Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Aug 28.
Published in final edited form as: Aggress Behav. 2025 Mar;51(2):e70021. doi: 10.1002/ab.70021

Exploring Measurement Strategies for Identifying Multiple-Perpetrator Sexual Violence: Higher Prevalence Rates Than Past Research

RaeAnn E Anderson 1,2,4,*, Erica L Goodman 1, Frances G Eby 1, Katya M Mickelson 1, Zoe D Peterson 3
PMCID: PMC12386137  NIHMSID: NIHMS2101890  PMID: 40014028

Abstract

Objective:

Prior research suggests that multiple-perpetrator sexual violence (e.g., gang rape) is very rare – reported by less than 2% of individuals. However, these estimates may be outdated and biased by measurement. The goal of this study was to explore the prevalence of multiple-perpetrator sexual violence and ways to measure it.

Methods:

Data were aggregated across several studies that shared similar methodology in recruiting online, anonymous samples of adults (Table 1). Study 1 evaluated victimization across five samples of 2491 participants. Study 2 evaluated perpetration across four samples of 2699 participants. All participants completed questionnaires in a randomized order.

Results:

Reported prevalence rates for victimization were higher than prior research (1.5%) and ranged between 2.8 – 10.9% (Table 2) with an average of 4.86% (see Table 3). Multiple-perpetrator victimization items were significantly correlated with other victimization items suggesting validity, phi = .182 - .371. Reported prevalence rates of perpetration ranged from 0.8% – 18.9% with an average of 6.43%. Multiple-perpetrator perpetration items were significantly correlated with other perpetration items, suggesting validity, phi = .190 - .582.

Conclusions:

Our findings suggest that multiple-perpetrator sexual violence is more common than previously suggested. Psychometric analyses suggest initial validity, utility, and acceptability for all examined measures that should be further investigated in primary data collections.

Keywords: sexual violence, assessment, multiple-perpetrator

Public Significance Statement:

Involvement in multiple-perpetrator sexual violence incidents is higher than previously suggested, ranging from 0.8% to 18.9% in this study compared to 1.5% in prior work. Future research should routinely examine multiple-perpetrator sexual violence to learn more.

Research on sexual violence has rapidly expanded since the 1970s (McCauley et al., 2019), as has public awareness about sexual violence (CBS News, 2017). Studies continue to demonstrate the lasting, negative impact of sexual violence can have on both mental and physical health (Dworkin et al., 2017; Koss, 1993). And sexual violence is common. Data suggest that at least one in five college women in the United States will experience sexual violence in college (Muehlenhard et al., 2017); these numbers are higher for marginalized groups such as LGBTQ+ individuals (de Heer & Jones, 2017; Trottier et al., 2021) and Indigenous communities (Rosay, 2016). Research also suggests that different groups may experience different forms and characteristics of violence (Rosay, 2016). One unique, but somewhat understudied form of sexual violence is multiple-perpetrator sexual violence (MPSV). Prior research suggests this form of sexual violence may be rare (Gidycz & Koss, 1990) but is especially likely to result in physical injury, posttraumatic stress symptoms, and increased risk for suicide (Raj et al., 2022; Ullman, 2007). However, much of the research that highlights the clinical impact of MPSV does not or cannot speak to how often MPSV occurs or how it should be measured. However, given the demonstrated clinical impact of MPSV (Raj et al., 2022), and the research documenting differences between majority and marginalized groups in experiences of sexual violence (Rosay, 2016), it is imperative that ongoing work investigates how to optimally measure sexual violence so as to accurately ascertain the scope and impact of these experiences across populations. This methodological study aims to focus on documenting and describing how MPSV has been measured in one archive of secondary data to explore whether MPSV is truly rare in college/community samples and whether there are any methodological features that facilitate or hinder recognition of MPSV. Specifically, we report prevalence rates, validity correlations (phi), utility (number of unique cases identified by the addition of that item), as well as measurement and sample type characteristics. Our goal is to demonstrate whether MPSV is actually rare as suggested by prior research (Gidycz & Koss, 1990) and highlight characteristics of measurement that may be useful for further investigating MPSV.

Terms

We use the term sexual violence as an umbrella term to encompass any form of sexual contact without consent, consistent with the Centers for Disease Control and Prevention terminology (CDC; Basile et al., 2014). Rape is the most severe form of sexual violence and is defined as penetration by means of incapacitation from substance use, physical force, or threats of physical force (Federal Bureau of Investigation, 2013). We use the term victimization to refer to the experience of those harmed by sexual violence and perpetration to refer to the behavior of those imposing sexual activity on another. Finally, we use the term multiple-perpetrator sexual violence to refer to any type of perpetration behavior by two or more individuals in the same incident. We crafted this definition from a combination of our experiences as scientists/clinicians, reviewing the literature, and measures reviewed and cited in this study. Our definition is consistent with the term “multiple perpetrator rape” as used in some of the key studies on MPSV (da Silva et al., 2015; Horvath & Kelly, 2009); however, we chose to use the term sexual violence instead of rape to be clear that we include non-penetrative acts.

Prior literature has used a wide variety of terms to refer to this same phenomenon, including gang rape, group sexual assault, and even more colloquially “running a train.” We chose to use the term multiple-perpetrator sexual violence to be precise and clear and to avoid any stereotypes associated with more colloquial terms. Notably, this term could encompass a wide range of co-offending behaviors and tactics. This could include incidents where the mere physical presence of another individual was perceived as coercive and the second individual did not perform any other coercive tactic beyond their presence being intimidating. Multiple-perpetrator incidents could also involve two or more individuals who each utilized a different tactic, such as one person being physically forceful while another person said bullying and demeaning comments. This would be common in many documented incidents wherein a group of individuals intimidates the individual and while some individuals do penetrate the victim, not all individuals do. For example, it could be made clear to a victim that escape is not possible by three individuals blocking the doorway/exit while one says demeaning comments. Thus, the tactic of physically blocking their escape was a group effort, only one person said demeaning comments but it added to the humiliation or fear of the situation. Further, even if not all individuals took part in the sex act—such as two of the three individuals holding the victim down while the third sexually assaulted the victim—all would still be considered perpetrators because they contributed to the assault taking place. This is consistent with other research on MPSV documenting the heterogeneity of MPSV and reflects the complexities that may occur with hierarchical group dynamics – there are widely varying degrees of complicity and roles by those perpetrating MPSV (Horvath & Woodhams, 2013).

Characteristics of Multiple-Perpetrator Incidents

Prior research has suggested that multiple-perpetrator incidents are more explicitly violent than single-perpetrator incidents (Tillman, 2011; Ullman, 2007) and reflect a greater degree of severity of violence (Jewkes et al., 2015). Multiple-perpetrator incidents are more likely to include stranger assailants, substance use, and weapons than single-perpetrator incidents (Gidycz & Koss, 1990; Harris et al., 2020; Ullman, 2007). Those victimized in this manner reported higher levels of fear (Gidycz & Koss, 1990), injury (Raj et al., 2022), were more likely to report symptoms of PTSD, and more likely to report suicidality (Raj et al., 2022) compared to single perpetrator victims (Ullman, 2007).

However, much of the extant research on multiple-perpetrator sexual violence comes from media studies, gender studies, and criminology. Media and gender studies have contributed significantly to the literature on MPSV, although naturally, the focus has been on how these crimes are communicated and the societal narratives and discourses constructed around these crimes (Grewal, 2016; Sela-Shayovitz, 2014). Entire books have been written on the nature of gender, sexuality, and fraternity gang rape specifically (Sanday, 1992) highlighting group dynamics (Porter & Alison, 2001). Notably, gender and media studies have examined MPSV in a variety of international contexts – including Australia, France, Israel, and India (Grewal, 2016; Krupa, 2015; Sela-Shayovitz, 2014) – suggesting the cross-cultural nature of MPSV. Yet, although this research highlights important societal trends and characteristics of incidents, it cannot speak to the prevalence rate of MPSV.

Criminological research has primarily utilized administrative records, such as police and hospital records demonstrating the violence of MPSV and sometimes the difficulty in prosecuting those who inflict MPSV (Bamford et al., 2016). One meta-analysis of multiple-perpetrator sexual violence included 15 studies and only two of the studies reviewed were not based on administrative records (Bamford et al., 2016). Although administrative records are valuable in understanding how institutions handle MPSV cases, this type of research cannot determine the frequency of MPSV nor evaluate the best measurement approaches because these records are collected with select subsamples (i.e., those willing to report to the police) and often without accompanying questionnaires that would facilitate measurement research.

