CRIMINOLOGICAL RESEARCH REPORT

Transcription

1 CRIMINOLOGICAL RESEARCH REPORT With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

2 Research Report Criminology 1 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

3 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs 2

4 Foreword... 5 B-CCENTRE Project objectives for the Criminological research track... 6 How to achieve both objectives?... 6 Executive Summary Studies Studies Conclusion Partners Applicant Organisation/Coordinator Manager Ann Mennens Researchers Prof. Dr. Stefan Bogaerts Marjolein Brusselaers Karel Demeyer Jenny Houtepen Marjolein Missler Janneke Schilder Jelle Sijtsema Publications - Abstracts Circumventing internet parental control and online pretending in primary school children: Family related differences and predictors The Effectiveness of an Intervention to promote Awareness and Reduce Online Risk Behaviour of Belgian Primary School Children The influence of parental supervision on online risk behaviour among young children: a gender perspective Privacy, Risk, Information protection and Social Network Site-Using Behaviour in a Sample of Flemish University Students From child pornography offending to child sexual abuse: A review of child pornography offender characteristics and risks for cross-over Being sexually attracted to minors: Sexual development, coping, and risk factors for offending in self-identified pedophiles Helping the police investigation: police investigation applied to sex offenders by using network sampling Full Text Articles Circumventing internet parental control and online pretending in primary school children: Family related differences and predictors The Effectiveness of an Intervention to promote Awareness and Reduce Online Risk Behaviour of Belgian Primary School Children With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

5 The influence of parental supervision on online risk behaviour among young children: a gender perspective Privacy, Risk, Information protection and Social Network Site-Using Behaviour in a Sample of Flemish University Students From child pornography offending to child sexual abuse: A review of child pornography offender characteristics and risks for cross-over Being sexually attracted to minors: Sexual development, coping, and risk factors for offending in self-identified pedophiles Training Tool for Police and Aid Workers Risks and Vulnerabilities in offences with Pedophilia and Child Pornography With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

6 Foreword The Belgian Cybercrime Centre of Excellence for Training, Research and Education is Belgium s central coordination, collaboration and knowledge sharing platform in the fight against cybercrime. B-CCENTRE coordinates research teams at various universities which collaborate across disciplines on specific cybercrime, cybersecurity and cyberforensics related topics in both fundamental and applied research activities. Together with experts from public sector and industry partners, the academic B-CCENTRE partners design and teach basic and advanced trainings on specific cybercrime topics and develop and implement awareness raising initiatives in Belgium. B-CCENTRE does not only focus its efforts on a national level, but engages in the fight against cybercrime beyond the Belgian borders through numerous contacts with similar centres abroad. B-CCENTRE is the Belgian node in the European network of Cybercrime Centres of Excellence and collaborates with the main European and international organisations dealing with cybercrime. It is sponsored by the Prevention of and Fight against Crime Programme of the European Union, under contract HOME/2010/ISEC/AG/INT-011, and co-funded by the academic partners, under the coordination of the KU Leuven. The B-CCENTRE started its activities in spring 2011 and has since launched and supported numerous activities to enhance knowledge and sharing of information and expertise related to cybercrime, digital forensics, cybersecurity, online behaviour and risks, privacy, data protection and other related topics in Belgium and beyond. This book provides an overview of the results of the criminological research performed in the frame of the EU sponsored B-CCENTRE project, 18 April November There is a similar publication on the results of the legal and of the technical research performed. These three publications are complementary to the B-CCENTRE report of activities. For further reading we refer to the B-CCENTRE publications section under the research tab on our website, where you can find the links to the articles published in the frame of the B-CCENTRE project. On the site you can also find information about B-CCENTRE partners and activities as well as an overview of relevant actors, education programmes and awareness raising activities in Belgium and other interesting leads. We wish you an interesting read and welcome your feedback on the work done. Ann MENNENS Manager B-CCENTRE KU Leuven iminds ICRI/CIR Sint-Michielsstraat 6, box 3443, BE-3000 Leuven [email protected] Disclaimer: All publications listed represent the opinion of their author(s) and do not represent the official position of the B-CCENTRE, nor of the European Commission on the topics discussed 5 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

7 B-CCENTRE Project objectives for the Criminological research track In the B-CCENTRE Project, Work package 4 (CR1 and CR2) deals with the criminological aspects of cybercrime. The studies made in this package had two objectives from criminological and forensicpsychological perspective: The first objective was to conduct fundamental and applied scientific research on cybercrime and child pornography (CR1); The second objective was to develop educational material, tools and best practices, and transferring this knowledge and application to relevant field organisations, such as forensic psychiatric clinics (policlinic and clinic) and the police (CR2). To achieve both objectives, collaboration was established between the University of Leuven; Leuvens Institute of Criminology - LinC (Belgium), and Tilburg University, Department of Developmental Psychology, research line Forensic Psychology (the Netherlands). Both universities cooperate with the police (police Rotterdam and Tilburg), justice (public prosecutor Rotterdam, Tilburg), and a self-help group for child sex offenders who didn t come into contact with the police, the justice system, and forensic psychiatric treatment centres. Further, cooperation was also organised with two forensic psychiatric centres (FPC) and with one policlinic forensic centre and daycare centre (FPC Gent and FPC de Kijvelanden, DOK) in Belgium and in the Netherlands. How to achieve both objectives? Study of online behaviour of primary school children (9-11) In preparation to reach both research objectives, a very large (n = 1437 children), and explorative, quasiexperimental and prospective research was performed in in primary schools to examine online behaviour of Flemish children between 9 and 11 years old. This research resulted in several studies. We opted to examine online behaviour of primary school children for several reasons; first, very young children are already (extremely) active on the internet; second, very young children are often seen as forgotten online victims notwithstanding they are likely to become victimized on the internet by peers or adults. In our study, we saw that about 10% of the children reported to have met in real life an unknown adult they had met on the internet, what can be seen as problematic, despite the fact that the data were obtained through self-report and children sometimes fantasize and exaggerate. Third, we talk constantly about online risk behaviour while we have little understanding of online baseline behaviour, especially of young children. We developed a questionnaire for primary school children to study normal and abnormal online behaviour. The questionnaire consists of about thirty questions concerning, e.g., age, grade, country of birth of mother and father, brothers and sisters, the use of internet, the use of a computer, tablet, Smartphone, whether they had online contact with family, friends or strangers, peers or adults; whether they ever met online strangers in real life; why they use the internet: for school, games, others; whether they were ever bullied on the internet or they had ever bullied someone else. 6 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

8 We constructed an intervention and examined, prospectively whether awareness of dangers on the internet increased and resulted in online behavioural changes (less risky behaviour). Why this study? Insight in cybercrime can only make sense if there is insight in basic behaviours on the internet (to start with young children as an important target group). A major problem is that parents, school, police and government often speak of risky online behaviour while very little is known about baseline online behaviour (of youngsters). Therefore, research on online baseline behaviour is necessary what can lead to school interventions and interventions of parents what is, however, not the primary objective of this research. This preliminary research resulted in three studies (see further in this report). Study of digital protection behaviour of young adults In addition, among a group of young adults (n = 488), all students at the Catholic University of Leuven (Computer Science, Law and Criminology), we examined how these young adults behave with online protection rules and securing their computer. Further we examined whether following (or not) protection guidelines and the way they secure their computers is related to personality traits. This resulted in one study that is reflected in this report. Study on predators of child pornography and offline sex offenders Subsequently, we conducted two important studies on predators of child pornography and offline sex offenders. A first study was a systematic review on offenders of child pornography. The main reason to perform this systematic review was, firstly the fact that child pornography offenders can be seen as a very heterogeneous group. Secondly, mental health practitioners must make a distinction between paedophiles and child molesters, and child molesters and rapists (minor and adult victims), because this distinction is very important for therapists (diagnosis, treatment indication and treatment tools), and it gives the police information about the chance of (sexual) recidivism. For the benefit of forensic treatment on the one hand, and police enforcement on the other hand, it was important to give clarity in this heterogeneous group for individual risk assessment, treatment planning and detection. A second, empirical study was conducted among a very exceptional group, namely 15 self-identified paedophiles that were recruited through a Belgian and Dutch self-help group. Through this study, we provided more insight into risk and protective factors of child sexual offending in paedophiles, and the development of paedophilic feelings in general by studying self-identified paedophiles outside clinical and forensic settings. This specific group of self-identified paedophiles can also be seen as a hidden population that is hardly known by mental health institutions and the police. Both studies are described in this report. With this study, we make the move to the last study that has this hidden population as target. Study on hidden population of sex offenders Hidden population of sex offenders is rather a large group; research shows that the chance of being caught as an (online) sex offender is rather low. This has to do with difficulties in detecting and arresting (online) sex offenders on the one hand. On the other hand the unwillingness and impossibility of victims to report to the police is relatively high (often because they are forced by the perpetrator not to report), and offers the police weak indicators to detect the suspect. This detection problem was the immediate reason to conduct the last study in which we introduced a social network tool for the police that can be very helpful in tracking down (online) sex offenders. The same tool can also be used by mental health practitioners to get more insights in 7 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

9 protective and risky network configuration of sex offenders and child pornography downloaders after discharge from prison or mental health institution. The results of this study can be found in this report. The results of these studies have led to a number of different products that can be classified into supportive tools and professional group-specific tools. 8 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

10 Supportive tools (1) Circumventing internet parental control and online pretending in primary school children: Family related differences and predictors (M.B.J. Brusselaers, S. Bogaerts, & K. Demeyer); (2) The Effectiveness of an Intervention to promote Awareness and Reduce Online Risk Behaviour of Belgian Primary School Children (J.D. Schilder, M.B.J. Brusselaers, & S. Bogaerts); (3) The influence of parental supervision on online risk behaviour among young children: a gender perspective (M.A. Missler & S. Bogaerts); (4) Privacy, Risk, Information protection and Social Network Site-Using Behaviour in a Sample of Flemish University Students (Z. Horváth, S. Bogaerts, & Karel Demeijer). Notwithstanding the importance of these studies, we see them rather as supportive; not just for parents and teachers, but also for police, public prosecution and mental health care. In the many conversations we had with these actors, it came forward that they mostly have to deal with problematic cases of victims and offenders, and that very little information is available about risky behaviour of children on social media and how an intervention can lead to behavioural change. Their interest in the behaviour of young children on the internet can be described as follow: "Children have no idea of dangers on the internet but behave in social media as if they are adults. These children can become the victims and perpetrators of tomorrow. Raising awareness at a very young age and see if an intervention actually helps is essential". Five products were developed specifically for the police and justice (detection, arrest and prosecution) and/or for forensic mental health. Professional group-specific tools (5) From child pornography offending to child sexual abuse: A review of child pornography offender characteristics and risks for cross-over (J.A.B.M. Houtepen, J.J. Sijtsema, & S. Bogaerts); (6) Being sexually attracted to minors: Sexual development, coping, and risk factors for offending in selfidentified pedophiles (J.A.B.M. Houtepen, J.J. Sijtsema, & S. Bogaerts); (7) PowerPoint presentations in Dutch, English and French that have been presented in Leuven and Rotterdam to a large group of mental health professionals, teachers and law enforcement (J.A.B.M. Houtepen, J.J. Sijtsema, & S. Bogaerts); (8) Helping the police investigation: police investigation applied to sex offenders by using network sampling (M. Spreen, S. Bogaerts, & M.A. Misller). (9) Instructions, guidelines and a step-by-step example of how to use UCINET to develop a network configuration to support police investigation and risk management of sex offenders and possessors of child pornography. These tools were presented in FPC Kijvelanden, DOK (the Netherlands) and recently in FPC Ghent (Belgium). Forensic Network Analysis is currently used in both Dutch organisations in the framework of risk management of (online) sex offenders. This tool shall also be used in spring of 2015 in the FPC Ghent. Forensic Network Analysis, more specific B-graph network sampling was presented to 4 detectives who were involved in the detection of (online) sex offenders and will hopefully be implemented in the near future by the Belgian and Dutch police. 9 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

