Victim Crossover Index Offending Patterns and Predictors in a Portuguese Sample

Abstract

Sex offenders who cross over in victims’ age, gender and relationship usually have a greater number of victims, which is associated with sexual recidivism. This investigation aimed to examine the prevalence of crossover index offending in Portugal, and to explore the predictive ability of sociodemographic and criminological variables on this outcome. A retrospective sample of 247 male individuals incarcerated for sex offenses in a Portuguese prison was drawn from official records. From those offenders with multiple victims (n = 94), 48% had victims of different age categories, 10% had both gendered victims, and 12% had intrafamilial and extrafamilial victims. Comparative statistics and logistic regressions were able to identify variables that distinguished noncrossover and crossover offenders and that predicted crossover, respectively. While likely underestimates of the prevalence of victim crossover, these findings are compared to previous international studies and provide a better understanding of the phenomenon.

Keywords

crossover sex offender polymorphism victim selection Sexual Abuse

Classification schemes, or typologies, usually discriminate sex offenders based on the type of offense perpetrated (e.g., rape, child sexual abuse, child pornography), and on the victim’s age (child or adult), gender (male or female), and relationship with the perpetrator (intrafamilial or extrafamilial; Bickley & Beech, 2001; Blasko, 2016; Lanning, 2010; Robertiello & Terry, 2007). These typologies help guide officials on risk assessment, treatment plans, and supervision strategies (Kleban, Chesin, Jeglic, & Mercado, 2013). For example, intrafamilial abusers may be subjected to lower supervision upon release based on the assumption that they were only a risk for their own family members. Extrafamilial child molesters may be restricted from frequenting places populated by children based on their preferred victim target. And rapist treatment programs may only feature modules on attitudes toward adult women.

However, there is a particular flaw underlying these typologies, as they assume that sex offenders are stable in their choice of victims and usually perpetrate the same type of offenses. Although the majority may remain relatively stable in their victim selection (Cann, Friendship, & Gonza, 2007; Guay, Proulx, Cusson, & Ouimet, 2001; Sim & Proeve, 2010), a significant portion cross over and are more heterogeneous in their offending patterns (Bourke & Hernandez, 2009; English, Jones, Patrick, & Pasini-Hill, 2003; Heil, Ahlmeyer, & Simons, 2003; Weinrott & Saylor, 1991).

Victim choice crossover or polymorphism occurs when an individual has victims of different age categories, of both genders or of disparate relationships to him (Heil et al., 2003). Studies that have analyzed sex offenders’ victims throughout their criminal career have indicated relative stability in their subsequent choice of victims of the same gender, but less so regarding disparate age categories (Guay et al., 2001; Kleban et al., 2013; Sim & Proeve, 2010).

Prevalence of victim crossover appears to vary based on methodology, as official records seem to substantially underestimate the rates when compared to self-reports (Heil et al., 2003; Weinrott & Saylor, 1991). Table 1 summarizes the prevalence rates of victim crossover across International studies. The significant prevalence rates of crossover sex offenders have lead Kleban and colleagues (2013) to argue that classifying a sex offender based on their current victim selection may underestimate their level of recidivism risk and facilitate their access to a wider range of potential victims. In addition, crossover offenders have higher recidivism scores in some actuarial measures (Cann et al., 2007; Olver, Wong, Nicholaichuk, & Gordon, 2007). Although, Stephens, Seto, Goodwill, and Cantor (2018) argue that this is related to their greater number of sexual offense victims, as that is associated with increased risk for sexual recidivism (R. L. Hanson & Bussière, 1998).

Table 1.

Summary of Victim Crossover Prevalence Rates Across International Studies.

Sources	Methodology	Crossover domain
Sources	Methodology	Age	Gender	Relationship
Weinrott and Saylor (1991) Washington, USA	Incarcerated rapists and child molesters (N = 99) Self-reports	32% of rapists had a child victim; and 12% of child molesters had an adult female victim	n/a	34% both extrafamilial and incest
Heil, Ahlmeyer, and Simons (2003) Colorado, USA	Incarcerated sex offenders (N = 489) Official records	7% (both <15 years and ⩾15 years)	9%	7%-17% various relations
Heil, Ahlmeyer, and Simons (2003) Colorado, USA	Self-reports and polygraph	70% (both <15 years and ⩾15 years)	36%	24%-80% various relations
English, Jones, Patrick, and Pasini-Hill (2003) (Three states, USA)	Incarcerated sex offenders receiving treatment (N = 180) Polygraph reports	39% (both child and adults)	31%	64% both extrafamilial and incest
Cann, Friendship, and Gonza (2007) England and Wales	Discharged serious sex offenders (N = 1.345) Official records	8% (both <16 years and adults)	9%	14% both intra and extrafamilial
Sim and Proeve (2010) South Australia	Sex offenders against children and adolescents at community-based treatment (N = 128) Official treatment reports	48% (preschool ⩽5 years, 6-12 years, and ⩾13 years)	22%	26% both intra and extrafamilial
Kleban, Chesin, Jeglic, and Mercado (2013) USA	Incarcerated sex offenders in a prison (N = 789, 89% against children and 6% against adults exclusively). Official conviction records Index offense	14% (child <13 years, adolescent 13-17 years, and adult >17 years)	13% had victims of both genders, 14	13% family member, acquaintance, or stranger
Kleban, Chesin, Jeglic, and Mercado (2013) USA	Criminal career	40% (child, adolescent, and adult)	21%	48% various relations

Accordingly, research has tried to understand what makes an individual cross over in their sexual offending patterns. Victims’ age crossover may be partially explained by the fact that, for some men, strong attraction to adults can coexist with a strong attraction to children, with the inverse being true as well, although it is more common for men to be attracted to adjacent age categories/levels of maturity (Bailey, Hsu, & Bernhard, 2016). An investigation conducted by Michaud and Proulx (2009) found that crossover offenders present less exclusive phallometric profiles than rapists or child molesters (i.e., they were sexually attracted to scenes that involved both consensual and nonconsensual sex with adult females, as well as to prepubescent females). Another study found that above 64% of the age crossover offenders in their sample scored high on a psychopathy measure, particularly on the factor associated with having a ruthless and callous personality (Porter et al., 2000). The authors suggested that those individuals may victimize different types of victims when the opportunities arise or when they feel bored.

Easier access to victims’ and available opportunities may also help explain those who abuse both intra and extrafamilial victims. Cann and colleagues (2007) observed that the longer an individual had been offending, the more likely they were of having engaged in relationship crossover. Heil and colleagues (2003) also posited that, although sex offenders may have a preferred victim pool, preference may change over time and may be expanded when that type is unavailable.

Regarding gender crossover, it appears that sexual attraction may be more fluid for those who are especially attracted to prepubescent children (Bailey et al., 2016). A study of 362 sex offenders in the United States (Levenson, Becker, & Morin, 2008) indicated that those with preschool-aged (⩽6 years old) victims were three times more likely to target both males and females, than those with older victims. Blanchard and colleagues (2012) postulated that this might be due to the shared similarities between prepubescent boys and girls, owing to their absence of secondary sex characteristics.

