Fantasy-Driven Versus Contact-Driven Users of Child Sexual Exploitation Material: Offender Classification and Implications for Their Risk Assessment

Abstract

Since the advent of the Internet, convictions for the possession, display, trading, and distribution of child sexual exploitation material (CSEM) have risen steadily, but little is known about their appropriate assessment and treatment, especially concerning their risk of reoffending. It has been suggested that a conceptual distinction of fantasy- versus contact-driven CSEM users might be of merit. Sixty-eight offenders recruited from sex offender treatment providers were assessed via an anonymous computer survey including a variety of clinical and risk-related variables; the findings showed differences in the psychological profiles between CSEM users and contact child sex offenders. Numerical and spatial methods of data analysis were used to identify subgroups of CSEM users; these confirmed the twofold distinction of fantasy- versus contact-driven offending. The spatial representation of participants identified three dimensions as crucial in the classification of these subgroups: direct sexual contact with a minor, possession of fantasy-generating material, and social contact with other users with a sexual interest in minors; potentially differentiating distinct offender subgroups with different risks and needs. The current study informed the development of an empirical model of CSEM users that could aid in the assessment of risk of reoffending and cross-over to contact sex offending.

Keywords

child pornography child sexual abuse cognitive distortions sexual abuse sexual offender child sexual exploitation material

Convictions for the possession, exchange, and/or production of child pornography, more accurately referred to as child sexual exploitation material (CSEM), have risen considerably; in the last year alone, the UK Crown Prosecution Service (2014) commenced 20,373 prosecutions for child abuse image offenses. Consequently, police forces, the courts, and prison/probation services are required to manage growing numbers of CSEM users despite little empirical data to aid decision making, particularly in regard to risk prioritization and risk management.

Standardized measures for contact child sex offenders (CSOs) have been developed (e.g., the Static-2002, Hanson & Thornton, 2003; the Risk Matrix 2000, Thornton et al., 2003) that facilitate risk classification according to the statistical likelihood of recidivism. However, these tools have not yet been successfully validated for non-contact CSEM populations and are generally poor predictors of actual recidivism within this group (Henshaw, Ogloff, & Clough, 2017; Middleton, Beech, & Mandeville-Norden, 2005; Middleton, Mandeville-Norden, & Hayes, 2009; Osborn, Elliott, Middleton, & Beech, 2010; Wakeling, Howard, & Barnett, 2011; Webb, Craissati, & Keen, 2007). For example, Osborn et al. (2010) demonstrated that the Risk Matrix 2000 and the Static-99 (Hanson & Thornton, 2000) considerably overestimated the risk of sexual reoffending among a sample of CSEMOs (offenders who use child sexual exploitation material; n = 73); even though none had reoffended within a follow-up period of 1.5 years to 4 years, all the individuals were classified as having elevated risk profiles using these measures. However, when adjustments were made to the scale (e.g., omitting the items stranger victim and non-contact offending), 72.6% were subsequently classified as low risk.

Inappropriate use of such measures thus presents both ethical and economic challenges: For example, professional ethical guidelines for psychologists require the use of “assessment instruments whose validity and reliability have been established for use with members of the population tested” (Section 9.02b; American Psychological Association, 2010). Moreover, risk misclassification can have serious implications for the individual (e.g., sentencing length, parental access), broader society (e.g., evidence-based public protection, social reintegration), and allocation of scarce rehabilitation resources (e.g., treatment provision). Consequently, use of such risk assessment tools with CSEMOs would require for parity between the qualitative and quantitative risk factors of CSOs and CSEMOs to be established.

However, CSEMOs and CSOs have consistently been found to show significant differences across a number of potentially risk-relevant variables; for example, CSEMOs score higher on measures of sexual deviance, report lower endorsement of cognitions relating to children and sex, and are less likely to have a history of offending behavior (Babchishin, Hanson, & Hermann, 2010; Babchishin, Hanson, & VanZuylen, 2015; Elliott, Beech, & Mandeville-Norden, 2012; Elliott, Beech, Mandeville-Norden, & Hayes, 2009; Webb et al., 2007). Indeed, many of the proxy-measures considered informative of risk, such as potentially offense-supportive cognitions, were originally developed for CSOs, and their suitability and risk relevance for CSEMOs has not been established (e.g., Howitt & Sheldon, 2007; Merdian, Curtis, Thakker, Wilson, & Boer, 2014). In addition, a reliance on conventional and established risk assessment tools may neglect key differences between the topography and function of contact versus non-contact offending, and the potential predictive utility of offense characteristics specific to CSEM offenses (such as the content, size, or level of engagement with the CSEM material; Glasgow, 2010; Osborn et al., 2010; Seto & Eke, 2015; Taylor, Holland, & Quayle, 2001). Finally, the primary focus of CSEM risk formulation is arguably not an individual’s risk of reoffending (e.g., committing a similar CSEM offense) but their risk of escalation to a direct contact sexual offense.

The above considerations support the need for accurate and comprehensive risk assessment tools for CSEMOs, based on an empirical understanding of the specific risks and needs of this offender group. Whereas initial developments in this area focused on aiding the police in CSEM case prioritization during the investigatory process (i.e., the Kent Internet Risk Assessment Tool; Long, Alison, & McManus, 2013), more recently, Seto and Eke (2015) published the Child Pornography Offender Risk Tool (C-PORT), a first attempt to develop an actuarial risk assessment tool for the prediction of reoffending among CSEMOs. However, similar to findings derived from existing risk measures, the C-PORT only demonstrated predictive accuracy for those offenders who had CSEM offenses combined with other offenses (including contact sex offenses), but was not able to significantly predict reoffending for exclusive CSEM offenders.

Taken together, these findings confirm the lack of parity between the offending and risk profiles of CSEMOs and CSOs, and, given the poor predictive validity of existing risk measures for this population, challenge the direct relationship between CSEM viewing and committing a contact sex offense against a minor. Indeed, in two comprehensive meta-analyses (Hanson & Babchishin, 2009; Seto, Hanson, & Babchishin, 2011), 12.2% of convicted CSEMOs were found to have a history of contact sex offending (not exclusively child victims), increasing to 17.3% when self-report data were included.¹ Prospectively, CSEMOs have also been shown to be less likely to recidivate with a contact sex offense, with Seto et al. (2011) identifying contact sex offense recidivism rates of 2% (1.5 to 6-year follow-up) among a sample of 2,630 online offenders. In contrast, follow-up studies on CSOs have reported recidivism rates of 11% to 20% (e.g., Hanson & Morton-Bourgon, 2009).

These data therefore suggest a conceptual distinction between at least two types of CSEM users: those whose usage appears to be confined to engagement with the images themselves (e.g., as a facilitator of sexual fantasy or collecting behavior) and those whose CSEM usage is functionally and directly related to contact sex offending (e.g., as a behavioral substitute, facilitatory factor, or product of a contact offense). These groups will subsequently be labeled as contact-driven and fantasy-driven, respectively, utilizing terminology introduced by Merdian, Curtis, Thakker, Wilson, and Boer (2013). In their conceptual model, Merdian et al. suggest these two groups present with distinct offending profiles and motivations, with differential criminogenic and treatment needs, and with the contact-driven subgroup potentially more closely resembling CSO populations. Consequently, established risk assessment and treatment methods may have more applicability for the contact-driven offender subgroup, but are likely to be of limited value for fantasy-driven offenders. However, to date, Merdian et al.’s model, and its implications for risk management, has not been empirically validated. Further research is required both to establish the veracity of the model and to define the nature of the risks and needs presented by the proposed subgroups therein. If confirmed, such a typology may act as a starting point for directing future research into more comprehensive and appropriate risk assessment tools for this heterogeneous population.

