The Etiology of Risk in Sexual Offenders: A Preliminary Model

Abstract

Sexual offender risk assessment practice is considered by many to be atheoretical. The identification of the most predictive risk factors and tools has typically overshadowed questions about etiology. To gain insight into the origins of criminal behavior among sexual offenders, we developed and validated an etiological model of risk based on the theoretical framework of Beech and Ward. Our model focused on persistence rather than onset, and encompassed both the sexual and nonsexual criminal activity of these offenders. It comprised two pathways. The first was characterized by sexual victimization, social isolation, and early deviant sexual fantasies. It led to a prolific involvement in sexual criminality (especially toward children) and predicted sexual recidivism. The second pathway was characterized by externalization problems, sexual promiscuity, and physical/psychological victimization, and was associated with nonsexual offending and serious sexual offenses directed (mostly) toward women. It predicted all types of recidivism.

Keywords

developmental pathways etiological model recidivism risk sexual offenders structural equation modeling

Sexual offender risk assessment practice is considered by many to be atheoretical. Faced with the lack of predictive validity of unstructured clinical judgment (Grove, Zald, Lebow, Snitz, & Nelson, 2000; Hanson & Morton-Bourgon, 2009; Meehl, 1954; Mossman, 1994), researchers and practitioners have massively adopted structured techniques—that is, actuarial assessment and structured professional judgment (SPJ)—to assess the risk of individuals with a history of sexual crime (McGrath, Cumming, Burchard, Zeoli, & Ellerby, 2010; Neal & Grisso, 2014). In these approaches, risk is estimated by aggregating the best empirically validated predictors of criminal recidivism (Hanson, 2009; Knight & Thornton, 2007). Actuarial scales such as the Static-99R (Hanson & Thornton, 2000; Helmus, Thornton, Hanson, & Babchishin, 2012) or Static-2002R (Hanson & Thornton, 2003; Helmus et al., 2012) are mostly checklists of discrete risk factors previously identified in meta-analyses (e.g., Hanson & Bussière, 1998; Hanson & Morton-Bourgon, 2005). As such, there is no overarching theory in currently available scales that explains how risk factors are developed through the life course, and how they interact to cause and maintain sexual offending (Heffernan & Ward, 2015).

Theoretical Frameworks of Risk in Sexual Offenders

Theory construction is paramount to the field because it guides empirical research as well as clinical practice (Heffernan & Ward, 2015). Even if theory is not inherently integrated into actuarial and SPJ scales, numerous authors have proposed theoretical frameworks that bridge the gap between “atheoretical” risk factors, psychological mechanisms, and etiology (Beech & Ward, 2004; Brouillette-Alarie, Babchishin, Hanson, & Helmus, 2016; Mann, Hanson, & Thornton, 2010). Among these efforts, Beech and Ward’s (2004) etiological model of risk is probably the most influential. Their model explains how developmental factors lead to psychological vulnerabilities, which when triggered by contextual features result in high-risk psychological states that can precipitate recidivism (see Figure 1).

Figure 1.

Beech and Ward’s (2004) etiological model of risk in sexual offenders.

Beech and Ward (2004) argue that risk factors should not be seen as statistical correlates without clinical meaning, but rather as observable symptoms of latent psychological propensities that are related to recidivism. This type of inductive reasoning is common in trait theories of personality (Cattell & Kline, 1977; Widiger & Costa, 2013), where overt patterns of behavior, thought, and emotion are considered manifestations of latent psychological traits (e.g., extraversion, neuroticism). Because static and stable risk factors in actuarial scales are mostly behavioral, it should be possible to use them to infer the broad psychological constructs responsible for recidivism risk (Brouillette-Alarie et al., 2016; Mann et al., 2010). In this context, items from static risk scales are seen as measures of past manifestations of psychological vulnerabilities, while items from stable risk scales are seen as more direct and current measures of those vulnerabilities (Beech & Ward, 2004). For example, a sexual offender who has boy victims (static risk factor) and phallometrically measured sexual arousal to children (stable risk factor) can be assumed to have a certain degree of pedophilia, a psychological propensity that is associated with sexual recidivism. Therefore, according to Beech and Ward (2004), static and stable risk factors represent two ways of measuring the same latent psychological dispositions; the distinction between them has heuristic value, but it is not clear that they are assessing fundamentally different psychological attributes.

Like most other psychological traits, risk-relevant psychological vulnerabilities can be assumed to have developmental antecedents (Beech & Ward, 2004). Etiological models of sexual coercion (e.g., Daversa & Knight, 2007; Malamuth, Sockloskie, Koss, & Tanaka, 1991) have identified numerous developmental factors associated with sexual violence. Having a good understanding of the developmental factors associated with the risk of sexual violence is vital in the field, as it enables primary and secondary prevention efforts (Knight & Sims-Knight, 2003). According to multiple authors, we are at the point where risk assessment should be woven into life course developmental perspectives (Beech & Ward, 2004; Knight & Thornton, 2007; Lussier & Davies, 2011; Lussier, Leclerc, Cale, & Proulx, 2007). However, developmental factors have mostly been left out of actuarial scales due to the limited (direct) association they have with recidivism (Hanson & Morton-Bourgon, 2005). The association could here be indirect: Developmental factors, while not directly related to recidivism, contribute to the crystallization of psychological vulnerabilities that are risk-relevant later in adulthood (Brouillette-Alarie, Longpré, & Proulx, 2014).

The last component of Beech and Ward’s (2004) etiological model of risk is acute risk factors, which are divided between triggering events and high-risk psychological states. When psychological vulnerabilities (e.g., pedophilia) are subjected to triggering events from the social environment (e.g., victim access), they produce acute mental states (e.g., sexual arousal in presence of a child) that can precipitate recidivism. Therefore, by emphasizing the role of social context, Beech and Ward (2004) make an important distinction between risk status and risk state (see also Douglas & Skeem, 2005). Risk status is determined by static and stable risk factors, and refers to long-lasting, interindividual differences in risk (e.g., Mr. X has a higher Static-99R score than Mr. Y). In turn, risk state is determined by acute risk factors, and refers to intraindividual variations in risk over time (e.g., Mr. X is currently at a higher risk of acting out than usual, because he just lost his job).

In sum, Beech and Ward (2004) have shown that static, stable, and acute risk factors from actuarial scales can be mapped onto an etiological model of offending in which developmental antecedents, psychological vulnerabilities, triggering events, and acute mental states interact to cause recidivism. However, to date, there has been no empirical validation of an etiological model of risk, no matter how interesting or relevant. Therefore, the objective of the present study was to operationalize and test an etiological model of risk in sexual offenders, based on Beech and Ward’s (2004) conceptual model. Because Beech and Ward’s (2004) proposition deals with broad concepts (e.g., there are multiple ways to determine the number of relevant psychological vulnerabilities and to define them), any model based on their theory is likely to be idiosyncratic. Therefore, it would be more accurate to say that we tested one of the many possible operationalizations of their conceptual model. The following section will describe the literature we used to inform model construction.

Literature Review

Psychological Vulnerabilities

Literature on the dimensions of risk in sexual offenders was used to operationalize psychological vulnerabilities. Over the last 15 years, many studies have sought to identify the latent psychological constructs in actuarial scales for sexual offenders, using factor analysis (e.g., Allen & Pflugradt, 2014; Barbaree, Langton, & Peacock, 2006; Brouillette-Alarie et al., 2016; Brouillette-Alarie & Proulx, 2013; Knight & Thornton, 2007; Olver et al., 2016; C. F. Roberts, Doren, & Thornton, 2002; Seto, 2005). The results of these studies, despite their methodological differences, have been surprisingly consistent. Three constructs are usually found (Brouillette-Alarie, Hanson, Babchishin, & Benbouriche, 2014).

