Using the Explicit and Implicit Sexual Interest Profile in applied forensic or clinical contexts
Alexander F. Schmidt and Derek Perkins
Paraphilic interests in forensic contexts
The assessment of sexual interests is a crucial diagnostic task in forensic assessments of sexual offenders. Particularly, paraphilic sexual interest in non-normative sexual objects or activities (i.e. as a motivating factor) together with antisocial dispositions (i.e. as a facilitating factor) are discussed as major risk factors for sexual offending (Seto, 2019). Both constructs are among the core underlying factors in widely-used actuarial recidivism risk assessment instruments such as the Static-99R or Static 2002R (Brouillette-Alarie, Proulx, & Hanson, 2018). Paraphilic interests have been meta-analytically identified as among the most valid predictors of sexual reoffending (Mann, Hanson, & Thornton, 2010) and there are self-reported cross-sectional links to sexual victimization of children in community males (Dombert et al., 2016; Klein, Schmidt, Turner, & Briken, 2015).
Usually, in applied forensic and clinical contexts sexual interests are inferred from behavioural observations based on known offending behaviour (for an overview of recent advances see Lehmann, Dahle, & Schmidt, 2018) in combination with self- and other-reports (Kalmus & Beech, 2005). However, these approaches are limited in terms of validity as the offending behaviour-paraphilic interest link is equivocal and self-reports are problematic due to the associated legal and social repercussions for the respondent (Jahnke, 2018). Not every individual with paedophilic sexual interest is sexually victimizing children (Dombert et al., 2016; Seto, 2018) and the majority of child sexual abusers does not exhibit sexual preferences in the sense of a paedophilic disorder (Schmidt, Mokros, & Banse, 2013). Moreover, with roughly one out of three diagnoses of paedophilic disorder being invalid (Mokros, Habermeyer, & Kiichenhoff, 2018), the diagnostic accuracy of routine care clinical paedophilia diagnoses falls short of the desired validity threshold for single case assessments informing decisions that need to balance civil freedom restrictions against public safety.
Indirect latency-based measures of sexual interest in children
In order to tackle the abovementioned problems with the more or less direct assessment ofparaphilic interests in applied forensic and clinical contexts, particularly indirect latency-based measures of sexual interest in children have seen a growing body of research over the last two decades. Several (social-) cognitive psychology paradigms have been validated within sexual offender samples (for recent overviews see Bartels, Gray, & Snowden, 2017; Schmidt, Banse, & Imhoff, 2015). However, most of these measures are in relatively early validation stages based only on few studies at most with the exception of Viewing Time (VT; e.g. Harris, Rice, Quinsey, & Chaplin, 1996) measures and Implicit Association Tests (IAT; e.g. Gray, Brown, MacCulloch, Smith, & Snowden, 2005) for which cumulative meta-analyses exist.
For VT measures the average meta-analytic effect size for discriminating child sexual abusers (as a proxy measure for sexual interest in children) from controls (Schmidt, Babchishin, & Lehmann, 2017) is d = 0.60, C/95% [0.51, 0.68], n = 2,705, k = 14. This increases once an optimal scoring algorithm based on a maximized sexual preference index for children over adults (described below) will be used (d = 1.03, CI95% [0.82, 1.25], n = 414, k = 7). Moreover, VT measures are meaningfully associated with offence-behavioural indicators (Screening Scale for Pedophilic Interests [SSPI]; Seto & Lalumiere, 2001), self-report, penile plethysmographic, and IAT measures of sexual interest in children (Schmidt et al., 2017) as well as other such indirect latency-based and oculomotoric measures (O Ciardha, Attard-Johnson, & Bindemann, 2018). Furthermore, it has been shown that a VT measure of sexual interest in children predicts sexual recidivism over a 15-year follow-up interval (Harrel’s c = .68; Gray et al., 2015). For IAT measures of sexual preferences for children over adults meta-analytic results show a similar average effect in discriminating child sexual abusers from controls (d = 0.63, CI95% [0.42, 0.83], n = 707, k = 12). However, predictive validity for sexual preference lATs remains untested to date. Notably, for comparison, penile plethysmographic measures for sexual interest in children yield similar meta-analytic validity estimates for contrasting sexual offenders against children from controls (d = 0.67, C/95% [0.50, 0.84], n = 6,734, k = 32) increasing to d = 1.01, CI95% [0.64, 1.37], n = 3,116, k = 13, if based on a z-standardized preference index and a mean predictive validity of d = 0.44, CI95% [0.32, 0.57], n = 1,961, k = 16 (McPhail et al., 2019).
The explicit and implicit sexual interest profile
From a psychometric perspective it is a fact that any single measure is amenable to personal and situational moderators. Therefore, resting diagnostic single-case decisions solely on single measures is suboptimal. A solution to this problem lies in the general diagnostic principle of convergence according to which diagnostic conclusions can be drawn with greater confidence if they are derived from several conceptually different, valid, and convergent measures. Based on this notion, we combined the empirically most promising indirect latency-based measures as well as a self-report questionnaire into a computer-based test battery of sexual interests in children and adults - the Explicit and Implicit Sexual Interest Profile (EISIP; Banse, Schmidt, & Ciarbour, 2010; note also the yet not forensically validated EISIP variant tapping into interests in sexual coercion; Larue et al., 2014). In the following we will describe the different EISIP measures in the order as they occur during assessment.
Viewing time measures
The VT measure consists of 80 (plus four practice trials) pictures of semi-nude White Caucasian male and female persons in bathing clothes presented in a prefixed random-order. All stimuli are computer-manipulated pictures from the Not-Real People picture set (Pacific Psychological Assessment Corporation, 2004) depicting non-existing humans. Stimuli are grouped into male and female stimuli with each eight exemplars across five sexual maturity stages (Tanner stages; Tanner, 1973; see Figure 4.1). Tanner stages are based on the physical development of primary and secondary sexual organs (shape, size, colour) and must not be mapped onto specific age bands due to individual differences in onset and course of pubertal development. Conventionally, Tanner stages are grouped into prepubescent (Tanner 1), peripubescent (Tanner 2, 3), and postpubescent (Tanner 4, 5) classes dovetailing with paedophilic, hebe-philic, and teleiophilic sexual interests, respectively. Respondents are asked to rate the subjective sexual attractiveness of the target stimuli on a five-point Likert scale ranging from 1 (‘sexually unexciting’) to 5 (‘sexually very exciting’) without time constraints while VT is unobtrusively recorded.