When researched, reported prevalence rates in North American community and college samples have often been low – around 1.5 to 2.0% of the sample (Gidycz & Koss, 1990; Tillman, 2011). Notably, in the only available epidemiological data in North America, Gidycz and Koss (1990) recorded a low prevalence rate (1.5%) in a large (N = 3187) national sample of college women. However, some community and clinical samples that specifically recruited women with victimization histories have reported a relatively high number of victims reporting MPSV—18 and 20% of victimized American women (Raj et al., 2022; Ullman, 2007). Research from other parts of the world suggests MPSV is not rare. One South African study using two items, “How many times have you and other men had sex with a woman at the same time when she didn’t consent to sex or you forced her?” and “How many times have you and other men had sex with a woman at the same time when she was too drunk to stop you?” found a perpetration prevalence rate of 8.8% in a sample of nearly 1700 randomly selected men (Jewkes et al., 2015).

In spite of the media studies, gender studies, criminology (Bamford et al., 2016), and international research (Jewkes et al., 2015) on the topic, MPSV has not been the focus of much social science research in North American community or college samples (da Silva et al., 2015). It appears to be assumed that this type of sexual violence is rare, and as such, it is generally not included in standard instruments used to determine history of sexual violence such as the short form Sexual Experiences Survey (SES-SFV/SFP; Koss et al., 2007), the Revised Conflicts Tactics Scale (Straus et al., 1996), or the Post-Refusal Sexual Persistence Scale (PRSPS: Struckman-Johnson et al., 2019)1. In a systematic review of perpetration prevalence rates which reviewed different forms of sexual violence, multiple-perpetrator violence is not mentioned (Anderson et al., 2021a). The U.S. Center for Disease Control’s National Intimate Partner and Sexual Violence Survey, an ongoing surveillance study, does not measure MPSV.

Measurement Issues

As alluded to, the assessment of multiple-perpetrator sexual violence has been highly variable when conducted at all. Although the SES-Long Form Victimization/Perpetration (SES-LFV/LFP: Koss et al., 2007) did include items specifically designed to assess multiple-perpetrator attacks, these items were not included in the SES short form and therefore seem to be rarely administered. For example, in a systematic review of 20 years of sexual perpetration research in North America, so few studies utilized the SES-LFP data from this measure was not able to be analyzed separately (Anderson et al., 2021a); nor was the long form cited as used in any of the studies included in a systematic review of 15 years of college campus victimization research (Fedina et al., 2016).

Prior to the SES-LFV/LFP, multiple-perpetrator incidents were identified by asking those who endorsed any victimization item follow-up questions about the characteristics of the events (Gidycz & Koss, 1985; Ullman, 2007). This is reminiscent of, but not the same as, an “incident report,” to use the language of surveillance research like the National Crime Victimization Survey (Langton et al., 2017). Another term for this approach to measurement has been called the two-stage model, wherein the first stage screens for incidents using one to two very broad, descriptive items suggestive of sexual violence. Then, the presence of sexual violence is confirmed in a second stage via more detailed and specific questions (Cook et al., 2011). However, some research has suggested that using a two-stage model decreases reports of sexual victimization (Cook et al., 2011; Fisher, 2009), whereas other research has found only minor differences between one- and two-stage approaches for victimization reports (Krebs et al., 2017). Only one study is available examining self-reports of perpetration using one- and two-stages, but it suggests that two-stage approaches reduce perpetration reports to falsely low levels (Krebs et al., 2007).

In addition to the one vs. two-stage approach issue, sexual violence measurement is very sensitive to item phrasing and wording (Anderson et al., 2021a). Even wording differences that seem relatively minor such as using the phrase “unwanted” instead of “non-consensual” can have significant effects (Hamby & Koss, 2003). In an experimental study varying only the terms used to describe consent, the condition using behaviorally specific descriptions instead of the term “unwanted” increased sexual victimization and perpetration prevalence rates (Rueff & Gross, 2017). Another wording issue is whether the sexual behavior—for example “oral sex”—or the tactic—for example, “took advantage of intoxication”—is described first in the item. Items that describe the tactic first (i.e., “took advantage of intoxication to make you have oral sex”) in comparison to items that describe the sexual behavior first (i.e., “made you have oral sex by taking advantage of intoxication”) are psychometrically equivalent in terms of validity but record double the prevalence rate (Abbey et al., 2021). Given the range of existing approaches and the lack of data suggesting one is clearly superior to another, research should examine multiple potential ways to assess multiple-perpetrator sexual victimization and perpetration and examine how the characteristics of measurement may influence reporting prevalence rates.

Current Study

Prior research suggests multiple-perpetrator sexual violence may be uniquely terrorizing for victims and more likely to cause negative health sequelae than other forms of sexual violence. However, there is very little data available on the frequency of MPSV, and there is no psychometrically established method for identifying multiple-perpetrator incidents. The current methodological study aims to improve our understanding of sexual violence by examining whether MPSV is actually rare (<2%) in college and community samples by reporting prevalence rates and comparing them to the only available epidemiological prevalence rates – those reported in Gidycz and Koss (1990). We also aim to evaluate several possible methods for measuring MPSV. The items tested in this study span the range of the highly specific and legally precise language of the SES-LFV to the less specific but potentially more gender-inclusive language of the PRSPS-P and similar questionnaires. In the spirit of inclusivity and given recent data challenging the idea that women rarely perpetrate (Stemple et al., 2017), we utilized samples with participants of all genders.

Aims 1 and 2: Document and compare the prevalence rate of multiple-perpetrator sexual violence victimization (Aim 1, Study 1) and perpetration (Aim 2, Study 2) in multiple college and community worker samples to the only available national epidemiological prevalence rates, those reported in Gidycz and Koss (1990). We will also examine whether multi-perpetrator items identify unique cases that would have otherwise not been identified as sexual violence (e.g., utility). We will also examine rates of missingness as a proxy for the acceptability of the items. Acceptability refers to participants’ attitudes regarding a procedure or treatment (Sekhon et al., 2017). We hypothesized that multiple-perpetrator victimization (H1) and perpetration items (H2) would positively correlate (phi ≈ .4 ; a moderately strong positive association, as reported in Abbey et al., 2021 and Anderson et al., 2020) , with a non-MPSV composite score derived from other items on the same questionnaire. We expected multiple-perpetrator items to be correlated with a composite score representing the remaining items for two reasons: One, sexual violence is on average, a repeated experience (Basile et al., 2022; Busch-Armendariz et al., 2015). Recent nationally representative data from the U.S. suggests that 77-81% of those who experience sexual violence report multiple incidents (Peterson et al., 2024). Secondly, because each item on the questionnaire represents a different tactic, it is possible and perhaps even likely that individuals who reported multiple perpetrators will also report additional tactics such as verbal coercion, or use of weapons (Ullman, 2007). We will track different methodological features such as one vs. two-stage measurement, tactic- vs. sexual behavior-first items, and sample type to examine how these features may be associated with reported MPSV prevalence rates.

Method

Overview

This paper presents the results of two studies from aggregated data, see Table 1. Study 1 focuses on victimization across five samples, see Table 1. Study 2 focuses on perpetration across four samples. Participants consented before being able to access the survey. Because of the sensitive nature of the data and the fact that participants did not consent to sharing at the time of data collection, the datasets are available upon request to the first author. Data outputs are available here: https://osf.io/fyqst/. Data were collected under the supervision of MASKED IRBs and in accordance with the American Psychological Association ethics code. The authors have no conflicts of interest to disclose.

Table 1.

Sample demographics

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7
Sample type MTurk MTurk college students college students college students college students MTurk
Dates of data collection April-May 2018 April-May 2018 Spring 2016-Fall 2017 Spring 2016-Fall 2017 Fall 2021-Spring 2022 Spring-Fall 2010 February-April 2022
American Region, university type and size - - Great Lakes, public, large Great Lakes, public, large Northern Plains, public, medium Lower Midwest, public, medium -
Total n 428 265 673 792 333 843 636
%, women 47.4 46.0 54.5 72.6 51.1 75.8 5.8◊
% men 51.2 53.6 44.3 26.7 41.1 23.8 94.2
% trans or gender nonbinary 1.2 0.4 1.1 9.5 3.3 0.3 -
% heterosexual 84.8 87.5 85.6 84.7 82.9 89.8 -
% bisexual 10.5 9.1 5.6 8.5 6.3 4.7 -
% gay or lesbian 3.7 1.0 4.9 3.6 - 3.1 -
% other sexualities 0.2 2.2 6.4 3.1 6.4 2.2 -
% White 82.5 78.5 85.6 81.7 87.1 61.6 81.0
% Black or African American 11.0 11.7 9.4 13.1 3.6 25.4 6.9
% Asian or Asian American 6.8 9.4 3.7 2.9 3.9 4.5 3.5
% Indigenous or Native American 1.2 0 1.0 0.8 3.6 0.7 0.2
% other racial identity - 1.6 5.2 5.6* - 3.3 1.2
% Hispanic or Latino/a 7.5 7.9 3.7 - 5.1 3.1 5.2
M, SD age, range 32.47(7.34), 19-71 32.11(7.41), 19-64 19.5(3.6) 20.0(1.72) 18-34 19.8(7.5) 23.49(6.7) 38.38(11.96)
Prior peer-reviewed publications Anderson, Doctor et al., (2023); Anderson, Garcia et al., (2022) Anderson, Goodman, et al., (2021) Anderson et al., (2020); Anderson, Carstens Namie et al., (2021);Anderson, Goodman et al., (2023) Anderson et al., (2022); Gesser et al., 2024; Anderson et al., (2020) - Buday & Peterson (2015) -
Open in a new tab

Note. *many of these participants reported Hispanic or Latino heritage. ◊participants were only asked their sex assigned at birth. MTurk participants were required to be North American adults, with high-quality work (e.g., little missing or invalid data) and high completion ratings from prior tasks (>90% task completion).

