1. Psychopathy as a construct and its structural model Psychopathy is a personality construct associated with dysfunction in several domains, including deficient perspective-taking (Brown et al., 2015; Decety et al., 2013; Meffert et al., 2013), aggression (e.g., Walters, 2003), substance overuse (Gustavson et al., 2007; Lynam et al., 2013), sexual offending (e.g., Knight & Guay, 2006), and criminal recidivism (Collison et al., 2016). Although psychopathy is not a personality disorder diagnosis included in the Diagnostic and Statistical Manual of Mental Disorders, it is a specifier for antisocial personality disorder in the Section III model and is arguably the first personality disorder conceptualized (American Psychiatric Association, 2013; Saleh et al., 2010). Psychopathy was initially conceived at the turn of the 19th century when Pinel (1806) identified a particular phenotype of patients, which he labeled with an affliction called manie sans de ́lire, who cognitively seemed to be entirely sane and cognizant of the negative impact of their behaviors, and yet continued to behave with moral disregard (Pinel, 1806). A century later, Maudsley wrote about a similar phenotype of people “destitute of moral sensibilities” and “with vicious inclinations” (Maudsley, 1895). Cleckley (1941), regarded as the forefather of more modern conceptualizations of psychopathy due to his seminal work, The Mask of Sanity, described the prominent features of ‘‘psychopathic personality’’ and its tendency to appear “normal” due to the “mask of sanity” presented by people with this personality (Cleckley, 1941). Building upon Cleckley’s theories, Hare (2003) has described people with psychopathy as interpersonally grandiose, egocentric, charming, manipulative, dominant, forceful, exploitative, callous, and cold-hearted; affectively shallow and labile, unable to form long-lasting personal bonds, devoid of principles or goals, and lacking empathy and genuine guilt or remorse; upholding an impulsive lifestyle, unstable, sensation-seeking; and readily violating social norms and failing to fulfill social obligations and responsibilities, both explicit and implied (Hare, 2003; Hare & Neumann, 2008; Lynam et al., 2011). The development of the Psychopathy Checklist and its revision (PCL, PCL-R; Hare, 1980, 1991, 2003) afforded a useful assessment that enabled research on the construct (Hare & Neumann, 2008). Stemming from a need to measure psychopathy outside of forensic settings, several self-report assessments have been developed, such as the Triarchic Psychopathy Measure (TriPM; Patrick et al., 2009), the Psychopathic Personality Inventory (PPI, Lilienfeld & Andrews, 1996; PPI-R, Lilienfeld & Widows, 2005), the Levenson Self-Report Psychopathy Scale (LSRP; Levenson et al., 1995), Self-Report Psychopathy Scale (SRP; Paulhus et al., 2009), and the Elemental Psychopathy Assessment (EPA; Lynam et al., 2011). These measures are multidimensional and assess highly similar dimensions. For example, empirically using basic trait dimensions from the Five-Factor Model of personality, the boldness, meanness, and disinhibition scales from the TriPM can be understood as low neuroticism and high extraversion, low agreeableness, and low conscientiousness, respectively (e.g., Collison et al., in press; Sleep et al., 2019; Miller et al., 2016; Hyatt et al., 2020). Subscales from other measures, including SRP Erratic Lifestyle, SRP Antisocial Behavior, PPI Self-Centered Impulsivity, and LSRP Factor 2, comprise low conscientiousness and agreeableness and high externalizing-based neuroticism (e.g., angry hostility; Lynam & Miller, 2015). Low agreeableness and conscientiousness also make up SRP Callous Affect, LSRP Factor 1, and SRP Interpersonal Manipulation (Lynam & Miller, 2015). In the present paper, we refer to these similar dimensions across psychopathy measures as boldness, meanness, and disinhibition, which make up the triarchic model of psychopathy (Patrick et al., 2009). Meanness in this model conveys lack of empathy and callousness. Disinhibition encompasses antisocial behavior and impulsivity. Boldness, comparable to Fearless Dominance in the PPI, represents traits involving assertiveness, social dominance, and emotional resilience. Researchers have generated scoring capabilities for alternative measures that render scores for the three domains as well, including for the Multidimensional Personality Questionnaire (MPQ-Tri; Brislin et al., 2015), the PPI-R (Hall et al., 2014), the Minnesota Multiphasic Personality Inventory–2–Restructured Form (Triarchic Psychopathy Scales; MMPI-2-RF-Tri; Sellbom et al., 2016) and the HEXACO Personality Inventory-Revised (HEXACO-Tri; Ruchensky et al., 2018). In the present study, we will use the TriPM as a tool to break each psychopathy measure into its constituent parts as it appears to be the most widely used self-report measure of psychopathy and has many studies containing correlations needed to decompose extant psychopathy measures. While the triarchic model of psychopathy, as operationalized in the