Beliefs About Consent and Sexual Assault Perpetration in a Longitudinal Study of College Men

Abstract

This study examined the mediating role of beliefs about both active and passive consent in the prospective associations between sexual assault (SA) risk factors and coercive, incapacitated, and forcible attempted/completed SA among college men. Participants were 471 college men who completed self-report surveys at the end of each of their 4 years of college. SA risk factors (risky behavior, rape-supportive beliefs and peer norms, personality traits, childhood adversity) were assessed at Wave 1, beliefs about consent were assessed at Wave 2, and perpetration was assessed at Waves 3 and 4. Multivariate regression models with bias-corrected bootstrapping assessed longitudinal mediation. SA risk factors were negatively associated with endorsement of active consent (verbal approval required) and positively associated with passive consent (assume “yes” until you hear a “no”), with strongest effects observed for coercive SA. Both types of beliefs about consent served as mediators between risk factors and perpetration. Findings suggest that prevention programs should include a focus on reducing SA risk factors, clarifying definitions of consent, and improving sexual communication.

Sexual assault (SA) includes nonconsensual sexual activity ranging from coercive sex to forcible and/or incapacitated rape. SA is a prevalent problem on college campuses, with approximately 20%–25% of women experiencing victimization during college (Fisher, Cullen, & Turner, 2000; Kilpatrick, Resnick, Ruggiero, Conoscenti, & McCauley, 2007; Krebs, Lindquist, Warner, Fisher, & Martin, 2007) and 10%–30% of men perpetrating SA (Abbey & McAuslan, 2004; Hines & Saudino, 2003; Thompson, Swartout, & Koss, 2013; White & Smith, 2004; Zinzow & Thompson, 2015b). SA is associated with myriad mental and physical health outcomes, including posttraumatic stress disorder, depression, substance use, and poor self-rated health (e.g., Golding, 1999; Resick, 1993; Tjaden & Thoennes, 2006). In order to mitigate the public health impact of SA, it is important to understand the factors associated with SA perpetration. The purpose of the current study was to examine the role of beliefs about consent as mediators between established SA risk factors and perpetration in college men.

Researchers have developed several theoretical models to explain SA perpetration. Malamuth and colleagues applied an ecological framework to understanding SA, which posits an interaction between embedded systems of influence on human development (Malamuth, Sockloskie, Koss, & Tanaka, 1991). These include: (a) the microsystem (e.g., individual factors and the home environment, including antisocial personality characteristics and parental violence); (b) the exosystem (e.g., peer group influences, including delinquent behavior, rape-supportive peer norms, and hostility toward women), and (c) the macrosystem (e.g., broader cultural values and belief systems, including traditional gender role norms). Within this context, Malamuth and colleagues focus on a confluence model of sexual aggression that describes how early adverse experiences lead to SA via two inter-related trajectories: (a) hostile attitudes and personality characteristics and (b) impersonal sex and promiscuity. Similarly, social cognitive models of aggression suggest that exposure to a social environment where hostility is modeled (e.g., rape-supportive beliefs and coercive behavior modeled by peers, family, media, or pornography) can increase violence-supportive schemas, self-efficacy for aggression, positive outcome expectations for aggression, and desirability for outcomes obtained through aggression (Bandura, 1986; Gannon, 2009). Within the confluence model and related social cognitive models of SA perpetration, SA-supportive attitudes and beliefs mediate the relation between early adverse experiences, personality characteristics, peer norms, and sexual aggression (Abbey, Jacques-Tiura, & LeBreton, 2011; Hanson & Harris, 2000; Malamuth et al., 1991). Abbey and colleagues later expanded the confluence model to include alcohol use, with childhood victimization, delinquency, and antisocial personality traits leading to heavy alcohol consumption (Abbey et al., 2011). Within this model, heavy alcohol consumption predicts impersonal sex and misperception of sexual intent, which are positively associated with sexual aggression.

In support of these models, research with college men has established childhood victimization, antisocial personality traits (conning/superficial charm, low empathy, impulsivity), multiple impersonal sexual partners, and alcohol use as risk factors for sexual aggression (Abbey, 2002; Abbey et al., 2011; Giancola, 2002; Koss & Dinero, 1988; Malamuth, 2003; Malamuth, Linz, Heavey, Barnes, & Acker, 1995; Ouimette & Riggs, 1998). Regarding peer influences on these trajectories, research has found that perceived peer norms such as peer approval of forced sex, peer pressure for sex, and perceived sanctions against SA play a role in SA perpetration (Abbey & McAuslan, 2004; Thompson, Koss, Kingree, Goree, & Rice, 2010).

Finally, social cognitive factors that serve as risk factors for perpetration include rape-supportive beliefs, hostility toward women, sexual compulsivity, and pornography exposure (Abbey & McAuslan, 2004; Abbey, McAuslan, & Ross, 1998; Beech, Friendship, Erikson, & Hanson, 2002; Davis, Norris, George, Martell, & Heiman, 2006; Malamuth, Addison, & Koss, 2000; Malamuth et al., 1995; Mann & Hollin, 2007; Thornton, 2002). Although these studies suggest that cognitive factors such as rape-supportive beliefs and misperception of sexual intent play a mediating role between more distal risk factors and perpetration, little research has focused on comprehension of or beliefs about consent as a risk factor for SA perpetration.

