Evidence for Multiple Classes of Sexually-Violent College Men

Abstract

Objective

Research on college sexual violence perpetration suggests there are multiple groups of male perpetrators. It is important to understand the distinctions between perpetrator subgroups to determine appropriate prevention strategies, as multiple strategies may be necessary to address multiple types of perpetrators. However, previous studies on subgroups of sexually-violent college men have relied on theoretically based distinctions, and there is currently no consensus on how to best classify perpetrators based on their sexually violent behaviors. The purpose of the current study was to take a data-driven approach to identifying subgroups of sexually-violent college men to help clarify: 1) the number and size of cohesive subgroups of sexually-violent college men, and 2) the types of behaviors that characterize each group.

Methods

1,982 college men across five universities in the U.S. self-reported their past sexually-violent behaviors, using the Sexual Experiences Survey (SES; Koss et al., 2007).

Results

Latent class analysis uncovered evidence for three groups: (1) a group unlikely to perpetrate any SV (88.6%); (2) a group likely to perpetrate SV using coercive tactics (verbal coercion or victim intoxication), but unlikely to use physical force (9.8%); and (3) a group likely to perpetrate the full range of SV (1.5%). Although the coercive tactics group was composed of men unlikely to use forceful tactics, it included the majority who attempted or completed rape based on legal definitions.

Conclusions

Our findings suggest that there are multiple, distinct perpetrator subgroups and signal the need for multiple prevention approaches, including approaches that address campus social norms.

Sexual violence (SV) is endemic among college students. Former U.S. President Obama (2014) convened a White House task force to protect students from sexual assault that led to the It’s On Us campaign against sexual violence, and the U.S. Department of Education recently investigated over 150 colleges and universities for potentially mishandling sexual misconduct cases (U.S. Department of Education, 2014). The government of Ontario, Canada recently passed sweeping legislation and launched the It’s Never Okay action plan to prevent sexual violence on college campuses (Office of the Premier, 2015). Furthermore, the respected student newspaper of Trinity College Dublin declared 2014 “the year we finally started talking about sexual assault on university campuses” (Page-Cowman, 2015). These efforts and discussions currently occur within cultures that define rape and rapists according to rape scripts, including a perception that rapes are committed by a small group of men who are unmistakably different from the average man (Ryan, 2011). Consequently, some survivors’ accounts that do not fit rape scripts are discounted, and their experiences reclassified as merely “bad hookups” (Epifano, 2012). College men who might truly have perpetrated SV may not face sanctions because their behavior does not seem prototypical (Author citation). We must better understand college sexual aggressors, and examine characteristics of different types of sexual aggressors, to inform appropriate and effective prevention and adjudication policies. To this point, however, there has not been a comprehensive, data-driven classification of college men based on the types of sexually-violent acts they perpetrate and the tactics they use to coerce sexual contact.

The Centers for Disease Control and Prevention defines SV as sexual activity perpetrated against someone who is unwilling or unable to consent (Basile, Smith, Breiding, Black, & Mahendra, 2014). Previous research has found that between 25% and 35% of college men report perpetrating SV (Abbey & McAuslan, 2004; Hines & Saudino, 2003; Koss, Gidycz, & Wisniewski, 1987; Author citation). Over 1 in 10 college men report engaging in oral, anal, or vaginal penetration of a person who is unable or unwilling to consent (Author citation), which is the Federal Bureau of Investigation’s definition of rape (FBI, 2013). Researchers have previously used theoretical or legal conceptualizations, rather than statistical modeling, to classify college men’s sexually-violent behavior (DeGue, DiLillo, & Scalora, 2010; Lyndon, White, & Kladec, 2007; Author citation). Nevertheless, there are several inconsistencies across classification systems, which has confounded attempts to understand characteristics of men in different perpetrator subgroups.

Findings based on these predetermined classification strategies suggest college men who commit SV using coercion (i.e., verbal coercion and other non-forceful tactics) outnumber men who use a combination of coercion and force tactics (i.e., physical force or threatening physical harm) by a 4:1 margin (Lyndon et al., 2007). However, researchers have used different criteria for what constitutes coercion or force, with alcohol-facilitated SV classified along with forceful tactics or verbally coercive tactics depending on the study (DeGue & DiLillo, 2004; Lyndon et al., 2007). The inconsistency in defining groups makes it difficult to draw conclusions regarding any meaningful differences that may exist between perpetrator groups.

