The Revised Screening Scale for Pedophilic Interests (SSPI–2): Development and Criterion-Related Validation

Abstract

The Screening Scale for Pedophilic Interests (SSPI) is a structured rating scale of four child victim characteristics: number, age, gender, and relationship of victims. Sexual offenders against children who score higher on the SSPI are more likely to be identified as having pedophilic sexual arousal and are more likely to sexually reoffend. Recent research suggests that child pornography offending is a valid and perhaps independent indicator of pedophilia. In this study, we examined whether child pornography offending would add to the criterion-related validity of the SSPI. In the construction sample of 950 offenders with child victims, the SSPI with a fifth item of child pornography (Screening Scale for Pedophilic Interests–2 [SSPI-2]) was significantly associated with phallometrically assessed sexual arousal to children. In a validation sample of 950 offenders with child victims, the SSPI-2 was again related to phallometrically assessed sexual arousal to children, outperforming the original SSPI. We recommend clinicians and researchers use the SSPI-2 as a structured method of assessing pedophilic sexual interests based on offending behavior.

Keywords

pedophilia sexual offending child pornography assessment diagnosis

The accurate assessment of sexual interest in boys or girls is central to appraisals of risk to sexually offend and to treatment planning for individuals who have committed sexual crimes against children. Assessment of pedophilia through interview or questionnaire can be helpful but self-report is limited by socially desirable responding, especially when the consequences are serious (e.g., denial of parole). Alternative assessment methods have been developed as a result, including measures based on looking time, reaction time, and genital response (reviewed in Kalmus & Beech, 2005; Seto, 2008, 2013). Researchers have also begun to examine assessment methods adapted from cognitive science, such as the Implicit Association Task, which assesses the strength of cognitive associations by comparing reaction times with different pairings of concepts (e.g., Babchishin, Nunes, & Hermann, 2013).

One of the best validated assessment tools for assessing pedophilic interests is phallometry. In the typical phallometric procedure, the subject is presented with stimuli (audio-recorded stories, photos, or videos) depicting persons of either sex and of various ages from prepubescent to pubescent children to adults. Penile responses are recorded in terms of changes in circumference or volume. Responses to children relative to adults consistently discriminate offenders against children from other men, including other sex offenders and nonsexual offenders (reviewed in Seto, 2008). Furthermore, phallometrically assessed sexual arousal to children is one of the strongest predictors of sexual recidivism among sex offenders, with a larger effect size (d = 0.37) than for other important risk factors such as psychopathy (d = 0.25), antisocial personality disorder (d = 0.29), and prior criminal history (d = 0.30; Hanson & Bussière, 1998; Hanson & Morton-Bourgon, 2005).

Although phallometric testing is supported empirically, its widespread application is limited by lack of adequate training and standardization across laboratories, perceptions about the intrusiveness of the procedure, and difficulties obtaining the sexual stimuli (depictions of nude children) that are required. Also, there are legal restrictions in the United States on the use of visual sexual stimuli depicting children (Prosecutorial Remedies and Other Tools to End the Exploitation of Children Today [PROTECT] Act of 2003; there is a legal exemption for possessing these stimuli for medical and educational purposes in Canada). Not surprisingly, therefore, phallometric testing is not commonly available, especially to service providers in less densely populated areas (McGrath, Cumming, Burchard, Zeoli, & Ellerby, 2009). In addition, phallometric testing is not permitted in the assessment of sexual offenders in some countries (e.g., Germany).

Responding to this gap, Seto and Lalumière (2001) developed the Screening Scale for Pedophilic Interests (SSPI) for clinical or research use when phallometric test results are unavailable. To develop a generalizable and scalable measure, they combined four well-established correlates of pedophilic penile responding that would usually be known to clinicians and researchers. These included the following: having any boy child victims, having multiple child victims, having any young child victims (below age 12), and having any extrafamilial child victims. Each SSPI item is scored as present or absent. Child victims were defined as below age 14 because the legal age of consent in Canada was 14 at that time. Extrafamiliality was defined as not son or daughter, step-son or step-daughter, nephew or niece, grandchild, and first cousins; the large majority of (intra)familial victims were offspring or step-offspring.

Seto and Lalumière (2001) planned to use unitary weights in scoring the SSPI, but having any boy victims was assigned a weight of 2 because exploratory analyses showed that it explained approximately twice the variance in relative penile response to children as each of the other three correlates. Information about child victims for the original SSPI was obtained from multiple sources, including official records regarding sexual victims and credible offender admissions of otherwise unreported child victims. Offenders with child victims could therefore have a SSPI score ranging from 0 to 5. An offender with a score of 0 would be an incest offender with a single female victim, aged 12 or 13. An offender with a score of 5 would be an offender with multiple child victims, at least one boy victim, at least one victim below the age of 12, and at least one extrafamilial child victim. In Seto and Lalumière’s construction sample of 1,113 offenders with child victims, the median SSPI score was 3 (M = 2.8, SD = 1.4). Individuals with a score of 5 were approximately 4 times more likely to show greater penile response to children than to adults than were individuals with a SSPI score of 0 (72% vs. 18%).

Thus, the SSPI can be viewed as a brief actuarial measure of pedophilic sexual interests, operationalized as sexual arousal to children as assessed by phallometry. SSPI scores were significantly and positively correlated with a continuously distributed index of penile response to children, r(1111) = .34, p < .001. A subsequent study found that the SSPI had a similar but smaller association with penile responses among adolescent sexual offenders with child victims (Seto, Murphy, Page, & Ennis, 2003). SSPI scores have also been found to predict sexual recidivism among adult offenders with child victims (Helmus, Ó Ciardha, & Seto, 2014; Seto, Harris, Rice, & Barbaree, 2004). Other investigators have also used the SSPI, finding it to correlate positively and significantly with other measures of pedophilic interest (e.g., Banse, Schmidt, & Clarbour, 2010; Canales, Olver, & Wong, 2009; Mokros, Dombert, Osterheider, Zappala, & Santilla, 2010).