TriPM, is important and has helped enable further research on the construct, it is not without issues. A primary problem recently empirically uncovered is the lack of distinguishability between disinhibition and meanness in the TriPM (Collison et al., 2021): findings indicate that they share notably overlapping content. Collison and colleagues (2021) offer that this indistinguishability is likely due to including items in each domain that cover explicit antisocial behaviors, which can be driven by either or both antagonistic/mean and impulsive/disinhibitory features. They contend that the antisocial behavior items should be moved to its own factor, as the optimal factor solution rendered from their analysis yielded an Antisociality factor. There also exists controversy regarding the relevance of boldness (and the equivalent in the PPI, fearless dominance) to the construct of psychopathy for a few reasons. First, it has shown weak convergence with many other psychopathy measures and subscales (Miller & Lynam, 2012; Murphy et al., 2016), including in its correlational profiles with important psychopathy external criteria (Crowe et al., 2021; Sleep et al., 2019; e.g., antisocial behavior, aggression, and substance use). Second, previous findings have indicated that boldness is more strongly associated with adaptive outcomes than maladaptive outcomes (e.g., Crego & Widiger, 2014; Gatner et al., 2016; Smith et al., 2013). Third, explanations for these issues (i.e., boldness will associate with externalizing behaviors when paired with higher scores on other psychopathy components such as meanness or disinhibition; boldness will evince curvilinear relations with maladaptive outcomes, such that moderate boldness is optimal and high boldness is problematic) have failed to find substantial support across empirical studies (see Crowe et al., 2021). 2. Sexual aggression Sexual aggression is defined generally as any method used to obtain some kind of sexual activity from an individual who is unwilling or not able to consent (Abbey et al., 2007; Koss et al., 1987; White & Smith, 2004). It can involve coercion in the form of threats or use of physical force, abuse of a position of power, capitalizing on a victim’s inability to resist, and verbal manipulation (Krahé, 2020). Globally, approximately 35.6% of women have experienced sexual aggression (García-Moreno et al., 2013). Less is known in terms of the global prevalence rate of sexual aggression victimization in men due to underreporting (Borumandnia et al., 2020) and the relative focus of research on female victims, but 17.1% of men in the United States have reported experiencing lifetime sexual violence (Smith et al., 2017). There are several risk factors for committing sexual aggression, including childhood sexual, emotional, and physical abuse (Abbey et al., 2006; Peterson et al., 2018), negative affect (i.e., anger, anxiety, depression; Peterson et al., 2018), hostile masculinity, psychopathy (including in college samples; DeGue & DiLillo, 2004; Knight & Guay, 2006; Kosson et al., 1997), impersonal sex (Abbey et al., 2011), a tendency to misinterpret women’s sexual cues (Abbey et al., 2011), delinquency (Abbey et al., 2012), alcohol consumption (Abbey et al., 2011), hypersexuality, sexual preoccupation, and sexual compulsivity (Knight & Graham, 2017). A highly researched area of sexual aggression is documented sexual offending, that is, sexual aggression that has led to legal ramifications for a perpetrator, including conviction. Legal classifications of sexual offenses include sexual battery (which includes rape—penetration and sodomy depending on state jurisdictions—in addition to non-consensual sexual contact), indecent exposure, child molestation, child pornography, and internet sex crimes. Examples of internet sex crimes include “sexting” with a minor, using personal information to extort nude images or sexual favors (e.g., claiming to already possess compromising photos of a victim to extort more sexually explicit photos), and online sex trafficking. Sexual offenders are a heterogeneous group (Seto, 2018; Ward & Beech, 2008). Researchers have cultivated distinct sexual offender profiles and typologies that represent different antisocial behaviors, motivations, and personality traits in order to facilitate identification of important etiological processes (Brown et al., 2015; Carvalho & Nobre, 2019; Parent et al., 2011, 2012; Walters et al., 2009; Woodworth et al., 2013). For example, child sexual offenders are often characterized by poor social skills and low self-esteem (e.g., feelings of inadequacy; Robertiello & Terry, 2007a), and often manipulate their victims into complying with sexual abuse using “grooming” methods (Craven et al., 2006). They are also likely to view children as sexual beings and think that their actions do not harm the victims (Mann & Barnett, 2013). There even is evidence of more specific phenotypes of child molesters, including affiliative (i.e., offending against relatives or when in a caregiver role) and non-affiliative (i.e., offending against strangers or acquaintances; Marshall et al., 2015; also see Strassberg DS et al., 2012). Affiliative child molestation is thought to be more common than non-affiliative child molestation and is typically marked by grooming and a seemingly nurturing relationship with a victim as opposed to blatantly coercive methods (Marshall et al., 2015; Vize et al., 2018). A recent meta-analysis indicated that child molesters differ from other sexual offenders in levels of trait agreeableness (Vize et al., 2018): while other forms of sexual aggression show a moderate negative relation with agreeableness, child molestation shows a positive moderate relation. Child molestation also has a stronger association with trait neuroticism compared to other antisocial behavior, including sexual aggression against adults, bullying, relational aggression, and reactive aggression (Vize et al., 2018). Finally, child molestation has evinced a highly similar personality profile to that of DSM-5 dependent personality disorder, which is marked by submissive and clingy behavior as well as a strong desire to be taken care of (Miller & Lynam, 2008; Vize et al., 2018). Another sexual offender profile suggested by research describes rapists: in contrast to child molesters, they are more likely to have coercive sexual fantasies and engage in violence when carrying out their crimes (Marshall, Laws, & Barbaree, 1990). Additionally, rapists are prone to hold negative views of women, such as maintaining that women are subordinate to males and enjoy coercive sexual advances (Kuznetsov et al., 1992). Moreover, some rapists are sexually motivated while others are motivated by anger or a yearning for domination and control (Robertiello & Terry, 2007). In some prototypes, rapists may feel rage and their sexual offending behavior serves to dilute psychologically threatening or emotionally uneasy states (Mann & Barnett, 2013). A final example of a sexual offender phenotype supported by research describes mixed offenders, who are defined by a combination of sexual offenses that includes child and adult victims. Their profile differentiates them from child sexual offenders and rapists in that they are characterized by thrill-seeking, a propensity to boredom, and other psychopathic features (Porter et al., 2001; Saleh et al., 2010). Some researchers contend that perpetrators against an assortment of age groups exhibit a more all-encompassing lack of empathy that extends beyond their conduct toward women and children (Cann et al., 2007). As such, while nearly all sexual offenders are putatively deficient in empathy (e.g., Mann & Barnett, 2013), the empathy deficits shown by mixed offenders are perhaps exceptionally grave and may be rooted in a more extensive, prominent psychopathic personality (Brown et al., 2015; Meloy, 2002). 3. Psychopathy and sexual aggression Research has indicated strong associations between psychopathy and sexual aggression (Knight & Guay, 2006, 2018). Research suggests that psychopathic criminals evince more sexually coercive behavior than non-psychopathic criminals (Coid, 1992; DeGue et al., 2010; Krupp et al., 2012) and that psychopathy rates are highest among rapists and mixed offenders and lowest in child molesters (Hare, 2003; Porter et al., 2009; Rice & Knight, 2019). Psychopathy is putatively a salient feature in particularly violent sexual offenders (e.g., Porter et al., 2003) and mixed offenders (Meloy, 2002). While psychopathy has been commonly associated at its global level with sexually aggressive acts, deconstructing psychopathy into underlying factors when studying how it associates with sexual aggression is an important pursuit. For example, previous studies have found sexual offenders, particularly child molesters, to report greater external locus of control and lower perceived control (Beck-sander, 1995; Fisher et al., 1999; Marsa et al., 2004) than other offenders and non-offenders, and control deficits have been linked to violent crimes (Baron, 2009; Baron & Forde, 2007). Given these findings, it is reasonable to hypothesize that sexual aggression, particularly child molestation, may associate strongly with disinhibition. Furthermore, sexual aggression is linked to the diminished ability to inhibit sexual drive in response to sexually coercive situations (Knight & Guay, 2006) as well as low conscientiousness, particularly in child molesters (see the meta-analysis by Vize et al., 2018). The same meta-analysis rendered a significant negative relation between agreeableness and sexual aggression but the opposite between agreeableness and child molestation, warranting corresponding hypotheses pertaining to sexual aggression and the triarchic psychopathy factor of meanness. Furthermore, sexual aggression is associated with lack of responsivity to others’ distress cues (Blair et al., 2005), reflecting a lack of empathy that is highly relevant to meanness. Sexual aggressors also exercise interpersonal dominance, a central feature of boldness, in coercive sexual situations (Wood et al., 