Other studies have classified men who use alcohol-facilitated SV as separate from men who use force or men who use verbal tactics only, and have found that they tend to report more antisocial tendencies, hostility toward women, and risky behavior compared to men who use verbal tactics exclusively (Abbey & Jacques-Tiura, 2011; Author citation). College students view SV by means of victim intoxication as less socially acceptable than verbal pressure, yet they see intoxication tactics as more socially acceptable than forceful tactics (Struckman-Johnson & Struckman-Johnson, 1991). Thus, differences between men who use verbal coercion, victim intoxication, or physical force may in part reflect willingness to violate certain social norms. These norms are likely to be descriptive as opposed to injunctive, as SV by verbal pressure or victim intoxication may not be sanctioned explicitly, but rather describe how some college men behave. That said, previous studies that differentiate perpetrators who use alcohol-facilitated tactics have used measures that define alcohol-facilitated tactics in inconsistent ways (e.g., deliberately giving the victim alcohol or drugs versus taking advantage of a victim who is passed out or otherwise incapacitated), again making it difficult to draw meaningful conclusions about the characteristics or makeup of this group. Additionally, although college perpetrator subgroups have primarily been distinguished by the tactics they use to perpetrate SV, perpetrators can also be classified by the severity of SV acts (e.g., unwanted touching or kissing versus intercourse; Abbey, Parkhill, Clinton-Sherrod, & Zawacki, 2007).

Given the disagreement in how to classify perpetrator subgroups based on their SV behaviors, a data-driven approach is needed to understand which types of behaviors are likely perpetrated by cohesive subgroups of perpetrators. Furthermore, unlike prior studies on perpetrator subgroups, the current study included data from five different universities across the United States, which increases generalizability of the findings and allows for a data-driven approach to analyzing low base rate sexual violence. We conducted the present study to answer three questions: 1) are there multiple subgroups of college men indicated by their self-reported types and tactics of SV perpetration? 2) if so, what behaviors distinguish group membership?, and 3) what percentage of the men in each subgroup reported attempted or completed rape? Although perceptions regarding college men who perpetrate rape and other forms of SV suggest one relatively homogeneous small group of men who perpetrate a wide range of SV (Lisak, 2011), based on the research reviewed, we hypothesized at least two groups of perpetrators would emerge from the analysis. By identifying cohesive subgroups of perpetrators based on their SV behaviors, our findings may help further our understanding of college perpetrator heterogeneity and facilitate prevention programing development and implementation that is targeted toward distinct perpetrator subgroups.

Method

Participants

Data used for this study came from six independent datasets collected at five different universities. Two datasets from the southeast region were collected at the same university. These universities were not selected specifically for this study, but rather, researchers at each university volunteered their data to be part of a repository of data using the revised version of the Sexual Experiences Survey – Short Form Perpetration (SES-SFP; Koss et al., 2007). Participant recruitment and research design procedures were similar across sites, but they did vary somewhat. See Table 1 for an overview of the study procedures for each of the six datasets that contributed to this data. Institutional review board approval for all study procedures was obtained at all institutions by the local investigators prior to data collection, and informed consent was obtained for all participants. Data from each university were gathered using convenience sampling; sample sizes and data collection stopping rules for the respective studies were determined by the individual investigators based on their primary research questions and statistical power considerations. Across sites, all participants were provided with a list of counseling resources at the end of the survey. The combined dataset initially included survey data from 2,203 undergraduate students. One of the datasets included 215 female participants, who were excluded from our analysis to maintain consistency with the extant literature on male-to-female SV perpetration. We also excluded 6 participants who were under the age of 18 at the time of survey completion. The final sample included 1,982 college men who were 19.33 (SD=2.5) years old on average. Across datasets, the mean age ranged from 18.55 to 21.35 years old, and there was considerable variability across datasets in racial/ethnic makeup (see Table 2).

Table 1.