The original idea of using viewing times as a measure related to sexual interest is based on Rosenzweig (1942) although the first application of a VT measure in the sense described here dates back to Zamansky (1956). It is a robust
Figure 4.1 Illustration of Tanner stages with exemplary pictures from the Not-Real People picture set (Pacific Psychological Assessment Corporation, 2004).
finding that the more sexually attractive stimuli are, the longer it takes to reach a rating decision in terms of their subjective sexual attractivity (Imhoff et al., 2010; Schmidt et al., 2017). Notably, these decision latencies are not caused by attentional engagement or holding effects (Imhoff, Barker, & Schmidt, 2019) as often supposed in the literature (Imhoff et al., 2010) but are the result of an experimentally corroborated feature checking process (Imhoff, Schmidt, Weiß, Young, & Banse, 2012). According to this notion, in order to decide whether a stimulus is sexually attractive, one needs to check a set of idiosyncratic features such as, for example, age, gender, and attractivity. Once one of these features can be ruled out, the rating process ends (i.e. a relatively faster process) whereas as long as features cannot be rejected, the decision process will need to take further features into account (i.e. a relatively slower process). Importantly, from this psychological process it follows that it is not the stimuli features that drive VT effects but rather the set of features that determine subjective sexual attractiveness. The latter set ultimately is a function of the rating task (Imhoff et al., 2012). From an applied perspective, this task-driven effect poses an important potential diagnostic validity threat to VT paradigms: Only as long as participants comply with the instructions to rate targets’ sexual attractiveness from their own personal perspective the measure produces meaningful results. On the contrary, whenever participants perform rating tasks not relevant to their own sexual perspective (e.g. rating stimuli from a vicarious perspective or determining exclusively a single stimulus property such as age or hair colour), latency patterns in standard VT paradigms become invalid (Imhoff et al., 2012; Pohl, Wolters, & Ponseti, 2016).
For diagnostic purposes, VT latencies (in milliseconds) are aggregated into 2 (Target Gender) x 5 (Tanner Stages) subscales indicating absolute sexual interest levels. Absolute measures have the advantage to illustrate absolute sexual interest differences across all target categories. Subtracting the maximum adult category (Tanner 4,5) from the maximum child category (Tanner 1,2,3) irrespective of target gender results in a relative sexual preference index1 that is independent of sexual orientation (i.e. sexual gender preferences). It indicates to what extent children are sexually preferred over adults. Relative preference indexes, however, must not be interpreted in absolute terms (i.e. they are not informative whether the identified sexual preference or non-preference is due to high or low absolute levels of sexual interest in the underlying target categories). As laid out above - and similar to PPG assessments (McPhail et al., 2019) - this preference index is the most valid measure when it comes to criterion group differentiation (Schmidt et al., 2017).
Implicit association tests
The EISIP consist of three different sexual preference IATs: a Men-Women IAT, a Girls-Women IAT, and a Boys-Men IAT. The first IAT assesses sexual preferences between adult females and males (i.e. sexual orientation) whereas the latter IATs tap into sexual age preferences for adults over children within both gender categories, respectively. Each IAT follows the original five-block order as described in Greenwald, McGhee, and Schwartz (1998) but with an increased number of trials. All blocks were presented in the same prefixed random order as described in the following utilizing pictorial stimuli from the VT measure. The first block of 40 trials is a categorization task of ten words (i.e. attribute categories) that have to be classified as sexually exciting (erotic, exciting, lustful, sensual, orgasm) or unexciting (dull, bland, indifferent, unexciting, boring) by pressing a left or right response key. In the second 40-trial block, ten pictures have to be assigned to the target categories man versus woman (girl vs. woman, boy vs. man, respectively in the other two IATs). In the third block, both tasks are mixed in alternating order. Four practice trials precede 80 test trials. The left response key has to be pressed for items belonging to the categories man or sexually unexciting as opposed to the right response key for items indicating woman or sexual excitement. The fourth block of 40 trials is similar to the second, but the key assignment is reversed. Finally, in the fifth block 80 of test trials (plus four practice trials), both tasks are again combined, however, with reversed target categories: Now the left response key has to be pressed for items relating to the categories man or sexually exciting, and the right response key for woman or sexually unexciting attributes (similarly adapted for the Boys-Men and Girls-Women IATs). Incorrect responses are indicated by an error message to the participant throughout all blocks without requiring a further correct response.
The IAT was originally developed by Greenwald et al. (1998). It is a latencybased measure that serves as an indicator ofthe strengths ofimplicit associations between attribute and target categories each arranged on bipolar dimensions. As laid out above, respondents classify word or picture stimuli that each represent one of these four categories (e.g. men vs. women as target categories and words representing sexual excitement vs. no excitement) using two response keys, each assigned to the crosswise combined target and attribute categories. In this double discrimination task, classifications are relatively faster (i.e. cognitively easier) if two closely associated target and attribute concepts share the same response key (compared to trials in which they are assigned to different response keys). The underlying psychological process is most likely an effect of response interference if two non-associated concepts share the same response key (Gawronski, Deutsch, & Banse, 2011).
The IATs are scored by calculating the difference between the mean response latencies of the critical third and fifth block, divided by the pooled standard deviation of response latencies (D-measure; Greenwald, Nosek, & Banaji, 2003). Only correct trials are included into the analysis. This standardized individual effect size measure - similar to Cohens d - reflects relative sexual preferences for one target category over the other and controls for individual differences in executive functioning abilities (De Houwer, Teige-Mocigemba, Spruyt, & Moors, 2009). In the EISIP higher sexual preference scores represent the magnitude of atypical or paraphilic preferences for men over women, girls over women, and boys over men, respectively.
Explicit Sexual Interest Questionnaire (ESIQ)
The Explicit Sexual Interest Questionnaire (ESIQ; Banse et al., 2010; see Table 4.1 for the slightly modified item set that differs from the originally
Table 4.1 Item overview of the revised Explicit Sexual Interest Questionnaire (ESIQ; Banse et al., 2010).
I have enjoyed orally stimulating X.
I have tongue kissed X.
I have enjoyed getting my private parts touched by X.
I have sexually caressed X.
I have had sexual intercourse with X.
I find it erotic if I see X’s beautiful chest.
I have daydreamed of having sex with X.
I find it erotic to see X’s body through the clothes.
I find it erotic to see X’s beautiful bottom.
I get excited when I imagine that X stimulates me.