11 Executive Summary Studies 1-4 Notwithstanding internet and social media tools offer many advantages for children and adults, there is always the risk to get hurt. Children are especially vulnerable for this, but so are adults. The internet and social media are very easily accessible and at first glance very innocent and easy in use. In other words, children and adults who have little understanding of risks or who are prone to be manipulated can very quickly become deceived on the internet without being aware of deceptions. Parents and teachers can explain children what the dangers of the internet are but very often, they not always have enough knowledge and technical skills about internet applications and risk, or have insufficient possibilities to observe the online behaviour of the children. The internet is everywhere, on computers, tablet, smartphone and is constantly within reach. Despite parental supervision, filter software and usage restrictions, some children try to avoid these support frameworks by circumventing parental control and by online pretending. We know very little of the prevalence and possible predictors of circumventing parental control and online pretending in youngsters, and whether this behaviour is different for boys and girls, differs between ages, family structure and the country of birth of the parent(s). A representative sample of 1437 Flemish primary school children was set up in cooperation with more than 45 primary schools in the Flemish part of Belgium. All children were 8-12 years old and followed the second (third year) or third grade (fifth year) of primary school. All children were written and orally questioned in class about their basic internet behaviour (f.e., number of hours on the internet per day and in the weekend, types of social media they were on, with whom they had contact on the internet, parental control and supervision, bullying behaviour on the internet, victim of bullying and knowledge of the parents of the internet). The survey took place in the schools and was coordinated by one or two researchers who were always in class. The questions were read by the researchers, and children could ask for information when something was unclear. We saw that circumvention of parental control and online pretending was reported by more than 50% of all children. We found that birth country of both parents resulted in significant differences in circumventing parental control on the one hand, and online pretending on the other hand. Parental control was lower for children with a parent born in Belgium compared to children with a non-eu originated parent, at least according to the reporting of the children. We found that the level of online pretending was significantly higher for male 5 th graders and for children from a non-traditional family. Secondly, just because children use the internet long before their first years of primary school and the importance to study online risky behaviour at early age, it seems to be important to understand parental attitudes and the role of parents as they, especially in primary school children, have very important influence on children especially between the ages of 8-12 years. At that age parents are more important than when children go to secondary school and the influence of peers and friends becomes more and more prominent. Therefore, we studied the influence of parental mediation, namely parental supervision on online risk behaviour among the whole sample of 1,437 children in both 3rd and 5th grade of primary school and investigated whether gender differences occur. We found very interesting gender differences and gender patterns. For girls, it was found that online risk behaviour was negatively associated with parental supervision and school grade. That means that for girls, online risky behaviour decreases when parental supervision increases, which was not found in boys. For boys, non-traditional family structures and children from the 5 th grade were significant predictors of online risk behaviour. Children from single-parent families and composite families, and older children (5 th grade), exhibited more online risky behaviour than children from more 10 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

12 traditional families and children of the 3 th grade. We also found that an active and communicative approach by parents was more effective in lowering online risk behaviour among girls than a passive and non-communicative approach. After four months, a second measurement was performed in approximately 20 schools (less than half of the original schools). The smaller sample for the second study was drawn from the complete sample of more than 1,400 children. Before conducting the second measurement, children were classified by the researchers into a high-risk and low-risk online group based on 12 indications of online risky behaviour. Both groups, together, counted over 800 children. After that, both groups were randomised classified into an intervention group and a control group; half of the children of the high-risk group received the intervention or control condition. The same happened in the low-risk group. The intervention group received information about danger behaviours on the internet. This information was given by the researchers by showing the children a PowerPoint presentation that took about 10 minutes. Each sheet was explained by the researcher without being suggestive. In contrast, the control group received neutral information about the internet, such as games on the internet, etc. and no information was given about online risks. The purpose of the intervention was to examine the awareness-effect and behavioural-effect after the intervention or control condition and to examine differences between the intervention and control group. The second measurement took place immediately after both conditions and was instructed and coordinated by one of the researchers in class. After about four months, all 800 children who were involved in the intervention/control conditions were questioned again (third measurement), to examine whether awareness/behaviour about online risks continued or not. The general hypothesis was that awareness of online risks would increase in the intervention group compared to the control group in both the high and lowrisk group. It was further assumed that between the intervention and the third measurement (four months following the intervention), a decrease of awareness was found and online risk behaviour would rise again. This increase would be especially observed in the high-risk group. We found that children in the intervention group compared to children in the control group were much more likely to be aware of online risks and although diminishing, online risk awareness was still noticeable four months after the intervention and remained higher in the intervention group than in the control group. This pattern was found, especially in the high-risk group, but also, to a lesser extent in the low-risk group. This finding was in line with the general hypothesis. However, against our expectations and in opposite with the hypothesis, the intervention did not seem to diminish the amount of online risk behaviour. Four months after the intervention, we found that children in the intervention group were even more likely to report online risk behaviour. This means that the intervention had an impact on the awareness of dangers on the internet, but no transition happens from awareness to behavioural changes. A gender difference was observed: girls were more likely to be aware of the online risks directly after the intervention than boys were. Girls and children in 3 th grade were less likely to behave in a risky manner when online. We also concentrated on online behaviour of young adults and more particularly of university students from computer science, law and criminology. For this study, an online survey was conducted among almost 500 students. We examined potential influential factors related to the degree to which students are concerned about online privacy, online risk and the way they are (or not) engaged in information protecting behaviour. We examined a series of predictors such as, unpleasant experiences on social network sites, experiences of privacy invasions, type of education (computer science, law and criminology), gender, age, and the personality trait extraversion. We assumed that age, gender, type of education, unpleasant experiences, extraversion, and privacy invasion significantly contributed to privacy concerns. More particularly, age, gender, type of education, unpleasant experiences, extraversion, and privacy invasion were expected to significantly contribute to risk concerns and to information protection. We found that students who scored 11 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

13 higher on extraversion, showed a larger online social network size than students who scored lower on extraversion. Results show that extraverted students had a larger online social network size, the majority of the participants had a Facebook profile, were most active on Facebook, used social network sites more than twice per day, were between 1 and 5 hours a week occupied with checking their profiles, had their total number of friends between 201 and 300, and updated their profile more than twice a week. Most of the participants had read the privacy policy but not carefully. Our results also showed that education type and privacy invasion were significant predictors of privacy concerns, with privacy invasion having the largest contribution. Students who studied computer science took more precautions on the internet than students who studied law. Criminology students scored the worst on security concerns. Studies 5-9 Following studies focussed on (online) child molesters, paedophiles and rapists to gather information and tools for mental health professionals, police and public prosecutor. As already mentioned in the objectives, our first goal was to understand typologies of deviant sexual behaviour and to gain more insight into online predators, paedophiles, child molesters, rapists and generalists. The latest group is e.g., online predators that also show cross-over sexual behaviour what means that they both commit crimes online and off-line. As it is known from numerous studies, child molesters commit very often sexual and non-sexual crimes, and both children and adults can be victimised. Otherwise, paedophiles can be characterised more as specialists, often limited to sex crimes and seeking for a relationship with children. They don t use violence and force very often. In our first study, we conducted a systematic review to give more insight in the heterogeneity of (online) child sex offenders in order to provide tools for individual risk assessment, treatment planning, police detection and information about recidivism rates. Based on reviewing offender characteristics, including demographics, socio-affective difficulties, cognitive distortions and psychosexual issues, we found that, despite individual differences, many child pornography offenders suffer from psychological difficulties in multiple areas of psychological, social and professional functioning. We found that risk factors of especially online offenders can be seen along the lines of two continua: the first continuum is characterised as features related to criminal behaviour; in the second continuum, the presence of sexual deviance/fantasy and cognitive distortions is apparent. Both continua gave insight into pathways and psychological differences between offender types, but also allow practitioners and police officers to detect variation in the severity of psychological difficulties within offender types, and therefore may enhance individual risk assessment, interviewing strategies and treatment. Further, information of the background of sex offenders also gives us more information in the severity of a crime and victim injuries. In this study, we also discussed cross-over behaviour, factors related to engagement with the internet, the offline environment and the combination of both online and offline sexual deviant behaviour. This information is important for mental health services and the police. A systematic review is very interesting to bring together a lot of research to distillate sexual deviant patterns (criminal and antisocial pathway and a more sexual deviance pathway), as we did above. However, a large group of sex offenders (online and offline) is missing in current research; most studies on (online) sex offenders took place in prisons and forensic mental health institutions. Most studies are carried out on well known and identified sex offenders. However, this group of known sex offenders is a selective group of sex offenders who have been arrested and forced to mandatory treatment. Only a very small group of (online) sex offenders asked voluntarily for mental health. 12 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

14 The group that is hidden from the police and mental health services is called the 'hidden population or the dark number group. These sex offenders barely have come in contact with the police and with mental health. It is, of course, the question what characteristics are typical in this group of unidentified sex offenders and whether this group differs from the detected group sex offenders. A study of this unidentified group is difficult because these individuals are difficult to trace but a study on unidentified sex offenders cannot be missing in this study. As explained earlier, we found a method to interview 15 self-identified sex offenders what offers a great contribution to both mental health and the police. By interviewing those 15 self-identified sex offenders, we provided more insight in risk- and protective factors for child sexual offending in paedophiles, and the development of paedophilic feelings in general. We studied self-identified paedophiles outside clinical and forensic settings. An important question is how they survive and what kind of skills they have to (self)-regulate their sexual and emotional feelings toward children. Two studies were conducted. In the first study, 15 paedophiles were interviewed about sexual development, coping, and ways to relief sexual arousal. Most paedophiles mentioned that they often became aware of their sexual attraction to children in early puberty, and that their sexual preference continued into adulthood. Many of them struggled with those paedophilic feelings and experienced psychological problems such as anxiety and feelings of depression. Many of them were less or more engaged in child sexual offending in puberty when they were still discovering their sexual feelings and sexual identity. Therefore, early recognition of risk factors and early start of interventions seem vital in preventing (online) sex offenses. What we found is that there is an urgent need for sufficient help to support paedophiles in their acceptance of paedophilic feelings, and in finding non-harmful ways to relief their sexual arousal without making victims. In our second study, we compared those 15 self-identified paedophiles to a control group of 62 adult males on risk factors for offending. We found that paedophiles were more likely to have a history of abuse in childhood and experienced more negative life events before the age of 16 years compared to the control group. These findings were in line with previous research in the last three decades. Numerous studies are published about victimisation and witnessing domestic violence, psychological, physical and sexual abuse in the family of origin. Compared with normal controls, sex offenders recalled, regardless of the type of sex offender, significantly more frequently victimisation of neglect and abuse in childhood and youth. Sex offenders reported significantly more insecure attachment styles with parents and caregivers in childhood and intimacy problems with romantic partners in adulthood. Assuming an intergenerational approach, it can be stated that there has been a reversal: especially boys who were victimised or have witnessed violence in their family in childhood and youth, are more at risk to become an aggressor in adulthood. This is also found in girls but to a much lesser extent. Finally, let s return to the group of 'hidden sex offenders' and the fact that the detection and monitoring of sex offenders is very difficult. The following recent Dutch press release is an example of this (July, 18 th 2014): Arrest of the Utrecht serial rapist: There is a big breakthrough in the case of the Utrecht serial rapist. A suspect is arrested after 20 years. It is a 51-year-old man from the vicinity of Utrecht. The suspect has been convicted of bicycle theft earlier this year. He had to give the police his DNA. It turned out that his DNA-profile matched with the DNA from three cases, two from 1995 and one from The match was found by the Dutch Forensic Institute on Tuesday 15 th July. The man was arrested and brought before the judge. The suspect is not a stranger to the police. Twice before, 13 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