Overview of study procedures for each dataset

Dataset	Study Procedure
Dataset 1.	This sample contained 132 eligible male students enrolled in a small Midwestern university. Participants were recruited for a study that investigated perceptions of sexually coercive behaviors and past experience with sexual violence. Participants completed anonymous surveys online and were compensated with course credit.
Dataset 2.	This sample contained 167 eligible male students enrolled in a large Southeastern university. Participants were recruited through a psychology department participant pool for a study examining bystander intervention. Participants completed anonymous surveys online and were compensated with course credit.
Dataset 3.	This sample contained 156 eligible male students enrolled in a large Southeastern university. Similar to dataset 2, these participants were recruited through a psychology department participant pool for a separate study examining bystander intervention. Participants completed anonymous surveys online and were compensated with course credit.
Dataset 4.	This sample contained 341 eligible male students enrolled in a medium-sized Southeastern university. These participants were recruited from a psychology department participant pool for a study involving peer networks and sexual violence. Participants completed anonymous surveys online and were compensated with course credit.
Dataset 5.	This sample contained 392 eligible male students enrolled in a large Southeastern university. These participants were recruited from a psychology department participant pool for a study on the relationship between psychosocial factors and sexual violence. Participants completed anonymous surveys online and were compensated with course credit.
Dataset 6.	This sample contained 794 eligible first-year male students at a large Southeastern university. These participants were recruited via email, flyers, and advertisements in the university newspaper for a longitudinal study on attitudes toward women and sexual behaviors. Participants completed paper-and-pencil surveys that were masked for anonymity using a random-digit ID code and were compensated $20. Data used for the current study came from the first assessment at the end of students’ first spring semester.

Table 2.

Participant demographics by dataset

Dataset	n	Mean Age (SD)	Caucasian	African Am.	Asian Am.	Hispanic	Other/ Not Identified	Location
1	132	21.35 (4.42)	124 (93.9%)	1 (0.8%)	1 (0.8%)	1 (0.8%)	5 (3.8%)	Midwest
2	167	21.21 (4.91)	52 (31.1%)	46 (27.5%)	47 (28.1%)	—	22 (13.2%)	Southeast
3	156	20.39 (3.76)	49 (31.4%)	51 (32.7%)	41 (26.3%)	—	15 (9.6%)	Southeast
4	341	18.55 (0.70)	207 (60.7%)	70 (20.5%)	26 (7.6%)	16 (4.7%)	22 (5.9%)	Southeast
5	392	19.55 (1.44)	278 (70.9%)	21 (5.4%)	57 (14.5%)	15 (3.8%)	21 (5.4%)	Southeast
6	794	18.58 (0.83)	710 (89.5%)	57 (7.2%)	0 (0%)	2 (0.3%)	25 (3.1%)	Southeast
Total	1982	19.33 (2.5)	1420 (71.6%)	246 (12.4%)	172 (8.7%)	34 (1.7%)	110 (5.5%)

Measure

Sexual violence was measured using the SES-SFP (Koss et al., 2007), which has been found to have strong internal consistency (α = .98), test-retest reliability, and predictive validity (Johnson, Murphy, & Gidycz, 2017). This scale consists of behaviorally-based items that assess seven different sexually-violent acts: unwanted kissing or touching, oral sex, vaginal penetration, anal penetration, attempted oral sex, attempted vaginal penetration, and attempted anal penetration; and five different SV tactics: verbal pressure, getting angry, taking advantage when the victim was too drunk or “out of it” to stop what was happening, threatening physical harm, or using physical force. Each act is cross-listed with each tactic for a total of 35 items. It is important to understand college men’s complete history of SV behavior; therefore, data included in this study indicated SV behavior perpetrated from age 14 to the time of the survey. Responses to each of the 35 items were dichotomized (1/0) reflecting whether a man did or did not perpetrate a particular combination of SV act and tactic within this assessment period.