Definitions of sexual consent vary widely, and generally refer to a person meeting the local age for consent, having the mental and physical capacity to consent, and having freely communicated willingness to participate in sexual activity (Rape, Abuse, and Incest National Network [RAINN], n.d.; Hickman & Muehlenhard, 1999). Consent that is freely given is not induced by coercion and includes overt actions or words indicating agreement to sexual activity (RAINN, n.d.). Unfortunately, lack of a consistent legal definition or clarification regarding which actions indicate agreement to sexual activity can leave room for ambiguity in sexual decision-making. Misperceptions of consent have been included in scales assessing rape-supportive beliefs, and therefore are likely to comprise the context in which SA occurs. For example, the Rape Supportive Beliefs Scale includes items such as “When a woman allows touching to get to a certain point, she is implicitly agreeing to have sex” and “If a woman is raped, often it’s because she didn’t say ‘no’ clearly enough” (Lonsway & Fitzgerald, 1995).

In addition to lack of clarity regarding the definition of consent, misperceptions of sexual intentions can contribute to SA. Studies of college men have found that misperception of a woman’s sexual interest has been associated with greater likelihood of SA perpetration (Abbey et al., 1998). Misperceptions of intent may be more likely to occur when the victim is intoxicated, has had prior sexual activity with the perpetrator, or has engaged in some level of sexual activity and then refused (Abbey, Zawacki, Buck, Clinton, & McAuslan, 2004). Men who endorse other forms of rape-supportive beliefs are also more likely to misinterpret a woman’s sexual interest (Abbey, Parkhill, Jacques-Tiura, & Saenz, 2009).

Only one known study has examined comprehension of sexual consent as a risk factor for SA perpetration (Warren, Swan, & Allen, 2015). Consent was assessed by presenting two vignettes involving coercive sexual behavior and asking participants to what extent they believed the behavior to be acceptable. Among 217 male college students, rape myth acceptance, peer support of SA, and conformity to masculine gender role norms were associated with comprehension of consent. Comprehension of consent mediated the relationship between these social cognitive variables and SA. A limitation of this and prior studies is that cross-sectional designs are unable to test causal connections between risk factors and SA perpetration. Therefore, longitudinal designs are needed to examine the role of beliefs about consent as a mediator between SA risk factors and perpetration. In addition, this study focused on a limited set of social and cognitive risk factors for SA. Further research is needed to investigate the influence of other established SA risk factors on beliefs about consent. These include personality characteristics and a broader range of peer norm variables. Studies have shown that SA risk factors differ between perpetrators employing different types of SA tactics (i.e., verbal coercion, incapacitation, physical force; Abbey & Jacques-Tiura, 2011; Zinzow & Thompson, 2015a), suggesting that research should also investigate how beliefs about consent may differentially relate to various forms of perpetration.

The purpose of the current study was to employ a longitudinal design to examine the prospective associations between SA risk factors, beliefs about consent and SA perpetration. We assessed the following empirically supported social cognitive risk factors for SA: pornography exposure, sexual compulsivity, peer approval of forced sex, peer pressure for sex, perceived sanctions for SA, rape-supportive beliefs, and hostility toward women. We also assessed personality traits that are associated with SA: lack of empathy, conning/superficial charm, and impulsivity. We assessed two forms of beliefs about consent: (a) active consent (always obtain verbal approval for sexual activity) and (b) passive consent (OK to continue until a partner indicates otherwise). We examined three SA tactic groups: (a) verbal coercion, (b) incapacitation, and (c) physical force. We expected that risk factors would be negatively associated with active consent and positively associated with passive consent. Given prior research suggesting a mediating role of beliefs about consent between risk factors and SA, we expected that both types of consent would perform as a mediator between this broader range of risk factors and SA perpetration.

METHOD

Participants and Procedures

The current study was part of a larger study on longitudinal trajectories of SA perpetration among college men (citations redacted for blind review). We recruited first-year male students by sending personal e-mails to all males enrolled as full-time, first-time students at a large southeastern university in March 2008 (N = 1,472). Students also were recruited with notices in the student newspaper and flyers distributed around campus. Students were invited to come to the student health center between 9:00 a.m. and 4:00 p.m. during the upcoming week to complete a confidential, 20- to 30-minute self-report survey on men’s attitudes and behaviors regarding relationships with women. They were paid $20.00 at Waves 1 and 2 and $25.00 at Waves 3 and 4 for their participation. Within 1 week, 800 students completed surveys after providing written informed consent.

