BACKGROUND
National survey findings indicate that at least 20% of American women are victimized by sexual violence in their lifetime (Kilpatrick, 1992; Tjaden & Thoennes, 2000). Prevalence rates appear to be even higher among college students (Krebs et al., 2009). Sexual aggression appears to be committed disproportionately (~80%) by men (Hequembourg et al., 2013). Roughly 25% to 60% of men acknowledge sexual acts in survey research that constitutes coercion or rape (Abbey et al., 2011; Parkhill & Abbey, 2008; Russell & King, 2016, 2020; Widman et al., 2013). This analysis focused on self-reported acts of oral, anal, or vaginal physical contact perpetrated against a nonconsenting victim. These acts constitute rape as defined by the U.S. Department of Justice (2012) and within the research literature (Koss et al., 2007). The criterion behavior in this analysis was measured by the Sexual Experiences Survey – Long Form Perpetration (Koss et al., 2007) and referred to as “completed rape” (CRAPE).
The confluence mediational model (Malamuth et al., 1995; Malamuth, 2003) identified hostile masculinity and impersonal sexuality as central risk factors in the development of sexual aggressiveness. Subsequent research expanded the model to include contributors such as adolescent delinquency, alcohol consumption, psychopathy, childhood maltreatment, the misperception of sexual intent, and other factors (Abbey et al., 2011; Jacques-Tiura et al., 2007; Malamuth, 2003; Parkhill & Abbey, 2008; Russell & King, 2016, 2020).
The potential role of the “dark tetrad” personality dimensions of narcissism, psychopathy, sadism, and Machiavellianism has been examined in a range of sexual aggression studies (Eman et al., 2022; Jones & Figueredo, 2013; Jones & Olderbak, 2014; Perenc, 2022; Pilch et al., 2020; Zeigler-Hill et al., 2016). While personality disorder diagnoses pose obvious maladjustment risks, the proposed DSM-5 hybrid model (American Psychiatric Association, 2013) encouraged greater attention to the constituent dimensional facets that underlie these diagnostic conditions (Huczewska et al., 2019; Krueger & Markon, 2014). The Personality Inventory for the DSM-5 (PID-5; Krueger et al., 2012) provided measures of the 5 trait domains and 25 constituent facet dimensions that mirror those advanced in the DSM-5 for further study.
PID-5 trait scores are often elevated among men who have perpetrated either physical or sexual violence. One analysis found higher scores on all 25 facet dimensions for violent prisoners compared to non-incarcerated controls (Adhiatma & Halim, 2016). A study of sex offender inmates found that men with high callousness, irresponsibility, intimacy avoidance, distractibility, and restricted affectivity were more violent then counterparts without those attributes (Somma et al., 2021). Men exhibiting polytactic sexual aggression (Norton-Baker et al., 2017) were found to score significantly higher than non-violent controls on all but five facet dimensions. The most substantive differences were found for callousness, deceitfulness, cognitive/perceptual dysregulation, grandiosity, and suspiciousness. Men in a community sample who described sexually aggressive ideation scored higher than normative counterparts on the PID-5 callousness, grandiosity, deceitfulness and cognitive/perceptual dysregulation scales (Kasowski & Anderson, 2020). This latter trait dimension was described in the DSM-5 as odd and unusual mentation including dissociation and altered states of consciousness. Errant information processes could predispose to the misinterpretation of social cues and acts of sexual aggression. The confluence model also posits that misperceived sexual intentions sometime elevate the risk of sexual aggression (Jacques-Tiura et al., 2007). Russell and King (2020) suggested that smiling or even eye contact could be misconstrued as flirtation by men who exhibit cognitive/perceptional dysregulation.
This study identified PID-5 items, facets, and domains associated with self-reported acts of rape. A specialty sexual aggression risk indicator was constructed and contrasted with PID-5 facet scores and the count of trait elevations (> 1 SD) on the inventory.
PARTICIPANTS AND PROCEDURE
PROCEDURE
Data collection relied on Amazon’s Mechanical Turk (MTurk) crowdsourcing research platform (Buhrmester et al., 2011; Clifford et al., 2015). This project was approved by the university institutional review board (University of North Dakota, approval no. IRB-202001-163) with informed consent required for all survey respondents.
PARTICIPANTS
The sample was restricted to men who completed the survey in the United States for financial compensation ($0.50). Online proxy/VPN detection software (https://iphub.info) was used to exclude users from international or dubious internet locations. Inclusion criteria required completion of at least 75% of the survey items.