Data Analysis Plan

We estimated a series of latent class models, ranging from one to four classes, using Mplus version 7.2 (Muthén & Muthén, 2015) with maximum likelihood estimation and the expectation maximization algorithm. Latent class analysis is a model-based analytic technique used to identify unobserved subgroups within a sample (Lanza & Rhoades, 2013). This type of analysis requires a large sample size, and we were able to perform this analysis by combining multiple datasets. Standard errors were adjusted for data clustering by dataset in all models. We used multiple statistical indicators to determine the appropriate number of subgroups. The sample-size adjusted Bayesian Information Criterion (BIC, lower score indicates better model fit) and adjusted Lo-Mendell-Rubin Likelihood Ratio Test (LRT, significant p-value indicates significantly improved fit over a model with 1 fewer class) are both accurate in determining latent class number based on simulations (Nylund, Asparouhov, & Muthén, 2007). Entropy values and classification probabilities indicated how decisively each model classified college men into their most likely latent class, with higher values indicating better classification quality. Substantive interpretation and model parsimony also helped determine the final class model. In the case of the present study, college men were classified based on their responses to the 35 SES items. Because men’s SES scores were dichotomized, the subgroups can be represented by members’ average probability of reporting each SV act and tactic combination.

Results

Descriptive Statistics

Across the combined sample of 1982 college men, 81.7% (n = 1,619) of college men did not report engaging in any form of SV. Categorized by the most severe act reported, 4.7% (n = 93) of men reported perpetrating unwanted touching or kissing; 2.1% (n = 41) attempted but did not complete verbal sexual coercion; 4.0% (n = 79) reported perpetrating verbal coercion; 1.5% (n = 30) reported attempted but not completed rape; and 6.1% (n = 120) reported perpetrating rape according to the FBI definition (2013). Rates of perpetration varied across datasets, with non-perpetrators constituting 66% of participants from dataset 3 and 89.5% of participants from dataset 5 (see Table 3).

Table 3.

Percentage of participants in each SES category (grouped by most severe SV type perpetrated) across the 6 samples

Dataset (Region)	Non- Perpetrator	Unwanted Sexual Contact	Attempted Verbal Coercion	Verbal Coercion	Attempted Rape	Rape
1 (Midwest US)	74.2%	9.8%	1.5%	4.5%	3.0%	6.8%
2 (Southeast US)	77.8%	4.2%	0.6%	3.6%	1.2%	12.6%
3 (Southeast US)	66.0%	9.0%	3.2%	9.6%	3.8%	8.3%
4 (Southeast US)	76.0%	7.0%	2.3%	4.7%	2.3%	7.6%
5 (Southeast US)	89.5%	3.3%	2.6%	2.0%	0.3%	2.3%
6 (Southeast US)	85.4%	2.8%	1.9%	3.5%	1.1%	5.3%

Model Selection

All models tested terminated normally by replicating the best loglikelihood value (see Table 4). Statistical indicators of model fit were inconsistent—although the p value associated with the one- versus two-class adjusted LRT was nonsignificant, there was a notable improvement (36.6%) in BIC. The three-class model fit the data better than the two-class model based upon both the adjusted LRT and BIC (16.7% improvement). Finally, although the BIC was slightly lower for the four-class model compared with the three-class model (2.2% improvement), the adjusted LRT suggested the additional estimated class did not significantly improve the model fit. In addition, two classes in the four-class model essentially displayed the same perpetration pattern. We therefore selected the three-class model based upon a combination of statistical indicators and interpretability of the class structure.

Table 4.

Estimated fit statistics for competing latent class models

	Number of Latent Classes
	One	Two	Three	Four
Adjusted BIC	17669.43	11053.54	9318.26	9117.25
Adjusted LRT	---	6750.13	1887.30	358.64
p value	---	1.00	< .001	.57
Entropy	---	.99	.98	.96
LC Probabilities	1.00	.96–1.00	.96–1.00	.95–1.00
Smallest Class, %	100%	6.0%	1.5%	1.3%

Note. BIC = Bayesian Information Criteria; LRT = Likelihood Ratio Test; LC = Latent Class.

Class 1: Low likelihood of SV perpetration

The first class comprised most of the men in the combined sample (88.6%, n = 1757) and had a low probability of perpetrating most forms of SV, with slightly elevated probabilities of engaging in unwanted sexual contact (see Figure 1). Of the 138 men in this class who perpetrated SV, 87.7% (121 of 138) engaged in verbally coercive or otherwise unwanted sexual contact that did not meet the definition of attempted or completed rape. Fewer than 1% of men in this class (15 of 1757) endorsed behavior that meets the legal definition of rape, which may reflect false positives that occur in self-reports of violent behavior (Hamby, 2005, 2009). All men in this group endorsed two or fewer SES items.