15 he was named in the research but there was little evidence to see and arrest him as a suspect. In 1995 and 1996, five women were raped in Utrecht and surrounding area. For a few years, it remained silent; until 2001, another rape took place by the same perpetrator. Justice brings the man also in relationship to 12 other attempts to rape. Knowledge about hidden populations is important for forensic and investigative psychology, criminology, police investigation, justice and mental health risk management. Hidden populations are f.e., drug users, illegal immigrants, illegal prostitutes, human traffickers, rapists and (online) child abusers. These groups succeed reasonably easy in hiding from the police because they seldom leave traces on the crime scene and because they are geographically mobile. Geographical distribution also requires cooperation with other regions what is not easy and requires coordination. The example of the serial rapist in Utrecht shows very well that it has lasted 20 years before the suspect could be arrested; in this specific case, thanks to a DNA match that was taken by coincidence after a bicycle theft. In this case, DNA traces were found but in many cases, tracks are destroyed by the offender and very often, victims are unable to observe the perpetrator and to give useful information to the police. Because the police very often has a lack of clear investigative directions, it s the question whether there are methods to detect specific groups of criminals, such as sex offenders. A method that is often applied in criminology is the capture-recapture method. This is an interesting method that is based on the assumption that a hidden population can be estimated by using registered individuals in this hidden population from, e.g., police databases or mental health databases. However, a problem that very often occurs is that the quality of databases is bad to perform capturerecapture method. This is very often the case, or because there are no databases at all. Because capturerecapture often fails, sampling networks can be seen as an interesting method and an alternative to help the police to find hidden targets, f.e., in the detection of unknown sex offenders. On the basis of a hypothetical example, a network sampling method was presented in this report that gives resources to the police to perform their investigation in a better and more structured way. We also included software and learning tools and the illustration of how this method works was explained step by step. Information on the software and learning tool is also in the report. 14 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

16 Conclusion This study can be seen as a multi-method study because we used multiple data sources: a prospective explorative research was conducted among about 1,400 children from primary school. The research design was explorative and prospective (repeated measurements). A second database was obtained through an online survey that consists of nearly 500 university students (computer science, law and criminology); a third database was set up consisting of 15 self-identified sex offenders. We used both quantitative and qualitative research methods (statistical analyses, interviews and systematic review). The delivered products: articles, PowerPoint presentations in different languages and the tool to perform social network analysis to use in police investigation research, can contribute to a better and more efficient detection of sex offenders by the police. We portrayed the first four studies as supportive studies. However, these studies are also vitally important for the police, public prosecution and mental professionals to gain insight into basic behaviours of young children and adults. The reason for that is that children of today very frequently consult social media, unknowingly, can be confronted with online risk behaviour initiated by themselves or by others. Those children can become victims and/or perpetrators of tomorrow. Prevention is better than cure. 15 With the financial support from the Prevention of and Fight against Crime Programme of the European Union European Commission Directorate-General Home Affairs

17 Partners Applicant Organisation/Coordinator KU Leuven KU Leuven is the largest academic institution in Belgium and one of the oldest European universities as it was founded in 1425 ( It is a research-intensive, internationally oriented university that carries out both fundamental and applied research. It is strongly inter- and multidisciplinary in focus and strives for international excellence. To this end, KU Leuven works together actively with its research partners at home and abroad. With a research expenditure of 365 million in 2012, the KU Leuven is a leading research university in Europe. KU Leuven is also a member of the League of European Research Universities (LERU), a group of twenty European research-intensive universities committed to the values of high-quality education in an internationally competitive research environment. More than 200 KU Leuven researchers are permanently working on information and communications technology related issues. They belong to different university departments with a strong tradition in multidisciplinary research on information and communications technology issues. The Interdisciplinary Centre for Law & ICT - ICRI ICRI is co-ordinating the activities of the Belgian Cybercrime Centre of Excellence for Training, Research and Education - B-CCENTRE. The Interdisciplinary Centre for Law & ICT ( is a research centre at the Faculty of Law of KU Leuven dedicated to advance and promote legal knowledge about the information society through research and teaching of the highest quality. ICRI is among the founding members of the LEUVEN Centre on Information and Communication Technology (LICT) and the Flemish ICT Research Institute iminds. Currently, ICRI is part of the iminds Security Department, a de facto one-stop-shop for ICT security research. ICRI is committed to contribute to a better and more efficient regulatory and policy framework for information & communication technologies (ICTs). Its research is focused on the design of innovative legal engineering techniques and is characterised by its intra- and interdisciplinary approach, constantly aspiring cross-fertilisation between legal, technical, economic and socio-cultural perspectives. By conducting ground-breaking legal research in a spirit of academic freedom and freedom of inquiry, ICRI aspires to a place among the centres of excellence in the area of law & ICT in Europe and beyond. LINC Launched in January 2007, the Leuven Institute of Criminology (LINC - is composed of about seventy professors and assistants involved in criminological research and education within the Department of Criminal Law and Criminology of the KU Leuven Faculty of Law. LINC is the most recent institutional incarnation of the criminological tradition in Leuven that started with the establishment of the "School for Criminology" in LINC intends to continue the Leuven tradition of combining solid research with a deep commitment to society. This goal is achieved through fundamental as well as policy-oriented research that is organised around 8 research lines. 16

18 Co-beneficiaries Universiteit van Tilburg (UVT) Department of Developmental Psychology and Forensic Psychology The department of Developmental Psychology at Tilburg University consists of a group of researchers interested in development and individual differences over the life span. With a focus on early infancy on the one hand, and adult developmental processes, on the other hand, the research group pools experts in different stages of development. Using a broad spectrum of research designs and methods, such as shortand long-term longitudinal studies, behavioural genetics, and neurophysiological methods, the research is aimed at describing, explaining, and predicting the development of individual differences. Important outcomes include attachment, social-emotional processes, stress reactivity, personality, and deviant/delinquent behaviour. 17

19 Manager Ann Mennens Ann Mennens is the Manager of the B-CCENTRE Project. She started working in September 2011 at ICRI, KU Leuven to organise the work of the Belgian Cybercrime Centre of Excellence for Training, Research and Education. She coordinates the activities of several academic research groups, public sector bodies and businesses in Belgium dealing with cybercrime. She initiates, supports and manages interdisciplinary research on cybercrime and cyber security, the development and teaching of basic and advanced cybercrime trainings. She is active in setting up and creating awareness raising initiatives related to safe online experiences, both for businesses and organisations, as well as the general public. She is representing the B-CCENTRE in conferences and working groups in Belgium, the EU and worldwide. She is one of the founding members of the Belgian Cyber Security Coalition, a coalition of public authorities, the academic world and the business sector joining forces against cybercrime in Belgium. It brings together more than 50 key players to share knowledge, raise awareness among citizens and businesses and issue recommendations for a more efficient policy. For over 20 years, she has led various projects in the field of Justice and Security, involving governmental and other actors from the EU Member States and beyond. The fight against crime and cooperation between judicial authorities and law enforcement in the EU, have been at the core of the projects under her management. She has a track record of creating networks and systems for cooperation, information exchange and dissemination and of organising training programmes for several target groups, in particular Police and Judiciary. 18

20 Researchers Prof. Dr. Stefan Bogaerts Between 2006 and October 2014, Stefan Bogaerts was one day a week appointed as professor of Criminology at the Research unit of Criminal Law and Criminology at KU Leuven, Belgium. He is one of the principal investigators of The Belgian Cybercrime Centre of Excellence for Training, Research and Education responsible for the criminological and psychological research. Since 2005, he was also part-time full professor of forensic psychology at Tilburg University, Department of Development Psychology, since 2010 responsible for the education in forensic psychology and since 2012 appointed as vice-chair of the Department. In 2014, he made the full switch to Tilburg University. In addition, he works as the head of research and treatment innovation, psychotherapist and supervisor at the Forensic Psychiatric Center Kijvelanden/DOK and FPC Gent, and is doing research on causes and deregulation processes of aggressive behaviour and victimisation. Since 2008, he examines online risk behaviour and vulnerability factors of online aggression and victimisation. Currently, he is involved in several (inter)national research projects on cybercrime and online sexual deviant behaviour, and concentrates on self-regulation and deregulation processes and the development of problematic personality. He wrote about 125 articles, chapters and books in the field of forensic psychology, victimology and criminology. Marjolein Brusselaers Marjolein B. J. Brusselaers works as a junior researcher at the Leuven Institute for Criminology (Catholic University of Leuven). Her research interests include the interconnection between online behaviour of minors, its psychological impact and the educational field. Within the B-CCENTRE project she investigates the link between online risk behaviour in primary school children and child/family characteristics as well as children s coping responses to unpleasant online experiences. Therefore she collaborates closely with Prof. Dr. Stefan Bogaerts, Janneke Schilder, MSc and Marjolein Missler, MSc. She is involved in preparing the material for presenting the B-CCENTRE research results. She holds a Bachelor s degree in Applied School and Pedagogical Psychology and a Master s degree in Clinical Psychology Specific Teacher Training in Human Sciences. Karel Demeyer Karel Demeyer obtained a Master s degree in both Criminological Sciences and Applied Computer Sciences. At ICRI KU Leuven he was part of the team setting up the B-CCENTRE. He worked at the Leuven Institute for Criminology (Catholic University of Leuven) from 2011 until During this period he was together with Professor Bogaerts responsible for the data collection for the B-CCENTRE research study and he was also involved in the data analysis. 19

21 Jenny Houtepen Jenny Houtepen, MSc. is a junior researcher at Tilburg University at the department of Developmental Psychology. After completing her master s in Forensic Psychology at Tilburg University she worked at Tilburg University to study child pornography offending. Her role in the B-CCENTRE is to identify risk factors for engaging in child pornography downloading, including demographics, socio-affective difficulties, cognitive distortions, and psychosexual difficulties. Furthermore, her research aims to identify risk for crossing over to hands-on offences in child pornography downloaders by focusing on processes of self-regulation. She collaborates closely with Dr. Jelle Sijtsema and Dr. Stefan Bogaerts. Marjolein Missler Marjolein A. Missler, MSc. is a junior researcher at the Department of Developmental Psychology at Tilburg University. She received her Master s degree in Clinical Psychology at Utrecht University. During her Master s, she worked with traumatised war victims in an internship at Centrum 45, the Dutch national centre for treatment of and research into the consequences of organised violence. She was supervised by Prof. dr. Stefan Bogaerts in a research project aimed at developing an aggression regulation training for forensic psychiatric outpatients. In her work for the B-CCENTRE, also supervised by Prof. dr. Stefan Bogaerts, she studies parental influences on online risk awareness and risk behaviour among school-aged children. In this project, she further collaborates with Marjolein Brusselaers, MSc. and Janneke Schilder, MSc. Janneke Schilder Janneke D. Schilder works as a junior researcher at the Department of Developmental and Forensic Psychology at Tilburg University. She received her Bachelor s degree in Social Psychology at the University of Utrecht and her Master s degree of Social Psychology at the University of Tilburg. Her research interests include the interconnection between online behaviour of minors, child development in educational settings, and family relations. Since July 2014 she is involved in the B- CCENTRE project where she closely collaborates with Prof. Dr. Stefan Bogaerts, Marjolein Missler, MSc, and Marjolein Brusselaers, MSc. Her role in the project is to study the internet risk behaviour in Flemish primary school children and the effectiveness of a school intervention to reduce online risk behaviour. 20