Research on what distinguishes victim crossovers from noncrossovers, and its predictors is still scarce (Cann et al., 2007). Besides their greater number of sexual offense victims (Sim & Proeve, 2010; Stephens et al., 2018), crossover offenders exhibit more convictions for offenses (including sexual), and have been shown to be younger at the time of their first conviction for a sexual offense and older when discharged from custody for their index offense (Cann et al., 2007). These two age-related variables were also able to significantly predict victim crossover, in a negative and positive way, respectively (Cann et al., 2007). Furthermore, Levenson and coworkers (2008) tested the predictive ability of substance abuse, antisocial personality disorder, psychopathy, pedophilia, major mental illness, having a victim younger than 6 years and being sexually abused as a child, on the probability of a sex offender having crossed over in victims’ gender, specifically. The authors found that the only variables that contributed to that outcome were having a victim younger than 6 years old and having a major mental illness. To our knowledge, no other variables have been studied in regards to victim crossover.

We speculate that other sociodemographic and criminological variables may contribute to the outcome of being a crossover offender, due to their general relation to other offending outcomes. For instance, marital status, employment, and substance abuse have been found to predict sexual offense versatility (McGloin, Sullivan, Piquero, & Pratt, 2007), while history of prior offenses has been associated with higher recidivism (R. L. Hanson & Bussière, 1998; R. K. Hanson & Morton-Bourgon, 2005). Denial of the offenses perpetrated has also been studied in relation to recidivism risk, but has achieved mixed results, as it appears that it may only be a significant risk factor for certain subgroups of sex offenders, although a pattern has yet to emerge (Mann, Hanson, & Thornton, 2010).

Victim crossover has yet to be studied outside Anglo-Saxon countries (i.e., Australia, Canada, United Kingdom, United States) and this is the first study using a sample from a Mediterranean country, namely Portugal. Anglo-Saxon and Mediterranean countries differ in various aspects, for example, in their cultural values and levels of intimacy (Ardila, 2013; Wierzbicka, 2003). Portugal is also one of the safest countries in the world (Institute for Economics & Peace, 2017) and, although Canada and Australia are also highly ranked, there is a larger disparity to the United Kingdom and the United States. In this sense, it is important to research offending patterns in other countries not previously studied and compare them to the existing literature for similarities and differences.

The present investigation seeks to fill this gap and provide information on the prevalence of victim crossover index offending using a Portuguese sample of incarcerated sex offenders. In addition, as an exploratory study, our purpose is to investigate the relationship between some sociodemographic and criminological variables on victim crossover in general and, specifically, on age crossover at index offenses. Understanding of the characteristics that distinguish sex offenders that abuse a greater number and variety of victims, as is the case of crossover offenders, is of paramount importance due to its related risk to sexual recidivism (R. L. Hanson & Bussière, 1998). Ultimately, knowledge of victim crossover predictors may help to improve the precision of risk assessment and aid in the management, supervision, and treatment plans of sexual offenders.

Method

The Portuguese Judicial System

According to the Portuguese General Office of Prison Services and Social Rehabilitation (Direção Geral de Reinserção e Serviços Prisionais [DGRSP], 2014a), at the end of 2014, 203 male individuals were serving sentences for rape and 275 for child sexual abuse/dependent youngster sexual abuse in the Portuguese Judicial System.¹ No official data regarding inmates incarcerated for other sex crimes is available, as their numbers were less significant at the time and, consequently, were not discriminated in the official reports.

In the Portuguese context, a child sexual abuse conviction can lead to sentences ranging from 1 to 10 years and a large portion of sex offenders can have their prison sentences suspended when its maximum length is equal to or less than 5 years (among other conditions). In those cases, they are instead required to comply with certain duties, like paying a fine and having to participate in sexology treatment programs. Unfortunately, there is no official data available that quantifies the number of individuals with suspended sentences by crime typology for the years 2014 and 2015 (DGRSP, 2014b, 2015). A recent newspaper article, citing the Portuguese Minister of Justice, mentioned that, between 2015 and 2016, there were 696 convictions for child sexual abuse and that 523 of those (i.e., about 75%) had their sentences suspended (“Tribunal de Braga dá como provado,” 2018). The article does not discriminate between male and female offenders, nor does it mention statistics for other sex crimes.

Participants

A retrospective convenience sample was drawn from a database of 261 male adult sex offender cases, collected from September 2014 to October 2015 at a Portuguese prison facility. These cases were from all the sex offenders that were incarcerated in the prison at that time, which was previously chosen due to its high concentration of sex offenders in its incarcerated population. Hence, the database featured, among others, 54% of the individuals incarcerated for child/dependent youngster sexual abuse and rape in the Portuguese Judicial System at the time (DGRSP, 2014a). The remaining incarcerated sex offenders in the country were spread across other Portuguese regions, which hindered our access to them.

In addition, the sample does not include those with suspended sentences. Therefore, it is not representative of all sex offenders in the Portuguese context. Rather, it reflects only those who committed more serious sexual offenses and that were serving prison sentences in that facility.

In 14 cases there was no information regarding the victims’ characteristics (i.e., age, gender, and relationship to the offender), which did not allow for crossover analyses, so these were subsequently removed from the sample. The four noncontact offenders (i.e., three convicted for child pornography and one for indecent exposure exclusively) were also excluded from the sample, following the same procedure used by previous authors (Heil et al., 2003; Levenson et al., 2008).

The final sample consisted of 243 male individuals serving sentences for contact sexual offenses¹: 37% (n = 89) for child sexual abuse, 37% (n = 91) for rape, 7% (n = 18) for dependent youngster sexual abuse, 9% (n = 22) for mixed contact sexual offenses (at least two of the above), and 10% (n = 23) were dual offenders of child pornography and contact sexual offenses. Their sentences stretched from 28 to 300 months (M = 110.18, SD = 53.34).

Participants’ age at the time of the assessment ranged from 21 to 76 (M = 45.15, SD = 11.38). The majority, 78% (n = 189), were Portuguese and 19% (n = 45) were of African descent, mainly from former Portuguese colonies; there were also three individuals from other European countries and six South Americans (predominantly from Brazil). Other sample descriptive information can be seen in Table 2.

Table 2.

Sample Description (N = 243).

Categories	Frequency	Valid percentage
Age at index offense M (SD)	37.24 (11.32)
Education level (years) M (SD)	5.93 (3.25)
Marital status
Single	65	26.7
Divorced/separated/widower	96	39.6
Married/common law	82	33.7
Labor situation
Unemployed	72	29.1
Employed	158	64.0
Retired	17	6.9
Social support (in prison)
Absence	43	20.5
Inconsistent	43	20.5
From intimate partner, family, and/or friends	124	59.0
Substance abuse
Absence	168	69.1
Alcohol	67	27.6
Drugs	8	3.3
Past criminal convictions
Absence	122	51.7
Presence	114	48.3
Nonviolent	82	73.2^a
Violent	53	47.3^a
Sexual	35	31.3^a
Behavior at index offense
Grooming/seduction	64	26.7^a
Intimidation/threats	130	54.2^a
Physical coercion	168	70.0^a
Attitudes about the index offense
Denies responsibility	91	41.0
Justifies/minimizes	106	47.7
Accepts responsibility	25	11.3

Percentages of cases in multiple response sets (the same participant can have more than one subcategory present so the total exceeds 100%).