Note. The original ESIQ items from Banse et al. (2010) have been slightly modified to better fit prototypical sexual behaviours that are shown across all target groups - both items mentioning penetrative acts have been replaced with the 'I have tongue kissed’ and ‘I have had sexual intercourse’ items. ’X’ refers to four different groups of stimuli and has to be replaced with ‘a young girl’, ‘a young boy’, 'a woman’, ‘a man', respectively.
published items and is currently in use) was purpose-designed in order to be able to compute absolute measures of sexual interest in children and adults (i.e. indicators of how sexually attractive certain target categories are that can be compared with the absolute indicators from the VT assessments) and relative sexual preference indexes (i.e. indicators of how much more children are sexually preferred over adults that comparable with the IAT results and the VT sexual preference index).
The 40-item ESIQ consists of each five self-report items tapping into basic (a) sexual behaviours and (b) sexual fantasies. Moreover, each of these items is assessed relating to four sexual target categories: (1) young girls, (2) young boys, (3) women, (4) men. All items refer to life-time prevalences starting from when respondents’ were 18 years old and are rated on a binary forced-choice (yes/no) scale, thus, indicating the rate of scale items acknowledged. Child categories are referring to prepubes-cent stimuli younger than 12 years as defined in the general instructions. Mean scales scores can be aggregated into 2 (Behaviours or Fantasies) x 4 (Target categories) subscales or - on a higher level - into the four sexual target categories only.
Due to the fact that at present a psychometrically flawless criterion for sexual interest in children is missing, validation attempts for such measures need to be tested against several possible proxy criteria of these paraphilic interests (Schmidt et al., 2017)? The EISIP has been shown to validly discriminate between multiple proxy criteria of sexual interest in children and adults: It distinguishes convicted child sexual abusers from various offender and non-offender control groups (Banse et al., 2010) including different groups of sexual offenders against children (i.e. extrafamilial child sexual abusers show more paedohebephilic sexual preferences than intrafamilial offenders; Schmidt, Gykiere, Vanhoeck, Mann, & Banse, 2014). Aggregated EISIP scores and the underlying single measures were meaningfully associated with clinicians’ diagnoses of paedophilic disorder (Schmidt, Bonus, & Banse, 2010), a phallometrically validated offence-behavioural measure (SSPI, Seto & Lalu-miere, 2001) and static actuarial risk assessment measures (as shown in Banse et al., 2010; Schmidt et al., 2014). Moreover, taxometric analyses among men with and without sexual interests in children have revealed a taxonic structure of paedohebephilic interests based on the EISIP - as well as only its latencybased measures - with roughly every fourth child sexual abuser exhibiting relatively marked paedohebephilic preferences as compared to non-paedo-hebephilic men (Schmidt et al., 2013). Notably, EISIP indirect latency-based measures are incrementally valid above and beyond self-report measures of sexual interest in children and adults as shown for selected non-denying (Banse et al., 2010) vs. routine care (i.e. containing a larger amount of denying child sexual abusers; Schmidt et al., 2010) samples of sexual offenders against children. Also, these studies revealed that the VT measures outperformed the IATs in multivariate group classifications based on the single EISIP measures (Banse et al., 2010; Schmidt et al., 2010, 2013).
Although, the EISIP seems to be unrelated to general socially desirable responding tendencies (Banse et al., 2010), it certainly can be wilfully altered - much like any psychological measure - if the underlying measurement rationale is known and the respondent has the cognitive abilities and the motivation to do so. Strikingly, among convicted contact child sexual abusers we have found that men who denied any sexual interest in children or any sexual behaviour involving minors on the ESIQ show decreased sexual preference indexes for children over adults on the EISIP indirect latency-based measures. This may look like corroborating the deliberate malleability of the EISIP. However, the SSPI scores of these paedophilic interest-denying sexual offenders against children were also lower than in the non-denying child sexual abusing controls (Figure 4.2). Taken together, this corroborates that deniers’ offending
Figure 4.2 Indirect latency-based EISIP measures |sexual preference indexes) and Screening Scale for Pédophilie Interest (SSPI) scores in convicted contact child sexual abusers who admitted vs. denied sexual fantasies or sexual behaviours involving children on the Explicit Sexual Interest Questionnaire (ESIQ; unpublished data).
behaviour (as an independent objective indicator) also indicated lesser sexual interest in children over adults underscoring the notion that their self-report is actually more valid than prototypically supposed in forensic (research) contexts.
In single case diagnostics - apart from a measures validity - reliability is of crucial concern (in fact, reliability limits the validity of psychological measures according to classical test theory). While latency-based measures are notorious for low reliabilities (i.e. internal consistencies, test-retest correlations), the EISIP consistently yields remarkably good reliability coefficients. In terms of internal consistencies (Cronbach’s as), the ESIQ behaviour and fantasy subscales range from .86 (good) to .96 (excellent; Banse et al., 2010) and .88 to .97 for the aggregated ESIQ sexual interest scales (Banse et al., 2010; Schmidt et al., 2014). The absolute VT measures range between .77 and .85 in samples including sexual offenders, non-sexual offenders, and non-offenders (Banse et al., 2010) and from .90 to .95 in exclusive child sexual abuser samples (Schmidt et al., 2014). The relative sexual preference IATs ranged from .79 to .89 with the exception of the Boys-Man IAT that reached values of .61 and .65 (Banse et al., 2010; Schmidt et al., 2014). The low internal consistency of the Boys-Men IAT is highly likely a result of a sampling artefact due to a lack of gay men who were not child sexual abusers in the samples. Therefore, virtually all participants were either sexually attracted to men and boys or to neither of them. In consequence, the IAT as a relative measure of sexual preference for one target category over the other could not detect substantial individual differences in its Boy-Man version (i.e. mostly near-zero scores leading to low variability) and hence, reliability was low.
Last but not least, recent data (Welsch, Schmidt, Tuner, & Rettenberger, 2019) revealed that classical relative test-retest reliabilities for the EISIP sexual preference scores were surprisingly high over a 14-day interval that had been chosen to rule out spontaneous change of sexual preferences in convicted child sexual abusers and community controls. Intraclass correlation coefficients (ICCs) for sexual preference indexes ranged from satisfactory to excellent (.66 IAT, .78 VT, .87 ESIQ, and .90 full EISIP aggregate score). However, more sophisticated absolute test-retest reliability analyses based on Bland-Altman plots found that quite large effects from roughly a third to three quarters of a standardized mean difference unit will be needed to rule out artificial change induced by transient measurement error. This means the EISIP is a particularly well-suited measure in case repeated assessments seek to elucidate rank order stability of sexual preferences for children over adults over time. However, if the actual amount of individual change is the main research question one will need to show relatively larger differences between measurement times in order to be sure that genuine change in sexual preferences has taken place (for a detailed discussion see Welsch et al., 2019).