The Present Study

The present study was conducted to improve the SSPI, taking advantage of accumulating evidence that possession of child pornography also predicts phallometrically assessed pedophilic sexual arousal (Blanchard et al., 2007; Seto, Cantor, & Blanchard, 2006). Seto et al. (2006) compared the phallometric responses of 100 child pornography offenders with a group of sex offenders who committed contact offenses against children (but had no child pornography charges) and with another comparison group of men with no evidence of sexual interest in children (no admitted or officially known sexual offenses involving children or child pornography). Of the child pornography offenders, 43 had also committed offenses directly against child victims; these men showed the greatest average sexual response to children. Overall, a majority (61%) of the child pornography offenders showed larger penile response to children over adults. In a subsequent study, using an overlapping sample of offenders, Blanchard et al. (2007) found that individuals who were identified as pedophiles or hebephiles (hebephilia represents sexual interest in pubescent children) on the basis of their phallometric responding had over 3 times the odds of having a history of child pornography offenses than those who were identified as sexually interested in adults (i.e., teleiophiles).

Here, we analyze a large sample of men charged with sexual offenses involving children or men who had committed both contact sexual offenses against children and used child pornography and who were not part of the SSPI development sample (Seto & Lalumière, 2001). We used this new sample to examine the performance of the original four SSPI items in conjunction with a new fifth item regarding child pornography offenses. This study also improves upon Seto and Lalumière (2001) in that all offenders were tested on a revised phallometric test that was first introduced in 1995 (Lykins et al., 2010). Based on the results of Seto et al. (2006) and Blanchard et al. (2007), we predicted that a child pornography item would add to the variance in phallometrically assessed sexual arousal to children explained by the four SSPI items.

Method

Participants

Participants were assessed at the Sexual Behaviours Clinic of the Centre for Addiction and Mental Health (SBC). Individuals were referred to the SBC if their sexual urges or behavior caused legal difficulties or clinically significant distress. The majority of patients referred to the SBC committed either contact or noncontact sexual offenses involving children, and most of the rest committed sexual offenses against adults.The referral source and purpose of the assessment varied considerably, contributing to the heterogeneity of the sample. Criminal justice personnel (e.g., courts, probation, or parole officers) were the most common source of referrals. When clients were referred by probation/parole or correctional institutions, the assessment was typically requested to inform risk management and treatment decisions. In a small number of cases, clients were referred by lawyers so that the report could be used in court proceedings (e.g., sentencing). There was a small proportion of patients who had not committed any sexual offenses or had committed otherwise undetected sex offenses and were self-referred for problematic sexual behaviors.

Patients were assessed between December 1996 and August 2011. The cutoff dates were chosen as to not overlap with the sample analyzed by Seto and Lalumière (2001). Portions of the subject pool in the present study overlap with those described in previous studies that compared offender groups, evaluated the validity of the phallometric procedure, or examined different research questions (Barbaree, Blanchard, & Langton, 2003; Blanchard, 2011; Blanchard & Barbaree, 2005; Blanchard, Klassen, Dickey, Kuban, & Blak, 2001; Blanchard, Kuban, et al., 2006, 2009; Blanchard et al., 2012; Blanchard, Lykins, et al., 2009; Lykins et al., 2010; Seto et al., 2006). Data from other questionnaire items and psychological testing were not analyzed or reported here.

The dataset included 3,343 men: 286 nonoffenders, 432 sex offenders with adult victims only (aged 17 or older), 179 sex offenders with victims aged 15 or 16 (referred to as “teen victims” herein), and 2,365 offenders who had committed a contact or noncontact crime involving one or more children younger than 15. The nonoffenders and offenders with only adult victims were included as a comparison group to demonstrate the discriminant validity of the phallometric test scores for age interests and to establish a priori specificity cutoff scores for the phallometric pedophilia index.

Eighty-one cases were excluded due to lack of victim information. Another 499 men who underwent phallometric testing were excluded due to invalid phallometric responses. Offenders who had committed child pornography offenses as their only sexual offense (n = 204) were also excluded because they could not be scored on the original SSPI. Forty-two nonoffenders were excluded because they were referred by a legal source, suggesting a legal issue involving problematic sexual behavior of some kind (e.g., an allegation of sexual contact with a child or use of child pornography). Finally, 40 nonoffenders and 13 offenders against adults were excluded because they self-reported a sexual interest in pubescent or prepubescent children but had no (known) child victims.

The final sample of 2,464 men comprised 166 (7%) nonoffenders, 249 (10%) sex offenders against adults, 149 (6%) sex offenders against teens, and 1,900 (77%) sex offenders against children. The sample mean age at the time of assessment was 38.3 years (SD = 13.5). Among the sex offenders, the average age at first sexual offense was 28.8 years (SD = 13.0), and the average age at the most recent sexual offense was 35.0 years (SD = 13.2). Eighty percent of the sample were Caucasian, 2% were Asian, 6% were Black, 4% were East Indian or Pakistani, 3% were Aboriginal, 1% were Pacific Islander or Filipino, and 5% were of another ethnic background. Two thirds (67%) had previously been married or in a common-law relationship, and 89% had no known major mental disorder diagnosis. In this sample, 45% were referred by probation/parole, 16% were referred by a lawyer, 27% were referred by a correctional institution, and 12% were self-referred. The results of the phallometric assessment were used clinically; however, offenders had the right to refuse release of the assessment results to the referral source. There was no confidentiality agreement regarding the clinical assessment information, which might also be released by subpoena.