The first construct, Persistence/Paraphilia, is defined by static items related to sexual criminality (e.g., prior sexual offenses) and paraphilic sexuality (e.g., child victims, noncontact sexual offenses; Brouillette-Alarie et al., 2016). It correlates with stable indicators that are reminiscent of the characteristics of fixated child molesters (Groth, Hobson, & Gary, 1982; Knight & Prentky, 1990), namely, deviant sexual interests/paraphilias (especially pedophilia), emotional identification with children, and grooming offending strategies (Brouillette-Alarie & Hanson, 2015; Brouillette-Alarie, Proulx, & Hanson, 2017).

The second construct, General Criminality, comprises static items that reflect the magnitude, violence, and diversity of criminal careers (e.g., number of prior sentencing occasions, prior nonsexual violence; Brouillette-Alarie et al., 2016). It correlates with features of antisocial personality disorder (American Psychiatric Association, 2013) and psychopathy (Hare, 2003), such as impulsivity, lack of empathy, and manipulativeness (Brouillette-Alarie & Hanson, 2015; Brouillette-Alarie et al., 2017).

The third, less internally consistent construct, Youthful Stranger Aggression, comprises static items related to young age, unrelated/unknown sexual victims, and violence in the index offense (Brouillette-Alarie et al., 2016). It correlates with stable indicators of sexual sadism and hostility, and was, therefore, interpreted by the authors as a general motivation to harm victims (Brouillette-Alarie & Hanson, 2015; Brouillette-Alarie et al., 2017).

The first construct exclusively predicts sexual recidivism and is more commonly found in child molesters than rapists (Brouillette-Alarie et al., 2016). The second and third constructs, in turn, predict all types of recidivism and are more common in rapists than child molesters. They are substantially correlated. As one of our reviewers suggested, it might be that General Criminality and Youthful Stranger Aggression are two sides of the same coin. The former could be an expression of antisocial tendencies at an older age, involving a more extensive criminal record, while the latter could be an expression of early onset antisociality, where charges have yet to accumulate. Early onset and serious offending are empirically related (Moffitt, 1993; Yessine & Bonta, 2008), which could explain the clustering of violence and youth in the Youthful Stranger Aggression construct.

Even though these dimensions of risk are not the most psychologically subtle (factor analysis is, after all, a data reduction technique), they nevertheless constitute a methodologically sound starting point for the operationalization of psychological vulnerabilities.

Developmental Antecedents

Two areas of research give insight on the developmental factors associated with the onset and persistence of (sexual) offending: etiological models of sexual coercion and developmental criminology. We identified five etiological models of sexual coercion. Three explained sexual violence toward women (Knight & Sims-Knight, 2003; Lussier, Proulx, & LeBlanc, 2005; Malamuth et al., 1991), one explained sexual violence toward children (Daversa & Knight, 2007), and one explained both types of sexual violence (Lussier et al., 2007). We also identified four articles that linked various developmental problems to the criminal activity of adult sexual offenders (Beauregard, Lussier, & Proulx, 2004; Lee, Jackson, Pattison, & Ward, 2002; Lussier, Beauregard, Proulx, & Nicole, 2005; Prentky et al., 1989). The aforementioned references were reviewed to identify the developmental factors that were most relevant to our study (see Table S1 in the supplementary material).

Three clusters of variables are commonly reported: childhood victimization (sexual, physical, and psychological), problematic sexual development (early first sexual experiences/sexual promiscuity, early deviant sexual fantasies/behaviors), and early antisocial tendencies (substance abuse, conduct disorder, juvenile nonsexual delinquency). These developmental factors are likely related to all types of sexual coercion, as they figure in models that explain sexual violence toward children and adult women. Some factors, however, are exclusively related to the criminal activity of rapists (e.g., hostile masculinity) or that of child molesters (e.g., social rejection/isolation, sexual inadequacy with age-appropriate peers). This suggests that different developmental pathways lead to different psychological vulnerabilities in adulthood (Lussier et al., 2007).

Acute Risk Factors

The recent interest in acute risk factors can be traced back to studies on the offending process of sexual offenders (e.g., Pithers, Marques, Gibat, & Marlatt, 1983; Proulx, Perreault, & Ouimet, 1999; Ward, Louden, Hudson, & Marshall, 1995). By analyzing the events that preceded sexual offenses, researchers have highlighted the characteristics that suggest that relapse is imminent. Most of the offending process studies were, however, based on a limited number of participants (n < 50), potentially limiting their generalizability. Years later, larger samples were used in studies such as the Dynamic Supervision Project (n = 749; Hanson, Harris, Scott, & Helmus, 2007). In the latter, seven acute risk factors were empirically validated: victim access, hostility, sexual preoccupation, rejection of supervision, emotional collapse, collapse of social supports, and substance abuse. According to Beech and Ward (2004), victim access, rejection of supervision, collapse of social supports, and substance abuse can be classified as triggering events/contextual risk factors, while hostility, sexual preoccupation, and emotional collapse correspond to acute psychological states.

It should be noted that in the data set used in the present study, acute risk factors preceding recidivism were unavailable (only events preceding the index offense were gathered). Therefore, our etiological model of risk is limited to the following three components: developmental factors, psychological vulnerabilities, and recidivism outcomes.

Method

Sample

The sample consisted of 613 men who were convicted of at least one contact sexual offense in Quebec between 1995 and 2000 (Proulx, Beauregard, Lussier, & Leclerc, 2014). This sample is a near population, as it includes 93.5% of all offenders who met the above criteria (the remaining 6.5% declined to participate in the study). These participants were all under federal supervision, which meant that they received a sentence of 2 or more years for their index offense. A minority (11.3%) of Canadian sexual offenders receive federal sentences; most receive noncustodial (51.3%) or provincial (37.4%) sentences of less than 2 years (Hanson, Lloyd, Helmus, & Thornton, 2012). Consequently, the risk of the participants in this study can be assumed to be higher than that of nonfederal offenders. Static-99R scores of our participants (M = 2.4, SD = 2.4) were, however, not much higher than those of more routine (unselected) samples (e.g., the Dynamic Supervision Project; M = 2.1, SD = 2.3; Hanson et al., 2012).

The mean age of the participants was 39.5 (SD = 12.1) years old. More than half (65.8%) were single at the time of their incarceration, 23.1% were in a relationship, and 11.2% were married. Most had high school education (89.2%), and a few went to college (4.7%) or university (6.1%). Among the 613 offenders, 355 were child molesters (their victims were all aged 15 years or less), 174 were rapists (their victims were all aged 16 years or more), 59 were mixed offenders (they had victims of both categories), and 25 could not be classified due to missing data.

Data Collection

Data were collected as participants went through their intake assessment at the Regional Reception Center (RRC) in Sainte-Anne-des-Plaines (Quebec, Canada), a maximum-security penitentiary of the Correctional Service of Canada. During their 6-week stay at the institution, they were evaluated by a multidisciplinary team of psychologists, psychiatrists, criminologists, sexologists, vocational training professionals, and correctional agents. This evaluation was not only carried out for research purposes, but it was also used by the correctional service to determine the level of supervision and the treatment targets for each offender before they were transferred to another institution.

Data were collected in a semistructured interview that followed the Computerized Sex Offender Questionnaire (CSOQ; St-Yves, Proulx, & McKibben, 1994). This questionnaire addresses developmental history, criminal career (including recidivism data), personality and mental health characteristics, general and sexual lifestyles (up to 1 year prior to the index offense), precrime factors (up to 48 hr prior to the index offense), and modus operandi (of the index offense). Most of the offenders also underwent phallometric and psychometric testing. When possible, self-disclosed information was compared with official records, the latter being authoritative. Criminal career data were extracted from the national records of the Royal Canadian Mounted Police’s Fingerprint System.

Interrater reliability tests were performed by the data collection team on 92 dichotomous variables. The mean Cohen’s kappa was .86 (SD = .18), indicating almost perfect agreement (Landis & Koch, 1977).