Testing single cases with the Explicit and Implicit Sexual Interest Profile
Single case assessments with the EISIP are fully computerized (i.e. standardized presentation and instructions are presented on a laptop during the assessment) and take about 30 minutes for the average participant. At present there are German, English, and Flemish/Dutch versions of the EISIP and further translations could be easily implemented. Respondents need basic reading and computer handling skills. The EISIP can be applied with prototypical (non-intellectually challenged) offender populations in forensic psychiatric and prison settings. Once the respondent has taken the assessment, there is a tool available that generates the EISIP profile directly from the assessment data (Figures 4.3 and 4.4).
Importantly, the EISIP is a measure that is designed to assess sexual interests in children and adults. Based on the outlined studies we can be quite certain that it is valid for this task, indeed. Moreover, it has been
Figure 4.3 EISIP profiles from a convicted child sexual abuser with a diagnosis of paedophilic disorder who had victimized multiple partly prepubescent and partly extrafamilial girls (upper panel: aggregated ESIQ scales, absolute VT indexes, and IAT preference scores; lower panel: ESIQ subscales, explicit picture ratings in the EISIP).
shown that the EISIP measures of paedo-, hebe-, and teleiophilic interests are cross-sectionally associated with forensically relevant risk factors for reoffending (Schmidt et al., 2014) and a similar VT measure such as used in the EISIP has been prospectively linked to recidivism risk in forensic samples (Gray et al., 2015). Although in cross-sectional studies with community males self-reported paedohebephilic interest and sexual offending against children have been shown to be related (Bailey, Bernhard, & Hsu, 2016; Dombert et al., 2016; Klein et al., 2015), the causal relationship, however,
Figure 4.4 EISIP profiles from a convicted violent offender with no history of sexual offending (upper panel: aggregated ESIQ scales, absolute VT indexes, and IAT preference scores; lower panel: ESIQ subscales, explicit picture ratings in the EISIP).
of child sexual abuse and paedohebephilic interests remains inconclusive in non-forensic contexts. Hence, application of the EISIP as a screening measure in non-offending community samples needs to carefully take into account these validity restrictions (e.g. Turner, Hoyer, Schmidt, Klein, & Briken, 2016). Based on these findings, the EISIP must not be used to determine whether a) any respondent has sexually offended in the past or b) to determine a non-offending respondent’s future risk for sexual victimization of children.3
Graphical assessment profile output
The graphical output of the assessment results can be generated via a simple computer program that allows to print and electronically store single case EISIP profiles in a standard graphic format. It produces two profiles on separate pages (Figures 4.3 and 4.4). On the left side margin the respective measurement categories are labelled whereas on the right side margin raw values for each measurement category are reported in order to give the diagnostician full transparency across the whole assessment outcomes. The metric used to graphically display the profile is arbitrary in case of the VT measure and must not be compared with the ESIQ and IAT profiles nor with different VT profiles from other cases.
Profile page 1 - aggregated self-report and indirect latency-based measures
The first profile on page one (upper panels of Figures 4.3 and 4.4) presents an overview of the ESIQ self-reports aggregated over behaviours and fantasies for the women, men, girls, and boys subscales, respectively. The section labelled ‘Questionnaire’ lists the rate of items per category that have been answered with ‘yes’ with scores varying between 0 and 1 (rates > 0.5 indicate that at least one sexual behaviour on this scales has necessarily been endorsed).
The first ten rows in the section labelled ‘Objective’ report the VT latencies across Tanner categories 1 to 5 for female and male stimuli. Raw mean latencies >10.000 ms will be indicated with an exclamation mark as in the validation studies all single latencies that exceed this threshold were truncated at 10.000 ms. However, in order for the diagnostician to get a more accurate overview of the assessment this has not been implemented in the graphical output. All VT categories are rescaled to a minimum category value of zero and a maximum of 1. The last three rows in the ‘Objective’ section outline the results from the sexual preference IAT assessments starting with the Woman/ Man IAT, followed by the Woman/Girl IAT and the Man/Boy IAT as indicated by the actual sexual preference D-scores. Higher values point to relatively stronger preferences for men over women (i.e. gay sexual orientation for male respondents), girls over women, and boys over men, respectively. The plotted range is restricted from -1 to 1, larger sexual preferences will be displayed as -1 or 1, respectively, and marked with a‘T’ (i.e. truncated) in the right margin displaying the raw values. However, sexual preference D-scores > 111 represent very large sexual preferences and are empirically rather seldom. Additionally, IAT mean error rates > 35% for the respective measures are indicated below the profile as these should be interpreted only with caution.
Profile page 2 - detailed self-reports
The second profile page (lower panels of Figures 4.3 and 4.4) displays all selfreported measures from the EISIP. It starts with the more detailed endorsement rates for the behaviour and fantasy ESIQ subscales for women, men, girls, and boys. These allow to assess dissociations of sexual behaviours and fantasies across sexual target categories. They are followed by the explicit mean picture ratings from the VT assessments, starting with female targets from Tanner 1 to Tanner 5 and followed by male targets in the same order. Raw values for rating plots range from zero to four and are rescaled from zero to one for the graphical output.
The EISIP does not provide norms to compare assessment results against. Hence, profile interpretations need to be based on ipsative comparisons of high and low points in the sexual interest profile. These can be used as a reference frame or baseline of the subjectively least and most sexually interesting target categories within and across the EISIP measures. Note that the VT and ESIQ measures represent absolute sexual interest indicators for each target category whereas the IAT measures are relative sexual preference measures of one target category over the other that cannot be compared directly against each other. Absolute sexual interest measures are informative about how interesting single target categories are. In contrast, relative sexual preference measures reveal how much more sexually interesting a specific target category is than another comparison target category (see the section on VT scoring above). Also, it should be noted that VT measures consistently turn out as more valid than the IATs which needs to be taken into account when interpreting the EISIP profiles.