Prior to data collection, local Institutional Review Board approval was obtained from the university and a Certificate of Confidentiality was obtained from the National Institutes of Health. After completing the surveys, participants deposited them (without consent forms attached) into a locked box, received payment for their participation, and were provided a referral sheet of counseling resources. Men who completed surveys at Wave 1 were contacted via e-mail to participate in follow-up surveys. At Waves 2, 3, and 4, participants were provided survey packets with confidential, unique codes that linked their surveys. Survey procedures were similar across waves. No personal identifiers were included on the surveys.

The initial sample consisted of 800 men. Five individuals were excluded because they were less than 18 years of age at the time of Wave 1 data collection. Eighty-two percent of the sample completed the Wave 2 survey, 75% completed the Wave 3 survey, and 72% completed the Wave 4 survey. Only males who completed all four waves of data collection (n = 471) comprised the analytic sample. Attrition analyses for all four time points indicated that participants with higher impulsivity were less likely to complete the Wave 2 and Wave 3 surveys, F(1, 794) = 6.26, p < .05; F(1, 794) = 10.14, p < .01. Higher agreement with passive consent was associated with attrition at Wave 3 F(1, 794) = 5.06, p < .05. No study variables were associated with attrition at Wave 4.

Participants’ average age was 18.56 years at Wave 1 (SD = .51) and most (89%) were White. The sample was representative of the population of first-year male students in terms of age and race based on data provided from the Office of Institutional Research.

Measures

To assess the prospective mediating effects of consent in the relation between risk factors and SA, variables were selected from different waves of data. Risk factors were assessed at Wave 1, beliefs about consent were assessed at Wave 2, and SA was assessed at Wave 3 and 4.

Sexual Assault.

SA was assessed using two different time boundaries for recall at Waves 3 and 4: during the junior year and during the senior year. The revised Sexual Experiences Survey (SES; Koss et al., 2007), the most widely used and validated measure of perpetration among college students, was used to assess SA. Prior studies have established good internal consistency reliability and validity for the SES; responses to the measure were highly correlated with face-to-face interviews (Koss & Gidycz, 1985; Koss, Gidycz, & Wisniewski, 1987; Koss & Oros, 1982). Twelve items assessed whether the participant used verbal coercion (e.g., lies, threatening to end the relationship, continued verbal pressure when she didn’t want to, showing displeasure, criticizing sexuality) to attempt or complete oral, vaginal, or anal penetration. Six items assessed whether the participant used incapacitation as a tactic to attempt or complete oral, vaginal, or anal penetration (i.e., “taking advantage when she was too drunk or out of it to stop what was happening.”). Twelve items assessed whether the participant used forcible tactics to engage in oral, vaginal, or anal penetration (e.g., threatening physical harm, holding her down, having a weapon). Consistent with prior studies, participants were grouped by the most coercive tactic that they used at Wave 3 or 4 (verbal coercion, incapacitation, or forcible; Abbey & Jacques-Tiura, 2011; Zinzow & Thompson, 2015b). Approximately one-fifth (22%, n = 106) of the analytic sample reported having perpetrated attempted/completed coerced sex or rape (oral, vaginal, or anal penetration) at Waves 3 or 4, with 11% (n = 53) reporting use of verbal coercion, 8% (n = 40) reporting incapacitation, and 3% (n = 13) reporting forcible tactics.

Consent.

Two items assessing beliefs about consent were derived from a prior measure of preferred means for obtaining consent (Humphreys & Herold, 2003). One item assessed beliefs about passive consent “In making sexual advances, it is okay to continue until a partner indicates otherwise (i.e., assume ‘yes’ until you hear a ‘no’).” A second item assessed active consent “Before making sexual advances, one should always ask for and obtain a verbal ‘yes’ to engage in any sexual activities (i.e., assume ‘no’ until you get a ‘yes’).” The original two-item measure asked participants to indicate which statement they agreed with more. A later study modified the items to allow participants to rate each item separately; this allowed a clearer differentiation between the constructs of passive and active consent. In this study, a principal components analysis found these items to load highly on a 23-item measure of sexual consent attitudes (Humphreys & Herold, 2007). Items were answered on a 1–4 scale ranging from “strongly disagree” to “strongly agree.” The mean for passive consent was 2.46 (SD = .73) and active consent was 2.74 (SD = .71). The two items were negatively correlated with each other (r = −.38, p < .001).

Risky Behavior.

Items assessed number of sexual partners, high-risk drinking, and substance use. Two items assessed how many sexual partners (oral sex or vaginal/anal sex) they had since age 14 (M = 2.21, SD = 4.07). The College Alcohol Survey assessed high-risk drinking with five items measuring quantity and frequency of alcohol use in the past 30 days. The National Institute on Alcohol Abuse and Alcoholism (NIAAA) recommends these items to assess drinking quantity and patterns (Dawson & Room, 2000), M = 2.20, SD = 2.07; α = .93. Two items assessed lifetime marijuana or other illegal drug use (“yes” or “no”; 41% of participants indicated prior use).

Rape-Supportive Beliefs and Norms.

