Do Psychological Features Distinguish Those Who Sexually Offend Against Different Age Groups From Those Who Are Stable in Victim Age?

Operationalizing Age Polymorphism

A challenge in research on victim age and age polymorphism is the banding of different age groups. There are differences across studies on the age ranges used to classify child and adult victims. Further compounding this issue is that researchers are often most interested in victim age as a proxy of sexual development (e.g., Stephens et al., 2018a). For example, researchers are often interested in whether a child victim physically resembles someone who is prepubescent due to its relevance to the diagnosis of pedophilic disorder. Fortunately, researchers have proposed age bands that roughly correspond to different stages of sexual development that could increase consistency across studies (e.g., Hames & Blanchard, 2012; Seto, 2017).

Most notably, Seto (2017) provided age ranges that roughly correspond to different stages of sexual development and sexual age interest: infants and toddlers (up to age 2), prepubescent children (3-10), pubescent children (11-14), postpubescent teenagers (15-16), young adults (18 to late 30s), middle aged adults (40s to late 50s), and older adults (60+). These age groups could be used in studies on age polymorphism; however, one must consider that the number of age groups could inflate the prevalence of age polymorphism. For example, if two age groups (child and adult) are included there is only one possibility for age polymorphism, whereas if three age groups (child, teen, and adult) are included there are two possibilities for age polymorphism. To date, most research has included two (child and adult) or three (child, teen, and adult) victim age groups and operationalized age polymorphism as crossover between any of these different age groups (e.g., Brown et al., 2015; Kleban et al., 2013; Porter et al., 2000).

The way previous studies have classified age polymorphism means that someone who offends against adjacent age groups, such as a prepubescent and pubescent child, and non-adjacent age groups, such as a prepubescent child and a young adult, would both be considered polymorphic. This is potentially problematic, as age polymorphism between adjacent age groups is expected for two reasons. First, there is greater similarity between adjacent than nonadjacent age groups. It is possible that men are offending against victims in adjacent age groups who physically resemble each other, even if they are classified in different age groups because age is an imperfect proxy of sexual development.

Furthermore, researchers can expect someone to have sexual interest in age groups that slightly deviate from one’s preferred target (Seto, 2017), which is largely ignored in research on age polymorphism with one exception (e.g., Saramago et al., 2020). We argue that nonadjacent age polymorphism is likely of greatest interest to researchers and clinicians because it increases the likelihood that victims are truly distinct in physical appearance and that the offending pattern is more erratic. Unsurprisingly, nonadjacent age polymorphism is relatively rare (25% of cases of polymorphism) compared with adjacent age polymorphism (75%; Saramago et al., 2020).

Although discussion of victim age classification may appear esoteric, it is important because it can influence findings on age polymorphism and its correlates. Overall, there is a need to better understand correlates of age polymorphism, as clinicians can be asked about the risk a person poses to someone that does not resemble previous victims. For example, child protection services may explicitly ask if someone with a young adult victim poses a risk to their young teenage daughter. Additionally, clinicians are often required to consider victimology as part of their risk formulation. Research on correlates of age polymorphism may result in increased confidence in these formulations. Lastly, it is possible that research on age polymorphism correlates may aid in the creation of a tool that identifies the likelihood of polymorphism. The present study examines correlates of age polymorphism and whether these correlates differ depending on the operationalization of polymorphism.

Correlates of Age Polymorphism

Present Study

The present study examines the association between age polymorphism and the following correlates: psychopathy, sexual preoccupation, multiple paraphilias, psychosis, and substance use disorders in a provincial sample of men who sexually offended. The study adds to the literature in several ways: First, we considered methodological issues that may have impacted previous findings on correlates of polymorphism by examining the impact of different operationalizations of age polymorphism (adjacent versus nonadjacent age polymorphism). It is possible that previous findings might be more evident using a stricter definition of age polymorphism that considers nonadjacent polymorphism. Second, many previous studies have not restricted their sample to men with two or more victims. This is important as differences between groups could be driven by the fact that men who are polymorphic have a greater number of victims (e.g., Stephens et al., 2018b). Lastly, we utilized a range of clinical data to examine several hypotheses of polymorphism in a single study to provide a more holistic examination, whereas previous studies have often examined correlates in isolation.