The current study represents an important first step by aiming to enhance empirical knowledge of the potentially different CSEM pathways (contact- vs. fantasy-driven). Following the above considerations, the study had two primary research goals: (a) to establish whether the offending profile of CSEMOs is demonstrably distinct to CSOs, and (b) to investigate whether discrete subgroups of CSEM users can be empirically differentiated. In the event that discrete groups could be identified, a secondary research aim was to explore subgroup membership in relation to conventional predictors of (contact) sex offending.

Method

Participants

Participants (adult males) were recruited from both community sex offender treatment centers and prison settings in New Zealand. Participants were eligible to take part in this study if they (a) had a history (conviction) and/or an interest (self-referral to treatment) in either sexual contact with a minor, and/or possession, distribution, or production of CSEM; (b) had a sufficient ability to read and write in English; (c) had no reported intellectual disability or mental health concerns; and (d) provided consent to participate. Participants were initially approached by professionals within their respective organizations; consequently, no information is available on the representativeness of this sample. Overall, 77 individuals participated in this study, nine of whom were removed due to large amounts of incomplete data or responses that indicated nonengagement with the test material. The final sample consisted of 68 participants; participants self-identified as CSEMOs (n = 22), as CSOs (n = 29), or as mixed offenders (MOs, individuals with both offense types; n = 17), based on two screening questions: (a) “As an adult, have you ever had sexual contact with a person younger than 16 years?” and (b) “Have you ever seen pornography that showed children under 18 years?” Demographic and offense characteristics of the study sample are depicted in Table 1; details of this sample have been described elsewhere (Merdian et al., 2014).

Table 1.

Characteristics of the Offender Samples.

	Total	CSEMOs	CSOs	MOs	Significance differences
	n = 68	n = 22	n = 29	n = 17	Significance differences
Demographics
Age (years)
M	43.43	41.82	41.29	45.56
SD	13.2	14.5	7.86	13.2
Mdn	42.5	39.5	42	44.5
Ethnicity
NZ Europe	57.35	77.27	41.38	58.82	CSOs > CSEMOs
Maori	27.94	4.55	44.83	29.41
Pacific Islander	1.47	4.55
Other	7.35	9.09	6.90	5.88
Education (years)
M	9.69	11.62	7.87	9.69	CSEMOs > CSOs
SD	5.29	5.18	5.12	4.09
Mdn	10	12	8	9.5
Income (NZD)^a
M	34,414	37,565	22,248	49,454	CSEMOs > CSOs, MOs > CSOs
SD	23,588	15,222	19,419	30,982
Mdn	31,000	36,000	19,000	50,000
Unemployed	16.18	9.09	24.14	11.76
Own business	23.53	27.27	20.69	23.53
Relationship status and sexual preference
Sexual preference
Females	73.53	86.36	72.41	58.82
Males	16.18	9.09	20.69	17.65
Both	10.29	4.55	6.9	23.53
Current partner
Sexual	30.88	36.36	20.69	41.18
Live-in	25	31.82	13.79	35.29
Stable partner
None	27.94	36.36	27.39	17.65
1	23.53	18.18	24.14	29.41
2	20.59	18.18	20.69	23.53
3+	27.94	27.27	27.59	29.41
Own children	52.94	40.91	51.72	70.59
Criminal activities
In prison	54.41	4.55	89.66	58.82	CSOs > CSEMOs
Sex offending
Adult victim
Current	10.29		20.69	5.88
Previous	5.88		10.34	5.88
>1	5.88		10.34	5.88
Minor victim
Current	54.41		89.66	64.71
Previous	35.29		58.62	41.18
>1	41.18		55.17	70.59

Source. This table has previously been published in Merdian, Curtis, Thakker, Wilson, and Boer (2014, p. 5).

Note. Figures denote percentage scores if not labeled otherwise. CSEMO = Offenders who use child sexual exploitation material; CSO = Contact sex offenders with child victims; MOs = Mixed offenders (both offense types). Inferential statistics conducted with an adjusted alpha of p < .025 (only immediate comparisons conducted to reduce total number of total comparisons).

Outliers removed.

Procedure

An anonymous computer survey was designed for the purpose of this study, assessing different areas of clinical interest and potential risk relevance, such as lifestyle, criminal history, offense details (e.g., victim characteristics), and psychological variables (e.g., impulsivity, thinking styles). The survey items were developed in three stages: (a) item construction based on a systematic review of the literature, (b) expert validation, and (c) a series of pilot studies to finalize the items. The final item pool consisted of 211 items, grouped into seven subsections: (a) personal life, (b) work and spare time activities, (c) Internet behavior, (d) general offending behavior, (e) content of CSEM, (f) engagement with CSEM, and (g) cognitive distortions relating to children and sex. The majority of items were dichotomized with the exception of the cognitive distortions items (n = 39), which were scored on a 5-point Likert-type scale (1= strongly disagree; 5 = strongly agree). The cognitive distortion items consisted of the Abel and Becker Cognition Scale (Abel et al., 1984) and incorporated items from the Children and Sexual Activities Inventory (Howitt & Sheldon, 2007; see Merdian et al., 2014, for more details on these tools).

Each participant completed the survey unaided using a portable computer, with full instructions appearing on screen. Supervisory and/or research staff were unable to observe participant responses but were available to answer questions if requested. Participants completed the study in a private room, either individually or in small groups, dependent on security arrangements.

Results

Differences in Offender Profiles

Analysis

As a large number of variables were tested on a comparably small sample of participants, methods of dimension reduction were used to limit the number of multivariate comparisons between the subgroups, namely, cluster analysis (CA; binary variables) and principal components analysis (PCA; Likert-type scaled variables). Only the items responded to by all participants were included in this analysis (i.e., 70 binary variables, 39 Likert-type scaled variables; excluding items on CSEM usage). The number of binary variables was reduced through a preliminary screening of tetrachoric correlations², thus merging items that were highly inter-correlated (r_tets > .70, indicating redundancy) and removing items that were weakly correlated with other items (r_tets < .30). This screening process resulted in 25 binary variables remaining and ensured that subsequent CA met conventional recommendations regarding the minimum participant-to-variable ratio (2:1; Formann, 1984). Multinomial regression analysis was used to weight the resulting item clusters in terms of their associations with offender subtypes, and examine their ability to discriminate or account for these subtypes. In the absence of a strong conceptual basis for variable selection and in view of the small sample size available, a stepwise approach was applied to variable selection; although not a preferred option when conducting theory-driven analyses, it was considered useful for preliminary identification of potential discriminating factors (i.e., exploratory analysis) while avoiding model overfitting.

Results

Hierarchical CA (average linkage method) on the tetrachoric correlation matrix for binary variables resulted in six item clusters; this cluster solution was validated by re-clustering three quarters of randomly selected cases (replicating a six-cluster solution, with 68% identical classification of variables) and administering CA with a fixed cluster solution (100% identical classification of variables). The results of the PCA on the cognitive distortion items have been reported in Merdian et al. (2014); in short, PCA revealed six item components and one outlier item. Table 2 shows the resulting item groupings and highlights significant differences between the offender subtypes.

Table 2.