Measures

Psychological vulnerabilities

Psychological vulnerabilities and their ability to predict recidivism were modeled on the basis of the three main dimensions of risk identified by the Static-99 research group (Brouillette-Alarie et al., 2016; Brouillette-Alarie & Hanson, 2015; Brouillette-Alarie et al., 2017). Each of these dimensions has static and stable indicators. The first latent variable, Sexual Criminality/Deviance, was measured by Persistence/Paraphilia (static) and three psychological scales: Sexual Interests Toward Children, Noncoercive Offending Strategies, and Multiple Paraphilias. The second latent variable, Antisocial Traits, was measured by General Criminality (static) and two psychological scales: Sexual Promiscuity and Noncriminal Antisocial Behavior. The third latent variable, Intent to Harm Victims, was measured by Youthful Stranger Aggression (static) and two psychological scales: Hostility Toward Women and Sexual Sadism (as measured by the Sexual Sadism Scale [SeSaS]; Mokros, Schilling, Eher, & Nitschke, 2012). All latent variables were associated with sexual recidivism, but only Antisocial Traits and Intent to Harm were associated with nonsexual types of recidivism. We allowed Antisocial Traits and Intent to Harm to covary with each other, as suggested by earlier studies (Brouillette-Alarie et al., 2016; Brouillette-Alarie et al., 2017). The constituents and internal consistency of each of these scales can be found in the supplementary material (Tables S2 and S3).

To ensure that the overlap between developmental factors and psychological vulnerabilities was minimal, only events experienced in adulthood were used to score psychological scales. In turn, only events that happened before the offender was 18 years old were used to score developmental factors.

Developmental antecedents

Commonly reported developmental factors in etiological studies of sexual coercion were operationalized with variables of the RRC data set. Exposure to violence and direct victimization were merged to avoid redundancy. Furthermore, because physical and psychological victimization were nearly collinear (r = .77, p < .001), they were merged into a single scale to prevent problems in the regression equations of our structural equation model. There were no other cases of collinearity in our data. The constituents and internal consistency of developmental factors can be found in the supplementary material (Table S4).

Developmental antecedents were integrated into the model based on preliminary analyses between developmental factors and static risk constructs. The magnitude of Pearson correlations was used to screen out developmental variables (see Table S5 in the supplementary material). Variables that did not share a correlation of at least .15 with a static risk construct were left out of the model. This cutoff meant that the following developmental factors were associated with Sexual Criminality/Deviance: Sexual Victimization, Early Sexual Deviance, and Social Rejection/Isolation. Antisocial Traits were associated with Physical/Psychological Victimization, Sexual Promiscuity, Conduct Disorder, Substance Abuse, nonsexual crimes in the family, and school indiscipline. Finally, Intent to Harm was associated with Parental Separations, Sexual Promiscuity, Conduct Disorder, Substance Abuse, Social Rejection/Isolation, school indiscipline, and having received psychiatric treatments.

Developmental factors were entered in the model as manifest variables, to prevent identification problems, which can occur when there is a limited number of indicators per latent variable (Kenny, 2012). Indeed, our developmental scales often comprised one or two variables. Furthermore, our data set did not comprise the necessary variables to operationalize neglect/lackluster parental care, permissive/chaotic parental discipline, sexual inadequacy with age-appropriate peers, and hostile masculinity. Thus, not all potential developmental antecedents were accounted for. Finally, sexual and nonsexual juvenile delinquency was intentionally left out of developmental variables, because the onset of criminal career was already covered by items of the static risk constructs.

Recidivism data

Recidivism data were collected in official national criminal records and encompassed both charges and convictions. Three recidivism outcomes were examined: sexual (contact and noncontact sexual crimes), nonsexual violent (any violent crime except those sexual in nature), and nonsexual nonviolent (all other crimes). Recidivism data were last collected during the summer of 2007, which led to an average follow-up time of 7.7 years (SD = 2.6). Because survival analyses are not part of structural equation modeling (SEM), a fixed follow-up time of 5 years was used for recidivism outcomes. This implies that participants without such a follow-up had missing data on these variables (107 missings for sexual recidivism, 111 for nonsexual violent recidivism, 109 for nonsexual nonviolent recidivism).

Analytical Strategy

SEM was chosen to test our etiological model of risk. SEM is a collection of statistical techniques that is appropriate for theory testing (Schumacker & Lomax, 2016; Tabachnick & Fidell, 2013). By combining confirmatory factor analysis and multiple regression models, it can simultaneously analyze how latent constructs are measured by manifest variables (the measurement model), and how these constructs are related to each other or other manifest variables (the structural model). Contrary to most statistical methods, SEM explicitly takes measurement error into account (Schumacker & Lomax, 2016).

Our structural equation model was conducted in Mplus 6.12 (L. K. Muthén & Muthén, 2010). Parameters of the model were estimated with the weighted least squares means and variance adjusted (WLSMV) method. This estimator was developed by the Mplus authors (B. Muthén, 1983, 1984) and is recommended when dichotomous and ordinal variables are used (Brown, 2006). Missing data were handled by pairwise deletion, the default Mplus algorithm when WLSMV estimation is used.

Model fit

The adequacy of the model was assessed using three fit indices. The root mean square error of approximation (RMSEA) assesses lack of fit in a model relative to a perfect saturated model (Tabachnick & Fidell, 2013). It should not exceed .06 (Hu & Bentler, 1999). The comparative fit index (CFI) assesses model fit relative to a baseline model where there are no relationships between items. Similarly, the Tucker–Lewis index (TLI) assesses model fit relative to a baseline model but takes model complexity into account (Tabachnick & Fidell, 2013). Recommended CFI and TLI values vary depending on authors, with .90 being the minimum and .95 being preferable (Hu & Bentler, 1999; Schumacker & Lomax, 2016).

Model modification

Model refinement was done with Mplus’s modification indices. Among the numerous suggestions made by the software, only those that had practical significance, substantive meaning, and an important effect on model fit were followed. In addition, each variable was individually removed from the model to see if the overall fit would be improved. This step was performed because modification indices cannot suggest the deletion of a variable, no matter how damaging for model fit. Finally, to make the model cleaner, any relationship that was found to be nonsignificant postmodification was removed from the model. If a variable found itself with no relationship, it was also deleted.

Results

Descriptive Statistics

Descriptive statistics of the RRC participants can be found in Table 1. They are presented for the full sample, rapists, and child molesters. To identify meaningful group differences, t tests, crosstabs, and corresponding effect size analyses were performed.

Table 1.

Descriptive Statistics.