The procedures described in this study were approved by the Center’s Institutional Review Board. Offenders were asked to review a consent form that asked whether their clinical assessment data could be used in future research studies and to sign it if they agreed. It was clearly explained that they could refuse to consent and it would have no bearing on their clinical assessment, their refusal would not be held against them, and that the referral source would not know whether they consented to their data being used in future research studies. Only data from individuals who consented for their data to be used for research are included here.

Group assignment

File and interview information about sexual offending history were used to categorize participants as follows: offenders against children younger than 15, which could include noncontact offenses such as exhibitionism; offenders against older teenagers (15-16 years old); and offenders against adults (17 and older). Group assignment was based on the youngest known victim. For example, an offender in the offenders against children group may have offended against adolescents or adults as well, but was assigned to this group if he had at least one child victim below age 15. The Screening Scale for Pedophilic Interests–2 (SSPI-2) was examined using the 1,900 offenders with at least one child victim. These offenders were randomly assigned to the development or validation sample, 950 cases each, using the random selection command in SPSS.

Measures

Semi-structured interview

Data were coded from a detailed, semi-structured sexual history interview used since 1995 to gather information on sexual offenses (including those not known to authorities), sexual partner preferences, psychiatric and medical diagnoses, and the ages of onset of a variety of sexual behaviors (e.g., age at first sexual intercourse). Information was also gathered from documents such as police or court reports that accompany referrals. The participant’s sexual offending history was checked against these official records. Thus, the SSPI-2 was coded from a combination of self-report and official records. When there was a discrepancy, the source providing more information about sexual offense victims was used. For example, if an offender denied committing a sexual offense resulting in a criminal charge, then the official record was used; if the offender admitted to a sexual offense that was previously undocumented then the self-report was used, given that some sexual offenses do not result in criminal charges and our assumption that offenders would not report victims that did not actually exist during a clinical assessment that might influence criminal justice decisions.

Nine percent of the offenders against children reported a moderate to strong sexual interest in prepubescent children. We retained offenders who denied any sexual interest in children if they had committed sexual offenses involving children, given that denial of such interests is common, especially in forensic contexts. Interview responses were used to exclude nonoffenders or sex offenders against adults who self-reported pedophilic or hebephilic sexual interests, however, to provide a homogeneous comparison group. Hebephilia represents sexual attraction to pubescent, rather than to prepubescent, children; there is overlap between pedophilic and hebephilic sexual interests, but they are distinct (Blanchard, Lykins, et al., 2009). During the interview, individuals were asked to rank various age and sex categories from 1 (strongest sexual interest) to 5 (no sexual interest). Men were excluded from the comparison group if they ranked prepubescent or pubescent children from 1 to 3 on this scale, indicating moderate to strong sexual interest.

Phallometric assessment

Participants’ penile responses were assessed using the volumetric phallometry method, which records changes in penile volume during presentation of different stimulus categories. Participants were presented with seven story categories, describing sexual interactions with individuals varying in developmental stage (prepubescent, pubescent, or adult) and sex (male or female), plus a neutral category. The audio-recorded narratives were presented in the second person and present tense by a female or male actor to match the sex of the person described in the story; neutral stories were alternately read by a male or female narrator. The sexual stories described the physical attributes of the depicted person and sexual interactions involving that person. In addition, photographs that corresponded to the stories were presented on three adjacent projector screens, which showed a nude model in full length (front and rear views) and a close-up view of the model’s genital area. Neutral stories were accompanied by photographs of landscapes.

The phallometric test took 1 to 2 hrs to complete and involved four blocks of seven trials. Each trial included three nude models, presented for 18 s each (total trial duration was 54 s). Change in blood volume was recorded 4 times per second. The interval between trials varied, depending on how long it took the participant to return to baseline penile volume. Steps were taken during phallometric assessment to discourage feigning of sexual interests. The phallometric technician was able to communicate with the offender during the assessment if any abnormalities were noted during the test (e.g., if the offender was attempting to increase his sexual response by voluntarily contracting his abdominal and perineal muscles: Freund, Watson, & Rienzo, 1988). A camera was aimed at the subject’s face during the assessment to ensure that he was attending to the stimuli. If he was looking away from stimuli during testing, the technician instructed him to redirect his attention to the stimuli. Finally, a neutral category (i.e., nonsexual interaction accompanied by landscape images) was included, and if the subject’s greatest response was to the neutral stimuli, his assessment data were not used, as it was assumed that they were feigning sexual interest or not aroused by any of the sexual stimuli. Lykins et al. (2010) compared the stimuli used in this phallometric protocol with a previous test using one screen with nude models walking toward the camera, and found that the present method elicited greater penile responses.

The phallometric scoring process has been extensively documented by Blanchard et al. (2001). The participant’s greatest level of change from the initial value for each of the 28 trials was calculated by examining the largest deviation from his baseline and the area under the trial’s complete curve, both of which were ipsatively transformed to z scores to account for individual differences in responsivity (Harris, Rice, Quinsey, Chaplin, & Earls, 1992). Thus, each participant had a mean response score of 0. These z scores were grouped into seven stimulus category scores by taking the average of the four scores for each of the seven stimulus categories; these were used as the continuous category scores in the present study. Lykins et al. (2010) showed that a minimum 1 cc sexual response (i.e., 1 cc increase in penile volume) was necessary to show diagnostic consistency. Thus, potential participants were excluded from analyses of phallometric data if their output index (the average of the three highest responses, regardless of stimulus category, but not including the neutral category) was below 1 cc.