Six-Cluster Solution Resulting From Hierarchical Cluster Analysis on Binary Variables and Significant Differences Between Offender Subtypes.

Item group	(Selected) items	CSEMO–CSO	CSEMO–MO	CSO–MO
Cluster 1: Social Exclusion and Escape	Being bullied in childhoodInterest in fantasy and science fictionInterest in second-life or third-person gamesViewing of deviant pornographyIllegal downloading“Internet addiction”	>		<
Cluster 2: Alienation and Sensitivity	Different self-image from other peopleSelf-harm in childhoodFind it easy to lieBeing irritable and aggressiveRegretful
Cluster 3: Intimacy Deficits	Struggled to find partnerNever been in a relationship
Cluster 4: Exposure to Adversity	Physically abused as childSexually abused as childRule-breaking as childImpulsiveAntisocialTransient
Cluster 5: Adjustment	Cope well with stressOwn childrenHave engaged in sex tourism
Cluster 6: Resourced	Advanced Information Technology equipmentCurrently in sexual relationship
Component 1: Sexual Objectification of Children	E.g., most children 13 years (or younger) would enjoy having sex with an adult, and it wouldn’t harm the child in the future.
Component 2: Justification	E.g., A man is justified in having sex with his children or step-children, if his wife doesn’t like sex.	<	<
Component 3: Children as Sexual Agents	E.g., When a young child asks an adult about sex, it means she (he) wants to see the adult’s sex organs or have sex with the adult.	<	<
Component 4: Denial of Sex Offender Status	E.g., An adult can tell whether having sex with a young child will emotionally damage the child in the future.
Component 5: Emphasis on Cognitive Elements	E.g., If a person is attracted to sex with children, he (she) should solve that problem themselves and not talk to professionals.
Component 6: Power and Entitlement	E.g., A person should have sex whenever it is needed.	<	<
(Component 7):	My daughter (son) or other young child knows that I will still love her (him) even if she (he) refuses to be sexual with me.

Note. Cluster labels are the result of a discussion between the researcher and two independent raters, that is, a layperson and a researcher experienced in the area of sexual crimes. >, <: significant difference. Inferential statistics conducted using Kruskal–Wallis tests and Mann–Whitney U tests with an adjusted alpha of p < .025 (only immediate comparisons conducted to reduce total number of comparisons). CSEMO = offenders who use child sexual exploitation material; CSO = contact sex offenders; MOs = mixed offenders.

Overall, significant differences between offender types were found on four item groupings: Social Exclusion and Escape (Cluster 1), Justification (Component 2), Children as Sexual Agents (Component 3), and Power and Entitlement (Component 6).

Employing a stepwise model (forward entry) on the main effects of the variable sum scores resulted in successful classification of 66% of participants (68% of CSEMOs, 86% of CSOs, and 29% of MOs correctly classified), based on three predictors: Social Exclusion and Escape (Cluster 1), Exposure to Adversity (Cluster 4), and Justification (Component 2). Notably, Exposure to Adversity emerged in this model despite nonsignificant group comparisons on this variable (p = .056), suggesting that the discriminatory contribution of this variable is more clearly seen when modeled in combination with (as an adjunct to) Social Exclusion and Escape and Justification (i.e., Exposure to Adversity explains significant variability in classification after accounting for contributions of the other explanatory variables). To specify, CSEMOs were more likely than CSOs to report social exclusion and escapism (including immersion in online and fantasy activities), odds ratio (OR) = 2.88, Wald χ²(1) = 14.69, p < .001. Similarly, MOs were also more likely to report Social Exclusion and Escapism than CSOs, OR = 1.62, Wald χ²(1) = 4.63, p = .031; although less likely than CSEMOs, OR = 0.56, Wald χ²(1) = 4.80, p = .029. In comparison with CSEMOs, both CSOs, OR = 1.84, Wald χ²(1) = 6.44, p = .011, and MOs, OR = 1.72, Wald χ²(1) = 4.64, p = .031, were more likely to report exposure to adversity. Finally, MOs were more likely than CSEMOs to justify their sexual behavior, OR = 1.50, Wald χ²(1) = 8.33, p = .004. The model revealed no significant outliers or influential cases and reached large effect sizes, R² = .45 (Cox & Snell), R² = .51 (Nagelkerke).

Summary

Group comparisons between offender subtypes resulted in a number of significant differences; online offenders were found to have stronger immersion and emotional significance related to their online activities in comparison with contact sex offenders. Furthermore, CSEMOs appeared to have stronger internal inhibitions toward contact offending: both MOs and CSOs reported significantly higher cognitions relating to (a) justifications of their sexual behaviors, (b) endorsement of children as sexual agents, and (c) sexual entitlement. These results are in line with the findings reported in the analysis. However, only Justification and Social Exclusion and Escape remained as significant predictors of offender subtypes, alongside Exposure to Adversity. In summary, these findings support the first hypothesis that CSEMOs have a psychological profile distinct to CSOs, potentially indicative of differing risk and treatment needs between individuals with and without contact sex offenses. The identified differences between MOs and CSEMOs further highlight the heterogeneity within the group of CSEM users, explored further below.

Classification of CSEM Subgroups

Analysis

It was examined whether different subgroups of CSEM users could be identified beyond the above distinction of individuals with and without contact sex offenses (MOs vs. CSEMOs). A number of additional items regarding details of the participants’ CSEM usage and general online activities were available for this analysis. Overall, 67 binary variables were used after exclusion of items with low levels of endorsement or low inter-correlations (as defined previously). Participants were grouped based on similarities and differences in their responses to these variables, and classification was achieved via a two-step process. In Step 1, Multidimensional Scaling (MDS; using the squared Euclidean distance measure for binary data) was conducted to represent the data in a small number of dimensions, reducing participants’ responses across 67 variables to a set of coordinates that describe the relative distances between participants. These MDS dimensions were further explored following regression-informed analyses proposed by Kruskal and Wish (1978) and Everitt and Rabe-Heskett (1997). In Step 2, hierarchical CA was used to identify groups of participants according to their proximity on the MDS-identified dimensions. Thus, CA was applied to a meaningfully reduced data set (facilitating interpretation of participant groupings) and an adequate participant-to-variable ratio was assured.

Results

MDS pointed to a fair fit of a three-dimensional solution (Stress S = .13248, Stress and squared correlation RSQ = .91838), suggesting that the participants can be validly mapped in a three-dimensional space. To explore the meaning of this space, content-based categories were identified and regressed over the three MDS dimensions, resulting in a best fit of Peer Networking (i.e., social contact with other adults with a sexual interest in minors) for Dimension 1 (β = .905), Contact Offending (i.e., having had or attempted to have sexual contact with a minor) for Dimension 2 (β = −.568), and Fantasy-Based Material (i.e., possession of other child-related material, such as cartoons/narratives describing child sexual contact, pictures from commercial clothing catalogs) for Dimension 3 (β = −.422). The MDS dimension axes lose orthogonality if rotated according to their β-weighting, which indicates inter-correlation between the MDS dimensions and suggests the existence of stronger distinguishers than the suggested item categories. Hierarchical CA (Ward’s method) was subsequently applied to cluster participants according to their profile of scores across the three MDS dimensions. The analysis identified three discrete groups; Figure 2 depicts the position of these participant groupings with respect to the MDS dimension axes.

Figure 1.

The three dimensions of child pornography offending.

Figure 2.

Two-dimensional MDS maps depicting the hierarchical cluster structure of offender classification (Ward’s method).