Participants

Seven different samples were recruited representing a total of nearly 4000 participants. See Table 1 for sample demographics and characteristics and Table 2 for a summary of items administered. All samples were recruited for other primary research questions (in order: sample 1 - Anderson, Garcia et al., 2022; Sample 2 - Anderson, Goodman et al., 2021; Sample 3 - Anderson & Delahanty, 2020; Sample 4 -Anderson, Carstens Namie et al., 2022; Sample 5 – not applicable, undergraduate research project, no citation available; Sample 6 – Budy & Peterson, 2015; Sample 7 - Mickelson, 2020); thus, this paper is secondary analysis.

Table 2.

Items, Study descriptions, Sample descriptions, and Prevalence Rates

Item Source Stage
type		 Item type Sample Prevalence
rate for item		 Prevalence
rate, violence◊		 phi Utility
Victimization Items:
Study 1
1. Has someone ever acted as part of a group of two or more people in order to: [seven sexual acts listed below] Authors; 2007 SES-LFV one T Sample 1: 428 MTurkers 10.9% 45.3% .371 1
Sample 2: 265 MTurkers 10.9% 54.7% .311 0
Sample 5: 333 college students 3.0% 37.5% .194 1
2. [Has someone….sexual activity] Acted together with another person (or people) to overwhelm you Authors; PRSPS-V one T Sample 1: 428 MTurkers 9.3% 58.3% .278 0
Sample 2: 265 MTurkers 5.3% 59.6% .194 0
3a. On any particular occasion, did more than one person do these unwanted things to you? Authors; SES two SB Sample 3: 673 college students 3.3% 43.3% .203 3
3b. On any particular occasion, did more than one person act together, at the same time, to do unwanted sexual things to you? Authors; SES two SB Sample 4: 792 college students* 2.8% 39.7% .182 2
Perpetration items: Study
2
4. Have you ever acted as part of a group of two or more people to do these things to someone in order to…[seven sexual acts listed] Authors; 2007 SES-LFP one T Sample 1: 428 MTurkers 7.2% 26.3% .452 3
Sample 4: 792 college students* 1.2% 15.0% .278 2
5. [Have you…sexual activity] Acted together with another person (or people) to overwhelm them Authors; PRSPS-P one T Sample 1: 428 MTurkers 6.2% 38.8% .190 3
6. [I had…sexual activity] Acting as part of a group of two or more people who did these things after someone objected or was unable to give consent. 2007 SES-LFP one SB Sample 6: 843 college students 0.8% 24.7% .160 0
Sample 7: 636 MTurkers 18.9% 38.7% .582 3
Open in a new tab

Note. SES = Sexual Experiences Survey, PRSPS = Post-Refusal Sexual Persistence Scales. ◊This represents the overall prevalence rate for any type of sexual victimization or perpetration for the entire sample. Unless otherwise noted, men and women were approximately equally represented in the sample. *72.4% women. Stage: one or two type. Item type indicates, SB = sexual behavior first or T = tactic-first. Phi values represent convergent validity correlations with a composite or other violence items on the parent questionnaire.

Measures

All multiple-perpetrator items are listed in Table 2. All multiple-perpetrator items were administered as part of an overall questionnaire assessing sexual violence. The parent studies often administered experimental versions of sexual violence questionnaires as part of larger measurement studies. The key characteristics of the questionnaires are summarized in Table 2 with citations to other published studies for further details.

Multiple-Perpetrator Items.

Given the use of multiple datasets and the goal of comparing different types of items, seven different versions of the same basic item were tested. We used an unedited version of the SES-LFP item, shown as Item 6 in Table 2. Item 6 reads, “[I had…sexual activity] Acting as part of a group of two or more people who did these things after someone objected or was unable to give consent.” All other items were experimental items created by the authors, though modeled from this SES-Long Form item. Thus, beyond Item 6 we did not test standardized items. Notably, test-retest reliability data is available for Items 1, 2, 4, and 5 in Anderson et al. (2021b) supplemental materials but no validity data is available.

Item 1 was a tactic-first victimization version of this SES-Long Form item. In addition to testing SES-based items, we also tested items added to the PRSPS but tried to word these similarly to the SES items. Item 2, a PRSPS victimization item reads, “acting together with another person (or people) to overwhelm them.” Items 3a and 3b are two-stage items administered as follow-up questions after completing the SES-SFV. Item 4 is a tactic-first version of the SES-LFP item. Item 5 is a PRSPS item for perpetration

Questionnaires assessing victimization history.

The Sexual Experiences Survey – Short Form Victimization (SES-SFV, Koss et al., 2007).

All SES-SFV items are compound—a combination of sexual behaviors (7) and tactics (5) resulting in 35 items. The traditional format describes a sexual behavior as the stem of the item (e.g., “A man put his penis into my vagina, or someone inserted fingers of objects.”). The behavior is followed by a series of five sub-stems describing tactics used to perpetrate the sexual activity or attempted sexual activity (Koss et al., 2007). This structure is inverted in the tactic-first approach with the tactics as the main stems and sexual behaviors as sub-stems that are repeated. The validity of the SES-SFV for victimization has been demonstrated by correlations with psychological distress for women (Johnson et al., 2017) and with reports of family violence victimization for both women and men (Anderson et al., 2018). Levels of test-retest reliability for the tactic-first SES were 80% or better, with kappas ranging from .6 - .7 (Anderson et al., 2021b), which is similar or slightly better reliability than the sexual-behavior first SES (Anderson et al., 2018; Johnson et al., 2017).

The Post-Refusal Sexual Persistence Scale – Victimization (PRSPS-V: Struckman-Johnson et al., 2003).

The PRSPS-V contains 19 items that assess sexual victimization across five types of tactics: enticement, verbal coercion, misuse of authority, alcohol/drugs, and physical force. The PRSPS is unique in its ability to determine potential gender differences because it was specifically designed to assess victimization in both genders. Gender differences could be especially relevant to multi-perpetrator items due to the social and group aspects of these incidents (Park & Kim, 2016). Struckman-Johnson et al. (2003) found good construct validity by comparing written incident reports to questionnaire responses. One test-retest reliability study of 100 men and women found acceptable levels of agreement (kappa ≈ .6), comparable to the SES, over a one-week interval (Anderson et al., 2022).

Questionnaires assessing perpetration history.

The 2007 SES – Short Form Perpetration and Long Form Perpetration (2007 SES-SFP/LFP: Koss et al., 2007).

The SES-SFP is modeled identically to the 2007 SES-SFV. The validity of the SES-SFP is evidenced by correlations with trait aggression (Anderson et al., 2017) and rape myth acceptance (Johnson et al., 2017). The reliability of the SES-SFP is good based on two samples of college men, with test-retest reliability at 82-91% agreement (Anderson et al., 2017; Johnson et al., 2017). The Long Form of the SES includes an additional eight tactics for each of the same sex acts for a total of 56 items.

The Post-Refusal Sexual Persistence Scale – Perpetration (Struckman-Johnson et al., 2003).

The PRSPS-P is modeled nearly identically to the PRSPS-V. PRSPS-P scores are correlated with psychopathy suggesting convergent validity (Khan et al., 2017). A study of approximately 100 men and women demonstrated adequate test-retest reliability over one week (Anderson et al., 2021b).

Scoring.

All violence questionnaires and items were scored dichotomously wherein any response greater than 1 or a “yes” was considered affirmative. Composite scores were created that reflected any affirmative response including (overall prevalence rates) and separate composites that excluded the multiple-perpetrator items (for phi correlational analyses).