The phallometric pedophilic index was calculated by subtracting the highest response to the adult categories from the highest response to the prepubescent children categories. Higher index scores indicate greater response to prepubescent children relative to adults. We did not distinguish male from female stimuli. Blanchard et al. (2001) reported that the specificity of the phallometric test used in this study was 96%; that is, very few men who had no known sexual interest or behavior involving children were diagnosed as pedophiles using a clinically derived cutoff of z = 0.25, indicating a 0.25 standard deviation higher response to child than to adult stimuli. Using this same cutoff, the sensitivity to detect pedophilia was 61% among men with three or more child victims. We examined this clinical cutoff but also examined the intuitively appealing cutoff of 0, which would indicate an equal response to prepubescent children and to adults.

SSPI-2

The SSPI-2 was constructed to assess pedophilic sexual interest among contact or noncontact sex offenders with child victims. Contact offenses are defined as sexual offenses in which there is physical contact or attempted contact with a victim. Noncontact offenses are defined as sexual offenses in which there is proximity with the victim but no (attempted) physical contact (e.g., exhibitionism). Offenders who had only committed child pornography offenses — except for those involving production of child pornography with actual children — were excluded because they had no direct or proximal contact with a child. The scoring for the SSPI-2 was similar to the scoring for the original SSPI. The SSPI-2 items were as follows: any boy victim younger than 15, more than one child victim younger than 15, any child victim below age 12, and any extrafamilial child victim younger than 15. Each of these items was again scored as present or absent, with the boy victim item initially having twice the weight of the other items. The fifth item was whether the person admitted to or was charged with a child pornography offense (possession, distribution, or production of child pornography, as defined in the Canadian Criminal Code¹). This item was also scored as present or absent. Most offenders who received a positive score on this item had charges for possession of child pornography; specific proportions were not recorded, however.

Results

Descriptive Information

Among the 1,900 sex offenders against children, approximately 30% had boy victims younger than 15, 51% had more than one victim younger than 15, 73% had a victim younger than age 12, 80% had an extrafamilial victim younger than 15, and 10% had committed a child pornography offense. The proportion with child pornography offenses might seem low, but our sample excludes those who only had committed child pornography offenses (i.e., they had no contact sexual offenses). Also, child pornography offenses were not as high a criminal justice priority until policy and legal changes occurred in the past decade (reviewed in Seto, 2013).

For offenders against children, the mean phallometric pedophilia index score was −0.36 (SD = 0.93) with acceptable skewness and kurtosis (i.e., less than |2| and less than |7|, respectively). In the sample of offenders with adult victims, the mean score on the phallometric pedophilia index was −1.03 (SD = 0.87), and for nonoffenders, the mean phallometric pedophilia index score was −1.04 (SD = 0.82), both indicating greater sexual response to adults than to children. Offenders with teen victims had a mean pedophilia index score of −0.76 (SD = 0.79), indicating greater sexual response to adults than to children.

A univariate ANOVA was used to examine group differences on the phallometric pedophilia index, F(3, 2460) = 67.35, p < .001, $η_{p}^{2}$ = .08. Results demonstrated that, as expected, nonoffenders did not differ from offenders with adult victims (p = 1.00, d = 0.01), but did score significantly less on the phallometric pedophilia index compared with offenders with child victims (p < .001, d = 0.78) or offenders against teen victims (p = .037, d = 0.35). A similar pattern was observed with the offenders against adult victims who scored significantly less on the phallometric pedophilic index compared with offenders against child victims (p < .001, d = 0.74) or offenders with teen victims (p = .025, d = 0.35). Finally, offenders with teen victims scored significantly less than offenders with child victims (p < .001, d = 0.46).

The group of nonoffenders and the group of offenders against adults were combined into a single comparison group, as they were in the original SSPI study. This combined comparison group permitted analysis of the specificity of the phallometric pedophilia index at two cutoff scores: z = 0.00 and z = 0.25. The specificity of the phallometric pedophilia index was 90% at z = 0.00 and 93% at z = 0.25; the concomitant sensitivities were 33% and 25%, respectively. These represent underestimates of sensitivity because not all offenders with child victims are pedophiles. In fact, a recent review by Seto (2013) suggested that approximately 50% to 60% of offenders with child victims are pedophiles.

As previously mentioned, Blanchard et al. (2001) found that the sensitivity of the phallometric test used in this study was 61% for sex offenders with three or more child victims. In the present sample, of those with three or more child victims below the age of 15, irrespective of the number of victims they might have in other age categories, 48% were pedophilic at the 0.00 cutoff and 37% using the 0.25 cutoff.

Development of the SSPI-2

Half (N = 950) of the offenders with child victims were used for the development phase. Spearman’s correlations were calculated among the SSPI-2 items and the phallometric pedophilia index (Table 1). As expected, the SSPI-2 total score was significantly and positively correlated with the pedophilia index, r(948) = .27, p < .001. A multiple regression analysis was conducted with the original four-item SSPI score entered first to predict the phallometric pedophilia index scores, F(1, 948) = 56.31, p < .001, R² = .06. The child pornography item was added in the second block, significantly improving the model, F(2, 948) = 53.23, p < .001, R² change = .05.

Table 1.

Intercorrelations Among SSPI-2 Items and the Pedophilia Index for the Development Sample (N = 950).