Variables (range)	Full sample(N = 613)	Child molesters(n = 355)	Rapists(n = 174)	Group differences
Variables (range)	M (SD)/%			Φ/Cramer’s V/r
Risk
Static-99R total score (−3-12)	2.39 (2.35)	1.83 (2.44)	3.27 (1.85)	.32***
Static-2002R total score (−2-13)	3.96 (2.54)	3.42 (2.64)	4.77 (2.01)	.28***
Persistence/paraphilia (0-7)	1.07 (1.47)	1.24 (1.58)	0.58 (1.04)	.24***
General criminality (0-6)	2.50 (1.92)	2.08 (1.78)	3.20 (1.95)	.29***
Youthful stranger aggression (−2-7)	1.74 (1.76)	1.19 (1.69)	2.71 (1.47)	.43***
Recidivism
Sexual recidivism at 5 years	8.5%	6.6%	9.8%	.06
Nonsexual violent recidivism at 5 years	15.7%	9.2%	30.5%	.27***
Nonsexual nonviolent recidivism at 5 years	23.4%	17.1%	37.6%	.22***
Psychological scales
Sexual Interests Toward Children (0-3)	1.02 (0.98)	1.54 (0.85)	0.02 (0.15)	.78***
Noncoercive Offending Strategies (0-2)	0.68 (0.76)	1.03 (0.73)	0.13 (0.40)	.61***
Multiple Paraphilias (0-1)	44.1%	49.7%	23.4%	.25***
Sexual Preoccupation (0-9)	2.93 (1.93)	2.73 (1.85)	3.27 (1.95)	.14**
Noncriminal Antisocial Behavior (0-8)	2.26 (1.85)	1.70 (1.58)	3.04 (1.92)	.36***
Hostility Toward Women (0-2)	0.49 (0.84)	0.27 (0.67)	0.78 (0.97)	.30***
Sexual Sadism (0-9)	1.36 (1.65)	0.72 (1.27)	2.67 (1.54)	.57***
Developmental antecedents
Sexual victimization (0-2)	0.62 (0.70)	0.68 (0.69)	0.42 (0.63)	.19***
Physical/psychological victimization (0-4)	1.99 (1.68)	1.94 (1.67)	2.00 (1.75)	.02
Parental separations (0-2)	0.73 (0.83)	0.70 (0.85)	0.77 (0.81)	.12*
Sexual promiscuity (0-6)	1.45 (1.40)	1.27 (1.34)	1.75 (1.38)	.17***
Early sexual deviance (0-2)	0.13 (0.39)	0.14 (0.39)	0.08 (0.33)	.10
Conduct disorder (0-5)	1.26 (1.33)	0.99 (1.22)	1.63 (1.37)	.24***
Substance abuse (0-6)	1.16 (1.73)	0.82 (1.45)	1.77 (2.02)	.26***
Social rejection/isolation (0-2)	0.87 (0.82)	0.86 (0.83)	0.81 (0.81)	.04
Self-harm (0-2)	0.12 (0.38)	0.09 (0.35)	0.13 (0.37)	.10
A parent or sibling has committed sexual crimes	8.7%	8.1%	6.4%	.03
A parent or sibling has committed nonsexual crimes	33.8%	31.2%	39.6%	.08
Undisciplined in school	37.1%	29.1%	49.7%	.20***
Educational level (0-3)	0.90 (0.74)	0.90 (0.79)	0.91 (0.67)	.01
Underwent treatment for a sexual problem	0.9%	0.9%	0.0%	.05
Underwent psychiatric treatment	6.3%	4.0%	11.1%	.14***

Note. For dichotomous variables, the Φ is reported. For trichotomous variables, the Cramer’s V is reported. For variables with 4+ categories, the r is reported. Effect sizes are all listed positively; the direction of the relationship can be seen in the means/%.

p < .05. **p < .01. ***p < .001.

Rapists were at higher risk (Static-99R/Static-2002R) than child molesters and, accordingly, tended to reoffend in a higher proportion. The difference was, however, not statistically significant for sexual recidivism. Compared with rapists, child molesters had higher scores on Persistence/Paraphilia and psychological scales related to sexuality (Sexual Interests Toward Children, Noncoercive Offending Strategies, and Multiple Paraphilias), except Sexual Preoccupation, which was more prevalent in rapists than child molesters. Child molesters had lower scores on General Criminality, Youthful Stranger Aggression, and psychological scales tapping into antisociality and violence (Noncriminal Antisocial Behavior, Hostility Toward Women, and Sexual Sadism). These group differences were in line with the existing literature (Olver, Wong, Nicholaichuk, & Gordon, 2007; Parent, Guay, & Knight, 2011).

Unlike static risk indicators and psychological scales, developmental antecedents were not heavily group-dependent. Most developmental factors were similarly prevalent in both types of offenders (Physical/Psychological Victimization, Early Sexual Deviance, Social Rejection/Isolation, Self-Harm, crime in the family, educational level, and treatment for a sexual problem). A few were group-dependent, but effect sizes were low (<.20; Sexual Victimization, Parental Separations, Sexual Promiscuity, and Psychiatric Treatment). That being said, developmental factors related to early antisocial behavior (Conduct Disorder, Substance Abuse, and school indiscipline) were more common in rapists than child molesters.

Initial Model

The initial etiological model of risk in sexual offenders can be found in Figure 2. SEM conventions are respected: manifest variables are represented by rectangles, and latent variables are represented by ovals (Tabachnick & Fidell, 2013). Arrows starting from latent variables pointing toward manifest variables refer to the measurement model (confirmatory factor analysis), while double-headed arrows mean that two variables are allowed to correlate. All other arrows represent a direct relationship (the structural model). To improve model clarity, manifest indicators of latent variables are colored in gray.

Figure 2.

Initial structural equation model.

Even though the initial model successfully converged, its fit indices were largely unsatisfactory. The RMSEA (.073), CFI (.522), and TLI (.426) all failed to meet the threshold for adequate fit, and multiple relationships were nonsignificant. Three steps were taken to improve our model: (a) six paths were added (three pertaining to the measurement model and three covariance connections), (b) two variables were deleted (Multiple Paraphilias and Sexual Preoccupation), and (c) two prediction links were removed because they did not reach statistical significance. These modifications emphasized the contrast between Sexual Criminality/Deviance and Intent to Harm. Indeed, the former is defined by deviant sexual interests toward atypical objects without explicit intent to harm, whereas the latter is defined by a motivation to inflict pain (Brouillette-Alarie et al., 2017). They also highlighted the overlap between general and sexual criminality, which is consistent with sexual offenders being more often generalists than specialists (Harris, Knight, Smallbone, & Dennison, 2011; Lussier, 2005).

Worthy of note, Antisocial Traits was removed as a predictor of sexual recidivism. Even though this is contrary to theory (Doren, 2004), this relationship was far from reaching statistical significance in our model. A mediation effect seemed to limit Antisocial Traits’ predictive validity. Indeed, removing Intent to Harm as a predictor of sexual recidivism made Antisocial Traits’ contribution significant, and vice versa. This lack of incremental validity could be attributed to the important shared variance between these constructs.

Three-Factor Model

The three-factor version of our model can be found in Figure 3. The modifications provided a substantially better fit for the model. The RMSEA (.034) was very good, and the CFI (.924) and TLI (.903) were acceptable to good. In addition, the new paths significantly improved the percentage of variance explained of sexual recidivism (from .18 to .33).

Figure 3.

Three-factor structural equation model.

Despite the presence of three latent psychological vulnerabilities, our model converged on two pathways. In the first one, Sexual Victimization, Early Sexual Deviance, and Social Rejection/Isolation led to Sexual Criminality/Deviance in adulthood, and then sexual recidivism. In the second one, a nexus of developmental variables (school indiscipline, Conduct Disorder, Substance Abuse, and Sexual Promiscuity) was associated with both Antisocial Traits and Intent to Harm Victims, which predicted all types of recidivism (notwithstanding the mediation effect). Some developmental variables were exclusively related to Intent to Harm Victims (Parental Separations, psychiatric treatment) or Antisocial Traits (Physical/Psychological Victimization, nonsexual crimes in the family).

Considering that Antisocial Traits and Intent to Harm Victims were substantially correlated, and had a similar etiology as well as similar predictive validity patterns, it seemed like they could be collapsed into a more parsimonious two-factor model. To do so, we merged their indicators into one latent variable: General/Violent Criminality. Developmental factors that were predictive of both Antisocial Traits and Intent to Harm Victims were linked with the new construct, which was then linked with all types of recidivism (see Figure 4).

Figure 4.

Two-factor structural equation model.

Modifications indices suggested that we add Youthful Stranger Aggression as an indicator of Sexual Criminality/Deviance (negative link), as well as a negative covariance connection between Noncriminal Antisocial Behavior in adulthood and Youthful Stranger Aggression. Even though the negative covariance connection appears counterintuitive at first, it might be that these scales characterize offenders at different age brackets, as suggested by one of our reviewers. The violence inherent to Youthful Stranger Aggression could negatively correlate with the more experienced, adult antisocial tendencies measured by the Noncriminal Antisocial Behavior scale.