The mean of an eight-item scale adapted from the Hostility Toward Women Scale (Check, 1984; Koss & Gaines, 1993; Thompson et al., 2013) assessed for hostility toward women. Internal consistency reliability of the full scale is .80 (Check, 1984). The adapted brief scale was developed based on items that had the largest correlations with sexual aggression in prior research (Koss & Gaines, 1993). Items were answered using a 1–5 response format, with higher scores reflecting higher levels of hostility (e.g., “Many times a woman appears to care, but really just wants to use me;” α = .90, M = 2.60, SD = .82). The 19-item Rape Myth Scale (Lonsway & Fitzgerald, 1995) was used to assess for rape-supportive beliefs. Items were answered using a 1–5 response format, with higher mean scores indicating higher levels of rape-supportive beliefs (e.g., “When women talk and act sexy, they are inviting rape;” α = .90; M = 2.24, SD = .62). Prior scale validation determined strong relationships between the Rape Myth Scale and hostility toward women, as well as adversarial sexual beliefs (Lonsway & Fitzgerald, 1995).

Three scales assessed peer approval of forced sex, perceived sanctions for SA, and peer pressure for sex. A six-item scale developed by Abbey and McAuslan (2004) assessed perceptions of one’s current set of friends’ approval of forced sex. Internal consistency reliability in a study of college students was .80 (Abbey & McAuslan, 2004). Participants responded on a 1–4 scale, with higher scores indicating greater perceptions that participants’ peers would approve of various strategies to obtain sex with a woman (e.g., “Do your friends approve of getting a woman drunk or high to have sex?” α = .78, M = 1.28, SD = .40). Three items assessed perceived sanctions for SA (Foshee, Linder, MacDougall, & Bangdiwala, 2001). Items were answered on a 1–4 scale, with lower scores indicating greater perceptions of negative sanctions (e.g., “Bad things happen to people who are sexually aggressive to girls”; α = .68, M = 3.18, SD = .65). Three items were used to assess perceived peer pressure from friends to have sex with women (Kanin, 1985). Items were answered on a 1–4 scale, with higher scores reflective of perceived pressure from friends to have sex with women (e.g., “Do your friends lack respect for guys who have never had sex?” α = .76, M = 1.67, SD = .69).

Antisocial Personality Traits.

Five scales assessed personality traits associated with SA. The mean of the 19-item Impulsivity Questionnaire (Eysenck, Pearson, Easting, & Allsopp, 1985) was used to assess for impulsive behaviors (e.g., “I do and say things without stopping to think”). The scale has demonstrated internal consistency reliability, as well as validity in relation to other self-report measures of venturesomeness and extraversion (Eysenck et al., 1985). Items were answered using a yes (1)/no (0) response format, and higher scores indicating higher levels of impulsivity (α = .79; M = .34, SD = .21). The charming and conning subscale (six items) from the Multidimensional Assessment of Sex and Aggression (MASA) was used to assess conning and superficial charm (Knight, Prentky, & Cerce, 1994). Items were answered on a scale of 1–5, with higher mean scores reflecting more superficial charm (e.g., “I can easily charm someone to do almost anything for me,” α = .75, M = 3.13; SD = .92). The Pervasive Anger subscale (eight items) of the MASA was used to assess pervasive anger (e.g., “I have to struggle to control my sexual thoughts and behavior,” α = .86, M = 2.39; SD = .68). MASA scale development on a sample of sex offenders demonstrated internal consistency (94% of scales with alpha greater than .80), test–retest reliability (86% of scales with .70 or higher test–rest reliability), and concurrent validity with archival files of sexual behavior has been established among a sample of sex offenders (Knight et al., 1994). Empathy was measured with the mean score on the Perspective Taking Scale of the Interpersonal Reactivity Index (Davis, 1980). The scale consists of four items scored on a 1–5 response format with higher scores reflecting lower empathy (e.g., “When I’m upset at someone, I usually try to ‘put myself in his shoes’ for a while.” Scale validation on a sample of college students demonstrated high test–retest (.61–.62) and internal consistency reliability of the Perspective Taking subscale (.75–.78; Davis, 1980). The mean of the 10-item Sexual Compulsivity Scale (Kalichman & Rompa, 2001) was used to assess for sexual preoccupations and intrusive thoughts (e.g., “I feel that sexual thoughts and feelings are stronger than I am”). Prior studies have reported high internal consistency reliability, as well as convergent and discriminant validity of the scale in community and HIV-positive samples (Kalichman & Rompa, 1995, 2001). Items were answered on a 1–4 scale, with higher scores indicating greater sexual compulsivity (α = .83; M = 1.43, SD = .40).

Childhood Adversity.