	Pedophilia index	Male victim	Multiple victims	Victim below 12	Extrafamilial victim	Child pornography
Pedophilia index	—	—	—	—	—	—
Male victim	.21***	—	—	—	—	—
Multiple victims	.19***	.28***	—	—	—	—
Victim below 12	.15***	.16***	.26***	—	—	—
Extrafamilial victim	.16***	.18***	.22***	−.05	—	—
Child pornography	.22***	.10**	.03	.03	.09**	—
SSPI-2 total score	.27***	.65***	.72***	.52***	.49***	.32***

Note. SSPI-2 = Screening Scale for Pedophilic Interests–2. Spearman’s rank-order correlations between SSPI-2 items, Pearson’s correlation for the SSPI-2 and Pedophilia Index.

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

The proportion of cases in the development sample identified as pedophilic at each value of the SSPI-2, for both cutoff scores, is reported in Table 2. The likelihood of being identified as pedophilic generally increased with greater SSPI-2 score; however, the relationship was not linear. Exploratory analysis suggested that the heavier weighting for the boy victim item (2 instead of 1 for the other items) was problematic. The bottom panel of Table 2 shows the distribution when all items are unweighted (i.e., scored 0 or 1). When using unweighted SSPI-2 items, the distribution was relatively normal, and scores on the phallometric pedophilia index increased steadily with scores on the SSPI-2. Using unweighted items, the mean score on the SSPI-2 for child sexual offenders was 2.47 (SD = 1.18), and the median score was 2, with scores ranging from 0 to 5.

Table 2.

Proportion of Offenders Identified as Pedophilic at Different Cutoff Scores, by SSPI-2 Score in the Development Sample (N = 950).

SSPI-2 score	n	Mean pedophilia index	Cutoff = 0 (%)	Cutoff = 0.25 (%)
0	30	−0.96	3	3
1	225	−0.56	23	18
2	199	−0.45	31	23
3	29	0.15	46	45
4	100	−0.67	21	14
5	320	−0.06	46	37
6	47	0.58	78	66

0	30	−0.96	3	3
1	195	−0.65	19	14
2	261	−0.49	29	21
3	259	−0.20	39	32
4	175	0.06	54	43
5	30	0.76	79	73

Note. The top panel represents the sensitivity when male victim is scored 0 or 2. The bottom panel represents the sensitivity for the development sample when all items are scored 0 or 1. SSPI-2 = Screening Scale for Pedophilic Interests–2.

To illustrate the magnitude of the association between SSPI-2 and penile responding, someone with a score of 2 or 3 on the SSPI-2 had between 2.4 and 2.6 times the odds of exceeding one of the pedophilia index cutoff scores we examined than someone with a score of 0 or 1; the odds ratio was between 6.1 and 6.9 for those who scored a 4 or 5 on the SSPI-2, again in comparison with those who scored 0 or 1.

Another way of describing the association between SSPI-2 score and penile responding is through examination of positive and negative predictive values. If SSPI-2 scores of 4 or 5 were considered to be evidence of pedophilic responding, and pedophilic responding was indicated by a pedophilia score greater than 0, the positive predictive value was 57.8% and the negative predictive value was 71.2%.

A receiver operator characteristic (ROC; Swets, 1988) analysis was used to establish the classification accuracy of the SSPI-2 using the two different index cutoff scores to identify a case as pedophilic. All ROC analyses were conducted using MedCalc, including the area under the curve (AUC) comparisons. At a cutoff of 0.0, which indicates an equal response to prepubescent children than to adults, the AUC value was 0.68, p < .001. This can be interpreted as a randomly selected pedophilic offender against children having a higher score on the SSPI-2 that a randomly selected nonpedophilic offender against children 68% of the time. In comparison, the original SSPI AUC value was 0.64, p < .001, at the cutoff of 0.0. AUC comparisons were conducted using the Delong Test, and the difference was significant, Z = 4.40, p < .001, indicating the SSPI-2 produced a significantly greater AUC compared with the original SSPI. At a cutoff score of 0.25, which was the clinical cutoff used by Blanchard et al. (2001), the SSPI-2 AUC value was 0.68, p < .001, and the SSPI AUC was 0.63, p < .001. The Delong Test was significant for the AUC comparison, Z = 4.42, p < .001.

ROC analysis was also conducted on the ability of the original SSPI and the SSPI-2 to predict self-reported moderate to strong sexual interest in children. The Delong Test comparing the original SSPI, AUC = 0.73, p < .001, and the SSPI-2, AUC = 0.78, p < .001, was significant, Z = 4.40, p < .001, suggesting that the SSPI-2 was significantly better than the SSPI at distinguishing those who admitted to sexual interest in prepubescent children. This can be compared with an AUC of 0.84 (p < .001) for the phallometric pedophilia index to predict self-reported sexual interest in children.

A logistic regression addressed whether the SSPI-2 added incrementally to the pedophilia index in its association with self-reported sexual interest in children. In the first block, the pedophilia index was entered; the model was significant, χ²(1) = 116.23, p < .001, R² = .27. With every unit increase on the pedophilia index, there was a 4.62 times greater odds of the individual self-reporting moderate to high sexual interest in children. In the second block, the SSPI-2 was added to the model and explained additional variance, χ²(1) = 166.28, p < .001, R² change = .10. The odds ratios were 3.39 and 2.58 for the pedophilia index and SSPI-2, respectively.

Validation Sample

In the validation sample, SSPI-2 was scored using unweighted yes/no scoring given the problems identified in the development sample and the strong performance of unweighted items. Spearman’s correlations were again calculated for the SSPI-2 items and the phallometric pedophilia index (Table 3). The SSPI-2 total score was significantly and positively correlated with the pedophilia index, r(948) = .25, p < .001. A multiple regression analysis was conducted with the four-item SSPI score entered first, F(1, 948) = 24.39, p < .001, R² = .03. The SSPI-2 child pornography item was then added, significantly improving the model, F(2, 948) = 17.12, p < .001, R² change = .01.

Table 3.