22 Jelle Sijtsema Dr. Jelle J. Sijtsema is an Assistant Professor at Tilburg University at the department of Developmental Psychology and is affiliated to the forensic care institute Het Dok/ De Kijvelanden. He received his PhD in Sociology at the University of Groningen on the topic of adolescent aggression. In his current research he examines associations between social relationships and antisocial behaviour, including aggression, delinquency, and antisocial personality traits. In particular, his research centers on processes of self-regulation and aims to study to what extent social relationships may ameliorate or exacerbate the behavioural consequences of poor self-regulation. His role in the B-CCENTRE is to study processes of self-regulation in child pornography downloaders in order to identify risk for hands-on offences. He collaborates closely with Jenny Houtepen, MSc. and Dr. Stefan Bogaerts. 21

23 Publications - Abstracts Circumventing internet parental control and online pretending in primary school children: Family related differences and predictors Brusselaers, M. B. J., Bogaerts, S., & Demeyer, K. The internet offers many advantages, but at same time it holds risks of harm for which children are especially vulnerable. Despite protective efforts of parents and websites (e.g., parental supervision, filter software, usage restrictions, minimum age), some children (try to) evade these interventions by circumventing parental control and by online pretending. Little is known about the prevalence, level and possible predictors of these types of secret behaviour in pre-adolescents and how they relate to gender, age, child s and parents country of birth and family structure. To get some insight in this matter, a representative sample of 1437 Flemish primary school children, aged 8-12 years old, were questioned about their internet behaviour. Findings show that circumvention of parental control as well as online pretending is done by more than half of all children. Birth country of both parents result in significant differences in both circumventing parental control and online pretending: values are lower for children with a parent born in Belgium opposed to those with a non-eu originated parent. The level of online pretending is also significantly higher for boys, 5th graders and children from a non-traditional family. Circumventing parental control and online pretending predict each other reciprocally. For circumvention the country of birth of the mother is also found to be a significant predictor. Gender, school grade and the birth country of the father significantly predict online pretending. Practical implications are discussed alongside directions for future research. The Effectiveness of an Intervention to promote Awareness and Reduce Online Risk Behaviour of Belgian Primary School Children Schilder, Janneke D; Brusselaers, Marjolein B. J.; Bogaerts, Stefan The current study explored the effect of a school based intervention on online risk awareness and behaviour of primary school children (grade 4 and 6) at two measurements. Children received a presentation about the possible risks on the internet (intervention group) or a presentation about internet applications without emphasis on risk (control group). Children in the intervention group were more likely to be aware of online risks. Although diminishing, this effect on online risk awareness was still noticeable four months after the intervention. However, against our expectations the intervention did not seem to diminish the amount of online risk behaviour. Four months after the intervention children who received the intervention were even more likely to report online risk behaviour. In line with our expectations a gender difference was observed, girls were more likely to be aware of the risks directly after the intervention than boys were. Additionally, girls and children in a lower grade were less likely to behave in a risky manner when online. Additional experimental research is needed to explore the effects of school-based interventions on actual online risk behaviour of young children. 22

24 The influence of parental supervision on online risk behaviour among young children: a gender perspective M.A. Missler & S. Bogaerts Children use the internet from their first years of primary school onwards. Therefore, there is a need to study online risk behaviour at this early age, as well as the ways in which parents can guide the online behaviour of their younger children. The aim of the current study was to detect the influence of a specific form of parental mediation, parental supervision, on online risk behaviour among 1437 Flemish children in the 3rd and 5th grade of primary school. We were particularly interested in possible gender differences. Results of the regression and path analyses showed interesting gender patterns. For girls, online risk behaviour was negatively influenced by parental supervision and by school grade. For boys, family structure and school grade were significant predictors. This implies that when studying online risk behaviour among younger children, gender differences should be taken into account. While an active and communicative approach by parents seems to be effective in lowering online risk behaviour among girls, for boys it appears necessary to explore other strategies. M.A. Missler & S. Bogaerts (2014) The influence of parental supervision on online risk behaviour among young children: a gender perspective Privacy, Risk, Information protection and Social Network Site-Using Behaviour in a Sample of Flemish University Students Horvath, Z., Bogaerts, S. & Demeyer, K. Hundreds of millions of people from all age categories use social network sites and share a wide range and enormous amount of information of themselves and others, such as personal details and connections with friends, often without realizing the potential privacy, security and profiling risks they take. Objectives: The current paper intends to provide a description of social network site-using behaviour of the participants and examine influential factors contributing to the degree to which university students are concerned about their privacy and risk, and the extent to which they engage in information protecting behaviour. More specifically, we intend to shed light on predictors, such as unpleasant experience when using social network sites, experience of privacy invasion, type of education (law, informatics and criminology), gender, age, and the personality trait extraversion in a sample of university students. Based on the relevant literature, the following hypotheses are addressed: (1) Age, gender, type of education, unpleasant experience, extraversion, and privacy invasion are expected to significantly contribute to the degree of privacy concerns. (2) Age, gender, type of education, unpleasant experience, extraversion, and privacy invasion are expected to significantly contribute to the degree of risk concerns. (3) Age, gender, type of education, unpleasant experience, extraversion, and privacy invasion are expected to significantly contribute to the degree of information protection. Results: The results regarding students social network site-using behaviour reveal that high extraversion traits lead to large online social network size, the majority of the participants have a Facebook profile, are most active on Facebook, use social network sites more than twice per day, spend between 1 and 5 hours a week with checking their profiles, have their total number of friend between 201 and 300, and update their profile more than twice a week. Most of the participants have read the privacy policy but not carefully. Most of the students who have read the privacy policy understood it as a whole. The lack of interest was generally the reason for not reading the privacy policy. The results suggest education type and privacy invasion to be significant predictors of the degree of privacy concerns, with privacy invasion having the largest contribution. 23

25 The findings also imply that education type, unpleasant experience and privacy invasion are significant predictors of the degree of risk concerns, with unpleasant experience having the largest contribution. The role of education type, age, extraversion in the degree of information protection was also found to be significant, with extraversion having the largest contribution. Conclusion: Further studies are required to investigate the relationship among these variables to raise people s awareness of privacy, risks, and information protection. Furthermore, the identified influential factors, such as experience of privacy invasion and education type should also be taken into account in future research. Suggested Citation: Horvath, Z. Bogaerts, S. & Demeyer, K. (2014). Privacy, Risk, Information protection and Social Network Site-Using Behaviour in a Sample of Flemish University Students. From child pornography offending to child sexual abuse: A review of child pornography offender characteristics and risks for cross-over Houtepen, J., Sijtsema, J., & Bogaerts, S. In this review concrete directions are provided for individual risk assessment, treatment planning, and future research on child pornography offending. First, based on reviewing offender characteristics, including demographics, socio-affective difficulties, cognitive distortions and psychosexual issues, it is concluded that despite individual differences, many child pornography offenders have psychological difficulties in multiple areas of functioning. Based on earlier child pornography offender typologies, it is proposed that risk factors of individual offenders can be viewed along the lines of two continua: (1) features related to criminal behaviour, and (2) sexual deviance/fantasy. These continua give insight into psychological differences between offender types, but also allow for variation in the severity of psychological difficulties within offender types and therefore may enhance individual risk assessment and treatment planning. Moreover, risk factors for cross-over are discussed, including individual characteristics, factors related to engagement with the internet, and the offline environment. Future research should focus on the integration of risk domains within particular offending types. Specifically, research is needed on non-offending pedophiles in order to gain more insight into the relationship between pedophilia and child pornography offending in general. Houtepen, J., Sijtsema, J., & Bogaerts, S. (2014). From child pornography offending to child sexual abuse: A review of child pornography offender characteristics and risks for cross-over. Aggression and Violent Behavior, 19(5), DOI: /j.avb Being sexually attracted to minors: Sexual development, coping, and risk factors for offending in self-identified pedophiles. Jenny A.B.M. Houtepen MSc, Dr. Jelle J. Sijtsema, & Prof. Dr. Stefan Bogaerts This article aimed to provide more insight into risk- and protective factors for child sexual offending in pedophiles, and the development of pedophilic feelings in general, by studying self-identified pedophiles outside clinical and forensic settings. Specifically, attention is drawn to aspects of self-regulation. Two studies were conducted. In the first study 15 pedophiles were interviewed about sexual development, coping, and ways to relief sexual arousal. Pedophiles stated they often became aware of their sexuality in early puberty, and that their sexual preference continued into adulthood. Many struggled with acknowledging pedophilic feelings and as a result experienced psychological difficulties such as anxiety. Furthermore, many engaged in child sexual offending in puberty when they were still discovering their feelings. Therefore, early recognition of risk factors and early start of interventions seems vital in preventing sex offenses. There is a need for sufficient 24

26 help to support pedophiles in (1) acceptance of pedophilic feelings, and (2) finding non-harmful ways to relief sexual arousal. In the second study pedophiles were compared to a matched control group of 62 adult males on risk factors for offending. As hypothesized, pedophiles were more likely to have a history of abuse in childhood, and experienced more negative life events before the age of 16. In contrast to the expectations, pedophiles were found to be more behaviourally inhibited (i.e., less impulsive). Results suggest that the development of pedophilic interest is, at least in part, influenced by contextual factors. Houtepen, J.;Sijtsema, J.; Bogaerts, S. (2014) Being sexually attracted to minors: Sexual development, coping, and risk factors for offending in self-identified pedophiles. Helping the police investigation: police investigation applied to sex offenders by using network sampling Spreen, M 1. & Bogaerts, S 2. Knowledge about a hidden population is important for forensic psychology and criminology. Hidden populations are for example, drug users, human traffickers and online child abusers. This group succeeds reasonably easy in hiding from the police. The question is whether there are methods to detect these groups of criminals. A method that is often applied in Criminology is the capture-recapture method. This method is based on the assumption that a hidden population can be estimated by using registered individuals in this hidden population from, e.g., police databases. However, a problem is that the quality of databases must be very good to perform capture-recapture, which is very often not the case, or because there are no databases at all. Because capture-recapture often fails, sampling networks are interesting alternatives to help the police to find hidden targets. On the basis of a hypothetical example, a network sampling method is presented that gives resources to the police to perform their investigation in a better and more structured way. We also refer to software and learning tools, and the illustration in this paper is conducted on the basis of this software and learning tool. 1 sociologist, PhD: head of research, Dr s. van Mesdag, forensic psychiatric centre Groningen; lector Stenden University, Leeuwarden. 2 Full professor of Forensic Psychology, Tilburg University and head research and treatment innovation FPC de Kijvelanden, Netherlands, FPC Gent, Belgium. 25

