Preventing College Sexual Victimization by Reducing Hookups: A Randomized Controlled Trial of a Personalized Normative Feedback Intervention

Abstract

Sexual activity, including hooking up, increases college women’s vulnerability to sexual victimization. Reducing hookups may reduce rates of sexual victimization among this vulnerable population. Because college students overestimate how frequently their peers hook up, correcting their misperceptions may lead to more accurate perceived social norms, and consequently, less hookup behavior. The study was designed as a randomized controlled trial of the efficacy of a brief, computer-administered personalized normative feedback (PNF) intervention regarding hookups during the first semester of college. We tested an indirect effects model in which PNF was hypothesized to predict perceiving fewer peer hookups, which were expected to predict fewer actual hookups and consequently, less sexual victimization during the first semester of college. Entering first-year women (N = 760) were randomly assigned to receive web-delivered PNF or no information. At the end of the semester, perceived number of hookups of others, number of hookups during the semester, and sexual victimization experiences were assessed. Women who received the intervention perceived that their peers engaged in significantly fewer hookups than did control women. Consistent with the proposed indirect effects model, intervention had a significant indirect effect on the odds of first semester victimization via lower perceived descriptive norms, which in turn predicted fewer hookups. The study provides proof of concept for the importance of hookups as a risk factor for sexual victimization and provides novel, preliminary support for intervention to change descriptive norms as a way of reducing hookups and consequently, sexual vulnerability.

College sexual assault is a prevalent public health issue, with about 1 in 5 women reporting such an experience during college (Muehlenhard et al. 2017). Development of efficacious prevention programs depends upon the ability to identify the risk factors that contribute to sexual vulnerability and to alter these pathways. Sexual activity, including hookups (i.e., sexual encounters without a commitment or relationship, Paul et al. 2000), increases vulnerability to sexual assault (e.g., Fielder et al. 2014). Hence, decreasing the number of hookups may reduce sexual victimization. In the present study we examined whether women randomly assigned to receive personalized normative feedback (PNF) regarding hookups at T1 would report lower hookup norms at T2, which would in turn predict fewer T2 hookups, and lower incidence of T2 sexual victimization.

Behavioral Risk Factors Associated with College Sexual Victimization Vulnerability

Although all college women are vulnerable to sexual victimization, some women are at elevated risk. Adolescent sexual victimization increases risk of revictimization in college (Carey et al. 2015), an association that reflects, at least in part, higher levels of drinking and sexual activity among previously victimized women (Kelley and Gidycz 2017; Neilson et al. 2018; Testa et al. 2010). Drinking and casual sexual activity are also strongly associated (see Claxton et al. 2015 for a meta-analytic review). Although heavy episodic drinking (HED) has received a great deal of attention as a risk factor for college sexual victimization (see Testa and Livingston 2018 for a review), casual sexual activity and specifically, hookups, may also serve as an important proximal risk factor for experiencing sexual victimization. Hookups are sexual encounters between strangers, friends, or acquaintances (people not in a relationship) that may or may not include intercourse (Paul et al. 2000). Greater involvement in hookups is associated cross-sectionally and longitudinally with sexual victimization (Fielder et al. 2014; Sutton et al. 2019; Testa et al. 2019a) and most college sexual victimization events occur within the context of a hookup (Flack et al. 2007; 2016). In describing their worst hookup experience, college women were much more likely than men (43% vs. 10.5%) to report that it included forced sexual behavior (Paul and Hayes 2002).