2015). Given the wide use of the triarchic model of psychopathy and the relevance of its components to sexual aggression, we believe it is useful to further study what drives the varying relations between the elements in different measures of psychopathy and various kinds of sexual aggression at a granular level. Psychopathy and sexual aggression account for significant societal issues (Marshall, 1992; Prentky & Burgess, 1990), such as costs related to incarceration, victim and perpetrator treatment, investigations, and judicial processes; harm to victims, families, and communities; and lost productive output. According to some research, psychopathic sexual offenders are more difficult to treat than non-psychopathic sexual offenders (e.g., Barbaree et al., 2006), largely because they are highly resistant to treatment, cause disruptions, and exhibit several negative behaviors during treatment (Hildebrand et al., 2004). They are also more likely to recidivate, and this prediction is driven most by PCL-R Factor 2 (lifestyle and antisocial) and Facet 4 (antisocial tendencies) compared to other PCL-R factor (interpersonal and affective) and facet scores (interpersonal style, deficient affective experience, impulsive behavioral style). The likelihood of sexual recidivism in psychopathic sexual offenders is even stronger when they exhibit high levels of sexual deviance as well (see Hawes et al., 2013 for a meta-analysis). Understanding underlying personality mechanisms would help inform treatment so as to anticipate and better address particular treatment barriers and psychological processes to decrease sexually aggressive behaviors. This same level of understanding would aid in other endeavors for decreasing or preventing further sexual aggression, including risk assessments, criminal profiling for legal investigations, early intervention in schools, and making legal decisions (e.g., determining terms of legal sentencing). Previous meta-analyses on sexual aggression and psychopathy have focused on psychopathy as a predictor of sexual recidivism, but no meta-analysis to date has examined the basic association between sexual aggression and general psychopathy. The absence of such a meta-analysis in the literature is surprising given the consistent support for a strong, positive association across many individual research studies. There also is no extant meta-analysis that has investigated the relationship between sexual aggression and the triarchic factors of psychopathy nor between different forms of sexual aggression (e.g., child molestation, rape, exhibitionism) and psychopathy. Without the clarity that such an investigation would offer, it is difficult to 1) determine whether the impact of psychopathy on sexually aggressive conduct can be applied across all dimensions of psychopathy and 2) understand the heterogeneity of personality and behavior that researchers have found to exist among those who sexually aggress. Accordingly, the aim of the current study is to conduct a meta-analysis to comprehensively inspect the association between psychopathy at the global and triarchic levels and sexual aggression (including according to sexual offense typologies). There are several potential benefits of breaking psychopathy down into its constituent elements when studying it in relation to sexual aggression or any given construct (Lynam & Miller, 2015). First, it enables comparison of extant psychopathy measures. For example, rather than having some vague sense that two psychopathy measures conceptualize psychopathy a bit differently, we could ascertain those differences based on the degree that measures emphasize certain dimensions of psychopathy that map across measures with sometimes different labels, such as boldness/fearless dominance. Second, it facilitates understanding of the factor structures of extant psychopathy measures and therefore how they measure psychopathy. Third, it provides a parsimonious way to study the epidemiology of psychopathy, including comorbidities (e.g., psychopathy is comorbid with other personality disorders to the degree that they measure similar traits) and how it differs across gender and age. Fourth, it allows for understanding of the underlying mechanisms of psychopathy, including deficits (e.g., cognitive, affective, behavioral) and advantages that lead some psychopathic individuals to function reasonably well, per some conceptualizations (e.g., Hare, 1999). Fifth, it permits a bottom-up approach to understanding and refining the construct of psychopathy through the examination of which elements are unnecessary, peripheral, and central to psychopathy (Lynam & Miller, 2015). Although formally meta-analyzing how the triarchic model of psychopathy relates to sexual aggression is not possible, because of the relative lack of studies directly examining this relation, it is feasible to infer the relation by capitalizing on previously found associations between the three psychopathy factors and commonly used psychopathy measures using a relative weights analysis (Tonidandel & LeBreton, 