Figure 1.

Latent classes of college men illustrated by estimated probabilities of perpetrating each SV type/tactic combination. On the x-axis: 1= unwanted touching; 2 = oral sex; 3 = vaginal sex; 4 = anal sex; 5 = attempted oral sex; 6 = attempted vaginal sex; 7 = attempted anal sex

Class 2: Perpetrators using verbal coercion and alcohol

The second class included 9.8% (n = 195) of the combined sample; all of these men reported perpetrating some form of SV. Specifically, 75 (38.5%) perpetrated rape, and an additional 28 (14.4%) attempted but did not complete rape (Table 5). As depicted in Figure 1, men in this subgroup, relative to men in the low likelihood class, were more likely to use verbal pressure, expressions of anger, or take advantage of victim intoxication to facilitate unwanted touching or kissing, oral sex, or vaginal sex. However, these men were unlikely to use physical force or threats of physical harm to facilitate any type of SV; they were also unlikely to attempt or complete unwanted anal penetration using any tactic. This subgroup contained the majority of men (68.7%, 103 of 150) who attempted or completed rape in the combined sample, a majority by exclusively taking advantage of victim intoxication (69.4%, 73 of 103).

Table 5.

Number of college men in each class who perpetrated SV, categorized by their most severe SV type perpetrated

Subgroup (n)a	Unwanted Contact, No. (%)b	Attempted Coercion, No. (%)b	Coercion, No. (%)b	Attempted Rape, No. (%)b	Rape, No. (%)b	Total, No. (%)b
Non-perpetrator or Unwanted Contact (1757)	86 (4.9)	21 (1.2)	14 (0.8)	2 (0.1)	15 (0.9)	138 (7.9)
Coercion Only (195)	7 (3.6)	20 (10.3)	65 (33.3)	28 (14.4)	75 (38.5)	195 (100.0)
Force and Coercion (30)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	30 (100.0)	30 (100.0)
Total	93 (25.6)	41 (11.3)	79 (21.8)	30 (8.3)	120 (33.1)	363

^aBased on most-likely class membership; will slightly differ from percentages based on posterior probabilities.

^bFirst three rows: percentage based on total class n; Last row: percentage based on total number who perpetrated SV (n=363)

Class 3: Perpetrators using all tactics

The third class was quite small, containing only 1.5% (n = 30) of the combined sample; however, all of these men reported perpetrating rape. The majority in this group endorsed perpetrating rape by force or threat of force (86.7%, 26 of 30). They accounted for a small yet important portion (20%, 30 of 150) of the men in the combined sample who attempted or completed rape (Table 5). Figure 1 illustrates that men in this class were likely to engage in a broad range of SV, encompassing multiple acts and tactics.

Discussion

The goal of this study was to determine whether there are classes of college men indicated by the type of SV they perpetrate and the tactics they use, and how these classes should be defined. Data from nearly 2,000 college men across five universities in the United States support the existence of three classes: (1) a class of college men with low likelihood of SV perpetration, (2) a class who is likely to use “coercive” tactics (i.e., verbal coercion or victim intoxication) to perpetrate a range of SV acts, and (3) a class that was relatively likely to perpetrate all SV acts and tactics. The coercive tactics class accounted for the majority of those who endorsed at least one attempted or completed rape, using mostly victim incapacitation tactics; but they were unlikely to perpetrate SV using force, threats of bodily harm, or to perpetrate unwanted anal sex.