Intercorrelations Among SSPI-2 Items and the Pedophilia Index for the Validation Sample (N = 950).

	Pedophilia index	Male victim	Multiple victims	Victim below 12	Extrafamilial victim	Child pornography
Pedophilia index	—	—	—	—	—	—
Male victim	.21***	—	—	—	—	—
Multiple victims	.14***	.29***	—	—	—	—
Victim below 12	.14***	.16***	.29***	—	—	—
Extrafamilial victim	.12***	.17***	.30***	−.004	—	—
Child	.10**	.02	.02	−.05	.03	—
pornography SSPI-2	.25***	.63***	.77***	.53***	.52***	.24***

Note. SSPI-2 = Screening Scale for Pedophilic Interests–2. Spearman’s rank-order correlations between SSPI-2 items, Pearson’s correlation for the SSPI-2 and Pedophilia Index.

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

The proportions of cases identified as pedophilic at each possible value of the SSPI-2, for the two index cutoff scores, are reported in Table 4. There was a steady increase in likelihood of being identified as pedophilic as SSPI-2 score increased. To again illustrate the magnitude of this effect, someone with a score of 2 or 3 on the SSPI-2 had between 1.6 and 2.2 times the odds of exceeding one of the pedophilia index cutoff scores than someone with a score of 0 or 1; the odds ratio was between 4.4 and 5.7 for those who scored a 4 or 5 on the SSPI-2, again in comparison with those who scored 0 or 1. The positive and negative predictive values for a SSPI-2 score of 4 or 5 and a pedophilia index cutoff of 0, corresponding to the same analysis for the development sample, were 53.2% and 73.9%, respectively.

Table 4.

Proportions of Offenders Identified as Pedophilic at Two Different Cutoff Scores, by SSPI-2 Score for the Validation Sample (N = 950).

SSPI-2 score	n	Mean pedophilia index	Cutoff = 0 (%)	Cutoff = 0.25 (%)
0	42	−0.62	20	12
1	193	−0.64	20	11
2	242	−0.51	28	22
3	270	−0.43	30	20
4	187	0.05	53	41
5	16	0.18	60	50

Note. All items are scored 0 or 1. SSPI-2 = Screening Scale for Pedophilic Interests–2.

ROC analyses were again used. At a cutoff of 0.0, the AUC value was 0.63, p < .001, compared with an AUC of 0.59 for the original SSPI, p < .001. The Delong Test for the AUC comparison was significant, Z = 4.32, p < .001, indicating the SSPI-2 produced a significantly greater AUC. At a cutoff score of 0.25, the SSPI-2 AUC value was 0.65, p < .001, whereas the original SSPI AUC was 0.61, p < .001. The Delong test was again significant, Z = 3.87, p < .001. ROC analysis was also conducted on the ability of the original SSPI and the SSPI-2 to predict self-admitted pedophilic interest. The Delong Test comparing the original SSPI AUC = 0.73, p < .001, and the SSPI-2, AUC = 0.78, p < .001, was significant, Z = 3.31, p = .001.

Discussion

In this study, we developed and validated the SSPI-2, a revision of the SSPI, incorporating a new item to capture variance associated with child pornography offending. All five SSPI-2 items were positively and significantly correlated with a phallometric index of sexual arousal to children relative to adults, though the child pornography item had lower intercorrelations with the other items as well as with the phallometric pedophilia index. These results are consistent with the idea that there are different ways to express pedophilic sexual interests (see Seto, 2013).

The SSPI-2 produced AUCs in the moderate effect size range for distinguishing those who responded above a designated pedophilia index cutoff of 0.0 or the 0.25 clinical cutoff examined by Blanchard et al. (2001). These results suggest that the SSPI-2 can be useful as a structured measure of pedophilic sexual interests, based on criminal history, when phallometric testing is not available (e.g., laboratories are not accessible, an individual refuses to be assessed or produces uninterpretable responses, analyzing archival data). The positive predictive values were lower than the negative predictive values in both samples, suggesting that a high score should be followed by further assessment of pedophilic sexual interests to reduce false positives, especially given the potentially serious consequences of being identified as pedophilic.

The SSPI is an acceptable substitute for phallometric test results in the Sex Offender Risk Appraisal Guide, a validated risk assessment measure for adult male sex offenders (Quinsey, Harris, Rice, & Cormier, 2006). We are also aware that the SSPI is routinely used in the clinical assessment of pedophilic sexual interests among sex offenders with child victims (e.g., Arkansas Department of Corrections, Missouri Department of Corrections, New York State Office of Mental Health). The current results suggest that the SSPI-2 might be clinically useful for the same purpose. The results of a logistic regression analysis indicate that both the SSPI-2 and phallometric testing results are associated with self-reported pedophilic interests. It is plausible that the combination of these two indicators might have incremental predictive validity for sexual recidivism as well.

The moderate correlation between SSPI-2 scores and the phallometric pedophilia index should be interpreted cautiously. The pedophilia index is taken from a single psychophysiological assessment that occurred fairly late in the sexual offending career of some participants, sometimes when the stakes were high (e.g., presentencing assessment, to determine treatment placement). The SSPI-2 score is based on behaviors that may have occurred many years before and over a long period in some cases. The fact that these two very different types of assessment are reliably correlated with each other speaks to the probative value of each.

Other behavioral correlates or indicators of pedophilia might add further to the discriminative or criterion-related validity of the five SSPI-2 items. One could test whether types of sexual offending acts (fondling vs. digital or penile penetration), having other types of victims, age of onset of sexual offending against children, or duration of sexual offending could add to the SSPI-2. In addition, there may be factors that suggest the absence of pedophilic sexual interests, such as number of adult sexual partners, sexual offenses against adults, or being in a long-term romantic relationship with an adult. Nonetheless, the five SSPI-2 items have the advantage of being readily available from official records, not dependent on self-report, and relatively easy to score. The incremental gain from adding more items has to be evaluated against the costs in terms of lower reliability, missing information, or more complex scoring.