Data Collection Procedure

To acquire permission to consult the offenders’ records, formal authorizations were previously requested to the Portuguese General Office of Prison Services and Social Rehabilitation, as well as the prison facility that housed the participants. The William James Center for Research approved this study. The American Psychological Association’s standards on the ethical treatment of participants were followed.

Data were extracted from the prison records of all individuals incarcerated for sex offenses in that facility by the leading researcher. The files generally contained details about the persons’ sociodemographics (date of birth, marital status, profession, labor situation, education level); psychiatric reports (diagnoses of substance abuse problems, cognitive deficits, personality traits, symptomology or paraphilias associated with mental disorders); index offense sentencing decisions, which featured details about the circumstances surrounding it (i.e., victim characteristics and offender’s behaviors at the offense); prison adjustment reports, which included management officials’ assessments on the quality of the offenders’ current social support and their attitudes toward the offense; and criminal conviction records. The latter only lists the type of crime for which the person was convicted and does not provide further information about the victims’ characteristics. As such, the data did not allow for analyses featuring historical crossover offending, but only regarding the index offense. With the exception of information pertaining to substance abuse, the psychiatric records were generally sparse in individual diagnoses (n_{personality traits} = 15, n_{symptomatology} = 14, n_{cognitive deficits} = 5, n_paraphilias = 4 [categories not mutually exclusive]). Hence, we were not able to include these variables in the analyses. Thus, our choice of variables was due, in part, to the available information in the participants’ files.

Dependent Variable

Victim crossover at index offenses

Victim crossover was detected solely at index offense for individuals with multiple victims, similarly to Kleban and colleagues’ (2013) first analysis, due to lack of details about past offenses’ victim characteristics. Like in that same study, the index offense was determined to be the sex offense for which the offender was incarcerated at the time of the assessment (between 2014 and 2015). While the majority of the victims at the index offenses spanned through time, there may have been a small portion that resulted from a single assault incident.

Victim crossover at index offense was detected when an offender had multiple victims of different age categories, both gendered victims or of disparate relationships. Victims’ age was coded into four categories based on the level of maturity, according to the Tanner stages (a scale of physical development of the body): prepubescent (⩽10 years old; Tanner 1), pubescent (11-14 years; Tanner 2-3), postpubescent (15-17 years; Tanner 4), and adults (⩾18 years; Tanner 5). While the fifth Tanner stage includes individuals of 17 years, we opted to use 18 years as that is the legal age of adulthood in Portugal. Intrafamilial victims were those with a biological or sociolegal relation (e.g., stepfather, step-uncle or step-grandfather) to the offender, and extrafamilial included acquaintances or strangers.

Independent Variables

Sociodemographic predictors

We include two interval and four categorical variables that measure various sociodemographic characteristics: offender’s age (in years) at index sexual offense; educational level (in number of years); marital status (coded as 1 = single, 2 = divorced/separated/widower, or 3 = married/living in common law); labor situation (coded as 1 = unemployed, 2 = employed, or 3 = retired); social support while in prison (coded as 1 = absence, 2 = inconsistent, or 3 = from intimate partner, family and/or friends). With the exception of social support while in prison, all other variables are related to the time of the index offense. More information about the coding of social support in prison can be found in Table A1 in Appendix. These variables have been generally able to discriminate between sex offender types (Bard et al., 1987; Craissati & Beech, 2004). Cann and colleagues (2007) have also found that age at first conviction for a sexual offense and age at first court appearance were both able to predict crossover offending.

Criminological predictors

We include interval (frequency) and categorical (coded as 0 = absence and 1 = presence) variables of nonviolent, violent, and sexual criminal convictions prior to the index offense; use of grooming/seduction, intimidation/threats, or physical coercive behaviors at index offense (coded as 0 = absence and 1 = presence); and two categorical variables for substance abuse (coded as 0 = absence, 1 = alcohol, or 2 = drugs) and attitudes about the index offense (coded as 1 = denies responsibility, 2 = justifies/minimizes, or 3 = accepts responsibility). Table A1 in Appendix contains more details about the coding rules. A history of past criminal behavior has been associated with higher recidivism (R. L. Hanson & Bussière, 1998; R. K. Hanson & Morton-Bourgon, 2005). Some studies have found that denial of the offense was related to recidivism in certain subgroups of sex offenders, although the patterns have differed across studies (Mann et al., 2010). An offender’s behavior at index offense has been shown to discriminate between different types of sex offenders (Bard et al., 1987; Craissati & Beech, 2004). And, while substance abuse, specifically, was unable to contribute to gender crossover versus noncrossover (Levenson et al., 2008), it has not been studied in relation to other types of victim crossover.

Statistical Analyses

As an exploratory study, all analyses were implemented after examination of the data. In accordance with Simmons, Nelson, and Simonsohn’s (2012) suggestion, we acknowledge that we report how the sample size was determined, all data exclusions, all manipulations, and all measures in the study.

The characterization of victim crossover at index offense was made through a frequency and percentage study. The sample for the analyses that featured victim crossover included only those contact offenders that had multiple index victims (n = 94) at index offense. In the sole instance of victims’ age crossover, the sample comprised 92 individuals, as there were two cases that did not specify this information and were omitted from the analyses that featured this subject.

Independent samples t tests were performed to examine the significance of the interval independent variables (i.e., offender’s age at index offense, years of education, frequency of past nonviolent, violent, and sexual convictions) on noncrossover (Group 1) and crossover (Group 2) offenders with multiple victims at index offenses. The normality assumption was validated with univariate Shapiro–Wilk tests (p ⩾ 0.05) and skewness and kurtosis (sk < 2, ku < 7) observations. The homogeneity of variances’ assumption was assessed in each group with the Levene’s test based on the median (p ⩾ 0.05). A significance level of α = 0.05 was considered for statistically significant differences. Cohen’s d was provided as a measure of the effects’ size, as described in Marôco (2011).

Chi-square tests were also executed to analyze the independence of the categorical sociodemographic (i.e., marital status, labor situation, social support in prison) and criminological variables (i.e., substance abuse; grooming, intimidation, or physically coercive behaviors at index offense; attitudes about the index offense) between noncrossover (Group 1) and crossover (Group 2) offenders with multiple victims at index offenses. A Type I (α) error probability of 0.05 was considered for the chi-square analyses. Cramer’s V was included as a measure of the effects’ size.

A logistic regression analysis was performed so as to assess the significance of sociodemographic (i.e., age at index offense, years of education, marital status, labor situation, social support in prison) and criminological predictors (i.e., presence of past nonviolent, violent and sexual criminal convictions; substance abuse; grooming, intimidation or physically coercive behaviors at index offense; attitudes about the index offense) on the probability that a sex offender with multiple victims at index offenses would be classified either as a noncrossover or crossover of any domain (age, gender, and relationship), as described in Marôco (2011). A second logistic regression was executed with the same variables to evaluate the probability that a sex offender had crossed over in victims’ age, specifically. The lesser cases of gender and relationship crossover did not allow for valid separate logistic regressions. Logistic regression assumptions were validated through graphical analysis of the residuals and diagnostics of influential cases (studentized residuals ⩽ .2, Cook’s distance < 1; Marôco, 2011). Four outlier observations were removed from the general crossover model, as it improved its significance and adjustment quality. In the victims’ age crossover model, two outliers were removed to increase its quality. Moreover, in both logistic regressions there were 20 missing values related to the independent variables that were not included in the analyses. Thus, the final sample for the victim crossover (all three domains) logistic regression was 70 and the one for the age crossover regression was 72.