There is an emerging consensus that men who commit SV are heterogeneous in regard to the timing of their assaults, the tactics they employ, and the risk and protective factors that predict their sexually-violent behavior (Abbey et al., 2012; Author citation; Thompson et al., 2013). Our findings lend further support for the existence of multiple distinct classes of perpetrators. Furthermore, our findings help advance research on perpetrator heterogeneity, as prior research on perpetrator subtypes based on SV behaviors had classified men using theoretically-based distinctions, with little consensus across studies for how to define the subtypes. For example, there was a lack of clarity regarding whether men who used alcohol-involved tactics should be classified along with those who used verbal tactics, those who used physical force, or as a separate subtype. Additionally, there had been inconsistencies in how victim intoxication as a tactic was defined across studies, and these definitions did not always match federal definitions for rape. In contrast, the revised SES contains a victim intoxication item in line with the FBI definition of rape, which allowed us to examine whether men who commit rape using different tactics (victim intoxication or force) form part of the same subgroup. In this latent class analysis, men who committed rape by victim intoxication without using force were frequently classified along with men who used verbal tactics, rather than men who committed rape by force. Theoretical approaches are valuable to understanding and classifying sexual assault perpetrators; the data-driven classifications suggested by the current results present an additional perspective that may help guide researchers and practitioners when conceptualizing the issue and weighing prevention strategies.

Our findings are in line with other past research that has uncovered a relatively small subgroup of men who report alarming SV patterns, in terms of their persistence across time or their wide range of violence perpetrated, including use of force tactics (Abbey et al., 2012; Author citation; Author citation). Although these men constitute the minority of SV perpetrators, their behavior mirrors rape scripts of consistent perpetrators that tend to use force to engage in nonconsensual sex (Ryan, 2011). This class of perpetrators who are likely to use physical force may be more visible to campus administrators because their assault patterns and characteristics increase the likelihood that they are reported to campus officials or law enforcement (Clay-Warner & McMahon-Howard, 2009). This might partially explain the assumption that all college men who perpetrate SV are similar to incarcerated rapists (Lisak, 2011).

Results from this study support previous findings that the majority of SV perpetrators use tactics other than threats of physical harm or force (Abbey & Jacques-Tiura, 2011; DeGue et al., 2010; Lyndon et al., 2007; Author citation). Given that college students perceive SV by means of verbal coercion or victim intoxication as more acceptable than forced sex (Struckman-Johnson & Struckman-Johnson, 1991; Warkentin & Gidycz, 2007), men in the coercive tactics group may internally or externally justify their behavior under the assumption that it is more permissible or less harmful than other forms of rape. This does not minimize the severity of these men’s behavior—many engage in nonconsensual sex by intentionally taking advantage of intoxicated women, which constitutes rape (FBI, 2013). The item that assessed SV via incapacitation explicitly asked about sexual encounters in which the respondent believed the other person was unable to consent (e.g., “I put my penis or I put my fingers or objects into a woman’s vagina without her consent by taking advantage when they were too drunk or out of it to stop what was happening.”) Thus, the concern that “drunken hookups” are misinterpreted as rape (Sokolow, 2015) is unfounded in this case. In fact, our results suggest that alcohol-facilitated SV deserves particular attention due to the finding that the majority of sexual assault perpetrators are likely to use these tactics.

Although the primary focus of this study was on perpetrators rather than non-perpetrators, approximately 8% of men who were classified into the low likelihood group perpetrated some form of SV, primarily unwanted contact. Although only a small minority of this class perpetrated rape (<1%), the classification of men who had perpetrated rape into a group with primarily non-perpetrators was surprising. Men in this group who endorsed rape endorsed only one item on the SES, which is atypical for men who commit rape, as prior studies have found that men who commit more severe SV acts tend to perpetrate additional acts and use multiple tactics (Author citation; Lyndon et al., 2007). More generally, across all forms of violent behavior, people are unlikely to perpetrate only extreme acts of violence; they typically demonstrate an escalation in violence severity across time (Elliot, 1994; Piquero, Jennings, & Barnes, 2012). The patterns of responding among men in the low likelihood group who reported rape do not match what we know about men who perpetrate rape, or people who perpetrate violence in general. Therefore, these specific self-reports should be interpreted with caution and may reflect measurement error or intentionally fabricated responses. This unexpected finding suggests that future SV research should explore in more depth the small group of college students who report rape, but no other SV behaviors, to assess the veracity of their self-report.

Limitations

Although generalizability of these findings is strengthened by the inclusion of data collected from male students attending universities in different regions of the United States, much of the data were collected from first and second-year students. This limitation likely explains the lower base rate of SV reported in this study compared with studies that surveyed men across four college years (White & Smith, 2004; Author citation). Additionally, the Bureau of Justice Statistics reports substantial variability in rates of perpetration across campuses (Krebs et al., 2016), which may also explain why the rates of perpetration in our study different from previous research. Further research is needed to clarify the causes of discrepancies in rates of SV across campuses.