We hypothesized that men who were age polymorphic would have a higher mean psychopathy score compared with men who were stable in victim age. We also examined differences in PCL-R factor scores between groups but did not have a pre-specified hypothesis about what we would find. We also expected mean elevated scores on relevant personality measures in the age polymorphism group compared with those who were stable in victim age. It was also hypothesized that sexual preoccupation and multiple paraphilias would be more common in those who were age polymorphic compared with those who were stable in victim age. Lastly, we hypothesized that psychosis and substance use disorders would be more prevalent in those who were age polymorphic compared with those who were stable in victim age and that they would show elevations on relevant personality measure subscales. Across all analyses, we expected that these differences would be particularly pronounced in the nonadjacent polymorphism group compared with the adjacent polymorphism group.

Sample

The study was conducted using an archival database of adults whose sexual behavior crossed legal boundaries. All the men in the sample were referred to the Forensic Sexual Behaviour Program in Nova Scotia, Canada. The clinic provides assessment and treatment services to adults across the province. Adults were referred to the clinic for a variety of reasons including for the purposes of receiving an assessment prior to sentencing or as part of their probation or conditional sentencing order to determine risk and eligibility for an outpatient treatment program.

The archival database consisted of 1,359 adults over the age of 18 who sexually offended and were assessed between 1998 and 2018. The sample in the present study was reduced to 387 adult men who had sexually offended against two or more victims and were included in analyses. We excluded women, those with only one victim, and/or those with child pornography offences only. Sample characteristics can be found in Table 1.

Measures

Procedure

All men attended a two-day comprehensive assessment with a registered psychologist or psychology trainee who was under the supervision of a registered psychologist. The assessment consisted of a semi-structured interview (PCL-R interview supplemented with additional questions), pencil-and-paper testing (e.g., MCMI-III, MMPI-2), and a three-hour penile plethysmography assessment. As part of the assessment, men were scored on several sexual and violent risk assessment tools. The archival database also included clinician judgement based on the entirety of the information (e.g., mental health diagnoses, recommendations). All men underwent the same assessment procedure, which is standardized at the clinic. In the present study, we have only used the variables that are included in the database that are relevant to our hypotheses.

Information collected during the assessment was entered by psychologists on a data entry sheet, which was subsequently entered into the archival database by an assessment technician who maintained the database. The archival database used in the present study has predominantly been used for quality-control initiatives and to understand the needs of the population to adjust clinical practice where necessary. The archival database has not been previously included in published research studies. We received approval from the research ethics board at the hospital to use the archival database for the present study.

Planned Analyses

To examine the difference in mean PCL-R, MMPI-2, and MCMI-III scores, we utilized univariate ANOVAs with post hoc Bonferroni tests. To examine the frequencies of sexual preoccupation, multiple paraphilias, and substance use disorders, we utilized chi square tests. Effect sizes are reported for all analyses.

Psychopathy

Despite our hypothesis, there was no significant mean group difference on the PCL-R total score, F(5, 230) = 0.73, p = .604, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .02 (see Table 3 for mean PCL-R total, Factor 1, and Factor 2 scores). The effect sizes were negligible to small when the adjacent age polymorphism group was compared with those who solely offended against victims 5 and under (d = 0.32), 6-11 (d = 0.02), 12-15 (d = 0.22), 16 or older (d = 0.11), and the nonadjacent age polymorphism group (d = 0.35). As expected, the nonadjacent age polymorphism group had the highest mean PCL-R score; however, effect sizes were negligible when compared with those who solely offended against victims 5 and under (d = 0.04) and 12-15 (d = 0.13). The effect sizes were small for the comparison of the nonadjacent age polymorphism group with those who solely offended against victims 6-11 (d = 0.34) and 16 or older (d = 0.24).

There were no significant group mean differences for Factor 1 PCL-R scores, F(5, 306) = 2.15, p = .060, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .03. Despite this finding, there was a medium effect size when comparing the adjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.59). Remaining effect sizes were negligible to small when the adjacent polymorphism group was compared with men who solely offended against victims 6-11 (d = 0.05), 12-15 (d = 0.28), 16 or older (d = 0.03), and the nonadjacent age polymorphism group (d = 0.24). Similarly, effect sizes were negligible or small for the comparison of the nonadjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.35), 6-11 (d = 0.29), 12-15 (d = 0.05), and 16 or older (d = 0.29).

There were no significant mean group differences on Factor 2 PCL-R scores, F(5, 235) = 0.54, p = .748, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .01. Effect sizes were negligible or small when comparing the adjacent polymorphism group with those who solely offended against victims 5 and under (d = 0.26), 6-11 (d = 0.002), 12-15 (d = 0.20), 16 or older (d = 0.18), and the nonadjacent age polymorphism group (d = 0.23). The effect sizes were in the negligible or small range for the comparison of the nonadjacent age polymorphism group with those who offended against victims 5 and under (d = 0.01), 6-11 (d = 0.24), 12-15 (d = 0.04), and 16 or older (d = 0.06).