The fit indices of the two-factor model were very similar to those of the three-factor model, if slightly inferior (RMSEA = .042, CFI = .918, TLI = .895). In return, the new model was more parsimonious, less redundant, and did not suffer from a mediation effect.

Discussion

The objectives of the present study were to operationalize and test an etiological model of risk in sexual offenders. Our study suggests that such an enterprise is viable: Fit indices of the final models were good, and the percentage of variance of dependent variables that was explained was substantial (.25 to .61). Juvenile developmental experiences led to psychological vulnerabilities in adulthood, and psychological vulnerabilities predicted recidivism outcomes in ways that were expected. We did not model direct links between developmental factors and recidivism, and modification indices did not suggest that we should have. Our results therefore confirm that developmental factors indirectly lead to recidivism, by contributing to the crystallization of risk-relevant psychological propensities.

Developmental Pathways to the Onset and Persistence of Offending

Analyses of developmental factors and psychological vulnerabilities highlighted two developmental pathways that were quite differentiated. In the first pathway, Sexual Victimization, Social Rejection/Isolation, and early manifestations of deviant sexual interests led to a prolific sexual criminality in adulthood, as indicated by the Sexual Criminality/Deviance construct, and to sexual recidivism. Sexual crimes tended to be directed toward children, be underlaid by pedophilic sexual fantasies, and feature low levels of coercion. They were an important predictor of future sexual offending, but not of other types of recidivism. The prototypical offender of the first pathway would be the fixated child molester, who is characterized by preferential pedophilic sexual interests, social isolation, emotional identification with children, grooming offending strategies, and high sexual recidivism risk (Groth et al., 1982; Knight & Prentky, 1990; Proulx et al., 1999).

The second pathway was constituted of a nexus of developmental variables (school indiscipline, Conduct Disorder, Substance Abuse, and Sexual Promiscuity) that was associated with both Antisocial Traits and Intent to Harm Victims (or General/Violent Criminality in the two-factor model). Whereas offenders in the first pathway seemed to be akin to children bullied in school (they were socially isolated and had low self-esteem; Carney & Merrell, 2001; Crick & Grotpeter, 1995; Nansel et al., 2001), offenders in the second pathway were likely the ones distributing the bullying. Indeed, they shared multiple characteristics with children who engage in bullying behaviors, such as poor academic achievement, conduct problems, aggressiveness, and substance abuse (Knight & Sims-Knight, 2003; Nansel et al., 2001; Natvig, Albrektsen, & Qvarnstrøm, 2001; Rigby & Slee, 1993; S. J. Roberts, Glod, Kim, & Hounchell, 2010; Wolke, Copeland, Angold, & Costello, 2013). As bullies tend to do, these juveniles went on to perpetrate various forms of crimes in adulthood (Ttofi, Farrington, Lösel, & Loeber, 2011).

The second pathway was associated with criminal versatility. Indeed, Antisocial Traits reflect the magnitude, violence, and diversity of criminal careers, and Intent to Harm Victims comprises static risk factors related to sexual (e.g., unrelated/stranger sexual victims) and nonsexual (e.g., index nonsexual violence) criminality. Intent to Harm Victims emphasizes the seriousness of sexual offenses rather than their repetition. Therefore, offenders characterized by the second pathway would be expected to have long, violent criminal records in which sexual crimes are not the defining feature. Antisocial Traits and Intent to Harm Victims (mostly) predicting all types of recidivism was consistent with this hypothesis. Descriptive statistics revealed, without surprise, that the second pathway was more common in rapists than child molesters.

While it is cognitively intuitive to associate each of the developmental pathways with specific offender types, one should remember that both pathways can apply to the same offender. While rarer, some child molesters do exhibit strong sadistic and antisocial tendencies (Beauregard, Proulx, & Leclerc, 2014; Groth et al., 1982; Knight & Prentky, 1990), and some rapists do become highly specialized in sexual crimes (e.g., serial rapists; Ciardha, 2015; de Heer, 2016; Slater, Woodhams, & Hamilton-Giachritsis, 2014). These offenders would be expected to have lived through the developmental hurdles of the first and second pathways, putting them at a particularly high risk of recidivism. Conversely, offenders who are at a very low risk could be characterized by neither developmental pathway. Because our etiological model is latent variable-centered, it accounts for a wide number of types.

Links Between Our Developmental Pathways and Etiological Models of Sexual Coercion

Our developmental pathways shared multiple similarities with earlier etiological models of sexual offending. First, as in Knight and Sims-Knight’s (2003) model of sexual coercion against women, the effects of physical victimization were clearly differentiated from those of sexual victimization. The former led to antisocial/psychopathic tendencies, while the latter led to deviant sexual fantasies. This differentiation was also found in Lussier et al.’s (2005) model of the criminal activity of sexual aggressors of women. Although those two models applied to rapists only, our model suggests that the same explanatory pathways can be generalized to all types of sexual offenders. Second, the co-occurrence of early antisocial tendencies and sexual promiscuity in the pathway that (mostly) explained the criminal sexual activity of rapists was reminiscent of Malamuth et al.’s (1991) confluence model of sexual aggression toward women. In their model, sexual promiscuity originates from general delinquency, more specifically, from the importance that sex has in the identity construction of delinquent youth. It was, therefore, not surprising to find Sexual Promiscuity in the antisocial developmental pathway rather than the one centered on pedophilic sexual deviance. Third, as in Lussier et al.’s (2005) model, the pathway focused on sexualization was predictive of sexual criminality, while the general deviance pathway was predictive of sexual and nonsexual criminality. Again, our model suggests that Lussier et al.’s (2005) pathways also apply to other types of sexual offenders.

Our model has advantages and disadvantages (see “Limitations” section) compared with previous etiological models of the criminal activity of sexual offenders. First, it differentiates between the onset (static risk constructs) and persistence (static risk constructs + recidivism outcomes) of offending, which previous models often confounded. For example, Knight and Sims-Knight (2003) define their dependent variable as the number of minor, moderate, and serious sexual offenses against women. Therefore, their model accounts for onset and repetition but does not clearly differentiate them. Second, our model factors in the criminal activity of all types of sexual offenders—not just of rapists or child molesters. However, it does not efface differences between them; by having two distinct pathways that are closely related to each type of sexual offender, it highlights the fact that their backgrounds might not be common. Third, our model accounts for the sexual and nonsexual criminal activity of sexual offenders. Considering that most crimes committed by sexual offenders are not sexual in nature (Harris et al., 2011; Lussier, 2005), and that their recidivism is more likely to be nonsexual than sexual (Hanson & Bussière, 1998; Hanson & Morton-Bourgon, 2005; Prentky, Lee, Knight, & Cerce, 1997), it is paramount for theoretical models to explain all these offenders’ potential criminal activity. Our model enables such an explanation, while clearly differentiating between the pathways associated with sexual and general criminality.

Limitations

The present study is not without limitations. First, our sample exclusively comprised federally sentenced sexual offenders. Therefore, it is not certain that results derived from it are generalizable to other correctional samples. That being said, the average risk of our sample was not much different from that of the Dynamic Supervision Project, which comprises both federal and nonfederal offenders (Hanson et al., 2012). Second, because CSOQ data were collected not only for research purposes but also for correctional decision making, participants could have considered hiding relevant information or trying to fake good (Furnham, 1986) during the assessment process. Although data from official records were compared with self-reported measures to identify malingerers, such comparisons were not always possible, or exhaustive.