The two-dimensional map depicting Dimensions 1 and 2 shows that Group 2 is characterized by greater social networking with other CSEM users; Group 2 is clearly separated from Group 3 on this dimension, with Group 1 occupying the interceding space. Group 3 is notably concentrated in a region of Dimension 2, indicating greater propensity toward contact offending. Examining the map depicting Dimensions 1 and 3, the separation of groups becomes most prominent: In addition to aforementioned between-group distances on Dimension 1, Dimension 3 reveals that participants in Group 1 are distinguished by their interest in fantasy-based CSEM.

Table 3 describes the offense profiles of the identified subgroups. Given the high percentage of offenders in Group 3 (n = 15) who engaged in contact sex offending, and given their generally lower level engagement in online CSEM use or networking, this group was labeled contact-driven (see Merdian et al., 2013). Groups 1 (n = 17) and 2 (n = 7) supported the existence of the fantasy-driven offense pathway, given the emphasis the individuals placed on sexual satisfaction derived from CSEM (82.4% vs. 100% reported to find CSEM sexually arousing). In contrast, it appears that Contact-Driven Users used CSEM only as an occasional outlet for sexual interest or exploration (20% reported they found CSEM sexually arousing).

Table 3.

Descriptive Information on CSEM User Subgroups Identified Via Cluster Analysis and Multidimensional Scaling.

Variables	Fantasy-driven users	Fantasy-driven users (social)	Contact-driven users
Offender type	13 CSEMOs, 4 MOs	5 CSEMOs,2 MOs	4 CSEMOs, 11 MOs
CSEM conviction	52.9% (1 reconvicted)	71.4% (1 reconvicted)	20.0%
	%	%	%
CSEM type
Digital images	94.1	100	60.0
Photos	23.5	14.3	6.7
Digital video	88.2	71.4	20.0
Video	0.0	14.3	0.0
Sound	5.9	14.3	13.3
Digital text	47.1	100	20.0
Magazines	11.8	14.3	26.7
CSEM content
Fictional	58.8	85.7	26.7
Preferably male	23.5	42.9	46.7
Young children/infants	29.4	71.4	0.0
Defined preferences	58.8	100	53.3
COPINE^a scale levels
Level 1	47.1	71.4	6.7
Level 2-4	88.2	71.4	40.0
Level 5	94.1	71.4	40.0
Level 6	100	100	86.7
Level 7	76.5	85.7	53.3
Level 8-9	76.5	85.7	46.7
Level 10 (sadistic)	29.4	71.4	6.7
Self-report offense motive
Material sexual arousing	82.4	100	20.0
Financial incentive	5.9	0.0	0.0
Own sexual trauma	23.5	0.0	0.0
Curiosity and sexual exploration	23.5	14.3	33.3
Stress relief	35.3	42.9	13.3
Sexual interest in minors	23.5	57.1	26.7
Desensitized to adult sexually explicit material	17.6	28.6	0.0
Statement against authority	0.0	28.6	0.0
	%	%	%
CSEM engagement
Paid for CSEM	35.3	14.3	6.7
Shared	23.5	85.7	0
Increase in CSEM contacts	5.9	71.4	6.7
≥10 hr per week	58.8	100	0
≥1 hr sorting collection	52.9	100	40
Saved on external device	64.7	85.7	0.0
Created hardcopies	23.5	57.1	6.7
Hidden	58.8	100	6.7
CSEM source
WWW	82.4	85.7	20.0
Chat	5.9	85.7	6.7
Newsgroup	23.5	71.4	6.7
File sharing	41.2	100	0.0
Email	0.0	42.9	6.7
Online supply/mail	0.0	0.0	13.3
Social networking
Online contact with minors	0.0	28.6	0.0
Online contact with adults sexually interested in children	23.5	100	13.3
Online contact with other CSEMOs	0.0	71.4	0.0
Offline contacts with adults sexually interested in children	11.8	100	6.7

Note. CSEM = child sexual exploitation material; CSEMO = offenders who use child sexual exploitation material; MOs = mixed offenders; WWW = World Wide Web.

The COPINE Scale (“Combating Paedophile Information Networks in Europe” Scale; Taylor et al., 2001) describes a typology of CSEM content, with ascending numbers depicting an increase in sexual explicitness and perceived victim impact.

Fantasy-Driven Users (Groups 1 and 2) described a more frequent and dynamic engagement with CSEM (e.g., searching for and sorting material) and reported sexually explicit material of broader variety and severity, including fantasy-based material such as fictional and narrative CSEM. Group 2 is partly differentiated from Group 1 by gradation; Group 2 reported greater preoccupation with CSEM and use of higher level material but most notably more extensive engagement with other users (embeddedness in peer networks) and minors through online channels.

Summary

Spatial and numerical classification methods led to the identification of three subtypes of CSEM users, with distinct profiles in their CSEM usage and online engagement. Social engagement with other CSEM users, contact sex offending, and usage of fantasy-based material were identified as the drivers of this classification. In summary, these findings evinced the existence of at least two distinct offense pathways, contact-driven and fantasy-driven CSEM users, in line with Merdian et al. (2013). A variant of this latter pathway was also apparent, marked by greater extremity and social embeddedness.

CSEM Subgroup Membership and Predictors of Contact Sex Reoffending

Analysis

An exploratory secondary analysis was conducted to model subgroup membership as a function of historical criminal activity and other conventional predictors of (contact) sex offending. As above, multinomial regression analysis was used to identify the variables that were differentially associated with group membership (i.e., those that have discriminatory and possible explanatory value); given the exploratory nature of the analysis and the small sample size, again a stepwise approach to variable selection was applied. For conventional predictors of contact sex offending, the following variables were deduced from the item pool: age below 25 years, intimacy deficits, criminal/antisocial lifestyle (childhood conduct issues, antisocial personality traits), treatment and supervision failures (sexual reoffending, more than one period of treatment for sexual behaviors), preference for male victims (contact, non-contact), sexual deviancy (deviant pornography other than CSEM, CSEM with extreme content, perception of CSEM as sexually arousing), and cognitive distortions. The breakdown of variables and distribution of scores can be seen in Table 4.

Table 4.

Criminal Activity of Offender Subgroups and Scores on Potential Predictor Variables Related to Contact Sex Offending.