The Children’s Perception of Interparental Conflict Scale (seven items) assessed witnessed interparental conflict with a 1–4 response format, with higher mean scores reflecting higher levels of conflict (e.g., “My parents got really mad when they argued,” α = .86, M = 1.71; SD = .56). Five items from the Adverse Child Experiences study (Dube, Williamson, Thompson, Felitti, & Anda, 2004) were used to assess adverse childhood experiences (e.g., lived with someone who was a problem drinker or alcoholic). Participants answered “yes” (coded “1”) or “no” (coded “0”) to each experience (α = .58, M = .43; SD = .85). Two behaviorally specific items assessed child sexual abuse (sexual activity with someone 5 years older before age 18, or physically forced sexual activity with someone less than 5 years older before age 18; Briere & Runtz, 1990). Eleven percent of participants endorsed at least one of these items.

Data Analytic Strategy

We first conducted ANOVAs to examine differences between SA tactic groups (coercive, incapacitated, forcible) on both active and passive consent. Tukey’s LSD post-hoc analyses were used to conduct pairwise comparisons between tactic groups. Next, multivariate regression models using PROCESS (Hayes, 2013) with bias-corrected bootstrapping based on 5,000 bootstrap samples assessed longitudinal mediation. To simplify the number of predictor variables, we applied principal components analysis, with Wave 1 variables from the entire sample. This resulted in four sets of risk factors: (a) risky behavior; (b) rape-supportive beliefs and norms; (c) antisocial personality traits; and (d) childhood adversity (see [citations redacted for blind review] for procedure and factor loadings). The mediating roles of both passive consent and active consent were examined, with separate analyses conducted for each of the four sets of risk factors. Mediating models were examined for each SA tactic group (coercive, incapacitated, and forcible), with nonperpetrators serving as the reference group. Factor scores for Wave 1 risk factors were included as predictors, beliefs about consent at Wave 2 were included as mediators, and perpetration of SA at Wave 3 or 4 was included as the outcome. Race and age were entered into the models as covariates.

RESULTS

ANOVA Comparisons

Results of ANOVAs demonstrated overall significant differences when tactic groups were compared on both active and passive consent (Table 1). The coercive SA perpetration group demonstrated the lowest mean on active consent (i.e., low agreement that one should assume “no” until hearing a “yes”) and highest mean on passive consent (i.e., high agreement that one should assume “yes” until hearing a “no”). Pairwise comparisons indicated that the coercive SA group significantly differed from nonperpetrators on active consent. Both the coercive and incapacitated SA groups endorsed significantly higher levels of passive consent than the nonperpetrator and forcible groups.

TABLE 1.

Results of ANOVA Comparisons of SA Tactic Groups on Active and Passive Consent

Tactic Group	Active Consent			Passive Consent
	M	SD	F	M	SD	F
Nonperpetrator	2.80a	0.69	4.48**	2.34a	0.72	10.31***
Coercive SA	2.44b	0.71		2.88b	0.70
Incapacitated SA	2.61a,b	0.64		2.72b	0.66
Forcible SA	2.73a,b	0.79		2.18a	1.08

Note. SA = sexual assault.

Means with differing superscripts were significantly different at p < .01.

^**p < .01.

^***p < .001.

Mediation Analyses

As shown in Table 2, beliefs in the importance of active consent were prospectively and negatively associated with all four sets of SA risk factors for the coercive SA models; active consent was also associated with all risk factors except childhood adversity in the incapacitated and forcible SA models. As shown in Table 3, passive consent was positively associated with risky behavior and rape-supportive beliefs in all models, and with antisocial traits in the coercive and incapacitated SA models.

TABLE 2.

Model Coefficients for the Mediating Role of Active Consent

	Variable	Coercive SA				R 2	Incapacitated SA				R 2	Forcible SA				R 2
		Active consent (M)		Coercive SA (Y)			Active consent (M)		Incapacitated SA (Y)			Active consent (M)		Forcible SA (Y)
		b	SE	b	SE		b	SE	b	SE		b	SE	b	SE
Model 1	Risky behavior (X)	−.21***	1.25	.63***	.16	.09	−.17***	.03	.78**	.18	.09	−.17***	.03	1.18***	.30	.09
	Active consent (M)	—	—	−.70**	.26		—	—	−.12	.31		—	—	.26	.53
	Indirect effect of X on Y through M	.15 (.03, .29) *					.02 (−.09, .14)					−.04 (−.32, .19)
Model 2	Rape-supportive beliefs (X)	−.14***	.03	.76***	.17	.04	−.09**	.04	.71***	.21	.05	−.08*	.04	1.00***	.30	.04
	Active consent (M)	—	—	−.72**	.25		—	—	−.25	.28		—	—	−.12	.47
	Indirect effect of X on Y through M	.10 (.02, .21) *					.02 (−.02, .10)					.01 (−.09, .13)
Model 3	Antisocial personality traits (X)	−.11**	.03	.28	.16	.03	−.09*	.03	.25	.18	.04	−.08*	.03	.81**	.28	.04
	Active consent (M)	—	—	−.88***	.25		—	—	−.34	.28		—	—	−.03	.47
	Indirect effect of X on Y through M	.09 (.03, .20) *					.03 (−.01, .11)					.002 (−.10, .11)
Model 4	Childhood adversity	−.08*	.04	.40**	.15	.02	−.05	.04	.47**	.17	.03	−.03	.04	.60**	.21	.03
	Active consent (M)			−.88***	.25				−.39	.28				−.31	.47
	Indirect effect of X on Y through M	.07 (.01, .19) *					.02 (−.01, .09)					.01 (−.02, .12)

Note. M = mediator; SA = sexual assault; X = predictor; Y = outcome.