Procedures

The number of co-administered questionnaires in the Studies ranged from three to five, with the exception of Samples 4 and 5, where 15 co-occurring questionnaires of various lengths were administered. In all studies, demographics were counted as a single questionnaire. All studies were advertised in simple, straightforward language such as “questionnaires about sexual behavior” or “questionnaires about alcohol.”

Data Cleaning and Analytic Plan

If a participant did not complete at least one item on the sexual violence questionnaire they were excluded. This was highly consistent with comparison to attention check items (agreement > 90%). Chi-squares were computed to ascertain prevalence rates of endorsement, relationships to demographics, utility (i.e., unique cases) and missing data. Only significant demographic relationships are reported for space. Given the small number of non-cisgender participants, we did not test for cis- vs. transgender differences. Because participants could endorse more than one racial or ethnic identity we tested for these differences for the endorsed vs. non-endorsed identity (e.g., White vs. non-White). Finally, we tested for sexual orientation identity differences. Phi correlations (binary variables, or “present or not”) were computed to evaluate validity due to the use of dichotomous scoring (Guilford, 1941). Phi correlations are reported with a composite of the traditional items for the same measure excluding the multiple-perpetrator item. Rates of missing data for the multiple-perpetrator items were compared to the rate of missing data for composite variable representing the overall sexual violence questionnaire approximating acceptability in that participants are more likely to skip items that make them uncomfortable or that are difficult to understand (Fongwa et al., 2010). Sample 6 did not use composites in comparisons rather compared to one similar item consistent with the way the overall questionnaire was designed in that study.

Results

Prevalence rates for each item and overall violence prevalence rate (victimization or perpetration: Prevalence rate, violence) for each sample are also reported in Table 2. Prevalence rates are averaged for each item type and sample type and compared in Table 3. The designation of primary vs. aggregate sample was based mostly on timing of data collection. The percentage of violence affected (victimized or perpetrated) participants who reported MPSV is reported below for context but not listed in Table 2 for space.

Table 3.

Comparison of Averaged Prevalence Rates by Item-type across Studies

Item or Sample Type Average Prevalence
Rate, Total N		 Statistical
Comparisons		 Comparison to Gidycz
& Koss, 1.5%
Study 1
Victimization
A. One-stage SES 8.3%, 1,026 A=B; χ2(1) = 0.57, p = 0.451 χ2(1) = 117.39, p < 0.0001
B. One-stage PRSPS 7.3%, 693 B>C; χ2(1) = 19.51, p < 0.0001 χ2(1) = 77.45, p < 0.0001
C. Two-stage SES 3.1%, 1,465 A>C; χ2(1) = 32.83, p < 0.0001 χ2(1) = 13.08, p = 0.0003
D. Sexual-behavior first 3.1%, 1465 — (same as row C)
E. Tactic-first 7.2%, 1,026 E>D, χ2(1) = 22.24, p < 0.0001 χ2(1) = 89.89, p < 0.0001
AA. college sample 3.0%, 1,798 — χ2(1) = 12.93, p = 0.0003
BB. M Turk sample 9.1%, 1,386 AA< BB, χ2(1) = 54.57, p < 0.0001 χ2(1) = 152.46, p < 0.0003
Total average 4.86%, 2,491 — χ2(1) = 54.69, p < 0.0001
Study 2 Perpetration
W. Sexual behavior-first 12.6%, 1,479 —
X. Tactic-first 3.4%, 1,220 W>X, χ2(1) = 73.19, p < 0.0001 —
Y. One-stage SES 7.0%, 2,699 — —
Z. One-stage PRSPS 6.2%, 428 Z=Y; χ2(1) = 0.368, p = .5439 —
WW. college sample 1.0%, 1,635 — —
XX. M Turk sample 10.7%. 1,492 WW < XX, χ2(1) = 138.15, p < 0.0001 —
Total average 6.43%, 2,699 — χ2(1) = 98.12, p < 0.0001
Open in a new tab

Note. SES = Sexual Experiences Survey, PRSPS = Post-Refusal Sexual Persistence Scale. When a sample included more than one measure with the same feature, the average prevalence rate across the measures was used in characteristic comparisons. When a sample included more than one measure, the average prevalence rate across the measures was used to calculate the total victimization and perpetration averages.

Victimization Findings: Prevalence Rates, Validity, Utility, and Missingness.

Study 1 – one stage items

Sample 1.

Using Item 1, “Has someone ever acted as part of a group of two or more people…” (see Table 2), 10.9% of the total sample (n = 48) and 24.4% of victimized participants experienced multiple-perpetrator victimization. Regarding validity, the multiple-perpetrator victimization item was significantly correlated with a composite of other SES victimization items, phi = .371, p < .001. Regarding unique cases, a cross-tabulation comparing the cases identified by all other SES victimization items and the multiple-perpetrator item indicated that the addition of this item identified one unique case that would have otherwise been labeled nonvictimized. There were no statistically significant differences in the rate of missing data (χ2(1) = 2.02, p = 0.155) on the multiple-perpetrator item (5.6% missing) compared to a composite measure of any missing items on the traditional 2007 SES (3.6% missing).

Using Item 2, “Has someone…acted together with another person (or people) to overwhelm you,” (see Table 2), the prevalence rate of multiple-perpetrator victimization was 9.3% (n = 41) or 16.8% of the victims according to the PRSPS-V. Black individuals (n = 47) were more likely to report multiple-perpetrator attacks than non-black individuals, χ2(1) = 4.19, p = 0.041. Regarding validity, the multiple-perpetrator item was significantly correlated with a composite of other PRSPS victimization responses, phi = .278, p < .001. As for utility, this item did not identify any new cases not identified by other PRSPS-V items. Concerning missing data, 4.8% of the sample had missing data on the multiple-perpetrator question compared to 0% on the traditional PRSPS-V items, which was significantly different, χ2(1) = 11.90, p = 0.0006.

Aggregate Data, Sample 2.

Using Item 1, “Has someone ever acted as part of a group of two or more people…” 10.9% (n = 29) reported experiencing multiple-perpetrator victimization, and 23.93% of victimized participants reported experiencing multiple-perpetrator victimization at least one time. Those identifying as Hispanic/Latinx (n = 21) were more likely to report experiencing multiple-perpetrator victimization than non-Hispanic/Latinx individuals, χ2 (1) = 17.253, p < .001. The multiple-perpetrator victimization item was significantly correlated with other SES victimization items, phi = .311, p < .001. The multiple-perpetrator item did not identify any new cases of victimization that the traditional items did not. There was no significant difference in the rate of missing data for the traditional (0%) vs. multiple-perpetrator item (0.75%).

Using Item 2, “Has someone…acted together with another person (or people) to overwhelm you,” 5.3% of participants (n = 14) reported experiencing multiple-perpetrator victimization, representing 8.9% of the victimized. The only group difference in reporting history of multiple-perpetrator victimization was that those who identified as Hispanic/Latinx (n = 27) were more likely to report multiple-perpetrator victimization than those who did not, χ2(1) = 8.58, p = .003. The multiple-perpetrator item was significantly correlated with a composite of other PRSPS victimization items, phi = .194, p = .002. The multiple-perpetrator item identified no unique cases. There were no differences in the rates of missing data, which were less than 1%.

Aggregate Data, Sample 5.

Using Item 1, “Has someone ever acted as part of a group of two or more people…,” we found that 3.0% of college students endorsed multiple-perpetrator victimization, which was 8.0% of the victimized. Women were more likely to report experiencing multiple-perpetrator victimization than men, χ2(2) = 13.066, p = .004. There were no other demographic differences in endorsement. The multiple-perpetrator item was correlated with the endorsement of other SES victimization, phi = .194, p < .001. The item uniquely identified one case of victimization that would have been otherwise missed. The rate of missing data was equivalent to the average rate of other items, which was zero.

Study 1 – two-stage items

Sample 3.

In a college sample, item 3a, “On any particular occasion, did more than one person do these unwanted things to you?” 3.7% (n = 25) of participants reported experiencing multiple-perpetrator-instigated victimization, representing 9.9% of the victimized. There were no demographic differences in endorsement. However, an inspection of another follow-up question, “If so, how many people did these unwanted sexual things to you?” suggests some possible confusion with this item, as six respondents who endorsed the multiple-perpetrator item answered 0 or 1, suggesting that 2.8% (n = 19) is a more accurate prevalence rate for this sample. Responses to this item were positively correlated with other SES victimization items, phi = .203, p < .001. The multiple-perpetrator item identified three unique cases that were not identified by the traditional victimization questions. Twenty individuals (2.97% of total sample) had missing data on the multiple-perpetrator item which was significantly more compared to stage-one SES items, χ2(1) = 20.27, p < 0.0001.

Aggregate Data, Sample 4.