All data included in this study were cross-sectional, retrospective self-reports of SV behavior perpetrated either before or during college. Although these findings shed new light on college men’s previous SV, it is not possible to determine acts perpetrated before versus during college from these data. Furthermore, all retrospective self-reports involve a certain degree of error, particularly self-reports of violence, a low base rate behavior that is susceptible to false positives (Hamby, 2005, 2009).

Research Implications

Our findings provide an additional approach to SV perpetrator classification beyond theoretically-defined classifications. By using a data-driven approach, we were able to identify typical behaviors of two distinct subgroups of college men, both likely to perpetrate SV. Our findings may help guide future research on campus SV perpetration; for example, the current findings suggest college men who use victim intoxication as a tactic, but not force, may be classified along with men who use only verbal tactics.

Future research should build upon the findings from the current study by exploring the predictors of group membership, to help facilitate the development of targeted prevention policies. For example, research could explore the role of individual attitudes toward women or perceived peer attitudes in predicting perpetrator group classification. In addition, future research should model longitudinal data on college men’s SV using latent transition analysis (Collins & Lanza, 2013) to determine if men change subgroups either during their transition to college or across the college years.

Clinical and Policy Implications

We found that most college men who perpetrate attempted or completed rape do not use physical force. This finding adds to the growing evidence that the majority of SV perpetrators may remain unreported, undetected, and unsanctioned due to their tendency to refrain from using physical force (Clay-Warner & McMahon-Howard, 2009; Du Mont, Miller, & Myhr, 2003), the relatively brief time frame within which they perpetrate (Author citation), or alcohol or drug use at the time of the assault by the victim, perpetrator, or both (Abbey et al., 2012; Author citation). Additionally, they may believe their behavior to be less reprehensible than forcible rape, as coercive tactics are seen as more acceptable than force (Struckman-Johnson & Struckman-Johnson, 1991; Warkentin & Gidycz, 2007). Therefore, our findings support SV prevention programming—not only for high-risk men, but for all students who play a role in developing and maintaining sexual norms—through expanding college students’ understanding of consensual sexual behavior to more closely match current public health (Author citation) and legal definitions (FBI, 2013).

Social norms correction campaigns may help to challenge misperceptions regarding peer approval of certain forms of SV, like the use of verbal coercion or taking advantage of a victim who is intoxicated (Berkowitz, 2010; Gidycz, Orchowski, & Berkowitz, 2011). Additionally, a primary aim of many bystander education programs, in addition to training students how to effectively intervene when they encounter SV, is to proactively create a campus climate intolerant of all forms SV (Banyard, Moynihan, & Plante, 2007; Gidycz et al., 2011; Katz & Moore, 2013). Nevertheless, changing problematic campus norms, what has been termed “rape culture” (e.g., Page-Cowman, 2015), may not directly influence the small yet dangerous group of men whose behavior does not appear to be constrained by social norms, as they are likely to perpetrate using forceful tactics that are overwhelmingly perceived as unacceptable by college students (Struckman-Johnson & Struckman-Johnson, 1991). This signals the need for peer-level interventions, such as the direct intervention component of bystander education programs (Banyard et al. 2007; Gidycz et al., 2011; Katz & Moore, 2013) as well as campus-level intervention programs, like increased monitoring of areas on campus where SV is more likely to occur (Taylor, Stein, Mumford, & Woods, 2013).

Conclusions

Colleges and universities are scrambling to stem the tide of campus SV by revisiting relevant policies and procedures and bolstering prevention programming. These findings indicate that institutions should adopt a multifaceted approach to campus SV response and prevention. Although some college men perpetrate a seemingly uninhibited range of SV, the majority of college men who commit SV draw well-defined limits on their behavior, and these limits coincide with college-student norms regarding sexual behavior, violence, and substance use (Warkentin & Gidycz, 2007). These findings indicate the need for policies and programs informed by psychological perspectives focused on (1) changing campus social norms regarding SV and substance use, and (2) campus-level interventions to increase bystander intervention and address dangerous areas on campuses and in surrounding communities.