All the analyses were performed using the IBM SPSS Statistics software (v. 23.0, SPSS Inc., Chicago, Il., USA).

Results

Prevalence of Victim Crossover at Index Offense

In our sample of contact offenders, 39% (n = 94) had multiple victims at index offense. Figure 1 presents the prevalence of the detected crossover within this sample, based on victim characteristics at index sex offense. Totally, 10 individuals crossed over in more than one category, but there was no one who crossed over in all three domains.

Figure 1.

Venn diagram with victim crossover at index offenses’ sample (n = 55).

Of note, five individuals with multiple index victims that were presently categorized as noncrossovers had a criminal record for past sexual offenses, although there was no data regarding those victims’ characteristics. Hence, from a lifetime perspective, it is possible that some of these individuals could have crossed over in any domain.

Roughly 75% of those who crossed over in the age domain at index offense chose victims in adjacent age categories. Figure 2 demonstrates the distribution of the targeted victims’ ages, within the age crossover sample (n = 44).

Figure 2.

Representation of the distribution of the targeted victims’ ages within the age crossover at index offenses’ sample (n = 44).

All the gender crossover offenders had prepubescent victims, but some of them were also age crossovers: besides prepubescent victims, four also had pubescent ones, another had a postpubescent, and one had an adult victim.

To compare our findings to those of other studies of victims’ age crossover, we conducted three supplemental analyses on the prevalence of age crossover at index offense. In the first one, we used three age categories (<13 years, 13-17 years, and >17 years) and found rates of age crossover of 30% (n = 28). In the second one, we compared two other different age categories (>16 years and ⩾16 years) and found crossover rates of 14% (n = 13). And in the third one, we compared two other age categories based solely on the legal age of adulthood (>18 years and ⩾18 years) and observed age crossover rates of approximately 8% (n = 7).

Comparison of Index Crossover and Noncrossover Offenders

Noncrossover (n = 38) and crossover offenders (n = 54) in victims’ choice (including the age, gender, and relationship domains) at index offense were compared on a range of sociodemographic and criminological variables. Independent samples t tests revealed no statistically significant differences between the two groups for age at index offense, t(89) = –.182, p = .856, d = –.04; years of education, t(90) = .038, p = .970, d = .01; and frequency of past nonviolent, t(50.298) = 1.894, p = .064, d = .45, or sexual, t(89) = –.578, p = .565, d = –.12, convictions. The only significant variable was past violent convictions, which had a moderate effect, t(51.227) = 2.008, p = .050, d = .48. The results indicated that noncrossover offenders had a higher frequency of prior violent convictions (M = 1.11, SD = 1.776, SEM = .292), as opposed to crossover offenders (M = .46, SD = .985, SEM = .134).

Chi-square analyses with Fisher’s exact test showed no statistically significant differences between noncrossover and crossover offenders regarding marital status, χ²(2) = 1.943; p = .379; N = 92; ϕ_C = .145; labor situation, χ²(2) = 2.538, p = .281, N = 92, ϕ_C = .166; social support in prison, χ²(2) = .701, p = .704, N = 82, ϕ_C = .092; use of grooming, χ²(1) = 2.241, p = .134, N = 91, ϕ_C = .157, intimidation, χ²(1) = .008, p = .929, N = 91, ϕ_C = .009, or physical force, χ²(1) = .081, p = .777, N = 91, ϕ_C = .030, at index offense; and attitudes about the index offense, χ²(2) = .911, p = .634, N = 84, ϕ_C = .104. However, substance abuse was dependent on the type of offender, χ²(2) = 8.480, p = .014, N = 92, ϕ_C = .304. In this sense, a higher incidence of alcohol abuse was observed in the crossover group (n_D = 14; 25.9%), when compared to the noncrossovers (n_D = 3; 7.9%); and all the drug abusers were noncrossover offenders (n_D = 3; 7.9%), as opposed to the crossovers (n_D = 0; 0%).

Predictors of Victim Crossover at Index Offense

A logistic regression was performed to ascertain the effects of a wide range of sociodemographic and criminological variables on the likelihood that the offenders with multiple victims at index offense had crossed over in any domain (age, gender, and relationship). After removal of four outliers, the adjusted model was statistically significantly, G²(18) = 34.212, p = .012, χ²_Wald (8) = 10.52, p = .23. The model explained between 39% (R²_CS) and 53% (R²_N) of the variance in crossover offending and correctly classified 81% of the cases. It also exhibited high sensitivity (87%) and specificity (72%), as well as good discriminant ability, receiver operating characteristic curve (ROC) c = .882, SE = .043, p < .001, 95% confidence interval (CI) [.798, .965]. Table 3 shows the coefficients of the logistic regression model with all significant predictors, with absence being the reference outcome category in dichotomous variables. Being divorced/separated/widower, being employed prior to imprisonment and using alcohol increased the probability of being a crossover offender. However, increasing age at index offense was associated with a reduction in the chance of having crossed over. No other variables influenced the model probability.

Table 3.

Logistic Regression Model on the Prediction of Victim Crossover Index Offending (n = 70).

	β (SE)	χ²_Wald	df	Exp(β)	95% CI for Exp(β)
	β (SE)	χ²_Wald	df	Exp(β)	Lower	Upper
Age at index offense	−0.081* (0.041)	3.918	1	0.922	0.851	0.999
Marital status
Single^a		4.859	2
Divorced/separated/widower	2.131* (1.020)	4.363	1	8.424	1.140	62.232
Married/common law	1.995 (1.390)	2.061	1	7.351	0.483	111.979
Years of education	0.016 (0.130)	0.016	1	1.017	0.788	1.312
Labor situation
Unemployed^a		7.191	2
Employed	2.925* (1.110)	6.943	1	18.630	2.115	164.077
Retired	2.338 (1.690)	1.915	1	10.363	0.378	284.372
Social support (in prison)
Absence^a		3.183	2
Inconsistent	1.786 (1.275)	1.961	1	5.964	4.90	72.642
From intimate partner, family, friends	0.078 (1.156)	0.005	1	1.081	0.112	10.426
Substance abuse
Absence^a		5.743	2
Alcohol	3.656* (1.526)	5.743	1	38.702	1.946	769.581
Drugs	−21.405 (28,222.593)	0.000	1	0.000	0.000
Past nonviolent convictions	0.464 (0.990)	0.220	1	1.590	0.229	11.062
Past violent convictions	−0.972 (0.909)	1.142	1	0.378	0.064	2.249
Past sexual convictions	−0.528 (1.027)	0.264	1	0.590	0.079	4.418
Grooming/seduction during offense	1.859 (0.953)	3.801	1	6.416	0.990	41.575
Intimidation/threats during offense	−0.104 (0.822)	0.016	1	0.901	0.180	4.510
Physical force during offense	1.546 (0.970)	2.538	1	4.692	0.700	31.431
Attitudes about the offense
Denies responsibility^a		2.615	2
Justifies/minimizes	−0.703 (0.978)	0.517	1	0.495	0.073	3.364
Accept responsibility	2.235 (1.847)	1.463	1	9.343	0.250	349.156

Note. Victim crossover included the age, gender, and relationship domains. There were 20 missing values in the analysis and four outliers that were removed to improve the model’s quality. CI = confidence intervals.