In a sample of college students, using item 3b, “On any particular occasion, did more than one person act together, at the same time, to do unwanted sexual things to you?” 3.3% (n = 26) reported experiencing multiple-perpetrator victimization, representing 7.69% of the victimized. There were no demographic differences in endorsement. Responses to this item were positively correlated with other traditional victimization items, phi = .182, p < .001. The multiple-perpetrator item identified two unique cases that were not identified by the traditional victimization questions. Thirty individuals (3.8% of the total sample) had missing data on the multiple-perpetrator follow-up item compared to zero missing data on the stage-one items, a significant difference, χ2(1) = 30.66, p < 0.0001.

Comparison across item, sample types and to prior data

Prevalence rates were combined across samples, see Table 3 to assess how questionnaire or sample level differences may impact prevalence rates. The average prevalence rate of multiple-perpetrator victimization for the one-stage items (Items 1 and 2) was 7.8%. For the two-stage items (Items 3a and 3b), the average prevalence rate of multiple-perpetrator victimization was 3.1%, significantly lower, χ2(1) = 32.83, p < 0.0001. We were unable to truly test the difference between sexual-behavior first items and tactic-first items because our sexual-behavior first items were also all two-stage items; although there was a significant difference between sexual-behavior first two-stage items (Items 3a, 3b, 3.1%) and tactic-first one-stage items (Items 1, 2, 7.2%), χ2(1) = 22.24, p < 0.0001. There were sample-type differences with college student samples (3.0%) reporting lower levels than MTurk samples (9.1%), χ2(1) = 54.57, p < 0.0001. Prevalence rates combined across item types (to increase sample size) were compared to Gidycz and Koss’s (1990) national data prevalence rate of 1.5%. Every comparison was higher than rates reported in Gidycz and Koss (see Table 3).

Study 2 – Perpetration Findings: Prevalence Rates, Validity, Utility, and Missingness.

Sample 1.

In an MTurk sample, using Item 4, “Have you ever acted as part of a group or two or more people to do…” 7.18% (n = 30) of the sample reported participating in multiple-perpetrator attacks or 27.03% of those with a perpetration history. Black individuals were more likely to report this behavior than non-Black individuals, χ2(1) = 5.123, p = 0.024. White individuals were less likely to report this behavior than non-White individuals, χ2 (1) =5.642, p = 0.018. The multiple-perpetrator item was significantly correlated with a composite item representing all other SES perpetration items, phi = .452, p < .001. This item identified three unique cases. There was no significant difference in rates of missing data; 3.9% of samples had missing data on the multiple-perpetrator item compared to 2.8% for the composite perpetration average.

In Sample 1, using Item 5, “…acted together with another person (or people) to overwhelm them…” 6.16% (n = 26) of the sample reported participating in multiple-perpetrator attacks; or 15.85% of those with a perpetration history. Black individuals (n = 47) were more likely to report this behavior than non-Black individuals, χ2(1) = 8.073, p = 0.004. White individuals were less likely to report this behavior than non-White individuals, χ2(1) =4.601, p = 0.032. Responses on the PRSPS-P multiple-perpetrator item were significantly correlated with endorsing any other perpetration items; phi = .278, p < .001. This item identified three unique, additional cases.

Seventeen participants (3.9% of the sample) had missing data on the multiple-perpetrator question compared to no missing data on the other PRSPS-P items, a significant difference, χ2(1) = 17.004, p < 0.0001.

Aggregate Data, Sample 4.

Using Item 4, “Have you ever acted as part of a group of two or more people…” in a college sample, 1.2% (n = 9) of the sample reported participating in multiple-perpetrator attacks, 7.69% of those with a perpetration history. Individuals who identified their race as “other” (n = 44) were more likely to report multiple-perpetrator tactics than those who did not, χ2 (1) = 4.722, p = .030. The multiple-perpetrator item was significantly correlated with a composite of other perpetration items, phi = .190, p < .001. This item identified two unique cases that the traditional items did not. There was no missing data to compare.

Sample 6.

Using item 6, “…acting as part of a group or two or more people who did these things after someone objected or was unable to give consent…” in a college sample, seven individuals (0.8%) reported participating in at least one multiple-perpetrator act. Given the low rates of endorsement, comparisons as a function of race, sexual identity, and gender were not conducted. Endorsement on a multi-perpetrator item was significantly correlated with endorsement on another SES item, phi = .160, p < .001. No unique cases were identified with the multiple-perpetrator items. We compared missing data on the multiple-perpetrator item with missing data on a single verbal coercion tactic (“telling, lies, threatening to end the relationship threatening to spread rumors…”. The missing data was still less on the multiple-perpetrator item (7.5%) than on the verbal coercion item (9.7%), p = .02.

Aggregate Data, Sample 7.

Using item 6, “….acting as part of a group of two or more people who did these things after someone objected or was unable to give consent…” in an MTurk sample, 120 individuals (18.9%) reported participating in at least one multiple-perpetrator act. Those assigned female at birth were more likely to report multiple-perpetrator incidents than those assigned male, χ2(1) = 3.84, p < 0.0001. White individuals were less likely to report this behavior than non-White individuals, χ2(1) = 4.450, p = 0.035. The multiple-perpetrator items were significantly correlated with a composite of other perpetration items, phi = .582, p < .001. The multiple-perpetrator items added 3 unique cases. There were no differences in rates of missingness on this item versus the other perpetration items, which were uniformly very low (<1%), χ2 (1) = .955, p = .955.

Comparison across items and sample types

There was a significant difference between sexual behavior-first (Item 6, 12.6%) and tactic-first items (Items 4, 5, 3.4%), χ2(1) = 73.19, p < .0001; sexual behavior-first items had a significantly higher prevalence than tactic-first items. Items based on the PRSPS (Item 5) did not differ in endorsement rates than SES-based items (Items 4, 6), χ2(1) = 0.368, p = .544. MTurk workers reported more multiple-perpetrator-based perpetration than college students, χ2(1) = 128.15, p < 0.0001. We were unable to compare to any national or large-scale data as this does not exist to our knowledge.

Discussion

This study aimed to examine whether MPSV is rare in college/community populations and examined several different strategies for assessing the history of multiple-perpetrator-involved sexual violence. Rates of multiple-perpetrator victimization ranged from 2.8 to 10.9% of the sample, and perpetration rates ranged from 0.8 to 18.9%, suggesting MPSV is not rare as suggested in prior research (Gidycz & Koss, 1990; Tillman 2011), even in college samples. Further, findings suggested that MPSV is more common than suggested by the one prior epidemiological study that measured and reported the prevalence of MPSV specifically (1.5%: Gidycz & Koss, 1990 vs. 4.86% on average for victimization and 6.43% for perpetration in this study). Our examination of several different items to measure MPSV suggested preliminary evidence of validity and acceptability for all items tested, including both victimization and perpetration items, highlighting several different ways future researchers can include MPSV in their studies.

Victimization Findings

Prevalence rates for multiple-perpetrator victimization were higher than much of the prior research, which has suggested a range of 1-2% whereas the overall average in this study was 4.86% with a range of 2.8 to 10.9% (Gidycz & Koss, 1990; Aim 1). Thus, multiple-perpetrator sexual victimization is not rare in college or community (i.e., MTurk) samples and warrants greater inclusion and attention than is currently given. Considering how to measure MPSV, one-stage approaches produced higher prevalence rates and phi correlations were numerically slightly stronger for the SES 2007 one-stage items (.311-.371) in comparison to the two-stage versions of these items (.182-.203). This is especially notable given that more participants were recruited for the two-stage studies. Another potential downside of the two-stage items is that multiple participants appeared to potentially misunderstand the item, as evidenced by their contrary reports in a second follow-up question, “ If so, how many people did these unwanted sexual things to you?” where six respondents answered 0 or 1. Alternatively, this may reflect sample differences with college students simply experiencing less multiple-perpetrator violence than MTurk workers who did not receive the two-stage items. Of course, it is also possible that participants misunderstood the one-stage question, but that was harder to evaluate. Validity correlations were actually slightly lower overall than hypothesized (H1) which may reflect the unique nature of MPSV; that is co-occurs less frequently than other forms of sexual violence. Alternatively, this may suggest there are even better ways to measure MPSV than those tested here. Finally, the items appeared to be at least somewhat acceptable with low missing data <5%.