Case 1: paedophilic sexual offender against children
The assessment of the first case presented in Figure 4.2 stems from a convicted male child sexual abuser who had been diagnosed with a paedophilic disorder by the treating clinicians. He had offended against multiple female child victims including victims below twelve years of age and extrafamilial victims. His aggregated self-reported sexual interests (upper panel in Figure 4.2) indicate that women are his preferred sexual target category followed by somewhat lesser sexual interest in girls. Male targets have been identified as sexually unappealing. This is corroborated by the VT measures that show clearly decreased sexual interest levels for any male target category in comparison to any female target category. Moreover, Tanner 4 (i.e. postpubescent) females yield the second highest decision latencies (note that Tanner 4 and Tanner 2 females showed raw mean latencies > 10.000 ms which are roughly twice as long as for the least interesting target category of Tanner 5 males). However, in contrast to the self-reported sexual interest maximum for women VT measures indicate a maximum interest level for Tanner 2 females. Also, note the sharp decline from Tanner 4 to Tanner 5 females. This sexual preference for girls over woman is backed up by the Woman/Girl IAT showing also a slight relative preference for girls over woman (with a D-score of 0.19 which can be interpreted similar to Cohens d in terms of effect size conventions; Cohen, 1988). The sexual orientation Woman/Man IAT indicates heterosexual preferences among adult sexual targets. In line with this finding, there is no sexual preference for men over boys (D-score -0.05). Note that this does not necessarily mean that men and boys are sexually irrelevant as it could also be the case that both target categories are highly interesting to this respondent. Both cases of correspondingly high or low absolute sexual interests in the underlying target categories would result in a sexual preference score of zero. The VT results, however, corroborate the latter interpretation of generally low absolute sexual interest levels in any male target category.
This diagnostic picture is further elucidated taking into account the assessment results from the second profile page (lower panel in Figure 4.2). The explicit picture ratings also fit in with the former findings of a clear heterosexual orientation (as indicated by zero ratings for any male picture). Strikingly, the respondent acknowledges sexual attraction for any of the female pictures across the whole sexual maturity range (with an apparent linear increase from Tanner 1 to Tanner 5). This dissociation between the VT assessment and the self-reported indications might be explained by the fact that the ESIQ woman behaviour subscale shows no sexual experience with adult women at all as opposed to a maximum of sexual fantasies directed towards adult females. Clinically, this might indicate difficulties in engaging in intimate relationships with adults although considerable interest in sexual contacts with women might be present. This might be followed up in the post-assessment interview. Notably, the respondent also acknowledges sexual fantasies involving prepubescent girls but denies any sexual contacts with girls with the former clearly indicating paedophilic sexual interest but the latter contradicting his actual conviction for child sexual abuse (there might be meaningful reasons that explain this minimization of offending behaviour; e.g. Maruna & Mann, 2006).
Case 2: violent non-sexual offender
The second case presented in Figure 4.4 reports an EISIP profile from a male violent offender with no official history of sexual offending. The first profile page (upper panel in Figure 4.4) indicates exclusive sexual interest in adult women in the aggregated ESIQ self-report matched by clear sexual preferences for women over men and women over girls in the IATs (D-scores -0.92 and -0.45, respectively). This is well corroborated by the VT profile showing the prototypical ‘normality spike’ that indicates a strong heterosexual preference for postpubescent Tanner 4 and Tanner 5 females over any other sexual target category. Finally, the Man/Boy IAT yields a mild preference of men over boys which might be explained by the fact that the respondent exhibits stronger general implicit associations between adults and sex than for children and sex although the VT measures and the ESIQ self-reports showed low absolute sexual interest levels in any male sexual target category. The second profile page fully underscores the outlined diagnostic pattern of exclusive sexual preferences for postpubescent females.
Feedbacking assessment results in therapeutic contexts
Particularly in therapeutic contexts, feedbacking the EISIP might be the most central aspect of the sexual interest assessment as here a fundamental part of the foundation for the following therapeutic interventions and working alliance will be laid. As usual in therapeutic contexts, the EISIP should be used to foster a collaborative process and not as a ‘truth detector’ or psychological ‘x-ray machine’. For therapeutic purposes it might be more important to get a sense of what the respondent concludes for himself from the assessment. Hence, the feedback should try to elicit the subjective meaning and hypotheses that the assessed individual comes up with when given information what the EISIP might indicate about his sexual interests from the view of the therapist. In clinical contexts it might thus be helpful to start with sexually healthy or unproblematic sexual interests as these might be used as therapeutic resources for the following process of strengthening non-deviant sexual interests and/or managing paedohebephilic interests (this shall not imply that it is impossible in any case to change manifest paedohebephilic sexual preferences). It might also be worthwhile to elucidate patterns of temporal stability for the given sexual interest pattern and to explore boundary conditions or situational factors under which healthy and deviant sexual interests over the life course have been increased or decreased. Moreover, possible early sexual victimization experiences of the respondent himself should be explored in the sense of how these might relate to the specific pattern of EISIP results.
Importantly, we advise against the direct use of the actual profile printout when feedbacking the results as the raw values and profile labels might elicit questions concerning the underlying measures. The actual names of underlying tasks nor the fact that latencies are recorded during the assessment must not be mentioned as this unduly reveals the measurement rationale to the respondent and thus will prevent further use of the EISIP in applied contexts in the longer run. Moreover, one should never mention VT or the exact nature of the dependent variables used in the EISIP. It is bad enough that the DSM-5 (American Psychiatric Association, 2013) mentions VT as a valid diagnostic tool in the introductory section on paedophilic disorder giving away the rationale of these tasks. Note that although this might sound as if we were actively deceiving respondents neither do diagnosticians/clinicians (a) regularly reveal the actual scoring keys of any diagnostic interview or psychometric scale to the respondents in order to protect their assessments’ future diagnostic potential, nor (b) is the EISIP assessment particularly deceptive on what it actually assesses. It is rather clear from the first trial on, what the assessment purpose of the EISIP is (namely, sexual interest in children and adults). Certainly - as with any other measure as well - every respondent is free to deny an EISIP assessment or stop taking part in it during the assessment process (as well as trying to get access to the published literature on any assessment he is interested in). An ethically satisficing way of explaining the underlying measurement rationale while at the same time protecting the EISIP s validity might be to refer to individual differences in information processing that are exhibited in classification and sorting tasks (without mentioning response latencies of course) as laid out in the next section on how to apply the EISIP in court.
Using the Explicit and Implicit Sexual Interest Profile in court
Forensic psychological assessments for Court proceedings (criminal or civil) and for other, quasi-judicial, processes such as Mental Health Review Tribunals and Prison Parole Boards require a comprehensive and multifaceted approach. It is recognized that no single type of assessment on its own is likely to be valid or robust against test counter measures in responding to the questions typically needing to be addressed. In much the same way that the EISIP comprises several elements from which a triangulation of findings can be achieved, using the EISIP in court and similar proceedings also requires this test to be used in combination with other assessments. These include a thorough file review, interviews with the person concerned (patient, prisoner, defendant etc.) and with relevant others (spouse, parents, friends etc.), together with findings from other modalities of assessment, such as psychometric, psychophysiological, or observed behavioural responses.