27 Full Text Articles Circumventing internet parental control and online pretending in primary school children: Family related differences and predictors Marjolein B. J. Brusselaers, Stefan Bogaerts, Karel Demeyer Catholic University of Leuven, Department of Criminal Law and Criminology, Hooverplein 10, 3000 Leuven, Belgium Tilburg University, Department of Developmental and Forensic Psychology, Warandelaan 2, 5000 Tilburg, the Netherlands Introduction In these modern times a world without the internet is unimaginable. Children grow up with this medium and consider its existence as self-evident. Already in kindergarten they take their first steps into the online world (Teuwen, De Grooff, & Zaman, 2012). Young people use the internet mostly for entertainment, education and edutainment (Livingstone, 2003). Despite all the advantages the internet offers, we also have to be aware of the possible risks of this virtual world for minors. Livingstone and Haddon (2009) give a good overview of the online risks for children and adolescents (see Table 1). They have classified the risks of harm to children from their online activities along four dimensions (commercial, aggressive, sexual and values) and three types (content: child is a recipient of mass-distributed content, contact: child as participant in interaction, mostly driven by adults and conduct: child as actor in an interaction in which he may be initiator). Table 1 Classification of online risks of harm Content: Contact: Conduct: child as recipient child as participant child as actor Commercial Advertising, spam, sponsorship Tracking/ harvesting personal information Gambling, illegal downloads, hacking Aggressive Violent/ gruesome/hateful content Being bullied, harassed or stalked Bullying or harassing another Sexual Pornographic/harmful sexual content Meeting strangers, being groomed Creating/ uploading pornographic material Values Racist, biased info/ advice Self-harm, unwelcome persuasion Providing advice e.g., suicide/pro-anorexia Source: Livingstone & Haddon, Considering this broad spectrum of online risks, parents have a responsibility to act upon these virtual threats to protect young people s safety (but still allow them to explore the cyber world). Especially children under the age of 12 need this guidance to account for their less developed skills to assess and manage the internet and its risks (Hasebrink, Livingstone, Haddon, & Ólafsson, 2009). Parents can use different approaches to mediate the internet behaviour of their children, e.g., staying around or watching the child when he surfs on 26

28 the internet, limiting online time, restricting certain online activities, talking about the internet with the child, using software to filter or block questionable websites, checking internet history afterwards (e.g., Cole, 2001; Livingstone & Helsper, 2008; Valcke, Bonte, De Wever, & Rots, 2010). These parental actions are categorized differently among researchers. Since Belgium belongs to the group of European countries that mostly protect children from risks of harm by using restrictive mediation approaches (Helsper, Kalmus, Hasebrink, Sagvari, & de Haan, 2013), we endorse the three different types of (restrictive) internet behavioural control mentioned by The Internet Parenting Style Instrument (Valcke et al., 2010): supervision, stopping internet usage and setting internet usage rules. Parental supervision refers to being around when the child uses the internet, watching along while the child surfs the internet, checking the history of the child s internet usage and using filter-software to block certain internet content. A more invasive way of parental control is stopping certain internet usage (e.g., when the child visits a less suitable website or is chatting). Lastly parents can set rules about when and how long the child is permitted to go on the internet or define restrictions about internet activities and type of websites (Valcke et al., 2010). Furthermore, apart from the parental efforts to educate children about safe internet use, websites and internet applications also want to provide a nice and safe environment for their users. They do so by implementing a statement of rights and responsibilities and/or obliging users to sign a user agreement in which conditions and terms of use are explained. Facebook (2013) and Youtube (2010), two of the most popular social media in Flemish children and youngsters (Mediaraven & Linc, 2014), set rules about privacy, content and sharing information, users age (e.g., the minimum age to create an account on Facebook is set at 13), safety (e.g., bullying or harassing another user is not allowed), registration and account security (e.g., you are not allowed to create an account for someone else without permission) and protecting other people s rights. Despite efforts of parents and internet policies to encourage safe internet behaviour in children and adolescents, research shows that some youngsters ignore or circumvent rules and restrictions concerning their internet usage. In the following, studies on two sorts of secret and disobedient internet behaviour will be discussed: circumventing (internet) parental control and online pretending. Empirical and theoretical base Research on circumventing parental control The UK Children Go Online (UKCGO) survey showed that 69% of 9-17 year olds who use the internet at least weekly are concerned about the parental supervision of their children s internet usage (checking their , blocking websites and checking their internet use with and without their knowledge). As a result, almost two thirds (65%) of these concerned youngsters have taken some privacy protecting actions (e.g., deleting e- mails and history, minimizing a window when someone came into the room), although 46% of parents claim that there is filtering software on the computer used by their child (Livingstone & Bober, 2004). In line with these results, a Belgian study revealed that 39,3% of teenagers sometimes to always delete the list of visited websites, while only 14% of their parents suspect this circumventing behaviour and 22,2% of parents admit having no knowledge about this (Walrave, Lenaerts, & De Moor, 2009a). When looked at manipulating or overriding computer filters or blocks in year olds in the USA, this sneaky behaviour was found to be significantly more likely in boys (32.6%) than girls (15.7%), and this while more girls (60.2%) than boys (49.2%) had a filter or block on the computer they used (Dowell, Burgess, & Cavanaugh, 2009). Also in the USA, Cole (2001) found that, although parental supervision and internet usage rules were used and two thirds (67%) of children surveyed had to ask permission to access the internet, more than fifty percent of year olds did not tell their parents everything they do on the internet. Further, parents and children seem to have a different understanding of rules concerning internet usage. In general parents state more often than teenagers that certain agreements concerning internet usage are in force (Duerager & Livingstone, 2012; Livingstone, 2007; Walrave, Lenaerts, & De Moor, 2009b). The UKCGO 27

29 survey in 9-19 year olds revealed that more parents (86%) than children (49%) say children must not share personal information with others, but only 5% of the parents think children do this while 46% of the children say they do so. Similarly, more children than parents say that they (the child) have visited a chat room, while more parents than children say they are not allowed to (Livingstone & Bober, 2004). However, despite the difference in perception between both generations, it is clear that those rules of which teenagers most often say that they are prioritized, are the same as those of which parents most often claim to be in force (Walrave et al., 2009b). Besides differences in perception of rules, sometimes children just ignore what their parents say about using the internet. More than one third of 9-16 year olds admit they do this a lot (7%) or a little (29%) (Duerager & Livingstone, 2012). Although teenagers have to obide to fewer internet rules, they still circumvent these (fewer) rules. Some Belgian teenagers whose parents forbid the following online activities ignore their parents restrictions: 6,5% share personal information, 10% cross allowed time online, 2,1% participate in online contests, 2,9% visit forbidden websites, 3,1% surfs the web when home alone, 5,4% download files (Walrave et al., 2009b). The number of youths crossing these parental rules seems rather low, but we have to take into account that also less teenagers have to obide to internet rules (thus not needing to evade those rules). These findings show that (well-meant) parental control on internet use is regularly bypassed by youngsters. When parental control is undermined, it leaves parents unaware of (a part of) their children s online activities and thus it becomes more difficult to protect them from online risks in an adequate way. Research on online pretending The internet creates an opportunity to lie about identity characteristics and to behave differently than in real life interactions (Harman, Hansen, Cochran, & Lindsey, 2005). A study with young adolescents (12-15 years old) showed that 51% had misrepresented themselves online at least a couple of times. Within this group with pretending experience, the majority had pretended to be older (Gross, 2004). Interestingly, a Dutch study concluded that 9-12 year olds opposed to year olds are significantly more likely to use the internet for identity exploration, mainly through impersonating an older person (Valkenburg, Peter, & Schouten, 2005). However, online lying about your age is also used for more obvious purposes. Cassidy, Jackson and Brown (2009) found that 52% of children between the age of 11 and 15 admitted they often pretend to be older on the internet, with 21% of this group revealing they do this to circumvent age restrictions for adult websites. Furthermore, most social network sites (e.g., Facebook) only allow registration to users from the age of 13 and up, thus primary school children have to pretend to be older to get access to such sites. The EU Kids Online Report (Livingstone, Ólafsson, & Staksrud, 2011) shows that significant numbers of underage children break the Terms and Conditions of Service of social network sites by setting up their own profiles: in Belgium 20% of 9-12 year olds use the social network site Facebook. Worrisome is that more than one third (35%) of this Belgian group have their profile set to public, of which 20% displays their address or phone number, and 29% displays the school they attend. Furthermore, within the group of year olds using Facebook only 55% is able to change security settings and 61% is able to block another user (Livingstone et al., 2011). When looking at children who pretend to be someone else (they know) online, the EU Kids Online Survey reports that 16% of the 9-16 year olds confirm that they have pretended to be a different kind of person on the internet from what they really are (Livingstone, Haddon, Görzig, & Ólafsson, 2011). Cassidy and others (2009) found that 2% of 11 year olds and 7% of 12 year olds took someone s name and pretended to be that person online. This percentage is even higher in youth aged 13 (21%), 14 (20%) and 15 (19%). In her study, Gross (2004) also explored the motivation and the context of this identity misrepresentation. Of all 11- to 15- year olds who reported pretending to be someone else, 48% did so to fool other friends. A majority of children (57%) pretends to be someone else in company of friends. It is somehow reassuring that a lot of children fake their identity online only to play a joke on friends, but we have to take into account that we do not know how this friend feels about this joke. More worrisome is when identity theft occurs, resulting in real abuse or harm. The EU Kids Online Survey found that 7% of year olds have already been a victim of somebody using their password to access their information or to pretend to be them online. They report a 15-year old girl saying Girlfriends who I thought my friends, have been awful. They took my identity to have my boyfriend 28

30 (Livingstone, Haddon, Görzig, & Ólafsson, 2011, p.100). A study on cyberbullying also reported that children often give out their password to friends who in turn abuse this information and, pretending to be that person, to cyberbully others (Mishna, Saini, & Solomon, 2009). Among year olds almost one in four reports that they occasionally had been a victim of online identity theft, whereby the thief pretended to be them to disgrace their reputation (Cassidy, 2009). Circumventing parental control and online pretending in relation to child/family characteristics Having discussed available research on circumventing parental control and online pretending in minors it is clear that the majority of studies focused on adolescents and European countries. Only a few studies include Belgian participants. Moreover, circumventing parental control and internet faking are not the main focus in a lot of studies. As a result few is known about the relationship between these types of secret internet behaviour and variables such as gender, age, family structure and child s and parents country of birth. However, from a developmental and sociological point of view some expectations about circumventing parental control and online pretending can be set. As children get older they begin to assert their autonomy and are less compliant to parental instructions than young children (Pasquier, 2001). In line with this developmental change, we assume that more older children will engage in circumventing parental control and online pretending than younger children and that also the level of these risky behaviours will be higher in older children. Since both boys and girls go through this process of enlarging their social environment (Haselager, 2004) we assume that no remarkable differences resulting from gender will be found. Concerning family structure, the Parental Absence Model (Gottfredson & Hirschi, 1990) suggests that single parent households are less effective in producing positive child outcomes than dual parenting because half of the parental unit is absent. Of course the parent-child attachment also plays an important role in achieving beneficial effects for the child, but during parental absence the parent is unable to control and supervise in a proper way, thus it is hard to maintain a high level of attachment (Rebellon, 2002). Apel and Kaukinen (2008) as well as Rebellon (2002) suggest that the consequence of this part time presence may result in more motivation and opportunity for the child to engage in unconventional activities. A recent study of Notten (2013) supports this theory: children in a single-parent family engage in online risk behaviour more frequently than those in a two parent family. On the base of this social control perspective we assume that both the prevalence and the level of circumventing parental control and online pretending will be higher in children who live in a single-parent household (including co-parenting) opposed to children from a traditional twoparent family. In the EU Kids Online study was found that online risk behaviour of children and adolescents can differ strongly between European countries (Livingstone & Haddon, 2009). However, instead of comparing online behaviour between countries, it would be interesting to explore if within one country (Belgium) online behaviour of children differs by country of birth or ethnic origin of the child. This way, it may be possible to detect ethnic vulnerability characteristics for circumventing parental control and online pretending. Country of birth of both parents can be used as an indicator of ethnic origin (Central Bureau of Statistics, 2000). No hypotheses are made however, because of the lack of available comparable studies. Aims of the present study This study will focus on circumventing internet parental control and online pretending in primary school children since research on this subject and within this age range is scarce though interesting to concentrate on. Some studies show that parental control is significantly related to lower levels of unsafe internet behaviour (Helsper et al., 2013; Valcke, Schellens, Van Keer, & Gerarts, 2007). Of course, when parental control is circumvented by children, these effects are undermined, thus the chance of being exposed to online risks of harm rises. Similarly, when children pretend to be older to get access to certain websites, they are more 29