Considering the item content itself, we cannot definitively say one item is better without more indepth research wherein participants describe incidents in detail. However, the item content and not just the measurement characteristics (one vs. two-stage) may be related to endorsement rates. The one-stage, sexual behavior-first victimization item, “Has someone ever acted as part of a group of two or more people in order to…[sex acts listed],” could encompass a very wide range of behavior, consistent with definitions of MPSV therefore resulting in higher prevalence rates. In contrast, the two-stage item, “On any particular occasion, did more than one person do these unwanted things to you” given after a sexual behavior-first questionnaire likely focused participants’ attention and recall to the sexual behaviors experienced, rather than the tactics. Prior research suggests sexual behavior-first items may unintentionally suppress accurate recall by directing attention to the large category of sexual activity – including consensual incidents – that then must be sorted internally. Further, it is possible that, because this item was given in the context of a sexual behavior-first questionnaire, participants also interpreted this question as being only asked about the perpetrators who did the sexual things they endorsed. Yet research suggests that the behaviors of perpetrators of MPSV range widely, and it could be that some perpetrators used the tactics described (e.g.,, overwhelmed you) whereas others may not have physically touched the participant but may have been compliant in the act. Therefore, not endorsing the item would be an accurate report while simultaneously inconsistent with the definition of MPSV that was used in this paper.

Perpetration Findings

Perpetration findings mirrored victimization findings in that multiple-perpetrator sexual perpetration was not rare, was more common than prior epidemiological research would suggest, and all of the items tested appeared to be valid and acceptable. Again, a large range of prevalence rates was estimated with 0.8 to 18.9% prevalence of multiple-perpetrator perpetration reported with an average of 6.43%. Even if the outlier value of 18.9% was excluded, prevalence rates were still higher than 1.5% as suggested by prior research, demonstrating that MPSV deserves greater attention in research on sexual violence with college and community samples.

Somewhat surprisingly given prior research, tactic-first items were not associated with higher prevalence rates than sexual behavior-first items. Indeed, sexual behavior-first items were on average twice as high (12.6 vs. 3.6%). It is unclear why the effect of tactic-first items was not evident for perpetration findings in this study. We are truly puzzled but do wish to note the very wide range of the sexual behavior-first prevalences - .8 and 18.9% compared to the tactic-first – 1.2, 6.2 and 7.2%. Perpetration findings also revealed a difference between college and MTurk samples in prevalence rates, with MTurk samples reporting higher prevalence rates and also higher ages. Validity correlations, as estimated by phi, ranged from .190 to .582. Missing data tended to be higher for perpetration items than victimization items; however, rates of missing data were still low, suggesting the reasonable acceptability overall.

Considering the item content, Item 4 used very broad language, “Acted as part of a group of two or more people…[sexual acts]” whereas Item 5 was a little less broad and included intent language, “…[Sexual acts]…acted together with another person (or people) to overwhelm them.” Yet, these items seemed to produce comparable prevalence rates. We suggest future research examine how participants interpret the “to overwhelm” language. Although MPSV is associated with a higher degree of violence, pre-planned intent is uncommon in other sexual violence incidents in this population, such that perpetrators appear to be more negligent and self-absorbed than intending to harm (Abbey & Jacques-Tiura, 2010). Item 6, which includes consent language, some might hypothesize would suppress rates by including the phrase “who did these things after someone objected or was unable to consent.” Prior research suggests that just doing the sex act or surprising someone with the act is a very common tactic (Peterson et al., 2024), which appears to be excluded by this language and may explain the very low prevalence rate reported by the Sample 6 college students. Further, some research suggests that using the term "consent" within the item could suppress responding by triggering social desirability because the behaviors are then interpreted to be more “harsh” (Strang & Peterson, 2017). Yet, this same item produced a very high prevalence rate in MTurk workers (18.9%) suggesting either these measurement features (surprise tactic, consent language) are less relevant in the context of MPSV, or there is something unique about MTurk workers.

Demographic differences

Surprisingly, we found very few gender differences in endorsement of multiple-perpetrator items nor any differences related to sexual orientation (majority identity vs. minority identity). Across the seven items tested, we only found gender differences in endorsement twice, which is somewhat surprising given well-established gender differences in sexual violence (Peterson et al., 2024). This may reflect the great heterogeneity of the incidents likely captured using these particular items in these samples. If we were able to focus in specifically on large group- or force-involved incidents, gender or sexual identity differences may have emerged.

Differences in endorsement related to race or ethnic identity were not found in every Item or sample but were repeatedly found across victimization and perpetration studies. Although the pattern is not particularly clear as to which racial identities or items across these studies, it does suggest overall that there are racial/ethnical differences in the experience of MPSV, which is consistent with other scholarship (Grewal, 2016). Although this study is unable to say why there are racial/ethnic differences or exactly what they look like, this finding may be related to greater odds of experiencing community or gang violence and/or different community norms for group behavior. MPSV may be related to greater social ties within these communities due to multiple forms of marginalization (Freng & Esbensen, 2007) or may reflect vulnerability associated with racism-related trauma (Bryant-Davis & Ocampo, 2005). It is also possible that race is actually functioning as a proxy for other demographic variables, such as economic status given that violence is more likely to affect the economically disadvantaged and much of the cited research comes from Westernized nations wherein people of color are marginalized (Grewal, 2016). We suggest that the higher rate of violence reported by MTurk participants likely reflects either age, as MTurk workers were on average 11-12 years older or socioeconomic marginalization. The trauma of experiencing sexual victimization sometimes decreases individuals’ economic activity (Byrne et al., 1999), which may drive individuals with a victimization history towards the type of remote-, ad-hoc employment of MTurk, which has the benefit of being flexible and home-based.

Future Research

Our data suggest that multiple-perpetrator incidents are more common than previously suspected (Gidycz & Koss, 1990; Tillman, 2001) suggesting a need for further research on MPSV. Better understanding the sequence of events during multiple-perpetrator incidents is important for refining the items tested here. The items tested here could capture a wide range of types of incidents – a good thing for being consistent with definition of MPSV and reflecting the true heterogeneity of MPSV – but a challenge in understanding exactly what occurred in any one incident. For example, the current items would not be able to clarify incidents in which a group was used to intimate the victim but different individuals within that group used different tactics, such as one being physically violent while another cheered them on. This type of incident may be related to seemingly inconsistent two-stage item data. It is unclear using two-stage items if participants felt that none of the other tactics described captured their experience (e.g., unique intimidation caused by the presence of multiple perpetrators) or whether the wording of the other items appeared to exclude the possibility of multiple perpetrators. However, even using a one-stage item, there are still many questions to be answered about the variety of tactics that may appear within a single incident of MPSV, how to capture that potential variability in a questionnaire, and the minimal elements necessary to identify MPSV cases.

Within our own data, very few unique new cases were identified by multiple-perpetrator items (i.e., utility), in contrast to studies such as Anderson et al., 2020, and Peterson et al., 2024 which showed how including other new content items was associated with high utility. Low utility suggests that participants almost always endorsed items in addition to the MPSV item. Thus, it is unclear whether multiple-perpetrator tactics tested here “acted together with another person” were used simultaneously with other tactics like physical force or verbal coercion or whether the endorsement of multiple items represents separate incidents. Given how common multiple victimization is both interpretations seem equally probable. Better understanding the sequence of behaviors during MPSV would illuminate this issue. Although the utility of the MPSV items was low, the violent and traumatic nature of these incidents warrants further attention. From a measurement perspective, the wide range of items tested all appeared to be a reasonably valid way to assess multiple-perpetrator incidents with validity correlations somewhat favoring a one-stage approach.

We also recommend that future research rely upon the nomological network of sexual violence and include clinical and attitudinal constructs. This type of convergent validity work is especially important because correlations with other victimization items are inherently limited; victimization incidents are not necessarily related. We also suggest that research examine the potentially unique factors that lead up to MPSV. The presence of multiple perpetrators implies a level of social acceptance and potential planning that may be unique.

Limitations

This study was a secondary data analysis and findings should be confirmed in primary data collection wherein random assignment to different items in the same sample could be conducted to clarify potential confounds identified in this paper. There is only one prior epidemiological dataset to compare our prevalence rates to and the study was conducted decades ago; time may also be an adequate explanation for the differences in prevalence rates between our data and Gidycz & Koss (1990). Further, there is no epidemiological perpetration data available, to our knowledge; thus one of our comparisons to Gidycz & Koss 1990 is flawed, comparing victimization data to perpetration data. There was no complete crossing of sample type by item type; the differences between one vs. two-stage, tactic-first, and sample type are confounded in this paper. Indeed, we did not test sexual behavior-first victimization items at all given that the SES is completely moving away from this model (Peterson et al., 2024). Further, some of the items tested would not be consistent with legal definitions (i.e., PRSPS-based items, two-stage items), which need to tie specific tactics directly to specific sexual behaviors and, in the case of MPSV, specific perpetrators. This represents a unique challenge in MPSV research in that, by the very nature of multiple perpetrators being involved, it becomes more complex to directly link behaviors with individuals.