Hookups may be particularly risky because underlying norms and definitions are ambiguous, increasing the likelihood of misperceiving the other’s sexual interest (Lewis et al. 2013; Lovejoy 2015). Greater sexual misperception on dates is associated with sexual victimization on those dates (Abbey et al. 2001; Muehlenhard and Linton 1987) and dates that ended in sexual aggression compared to non-aggressive dates were more likely to include sexual activity (Harrington and Leitenberg 1994; Yeater et al. 2008). These findings suggest that initially consensual sexual behavior may be misinterpreted as indicating a desire for intercourse, leading to sexual advances that are unwanted by the recipient. College men reported using higher levels of sexually aggressive tactics in sexual encounters with new partners than in encounters with previous sexual partners (Testa et al. 2015). Misperception may be exacerbated by the use of alcohol, which is present in the majority of hookups (Fielder and Carey 2010; LaBrie et al. 2014). Alcohol increases men’s sexual misperception (Farris et al. 2010) and sexual aggression within sexual encounters (Testa et al. 2019b) and increases women’s sexual vulnerability by impairing the ability to recognize risk cues and respond effectively to unwanted sexual advances (Davis et al. 2009; Melkonian and Ham 2018). Greater consumption of alcohol prior to hooking up increases the likelihood that the hookup culminates in rape (Ford 2017). For these multiple reasons, engaging in more frequent hookups is likely to increase women’s vulnerability to experiencing sexual victimization.

The Role of Social Norms in College Risk Behaviors

The transition from high school to college involves increases in drinking and sexual activity for many students (Fromme et al. 2008). Interpersonal processes and perceived social norms have been shown to play an important role in alcohol use (Borsari and Carey 2001). Students typically believe that other students drink more than they actually do and these inflated norms subsequently predict heavier drinking (Kypri and Langley 2003; Read et al. 2002). Although there has been less research focused on the role of social norms and normative misperceptions of sexual behavior, similar processes appear to operate. College students tend to overestimate the sexual activity of others (Barriger and Vélez-Blasini 2013; Holman and Sillars 2012; Lambert et al. 2003) and these perceived social norms positively predict one’s own risky sexual behavior (Lewis et al. 2007a; 2014a).

Consistent with the importance of descriptive social norms in shaping one’s own behavior, interventions that involve altering social norms have emerged as an efficacious means of changing behaviors (Miller and Prentice 2016) and have been used to reduce HED in college and other populations. Interventions designed to alter normative perceptions of HED, primarily via personalized normative feedback (PNF), do in fact change these perceptions and consequently result in lower frequency of HED, although effect sizes are typically small (see Prestwich et al. 2016; Sheeran et al. 2016 for reviews). Consistent with the proposed mechanism underlying PNF, meta-analysis has identified descriptive social norms as the mechanism underlying the effect of PNF on subsequent drinking (Reid and Carey, 2015).

Although PNF has been used successfully for many years to reduce drinking via altering social norms, few studies have used such an approach to change social norms around risky sexual behaviors. Yet, normative perceptions about others’ sexual behavior contribute to risky behavior (Lewis et al. 2007a; 2014a) and sexual vulnerability. For example, the effects of prior sexual victimization on college victimization were partially mediated by social norms regarding hookups (Norris et al. 2018). Importantly, studies that have tested the proposed mediating mechanism provide evidence that providing PNF regarding alcohol-related sexual risk behaviors reduces subsequent risk behaviors via its effects on perceived social norms for sexual behavior (Lewis et al. 2014b; 2018). Because normative perceptions regarding others’ hookups and risky sexual activity are associated with one’s own sexual activity, correcting the misperception that others are engaging in more hookups than they actually are may serve to reduce hookups and, consequently, vulnerability to sexual victimization.

Preventive Intervention Early in College

First-year college women are at elevated risk of sexual victimization relative to older students (Cranney 2015). Consistent with Event Specific Prevention (Lee et al. 2014; Neighbors et al. 2007; 2012), the present study was designed to reduce harm during this known window of risk. Provision of PNF regarding hookups was believed to be particularly well-suited to intervention during the early weeks of college. Because social norms regarding hookups and other behaviors are just developing during this transition period, they are likely to be malleable. Accurate information about others’ sexual activity is likely to be even more lacking than accurate information about others’ drinking, since sex, unlike drinking, typically occurs in private.