Exclusions (n = 40) were made for respondents who scored higher than 17.5 on the PID-5 Response Inconsistency Scale (INC; Bagby & Sellbom, 2018; Keeley et al., 2016). Respondents affirming more than 20 of the 33 CRAPE items were outliers (top distribution 10%) and were also excluded from analysis (n = 48). All but one respondent in this subset endorsed more than five CRAPE items in a row with only one retained respondent showing the same repetitive pattern. Cuts (n = 60) were made as well for respondents who failed either a palindrome attention check or motivation standard (“Now that you have completed this survey, will you provide a final summary regarding your honesty and general attentiveness in responding?”). The exclusion rate was 21.1%.
The final sample of men (N = 517) varied in age (M = 39.12, SD = 11.78, range = 19-78) and ethnicity (White – 81.4%, Hispanic – 5.2%, Black – 6.8%, Asian – 3.5%; biracial or other – 3.1%). White respondents were overrepresented in this sample (United States Census Bureau, 2023: White – 58.9%, Hispanic – 19.1%, Black – 13.6%, Asian – 6.3%, biracial or other – 2.1%). Respondents did not differ significantly in ethnicity from those excluded from the final analysis [INC criterion, χ2(7) = 6.00, p = .540; CRAPE criterion, χ2(7) = 5.32, p = .621; palindrome/motivation criterion, χ2(7) = 2.40, p = .934].
MEASURES
The Personality Inventory for DSM-5 (PID-5). The PID-5 (Krueger et al., 2012; Krueger & Markon, 2014) is a 220-item inventory that relies on a four-point metric to generate dimensional scores on 5 trait domains and 25 personality facets. Facet scores require that 75% of facet constituent items are completed by the respondent. Random responding has been shown to elevate PID-5 facet and domain scores (Bagby & Sellbom, 2018), but a response inconsistency scale using 20 highly correlated pairs of near identical item content has been shown to minimize random responding as a source of error (INC; Keeley et al., 2016). Internal consistency and concurrent validity support for the PID-5 trait dimensions have come from many sources (Al-Dajani et al., 2016; Fossati et al., 2013; Norton-Baker et al., 2017; Russell & King, 2020). Minimal changes (Mdn d = .12) were found in PID-5 trait scores across one 17-month follow-up (Wright et al., 2015).
Sexual Experiences Survey – Long Form Perpetration (SES-LFP). The SES-LFP (Koss et al., 2007) asked respondents about attempted and completed sexual acts without victim consent that included noncontact intrusions (voyeurism, exposure, harassment, public masturbation, sexting, etc.), coercion (nonphysical threats and pressure), frotteurism (unwanted physical contact), and rape (oral, anal, or vaginal penetration). The focus of this analysis was the 33 acts of rape that were sampled using a four-point frequency metric (0, 1, 2, 3+). Items were then dichotomized to provide a count of the number of different acts that occurred at least once since age 14.
Self-reported SES rape disclosures by men in the general population have been common (43%, Abbey et al., 2011; 59%, Widman et al., 2013). Exclusion criteria have not been established to screen SES-LFP protocols, and sexual aggression researchers have relied on idiosyncratic methods to identify dubious response patterns. SES-LFP subscale internal consistencies (α ~ .80) and test-retest reliabilities (r =. 93) have been established (Cecil & Matson, 2006; Norton-Baker et al., 2017; Koss & Gidycz, 1985; Sisco & Koss, 2006).
Support for the validity of sexual aggression self-reports continues to emerge in the literature (Anderson et al., 2017; Johnson et al., 2017; Yapp & Quayle, 2018).
ANALYTIC LOGIC
Reliance on sequential dimensional and binary versions of predictor or criterion variables has been supported for markedly skewed distributions (DeCoster et al., 2011; Farrington & Loeber, 2000). Any single self-reported act of forced oral, anal, or vaginal physical contact constitutes a consequential outcome in and of itself. A count was generated of the 33 SES-LFP items that constituted acts of completed rape (CRAPE). Standardized rape frequency scores (CRAPEz) were then generated in addition to a categorical (CRAPEg) indicator defined by CRAPE scores above or equal to 0.
PID-5 facet scores were analyzed as well using both standardized dimensional (TRAITz) and categorical (TRAITg) score predictors. Reliance on a standard deviation cutoff has been recommended for the interpretation of the PID-5 elevations (Samuel et al., 2013). Categorical trait scores were differentiated on the basis of raw scores above +1 SD (~ 85th percentile).