Our model itself also has numerous imperfections that need to be mentioned. First, psychological scales were sometimes based on a scarce number of indicators. For example, Hostility Toward Women was based on two self-reported dichotomous indicators of conflicts with women. Second, some psychological features such as sexual coping or implicit theories were not available in our data set. The same was true for a small number of developmental variables (e.g., neglect/lackluster parental care). Combined with the lack of acute risk factors, our etiological model of risk in sexual offenders was certainly not perfectly specified. Despite these issues, its fit indices were acceptable to good. Nevertheless, basic SEM principles suggest that they could be improved with a more exhaustive model. Third, compared with earlier etiological models of sexual coercion, our model did not operationalize developmental variables as latent variables. This was the result of building the model around psychological vulnerabilities rather than developmental variables, and trying to avoid identification issues in SEM. Indeed, psychological vulnerabilities were modeled with clear theoretical a prioris, while developmental variables were integrated based on exploratory statistical analyses. Doing so may have undermined the parsimony of the model. Finally, because our model encompassed all types of sexual offenders, it may have highlighted effects that are by-products of the contrast between child molesters and rapists. For example, the presence of sadism in the “rapist pathway” could simply be due to noncoercive sex with adults being legal, meaning that very few nonviolent sexual offenses toward women were registered in the data set. This is in contrast with sexual offenses against children, which feature varying levels of coercion.

Another important limitation pertains to recidivism data. Because recidivism was split into three outcomes (sexual, nonsexual violent, nonsexual nonviolent) and exposure time during the follow-up period could not be accounted for, offenders who committed one type of recidivism could have had limited opportunities to engage in other types of recidivism because of reincarceration. Therefore, sexually reoffending may have precluded participants from committing nonsexual and nonsexual violent recidivism, and vice versa. Because of that, the links between the two pathways and the three recidivism outcomes could have been biased.

The research design of the RRC data set also comes with limitations. Even though our model is etiological in nature, it is still based on retrospective data. Thus, it is incapable of highlighting causal relationships. Doing so would require longitudinal data, which our data set did not comprise. This has implications for the mediation effect that was found in the three-factor model. Indeed, mediation effects found in cross-sectional research designs may not hold in longitudinal analyses (Maxwell, Cole, & Mitchell, 2011). Thus, the mediation effect between Antisocial Traits and Intent to Harm Victims should not be given an excessive amount of thought, as it could be a statistical artifact devoid of substantive meaning.

Finally, it is important to note that the authors (Beech & Ward, 2004) responsible for the etiological model of risk that inspired ours have since published articles that challenge the notion of dynamic risk factors, especially stable ones (e.g., Ward, 2016; Ward & Beech, 2015; Ward & Fortune, 2016). They argue that dynamic risk factors are composite constructs that offer poor/incomplete causal explanations of offending. In that sense, they are more suited to prediction than theory building or clinical intervention (Heffernan & Ward, 2015). Because stable risk factors are one of the core components of our model, the same criticisms probably apply. On that, we do not disagree: Our etiological model is quantitative and predictive, and, therefore, provides limited explanations of the process that leads to persistence in offending. Such an endeavor would probably be best served by qualitative research designs.

Conclusion

The present article described the development and validation of an etiological model of risk in sexual offenders, based on the theoretical framework of Beech and Ward (2004). By combining different areas of research, namely, developmental models of sexual coercion and studies about the dimensions of risk in sexual offenders, we were able to develop a coherent model of the criminal activity of sexual offenders. Compared with existing models, ours focused on persistence rather than onset, and encompassed both the sexual and nonsexual criminal activity of these offenders. It comprised two main pathways. The first was characterized by sexual victimization, social isolation, and early deviant sexual fantasies. It led to a prolific involvement in sexual criminality, especially toward children, and predicted sexual recidivism. The second pathway was characterized by externalization problems, sexual promiscuity, physical/psychological victimization, and was associated with nonsexual delinquency and serious sexual offenses directed (mostly) toward women. It predicted all types of recidivism.

Supplemental Material

Supplementary_Material_-_Review_2 – Supplemental material for The Etiology of Risk in Sexual Offenders: A Preliminary Model

Supplemental material, Supplementary_Material_-_Review_2 for The Etiology of Risk in Sexual Offenders: A Preliminary Model by Sébastien Brouillette-Alarie and Jean Proulx in Sexual Abuse: A Journal of Research and Treatment

Footnotes

Authors’ Note

The authors take responsibility for the integrity of the data and the accuracy of the data analyses, and have made every effort to avoid inflating statistically significant results.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplemental Material

Supplemental material for this article is available online.

References

Allen

B. P.

Pflugradt

D. M.

(2014). An exploration of the latent constructs of the STATIC-99. International Journal of Offender Therapy and Comparative Criminology, 58, 1376-1388. doi:10.1177/0306624X13496046

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Arlington, VA: American Psychiatric Publishing.

Barbaree

H. E.

Langton

C. M.

Peacock

E. J.

(2006). The factor structure of static actuarial items: Its relation to prediction. Sexual Abuse: A Journal of Research and Treatment, 18, 207-226. doi:10.s11194-006-9011-6

Beauregard

Lussier

Proulx

(2004). An exploration of developmental factors related to deviant sexual preferences among adult rapists. Sexual Abuse: A Journal of Research and Treatment, 16, 151-161. doi:10.1023/B: SEBU.0000023063.94781.bd

Beauregard

Proulx

Leclerc

(2014). Offending pathways: The role of lifestyle and precrime factors in extrafamilial child molesters. In Proulx

Beauregard

Lussier

Leclerc

(Eds.), Pathways to sexual aggression (pp. 137-155). Abingdon, UK: Routledge.

Beech

A. R.

Ward

(2004). The integration of etiology and risk in sexual offenders: A theoretical framework. Aggression and Violent Behavior, 10, 31-63. doi:10.1016/j.avb.2003.08.002

Brouillette-Alarie

Babchishin

K. M.

Hanson

R. K.

Helmus

L.-M.

(2016). Latent constructs of the Static-99R and Static-2002R: A three-factor solution. Assessment, 23, 96-111. doi:10.1177/1073191114568114

Brouillette-Alarie

Hanson

R. K.

(2015). Comparaison de deux mesures d’évaluation du risque de récidive des délinquants sexuels [Comparison of two measures of recidivism risk assessment of sexual offenders]. Canadian Journal of Behavioural Science, 47, 292-304. doi:10.1037/cbs0000019

Brouillette-Alarie

Hanson

R. K.

Babchishin

K. M.

Benbouriche

(2014). De la prédiction à la compréhension: Recension des dimensions psychologiques de la Statique-99 [From prediction to understanding: A literature review of the psychological dimensions of the Static-99]. Pratiques Psychologiques, 20, 1-19. doi:10.1016/j.prps.2013.12.001

10.

Brouillette-Alarie

Longpré

Proulx

(2014, October). Can developmental history be linked to risk-relevant propensities? Towards an etiological model of the recidivism of sex offenders. Poster session presented at the meeting of the Association for the Treatment of Sexual Abusers, San Diego, CA.

11.

Brouillette-Alarie

Proulx

(2013). Predictive validity of the Static-99R and its dimensions. Journal of Sexual Aggression, 19, 311-328. doi:10.1080/13552600.2012.747630

12.

Brouillette-Alarie

Proulx

Hanson

R. K.

(2017). Three central dimensions of sexual recidivism risk: Understanding the latent constructs of Static-99R and Static-2002R. Sexual Abuse: A Journal of Research and Treatment, 30, 676-704. doi:10.1177/1079063217691965

13.

Brown

(2006). Confirmatory factor analysis for applied research. New York, NY: Guilford Press.

14.

Carney

A. G.

Merrell

K. W.

(2001). Bullying in schools: Perspective on understanding and preventing an international problem. School Psychology International, 22, 364-382. doi:10.1177/0143034301223011

15.

Cattell

R. B.

Kline

(1977). The scientific analysis of personality and motivation. New York, NY: Academic Books.

16.

Ciardha

(2015). Experts in rape: Evaluating the evidence for a novice-to-expert continuum in the offense behavior and cognition of sexual offenders. Aggression and Violent Behavior, 20, 26-32. doi:10.1016/j.avb.2014.12.003

17.