Indirect effect statistics reflect coefficient estimate and bias-corrected bootstrap confidence intervals. Covariates included age and race. For coercive SA models, N = 404. Incapacitated SA models: N = 390. Forcible SA models: N = 370.

^*p < .05.

^**p < .01.

^***p < .001.

TABLE 3.

Model Coefficients for the Mediating Role of Passive Consent

	Variable	Coercive SA					Incapacitated SA					Forcible SA
		Passive consent (M)		Coercive SA (Y)			Passive consent (M)		Incapacitated SA (Y)			Passive consent (M)		Forcible SA (Y)
		b	SE	b	SE	R 2	b	SE	b	SE	R 2	b	SE	b	SE	R 2
Model 1	Risky behavior (X)	.23***	.04	.59***	.17	.10	.23***	.04	.67***	.18	.09	.20***	.04	1.47***	.35	.08
	Passive consent (M)			.88**	.27				.60	.31				−1.07*	.49
	Indirect effect of X on Y through M	.20 (.06, .39) *					.14 (−.002, .33)					−.22(−.58, .08)
Model 2	Rape-supportive beliefs (X)	.16***	.04	.72***	.17	.06	.13***	.04	.64**	.21	.03	.11**	.04	1.07***	.30	.02
	Passive consent (M)			.88**	.27				.75*	.29				−.46	.40
	Indirect effect of X on Y through M	.14 (.04, .29) *					.10 (.02, .25) *					−.05(−.22, .07)
Model 3	Antisocial personality traits (X)	.08*	.04	.30	.16	.02	.08*	.04	.22	.18	.01	.06	.04	.88**	.29	.01
	Passive consent (M)			1.08***	.26				.86**	.29				−.47	.42
	Indirect effect of X on Y through M	.09 (.01, .21) *					.07 (.01, .18) *					−.03(−.20,.03)
Model 4	Childhood adversity	.06	.04	.40**	.15	.02	.02	.04	.51**	.17	.002	.01	.04	.58**	.21	.03
	Passive consent (M)			1.08***	.27				.96**	.30				−.28	.43
	Indirect effect of X on Y through M	.07 (−.01, .20)					.02 (−.05, .13)					−.003 (−.11, .04)

Note. M = mediator; SA = sexual assault; X = predictor; Y = outcome.

Indirect effect statistics reflect coefficient estimate and bias-corrected bootstrap confidence intervals. Covariates included age and race. For coercive SA models, N = 404. Incapacitated SA models: N = 390. Forcible SA models: N = 370.

^*p < .05.

^**p < .01.

^***p < .001.

Active consent was prospectively and negatively associated with coercive SA in each of the four risk factor models (Table 2). Passive consent was positively associated with coercive SA in each of the four risk factor models. Passive consent was also positively associated with incapacitated SA in all risk factor models except risky behavior. Passive consent was only significantly associated with forcible SA in the risky behavior model and, by contrast, showed a negative relationship with forcible perpetration (Table 3).

Examining the indirect effects revealed support for the mediating role of active consent in the associations between all four risk factor scores and coercive SA. Active consent did not play a significant mediating role for incapacitated or forcible SA. Analyses demonstrated a significant mediating role for passive consent in the relation between the first three sets of risk factors (risky behavior, rape-supportive beliefs, antisocial traits) and coercive SA. We also found support for significant indirect effects for passive consent in the relation between rape-supportive beliefs, antisocial traits, and incapacitated SA. Results did not support a mediating role for passive consent in the relation between risk factors and forcible SA.

DISCUSSION

As hypothesized, SA risk factors were generally negatively associated with endorsement of active consent and positively associated with endorsement of passive consent. These sets of risk factors included risky behavior (alcohol/substance use and multiple sexual partners), rape-supportive beliefs/peer norms, antisocial personality traits (impulsivity, anger, conning, low empathy), and childhood adversity (parental conflict, abuse, neglect). Our findings are consistent with prior research implicating the importance of social cognitive variables such as rape myth acceptance and peer acceptance of SA in determining beliefs about consent (Warren et al., 2015). The current study extended this research by employing a longitudinal model and expanding the assessment of peer norms to include peer pressure for sex and perceived sanctions for coercive sex. Results suggest that a peer culture that is supportive of coercive sex, is hostile toward women, and minimizes sanctions for coercive sex, may represent the milieu in which problematic beliefs about consent develop.