As noted earlier, the EISIP is designed for, and is successful in identifying patterns of sexual interest across different age groups and gender. On its own, it is unable to, and is therefore inappropriate for, explaining past sexual offending or assessing risks of future offending. However, given that paedophilic sexual interest is one of two major risk factors for sexual recidivism (Hanson & Morton-Bourgon, 2005; Seto, 2019) - the second being general antisociality - the EISIP, when combined with measures of other relevant factors, such as antisocial attitudes, patterns of socio-sexual behaviour, opportunities to offend and the absence of barriers to offending, can have a useful part to play within Court or similar proceedings. Issues to be addressed will typically include (a) psychological case formulation of the behaviour of concern - child sexual abuse, sexual violence against adults etc. - (b) assessment of likely future risk, (c) identification of treatment needs and (d) analysis of data on treatment progress. It is important that these assessments cover an appropriately wide range of domains, combine different assessment modalities and adopt a hypothesis-testing and hypothesis-generating approach. Within this context, information from EISIP assessments can be very helpful. The fact that an EISIP interpretation profile can be instantaneously produced, in a format that can be quickly interpreted by the assessor, can aid assessment by feeding back and discussing the findings, in the context of other information, within the assessment session.
In the second authors clinical and forensic experience, this process of testing, feedback and discussion often leads to disclosures and acknowledgements that can assist in clarifying risks, needs, and future offence risk management strategies. Where the findings from the EISIP’s various subsections align -
i.e. where self-report on the ESIQ, viewing time and IAT findings are all pointing in the same direction - this increases confidence in the results. Where this is not the case, differences in the sub-test findings can be discussed with the subject and this often provides information useful to the assessment - for example possible explanations for these differences or the generation of further hypotheses to be tested. It can of course also arise that the two objective measures align - for example to suggest sexual interest in pre-pubescent girls - but are discrepant from the subjects self-report - for example a sole sexual interest in women. This can in turn raise questions about the test subject’s insight or test faking, for example. This would be particularly so if known past behaviours (such as a previous conviction for child sexual abuse or child sexual exploitation material) and fantasies (for example, child fantasies disclosed during previous assessments or in the subject’s private writings) are not acknowledged in the ESIQ. In such cases, while it is not possible to categorically assert that the EISIP’s objective measures ‘prove’ a subject’s sexual interests, discrepancy between these and the ESIQ, especially where the ESIQ is also discrepant with known history, increases confidence in a hypothesis of minimization of sexual interest in children or denial of past behaviours. Again, while this does not itself help quantify risk, it both highlights areas for further exploration
(such as motives for minimization: fear, shame, self-interest) and points to aspects that might be most pertinent in generating an offence formulation and, through this, identifying relative risks and needs.
The following two cases - anonymized by altering specific, potentially identifying features - illustrate some of these points in the incorporation of EISIP testing into wider psychological assessments for legal proceedings.
Mr A, aged 28, was a named party in Family Court proceedings in which the two pre-pubescent children of his partner Ms X had been put at risk by Ms X’s prior association with a convicted child sexual abuser who had targeted her online, seemingly to access her two young daughters. Mr A, who had entered the relationship with Ms X shortly after this, but before police and social care became involved, was therefore also subject to investigation. As part of this, his computer and mobile devices were subject to forensic analysis, revealing that he had accessed pornography, including ‘teen material recovered from his computer. There was concern that he too might be a risk to the children.
As part of the psychological evaluation of Mr A for the Court, file review and collateral interviews were carried out, Mr A undertook various interviews and psychometrics assessments, including the EISIP, and the full contents of his pornography collection were sought, obtained, and analysed. The question was whether his association with Ms X and her children, his minimizations of the risk her daughters had faced - in an apparent attempt to protect her from police scrutiny - and his possession of‘teen’ material, meant that he too might be a risk to the girls.
On the ESIQ, he indicated only adult heterosexual behaviours and fantasies, which could of course be a faked response. On the EISIP’s VT measure of sexual interest, his profile indicated an interest in post-pubescent females, but one which was higher for post-pubescent teenaged females than for adult females. On the implicit association test, his profile indicated preferences for women rather than girls.
Mr As pornography collection, while containing some teen material, also contained other categories of sexual interest in adult females, such as ‘lesbian’, ‘uniformed women’, older females’, which in total exceeded the quantity of teen material; he was not therefore exclusively focused on teens. There was no pre-pubescent material recovered, which was consistent with his own account and the EISIP findings. In relation to general antisociality (the second major dynamic risk factor for sexual offending; Hanson & Morton-Bourgon, 2005; Seto, 2019), he had only one conviction, for a driving offence, there were no indications of generalized antisocial behaviour. He appeared desperate to maintain his relationship with Ms X following past episodes of partners leaving him, because he was regarded as too passive and unassertive, qualities that Ms X appeared to value, and the Court accepted that this was the main but misguided reason he tried to minimize the risk to the girls. The couple successful undertook some individual and joint therapy to address aspects of the risk and need assessment, to which the EISIP had made a useful contribution.
Mr B, aged 34, was a recently married man who had been involved in live previous adult heterosexual relationships. During a period of work stress and marital problems, he increasingly accessed adult pornography, seemingly to distract from, and soothe himself in response to his other problems. During this period, his pornography access drifted to younger female images. He was arrested for possessing illegal child material of girls in the range 12-16 years.
Mr B recalled early and very exciting sexual activity with a 14-year-old girl when he was aged 13, which he contrasted to the more awkward and less exciting sexual activity with his first 18-year-old girlfriend when he was 20. She left him for another man, which resulted in him being cautious about forming further relationships and he had a period of casual sexual encounters. He eventually entered a relationship with a young woman of his own age whom he met at work and they married.
His lifelong difficulties discussing his emotions appeared to have combined with his marital and work related problems and he recaptured earlier experiences of sexual excitement in viewing online adult pornography during which he was drawn to younger, teen images, which he said during assessment reminded him of his early sexual experiences at age 13.
Mr B took part in a psychological evaluation for the criminal court, in which a formulation of his offending and assessment of future offending risk were required. In addition to file review, interviews and psychometric testing of cognitive, personality and mental health functioning, he undertook the EISIP. On the ESIQ, he acknowledged some previous fantasies of girls in addition to his predominantly adult heterosexual fantasies and behaviour. He rated Tanner 4 teenaged girls as the most sexually exciting category. On the VT, he showed similar response to pubescent and post-pubescent females (Tanner 3 - Tanner 5), but not to pre-pubescent children. On the IATs, his sexual orientation was predominantly adult female.