Linear and logistic regression analyses were used to identify trait dimensions that accounted for unshared variance in CRAPEz and CRAPEg scores. The count of PID-5 trait elevations (> 1 SD) was also examined as an additional risk indicator. A collateral analysis was conducted to identify PID-5 items (Sexual Aggression Risk Index) that accounted for unshared variance in CRAPEg scores.
RESULTS
The prevalence of CRAPE scores above zero in this sample was 9.3%. A subset of 3.1% of men responded affirmatively to an even more direct question about past sexual violations of nonconsenting partner(s). “Do you believe you may have ever raped someone?”
Descriptive and correlation statistics are provided in the Supplementary materials. Most of the 25 PID-5 trait dimensions were correlated significantly with CRAPEz (92%) and CRAPEg (96%) scores. The majority (n = 17) of these facet dimensions were associated with both the CRAPEz and CRAPEg indicators. While the dimensional coefficients were modest in size (Mr = .16, range = .06 to .27), the odds ratios for CRAPEg scores were often sizable (MOR = 2.90, range = 1.11 to 6.92).
None of the PID-5 trait dimensions accounted for unshared variance in dimensional CRAPEz scores. Elevations (> 1 SD) on four PID-5 scales (intimacy avoidance, cognitive/perceptual dysregulation, attention seeking, and deceitfulness) were associated with the CRAPEg classification in the logistic regression model (Table 1). The model was significant, Wald (1) = 206.59, p < .001 (Nagelkerke R2 = .27; Cox & Snell R2 = .13). Goodness of fit concerns were not identified (Hosmer & Lemeshow, χ2(8) = 6.26, p = .618).
Table 1
Logistic regression to identify PID-5 risk indicators for categorical CRAPEg scores
[i] Note. PID-5 – Personality Inventory for the DSM-5. The criterion measure in this model was CRAPEg. Trait predictors were dichoto mized (> 1 SD versus remaining sample). Respondent age analyzed as a dimensional variable. Exp(B) reversed for negative coefficients to enhance interpretability. N = 472.
The count of PID-5 trait elevations (> 1 SD) was tested as a risk indicator in the sample (Table 2). Trait elevation counts were correlated with both CRAPEz (r = .20, p < .001) and CRAPEg (r = .21, p < .001) scores. The odds of a respondent acknowledging one or more acts of completed rape increased with the trait elevation counts in a graded fashion.
Table 2
PID-5 trait elevation count and SES-LFP rape history (CRAPEg)
| Elevation count | n | Sensitivity | r | OR |
|---|---|---|---|---|
| > 0 | 315 | 11.7% | .11 | 2.31* |
| > 1 | 242 | 14.0% | .15 | 3.05*** |
| > 2 | 181 | 17.1% | .20 | 3.88*** |
| > 3 | 140 | 18.6% | .20 | 3.68*** |
| > 4 | 91 | 24.2% | .24 | 4.91*** |
| > 5 | 70 | 24.3% | .21 | 4.30*** |
[i] Note. PID-5 – Personality Inventory for the DSM-5; SES-LFP – Sexual Experiences Survey-Long Form Perpetration. These 11 PID-5 items each accounted for unshared variance in the CRAPEg classification. Criterion group in each analysis defined by CRAPE scores over 0. The CRAPE base rate was 5.4% when the trait elevation count was zero (n = 203). Sensitivity – percent of rape cohort detected by the elevation count cutoff. *p < .05, ***p < .001.
None of the PID-5 domains accounted for variance in CRAPEz that exceeded their facet constituents (antagonism, r = .13, p = .003; negative affect, r = .10, p = .027; detachment, r = .13, p = .003; disinhibition, r = .22, p < .001; psychoticism, r = .20, p < .001).
SEXUAL AGGRESSION RISK INDEX (SARI)
A subset of PID-5 items (147/220, 66.8%) were correlated (p < .001) with the CRAPEg classification. Eleven items accounted for unshared variance in a logistic regression of CRAPEg scores. These items were derived from the callousness (19. “I really don’t care if I make people suffer”; 198. “I sometimes hit people to remind them who’s in charge”; 153. “I don’t care if my actions hurt others”), risk taking (3. “People would describe me as reckless”; 39. “I don’t think about getting hurt when doing things that might be dangerous”), attention seeking (191. “I crave attention”), distractibility (132. “I am easily distracted”), unusual beliefs/experiences (99. “I sometimes have heard things that others couldn’t hear”), and cognitive/perceptual dysregulation (83. “I often can’t control what I think about”; 36. “I have trouble telling the difference between dreams and waking life”; 77. “Sometimes when I look at a familiar object, it’s somehow like I’m seeing it for the first time”) dimensions. The SARI index was internally consistent (α = .86) and correlated with both CRAPEz (r = .23, p < .001) and CRAPEg (r = .27, p < .001) scores. Elevated (> 1 SD) SARI scores sharply raised the odds of self-reported rape (OR = 8.25, p < .001). Over 32% of the men with SARI scores above +1.0 reported sexual assault histories.