Limitations

Given that analyses relied on archival data and, for the most part, on officially recorded information about sexual offense histories, we had incomplete data on the sexual behavior of the individuals in this study. It is possible that some offenders had child victims or involvement with child pornography of which we are not aware. Such individuals would have SSPI-2 scores that underestimate their actual pedophilic sexual interests.

One limitation of our child pornography offending variable is that it did not make a distinction between individuals who were charged for offenses involving images of adolescents who were below the age of 18, thus meeting the legal definition of child pornography in Canada or the United States, and those charged for offenses involving images of prepubescent or pubescent children, where the question of pedophilia or hebephilia would be more germane. We suspect that the contribution of child pornography offending would be stronger if we had additional information about the nature of the child pornography content. Recently, for example, we demonstrated that having greater interest in images of boys relative to girls is associated with a greater likelihood of sexual recidivism (Seto & Eke, 2015). At the same time, the most common child pornography images are of prepubescent children, and the large majority of child pornography offenders possess images of prepubescent children (reviewed in Seto, 2013).

Despite being the gold standard, phallometric testing can be subject to minimal or invalid responding in the laboratory or attempts to fake (e.g., by suppressing sexual arousal to a preferred category). Phallometric assessment of sexual interests is viewed as the gold standard in the assessment of pedophilia because it is objective, has good sensitivity and specificity, and predicts sexual recidivism. The specific phallometric protocol used in the present study is well standardized and one of, if not the, most well-researched in the phallometric literature. Other objective measures of sexual interest show promise, including measures based on viewing time and neuroimaging, and future research could evaluate the SSPI-2 against these measures (e.g., Mokros et al., 2010).

Conclusions and Future Directions

As we have noted, the SSPI-2 has several advantages as a measure of sexual interest in children: The SSPI-2 can be reliably coded using self-report and file information about sexual offense history; it is positively and significantly correlated with a phallometric index of sexual arousal to children; and the SSPI-2 has incremental validity with phallometrically assessed sexual arousal to children in explaining variance in self-reported pedophilic sexual interests. Potential disadvantages are that both self-reported and official information about offense history can be incomplete (e.g., if offenders do not reveal undetected sexual offenses) and, as we suggested in the development of its predecessor, the SSPI, offense history might not accurately reflect pedophilic interests for individuals who have not had sufficient opportunity. At this time, we recommend the SSPI-2 for clinical and research purposes, as a measure of pedophilic sexual interests that could be used in conjunction with, or as a substitute for, self-reported sexual interests in children and phallometric testing of sexual age interests.

It would be helpful to cross-validate the SSPI-2 on other large, independent samples of potentially pedophilic individuals with data on phallometric responses. A related line for further research would examine agreement across different assessment methods and measures, as has been done in comparing clinician diagnosis of pedophilia with phallometric testing and with the SSPI (e.g., Moulden, Firestone, Kingston, & Bradford, 2009). Another important question for future research is to examine whether the SSPI-2 would predict sexual recidivism in longitudinal studies, as did the original SSPI (Helmus et al., 2014; Seto et al., 2004). We recently completed a study showing that the SSPI-2 significantly predicted sexual recidivism in a sample of sex offenders deemed to be unsuitable for civil management in New York state (Seto, Sandler, & Freeman, 2015).

Footnotes

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Michael C. Seto declared that this submission was reviewed independent of him as an Editor In Chief.

Funding

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

Notes

References

Babchishin

K. M.

Nunes

K. L.

Hermann

C. A.

(2013). The validity of Implicit Association Test (IAT) measures of sexual attraction to children: A meta-analysis. Archives of Sexual Behavior, 42, 487-499. doi:10.1007/s10508-012-0022-8

Banse

Schmidt

A. F.

Clarbour

(2010). Indirect measures of sexual interest in child sex offenders: A multimethod approach. Criminal Justice and Behavior, 37, 319-335. doi:10.1177/0093854809357598

Barbaree

H. E.

Blanchard

Langton

C. M.

(2003). The development of sexual aggression through the life span: The effect of age on sexual arousal and recidivism among sex offenders. Annals of the New York Academy of Sciences, 989, 59-71.

Blanchard

(2011). Misdiagnoses of pedohebephilia using victim count: A reply to Wollert and Cramer (2011). Archives of Sexual Behavior, 40, 1081-1088. doi:10.1007/s10508-011-9835-0

Blanchard

Barbaree

H. E.

(2005). The strength of sexual arousal as a function of the age of the sex offender: Comparisons among pedophiles, hebephiles, and teleiophiles. Sexual Abuse: A Journal of Research and Treatment, 17, 441-456. doi:10.1177/107906320501700407

Blanchard

Klassen

Dickey

Kuban

M. E.

Blak

(2001). Sensitivity and specificity of the phallometric test for pedophilia in nonadmitting sex offenders. Psychological Assessment, 13, 118-216. doi:10.1037/1040-3590.13.1.118

Blanchard

Kolla

N. J.

Cantor

J. M.

Klassen

P. E.

Dickey

Kuban

M. E.

Blak

(2007). IQ, handedness, and pedophilia in adult male patients stratified by referral source. Sexual Abuse: A Journal of Research and Treatment, 19, 285-309. doi:10.1177/107906320701900307

Blanchard

Kuban

M. E.

Blak

Cantor

J. M.

Klassen

Dickey

(2006). Phallometric comparison of pedophilic interest in nonadmitting sexual offenders against stepdaughters, biological daughters, other biologically related girls, and unrelated girls. Sexual Abuse: A Journal of Research and Treatment, 18, 1-14. doi:10.1007/s11194-006-9000-9

Blanchard

Kuban

M. E.

Blak

Cantor

J. M.

Klassen

P. E.

Dickey

(2009). Absolute versus relative ascertainment of pedophilia in men. Sexual Abuse: A Journal of Research and Treatment, 21, 431-441. doi:10.1177/1079063209347906

10.