Category of reference

p < .05.

A new statistically significant model was adjusted, G²(7) = 21.557, p = .003, χ²_Wald (8) = 4.292, p = .83, R²_CS = .22, R²_N = .30, with only the previously significant variables: age at index offense, βage at index offense = –.015, χ²_Wald (1) = .355, p = .551, OR = .986, 95% CI for OR [.939, 1.034]; marital status, βDivorced/separated/widower = .861, χ²_Wald (2) = 1.944, p = .163, OR = 2.366, 95% CI for OR [.705, 7.935]; labor situation, βEmployed = 1.478, χ²_Wald (2) = 5.536, p = .019, OR = 4.384, 95% CI for OR [1.280, 15.018]; and substance abuse, βalcohol abuse = 2.611, χ²_Wald (2) = 5.232, p = .022, OR = 13.610, 95% CI for OR [1.453, 127.473]. In the adjusted model, age at index offense and being divorced/separated/widower were no longer significant. Being employed and abusing alcohol were the only consistently significant predictors of victim crossover in general.

A second logistic regression was performed to analyze the effects of the same independent variables on the likelihood that the offenders with multiple victims crossed over in victims’ age, specifically. Following the removal of two outliers, the adjusted model was statistically significantly, G²(18) = 38.598, p = .003, χ²_Wald (8) = 6.799, p = .558. The model explained between 42% (R²_CS) and 56% (R²_N) of the variance in age crossover and correctly classified 78% of the cases. It also exhibited high sensitivity (77%) and specificity (79%), as well as good discriminant ability (ROC c = .880, SE = .039, p < .001, 95% CI [.7,803, .956]). Table 4 shows the coefficients of the logistic regression model with all predictors, with absence being the reference outcome category in dichotomous variables. Increasing years of education (prior to imprisonment), abusing alcohol, and using grooming/seduction strategies and/or physically coercing the victim at index offense improved the probability of having crossed over in victims’ age. No other variables influenced the model probability.

Table 4.

Logistic Regression Model on the Prediction of Victims’ Age Crossover at Index Offenses (n = 72).

	β (SE)	χ²_Wald	df	Exp(β)	95% CI for Exp(β)
	β (SE)	χ²_Wald	df	Exp(β)	Lower	Upper
Age at index offense	−0.068 (0.039)	3.129	1	5.558	0.779	39.665
Marital status
Single^a		2.787	2
Divorced/separated/widower	1.416 (0.922)	2.358	1	4.121	0.676	25.112
Married/common law	1.620 (1.256)	1.664	1	5.053	0.431	59.224
Years of education	0.303* (0.133)	5.158	1	1.354	1.042	1.759
Labor situation
Unemployed^a		1.183	2
Employed	1.009 (0.951)	1.126	1	2.743	0.425	17.697
Retired	1.383 (1.708)	0.656	1	3.987	0.140	113.268
Social support (in prison)
Absence^a		2.497	2
Inconsistent	1.722 (1.211)	2.021	1	5.595	0.521	60.093
From intimate partner, family and/or friends	0.583 (1.137)	0.263	1	1.792	0.193	16.639
Substance abuse
Absence^a		11.310	2
Alcohol	4.845** (1.441)	11.310	1	127.122	7.549	214.552
Drugs	−23.107 (26,574.452)	0.000	1	0.000	0.000
Past nonviolent convictions	1.715 (1.003)	2.926	1	5.558	0.779	39.665
Past violent convictions	−1.384 (0.929)	2.222	1	0.250	0.041	1.546
Past sexual convictions	5.551 (0.993)	0.308	1	0.576	0.082	4.036
Grooming/seduction during offense	1.981* (0.950)	4.349	1	7.251	1.127	46.669
Intimidation/threats during offense	0.356 (0.743)	0.230	1	1.428	0.333	6.127
Physical force during the offense	1.922* (0.887)	4.697	1	6.835	1.202	38.868
Attitudes about the offense
Denies responsibility^a		2.316	2
Justifies/minimizes	0.133 (0.848)	0.024	1	1.142	0.217	6.016
Accepts responsibility	2.332 (1.574)	2.194	1	10.298	0.471	225.321

Note. There were 20 missing values in the analysis and two outliers that were removed to improve the model’s quality. CI. = confidence intervals.

Category of reference.

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

A new statistically significant model was adjusted, G²(5) = 27.943, p = < .001, χ²_Wald (8) = 8.923, p = .35, R²_CS = .27, R²_N = .36, with only the previously significant variables: education, βeducation= .151, χ²_Wald (1) = 4.145, p = .042, OR = 1.163, 95% CI for OR [1.006, 1.346]; substance abuse, βalcohol abuse = 2.863, χ²_Wald (2) = 10.831, p = .001, OR = 17.512, 95% CI for OR [3.183, 96.337]; grooming behaviors, βgrooming= 1.455, χ²_Wald (1) = 5.871, p = .015, OR = 4.286, 95% CI for OR [1.321, 13.908]; and physical coercion, βphysical coercion = 1.188, χ²_Wald (1) = 3.853, p = .050, OR = 3.281, 95% CI for OR [1.002, 10.743]. In the newly adjusted model, all variables maintained their significant patterns.

Discussion

Victim crossover in sex offenses has been associated with having a larger number of victims (Sim & Proeve, 2010; Stephens et al., 2018), which is a risk factor for sexual recidivism (R. L. Hanson & Bussière, 1998). It is also a field lacking extensive research (Cann et al., 2007), particularly in non-Anglo-Saxon countries. The aim of this investigation was to contribute to a better understanding of this occurrence, using an incarcerated sample of male sex offenders in Portugal. Due to insufficient information regarding past victims’ characteristics, we were only able to focus on crossover index offending. Hence, we first examined the prevalence of victim crossover at index offense. We then explored group comparisons of noncrossover and crossover index offenders on a range of sociodemographic and criminological variables. Finally, we analyzed those variables’ predictive ability on victim index crossover in general and, specifically, on age crossover.

Despite only assessing victim crossover at index offense, our findings shared some resemblances to other International research that also relied on official criminal conviction records. By design, our study shares the closest similarities to Kleban and colleagues’ (2013) investigation that also analyzed victim crossover at index offense.

In our Portuguese sample of 243 incarcerated sexual offenders, 39% (n = 94) had multiple index victims, which is similar to the 35% observed by Kleban and colleagues (2013) in the United States. Among those, we found that 10% had victims of both genders and 12% had disparate relationships (intrafamilial and extrafamilial) with them. Our gender and relationship crossover rates are akin to those of Kleban and colleagues (2013)—13% in both domains—and to those of a previous study in the United Kingdom (Cann et al., 2007)—9% gender and 14% relationship crossover. Thus, it appears that gender and relationship crossover prevalence is relatively stable across different Western countries, which is in accordance to past research that used transitional matrices to analyze offenders’ progression of victim selection (Guay et al., 2001; Sim & Proeve, 2010).