The present randomized controlled trial was designed as an initial test of the hypothesis that providing college freshmen women with accurate social norms regarding the prevalence of hooking up among other first-year women at that university will reduce the number of hookups and consequently, lower the prevalence of sexual victimization. We hypothesized that women who were provided with PNF showing the actual number of hookups of other first semester freshmen women would subsequently report: 1) lower perceived descriptive hookup norms, 2) fewer hookups during the first semester, and 3) lower rates of sexual victimization during the first semester. Consistent with our theoretical framework, we tested the hypotheses via an indirect effects model (see Figure 1).

Figure 1.

Indirect Effects of PNF Intervention on Sexual Victimization

*** p < .001, ** p < .01, * p < .05

Path coefficients (b and standard error, S.E.) are unstandardized.

Note. Model controlled for potential effects of T1 victimization and T1 HED on T2 perceived hookups, T2 hookups and T2 victimization and for the effect of living with parents on victimization.

Because both prior sexual victimization and HED are robustly associated with sexual victimization (Messman-Moore and McConnell 2018; Testa and Livingston 2018), we controlled for the effects of both variables on the occurrence of first semester sexual victimization. Some prior sexual victimization interventions have been more (Gilmore et al. 2015) or less efficacious (Clinton-Sherrod et al. 2011) for women who were previously victimized or for heavier drinkers (Rothman and Silverman 2007). Thus, to consider whether the effects of the intervention differed for women who had previously been sexually victimized or who were heavier drinkers, we tested whether intervention interacted with either prior victimization or HED.

Method

Participants and Procedures

Participants consisted of 760 female first-year students from a large public Northeastern University. Participant flow is presented in Figure 2. Email addresses for all entering female students who were United States residents and aged 18 or 19 (N = 1,203) were obtained from the Registrar. In early September 2018, emails were sent inviting students to participate in two online surveys during the semester. Women who did not respond to the initial email were sent up to 4 additional emails over the next 3 weeks. Most (75%) completed their survey within one week and response rate was 63.2%. Ethnic/racial distribution of the sample was similar to that of the first-year class as a whole: 56.8% non-Hispanic white, 21.0% Asian, 9.5% African-American, 8.6% Hispanic compared with 54.0% non-Hispanic white, 18.1% Asian, 8.6% African-American, and 9.1% Hispanic for the first-year class. Most lived in University housing (76.8%), similar to the proportion of the class as a whole (75%). Most identified as heterosexual (83.2%), with the remainder identifying as lesbian (2.4%) or bisexual/questioning (14.4%).

Figure 2.

Participant Flow Chart

In mid-November, women who completed the T1 survey were invited by email to complete the follow-up Time 2 (T2) survey. Women who did not respond to the initial email were sent up to 4 emails reminding them to complete the survey by the last day of Fall classes. Of 760 women who completed the T1 survey, 654 (86.1%) completed the follow-up (T2) survey.

Email invitations contained a unique link to a secure website which could be entered with University ID. After providing electronic informed consent, participants completed a 15-20 minute online survey regarding their sexual behavior, alcohol use, perceived norms, and sexual victimization. Participants were compensated $25 in Campus Cash for completion of each survey. All procedures were approved by the University Institutional Review Board.

Randomization and Intervention

Participants were randomly assigned via computerized algorithm embedded within the T1 survey to receive personalized normative feedback (PNF) regarding how their hookups compared with those of other University first-year women (n = 377) or no information (n = 383). Participants were randomly assigned to condition regardless of sexual identity and the intervention used gender-neutral language. PNF was presented immediately following the T1 measures described below. Because social norms based on closer reference groups are more influential (Lewis et al. 2007b), the feedback provided included female and first-year-specific norms that were collected from an equivalent sample of women who entered the University the previous year. The PNF presented to women in the intervention condition was modeled after that used in previously efficacious interventions (Lewis et al. 2014b) and based on the Brief Alcohol Screening and Intervention for College Students (BASICS, Dimeff et al. 1999). First, women were presented with a screen stating: “We have conducted surveys regarding the social and sexual behaviors of hundreds of [University] students over the years and find that students often want to know how they compare with others. On the next screen, we’ll provide you with personalized information on how some of your behaviors compare with those of other [University] freshman women”.