Crick

N. R.

Grotpeter

J. K.

(1995). Relational aggression, gender, and social–psychological adjustment. Child Development, 66, 710-722. doi:10.1111/j.1467-8624.1995.tb00900.x

18.

Daversa

M. T.

Knight

R. A.

(2007). A structural examination of the predictors of sexual coercion against children in adolescent sexual offenders. Criminal Justice and Behavior, 34, 1313-1333. doi:10.1177/0093854807302411

19.

de Heer

. (2016). A snapshot of serial rape: An investigation of criminal sophistication and use of force on victim injury and severity of the assault. Journal of Interpersonal Violence, 31, 598-619. doi:10.1177/0886260514556110

20.

Doren

D. M.

(2004). Toward a multidimensional model for sexual recidivism risk. Journal of Interpersonal Violence, 19, 835-856. doi:10.1177/0886260504266882

21.

Douglas

K. S.

Skeem

J. L.

(2005). Violence risk assessment: Getting specific about being dynamic. Psychology, Public Policy, and Law, 11, 347-383. doi:10.1037/1076-8971.11.3.347

22.

Furnham

(1986). Response bias, social desirability and dissimulation. Personality and Individual Differences, 7, 385-400. doi:10.1016/0191-8869(86)90014-0

23.

Gadermann

A. M.

Guhn

Zumbo

B. D.

(2012). Estimating ordinal reliability for Likert-type and ordinal item response data: A conceptual, empirical, and practical guide. Practical Assessment, Research & Evaluation, 17(3), 1-13.

24.

Groth

A. N.

Hobson

W. F.

Gary

T. S.

(1982). The child molester: Clinical observations. Journal of Social Work and Human Sexuality, 1, 129-144. doi:10.1300/J291v01n01_08

25.

Grove

W. M.

Zald

D. H.

Lebow

B. S.

Snitz

B. E.

Nelson

(2000). Clinical versus mechanical prediction: A meta-analysis. Psychological Assessment, 12, 19-30. doi:10.1037/1040-3590.12.1.19

26.

Hanson

R. K.

(2009). The psychological assessment of risk for crime and violence. Canadian Psychology, 50, 172-182. doi:10.1037/a0015726

27.

Hanson

R. K.

Bussière

M. T.

(1998). Predicting relapse: A meta-analysis of sexual offender recidivism. Journal of Consulting and Clinical Psychology, 66, 348-362. doi:10.1037/0022-006X.66.2.348

28.

Hanson

R. K.

Harris

A. J. R.

Scott

Helmus

(2007). Assessing the risk of sexual offenders on community supervision: The dynamic supervision project (Publication No. 2007-05). Ottawa, Ontario: Public Safety Canada.

29.

Hanson

R. K.

Lloyd

C. D.

Helmus

Thornton

(2012). Developing non-arbitrary metrics for risk communication: Percentile ranks for the Static-99/R and Static-2002/R Sexual Offender Risk Scales. International Journal of Forensic Mental Health, 11, 9-23. doi:10.1080/14999013.2012.667511

30.

Hanson

R. K.

Morton-Bourgon

K. E.

(2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73, 1154-1163. doi:10.1037/0022-006X.73.6.1154

31.

Hanson

R. K.

Morton-Bourgon

K. E.

(2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 119 prediction studies. Psychological Assessment, 21, 1-21. doi:10.1037/a0014421

32.

Hanson

R. K.

Thornton

(2000). Improving risk assessment for sexual offenders: A comparison of three actuarial scales. Law and Human Behavior, 24, 119-136. doi:10.1023/A:1005482921333

33.

Hanson

R. K.

Thornton

(2003). Notes on the development of Static-2002 (Publication No. 2003-01). Ottawa, Ontario: Department of the Solicitor General of Canada.

34.

Hare

R. D.

(2003). Manual for the Revised Psychopathy Checklist (2nd ed.). Toronto, Ontario, Canada: Multi-Health Systems.

35.

Harris

D. A.

Knight

R. A.

Smallbone

Dennison

(2011). Postrelease specialization and versatility in sexual offenders referred for civil commitment. Sexual Abuse: A Journal of Research and Treatment, 23, 243-259. doi:10.1177/1079063210384276

36.

Heffernan

Ward

(2015). The conceptualization of dynamic risk factors in child sex offenders: An agency model. Aggression and Violent Behavior, 24, 250-260. doi:10.1016/j.avb.2015.07.001

37.

Helmus

Thornton

Hanson

R. K.

Babchishin

R. K.

(2012). Improving the predictive accuracy of Static-99 and Static-2002 with older sex offenders: Revised age weights. Sexual Abuse: A Journal of Research and Treatment, 24, 64-101. doi:10.1177/1079063211409951

38.

L.-t.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1-55. doi:10.1080/10705519909540118

39.

Kenny

D. A.

(2012). Identification. Retrieved from http://davidakenny.net/cm/identify_formal.htm

40.

Knight

R. A.

Prentky

R. A.

(1990). Classifying sexual offenders: The development and corroboration of taxonomic models. In Marshall

W. L.

Laws

D. R.

Barbaree

H. E.

(Eds.), Handbook of sexual assault: Issues, theories, and treatment of the offender (pp. 23-52). New York, NY: Plenum Press.

41.

Knight

R. A.

Sims-Knight

(2003). Developmental antecedents of sexual coercion against women: Testing of alternative hypotheses with structural equation modeling. In Prentky

R. A.

Janus

Seto

M. C.

(Eds.), Sexual coercion: Understanding and management (pp. 72-85). New York: New York Academy of Sciences.

42.

Knight

R. A.

Thornton

(2007). Evaluating and improving risk assessment schemes for sexual recidivism: A long-term follow-up of convicted sexual offenders (Publication No. 217618). Retrieved from https://www.ncjrs.gov/pdffiles1/nij/grants/217618.pdf

43.

Landis

J. R.

Koch

G. G.

(1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159-174. doi:10.2307/2529310

44.

Lee

J. K. P.

Jackson

H. J.

Pattison

Ward

(2002). Developmental risk factors for sexual offending. Child Abuse & Neglect, 26, 73-92. doi:10.1016/S0145-2134(01)00304-0

45.

Lussier

(2005). The criminal activity of sexual offenders in adulthood: Revisiting the specialization debate. Sexual Abuse: A Journal of Research and Treatment, 17, 269-292. doi:10.1007/s11194-005-5057-0

46.

Lussier

Beauregard

Proulx

Nicole

(2005). Developmental factors related to deviant sexual preferences in child molesters. Journal of Interpersonal Violence, 20, 999-1017. doi:10.1177/0886260505278024

47.

Lussier

Davies

(2011). A person-oriented perspective on sexual offenders, offending trajectories, and risk of recidivism: A new challenge for policymakers, risk assessors, and actuarial prediction? Psychology, Public Policy, and Law, 17, 530-561. doi:10.1037/a0024388

48.

Lussier

Leclerc

Cale

Proulx

(2007). Developmental pathways of deviance in sexual aggressors. Criminal Justice and Behavior, 34, 1441-1462. doi:10.1177/0093854807306350

49.

Lussier

Proulx

LeBlanc

(2005). Criminal propensity, deviant sexual interests and criminal activity of sexual aggressors against women: A comparison of explanatory models. Criminology, 43, 249-282. doi:10.1111/j.0011-1348.2005.00008.x

50.

Malamuth

N. M.

Sockloskie

R. J.

Koss

M. P.

Tanaka

J. S.

(1991). Characteristics of aggressors against women: Testing a model using a national sample of college students. Journal of Consulting and Clinical Psychology, 59, 670-681. doi:10.1037//0022-006X.59.5.670

51.

Mann

R. E.

Hanson

R. K.

Thornton

(2010). Assessing risk for sexual recidivism: Some proposals on the nature of psychologically meaningful risk factors. Sexual Abuse: A Journal of Research and Treatment, 22, 191-217. doi:10.1177/1079063210366039

52.