Implications for Practice and Policy

For those working with individuals who have been harmed, our findings suggest multiple-perpetrator incidents may be more common than previously thought and providing this information to patients may help alleviate shame and self-stigma. For those working with perpetrators, assessing for multiple-perpetrator incidents may reveal specific circumstances or peer groups that are risk factors and could provide insights into risk and prevention planning. Given potential racial and ethnic differences in this form of violence, clinicians should seek to carefully understand the social ecology in which the event occurred. In terms of policy, the results of this study suggest that providing funding for basic scientific research is necessary to better understand the scope and forms of sexual violence.

Conclusion

This study found that MPSV is more common than suggested by prior research (1.5%) – on average 4.86 of college/community individuals reported MPSV victimization and 6.43% reported MPSV perpetration. We also found evidence of validity and acceptability for several different items that could be used to examine MPSV in future research. We recommend further research into MPSV in college and community samples and further research examining the optimal measurement of MPSV.

Acknowledgments

The authors thank Ellei Burmeister for sharing data from her undergraduate research project.

Funding

Dr. Anderson’s work was supported by the National Institute on Alcohol Abuse and Alcoholism (Grant No. 5K01AA026643). The content is solely the responsibility of the authors and does not necessarily reflect the official views of the funding agency. Data collection for Samples 1 and 2 was supported by a small grant from the Applied Psychology Center at Kent State University.

Footnotes

Conflict of Interest

The authors declare no conflicts of interest.

Research Ethics

Participants provided informed consent prior to accessing the study survey. Data were collected under the oversight of Institutional Review Boards at Kent State University, University of North Dakota, and University of Missouri–St. Louis. All research was conducted in accordance with the ethical standards of the American Psychological Association.

Data Availability

Due to the sensitive nature of the data and the lack of participant consent for public sharing at the time of collection, full datasets are available upon request for confirmation purposes. Data outputs can be accessed here: https://osf.io/fyqst/