We also conducted supplementary analyses with relevant personality measure subscales (see Table 3). There were no group differences on the MMPI-2 Psychopathic Deviate scale, F(5, 285) = 1.60, p = .161, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .03. Effect sizes were negligible to small when comparing the adjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.47), 6-11 (d = 0.09), 12-15 (d = 0.02), 16 or older (d = 0.20), and the nonadjacent age polymorphism group (d = 0.09). The effect sizes were negligible or small for the comparison of the nonadjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.43), 6-11 (d = 0.002), 12-15 (d = 0.08), and 16 or older (d = 0.08).

Similarly, there were no group differences on the MCMI-III Antisocial subscale, F(5, 279) = 1.28, p = .273, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .02 (see Table 3). The effect sizes were negligible or small for the comparison of the men in the adjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.01), 6-11 (d = 0.29), 12-15 (d = 0.23), 16 or older (d = 0.13), and the nonadjacent age polymorphism group (d = 0.26). Effect sizes were similar when comparing men in the nonadjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.24), 6-11 (d = 0.02), 12-15 (d = 0.01), and 16 or older (d = 0.39).

Atypical Sexual Interests

Despite our hypotheses, the results were not significant for sexual preoccupation, χ²(5) = 5.11, p = .403, V = .13, or multiple paraphilias, χ²(10) = 3.25, p = .975, V = .07. The percentages and standardized residuals are depicted in Table 4.

Psychosis and Substance Use Disorders

Due to the small number of individuals in the sample (n = 3) with a psychotic spectrum disorder, we were unable to examine this variable as originally planned. As such, we relied on the MMPI-2 Schizophrenia scale to examine the association between age polymorphism and psychosis, F(5, 285) = 1.67, p = .143, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .03. Effect sizes were negligible or small when comparing the adjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.43), 6-11 (d = 0.13), 12-15 (d = 0.21), 16 or older (d = 0.06), and the nonadjacent age polymorphism group (d = 0.26). Effect sizes were negligible or small for the comparison of the nonadjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.14), 6-11 (d = 0.16), 16 or older (d = 0.33), except for the comparison to those who solely offended against victims 12-15 (d = 0.53).

We next examined the presence of substance use disorders, χ²(5) = 10.04, p = .074, V = .16 (see Table 4). We supplemented these analyses by examining the mean difference on the MCMI-III Alcohol and Drug subscales. For the Alcohol subscale, there were no significant group differences, F(5, 277) = 1.16, p = .329, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .02 (see Table 3). The effect sizes were negligible or small when comparing the adjacent age polymorphism group with those who solely offended against victims 5 and under (d = 0.32), 6-11 (d = 0.14), 12-15 (d = 0.04), 16 or older (d = 0.32), and the nonadjacent age polymorphism group (d = 0.48). The nonadjacent age polymorphism group had the highest mean score on the Alcohol subscale, but the effect sizes were negligible when compared with the mean score for those who solely offended against victims 5 and under (d = 0.14) and 16 or older (d = 0.15). The effect sizes were small for the comparison of the nonadjacent age polymorphism group with those who solely offended against victims 6-11 (d = 0.27) and 12-15 (d = 0.39).

For the MCMI III - Drug subscale, there was a significant group difference, F(5, 279) = 3.10, p = .010, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .05 (see Table 3). Bonferroni post hoc analyses indicated a significant mean difference such that those who solely offended against victims 16 or older had higher mean scores than those who solely offended against victims 6-11 (p = .019, d = 0.57). None of the other post hoc comparisons were significant. Effect sizes were negligible or small when comparing the adjacent age polymorphism group with those who only offended against victims 5 and under (d = 0.32), 6-11 (d = 0.10), 12-15 (d = 0.02), and the men in the nonadjacent age polymorphism group (d = 0.43). There was a medium sized effect when comparing the adjacent age polymorphism group to those who solely offended against victims 16 or older (d = 0.53). The effect sizes were negligible or small when the non-adjacent age polymorphism group was compared with those who solely offended against victims 5 and under (d = 0.10), 6-11 (d = 0.47), 12-15 (d = 0.39), and 16 or older (d = 0.08).