The PNF that followed presented information in bar graph form showing one’s own number of hookups and perceived hookups of other university first-year women (both reported in the T1 survey). Upon clicking a button (“see how you compare”), the participant was shown a third bar representing the actual number of hookups of other first-year women. Information was presented separately for hookups with alcohol and then for hookups without alcohol (determined by subtracting hookups with alcohol from total hookups). Participants were allowed to advance to the next screen even without clicking the “see how you compare” button and viewing the graphical presentation of the normative information. However, on the next screen, all intervention participants were shown in text form the actual mean number of hookups reported by other first-year women (M = 1.09 for hookups with alcohol, M = 0.80 for hookups without alcohol). This information was identical to the normative information presented in graphical form on the previous screen. The text screens also stated that “70% of freshman women reported never hooking up while drinking” and “68% of freshman women reported never hooking up when not drinking”. This wording was used to normalize low risk behavior. Thus, all women in the intervention condition, regardless of whether or not they clicked on the comparison button, were shown normative information on others’ hookups. The proportion who opted to click provided an estimate of interest in this normative information and a comparison of those who clicked versus did not click allowed us to examine whether this information impacted the efficacy of the intervention.

Measures

Hookups and Sexual Activity.

Hookups were assessed at T1 and T2. Participants were first provided with a definition (Paul et al. 2000): “A hookup is a sexual encounter between strangers, friends, or acquaintances – people not in a relationship with each other. Some physical interaction (e.g., kissing) is typical but it may or may not involve sexual intercourse”. Then they were asked “How many hookups have you had since you started college?” followed by “How many of these hookups occurred when you were drinking alcohol?” The definition was then repeated and participants were asked “How many hookups do you think the average [University] freshman woman has had since starting college?” followed by “How many of these hookups do you think occurred while or just after drinking alcohol?” Subtracting the number of hookups with alcohol from the total number of hookups provided the number of hookups without alcohol. Women were also asked how many hookups they expected to have during the semester (at T1 only), how many sexual partners since starting college, number of partners with whom they had sex on only one occasion, and number of partners with whom they had sex without interest in a committed relationship. Because items were open-ended, they yielded a few extreme values and hence were Winsorized at the 97^th percentile.

Heavy Episodic Drinking.

At T1 and T2, women were asked how many days out of 30 they typically drank alcohol. Those who drank at least once per month were asked how many drinks they typically consumed when they drank, how many days (out of 30) they drank 4 or more drinks at a setting, and how many days (out of 30) they drank to intoxication. The last two items were highly correlated (r = 0.87 at T1 and r = 0.88 at T2) and averaged to create a composite HED frequency measure (Testa et al. 2009). This measure was also Winsorized.

Sexual Victimization.

At T1, a 12-item brief version of the Revised Sexual Experiences Survey (Testa et al. 2019a) was used to assess sexual victimization experiences since age 14. Three perpetrator tactics (verbal coercion, physical harm or threats of harm, incapacitation) were followed by four potential outcomes (unwanted contact, attempted intercourse, completed vaginal intercourse, and oral sex or other sex acts). The measure was shortened from the 20-item version used by Carey et al. (2015) and Testa et al. (2010) by combining physical force with threats of force and combining oral sex with other sex acts. Women reported the number of times they had experienced each item (never, once, or two or more times; scored 0, 1, 2) since age 14. Women who reported any item were scored as experiencing T1 victimization. We also computed a severity score based on the most severe act of aggression experienced (see Testa et al. 2010) scored as follows: no victimization (0), contact (1), attempted coercion (2), coercion (3), attempted rape (4), completed rape (5).