DISCUSSION
The primary question of this analysis was the extent to which PID-5 item/facet/domain scores would account for variance in CRAPE scores. Over two-thirds of the PID-5 item and facet scores were significantly associated with self-reported acts of rape. The number of PID-5 facet elevations (> 1 SD) also provided a link with odds ratios that rose in a graded fashion. The 11-item Sexual Aggression Risk Index (SARI) was found to approximate or exceed the count of PID-5 trait elevations as a risk indicator, and a number of individual facets were closely associated with the criterion measures.
Sexual aggressiveness may be portended by a range of personality traits. Callousness has been linked to sexual aggression in multiple analyses (Norton-Baker et al., 2017; Somma et al., 2021). Deceitfulness has been found to be higher in men with sexually aggressive ideation (Kasowski & Anderson, 2020) and polytactic sexual pursuit strategies (Norton-Baker et al., 2017). Cognitive/perceptual dysregulation was identified previously as a risk factor for hostile masculinity and the possible misinterpretation of interpersonal cues among male sexual aggressors (Russell & King, 2020).
Data from this analysis offer more specific links between specific PID-5 items and traits with self-reported acts of rape. These attributes extended beyond the dark tetrad (Eman et al., 2022; Jones & Figueredo, 2013; Jones & Olderbak, 2014; Perenc, 2022; Pilch et al., 2020; Zeigler-Hill et al., 2016). The PID-5 provides measures of the constituent attributes believed to elevate the developmental trajectories of personality disturbance. For example, antisocial personality disorder in the proposed hybrid model (American Psychiatric Association, 2013) is conceptualized as an amalgam of manipulativeness, callousness, deceitfulness, hostility, risk taking, impulsivity, and irresponsibility. These associations are generally consistent with the confluence model (Malamuth et al., 1995; Malamuth, 2003), with added attention to peripheral attributes such as cognitive/perceptual dysregulation, deceitfulness, attention seeking, and intimacy avoidance that may account for unshared variance in sexual aggression.
Sexual aggression has posed investigative challenges since the behavior constitutes illegal acts potentially subject to criminal sanction. The SES perpetrator forms have provided an imperfect but valuable method of estimating the extent to which this behavior occurs in the general public. Base rates as high as 60% have been self-reported by men for selected nonconsensual acts of coercion and/or rape (Abbey et al., 2011; Parkhill & Abbey, 2008; Russell & King, 2016, 2020; Widman et al., 2013). Why would male respondents so casually disclose acts of oral, anal, or vaginal penetration perpetrated on nonconsenting victims? Are they accurately reading the survey items? Do they realize or care that they may be characterized as sexual assailants and rapists? The answer is likely to differ among the respondents, and analytic advances will be needed to differentiate between honest, inattentive, contemptuous, oblivious, and other mindsets that contribute to the high base rates in the SES-LFP literature. Exclusions were made in this analysis to minimize random and inattentive responding on both the personality and sexual assault inventories. Even with quality control efforts, interpretive challenges remain for individual protocols examined in isolation. While respondent motives may differ for any survey protocol, the fact remains that men with personality disorder trait elevations appear far more likely to acknowledge past behavior that constitutes sexual assault.
The PID-5 may provide an exemplar inventory designed around the emerging hybrid trait dimensions delineated in the DSM-5 (American Psychiatric Association, 2013). Practitioners have relied historically on personality profiles to identify clients who show elevated, but unclearly specified, risks of general psychological maladjustment. The availability of these more precise trait indicators may prompt practitioners to conduct more specialized analyses of the sexual histories of selected high-risk clients. Additional research would be needed to derive more precise guidelines regarding the threshold scores that warrant greatest concern.
LIMITATIONS AND FUTURE DIRECTIONS
These self-report data were retrospective and unverified. SES-LFP frequencies could be inflated since multiple target acts could occur in a single sexual encounter. Differentiations were also not made on the basis of the severity or consequence of the sexual act. The dates and timing of the self-reported acts of rape were not documented. Causality and directionality cannot be established from cross-sectional findings, and inferences may not generalize fully to population subsets that differ from this sample as composition. Future studies should continue to examine personality traits as central components in the developmental trajectory of sexual aggression.
Supplementary materials are available on the journal’s website.