31 likely to come across age inappropriate or harmful content. As mentioned above children who pretend to be someone else (they know) can also harm other children through this identity transformation or theft. We first want to explore how secret online behaviour in the form of circumventing parental control and online pretending is related to gender, age, birth country, origin of both parents and family structure. Based on the discussed literature, we assume that older children and children from non-traditional families engage more in circumventing parental control and online pretending. Apart from the higher prevalence in these groups, also higher levels of both types of secret online behaviour are expected for older children and children who live in non-traditional families. We do not expect differences resulting from gender. No assumptions are made regarding the child s and parents birth country. Next, we also want to shed some light on possible (risk or protecting) predictors of circumventing parental control and online pretending. This will happen in a more explorative way. The following research questions are addressed: 1. Does the prevalence of circumventing parental controls and online pretending differ by gender, age, family structure, birth country and/or ethnic origin? 2. Does the level of circumventing parental control and internet faking differ by gender, age, family structure, birth country and/or ethnic origin? 3. Can circumventing parental control and online pretending be predicted by gender, age, family structure, birth country and/or ethnic origin? The exploration of these research questions might help us to identify child or family characteristics that are related to an increased level of circumventing parental control or online pretending. This knowledge can lead to more efficient and focused preventive actions, psycho education and interventions concerning parental mediation and online risks. Method Participants A quasi representative number of primary school children were questioned in class, including pupils from all three educational networks and all but one provinces in Flanders (see Table 2). We only included 3 rd and 5 th graders because at a younger age it is difficult to question children given their less developed reading and comprehension skills. Special education was also not included. 30

32 Table 2 Division of the actual and surveyed number of primary school children across provinces (geographically) and educational networks a Nr of children in % of children in Nr of children surveyed % of children surveyed Provinces Antwerp ,9 Brussels ,6 Hainaut Limburg ,2 East Flanders ,6 Flemish Brabant ,4 West Flanders ,2 Total Educational networks Community education ,9 Subsidized publicly run education ,3 Subsidized privately run education ,8 Total a Source: based on figures of the Flemish Authorities (2011). The research sample consisted of 1437 Flemish primary school children, of whom 29,9% are third graders (n = 430) and 70,1% are fifth graders (n = 1007). The research group consists of 50,1% boys (n = 720) and 49,5% girls (n = 711); for 0,4% of the children, gender information is missing (n = 6). The majority of children (80,7%) are born in 2001 (n = 793) and 2003 (n = 359) and are therefore on the expected level of education for their age. Almost 20% are in a higher (1,9%) or lower grade (16,2% one grade below; 1,3% two grades below), compared to the group of children born in the same year. The average age (an approximation based on year of birth) of the total group of respondents is 10.5 years (SD = 1.05; range = 8-13). Nearly all children (94,7%) report to be born in Belgium. A smaller group (5,3%) is born abroad (e.g., the Netherlands, Russia, France, Armenia). For 8 of 1437 children (0,6%), no data are available (no answer or I don t know ). Procedure Power analysis showed that at least 1400 respondents would be needed to be able to conduct powerful statistical analyses (effect size = 0.30; confidence interval = 95%, alpha =.05). Because the actual numbers of pupils per grade of a school could vary during the research due to absence of pupils, we aimed at 110% of the desired 1400 respondents: 1540 in total. In order to obtain a research group representatively divided across provinces and educational networks, we used the following formula: 1540 * <PctPupilsInProvince> * <PctPupilsInNetwork> children. Schools had to be selected randomly until the number of 3rd and 5 th graders combined was at least Important to report is that since we selected a whole grade of a school and not individual pupils, the acquired number of pupils for every combination of province x network do not perfectly fit the calculated number of pupils needed for every province x educational network. Therefore, the sample is only quasi representative for all Flemish primary school children. In the last month of 2011, the sampling and recruitment of schools was conducted. For the sampling, we first randomly assigned each Flemish primary school a unique random number between 0 and 1. The list of schools, obtained from the Ministry of Education was then sorted based on this random number from small 31

33 to large. Subsequently, for every category of schools needed (a combination of province and network), the schools of the regarding category appearing first on the sorted list were contacted by and telephone to ask for their participation in our research and to ask for the actual number of pupils in 3rd and 5 th grade. This procedure was followed until we had contacted 87 schools with a total response level of 92% of which 45% negative and 55% positive responded. Out of the list of schools who responded positively, we made a selection of 39 schools with a total of 1546 children, quasi representatively distributed across provinces and the three educational networks. A passive informed consent was used to inform the parents about our research, which was distributed by the schools. Prior to the questioning, we conducted a pilot study with 78 3rd and 5 th graders of one school to check for usability, comprehension of language and degree of difficulty. Some changes were made regarding language, structure and instructions. Between April and June 2012 we surveyed all of the selected Flemish schools. The written questionnaires were assessed anonymously, classically and in the presence of at least one researcher. On every form, a unique code was printed which referred to the grade and school so the child did not have to provide this information. Measures Independent variables. To define gender children had to circle one of the depicted children: a boy (= 0) or a girl (= 1). The school grade could easily be derived from the code printed on every form (0 = 3rd grade; 1 = 5 th grade). To determine the birth country of the child, six different answering categories were offered (Belgium, Italy, Turkey, Morocco, other country fill in, I don t know), which were rescaled into three categories (0 = Belgium, 1 = member state of the European Union (EU), 2 = no member state of the EU). The country of birth of mother and father were also asked in order to get some insight in the ethnic origin (Central Bureau of Statistics (CBS), 2000; Roggemans, Cops, & Kolijn, 2013). Both variables were measured and recoded in the same way as the child s birth country. Lastly, the child was asked where he normally lives (family structure). From the six original answering possibilities (with my mom and dad, alternatively with my mom or my dad, with my mom, with my dad, with someone else from my family, with someone who is not family), the two last ones were not included in the recoded categories (0 = traditional two parent family, 1 = non-traditional single parent family) since the aim is to compare traditional two parent families with single parent families. Dependent variables. After surveys had been conducted, two principal axis factor analyses were performed to look for underlying constructs and therefore simplify data interpretation. The factor analyses addressed items measuring two types of online pretending and three types of evading parental involvement. For both factor analyses, orthogonal rotation (Varimax) was used. Since the purpose of this factor analysis was interpretative, a cut-off criterion of 0.4 was used for factor loadings (Stevens, 1992). This resulted in the identification of two factors, which we named circumventing parental control and online pretending (Table 3). Circumventing parental control (Cronbach s α = 0.498; SD = 0.885) reflects to which amount (in how many ways) the child evades internet parental control (0 = no circumvention, 1 = little circumvention, 2 = average circumvention, 3 = a lot circumvention). This scale was formed by summing up the scores on the following three items: having one or more addresses, chat names or profiles parents do not know about (1), often using the internet in absence of the parents (2) and using certain code words while chatting to warn the others when parents are watching along (3). For every item the child had to mark true (= 1), not true (= 0), or I prefer not to answer (= coded as missing. Online pretending (Cronbach s α = 0.483; SD = 1.063) measures to which amount the child pretends online (0 = no pretending, 1 = very little pretending, 2 = little pretending, 3 = frequent pretending, 4 = very frequent pretending). This total score is based on two items that question if and to which amount the child pretends to be someone else he knows on the internet and if and to which amount the child has already pretended 32

34 to be older than he really is (e.g., in order to play a game or construct a profile). For both items of sneaky behaviour three answers are possible: yes, I often do this (= 2), yes, I did this once (= 1) and no (= 0). Cases with a missing answer on at least one of the scale items were excluded in the analyses to avoid underestimation of circumventing parental control or online pretending. Cronbach s alpha shows the reliability of both scales. The values are rather low (at.50) and do not meet the commonly used.70 criterion for internal consistency. However, the interpretation of this threshold does not need to be absolute. In fact, we have to take into account that Cronbach s alpha is calculated based on the intercorrelations between all scale items which in their turn depend on the number of items. Hence, scales consisting of few items as in the present case have a lower Cronbach s alpha (Cortina, 1993). Table 3 Factor Matrix for Circumventing Parental Control and Online Pretending Item Circumventing parental control Factor loading Online pretending Factor loading Do you have 1 or more addresses, chat aliases or profiles your parents don t know about? Do you use the internet if your parents are away? Do you use "code"-language on chat to let the other party know your parents are watching along? Do you sometimes pretend to be someone else (you know) online? Do you sometimes pretend to be older than you really are? Eigenvalue Variance explained % % Cronbach s α Statistical analyses All data analyses were performed by SPSS Statistics 22. Differences in prevalence of circumventing parental control and online pretending between gender, school grades, birth country, parents origin and family structure were explored using a non-parametric χ² test. Therefore, both dependent variables were dichotomously recoded (presence or absence of both types of behaviour). Analyses of variance were applied to seek for differences in levels of circumventing parental control and online pretending by gender, school grade, birth country, origin of father and mother and family structure. Finally, linear regression analyses 33