The same sexual victimization measure, with items phrased as “since beginning college”, was used to assess T2 experiences, our primary outcome. For any item that was reported, women were asked the date of the most recent experience. Because the T2 victimization measure was intended to evaluate the efficacy of the T1 intervention, we excluded any experiences that occurred prior to the date on which T1 was completed (and were presumably already reported at T1). Thus, T2 victimization includes college experiences that occurred after the date of the T1 assessment and intervention.

Living With Parents.

Because living with parents may be protective against sexual victimization (Tyler et al. 2017) we asked women where they lived during the school year (1 = with parents or other relatives; 0 = dormitory or campus apartment).

Analytic Plan

We hypothesized that PNF intervention would predict lower perceived hookups at T2, which would in turn predict fewer T2 hookups, and lower incidence of T2 sexual victimization. We tested the indirect effects of PNF on T2 sexual victimization using a sequential mediator path analysis model in Mplus Version 8.2 (Muthén et al. 2016; Muthén and Muthén 2017). We used maximum likelihood estimation with robust standard errors. The statistical significance of the indirect effect was tested using 50,000 bootstrap draws to estimate precisely the 95% confidence intervals (CI) determined from the lower and upper 2.5 percentiles (Hayes 2013; MacKinnon et al. 2004). T1 sexual victimization (since age 14) and T1 HED were included as covariates, allowing us to control for their influence on the outcome (T2 victimization) and on the two mediators (i.e., T2 perceptions of others’ hookups, T2 hookups) in the model (see Figure 1).

Results

Preliminary Analyses

Of 760 women who completed T1 in September, 654 (86.1%) completed T2 in November and are included in analyses. Completers and non-completers were compared on T1 demographics, intervention status, prior victimization, perceived hookups, actual hookups, and HED. Compared with completers, non-completers were more likely to live with parents (29.1% vs. 19.9%), χ²(1) = 4.515, p = .034, and reported more HED days, M = 2.24 (SD = 2.94) vs. M = 1.68 (SD = 2.67), t(757) = 1.976, p = .049. We controlled for the effect of living with parents on victimization and for the effects of HED on all model variables (see below). Intervention (n = 377) and control group women (n = 383) did not differ on T1 sexual victimization, hookups, or perceived hookups, suggesting that randomization was successful.

As described above, women could choose whether to display normative information regarding actual hookups reported by their peers in graphical form (all saw the information in text form). This allowed us to assess whether opting to view this graphical information increased the efficacy of the minimal intervention. Roughly half in the intervention condition opted to view normative information on hookups with alcohol (n = 174/377) and hookups without alcohol (n = 166/377). Most clicked on both screens (n = 153) or neither screen (n = 190), but some (34) clicked on one or the other. Within the intervention condition, we compared those who clicked on at least one screen (n = 187) with those who did not (n = 190) on T2 perceived hookups of others, T2 actual hookups, and T2 sexual victimization but found no differences. Because the higher dosage of information (or interest in this information) failed to have a measurable effect, we did not consider this variable further and analyzed all intervention condition participants together.

Descriptive Information

At T1, women reported an average of 0.37 hookups (SD = 0.80) and expected, on average, to experience 1.87 hookups (SD = 2.65) during the semester. However, on average, women believed that others had 5.18 (SD = 3.49) hookups, thus most (742/759, 97.8%) believed that others hooked up more than they themselves did. The actual norm for hookups with alcohol (M = 1.09) was overestimated by 79.9% of women and the norm for hookups without alcohol (M = 0.80) by 72.5%. Thus, for most women, the normative feedback countered the perception that other women were hooking up more frequently than they actually were. As shown in Table 1, at T2, women who received the PNF intervention believed that others experienced significantly fewer hookups (M = 3.65, SD = 2.43) than did women in the control group (M = 4.29, SD = 2.85) but did not differ on T2 hookups or T2 sexual victimization.

Table 1.