However, we observed a much higher age crossover rate (48%), when compared to the 8% in Cann and colleagues’ (2007) and the 14% of Kleban and coworkers’ (2013) studies. This difference may be partially explained by our use of four categories for victims’ ages (⩽10 years old, 11-14 years, 15-17 years, ⩾18 years). Cann and coworkers (2007) used only two categories (<16 years and adults), whereas Kleban and colleagues (2013) used three (<13 years, 13-17 years, >17 years). To compare our sample’s results with theirs, we conducted additional analyses on age crossover prevalence, using both studies’ age categories. When using the two categories, <16 years and adults, we found age crossover rates of 14%; and 30% when using the aforementioned three (<13 years, 13-17 years, >17 years). Hence, we still found double the amount of age crossover prevalence in our Portuguese sample. We wonder if this is a characteristic specific to our country or if we used a somewhat different definition of index offense than Kleban and colleagues’ (2013). Ours include multiple victims from offenses that spanned through time, but there may be also a small portion that originated from a single assault incident, which we cannot confirm at this time. However, as those authors did not specify the source of their index offense multiple victims, we cannot be sure if the disparities are related to that. Further investigation that discriminates these two dynamics in victim crossover is needed. Replication of these findings in other Mediterranean countries is also desirable.

When analyzing the age crossovers’ choice of victims, we found that 75% of those were from adjacent age categories, which is a common occurrence (Bailey et al., 2016). Consequently, 25% of the age crossovers had victims of remote (e.g., prepubescent and adult) and broader (e.g., pubescent, postpubescent, and adult) age categories. Previously, Michaud and Proulx (2009) observed that crossover offenders had less exclusive phallometric profiles than rapists or child molesters, so our findings are not surprising.

On another note, one might question if crossover in adjacent age categories is a valid indicator of actual crossover, as age delimitation does not necessarily mean different levels of development, even if we did use the Tanner stages of physical maturity and its age-related gradients (prepubescents, pubescents, postpubescents, and adults). Age crossover rates were influenced by the number of categories used to operationalize the phenomenon. As the number of categories increased so did the rates. Hence, we found lower prevalence of age crossover when using fewer categories (<16 years and ⩾16 years, and child vs. adolescent vs. adult), as stated above. The smallest rate (8%) was detected when making comparisons based solely on the legal age of adulthood (>18 years and ⩾18 years), although we believe this one disregards physical maturity level and may not be an appropriate measure for age crossover analyses. Even with fewer categories, our age crossover rates were still twice the amount of the aforementioned studies.

As in Levenson and coworkers’ (2008) US study, our results corroborate that gender crossover appears to be more common for those individuals who target prepubescent victims. In our sample, all those who had both gendered victims (n = 9) abused prepubescent ones, which also supports Bailey and colleagues’ (2016) findings that gender attraction is more fluid for those especially attracted to prepubescents. Blanchard and colleagues (2012) previously theorized that this may be due to the fact that prepubescent boys and girls share more physical similarities and do not have their secondary sex characteristics developed. Of significance, six of those nine gender crossovers also had victims of other age categories besides the aforementioned, with the majority (n = 4) having pubescent ones as well. Only two offenders targeted victims of more remote levels of maturity, demonstrating less exclusive victimization patterns.

Although our prevalence rates of victim crossover at index offense are similar to other studies that used official conviction records, it is important to note that our methodology may be seriously undervaluing the occurrence (Heil et al., 2003; Kleban et al., 2013; Weinrott & Saylor, 1991). For instance, after including the offenders’ past sexual convictions together with their index ones, Kleban and colleagues (2013) found much higher prevalence rates of crossover offending in their sample. Presently, almost half of the sex offenders in our sample had at least one past conviction. Of those, 32% were of a sexual nature. As we had no information on the exact characteristics of those past victims, we were not able to provide a more detailed analysis of crossover across the offenders’ criminal careers. Thus, the actual crossover rates of sex offenders in Portugal may be much higher. This leads us to agree with Kleban and colleagues (2013), that a significant portion of sex offenders are not stable in their victim selection. Consequently, officials should bear this in mind when deciding on the risk level of a sex offender. If based solely on whether the offender targeted adults, or children related to them at the time, they may be underestimating their risk level and facilitating their access to a wider range of potential victims.

Past investigations have called for studies to ascertain the individual characteristics and situational factors related to victim crossover, to aid officials with essential information for effective sex offender risk assessment, treatment, and management (e.g., Cann et al., 2007; Kleban et al., 2013). Our study was able to make a contribution in this field by identifying some of the variables that distinguish crossover and noncrossover offenders in general. We conducted exploratory group comparisons on a range of sociodemographic and criminological variables, which the literature had shown were related to criminal behavior (R. L. Hanson & Bussière, 1998; R. K. Hanson & Morton-Bourgon, 2005; McGloin et al., 2007), and that were generally able to distinguish between sex offender types (Bard et al., 1987; Craissati & Beech, 2004). Our findings revealed that a higher percentage of crossovers abused alcohol at index offense, whereas all the drug abusers were noncrossovers and they also had a higher frequency of past violent criminal convictions. An earlier study from the United Kingdom (Cann et al., 2007) observed that age at first conviction for a sexual offense and age at discharge from custody for the index offense successfully discriminated crossover and noncrossovers. While not equivalent measures, age at index offense did not differentiate between these two groups in our present sample. Overall, both groups shared more similarities than differences, as they were also nondistinguishable in years of education; in frequency of past nonviolent or sexual convictions; in marital status or labor situation prior to imprisonment; social support in prison; use of grooming, intimidation or physically coercive behaviors at index offense; and in attitudes about the index offense.

We also tested the predictive ability of the aforementioned sociodemographic and criminological variables on victim crossover at index offense in general and, specifically, on age crossover. Being young at index offense was one of the variables that successfully predicted victim crossover in general, but not in the age domain, particularly. Previously, Cann and colleagues (2007) found that age at discharge from custody and age at first sexual conviction were able to predict age crossover, but were strongest when predicting relationship crossover. As mentioned above, as we used different measures of age, our results are noncomparable. Nevertheless, our finding that age at index offense increased the likelihood of being a victim crossover in general, and the results from those authors, lead us to wonder if an offender’s age has a bigger impact in relationship crossover specifically. As we had too few crossovers in the relationship domain we were not able to test this hypothesis. Further research is needed to clarify this matter.

We observed that being divorced/separated/widower and being employed predicted victim crossover in general. Previous research stated that changes in life’s circumstances can create opportunities to engage in some types of offenses (Gottfredson, 2005; Horney, Osgood, & Marshall, 1995). Perhaps not having a family or being employed allows crossovers to have more contact with a wider assortment of victims and facilitates their offending opportunities. However, age crossover, specifically, was predicted by higher frequency of education and not by marital status or labor situation. Better understanding of this dynamic is necessary.