Variable	Fantasy-driven users	Fantasy-driven users (social)	Contact-driven users
Variable	(n = 17)	(n = 7)	(n = 15)
Criminal activity
Conviction CSEM (maximum 2)	M = 0.6 (SD = 0.6)Mdn = 1	M = 0.9 (SD = 0.7)Mdn = 1	M = 0.2 (SD = 0.4)Mdn = 1
Current conviction	52.9% (9)	71.4% (5)	20.0% (3)
Previous conviction	5.9% (1)	14.3% (1)	0
Sexual contact minor	23.5% (4)	28.6% (2)	73.3% (11)
Sexual offense minor (maximum 3)	M = 0.4 (SD = 1.0)Mdn = 0	M = 0.3 (SD = 0.8)Mdn = 0	M = 1.3 (SD = 1.2)Mdn = 2
Current conviction	17.6% (3)	14.3% (1)	46.7% (7)
Previous conviction	11.8% (2)	0	33.3% (5)
Victims > 1	17.6% (3)	14.3% (1)	53.3% (8)
Production CSEM	5.9% (1)	28.6% (2)	13.3% (2)
Sexual offense adult (maximum 3)^a	0	M = 0.3 (SD = 0.8)Mdn = 0	M = 0.1 (SD = 0.3)Mdn = 0
Current conviction	0	0	6.7% (1)
Violent offending (maximum 2)	M = 0.2 (SD = 0.7)Mdn = 0	0	M = 0.7 (SD = 0.9)Mdn = 0
Conviction	11.8% (2)	0	40.0% (6)
Use of weapon	11.8% (2)	0	26.7% (4)
Non-violent offending	23.5% (4)	14.3% (1)	46.7% (7)
Online offending	52.9% (9)	100% (7)	26.7% (4)
Sum (maximum 15)	M = 3.2 (SD = 2.4)Mdn = 3	M = 2.0 (SD = 1.2)Mdn = 2	M = 3.5 (SD = 2.8)Mdn = 2
Conventional risk factors for sex offending
Age < 25 years	12.5% (2)	14.3% (1)	0
Intimacy deficits (maximum 3)	M = 1.3 (SD = 1.2)Mdn = 1	M = 2.0 (SD = 0.8)Mdn = 2	M = 0.8 (SD = 0.7)Mdn = 1
Criminal/antisocial lifestyle (maximum 11)	M = 4.4 (SD = 2.5)Mdn = 4	M = 3.7 (SD = 1.5)Mdn = 4	M = 5.3 (SD = 2.7)Mdn = 5
Childhood conduct (maximum 3)	M = 1.4 (SD = 1.1)Mdn = 1	M = 1.4 (SD = 0.9)Mdn = 2	M = 1.7 (SD = 1.3)Mdn = 2
Antisocial personality (maximum 10)	M = 2.9 (SD = 1.6)Mdn = 3	M = 2.1 (SD = 1.5)Mdn = 2	M = 3.3 (SD = 1.2)Mdn = 3
Domestic abuse	5.9% (1)	5.9% (1)	40.0% (6)
Treatment/supervision failures (maximum 3)	M = 0.4 (SD = 0.7)Mdn = 0	M = 0.4 (SD = 0.5)Mdn = 0	M = 0.5 (SD = 0.7)Mdn = 0
Reconviction contact so (minor)	11.8% (2)	0	33.3% (5)
Reconviction CSEM	0	14.3% (1)	0
Treatment period > 1	29.4% (5)	28.6% (2)	20.0% (3)
Male victim (maximum 2)	M = 0.4 (SD = 0.7)Mdn = 0	M = 0.6 (SD = 0.8)Mdn = 0	M = 0.8 (SD = 0.9)Mdn = 0
Contact offending	11.8% (2)	14.3% (1)	33.3% (5)
Preferably male CSEM	23.5% (4)	42.9% (3)	46.7% (7)
Sexual deviancy (maximum 14)	M = 8.4 (SD = 3.2)Mdn = 9	M = 10.9 (SD = 3.1)Mdn = 12	M = 5.4 (SD = 3.5)Mdn = 7
Deviant pornography (maximum 3)	M = 1.9 (SD = 1.1)Mdn = 2	M = 2.4 (SD = 0.5)Mdn = 2	M = 1.7 (SD = 1.2)Mdn = 2
CSEM with extreme content (maximum 4)	M = 0.9 (SD = 1.0)Mdn = 1	M = 2.1 (SD = 1.3)Mdn = 2	0
Level COPINE scale^b (maximum 6)	M = 4.8 (SD = 2.0)Mdn = 6	M = 5.3 (SD = 1.9)Mdn = 6	M = 3.3 (SD = 2.5)Mdn = 3
CSEM arousing	82.4% (14)	100% (7)	20.0% (3)
Cognitive distortions (maximum 195)	M = 71.5 (SD = 28.2)Mdn = 69	M = 57.9 (SD = 20.1)Mdn = 49	M = 91.3 (SD = 34.4)Mdn = 80
Children as sexual objects (maximum 65)	M = 22.9 (SD = 10.5)Mdn = 22	M = 18.6 (SD = 9.4)Mdn = 14	M = 29.9 (SD = 13.7)Mdn = 24
Justification (maximum 25)	M = 8.0 (SD = 4.0)Mdn = 6	M = 5.9 (SD = 1.5)Mdn = 5	M = 9.5 (SD = 4.3)Mdn = 8
Children as sexual agents (maximum 25)	M = 8.2 (SD = 3.6)Mdn = 7	M = 5.7 (SD = 1.9)Mdn = 5	M = 11.3 (SD = 4.7)Mdn = 10
Denial of sex offender status (maximum 30)	M = 12.5 (SD = 5.4)Mdn = 12	M = 12.3 (SD = 6.0)Mdn = 10	M = 16.3 (SD = 5.7)Mdn = 17
Emphasis on cognitive element (maximum 20)	M = 7.6 (SD = 2.9)Mdn = 8	M = 6.4 (SD = 1.8)Mdn = 7	M = 9.1 (SD = 2.7)Mdn = 8
Power and entitlement (maximum 25)	M = 9.6 (SD = 4.2)Mdn = 9	M = 7.3 (SD = 3.1)Mdn = 6	M = 11.3 (SD = 5.5)Mdn = 9
(Component 7; maximum 5)	M = 2.7 (SD = 1.6)Mdn = 2	M = 1.7 (SD = 1.5)Mdn = 1	M = 3.3 (SD = 1.4)Mdn = 3
Sum (maximum 230)	M = 85.5 (SD = 31.1)Mdn = 71	M = 70.4 (SD = 19.6)Mdn = 66	M = 108.9 (SD = 36.9); Mdn = 98

Note. Number of subjects (n) is listed in brackets next to percentage rates. None of the participants reported grooming activities, hence this category was removed. CSEM = child sexual exploitation material.

Sex offending against an adult: No participants scored on “previous convictions” or “more than one victim” and these variables are thus not displayed.

Higher levels of the COPINE scale are weighted more than low levels.

Results

A stepwise model (forward entry) resulted in correct subtype classification of 72% of CSEM users, based on three explanatory variables: sexual deviance, intimacy deficits, and viewing children as sexual agents. In terms of specific contrasts, Fantasy-Driven Users (Groups 1 and 2) were more likely than the Contact-Driven Users (Group 3) to use and report arousal to deviant sexually explicit material—Group 1: OR = 1.41, Wald χ²(1) = 4.88, p = .027; Group 2: OR = 2.19, Wald χ²(1) = 6.51, p = .011. Similarly, Fantasy-Driven Users (Groups 1 and 2) were more likely than Contact-Driven Users (Group 3) to report intimacy deficits—Group 1: OR = 4.51, Wald χ²(1) = 3.98, p = .046; Group 2: OR = 12.28, Wald χ²(1) = 6.85, p = .009. Finally, the Fantasy-Driven Users were less likely than Contact-Driven Users to view children as competent sexual agents—Group 1: OR = 0.76, Wald χ²(1) = 3.98, p = .046; Group 2: OR = 0.45, Wald χ²(1) = 5.43, p = .020. The model was able to successfully distinguish between fantasy-driven (Groups 1 and 2) and contact-driven groupings (Group 3), with 80% correct classifications of participants into Group 3. However, the explanatory variables did not reach significance (independently or in combination) for differentiating between the two fantasy-driven groupings (Group 1 vs. Group 2; ps ≥ .074); this was reflected in the relatively low percentage of correct classifications for participants in Group 2 (57%; with 43% misclassified as members of Group 1). Overall, the model revealed no significant outliers or influential cases and reached large effect sizes, R² = .58 (Cox & Snell), R² = .66 (Nagelkerke).