35 shed some light on the last research question, exploring the predictive power of the described independent variables for circumventing parental control and online pretending. Throughout regression analysis, native Belgians were used as reference group for variables involving different ethnic categories. A p value of.05 was put forward as the significance level in all analyses. Results Prevalence of circumventing parental control and online pretending in relation to child/family characteristics The findings show that more than half of all children (53.5%) evade parental control in at least one way. The most reported type is often using the internet in absence of the parents (45.8%), followed by having one or more addresses, chat names or profiles parents do not know about (18.7%) and using certain code words while chatting to warn the others when parents are watching along (15.1%). More than one third (34.2%) evades parental control in one of these three ways, while others combine two (13.5%) or three types (5.7%) of circumvention. School grade, χ²(1) = 4.440, p <.05, origin of the mother, χ²(2) = 7.164, p <.05, and origin of the father, χ²(2) = , p <.01, are significantly related to circumventing parental control. More than half of 5 th graders (55.3%) but less than half of 3rd graders (48.7%) hide online behaviour from their parents. Regarding birth country of the mother, a Belgian origin seems to be related to a lower chance of circumventing parental control: more than three out of five children with a mother born in the EU (61%) or outside the EU (61.5%) opposed to 52.2% of children with a Belgian mother avoids parental involvement. Children with a father born outside the EU are more likely (64%) than those with parents with an EU (51.6%) or a Belgian (51.5%) origin to circumvent parental control (36%). However, non-significant differences in prevalence of circumventing parental control were found for birth country of the child. Lastly, though more boys than girls and more children of a non-traditional family than of a traditional family circumvent parental control, these differences are also not significant. Analyses concerning online pretending point out that 58% of all children have already at least once conducted this secret behaviour. Most children report that they have only one time pretended to be older or to be someone else he knows (31.1%). A minority of children regularly pretend to be older and to be someone else he knows (2.9%). Another small group (6.9%) has already pretended in both ways and keeps engaging in one type of pretending regularly. The remaining 17.1% regularly pretends in one way or has already pretended once in both ways. In line with circumventing parental control, a significant relationship exists between the prevalence of online pretending and school grade, χ²(1) = , p <.001. More 5 th graders (64.1%) than 3 rd graders (43.5%) have already at least once pretended online. Contrary to circumvention of parental control, gender and family structure do relate to significant differences in prevalence of online pretending, respectively χ²(1) = 8.780, p <.01 and χ²(1) = 6.291, p <.05. More boys (61.8%) than girls (54%) and more children who live in a single parent family (63.9%) opposed to a traditional one (55.9%) have at least one time pretended on the internet. For birth country of the child and origin of both parents no significant results were found. Level of circumventing parental control and online pretending in relation to child/family characteristics By means of analyses of variance, we could detect whether specific child or family characteristics are related to differences in the level of circumventing parental control or online pretending. For multiple comparison, Tukey and Games-Howell tests were used, dependent on the (in)equality of variances. Table 4 displays the means of circumventing parental control and online pretending for all categories of the independent variables. 34

36 Table 4 Characteristics of respondents (N=1437) and mean values for circumventing parental control and online pretending. Variables N % Mean CPC a Mean OP b Sex 1431 Boy *** Girl *** School grade rd grade *** 5 th grade *** Family structure 1419 Traditional * Non-traditional c * Birth country child 1429 Belgium EU country Non-EU country Birth mother country 1412 Belgium ** 0.93* EU country Non-EU country ** 1.19* Birth country father 1391 Belgium *** 0.90*** EU country Non-EU country *** 1.29*** a CPC, circumventing parental control (score range 0-3); b OP, online pretending (score range 0-4); c Nontraditional, single-parent family or co-parenting. Note. *p <.05; **p <.01; ***p <

37 Across all pupils, the general mean of circumventing parental control is 0.78 which approximately corresponds with little circumvention (score 1). No significant differences in online pretending were found according to gender, school grade, family structure and country of birth of the child. However, the country of origin of both parents clearly plays a significant role, F (2,1272) = 7.382, p =.001 for the mother and F (2,1257) = 8.141, p =.000 for the father. A post hoc analysis (Tukey s multiple comparison test), reveals that children with a mother born outside the European Union engage in considerably more circumvention of parental control than children with a mother born in Belgium, p =.003. The same is true for the origin of the father, p =.000. When it comes to online pretending in general, children pretend very little (0.98), thereby meaning they have already lied about themselves online once. Opposed to the insignificant influence on circumventing parental control, gender significantly affects the level of online pretending, F (1,1394) = , p =.000. Boys seem to pretend online considerably more than girls. Also school grade results in significant differences in online pretending, F (1,1399) = , p =.000, with children in 5 th grade engaging in this behaviour to a higher degree than 3 rd graders. Children who live in a traditional family differ in their degree of online pretending as compared to children from a non-traditional family, F (1, 1381) = 3.999, p =.046. The latter group engages in more online pretending. No significant differences result from birth country of the child. In line with the findings for circumventing parental control, the origin of the mother and the father results in significant differences in online pretending, respectively F (2,1375) = 5.892, p =.003 and F (2,1356) = , p =.000. Games-Howell s post hoc comparison analyses reveal that a child with a parent born outside the EU engages in considerably more online pretending than a child with a parent born in Belgium, p =.013 for the mother s country of birth and p =.000 for the father s country of birth. Child/family predictors of circumventing parental control and online pretending Prior to the linear regression analyses, a correlation matrix was constructed to control for multicollinearity (Table 5). Since the birth country variables consist of three nominal categories, Cramer s V was used to examine the relationship between those variables and the other dependent and independent variables. No cases of collinearity were found. 36

38 Table 5 Correlation matrix of all dependent and independent variables (Pearson s r and Cramer s V) Variables Sex - 2. School grade a * - 3. Family structure.027 a.013 a - 4. Birth country child.044 b.023 b.068 b - 5. Birth country mother.009 b.072 b *.053 b.365 b *** - 6. Birth country father.055 b.056 b.041 b.290 b ***.551 b *** - 7. Circumventing parental control a.040 a.002 a.033 b.083 b **.086 b ** 8. Online pretending.107 a ***.121 a *** a.101 b ***.092 b **.127 b *** a Pearson s r; b Cramer s V: this statistic is only used to measure associations with birth country of the child, of the mother and of the father. Note. *p <.05; **p <.01; ***p <.001. The following predictors were simultaneously included in the linear regression: gender, school grade, family structure and dummy variables for parents origin and for child s birth country (Belgium, EU country, non-eu country). In the first linear regression circumventing parental control was included as the outcome variable and online pretending was added as a predictor (Table 6). In the second regression analysis online pretending was set as the outcome variable and circumventing parental control was included as a predictor (Table 7). Regression analysis identifies birth country of the mother as a significant predictor for circumventing parental control. Children with a mother born in a EU-country engage in more circumvention of parental control than children with a mother born in Belgium (b =.257, p <.05). Online pretending seems to have a very high predictive power (b =.390, p <.001). The more a child pretends online, the more he circumvents parental control. Gender, school grade, birth country of the child, birth country of the father and living situation do not make a significant contribution in predicting circumvention of parental control. 37

39 Table 6 Summary of multiple regression analysis for circumventing parental control Variables B SE (B) Beta p Constant Sex School grade Family structure Birth country Belgium vs. EU Birth country Belgium vs. non-eu Birth country father Belgium vs. EU Birth country father Belgium vs. non-eu Birth country mother Belgium vs. EU.257* Birth country mother Belgium vs. non-eu Online pretending.390*** Note. *p <.05; ***p <.001. The regression output shows four variables that significantly predict online pretending (see Table 7). Boys engage in higher levels of online pretending than girls (b =.179, p <.01), while we predicted an ex aequo for gender. School grade also forms a significant predictor of the amount of online pretending (b =.139, p <.001). In line with our hypothesis, 5 th graders show higher levels of online pretending opposed to 3rd graders. Furthermore, the origin of the father has a significant positive relationship with online pretending (b =.289, p <.01). Children with a father born outside the EU show more online pretending behaviour than children with a Belgian father. An interesting finding is that circumventing parental control is found to be a significant positive predictor of online pretending (b =.545, p <.001), meaning that both variables are reciprocally related. Higher levels of circumventing parental control predict higher levels of online pretending and vice versa. Birth country, origin of the mother and family structure do not make a significant contribution in predicting online pretending. 38

40 Table 7 Summary of multiple regression analysis for online pretending Variables B SE (B) Beta p Constant Sex -.179** School grade.139*** Family structure Birth country Belgium vs. EU Birth country Belgium vs. non-eu Birth country father Belgium vs. EU Birth country father Belgium vs. non-eu.289** Birth country mother Belgium vs. EU Birth country mother Belgium vs. non-eu Circumventing parental control.545*** Note. **p <.01; ***p <.001. Discussion The objective of this study was to explore in which way gender, age, birth country, origin of both parents and family structure relate to circumventing parental control and online pretending. The first and second research question examined differences in prevalence (1) and in level (2) of circumventing parental control and online pretending by child/family characteristics. Since hypotheses for both questions of interest were made on the base of the same developmental and sociological frameworks, the research findings for these questions are discussed together, per child/family characteristic. Findings indicate that school grade results in differences in prevalence of circumventing parental control as well as of online pretending. More 5 th graders than 3 rd graders engage in these behaviours. The level of online pretending, but not of circumventing parental control, is also influenced by school grade, with 5 th graders showing higher levels of online pretending than 3 rd graders. The majority of these findings concerning school grade are consistent with the developmental urge for autonomy as children grow older (Pasquier, 2001). We expected this developmental process to be the same for both genders but the present data do not entirely support our assumption. Gender does not influence the prevalence and level of circumventing parental control. But inconsistent with our hypothesis, gender does result in differences in prevalence as well as in the level of online pretending. Not only do more boys than girls pretend online, boys also pretend more frequently on the internet than girls. An explanation could be found in that girls possibly experience more parental supervision than boys (Livingstone & Helsper, 2008), thus perceiving less opportunity to engage in online pretending (without getting caught by their parents). However, future research must shed more light on this gender difference. 39

41 Inconsistent results are also found as to the relationship between family structure and both types of secret behaviour. The present study reflects that more children from non-traditional than from traditional families pretend online and that the first group does so at a higher level. For circumventing parental control no differences in prevalence nor level were found resulting from family structure. While the results for online pretending support the suggestion of more deviant behaviour in children from a non-traditional family, resulting from less parental control and supervision (Apel & Kaukinen, 2008; Rebellon, 2002), findings for circumventing parental control contradict this Parental Absence Model (Gottfredson & Hirschi, 1990). It may be that the level of difficulty of the questioned disobedient internet behaviours might be of influence. Circumventing parental control (e.g. using code words, secretly having an address or going on the internet in absence of the parents) generally is less noticeable and easier to perform than online pretending (e.g. visiting websites or games with age restriction, online identity misrepresentation). The latter behaviour is more obvious and complex, so it only stays unnoticed if you have enough opportunity (parental absence) to engage in this behaviour. The explorative analysis revealed no differences in prevalence nor level of both secret online behaviours resulting from child s birth country. However, the origin of both parents relates strongly to the prevalence and level of circumventing parental control. The number of children that circumvents parental control is higher for the group with a mother born abroad (EU or non-eu) or a father born outside the EU, opposed to the group with a parent born in Belgium. Also, children with a parent born outside the EU circumvent parental control to a higher degree than children with a parent born in Belgium. For online pretending, country of birth of both parents only results in differences in level and not in prevalence of online pretending. Children with a parent born outside the EU engage in online pretending more frequently than children with a parent of Belgian origin. Although ethnic origin seems to be strongly related to circumventing parental control and online pretending, there is most likely no direct relationship. It is possible that the internet parenting style of parents born outside the EU is less effective because they have less internet experience due to limited internet access and modernization in their country of birth. In fact, various studies on the digital divide found that significant differences in internet experience, internet usage and internet attitude explain the relationship between migration status of the parents and internet parenting styles (Valcke et al., 2010; Wang, Bianchi, & Raley, 2005). The third research question was of a more exploratory nature and examined whether circumventing parental control and online pretending can be predicted by child/family characteristics. Linear regression analysis reveals a strong reciprocal relationship between circumventing parental control and online pretending: the more a child circumvents parental control, the more he pretends online and vice versa. This finding makes sense: conducting certain risky or secret internet behaviour will probably lower the threshold to engage in other types of online risk behaviour. The mother s country of birth contributes to the prediction of circumventing parental control, though its predictive power is small. Children with a mother born in the EU show more circumventing of parental control than children with a Belgian mother. Gender, school grade and birth country of the father are identified as strong predictors of online pretending, thus confirming the existence of relationships between these three child characteristics and online pretending which were already found in the previous research questions. Being a boy or a 5 th grader predicts higher levels of online pretending than respectively being a girl or a 3 rd grader. Also having a father originated outside the EU is associated with more online pretending opposed to having a Belgian originated father. 40