Comparison of Intervention Group and Control Group on Key Variables

Variable	Group						t(df)/ χ2(df)
	Intervention group (n = 377; 49.6%)			Control group (n = 383; 50.4%)
	M/%	SD	n	M/%	SD	n
T1 HED frequency	1.80	2.80	377	1.71	2.63	382	−0.424 (757)
T1 perceived hookups of other women	5.03	3.43	377	5.32	3.56	383	1.142 (758)
T1 hookups since started college	0.38	0.80	377	0.36	0.80	382	−0.448 (757)
T1 any sexual victimization since age 14	115/376 (30.6%)			113/381 (29.7%)			0.77 (1)
T2 HED frequency	1.60	330	1.59	2.34	333	−0.047 (661)
T2 perceived hookups of other women	3.65	329	4.29	2.85	329	3.104 (656)**
T2 hookups since started college	1.09	2.17	329	1.13	2.30	329	0.192 (656)
T2 sexual victimization since intervention	36/325 (11.1%)			40/328 (12.2%)			0.198 (1)

^***p < .001

^**p < .01

^*p < .05

Indirect Effects Model

Primary analyses involved testing the proposed sequential mediator model of the effects of PNF on occurrence of T2 sexual victimization as mediated via perceived hookup norms and own hookups. We tested the model controlling for effects of T1 victimization and T1 HED on the outcome (T2 victimization) and on the two mediators (T2 perceived hookups, T2 hookups). All paths and effect sizes are presented in Table 2. As depicted in Figure 1, intervention (vs. control) predicted perceiving significantly fewer hookups by other first-year women at T2. These perceptions were positively associated with one’s own hookups, which in turn, positively predicted the occurrence of T2 sexual victimization. The indirect effect of intervention on T2 sexual victimization via perceptions of others’ hookups and one’s own hookups was significant, b = −0.022 (SE = 0.010), p = .020, effect size = 0.08. Standardized mediation effect-size measures were used to compute the indirect effect (Miočević et al. 2018). The statistical significance of the indirect effect was confirmed using 50,000 bootstrap draws (95% CI = −0.023, −0.003). The proposed model explained 19.2% of the variance in T2 sexual victimization.

Table 2.

Results for Sequential Mediator Model of the Effects of PNF on Sexual Victimization via Perceived Hookups and One’s Own Hookups

Model	Estimate (S.E.)	95% CI	Standardized effect size	R2
T2 perceived hookups
Intercept	4.122 (0.168)***	[3.793, 4.451]	−0.25	0.051
← PNF	−0.667 (0.204)**	[−1.067, −0.268]
← T1 sexual victimization	0.657 (0.247)**	[0.173, 1.141]	0.25
← T1 HED	0.135 (0.044)**	[0.050, 0.221]	0.13
T2 one’s own hookups
Intercept	0.183 (0.184)	[−0.178, 0.544]	0.02
← PNF	0.036 (0.154)	[−0.265, 0.338]
← T2 perceived hookups	0.171 (0.038)***	[0.096, 0.246]	0.20	0.254
← T1 sexual victimization	0.774 (0.206)***	[0.370, 1.178]	0.35
← T1 HED	0.305 (0.043)***	[0.221, 0.389]	0.37
T2 sexual victimization			Odds ratio
Intercept	1.309 (0.144)***	[2.062, 3.220]
← PNF	−0.149 (0.266)	[−0.698, 0.328]	0.862
← T2 perceived hookups	−0.029 (0.054)	[−0.138, 0.071]	0.971	0.192
← T2 one’s own hookups	0.196 (0.051)***	[0.107, 0.309]	1.217
← T1 sexual victimization	1.173 (0.281)***	[0.848, 1.910]	3.231
← T1 HED	0.083 (0.045)	[0.006, 0.178]	1.087
← T1 living with parents	−0.286 (0.383)	[−1.150, 0.363]	0.751
Effects from PNF to T2 sexual victimization
Total	−0.145 (0.270)	[−0.674, 0.385]
Total indirect	0.004 (0.046)	[−0.087, 0.095]
Specific indirect
PNF → T2 perceived hookups → T2 own hookups → T2 sexual victimization	−0.022 (0.010)*	[−0.041, −0.004]	0.08

^***p < .001

^**p < .01

^*p < .05

We then tested whether intervention status interacted with either T1 sexual victimization status or T1 HED. Neither the intervention X victimization (p = .352) nor the intervention X HED interaction (p = .418) emerged as significant and inclusion of the interaction terms did not alter any other pathways depicted in Figure 1. Thus, intervention effects were equivalent for women with and without prior victimization and regardless of initial level of HED.