Summary

The above analyses further supported the broad distinction between fantasy-driven and contact-driven users, and suggest that these groups may be distinguished in terms of their scores across multiple conventional risk factors for offending. Contact-Driven Users were more likely to endorse cognitive distortions around the sexual agency of children, suggesting that relative to the other identified (fantasy-driven) subgroups, these users may be more comparable with conventional child sex offenders in their (post hoc) propensity to justify/rationalize child-directed sexual behavior. Fantasy-Driven Users were broadly distinguished from Contact-Driven Users by their relative difficulties in establishing and maintaining interpersonal relationships (intimacy deficits) and the intensity of their use (and arousal by) of deviant sexually explicit material (including use of more extreme CSEM). The two subgroupings of Fantasy-Driven Users were not significantly differentiated by their scores on variables gauging criminal history or conventional risk factors. However, relative to Group 1, Group-2 users seemed to tend toward more extreme scores with regard to their sexual deviance and intimacy deficits. The above analysis indicates that the diverse subgroups of CSEMOs have discriminating clinical (and potential risk) characteristics; however, the relationship between these potential risk predictors and actual recidivism risk could not be established with the data available for this study.

Discussion

The current study highlighted potential markers to differentiate individuals with contact and non-contact sex offenses, confirmed the previously identified heterogeneity in clinical and risk-related characteristics among CSEM users, and provided initial empirical support for Merdian et al.’s (2013) conceptual distinction of fantasy and contact-driven CSEM users.

Differences Between Contact and Non-Contact Sex Offenders

In contrast to contact sex offenders, CSEM users (CSEMOs and MOs) appeared to be more reliant on indirect means for achieving sexual and social stimulation, as indicated by elevated reporting on items related to Cluster 1, Social Exclusion and Escape (e.g., interest in second-life or third-person games, viewing sexually explicit material other than CSEM, excessive and problematic Internet use). In line with previous research (e.g., Seto, Reeves, & Jung, 2010; Taylor & Quayle, 2003), these findings indicate that the online environment appears to serve both appetitive and avoidant functions for CSEM users: facilitating distal social and sexual engagement while also providing a means to avoid the stresses of the offline world. However, whereas in the current study. the items within this cluster were able to differentiate between contact and non-contact sex offenders, it is unclear at present whether this cluster entails specific risk-related propensities or simply reflects traits of excessive (but legal) Internet users.

Interestingly, although offender subgroups initially did not differentiate on items relating to Cluster 4, Exposure to Adversity (e.g., abuse experience as a child), this cluster became a significant predictor of contact sex offending when modeled alongside Cluster 1: Social Exclusion and Escape. Although highly speculative at this stage, it could be hypothesized that, in contrast to exclusive CSEM users who tend to cope with adverse experiences through an avoidant behavioral style (e.g., escape, immersion in online and offline fantasy), contact sex offenders may have a greater tendency to “act out” (e.g., rule-breaking in childhood, impulsive decision making) in response to adverse life events.

The results also highlighted the significant role of cognitive distortions in differentiating between contact and non-contact sex offending. Specifically, exclusive CSEM users were found to report fewer cognitive distortions regarding children and sex, provided less justification for their offending behavior, and demonstrated less sexual entitlement than contact and mixed sex offenders. Notably, items relating to Justification had the strongest discriminatory power in grouping exclusive from mixed CSEM users. The lack of endorsement of these cognitions (or reporting thereof) may thus be indicative of internal inhibitions toward contact sex offending (see also Babchishin et al., 2015). These findings also further challenge the applicability of existing measures of cognitive distortions supportive of child sexual abuse for CSEM users, and highlight a need for the systematic development and validation of CSEM-specific cognitive distortion tools (e.g., Howitt & Sheldon, 2007; O’Brien & Webster, 2007).

Identification and Profile of CSEM Subgroups

In partial support of the conceptual model proposed by Merdian et al. (2013), spatial and numerical classification methods led to the identification of three subtypes of CSEM users. Broadly, the participants split into two subgroups: individuals with a primary interest in direct sexual contact with a minor (Contact-Driven Users; n = 15) and individuals whose offending behavior appeared focused on their CSEM usage (Fantasy-Driven Users; n = 24). However, within the latter, a further subgroup emerged (n = 7) differentiated through the possession of more extreme material (e.g., lower victim age, higher level of sexual explicitness) and higher social involvement with other CSEM users. In line with the profile described above, individuals on the fantasy-driven pathway (both subgroups) demonstrated a preference for indirect means of sexual satisfaction, reporting high levels of intimacy deficits, and more frequent use of sexually deviant materials (CSEM and other). The findings further confirmed the potential function of cognitive distortions to discriminate between contact- and fantasy-driven users, with the former being more likely to report viewing children as sexual agents. While these established predictors of contact sex offending (i.e., intimacy deficits, deviant sexual arousal, and viewing children as sexual agents) had some discriminatory power between contact- and fantasy-driven users, it is noteworthy that the majority of established risk factors assessed within the current study (see Table 4) were not useful for differentiating CSEM users with and without a primary motivation to engage in direct sexual contact with a minor.

While these findings may have important implications for the (risk) assessment and management of CSEM users, it is important to first highlight the influence of methodological and contextual factors on the data collected and the (statistical) interpretation thereof. The study relies on the use of self-report data in regard to both the questionnaires the participants completed and the offending behaviors reported. While there are perennial issues integrating the reliability and veracity of such data, this method of data collection can potentially facilitate the identification of previously undetected offending behaviors, an important consideration given the disparity between official and self-reported crime data reported for this population (e.g., Beier et al., 2014; Beier et al., 2009; Grundmann, Neutze, & Beier, 2010; Neutze et al., 2011; Seto et al., 2011). Furthermore, an anonymized computer survey was used to maximize and facilitate honest responding. Indeed, while CSEMOs are generally found to report a low criminal activity beyond their CSEM offending (e.g., Faust, Bickart, Renaud, & Camp, 2015; Seto & Eke, 2015; Seto et al., 2011), the current sample reported a higher occurrence of direct sexual contact with a minor (43.6%) than identified in other research samples (see Babchishin et al., 2010; Babchishin et al., 2015); however, this number reduced significantly when only official conviction data were considered (2.6%).

The exploratory nature of this study in combination with the very small sample size also limited the choice and applicability of methods of data analysis, and the generalizability of the results. For example, the relatively low level of statistical power led to the removal of a large amount of information due to low inter-correlations between measured variables. However, effects that can be found under these (statistical) circumstances are assumed to be replicated with larger sample sizes. It was thus considered more likely to miss effects with the current study design (Type-2 error) than to identify false effects (Type-1 error).

Contact-Driven and Fantasy-Driven Pathways to CSEM Offending

Despite these caveats, the findings from the current study provide initial empirical evidence for the conceptual distinction between contact-driven and fantasy-driven CSEM users (see Merdian et al., 2013). To clarify, the twofold distinction differentiates users of CSEM based on their main source of sexual satisfaction, namely, direct sexual contact with a minor, or fantasies thereof, with individuals on each pathway being characterized by the specific needs they aim to fulfill with their online offending behavior. The function of the Internet for the individual has previously been identified as a crucial aspect in the assessment of online offenders (see Caple, 2008; Sheldon & Howitt, 2007; Surjadi, Bullens, Van Horn, & Bogaerts, 2010; Taylor & Quayle, 2003), and the current findings further support this notion.