2015). A relative weights analysis fully partitions the variance accounted for in an outcome variable by predictors; the resultant percentages will denote how much boldness, meanness, and disinhibition is contained in a given psychopathy measure. Relative weights of these three factors based on the TriPM in each of the frequently used psychopathy measures (i.e., PCL-R, PPI-R (both original and Tri versions), MMPI-2-RF-Tri, MPQ-Tri, HEXACO-Tri, Levenson Self-Report Psychopathy Scale (LSRP), Self-Report Psychopathy Scale (SRP), Interpersonal Measure of Psychopathy (IM-P), and Elemental Psychopathy Assessment (EPA), and NEO Personality Inventory-Revised Triarchic Scales (NEO-Tri)) can be obtained from correlation matrices from existing studies (e.g., TriPM and PCL-R: Venables et al., 2014, N = 169; TriPM and PPI-R original: van Dongen et al., 2017, N = 360; TriPM and PPI-R-Tri: Hall et al., 2014, N = 631; TriPM and MMPI-2-RF-Tri: Sellbom et al., 2016, N = 278; TriPM and MPQ-Tri, Brislin et al., 2015, N = 176; TriPM and HEXACO-Tri, Ruchensky et al., 2018, N = 545; TriPM and LSRP original, Weiss et al., 2021, N = 432; TriPM and SRP: Weiss et al., 2021, N = 432; TriPM and IM-P: Yoon et al., 2022, N = 152; TriPM and EPA, Weiss et al., 2021, N = 432; and TriPM and NEO-Tri: Collison et al., 2021, N = 431). The resultant percentages can help estimate how effect sizes of psychopathy on sexual aggression vary according to these relative weights across various measures of psychopathy. Despite previous difficulties with distinguishing meanness and disinhibition empirically and debate about the role of boldness in psychopathy, we hope to investigate what insights can be gleaned from examining triarchic relations with sexual aggression for the first time meta-analytically. We do expect that Boldness will be distinguishable from the other two factors and will be curious to see how Meanness and Disinhibition will behave in this meta-analysis. The proposed preregistered study comprises three parts: 1) a meta-analysis of general psychopathy’s association with sexual aggression. If a study only contains subscale scores, we will average those scores to produce a total score to use for analyses except for the relative weights moderator analysis in which case such studies will be excluded. Anticipating that some studies will have multiple measures of psychopathy or of sexual aggression, we will include study number as a random effect in our analysis; 2) moderation analyses of how general psychopathy’s association with general sexual aggression varies according to the following moderators: a) percentage of each of the three trait domains of psychopathy in the psychopathy measure used, b) type of sample used (e.g., forensic, community), c) source of information for psychopathy and sexual aggression measure used, separately (e.g., self-report, clinician ratings), d) age (mean age), e) gender (coded by percentage), f) ethnicity (coded by percentage), g) sample’s country of origin, h) measure of psychopathy (e.g., PCL-R, TriPM, SRP), i) measure of sexual aggression (e.g., dichotomous conviction presence, Sexual Experiences Survey (perpetration form), and j) victim class of perpetrators (adult, child, mixed). Because we anticipate that several moderators (i.e., sexual offense type, victim class, psychopathy measure, sexual aggression measure, and percentage of Boldness, Meanness, and Disinhibition in the psychopathy scales) will be highly correlated with each other, we will conduct a separate moderation analysis for each of these variables; and 3) separate meta-analyses of general psychopathy’s association with sexual aggression in each sexual offense type (e.g., sexual battery, indecent exposure). The sexual offense types will be based on common legal categories (i.e., sexual battery, indecent exposure, child molestation, child pornography, and internet sex crimes) that have been used in the literature. To our knowledge, there is no established sexual aggression coding scheme used by researchers; however, we will use the category of verbal sexual aggression in addition to the physical sexual aggression types informed by common legal categories and prior research to attempt to capture the full continuum of sexual aggression. We will require that the number of effect sizes for a particular sexual offense type needs to be sufficient for meta-analytic analyses (i.e., the cutoff will be set at k≥3, as done in a related meta-analysis conducted by Vize et al., 2018). If there are sufficient studies for such meta-analyses, the same moderation analyses as those conducted for the main meta-analysis (Part 1) will be conducted in each sexual offense type. A moderation analysis approach will be used to study the association between psychopathy triarchic trait domains and sexual aggression because it will help maximize the number of studies that can be used in the analysis and will allow the examination of how the effect of psychopathy on sexual aggression varies according to the measured percentages of higher-order psychopathy factors.