We also found that alcohol abuse contributed to being a crossover offender in general and, particularly, in the age domain. It is possible that alcohol abuse disinhibits crossovers into perceiving the existence of more criminal opportunities and also permits them to interact with more deviant peers (McGloin et al., 2007). Nonetheless, this result is different from Levenson and colleagues’ (2008), who did not find that substance abuse influenced gender crossover. While they only examined gender crossover, we analyzed age and victim crossover in general. Perhaps, gender crossover, specifically, is not influenced by substance abuse. We cannot be sure of this assumption, as we did not test it due to the low number of gender crossovers in our sample. Moreover, we assessed victim crossover at index offense based on the conviction records of incarcerated offenders, whereas they collected data from psychological evaluation reports of sexually violent predators in civil commitment. Different methodology may help explain our contrasting results.

Our findings also indicated that age crossovers, specifically, tended to use grooming/seduction behaviors at index offense, and were also likely to resort to physical coercion when the former was not a successful strategy. It is possible that grooming tactics allows crossovers to engage with a wider variety of potential victims.

While frequency of past violent criminal convictions was able to discriminate between crossovers and noncrossovers in the group comparisons, as stated above, having a past violent conviction was not able to contribute to either outcome in the logistic regressions. This discrepancy may be due to the different measures used in both analyses. In the former, we used an interval variable of frequency of past violent convictions, but in the logistic regressions we converted it into a categorical variable coded for absence or presence. Hence, it seems that the presence of past violent convictions alone does not influence victim crossover, but only having a higher frequency of those types of convictions.

Of importance, both regressions’ confidence intervals were too wide, which reflects the fact that our sample size is small. The adjusted model of age crossover with only the originally significant variables was able to replicate the same patterns, which gives us a little more confidence regarding our interpretation of the findings. However, the adjusted model of general victim crossover was unable to fully replicate the original results, with being employed and abusing alcohol as the only significant predictors. Therefore, we strongly recommend replication of our findings with a larger sample.

Study Limitations

The present study has several limitations that should be considered when interpreting our results. First, our data were retrieved from official conviction records, which has been shown to substantially underestimate crossover rates (Heil et al., 2003; Weinrott & Saylor, 1991). Moreover, while the majority of previous research has examined victim crossover through an offender’s criminal career, we only focused on the index offenses (i.e., sex offenses for which they were incarcerated) due to lack of information regarding their past sexual victims’ characteristics. There were 19 individuals with single index victims that were excluded from our analyses who had a prior criminal record for sex offenses. In addition, five of those categorized as noncrossovers also had victims from previous sex offenses. In both instances, some of those individuals could actually be crossovers, although we were unable to confirm this as we had no knowledge regarding their past victims’ characteristics. This not only underestimates our victim crossover prevalence rates but may have also contributed to smaller effect sizes in our group comparison analyses. Hence, the present study concludes that at least 35% of the incarcerated serious sex offenders in Portugal have crossed over in some type of victim selection.

The reduced sample size of multiple index victims also hinders the regression power of our analyses. Of note, this small sample size reflects the low number of sex offenders incarcerated in Portugal (“Tribunal de Braga dá como provado,” 2018), and our overall lower crime rates when compared to other countries. We did analyze 54% of all incarcerated sex offenders in Portugal during 2014 and 2015 (DGRSP, 2014a).

We also could not distinguish between those offenses with multiple victims that spanned through time and those that resulted from single assault incidents. All these have an impact on our findings’ replication and comparison with other International studies that used disparate methodology. Of note, our study is still comparable to Kleban and colleagues’ (2013), who also analyzed crossover at index offense based on official conviction records and did not specify if this spanned through time or if it resulted from a single assault incident.

In addition, our sample did not include sex offenders with suspended prison sentences, so our findings cannot be generalized to encompass all convicted sex offenders in the Portuguese Judicial System. Rather, they only represent those that perpetrated more serious sexual offenses and were incarcerated during 2014 and 2015 at that correctional facility. Future research should include those sex offenders with suspended sentences so as to provide a more accurate depiction of this population in general.

Another limitation lies in the fact that the coding of the variables’ information was solely done by the leading researcher, which does not allow for analysis of interrater reliability, therefore affecting the generalization of our findings. Finally, the cross-sectional nature of our study only permits the classification of these offenders as noncrossover or crossover based on the data at that point in time. Some of those classified as noncrossover offenders may indeed crossover in the future, which we would only be able to evaluate in a longitudinal study.

Implications and Future Research

Despite our data’s limitations, these findings are a first indication of sexual crossover in Portugal. In general, we did not find significant differences in the prevalence of victim crossover from preceding research in other countries that would indicate singularity in the Portuguese culture. The sole exception was age crossover which had much higher rates despite being limited to index offense information, retrieved from official records.

Officials should bear in mind that a significant portion of sex offenders will cross over in victims’ type, either due to less exclusive preferences (Michaud & Proulx, 2009) or due to a greater ability to disregard those if their preferred victim type is not accessible when the opportunities arise (Heil et al., 2003). Our results partially support these hypotheses, as we found that a subset of age crossovers assaulted victims of remote or broader age categories, and that certain life circumstances (marital status and labor situation) were able to predict victim crossover at index offense.

Future research should replicate our findings using a more robust measure of victim crossover (across criminal careers while relying on self-reports in conjunction with official records). In addition, an examination of victim crossover among less “serious” sex offenders with suspended sentences will contribute to a better understanding of this phenomenon. We strongly recommend that future research try and replicate our findings with the same sociodemographic and criminological variables, while using a larger sample. As having victims in adjacent age categories does not necessarily reflect “actual” crossover, we suggest conducting these same analyses while operationalizing age crossover solely as having victims in nonadjacent and/or broader age categories. We also recommend analyzing the role of other psychometric and situational variables that may contribute to victim crossover in general, and to specific types of crossover (e.g., relationship). Identifying those characteristics that distinguish crossovers from noncrossovers would improve risk assessment and facilitate the development of more effective treatment plans and supervision strategies for sex offenders.

Footnotes

Appendix

Acknowledgements

The authors would like to acknowledge the assistance of the Portuguese General Office of Prison Services and Social Rehabilitation (Direção Geral de Serviços Prisionais e Reinserção Social, 2015) and the prison establishment where the study was conducted. The authors are grateful to reviewers whose comments and insight have added to this paper.

Declaration of Conflicting Interests

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

Funding

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

ORCID iD

Mariana A. Saramago

Notes

References

Ardila

(2013). The impact of culture on neuropsychological test. In Uzzell

B. P.

Ponton

Ardila

(Eds.), International handbook of cross-cultural neuropsychology (pp. 23-44). New York, NY: Psychology Press.