Beyond the exploration of their sexual needs, the Internet appears to have a more elaborate function for individuals on the fantasy-driven pathway, such as satisfying a preference for sexually extreme content and providing a means for establishing social connectedness. The current study identified a number of critical defining variables for this group, most notably high intimacy deficits, sexual deviance, and low endorsement of children as sexual agents. In contrast, individuals on the contact-driven pathway showed some similarity with the conventional profile of sex offenders (e.g., history of confrontational offending, antisociality). In his review on CSEM offenders, Seto (2013) pointed to the importance of antisociality for users with a history of contact sex offending, a finding subsequently confirmed by Babchishin et al. (2015). From the 15 users identified as contact-driven offenders in the current study, four had not engaged in direct sexual contact with a minor. Their placement on the contact-driven pathway could therefore represent a misclassification error, or be a potential indicator of an enhanced risk of cross-over, although this cannot be confirmed with the current data. Of particular note is that no individual within the contact-driven group reported contacting minors online; however, almost half of the fantasy-driven users had engaged in online contact with a minor. While this may indicate that individuals on the contact-driven pathway underreported the use of the Internet to gain access to victims, or that fantasy-driven users underreported the commission of offline offending, this finding could also indicate that online contact-behavior in itself is not strongly related to the commission of an offline contact sex offense, but may have fantasy-facilitative functions (as reported by Briggs, Simon, & Simonson, 2011). Although this may suggest a higher likelihood for offenders on the contact-driven pathway to cross-over or recidivate with a contact sex offense (in line with Seto & Eke, 2015), the current study did now allow for any conclusive statements on risk probability to be made. Nevertheless, while established risk measures have been shown to have some predictive validity for CSEM users with a history of contact sex offenses, the empirical evidence to date indicates that they have limited value for the assessment of exclusive CSEM users (e.g., Osborn et al., 2010; Seto & Eke, 2015; Wakeling et al., 2011). Overall, these apparent differences between contact sex offenders and CSEM users in general, and between the two pathways of CSEM offending specifically, point to a need for a critical review of current clinical and risk assessments for CSEM users, specifically those on the fantasy-driven pathway.

Where to From Here?

The findings presented here provide initial empirical support for the contact-driven versus fantasy-driven distinction of CSEMOs previously outlined by Merdian et al. (2013). While this development is important, it is clear further research on the classification and risk assessment of CSEM offenders is needed. Specifically, contextual aspects of CSEM offending require empirical exploration, such as what actually constitutes “deviant” online sexual behavior, how and why sexual fantasies regarding minors develop, the relationship between online fantasy and behavioral enactment (e.g., see Bartels & Gannon, 2011; Dombert et al., 2015), and how these factors may be assessed and measured (e.g., see Ortigue, Patel, & Bianchi-Demicheli, 2009; Waismann, Fenwick, Wilson, Hewett, & Lumsden, 2003). Furthermore, the interplay between sexual fantasy and social contact with other offenders with a sexual interest in minors in the offending of CSEM users needs to be clarified. Above all, it still remains to be established whether CSEM offending constitutes an inherently different type of offending, or a different dimension of sex offending.

Footnotes

Authors’ Note

Douglas P. Boer is now at the Faculty of Health, University of Canberra, Australia.

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.

Notes

References

Abel

G. G.

Becker

J. V.

Cunningham-Rathner

Rouleau

J. L.

Kaplan

Reich

(1984). Treatment manual: The treatment of child molesters. Atlanta, GA: Emory University.

American Psychological Association. (2010). Ethical principles of psychologists and code of conduct. Retrieved from http://www.apa.org/ethics/code/

Babchishin

K. M.

Hanson

R. K.

Hermann

C. A.

(2010). The characteristics of online sex offenders: A meta-analysis. Sexual Abuse: A Journal of Research and Treatment, 23, 92-123. doi:10.1177/1079063210370708

Babchishin

K. M.

Hanson

R. K.

VanZuylen

(2015). Online child pornography offenders are different: A meta-analysis of the characteristics of online and offline sex offenders against children. Archives of Sexual Behavior, 44, 45-66. doi:10.1007/s10508-014-0270-x

Bartels

R. M.

Gannon

T. A.

(2011). Understanding the sexual fantasies of sex offenders and their correlates. Aggression and Violent Behavior, 16, 551-561. doi:10.1016/j.avb2011.08.002

Beier

K. M.

Grundmann

Kuhle

L. F.

Scherner

Konrad

Amelung

(2014). The German Dunkelfeld Project: A pilot study to prevent child sexual abuse and the use of child abusive images. Journal of Sexual Medicine, 12, 529-542. doi:10.1111/jsm. 12785

Beier

K. M.

Neutze

Mundt

J. A.

Ahlers

C. J.

Goecker

Konrad

Schaefer

G. A.

(2009). Encouraging self-identified paedophiles and hebephiles to seek professional help: First results of the Prevention Project Dunkelfeld (PPD). Child Abuse & Neglect, 33, 545-549. doi:10.1016/j.chiabu.2009.04.002

Briggs

Simon

W. T.

Simonson

(2011). An exploratory study of internet-initiated sexual offenses and the chat room sex offender: Has the internet enabled a new typology of sex offender? Sexual Abuse: A Journal of Research and Treatment, 23, 72-91. doi:10.1177/1079063210384275

Buschman

Bogaerts

Foulger

Wilcox

Sosnowski

Cushman

(2010). Sexual history disclosure polygraph examinations with cybercrime offences: A first Dutch explorative study. International Journal of Offender Therapy and Comparative Criminology, 54, 395-411. doi:10.1177/0306624X09334942

10.

Caple

(2008). A comparison of the characteristics and motivations of abusing and non-abusing child pornography offenders (Unpublished doctoral dissertation). James Cook University, Townsville, Queensland, Australia.

11.

Crown Prosecution Service. (2014). Violence against women and girls crime report 2013-14. Retrieved from http://www.cps.gov.uk/publications/docs/cps_vawg_report_2014.pdf

12.

Dombert

Schmidt

A. F.

Banse

Briken

Hoyer

Neutze

Osterheider

(2015). How common is men’s self-reported sexual interest in prepubescent children? Journal of Sex Research, 4, 1-10. Retrieved from https://orbilu.uni.lu/bitstream/10993/20033/1/Dombert_Schmidt_et_al.2015.Sexual_interest_in_children_JSR.pdf

13.

Elliott

I. A.

Beech

A. R.

Mandeville-Norden

(2012). Psychological profiles of internet, contact, and mixed internet/contact sex offenders. Sexual Abuse: A Journal of Research and Treatment, 25, 3-10. doi:10.1177/1079063212439426

14.

Elliott

I. A.

Beech

A. R.

Mandeville-Norden

Hayes

(2009). Psychological profiles of internet sex offenders: Comparisons with contact sex offenders. Sexual Abuse: A Journal of Research and Treatment, 21, 76-92. doi:10.1177/1079063208326929

15.

Everitt

B. S.

Rabe-Heskett

(1997). The analysis of proximity data. London, England: Arnold.

16.

Faust

Bickart

Renaud

Camp

(2015). Child pornography possessors and child contact sex offenders: A multilevel comparison of demographic characteristics and rates of recidivism. Sexual Abuse: A Journal of Research and Treatment, 15, 5460-5478. doi:10.1177/1079063214521469

17.

Formann

A. K.

(1984). The latent class analysis: Introduction to theory and application. Weinheim, Germany: Beltz.

18.

