risk Level ×  Alcohol Condition	12.28**	0.22	0.61	0.01	4.52*	0.08

Note: df = (1, 58);

^*p < .05,

^**p < .01,

^***p < .001

Discussion

The primary aim of this research was to develop a new measure for assessing women’s perceptions of risk for alcohol-related SA, which included three videos that increased the risk level of ambiguous and clear cues for SA (low, moderate, high). We successfully developed the measure using information collected through focus groups with young female heavy drinkers. These young women provided detailed descriptions about behavioral and environmental cues that they experienced during heterosexual social interactions that made them feel annoyed, uncomfortable or frightened for their sexual safety.

Establishing the reliability and validity of a new measure is an ongoing process. We began this process through the use of an expert panel and two validation studies. We found evidence for the reliability, realism, and the convergent construct validity of the original videos and the revised videos during the initial and follow-up phases of Study 2. We found support for our first hypothesis (H1), women assigned to the different video risk groups perceived increased risk on the three outcome measures of risk perception (i.e., key presses indicating discomfort, leaving the scenario, impression of the male) during both the initial and follow-up validation. During the initial validation women who viewed the moderate and high risk videos perceived greater risk than women who viewed the low risk videos; however, the moderate and high risk videos were not found to be significantly different from each other. The moderate and high risk videos were found to be different from each other after revision (i.e., follow-up validation). Thus, we believe that the revision of the videos provided a clearer distinction of the difference between the portrayal of the moderate and high risk cues. Clearly, additional studies with larger samples are needed to determine whether all three videos, low, moderate, and high are distinct from each other. In further support of validity, the responses from our experts in the field were consistent, that is, they all found the video, characters, and scenario, to be a reasonable depiction of a heterosexual drinking interaction with the potential to become a SA. In addition, as hypothesized they were able to identify the subtle and clear risk cues for SA (H3). As hypothesized women found all three videos to be equally realistic (H2), and in support of H4 the realism significantly increased suggesting that the revised videos were an improvement over the original videos.

The experimental alcohol administration study was used to provide additional evidence of the convergent construct validity of the measure by assessing the influence of acute alcohol intoxication on women’s perception of risk when presented with the low and high risk videos. We hypothesized an interaction between alcohol and video risk level on risk perception such that women assigned to the alcohol condition would be less likely to perceive risk cues for SA when watching the high risk video only compared to women in the no alcohol condition (H5). This hypothesis was supported for two of the outcome measures, key presses and impression of the male character, indicating that alcohol attenuated women’s ability to perceive risk cues. Although women were less likely to indicate a desire to leave the high risk scenario after drinking, this difference was not significant. Given the moderate to large main effects for alcohol for the two significant outcomes, this lack of a significant interaction was most likely due to our small sample size and a lack of sufficient power for detecting an interaction of small to modest effect size.

Limitations

It should be noted that our findings represent the development and initial tests for evidence of validity of a new video measure for assessing women’s risk perception for alcohol-related SA; therefore, more extensive testing of the measure with larger samples is needed. Our findings across these studies provide preliminary evidence to support the convergent validity of the measure for assessing women’s perception of risk or SA. Only one set of videos was developed using one primary female and one primary male character of White, non-Hispanic ethnicity. Although not hypothesized, racial differences were found for one of the risk perception outcome measures (i.e., key presses). It was the case that all of the women assigned to the high risk video during the initial validation who had experienced CSA were African American or of mixed-race. Therefore, CSA history may have confounded the effect of ethnicity on risk perception as measured by key presses. Clearly, additional studies using the current videos are needed to assess the potential for different cultural impressions of the risk cues (i.e., the way that the male complimented the female, the way touch was perceived). Future work is planned to develop additional sets of videos with African American actors and Hispanic/Latino actors following exploration of the cultural interpretation of the current videos (i.e., the language used in the scripts, meaning of touch, nonverbal cues, etc.) through focus groups with different ethnic groups of women. Our participants were heavy drinkers with substantial histories of SA; this group of women is at highest risk for SA and is the group that would be most likely to benefit from SA preventive intervention programs. Our findings may not be generalizable to all young women and thus, additional studies with more diverse young female populations are needed to obtain evidence for the validity of the measure. It should be noted that all modes of scenario presentation have advantages and disadvantages. Videos are limited by presenting a specific set of characters and scenes, if these are not representative of participants’ experiences, validity may be compromised. However, videos provide a strong format for presenting behavioral and environmental cues observed during social interactions, and thus provide a strong ecologically valid alternative to in-person role-play techniques.

Research Implications

Our findings suggest that this is an acceptable measure, with good preliminary evidence of consistency and construct validity for evaluating young, female, heavy drinkers’ ability to perceive ambiguous and clear risk cues for SA. Future studies should include validation with different populations of women (e.g., college women, urban, rural, etc.) and larger samples. Although there is no “Gold Standard” with which to compare this measure, it may be worth comparing responses on this measure to more standardized measures of perception such as the Perception of NonVerbal Sensitivity Test (PONS; Rosenthal, Hall, DiMatteo, Rogers, & Archer, 1979), and the Diagnostic Analysis of Nonverbal Accuracy, (DANVA; Nowicki & Duke, 1994), as well as future likelihood of SA. Finally, the measure is easily disseminated, requires minimal training to use, and is generalizable to heavy drinking women at high risk for alcohol-related SA.

Clinical and Policy Implications

In a previous experimental alcohol administration study, we found that women in a high dose alcohol condition (.08 BAC) were less likely to perceive risky behavioral cues enacted by a male they had just met and were more likely to see relationship potential with this same male compared with women in a low dose alcohol condition (Parks, Hequembourg, & Dearing, 2008). We found additional differences based on women’s histories of CSA, such that women with histories of CSA were less likely to notice the risk cues and more likely to view the male positively. The laboratory paradigm was labor intensive in its use of male confederates, who require extensive training and continued testing for reliability. The new video measure could be substituted in laboratory paradigms to provide a standardized scenario for use in intervention programs designed to increase women’s awareness of the changes that can occur in their perception of risk as a result of the impairing effects of alcohol or as a result of previous sexual traumas. The videos are unique in escalating from low to moderate to high risk, and through each including subtle, as well as clear risk cues for SA, therefore, they provide the opportunity for use as an intervention tool for recognizing early (i.e., more ambiguous) precursors to SA, that often occur during social drinking interactions. As with all research of this nature, the victim is never to blame for her behavior or her victimization. It is always the responsibility of the perpetrator. However, it is our hope that with greater knowledge, we can help empower women so that their risk for SA is minimized.