Blanchard

Kuban

M. E.

Blak

Klassen

P. E.

Dickey

Cantor

J. M.

(2012). Sexual attraction to others: A comparison of two models of alloerotic responding in men. Archives of Sexual Behavior, 41, 13-29. doi:10.1007/s10508-010-9675-3

11.

Blanchard

Lykins

A. D.

Wherrett

Kuban

M. E.

Cantor

J. M.

Blak

. . . Klassen

P. E.

(2009). Pedophilia, hebephilia, and the DSM-V. Archives of Sexual Behavior, 38, 335-350. doi:10.1007/s10508-008-9399-9

12.

Canales

D. D.

Olver

M. E.

Wong

S. C. P.

(2009). Construct validity of the Violence Risk Scale—Sexual Offender version for measuring sexual deviance. Sexual Abuse: A Journal of Research and Treatment, 21, 474-492. doi:10.1177/1079063209344990

13.

Freund

Watson

Rienzo

(1988). Signs of feigning in the phallometric test. Behaviour Research and Therapy, 26, 105-112.

14.

Hanson

R. K.

Bussière

M. T.

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

15.

Hanson

R. K.

Morton-Bourgon

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

16.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

Chaplin

T. C.

Earls

(1992). Maximizing the discriminant validity of phallometric assessment data. Psychological Assessment, 4, 502-511. doi:10.1037/1040-3590.4.4.502

17.

Helmus

O’ Ciardha

Seto

M. C.

(2015). The screening scale for pedophilic interests (SSPI): Construct, predictive, and incremental validity. Law and Human Behavior, 39, 3-43. doi:10.1037/lhb0000099

18.

Kalmus

Beech

A. R.

(2005). Forensic assessment of sexual interest: A review. Aggression and Violent Behavior, 10, 193-217.

19.

Lykins

A. D.

Cantor

J. M.

Kuban

M. E.

Blak

Dickey

Klassen

P. E.

Blanchard

(2010). The relation between peak response magnitudes and agreement in diagnoses obtained from two different phallometric tests for pedophilia. Sexual Abuse: A Journal of Research and Treatment, 22, 41-57. doi:10.1177/1079063209352094

20.

McGrath

R. J.

Cumming

G. F.

Burchard

B. L.

Zeoli

Ellerby

(2009). Current practices and emerging trends in sexual abuser management. Brandon, VT: Safer Society.

21.

Mokros

Dombert

Osterheider

Zappala

Santilla

(2010). Assessment of pedophilic sexual interest with an attentional choice reaction time task. Archives of Sexual Behavior, 39, 1081-1090. doi:10.1007/s10508-009-9530-6

22.

Moulden

H. M.

Firestone

Kingston

Bradford

(2009). Recidivism in pedophiles: An investigation using different diagnostic methods. The Journal of Forensic Psychiatry & Psychology, 20, 68-701. doi:10.1080/14789940903174055

23.

Prosecutorial Remedies and Other Tools to End the Exploitation of Children Today (PROTECT) Act of 2003. (2003). www.GovTrack.us. Retrieved from http://www.govtrack.us/congress/bills/108/s151

24.

Quinsey

V. L.

Harris

G. T.

Rice

M. E.

Cormier

C. A.

(2006). Violent offenders: Appraising and managing risk (2nd ed.). Washington, DC: American Psychological Association.

25.

Seto

M. C.

(2008). Pedophilia and sexual offending against children: Theory, assessment, and intervention. Washington, DC: American Psychological Association.

26.

Seto

M. C.

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

27.

Seto

M. C.

Cantor

J. M.

Blanchard

(2006). Child pornography offenses are a valid diagnostic indicator of pedophilia. Journal of Abnormal Psychology, 115, 610-615. doi:10.1037/0021-843X.115.3.610

28.

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 Behavior, 39, 416-429.

29.

Seto

M. C.

Harris

G. T.

Rice

M. E.

Barbaree

H. E.

(2004). The Screening Scale for Pedophilic Interests predicts recidivism among adult sex offenders with child victims. Archives of Sexual Behavior, 33, 455-466. doi:10.1023/B:ASEB.0000037426.55935.9c

30.

Seto

M. C.

Lalumière

M. L.

(2001). A brief screening scale to identify pedophilic interests among child molesters. Sexual Abuse: A Journal of Research and Treatment, 13, 15-25. doi:10.1177/107906320101300103

31.

Seto

M. C.

Murphy

W. D.

Page

Ennis

(2003). Detecting anomalous sexual interests among juvenile sex offenders. In Prentky

R. A.

Janus

E. S.

Seto

M. C.

(Eds.), Annals of the New York academy of sciences, Vol. 989: Understanding and managing sexually coercive behavior (pp. 118-130). New York: New York Academy of Sciences.

32.

Seto

M. C.

Sandler

J. C.

Freeman

N. J.

(2015). The revised Screening Scale for Pedophilic Interests-2 (SSPI-2): Predictive and concurrent validity. Manuscript submitted for publication.

33.

Swets

J. A.

(1988). Measuring the accuracy of diagnostic systems. Science, 240,1285-1293.