Glasgow

(2010). The potential of digital evidence to contribute to risk assessment of internet offenders. Journal of Sexual Aggression, 16, 87-106. doi:10.1080/13552600903428839

19.

Grundmann

Neutze

Beier

K. M.

(2010, September). Psychopathic characteristics and sexual deviancy in pedophilic and hebephilic child pornography offenders. Paper presented at the 11th IATSO Conference, Oslo, Norway.

20.

Hanson

R. K.

Babchishin

K. M.

(2009, April). How should we advance our knowledge of risk assessment for internet sex offenders? Position Paper prepared for the G8 Global Symposium “Global symposium for examining the relationship between online and offline offences and preventing the sexual exploitation of children,” University of North Carolina, Chapel Hill.

21.

Hanson

R. K.

Morton-Bourgon

K. E.

(2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 118 prediction studies. Psychological Assessment, 21, 1-21. doi:10.1037/a0014421

22.

Hanson

R. K.

Thornton

(2000). Improving risk assessments for sex offenders: A comparison of three actuarial scales. Law and Human Behavior, 24(1), 119-136. doi:10.1023/A:1005482921333

23.

Hanson

R. K.

Thornton

(2003). Notes on the development of the Static-2002 (User Report 2003-01). Ottawa, Ontario: Solicitor General Canada.

24.

Henshaw

Ogloff

J. R.

Clough

J. A.

(2017). Looking beyond the screen: A critical review of the literature on the online child pornography offender. Sexual Abuse: A Journal of Research and Treatment, 29(5), 416–445.

25.

Howitt

Sheldon

(2007). The role of cognitive distortions in paedophilic offending: Internet and contact offenders compared. Psychology, Crime & Law, 13, 469-486. doi:10.1080/10683160601060564

26.

Kruskal

J. B.

Wish

(1978). Multidimensional scaling (Sage University Paper series on Quantitative Applications in the Social Sciences, 07-11). Beverly Hills, CA: Sage.

27.

Long

M. L.

Alison

L. A.

McManus

M. A.

(2013). Child pornography and likelihood of contact abuse: A comparison between contact child sexual offenders and non-contact offenders. Sexual Abuse: A Journal of Research and Treatment, 25, 370-395. doi:10.1177/1079063212464398

28.

Merdian

H. L.

Curtis

Thakker

Wilson

Boer

D. P.

(2013). The three dimensions of online child pornography offending. Journal of Sexual Aggression, 19, 121-132. doi:10.1080/13552600.2011.611898

29.

Merdian

H. L.

Curtis

Thakker

Wilson

Boer

D. P.

(2014). The endorsement of cognitive distortions: Comparing child pornography offenders and contact sex offenders. Psychology, Crime & Law, 20, 971-993. doi:10.1080/1068316X.2014.902454

30.

Middleton

Beech

Mandeville-Norden

(2005). What sort of person could do that? Psychological profiles of internet pornography users. In Quayle

Taylor

(Eds.), Viewing child pornography on the internet: Understanding the offence, managing the offender, helping the victims (pp. 99-107). Dorset, UK: Russell House.

31.

Middleton

Mandeville-Norden

Hayes

(2009). Does treatment work with internet sex offenders? Emerging findings from the Internet Sex Offender Treatment Programme (i-SOTP). Journal of Sexual Aggression, 15, 5-19. doi:10.1080/13552600802673444

32.

Neutze

Seto

M. C.

Schaefer

G. A.

Mundt

I. A.

Beier

K. M.

(2011). Predictors of child pornography offenses and child sexual abuse in a community sample of pedophiles and hebephiles. Sexual Abuse: A Journal of Research and Treatment, 23, 212-242. doi:10.1177/1079063210382043

33.

O’Brien

M. D.

Webster

S. D.

(2007). The construction and preliminary validation of the Internet Behaviours and Attitudes Questionnaire (IBAQ). Sexual Abuse: A Journal of Research and Treatment, 19, 237-256. doi:10.1007/s11194-007-9057-0

34.

Ortigue

Patel

Bianchi-Demicheli

(2009). New Electroencephalogram (EEG) Neuroimaging Methods of analyzing brain activity applicable to the study of human sexual response. Journal of Sexual Medicine, 6, 1830-1845. doi:10.1111/j.1743-6109.2009.01271.x

35.

Osborn

Elliott

I. A.

Middleton

Beech

A. R.

(2010). The use of actuarial risk assessment measures with UK internet child pornography offenders. Journal of Sexual Aggression, Conflict and Peace Research, 2(3), 16-24. doi:10.5042/jacpr.2010.0333

36.

Seto

M. C.

(2013). Internet sex offenders. Washington, DC: American Psychological Association.

37.

Seto

M. C.

Eke

A. W.

(2015). Predicting recidivism among adult male child pornography offenders: Development of the Child Pornography Offender Risk Tool (CPORT). Law and Human Behaviour, 39, 416-429. doi:10.1037/lhb0000128.

38.

Seto

M. C.

Hanson

R. K.

Babchishin

K. M.

(2011). Contact sexual offending by men with online sexual offenses. Sexual Abuse: A Journal of Research and Treatment, 23, 124-145. doi:10.1177/1079063210369013

39.

Seto

M. C.

Reeves

Jung

(2010). Explanations given by child pornography offenders for their crimes. Journal of Sexual Aggression, 16, 169-180. doi:10.1080/13552600903572396

40.

Sheldon

Howitt

(2007). Sex offenders and the internet. Chichester, UK: West Sussex.

41.

Surjadi

Bullens

Van Horn

Bogaerts

(2010). Internet offending: Sexual and non-sexual functions within a Dutch sample. Journal of Sexual Aggression, 16, 47-58. doi:10.1080/13552600903470054

42.

Taylor

Holland

Quayle

(2001). Typology of paedophile picture collections. The Police Journal, 74, 97-107.

43.

Taylor

Quayle

(2003). Child pornography: An internet crime. Hove, UK: Brunner-Routledge.

44.

Thornton

Mann

Webster

Blud

Travers

Friendship

Erikson

(2003). Distinguishing and combining risks for sexual and violent recidivism. In Prentky

Janus

Seto

Burgess

A. W.

(Eds.), Annals of the New York Academy of Sciences: Vol. 989. Sexually coercive behaviour: Understanding and management (pp. 225-235). New York: New York Academy of Sciences.

45.

Waismann

Fenwick

P. B. C.

Wilson

G. D.

Hewett

T. D.

Lumsden

(2003). EEG responses to visual erotic stimuli in men with normal and paraphilic interests. Archives of Sexual Behavior, 32, 135-144. doi:10.1023/A:1022448308791

46.

Wakeling

H. C.

Howard

Barnett

(2011). Comparing the validity of the RM2000 Scales and OGRS3 for predicting recidivism by internet sex offenders. Sexual Abuse: A Journal of Research and Treatment, 23, 146-168. doi:10.1177/1079063210375974

47.

Webb

Craissati

Keen

(2007). Characteristics of internet child pornography offenders: A comparison with child molesters. Sexual Abuse: A Journal of Research and Treatment, 19, 449-465. doi:10.1007/s11194-007-9063-2

48.

Wood

J. M.

Seto

M. C.

Flynn

Wilson-Cotton

Dedmon

(2009, October). Is it “just” pictures? The use of polygraph with internet offenders who deny abusive sexual contact. Poster presented at the 28th Annual Conference of the Association for the Treatment of Sexual Abusers, Dallas, TX.