Bailey

J. M.

Hsu

K. J.

Bernhard

P. A.

(2016). An Internet study of men sexually attracted to children: Sexual attraction patterns. Journal of Abnormal Psychology, 125, 976-988. doi:10.1037/abn0000212

Bard

L. A.

Carter

D. L.

Cerce

D. D.

Knight

R. A.

Rosenberg

Schneider

(1987). A descriptive study of rapists and child molesters: Developmental, clinical, and criminal characteristics. Behavioral Sciences & the Law, 5, 203-220. doi:10.1002/bsl.2370050211

Bickley

Beech

A. R.

(2001). Classifying child abusers: Its relevance to theory and clinical practice. International Journal of Offender Therapy and Comparative Criminology, 45, 51-69. doi:10.1177/0306624X01451004

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

Blasko

B. L.

(2016). Overview of sexual offender typologies, recidivism, and treatment. In Jeglic

E. L.

Calkins

(Eds.), Sexual violence: Evidence based policy and prevention (pp. 11-29). Switzerland: Springer. doi:10.1007/978-3-319-44504-5_2

Bourke

M. L.

Hernandez

A. E.

(2009). The “Butner study” redux: A report of the incidence of hands-on child victimization by child pornography offenders. Journal of Family Violence, 24, 183-191. doi:10.1007/s10896-008-9219-y

Cann

Friendship

Gonza

(2007). Assessing crossover in a sample of sexual offenders with multiple victims. Legal and Criminological Psychology, 12, 149-163. doi:10.1348/135532506X112439

Craissati

Beech

(2004). The characteristics of a geographical sample of convicted rapists: Sexual victimization and compliance in comparison to child molesters. Journal of Interpersonal Violence, 19, 371-388. doi:10.1177/0886260503262078

10.

Direção Geral de Reinserção e Serviços Prisionais [General Office of Prison Services and Social Rehabilitation]. (2014a). Estatísticas do ano 2014: Reclusos condenados existentes em 31 de dezembro, segundo o sexo, os escalões de idade e a nacionalidade, por crimes. Retrieved from http://www.dgsp.mj.pt/paginas/estatisticas/estatisticas_arq_mostra.php?id=2014

11.

Direção Geral de Reinserção e Serviços Prisionais [General Office of Prison Services and Social Rehabilitation]. (2014b). Relatório estatístico anual 2014. Retrieved from http://www.dgrs.mj.pt/web/rs/estat

12.

Direção Geral de Reinserção e Serviços Prisionais [General Office of Prison Services and Social Rehabilitation]. (2015). Relatório estatístico anual 2015. Retrieved from http://www.dgrs.mj.pt/web/rs/estat

13.

English

Jones

Patrick

Pasini-Hill

(2003). Sexual offender containment: Use of the post-conviction polygraph. Annals of New York Academy of Sciences, 989, 411-427. doi:10.1111/j.1749-6632.2003.tb07322.x

14.

Gottfredson

M. R.

(2005). Offender classifications and treatment effects in developmental criminology: A propensity/event consideration. Annals of the American Academy of Political and Social Science, 602, 46-57. doi:10.1177/0002716205279937

15.

Guay

J. P.

Proulx

Cusson

Ouimet

(2001). Victim-choice polymorphia among serious sex offenders. Archives of Sexual Behavior, 30, 521-533. doi:10.1023/A:1010291201588

16.

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

17.

Hanson

R. L.

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

18.

Heil

Ahlmeyer

Simons

(2003). Crossover sexual offenses. Sexual Abuse: A Journal of Research and Treatment, 15, 221-236. doi:10.1177/107906320301500401

19.

Horney

Osgood

D. W.

Marshall

I. H.

(1995). Criminal careers in the short-term: Intra-individual variability in crime and its relation to local life circumstances. American Sociological Review, 60, 655-673. doi:10.2307/2096316

20.

Institute for Economics & Peace. (2017). Global Peace Index 2017: Measuring peace in a complex world. Retrieved from http://visionofhumanity.org/app/uploads/2017/06/GPI17-Report.pdf

21.

Kleban

Chesin

M. S.

Jeglic

E. L.

Mercado

C. C.

(2013). An exploration of crossover sexual offending. Sexual Abuse: A Journal of Research and Treatment, 25, 427-443. doi:10.1177/1079063212464397

22.

Lanning

K. V.

(2010). Child molesters: A behavioral analysis (5th ed.). Retrieved from https://www.icmec.org/wp-content/uploads/2015/10/US-NCMEC-OJJDP-Child-Molesters-A-Behavioral-Analysis-Lanning-2010.pdf

23.

Levenson

J. S.

Becker

Morin

J. W.

(2008). The relationship between victim age and gender crossover among sex offenders. Sexual Abuse: A Journal of Research and Treatment, 20, 43-60. doi:10.1177/1079063208314819

24.

Mann

R. E.

Hanson

R. K.

Thornton

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

25.

Marôco

(2011). Análise estatística com o SPSS Statistics [Statistical analysis using SPSS Statistics] (5ª Edição). Pero Pinheiro, Portugal: ReportNumber, Lda.

26.

McGloin

Sullivan

Piquero

Pratt

(2007). Local life circumstances and offending specialization/versatility: Comparing opportunity and propensity models. Journal of Research in Crime & Delinquency, 44, 321-346. doi:10.1177/0022427807302664

27.

Michaud

Proulx

(2009). Penile-response profiles of sexual aggressors during phallometric testing. Sexual Abuse: A Journal of Research and Treatment, 21, 308-334. doi:10.1177/1079063209342073

28.

Olver

M. E.

Wong

S. C.

Nicholaichuk

Gordon

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

29.

Porter

Fairweather

Drugge

Birt

Hervé

Boer

(2000). Profiles of psychopathy in incarcerated sexual offenders. Criminal Justice and Behavior, 27, 216-233. doi:10.1177%2F0093854800027002005.

30.

Robertiello

Terry

K. J.

(2007). Can we profile sex offenders? A review of sex offender typologies. Aggression and Violent Behavior, 12, 508-518. doi:10.1016/j.avb.2007.02.010

31.

Sanderson

(2004). The seduction of children: Empowering parents and teachers to protect children from child sexual abuse. London, England: Jessica Kingsley.

32.

Sim

D. J.

Proeve

(2010). Crossover and stability of victim type in child molesters. Legal and Criminological Psychology, 15, 401-413. doi:10.1348/135532509X473869

33.

Simmons

Nelson

Simonsohn

(2012). A 21 word solution. SPSP Dialogue Newsletter. doi:10.2139/ssrn.2160588

34.

Stephens

Seto

M. C.

Goodwill

A. M.

Cantor

J. M.

(2018). The relationships between victim age, gender, and relationship polymorphism and sexual recidivism. Sexual Abuse: A Journal of Research and Treatment, 30, 132-146. doi:10.1177/1079063216630983

35.

Tribunal de Braga dá como provado abusos sexuais a criança mas aplica pena suspensa a pedófilo. (2018, April 5). Expresso. Retrieved from http://expresso.sapo.pt/revista-de-imprensa/2018-04-05-Tribunal-de-Braga-da-como-provado-abusos-sexuais-a-crianca-mas-aplica-pena-suspensa-a-pedofilo#gs.9=qkSMQ

36.

Weinrott

M. R.

Saylor

(1991). Self-report of crimes committed by sex offenders. Journal of Interpersonal Violence, 6, 286-300. doi:10.1177/088626091006003002

37.

Wierzbicka

. (2003). Cross-cultural pragmatics: The semantics of human interaction (2nd ed.). Berlin, Germany: Mouton de Gruyter.