Another contribution of the study was the inclusion of risky behavior, antisocial personality traits, and childhood adversity as predictors. Behavior such as risky drinking can lead to students engaging in sexual activities when one or both partners are unable to consent due to incapacitation. Our results suggest that such behaviors are intertwined with poor knowledge about consent, and that perceptions of consent can play a mediating role in the relation between risky behavior and various forms of perpetration. Similarly, antisocial personality traits were also associated with problematic perceptions of consent that in turn, increased risk for SA perpetration. These traits may develop out of an interaction between biological factors, such as genetic predispositions, and a sociocultural environment that condones violence (e.g., exposure to parental aggression or violence-supportive peer norms; Knight & Sims-Knight, 2003). For example, experiences of physical and verbal abuse have been associated with higher presence of psychopathy traits, including deceitful personality, emotional detachment, impulsivity, and aggressive tendencies (Knight & Sims-Knight, 2003). Our results are consistent with literature connecting childhood adversity to aggressive behavior, although the adversity factor only played a significant role in the passive consent mediation model for coercive SA. The more proximal predictors of rape-supportive beliefs and norms, risky behavior, and personality traits showed the strongest associations with beliefs about consent.

Our findings highlight the importance of both beliefs about passive consent and beliefs about active consent as mediators in the relation between SA risk factors and perpetration. Passive consent played a mediating role for both coercive and incapacitated SA models, whereas active consent was only a significant mediator for coercive SA models. Therefore, beliefs in passive consent, or that consent can be granted through body language without active verbal consent, appear to play an important role in nonforcible forms of SA perpetration. This is consistent with literature suggesting that ambiguous sexual communication and misperceptions of sexual intent are significant factors leading to nonconsensual sex (Abbey et al., 1998). The role of passive communication in aggressive behavior has also been discussed in the broader literature on indirect and passive aggression. Indirect aggression has been defined as intention to harm a person through a circumvent or covert manner, wherein the aggressor is less likely to be identified as an aggressor (Björkqvist, 1994). Passive aggression is a form of indirect aggression that has been defined as harming others through nonresponsiveness (Richardson, 2014). In the case of SA, lack of responsiveness to indirect cues that a partner wishes to stop sexual activity, or lack of attempts to solicit active consent, could be understood as forms of indirect or passive aggression.

A focus on this indirect form of aggression is also consistent with the importance of passive consent in relation to nonforcible SA. In fact, perpetrators of coercive and incapacitated SA reported higher means on passive consent than perpetrators of forcible SA, suggesting that beliefs about passive consent are particularly important catalysts for nonforcible forms of assault. Even though they represented a small subsample, it is interesting to note that perpetrators of forcible SA were less likely to endorse problematic beliefs about consent than the other tactic groups, and beliefs about consent were generally unrelated to forcible perpetration. Prior research suggests that perpetrators of forcible assaults are more likely to be repeat assaulters characterized by higher levels of antisocial traits, hostility toward women, and childhood adversity than perpetrators of incapacitated and coercive SA (Abbey, Parkhill, Clinton-Sherrod, & Zawacki, 2007; Zinzow & Thompson, 2015b). Similarly, one study found that repeat assaulters exhibited more callous attitudes toward women and were more likely to have histories of adolescent delinquency, as opposed to single assaulters who often described the assault as a result of inexperience or misperceptions of consent that they later corrected (Abbey & McAuslan, 2004). Therefore, it is possible that antisocial traits and hostile attitudes are more salient predictors of perpetration for forcible SA perpetrators, whereas beliefs about consent play a more important role for the larger group of nonforcible SA perpetrators. Future studies should consider that different subtypes of perpetrators may require different forms of intervention to address varying SA risk factors. For example, prevention programming can target beliefs about consent to specifically address coercive and incapacitated SA, which are the most prevalent forms of SA perpetration.

Despite the research on unclear sexual communication as a risk factor for SA, there has been limited exploration into the potential protective role for understanding active consent, or the use of clear sexual communication. Given inconsistent definitions of consent, it will be important for prevention programs and policymakers to determine how active consent can be most accurately defined and promoted. Our findings suggest that increasing college men’s beliefs in the importance of active consent, defined as verbal approval for sexual activity, can mitigate the risk for SA perpetration. Furthermore, a focus on reducing other risk factors, such as rape-supportive peer norms and beliefs, could instigate a positive shift in beliefs about consent. For example, one prevention program that included components focusing on SA-supportive social norms and education on consent demonstrated a positive impact on college men’s accurate perception of consent (Gidycz, Orchowski, & Berkowitz, 2011). In our study, moderate correlations between active and passive consent indicated that they are two different, although related, constructs. This suggests that prevention programs may benefit from focusing on both types of beliefs about consent.

Limitations

Although our longitudinal design provides support for causal relationships among the variables, it is not possible to establish definitive support for causal relationships without use of an experimental design. Another limitation of the study was its reliance on self-report data, which can be subject to recall bias. Despite use of behaviorally specific assessments and assurances of confidentiality, participants may have been hesitant to accurately report on sexual behavior. The study was also limited by only obtaining self-report data from men. Therefore, we could not assess female perpetration or more importantly, women’s perceptions of consent. Because we wished to maintain prospective associations between variables for mediation analyses, we limited assessment of perpetration to Waves 3 and 4. However, this resulted in a limited number of perpetration cases, specifically for the incapacitated and forcible tactic groups. Therefore the power to detect relationships between study variables in these tactic group models may have been attenuated.