References

  1. Abbey A, Helmers BR, Jilani Z, McDaniel MC, & Benbouriche M (2021). Assessment of men’s sexual aggression against women: An experimental comparison of three versions of the sexual experiences survey. Psychology of Violence, 11(3), 253–263. 10.1037/vio0000378 [DOI] [Google Scholar]
  2. Anderson RE, Cahill SP, & Delahanty DL (2017). Initial evidence for the reliability and validity of the Sexual Experiences Survey-Short Form Perpetration (SES-SFP) in college men. Journal of Aggression, Maltreatment and Trauma, 26(6), 626–643. 10.1080/10926771.2017.1330296 [DOI] [Google Scholar]
  3. Anderson RE, & Delahanty DL (2020). Discrepant Responding across Measures of College Students’ Sexual Victimization Experiences: Conceptual Replication and Extension. The Journal of Sex Research, 57(5), 585–596. 10.1080/00224499.2019.1669135 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Anderson RE, Carstens Namie EM, & Goodman EL (2021). Valid for Who? A Preliminary investigation of the validity of two sexual victimization questionnaires in men and sexual minorities. American Journal of Criminal Justice, 46(1), 168–185. 10.1007/s12103-020-09589-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Anderson RE, Goodman EL, & Ciampaglia AM (2021). An Initial Test of the Tactic-First and Item-Order Hypotheses: Accounting for Response Discrepancies in Sexual Victimization Questionnaires. American Journal of Criminal Justice, 46, 149–167. 10.1007/s12103-020-09584-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Anderson RE, Garcia M, & Delahanty DL (2021). Test–retest reliabilities of four tactic-first sexual violence history questionnaires. Psychology of Violence, 11(6), 580–590. 10.1037/vio0000384 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Anderson RAE, Silver KE, Ciampaglia AM, Vitale AM, & Delahanty DL (2021). The Frequency of sexual perpetration in college men: A Systematic review of reported prevalence rates from 2000 to 2017. Trauma, Violence, and Abuse, 152483801986061–152483801986061. 10.1177/1524838019860619 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Anderson RE, Carstens Namie EM, Michel PK, & Delahanty DL (2022). Study title-based framing effects on reports of sexual violence and associated factors in college students. Journal of Interpersonal Violence, 37(17–18), NP15359–NP15383. https://journals.sagepub.com/doi/full/10.1177/08862605211016349 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Anderson RE, Doctor HN, & Piggott DM (2023). Measuring sexual violence perpetration acknowledgment: Testing the effects of label and response format. Aggressive Behavior, 49(5), 499–508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Anderson RE, Goodman EL, & Carstens Namie EM (2023). Participant acceptability of questionnaires impacts sexual victimization prevalence rates. Journal of Child Sexual Abuse, 32(6), 771–789. 10.1080/10538712.2023.2240778 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Bamford J, Chou S, & Browne KD (2016). A systematic review and meta-analysis of the characteristics of multiple perpetrator sexual offences. Aggression and Violent Behavior, 28, 82–94. 10.1016/j.avb.2016.04.001 [DOI] [Google Scholar]
  12. Basile KC, Smith S, Breiding MJ, Black MC, & Mahendra R (2014). Sexual Violence Surveillance: Uniform Definitions and Recommended Data Elements, Version 2.0. https://www.cdc.gov/violenceprevention/pdf/sv_surveillance_definitionsl-2009-a.pdf [Google Scholar]
  13. Basile KC, Smith SG, Kresnow M, Khatiwada S, & Leemis RW (2022). The National Intimate Partner and Sexual Violence Survey: 2016/2017 Report on Sexual Violence. Atlanta, GA: National Center for Injury Prevention and Control, Centers for Disease Control and Prevention. [Google Scholar]
  14. Bryant-Davis T, & Ocampo C (2005). Racist incident-based trauma. The Counseling Psychologist, 33(4), 479–500. 10.1177/0011000005276465 [DOI] [Google Scholar]
  15. Buday SK, & Peterson ZD (2015). Men’s and Women's Interpretation and Endorsement of Items Measuring Self-Reported Heterosexual Aggression. Journal of Sex Research, 52(9), 1042–1053. 10.1080/00224499.2014.967373 [DOI] [PubMed] [Google Scholar]
  16. Busch-Armendariz NB, Olaya-Rodriguez D, Kammer-Kerwick M, Wachter K, Sulley C, Anderson K, & Huslage M (2015). Health and Well-Being: Texas Statewide Sexual Assault Prevalence Study. 10.15781/T2Z92M [DOI] [Google Scholar]
  17. Byrne CA, Resnick HS, Kilpatrick DG, Best CL, & Saunders BE (1999). The socioeconomic impact of interpersonal violence on women. Journal of Consulting and Clinical Psychology, 67(3), 362–366. 10.1037/0022-006X.67.3.362 [DOI] [PubMed] [Google Scholar]
  18. Cook SL, Gidycz CA, Koss MP, & Murphy M (2011). Emerging issues in the measurement of rape victimization. Violence Against Women, 17(2), 2012–2018. 10.1177/1077801210397741 [DOI] [PubMed] [Google Scholar]
  19. da Silva T, Woodhams J, & Harkins L (2015). Multiple perpetrator rape: A critical review of existing explanatory theories. Aggression and Violent Behavior, 25, 150–158. 10.1016/j.avb.2015.07.017 [DOI] [Google Scholar]
  20. de Heer Brooke, & Jones Lynn. (2017). Measuring sexual violence on campus: climate surveys and vulnerable groups. Journal of School Violence, 16(2), 207–221. [Google Scholar]
  21. Dworkin ER, Menon SV, Bystrynski J, & Allen NE (2017). Sexual assault victimization and psychopathology: A review and meta-analysis. Clinical Psychology Review, 56, 65–81. 10.1016/j.cpr.2017.06.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Fedina L, Holmes JL, & Backes BL (2018). Campus sexual assault: A Systematic review of prevalence research from 2000 to 2015. Trauma, Violence, & Abuse, 19(1), 76–93. 10.1177/1524838016631129 [DOI] [PubMed] [Google Scholar]
  23. Fisher BS (2009). The Effects of survey question wording on rape estimates: Evidence from a quasi-experimental design. Violence Against Women, 15(2), 133147. 10.1177/1077801208329391 [DOI] [PubMed] [Google Scholar]
  24. Freng A, & Esbensen F (2007). Race and gang affiliation: An examination of multiple marginality. Justice Quarterly, 24(4), 600–628. doi: 10.1080/07418820701717136 [DOI] [Google Scholar]
  25. Fongwa MN, Setodji CM, Paz SH, Morales LS, Steers WN, & Hays RD (2010). Readability and missing data rates in CAHPS 2.0 Medicare Survey in African American and White Medicare respondents. Health Outcomes Research in Medicine, 1(1), e39–e49. 10.1016/j.ehrm.2010.03.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Gesser N, Eby FG, & Anderson RE (2024) Motives for Sex and Sexual Perpetration in College Men: An Exploratory Study. Sex Abuse, 36(4)-486–506. doi: 10.1177/10790632231190080 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Gidycz CA, & Koss MP (1990). A comparison of group and individual sexual assault victims. Psychology of Women Quarterly, 14(3), 325–342. 10.1111/j.1471-6402.1990.tb00023.x [DOI] [Google Scholar]
  28. Grewal KK (2016). Racialised Gang Rape and the Reinforcement of Dominant Order: Discourses of Gender, Race and Nation (1st ed.). Routledge. 10.4324/9781315580548 [DOI] [Google Scholar]
  29. Hamby SL, & Koss MP (2003). Shades of gray: A Qualitative study of terms used in the measurement of sexual victimization. Psychology of Women Quarterly, 27(3), 243–255. 10.1111/1471-6402.00104 [DOI] [Google Scholar]
  30. Harris C, Ullman SE, Shepp V, & O’Callaghan E (2020). Multiple-perpetrator sexual assault: Correlates of PTSD and depressive symptoms in a sample of adult women. Journal of Sexual Aggression. 10.1080/13552600.2020.1823496 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Horvath MAH, & Kelly L (2009). Multiple perpetrator rape: Naming an offence and initial research findings. Journal of Sexual Aggression, 15(1), 83–96. [Google Scholar]
  32. Horvath M, & Woodhams J (Eds.). (2013). Handbook on the Study of Multiple Perpetrator Rape: A multidisciplinary response to an international problem. (1st ed.). Routledge. 10.4324/9780203083406 [DOI] [Google Scholar]
  33. Jewkes R, Sikweyiya Y, Dunkle K, & Morrel R (2015). Relationship between single and multiple perpetrator rape perpetration in South Africa: A comparison of risk factors in a population-based sample. BMC Public Health, 15, 616. 10.1186/s12889-015-1889-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Johnson SM, Murphy MJ, & Gidycz CA (2017). Reliability and validity of the Sexual Experiences Survey – Short Forms victimization and perpetration. Violence and Victims, 32(1), 78–92. 10.1891/0886-6708.VV-D-15-00110 [DOI] [PubMed] [Google Scholar]
  35. Khan R, Brewer G, Kim S, & Centifanti LCM (2017). Students, sex, and psychopathy: Borderline and psychopathy personality traits are differently related to women and men’s use of sexual coercion, partner poaching, and promiscuity. Personality and Individual Differences, 107, 72–77. 10.1016/j.paid.2016.11.027 [DOI] [Google Scholar]
  36. Koss MP, & Gidycz CA (1985). Sexual experiences survey: Reliability and validity. Journal of Consulting and Clinical Psychology, 53(3), 422–423. doi: 10.1037/0022-006x.53.3.422 [DOI] [PubMed] [Google Scholar]
  37. Krebs C, Lindquist C, Planty M, Langton L, Berzofsky ME, Asefnia N, Griggs A, Shook-Sa B, & Enders K (2017). Sensitivity of sexual victimization estimates to definitional and measurement decisions. Journal of Interpersonal Violence, 088626051774419–088626051774419. 10.1177/0886260517744190 [DOI] [PubMed] [Google Scholar]
  38. Langton L, Planty M, & Lynch JP (2017). Second major redesign of the National Crime Victimization Survey (NCVS). Criminology & Public Policy, 16(4), 1049–1074. 10.1111/1745-9133.12335 [DOI] [Google Scholar]
  39. McCauley HL, Campbell R, Buchanan NCT, & Moylan CA (2019). Advancing theory, methods, and dissemination in sexual violence research to build a more equitable future: An Intersectional, community-engaged approach. Violence Against Women, 25(16), 1906–1931. 10.1177/1077801219875823 [DOI] [PubMed] [Google Scholar]
  40. Mickelson KM (2020) .Maladjustment correlates associated with sexual perpetration tactics [Master’s thesis, University of North Dakota; ]. Theses and Dissertations. 3286. https://commons.und.edu/theses/3286/ [Google Scholar]
  41. Muehlenhard CL, Peterson ZD, Humphreys TP, & Jozkowski KN (2017). Evaluating the one-in-five statistic: Women’s risk of sexual assault while in college. Journal of Sex Research, 54, 4–5, pp. 549–576. 10.1080/00224499.2017.1295014 [DOI] [PubMed] [Google Scholar]
  42. Peterson ZD, Koss MP, Anderson RE (2024). Preliminary prevalence estimates of sexual exploitation as measured by the Sexual Experiences Survey-Victimization (SES-V) in a national U.S. sample. Journal of Sex Research, 61(6), 904–921. doi: 10.1080/00224499.2024.2359639 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Porter LE, & Alison LJ (2001). A Partially ordered scale of influence in violent group behavior: An Example from gang rape. Small Group Research, 32(4), 475–497. 10.1177/104649640103200405 [DOI] [Google Scholar]
  44. Raj A, Molnar J, Surya S, DiGirolamo S, Christian CW, Lavelle JM, & Wood JN (2022). Characterizing multiple-perpetrator sexual assaults in the adolescent female population. Journal of Pediatric and Adolescent Gynecology, 35(6), 659–661. 10.1016/j.jpag.2022.06.003 [DOI] [PubMed] [Google Scholar]
  45. Rosay AB (2016). Violence against American Indian and Alaska Native Women and Men: 2010 Findings from the National Intimate Partner and Sexual Violence Survey Item Type Report. U.S. Department of Justice, National Institute of Justice. http://hdl.handle.net/11122/7025 [Google Scholar]
  46. Rueff WT, & Gross AM (2017). Assessing sexual coercion: Survey wording differences and the victimization-perpetration discrepancy. Journal of Family Violence, 32(3), 325–331. 10.1007/s10896-016-9859-2 [DOI] [Google Scholar]
  47. Sanday PR (1992). Fraternity gang rape: Sex, brotherhood, and privilege on campus. NYU Press. [Google Scholar]
  48. Sekhon M, Cartwright M, & Francis JJ (2017). Acceptability of healthcare interventions: An overview of reviews and development of a theoretical framework. BMC Health Services Research, 17(1), 88–88. 10.1186/s12913-017-2031-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Sela-Shayovitz R. (2015). “They Are All Good Boys”: The role of the Israeli media in the social construction of gang rape. Feminist Media Studies, 15(3), 411–428. 10.1080/14680777.2014.993675 [DOI] [Google Scholar]
  50. Stemple L, Flores A, & Meyer IH (2017). Sexual victimization perpetrated by women: Federal data reveal surprising prevalence. Aggression and Violent Behavior, 34, 302–311. 10.1016/j.avb.2016.09.007 [DOI] [Google Scholar]
  51. Strang E, & Peterson ZD (2017). Unintentional misreporting on self-report measures of sexually aggressive behavior: An Interview study. The Journal of Sex Research, 54(8), 1–13. 10.1080/00224499.2017.1304519 [DOI] [PubMed] [Google Scholar]
  52. Straus MA, Hamby SL, Boney-McCoy S, & Sugarman DB (1996). The Revised Conflict Tactics Scales (CTS2): Development and preliminary psychometric data. Journal of Family Issues, 17(3), 283–316. 10.1177/019251396017003001 [DOI] [Google Scholar]
  53. Struckman-Johnson C, Anderson PB, Struckman‐Johnson D, & Smeaton G (2019). The Post-Refusal Sexual Persistence Scale. In Milhausen R, Sakuluk JK, Fisher TD, Davis CM, & Yarber W (Eds.), Handbook of Sexuality-Related Measures (4th ed.). Routledge. [Google Scholar]
  54. Tillman S. (2011). Multiple-perpetrator sexual assault: Risk factors, effects, and help-seeking. In Bryant-Davis T (Ed.), Surviving sexual violence: A guide to recovery and empowerment (pp. 68–76). Rowan & Littlefield. [Google Scholar]
  55. Trottier D, Nolet K, Benbouriche M, Bonneville V, Racine-Latulippe F, & Bergeron S (2021). Sexual violence perpetration and victimization: Providing prevalence rates for understudied populations. Violence and Gender, 10.1089/vio.2020.0037 [DOI] [Google Scholar]
  56. Ullman SE (2007). Comparing gang and individual rapes in a community sample of urban women. Violence and Victims, 22(1). 10.1891/vv-v22i1a003 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Due to the sensitive nature of the data and the lack of participant consent for public sharing at the time of collection, full datasets are available upon request for confirmation purposes. Data outputs can be accessed here: https://osf.io/fyqst/

RESOURCES