In the assessment interviews, which included a review of his psychometric test results including the EISP, Mr B acknowledged a sexual interest in pubescent girls. He had justified to himself that accessing such images online ‘did no harm’ as the material was ‘already out there’. He was very ashamed, fearful and contrite and responded well to psychoeducational sessions about the nature of, and harm created by the viewing of online indecent images of children or minors. He received a community disposal and he and his wife worked together to restore and enhance their relationship.
Mr C, aged 56, was a married man whose wife was aged 62. They had been married for 15 years and there were no children of the marriage. Mr C had three previous convictions involving pre-pubescent boys. At 26, he was convicted of indecent assault of a 10-year-old boy, who was the son of family friends. Mr C was placed on probation and maintained that he was suffering from work-related stress and the breakup of a relationship at the time. He said that he had accepted that he had touched the boy but had only done so as he was advised that this would prevent a prison sentence. He said that he had no sexual interest or preoccupation with boys.
His second conviction, at the age of 41, was just before his marriage. He was convicted of sexually touching two boys, aged 9 and 10, while working at his church Sunday school. He maintained that these boys had heard of his previous conviction and had made up the story as they were jealous of his attention to other children in the group. He went to prison for 12 months; he was not eligible for sexual offender treatment, and he and his wife moved from the city to a new rural home upon release.
Mr and Mrs C ran a local shop and, fourteen years after moving there, two boys aged 10 and 11 made complaints that Mr C had touched them sexually while they were helping him with gardening and odd jobs at the shop: they had responded to an advertisement for this work. Mr C was arrested and his computer searched but there was nothing incriminating found on it. Enquiries suggested that other boys may have been assaulted and those continued while Mr C was being assessed prior to the trial.
Mr C completed a psychological assessment, which drew on the court papers, including witness statements and forensic evidence; interviews with Mr and Mrs C; and psychological testing, including the EISIP. On the ESIQ, Mr C acknowledged only adult heterosexual behaviour and fantasies, but the VT measure and the lATs indicated otherwise. On VT, his responses peaked at Tanner 2 and Tanner 3 males (this is, boys aged approximately 10-14 years old), with some elevation to Tanner 5 adult male. On the lATs, his result on the male dimension was skewed well over to boys rather than men. On the female dimension his score was in the middle.
The interpretation of the EISIP findings was that Mr C showed a strong interest in peri-pubescent males and this was fed back to him during the assessment for his comments. His response was less about whether this was correct than about how this would be seen in the context of his trial. As more evidence of other possible offences came to light, Mr C asked to review the EISIP findings again. He said that he had no adult homosexual interest but said that he had been abused as a boy and he wondered if that could be the cause of the adult male Tanner 5 elevation. Without mentioning his previous denials about his sexual interest in boys, he said that the male Tanner 2 and Tanner 3 findings seemed accurate and he asked if there was anything that could be done about that. He said that his wife confided that she too had been sexually abused as a girl and that she was uncomfortable with sex, which he had found acceptable. After lengthy discussion, and in the context of other findings, Mr C was keen to pursue antilibidinal medication as an option for managing his sexual interest in boys. He was disinclined to other forms of therapy, partly because of his age and partly because he just wanted not to be tempted to be offending against boys again. He received three years imprisonment, was released and he had his wife moved again, into sheltered accommodation, where he continued to take antilibidinal medication.
We hope that this chapter on the potential use of the EISIP in forensic contexts is instructive. In summary, we believe that the EISIP (Banse et al., 2010) is a viable diagnostic adjunct to penile plethysmographic assessment procedures - the only other valid objective measure of paraphilic sexual interest that is available in applied clinical (forensic) contexts (McPhail et al., 2019). We hope that this necessarily short introduction into the practical use of the EISIP may aid in underscoring the diagnostic potential (as well as its limitations) for single case applications of this easy to apply computerized assessment battery.
- 1. This means that sexual interest and sexual preference are not used interchangeably in this chapter. Sexual interests - the broader conception - refers to absolute interest levels independent of other target categories whereas sexual preference - the narrower conception - is the result of the comparison (i.e., the difference) of two different target categories (see Schmidt et al., 2017 for a more detailed description).
- 2. Given the fact that only a minority of convicted child sexual abusers show genuine pae-dophilic preferences (Schmidt et al., 2013; Seto, 2018), validity estimates based on sexual offending samples will be highly likely to underestimate criterion validity.
- 3. At the present stage of research, this does not imply that the EISIP must not be used to assess paedohebephilic preferences in non-offending populations as long as one does not draw any firm conclusions on whether these sexual inclinations are causally linked to a possible onset of future sexual offending.
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Arlington, VA: Author.
Bailey, J. M., Bernhard, P. A., & Hsu, K. J. (2016). An Internet study of men sexually attracted to children: Correlates of sexual offending against children. Journal of Abnormal Psychology, 125(7), 989-1000.
Banse, R., Schmidt, A. E, & Ciarbour, J. (2010). Indirect measures of sexual interest in child sex offenders: A multimethod approach. Criminal Justice and Behavior, 37(3), 319-335.
Bartels, R. M., Gray, N. S., & Snowden, R. J. (2017). Indirect measures of deviant sexual interest. In L. A. Craig & M. Rettenberger (Eds.), The Wiley handbook on the theories, assessment, and treatment of sexual offending - Volume II: Assessment (pp. 965-993). Chichester, West Sussex, UK: Wiley-Blackwell.
Brouillette-Alarie, S„ Proulx, J., & Hanson, R. K. (2018). Three central dimensions of sexual recidivism risk: Understanding the latent constructs of Static-99R and Static-2002R. Sexual Abuse, 30(6), 676-704.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
De Houwer, J., Teige-Mocigemba, S„ Spruyt, A., & Moors, A. (2009). Implicit measures: A normative analysis and review. Psychological Bulletin, 135(3), 347-368.
Dombert, B„ Schmidt, A. E, Banse, R., Briken, P., Hoyer, J., Neutze, J., & Osterheider, M. (2016). How common is men’s self-reported sexual interest in prepubescent children? The Journal of Sex Research, 53(2), 214-223.
Gawronski, B., Deutsch, R., & Banse, R. (2011). Response interference tasks as indirect measures of automatic associations. In K. C. Klauer, C. Stahl, 8r A. Voss (Eds.), Cognitive methods in social psychology (pp. 78-123). New York, NY: Guilford Press.