Explanations for Present Findings

There are several possible reasons for our present findings. The first set of considerations involve methodological differences compared with previous studies. The present study limited the sample to those with two or more victims, because men who are polymorphic have two or more victims by definition and tend to have a higher number of victims than those who are stable in victim age selection (e.g., Parent et al., 2011; Stephens et al., 2017; Stephens et al., 2018b). Many of the previous studies have not limited their examination to an equivalent comparison group of men with two or more victims (e.g., Guay et al., 2001; Kleban et al., 2013; Olver & Wong, 2006), despite this potential confound (Stephens et al., 2018b). This may be of relevance to the present findings because it is plausible that some of these correlates may be elevated in men with two or more victims. For example, men with multiple paraphilic interests are more prolific in their sexual offending and would be expected to have a greater number of victims (Laws, 1994). As a result, it is possible that differences among those who are polymorphic could be accounted for by the number of victims, as opposed to differences in victim age selection.

Another methodological consideration is that the victim age bands used in the present study differ from past studies. Most of the research on age polymorphism and psychopathy has grouped men into those who offended against children and adults (age polymorphism defined as crossover between children and adults; e.g., Brown et al., 2015; Olver & Wong, 2006; Porter et al., 2000). We attempted to use the age bands that most closely matched those provided by Seto (2017), which resulted in a greater number of victim age groups than previous studies (e.g., Olver & Wong, 2006; Porter et al., 2000; Stephens et al., 2018b).

Additionally, researchers have started to consider the difference between adjacent and nonadjacent age polymorphism, which has not been widely considered apart from a recent study that found nonadjacent polymorphism is less common than adjacent polymorphism (Saramago et al., 2020). It is possible that the way we operationalized polymorphism impacted our results.

An alternative possibility is that there may not be differences between men who are age polymorphic and those who are stable in victim age selection, though future studies are needed before reaching a definitive conclusion. If the body of evidence suggests no differences when confounds are controlled for, the implications of this are significant, given that clinicians are often asked to opine on risk to victims that differ from the index offence as part of their risk formulation. Nonetheless, it is possible that there are other correlates that would distinguish men who are polymorphic from those who are stable in victim age selection that have not been identified. For example, men who are polymorphic may have less exclusive age interests (i.e., sexual interest spanning multiple age groups), which translates into their offending. Additionally, past offending history may be stronger predictors of polymorphism than psychological features. For example, those who are polymorphic had a greater number of victims (e.g., Parent et al., 2011; Stephens et al., 2017) and adolescent victims (e.g., Guay et al., 2001; Lussier et al., 2007; Stephens et al., 2017). Thus, specific features of men’s offending history may be of greater value than psychological factors when clinicians need to determine the risk men pose to victims that differ from their index offence.

Limitations and Strengths

A significant limitation in the present study is that we may not have had a complete picture of the offending history for those in our sample. Although we considered previous and index victims, whereas some previous studies have only considered index offence information (e.g., Cann et al., 2007; Saramago et al., 2020), this did not provide a fulsome picture of their offending history. In our sample, 39.5% of men admitted to additional sexual offences that were undetected; however, the age of these victims were not included in the archival database. Further, it is possible that evaluators were unaware of some victim information or that men did not disclose their complete offending history. These issues are not unique to our study and are present in other studies that have examined victim age in sexual offending.

Additionally, our sample size for some of the groups were relatively small. This is most notable for the men in our main group of interest, the nonadjacent age polymorphism group, as well as those who offended against victims 5 and under. Furthermore, we were unable to conduct analyses for psychotic spectrum disorders due to the small number of men in the sample who were diagnosed with such a disorder. Lastly, we did not have interrater reliability for our variables of interest.

Despite these limitations, there are several strengths in the present study. More specifically, this is the first study to examine various correlates of age polymorphism in a single provincial sample using a comprehensive clinical database. The archival database in the present study was optimal in that it included fulsome assessment information that considered clinician diagnoses and several well-established measures including the PCL-R, MMPI-2, and MCMI-III. For many of our correlates, we were able to include multiple measures that spanned the range from clinician judgement to more standardized measures.

Future Research Directions

There is a need for future research that synthesizes past results to establish the prevalence of age polymorphism and how different methodological considerations impact prevalence estimates. Second, there is a need for prospective research on age polymorphism to better understand its correlates. Optimally these studies would use a large and diverse sample and collect data from a carefully matched group of men who exhibit age polymorphism with men who are stable in victim age. The researcher could utilize a comprehensive battery to assess different correlates that they have selected based on a review of the literature. Further, these men could be followed over time in a study on recidivism to examine any subsequent age polymorphism. This type of research may enable us to develop a model to predict subsequent age polymorphism, which would have significant implications for clinicians who must consider risk to victims that differ from the index offence as part of their risk formulation.