We repeated the model substituting T2 sexual victimization severity for any victimization and using T1 victimization severity as the covariate. The indirect effect from intervention to sexual victimization severity via perceived hookup norms and actual hookups was also significant, b = −0.034 (SE = 0.014), p = .013, 95% CI = −0.063, −0.006 and paths were virtually identical to the any victimization model.

Discussion

We tested and found evidence for a theoretically-supported indirect effects model in which PNF regarding hookups reduced perceived descriptive hookup norms which in turn were associated with fewer actual hookups, and ultimately, with less first semester sexual victimization. There is substantial evidence that brief PNF interventions can reduce alcohol use (Tanner-Smith and Lipsey 2015) and some evidence that such interventions may prevent alcohol-related risky sexual behavior (Lewis et al. 2014b; 2018) via normative perceptions. However, to our knowledge, this is the first study that has tested and found support for a similar model specific to hookup norms and behavior. Of note, we tested the interaction between prior victimization and intervention and between initial HED and intervention and found no interactions. Thus, the intervention was equally efficacious for women with and without prior victimization and regardless of initial level of drinking.

The PNF intervention was intentionally minimal and half of the women assigned to the intervention condition saw even less normative information than the others (because they opted not to click to display the information on the graph). Nonetheless, the normative information had a small but measurable effect on later social norms, suggesting that newly-matriculated college women lack accurate information about others’ sexual behavior and can benefit from information that helps to correct the misperception that hookups are highly prevalent. The PNF approach is promising and easily disseminable for universities to adopt, although the modest effect – and lack of direct effects on actual hookups – suggests that increasing the strength and dosage of the feedback intervention may be needed to increase efficacy.

The significant indirect effect of PNF intervention on sexual victimization via hookups adds to a growing body of research pointing toward casual sexual behavior as a key mechanism in college women’s sexual vulnerability (Fielder et al. 2014; Sutton et al. 2019; Testa et al. 2019a; Tyler et al. 2017). Even after accounting for the significant effects of prior victimization and HED, hookup frequency was positively associated with sexual victimization. Although we did not examine the hookup-victimization link at the event level, hookups are a common context for sexual aggression (Flack et al. 2007) and women are more likely than men to report that their worst hookup involved forced sexual behavior (Paul and Hayes 2002). All hookups involve exposure to a potential perpetrator, typically in a private setting in which a woman’s initial sexual interest and behavior may be misperceived as a desire for more intimate behavior. In contrast, HED increases vulnerability only when it occurs in conjunction with exposure to a potential perpetrator. Our findings, though cross-sectional, are promising in suggesting that reducing hookups may be an efficacious means of preventing sexual victimization among college populations.

Limitations

The sequential model we tested was not fully longitudinal, limiting the ability to assert causality. Although this design decision in part reflected recognition of the first semester of college as a particularly high-risk period, it will be important to test whether effects are obtained longitudinally and sustained over a longer follow-up period. Moreover, effects, although significant, were small. We deliberately opted for a minimal, universal intervention that could be easily disseminated at low cost. It would be valuable to test whether increases in dosage or intervention strength boost efficacy. Nonetheless, primary prevention efforts during this critical ‘red zone,’ even with modest effects, can ultimately result in potentially high public health and clinical impact (Spoth et al. 2009). Finally, although findings may generalize to other college samples, it is important to consider whether the approach is efficacious for populations at even greater risk of sexual assault such as sexual minority women (Coulter et al. 2017) and non-college populations (Sinozich and Langton 2014).