Gray, N. S„ Brown, A. S„ MacCulloch, M. J., Smith, J., 8< Snowden, R. J. (2005). An implicit test of the associations between children and sex in pedophiles. Journal of Abnormal Psychology, 114(2), 304-308.
Gray, S. R., Abel, G. G„ Jordan, A., Garby, T, Wiegel, M„ & Harlow, N. (2015). Visual reaction time™ as a predictor of sexual offense recidivism. Sexual Abuse: A Journal of Research and Treatment, 27(2), 173-188.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74(6), 1464-1480.
Greenwald, A. G„ Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the implicit association test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85(2), 197-216.
Hanson, R. K., 8< Morton-Bourgon, K. E. (2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73(6), 1154-1163.
Harris, G. T, Rice, M. E., Quinsey, V. L, & Chaplin, T. C. (1996). Viewing time as a measure of sexual interest among child molesters and normal heterosexual men. Behaviour Research and Therapy, 34(4), 389-394.
Imhoff, R„ Barker, P., & Schmidt, A. F. (2019). To what extent do erotic images elicit visuo-spatial versus cognitive attentional processes? Consistent support for a (non-spatial) Sexual Content-Induced Delay. Archives of Sexual Behavior. Advance online publication.
Imhoff, R., Schmidt, A. F„ Nordsiek, U„ Luzar, C., Young, A. W„ & Banse, R. (2010). Viewing time effects revisited: Prolonged response latencies for sexually attractive targets under restricted task conditions. Archives of Sexual Behavior, 39(6), 1275-1288.
Imhoff, R„ Schmidt, A. E, Weiß, S„ Young, A. W, & Banse, R. (2012). Vicarious viewing time: Prolonged response latencies for sexually attractive targets as a function of task- or stimulusspecific processing. Archives of Sexual Behavior, 41(6), 1389-1401.
Jahnke, S. (2018). The stigma of pedophilia: Clinical and forensic implications. European Psychologist, 23(2), 144-153.
Kalmus, E., & Beech, A. R. (2005). Forensic assessment of sexual interest: A review. Aggression and Violent Behavior, 10(2), 193-217.
Klein, V., Schmidt, A. E, Turner, D., 8; Briken, P. (2015). Are sex drive and hypersexuality associated with pedophilic interest and child sexual abuse in a male community sample? PLoS One, 10(7), e0129730.
Larue, D„ Schmidt, A. E, Imhoff, R., Eggers, K„ Schönbrodt, F. D„ & Banse, R. (2014). Validation of direct and indirect measures of preference for sexualized violence. Psychological Assessment, 26(4), 1173-1183.
Lehmann, R. J. B„ Dahle, K.-P, & Schmidt, A. F. (2018). Primer on the contribution of crime scene behavior to the forensic assessment of sexual offenders. European Psychologist, 23(2), 154-166.
Mann, R. E., Hanson, R. K., 8< Thornton, D. (2010). Assessing risk for sexual recidivism: Some proposals on the nature of psychologically meaningful risk factors. Sexual Abuse, 22(2), 191-217.
Maruna, S., & Mann, R. E. (2006). A fundamental attribution error? Rethinking cognitive distortions. Legal and Criminological Psychology, 11(2), 155-177.
McPhail, I. V, Hermann, C. A., Fernane, S., Fernandez, Y. M„ Nunes, K. L., & Cantor, J. M. (2019). Validity in phallometric testing for sexual interests in children: A meta-analytic review. Assessment, 26(3), 535-551.
Mokros, A., Habermeyer, E., & Kiichenhoff, H. (2018). The uncertainty of psychological and psychiatric diagnoses. Psychological Assessment, 30(4), 556-560.
Ö Ciardha, C„ Attard-Johnson, J., & Bindemann, M. (2018). Latency-based and psychophysiological measures of sexual interest show convergent and concurrent validity. Archives of Sexual Behavior, 47(3), 637-649.
Pacific Psychological Assessment Cooperation. (2004). The Not-Real People stimulus set for assessment of sexual interest. Victoria, BC: Author.
Pohl, A., Wolters, A., 8; Ponseti, J. (2016). Investigating the task dependency of viewing time effects. The Journal of Sex Research, 53(8), 1027-1035.
Rosenzweig, S. (1942) The photoscope as an objective device for evaluating sexual interest. Psychosomatic Medicine, April 1942 - pp 150-158.
Schmidt, A. E, Babchishin, K. M„ 8r Lehmann, R. J. (2017). A meta-analysis of viewing time measures of sexual interest in children. Archives of Sexual Behavior, 46(1), 287-300.
Schmidt, A. E, Banse, R., & Imhoff, R. (2015). Indirect measures in forensic contexts. In F. J. R. van de Vijver 8< T. Ortner (Eds.), Behavior based assessment in personality, social, and applied psychology (pp. 173-194). Göttingen: Hogrefe.
Schmidt, A. E, Bonus, P„ & Banse, R. (2010, July). Indirect measures of sexual interest in child sex offenders: A multimethod approach and its clinical implications. Paper presented at the international summer conference in forensic psychiatry, Regensburg, Germany.
Schmidt, A. E, Gykiere, K„ Vanhoeck, K., Mann, R. E„ & Banse, R. (2014). Direct and indirect measures of sexual maturity preferences differentiate subtypes of child sexual abusers. Sexual Abuse, 26(2), 107-128.
Schmidt, A. E, Mokros, A., & Banse, R. (2013). Is pedophilic sexual preference continuous? A taxometric analysis based on direct and indirect measures. Psychological Assessment, 25(4), 1146-1153.
Seto, M. C. (2018). Pedophilia and sexual offending against children: Theory, assessment, and intervention (2nd ed.). Washington, DC: American Psychological Association.
Seto, M. C. (2019). The motivation-facilitation model of sexual offending. Sexual Abuse, 31(1), 3-24.
Seto, M. C„ & Lalumière, M. L. (2001). A brief screening scale to identify pédophilie interests among child molesters. Sexual Abuse: A Journal of Research and Treatment, 13(1), 15-25.
Tanner, J. M. (1973). Growing up. Scientific American, 229,34-43.
Turner, D„ Hoyer, J., Schmidt, A. E, Klein, V., & Briken, P. (2016). Risk factors for sexual offending in men working with children - A community based survey. Archives of Sexual Behavior, 45, 1851-1861.
Welsch, R„ Schmidt, A. E, Turner, D., & Rettenberger, M. (2019). Test-retest reliability and temporal agreement of direct and indirect sexual interest measures. Manuscript submitted for publication.
Zamansky, H. S. (1956). A technique for measuring homosexual tendencies. Journal of Personality, 24, 436-448.