Another limitation is that the consent measure was brief and only consisted of two items. Future studies should consider utilizing a more comprehensive measure of consent, such as the one proposed in the ARC3 survey (Administrator Researcher Campus Climate Collaborative [ARC3], 2015). This measure assesses beliefs that: (a) one should obtain consent at each step of a sexual encounter; (b) a partner can revoke consent at any point in time; (c) previous consent with a partner does not indicate future consent; (d) incapacitation means that consent is not possible; (e) being invited to a person’s place or being given mixed signals does not indicate consent. In particular, research indicates that alcohol involvement can increase misperceptions of sexual intent, and that sex while incapacitated is often not recognized as nonconsensual (Abbey et al., 2004; Littleton, Axsom, Breitkopf, & Berenson, 2006; Walsh et al., 2015). This highlights the importance of including measures that assess perceptions of consent in alcohol-involved incidents, as well as misperceptions of sexual intent.

Future Directions and Implications

Further studies are needed to examine women’s perception of consent and its relation to SA. Although it is important for prevention programs to emphasize bystander intervention and reducing risk for perpetration, education that focuses on increasing assertive sexual communication may help to reduce risk among potential victims (Gidycz, Lynn, et al., 2001). Findings from the current study indicate that clearer communication, including a focus on defining consent as continuous and active, could reduce the prevalence of SA. Furthermore, prevention programs will likely need to challenge perceptions that passive consent is sufficient for continued sexual activity. Finally, findings from this study highlight the importance of differentiating between SA tactics when investigating risk factors for SA, as well as means of preventing various forms of assault.

In addition, victims are generally less likely to acknowledge SA when it involves ambiguous definitions of rape that do not fit with stereotypic rape scripts (i.e., sex that does not involve physical force and/or is perpetrated by a known assailant; Littleton et al., 2006). Similarly, incidents involving passive consent or indirect aggression may not be considered to fit the stereotypic definition of SA or aggression in general. As a result, beliefs that passive consent is acceptable could lead to lower acknowledgment of instances of SA by both victims and perpetrators. Lack of acknowledgment of SA can lead to decreases in help-seeking behavior among victims (Walsh et al., 2015). Moreover, when victims disclose their experiences to support persons who fail to acknowledge the incident as SA, these invalidating reactions can lead to negative mental health consequences (Ullman & Filipas, 2001). Therefore, education is needed to explicate broader definitions of SA, to challenge acceptance of passive consent, and to emphasize the need for active consent. Such education could lead to increased acknowledgment of SA, help-seeking, and support for victims.

Another consideration for intervention involves personality characteristics associated with rape-supportive beliefs, perceptions of consent, and SA. The fact that personality traits played a role in the modeling of cognitions and perpetration in this study suggests that tailored interventions for high-risk individuals may be beneficial. Men with antisocial personality traits may require more intensive intervention. For example, some existing interventions incorporate components to improve empathy for women and SA victims (Banyard, Plante, & Moynihan, 2005). Alternatively, other prevention programs intervene at earlier developmental stages (e.g., middle and high school) and could potentially prevent solidification of antisocial characteristics, coercive behavior, and related beliefs (e.g., Foshee et al., 1996). Although certain interventions have been developed for high-risk groups such as fraternities (Foubert, 2000), further research is needed to determine whether tailored interventions offer an additive effect beyond universal approaches.

Finally, the important role of rape-supportive beliefs and peer norms in this study highlights the value of bystander interventions that engage peers in challenging rape-supportive beliefs, and in intervening when they observe risky situations. Such programs have demonstrated effectiveness in decreasing rape myth acceptance, increasing knowledge about SA, and improving willingness and confidence to intervene as a bystander (e.g., Banyard, Moynihan, & Plante, 2007; Coker et al., 2011; Gidycz, Layman, et al., 2001; Moynihan et al., 2015). Fewer studies have demonstrated the impact of prevention programs on SA victimization and perpetration. However, one study of college men who participated in a social norms and bystander education program showed that participants engaged in less sexual aggression and were less likely to associate with sexually aggressive peers, in comparison to a control group (Gidycz et al., 2011). Participants were also more likely to perceive their peers as willing to engage in bystander intervention, and were more likely to accurately label a nonconsensual sexual scenario as SA. A second study of an online intervention that focused on altering normative beliefs about SA and increasing bystander behavior also demonstrated positive effects on bystander intervention and SA perpetration (Salazar, Vivolo-Kantor, Hardin, & Berkowitz, 2014). In sum, important targets for prevention programming include clarification of definitions of SA and consent, teaching sexual and relationship communication skills, challenging peer norms, and increasing bystander intervention skills. The most efficacious prevention programs will likely extend beyond brief educational interventions and include prolonged efforts to change risk factors at interpersonal, community, and societal levels (DeGue et al., 2014).