Maxwell

S. E.

Cole

D. A.

Mitchell

M. A.

(2011). Bias in cross-sectional analyses of longitudinal mediation: Partial and complete mediation under an autoregressive model. Multivariate Behavioral Research, 46, 816-841. doi:10.1080/00273171.2011.606716

53.

McGrath

Cumming

Burchard

Zeoli

Ellerby

(2010). Current practices and emerging trends in sexual abuser management: The Safer Society 2009 North American survey. Brandon, VT: Safer Society Press.

54.

Meehl

P. E.

(1954). Clinical versus statistical prediction: A theoretical analysis and a review of the evidence. Minneapolis: University of Minnesota.

55.

Moffitt

T. E.

(1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701. doi:10.1037/0033-295X.100.4.674

56.

Mokros

Schilling

Eher

Nitschke

(2012). The Severe Sexual Sadism Scale: Cross-validation and scale properties. Psychological Assessment, 24, 764-769. doi:10.1037/a0026419

57.

Mossman

(1994). Assessing predictions of violence: Being accurate about accuracy. Journal of Consulting and Clinical Psychology, 62, 783-792. doi:10.1037/0022-006X.62.4.783

58.

Muthén

(1983). Latent variable structural equation modeling with categorical data. Journal of Econometrics, 22, 43-65. doi:10.1016/0304-4076(83)90093-3

59.

Muthén

(1984). A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators. Psychometrika, 49, 115-132. doi:10.1007/BF02294210

60.

Muthén

L. K.

Muthén

B. O.

(2010). Mplus user’s guide. Los Angeles, CA: Author.

61.

Nansel

T. R.

Overpeck

Ramani

S. P.

Pilla

R. S.

Ruan

W. J.

Simons-Morton

Scheidt

(2001). Bullying behaviors among U.S. youth: Prevalence and association with psychosocial adjustment. Journal of the American Medical Association, 285, 2094-2100. doi:10.1001/jama.285.16.2094

62.

Natvig

G. K.

Albrektsen

Qvarnstrøm

(2001). School-related stress experience as a risk factor for bullying behavior. Journal of Youth and Adolescence, 30, 561-575. doi:10.1023/A:1010448604838

63.

Neal

T. M.

Grisso

(2014). Assessment practices and expert judgment methods in forensic psychology and psychiatry: An international snapshot. Criminal Justice and Behavior, 41, 1406-1421. doi:10.1177/0093854814548449

64.

Olver

M. E.

Klepfisz

Stockdale

K. C.

Kingston

D. A.

Nicholaichuk

T. P.

Wong

S. C. P.

(2016). Some notes on the validation of VRS-SO static scores. Journal of Sexual Aggression, 22, 147-160. doi:10.1080/13552600.2015.1116625

65.

Olver

M. E.

Wong

S. C. P.

Nicholaichuk

Gordon

(2007). The validity and reliability of the Violence Risk Scale-Sexual Offender Version: Assessing sex offender risk and evaluating therapeutic change. Psychological Assessment, 19, 318-329. doi:10.1037/1040-3590.19.3.318

66.

Parent

Guay

J. P.

Knight

R. A.

(2011). An assessment of long-term risk of recidivism by adult sex offenders: One size doesn’t fit all. Criminal Justice and Behavior, 38, 188-209. doi:10.1177/0093854810388238

67.

Pithers

W. D.

Marques

J. K.

Gibat

C. C.

Marlatt

G. A.

(1983). Relapse prevention: A self-control model of treatment and maintenance of change for sexual aggressives. In Greer

Stuart

(Eds.), The sexual aggressor: Current perspectives on treatment (pp. 214-239). New York, NY: Van Nostrand Reinhold.

68.

Prentky

R. A.

Knight

R. A.

Sims-Knight

J. E.

Straus

Rokous

Cerce

(1989). Developmental antecedents of sexual aggression. Development and Psychopathology, 1, 153-169. doi:10.1017/S0954579400000328

69.

Prentky

R. A.

Lee

A. F. S.

Knight

R. A.

Cerce

(1997). Recidivism rates among child molesters and rapists: A methodological analysis. Law and Human Behavior, 21, 635-659. doi:10.1023/A:1024860714738

70.

Proulx

Beauregard

Lussier

Leclerc

(2014). Pathways to sexual aggression. Abingdon, UK: Routledge.

71.

Proulx

Perreault

Ouimet

(1999). Pathways in the offending process of extrafamilial sexual child molesters. Sexual Abuse: A Journal of Research and Treatment, 11, 117-129. doi:10.1007/BF02658842

72.

Rigby

Slee

P. T.

(1993). Dimensions of interpersonal relation among Australian children and implications for psychological well-being. Journal of Social Psychology, 133, 33-42. doi:10.1080/00224545.1993.9712116

73.

Roberts

C. F.

Doren

D. M.

Thornton

(2002). Dimensions associated with assessments of sex offender recidivism risk. Criminal Justice and Behavior, 29, 569-589. doi:10.1177/009385402236733

74.

Roberts

S. J.

Glod

C. A.

Kim

Hounchell

(2010). Relationships between aggression, depression, and alcohol, tobacco: Implications for healthcare providers in student health. Journal of the American Academy of Nurse Practitioners, 22, 369-375. doi:10.1111/j.1745-7599.2010.00521.x

75.

Schumacker

R. E.

Lomax

R. G.

(2016). A beginner’s guide to structural equation modeling (4th ed.). New York, NY: Routledge.

76.

Seto

M. C.

(2005). Is more better? Combining Actuarial Risk Scales to predict recidivism among adult sex offenders. Psychological Assessment, 17, 156-167. doi:10.1037/1040-3590.17.2.156

77.

Slater

Woodhams

Hamilton-Giachritsis

(2014). Can serial rapists be distinguished from one-off rapists? Behavioral Sciences & the Law, 32, 220-239. doi:10.1002/bsl.2096

78.

St-Yves

Proulx

McKibben

(1994). Questionnaire informatisé sur la délinquance sexuelle [Computerized Sex Offender Questionnaire]. Unpublished manuscript.

79.

Tabachnick

B. G.

Fidell

L. S.

(2013). Using multivariate statistics (6th ed.). Boston, MA: Pearson Education.

80.

Ttofi

M. M.

Farrington

D. P.

Lösel

Loeber

(2011). The predictive efficiency of school bullying versus later offending: A systematic/meta-analytic review of longitudinal studies. Criminal Behaviour and Mental Health, 21, 80-89. doi:10.1002/cbm.808

81.

Ward

(2016). Dynamic risk factors: Scientific kinds or predictive constructs. Psychology, Crime & Law, 22, 2-16. doi:10.1080/1068316X.2015.1109094

82.

Ward

Beech

A. R.

(2015). Dynamic risk factors: A theoretical dead-end? Psychology, Crime & Law, 21, 100-113. doi:10.1080/1068316X.2014.917854

83.

Ward

Fortune

(2016). The role of dynamic risk factors in the explanation of offending. Aggression and Violent Behavior, 29, 79-88. doi:10.1016/j.avb.2016.06.007

84.

Ward

Louden

Hudson

S. M.

Marshall

W. L.

(1995). A descriptive model of the offense chain for child molesters. Journal of Interpersonal Violence, 10, 452-472. doi:10.1177/088626095010004005

85.

Widiger

T. A.

Costa

P. T.

(2013). Personality disorders and the five-factor model of personality (3rd ed.). Washington, DC: American Psychological Association.

86.

Wolke

Copeland

W. E.

Angold

Costello

E. J.

(2013). Impact of bullying in childhood on adult health, wealth, crime, and social outcomes. Psychological Science, 24, 1958-1970. doi:10.1177/0956797613481608

87.

Yessine

A. K.

Bonta

(2008). Pathways to serious offending (Publication No. 2008-01). Ottawa, Ontario: Public Safety Canada.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.06 MB