42 Implications and limitations These research findings have some implications for preventive actions and future research. Since almost half of all 3 rd graders and more than half of 5 th graders engage in online behaviour that is not supervised (e.g., use the internet in absence of parents) or that possibly brings harm to themselves or others (e.g., pretending to be older or someone else), parents and teachers should be educated in primary school already about how to communicate with, guide and inform children about internet use and online risks of harm. Complementary, internet courses should be provided for parents who lack skills and confidence when using a computer. Livingstone (2007) found that only 69% of parents feel very or fairly comfortable using a computer opposed to 92% of their children, showing an intergenerational difference in experience with computers and the internet. This lack of knowledge about IT makes it more difficult for parents to educate their children about the internet, to protect them from its risks of harm and to intervene if needed. Parents originated outside the EU should be encouraged to participate in both education and intervention, since their children show higher levels of hidden internet behaviour than children with Belgian parents. Some limitations of the present research have to be taken into account. Overall, this study provides us with a first exploration of circumventing internet parental control and online pretending in Flemish primary school children. For this reason, we definitely acknowledge that further research is needed to complete and verify found results. Firstly, since children were asked if and how frequently they engage in disobedient or secret behaviour (e.g., using code words to warn for parents watching along, purposefully misrepresenting themselves), some degree of falsification in answers is possible, despite reassurance of anonymity. Secondly, we did not control for parents internet experience or internet knowledge when looking at origin of both parents and family structure. Though, these two parental variables can affect the way in which parents control internet behaviour of their children, thus influencing the opportunity of children to engage in secret internet behaviour. Thirdly, the measures in this study showed a rather low reliability. In case of replication the addition of more scale items should be considered. Lastly, for future research it could be interesting to explore the link between online and offline risk behaviours (e.g., does online and offline risk (seeking) behaviour spring from a same underlying factor). This could provide more insight in the nature and development of risk behaviour in general. Acknowledgements We would like to thank all schools and pupils for their participation in this study, as well as the B-CCENTRE for their support. This research was financially supported by the European Commission, the Belgian Government and the Catholic University of Leuven, grant number HOME/2010/ISEC/AG/INT-011. References Apel, R., & Kaukinen, C. (2008). On the relationship between family structure and antisocial behaviour: parental cohabitation and blended households. Criminology, 46(1), Cassidy, W., Jackson, M., & Brown, K. N. (2009). Sticks and stones can break my bones, but how can pixels hurt me? School Psychology International, 30(4), Central Bureau of Statistics (2000). Standaarddefinitie van allochtonen. [Standard definition of foreigners]. Index, 10, Cole, J. I. (2001). The UCLA Internet Report: Surveying the Digital Future Year 2. Los Angeles, CA: UCLA Center for Communication Policy. Retrieved on January 7, 2014 from 41

43 Cortina, J. M. (1993).What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology, 78, Dowell, E. B., Burgess, A. W., & Cavanaugh, D. J. (2009). Clustering of internet risk behaviours in a middle school student population. Journal School of Health, 79(11), Duerager, A., & Livingstone, S. (2012). How can parents support children s internet safety? LSE, London: EU Kids Online. Facebook (2013). Statement of Rights and Responsibilities. Retrieved on February 19, 2014 from. Flemish Authorities (2011) Flemish education in images. Evere: Van Muysewinkel. Retrieved on October 10, 2011 from /statistisch_ jaarboek_pdfs/5_71_90.pdf Gottfredson, M., & Hirschi, T. (1990). A general theory of crime. Stanford, CA: Stanford University Press. Gross, E. F. (2004). Adolescent Internet use: What we expect, what teens report. Applied Developmental Psychology, 25, Harman, J. P., Hansen, C. E., Cochran, M., & Lindsey, C. R. (2005). Liar, liar: Internet faking but not frequency of use affects social skills, self-esteem, social anxiety, and aggression. CyberPsychology & Behaviour, 8(1), 1-6. Hasebrink, U., Livingstone, S., Haddon, L., & Ólafsson, K. (2009). Comparing children s online opportunities and risks across Europe: cross-national comparisons for EU Kids Online. (2 nd ed.). LSE, London: EU Kids Online. Haselager, G. (2004). Kleuter en schoolkind [Toddler and child]. In F. J. Mönks, & A. M. Knoers. Ontwikkelingspsychologie: Inleiding tot de verschillende deelgebieden (pp ). [Developmental psychology: Introduction to the different subfields]. Assen: Koninklijke Van Gorcum. Helsper, E. J., Kalmus, V., Hasebrink, U., Sagvari, B., & de Haan, J. (2013). Country classification: Opportunities, risks, harm and parental mediation. LSE, London: EU Kids Online. Livingstone, S. (2003). Children's use of the internet: reflections on the emerging research agenda. New Media and Society, 5(2), Livingstone, S. (2007). Strategies of parental regulation in the media-rich home. Computers in Human Behaviour, 23; Livingstone, S., & Bober, M. (2004). UK children go online: Surveying the experiences of young people and their parents [online]. London: LSE Research Online. Livingstone, S., & Haddon, L. (2009). EU Kids Online: Final report. LSE, London: EU Kids Online. Livingstone, S., Haddon, L., Görzig, A., & Ólafsson, K. (2011). Risks and safety on the internet: the perspective of European children: full findings and policy implications from the EU Kids Online survey of 9-16 year olds and their parents in 25 countries. LSE, London: EU Kids Online EU. Livingstone, S., & Helsper, E. (2008). Parental mediation and children s Internet use. Journal of Broadcasting & Electronic Media, 52(4),

44 Livingstone, S., Ólafsson, K., & Staksrud, E. (2011). Social networking, age and privacy. LSE, London: EU Kids Online. Mishna, F., Saini, M., & Solomon, S. (2009). Ongoing and online: Children and youth's perceptions of cyber bullying. Children and Youth Services Review, 31, Notten, N. (2013). Risicogedrag en het wereldwijde web. De invloed van gezin en samenleving op online risicogedrag van adolescenten vanuit een Europees perspectief. [Risk behaviour and the world wide web. The influence of family and society on online risk behaviour of adolescents from a European perspective]. Mens & Maatschappij, 88(4), Pasquier, D. (2001). Media at home: Domestic interaction and regulation. In S. Livingstone, & M. Bovill (Eds.). Children and their changing media environment (pp ). Mahwah, NJ: L. Erlbaum Associates. Rebellon, C. J. (2002). Reconsidering the broken homes/delinquency relationship and exploring its mediating mechanism(s). Criminology, 40, Roggemans, L., Cops, D., & Kolijn, S. (2013). De Gentse en Antwerpse jeugd onderzocht [Examination of Ghent and Antwerp youth]. In N. Vettenburg, M. Elchardus, J. Put, & S. Pleysier (Eds.). Jong in Antwerpen en Gent. Bevindingen uit de JOP-monitor Antwerpen-Gent (pp.29-47). [Being young in Antwerp and Ghent. Findings from the JOP-monitor Antwerp-Ghent]. Leuven/Den Hague: Acco. Stevens, J. P. (1992). Applied multivariate statistics for the social sciences (2 nd ed.). Hillsdale, NJ: Erlbaum. Teuwen, J., De Groof, D., & Zaman, B. (2012). Flemish Preschoolers Online: A mixed-method approach to explore online use, preferences and the role of parents and siblings. Paper presented at the Etmaal van de Communicatiewetenschap, Leuven, Belgium. Retrieved on March 18, 2014 from. Valcke, M., Bonte, S., De Wever, B., & Rots, I. (2010). Internet parenting styles and the impact on the internet use of primary school children. Computers & Education, 55, Valcke, M., Schellens, T., Van Keer, H., & Gerarts, M. (2007). Primary school children s safe and unsafe use of the Internet at home and at school: An exploratory study. Computers in Human Behaviour, 23, Valkenburg, P. M., Peter, J., & Schouten, A. P. (2006). Friend networking sites and their relationship to adolescents well-being and social self-esteem. CyberPsychology & Behaviour, 9, Walrave, M., Lenaerts, S., & De Moor, S. (2009a) Ouderlijke controle en vrijheid (p.76). [Parental control and freedom]. In J. Bauwens, C., Pauwels, C., Lobet-Maris, Y., Poullet, & M., Walrave (eds.). Cyberteens, cyberrisks, cybertools. Tieners en ICT, risico's en opportuniteiten. [Cyberteens, cyberrisks, cybertools. Teenagers and ICT, risks and opportunities]. Ghent, Academia Press. Walrave, M., Lenaerts, S., & De Moor, S. (2009b) Afspraken tieners & ouders (pp ). [Agreements teenagers & parents]. In J. Bauwens, C., Pauwels, C., Lobet-Maris, Y., Poullet, & M., Walrave (eds.). Cyberteens, cyberrisks, cybertools. Tieners en ICT, risico's en opportuniteiten.[cyberteens, cyberrisks, cybertools. Teenagers and ICT, risks and opportunities]. Ghent, Academia Press. Wang, R., Bianchi, S., & Raley, S. (2005). Teenagers Internet use and family rules: a research note. Journal of Marriage and Family, 67, Youtube (2010). Terms of service. Retrieved on February 19, 2014 from. 43

45 The Effectiveness of an Intervention to promote Awareness and Reduce Online Risk Behaviour of Belgian Primary School Children Schilder, Janneke D; Brusselaers, Marjolein B. J.; Bogaerts, Stefan Department of Developmental and Forensic Psychology, Tilburg University. The Netherlands Department of Criminal Law and Criminology, Catholic University of Leuven, Belgium Belgian Cybercrime Centre of Excellence for Training, Research and Education, Leuven, Belgium Correspondence concerning this article should be addressed to: Stefan Bogaerts, Department of Developmental and Forensic Psychology, School of Social and Behavioural Sciences, Tilburg University, 5000 LE Tilburg, The Netherlands, Electronic Mail may be sent to: Abstract The current study explored the effect of a school based intervention on online risk awareness and behaviour of primary school children (grade 4 and 6) at two measurements. Children received a presentation about the possible risks on the internet (intervention group) or a presentation about internet applications without emphasis on risk (control group). Children in the intervention group were more likely to be aware of online risks. Although diminishing, this effect on online risk awareness was still noticeable four months after the intervention. However, against our expectations the intervention did not seem to diminish the amount of online risk behaviour. Four months after the intervention children who received the intervention were even more likely to report online risk behaviour. In line with our expectations a gender difference was observed, girls were more likely to be aware of the risks directly after the intervention than boys were. Additionally, girls and children in a lower grade were less likely to behave in a risky manner when online. Additional experimental research is needed to explore the effects of school-based interventions on actual online risk behaviour of young children. Keywords: School based intervention, online risk awareness, online risk behaviour Introduction Growing up is nowadays inextricably linked to going online. At the global level, there is an increase of 676.3% in internet users since 2000 (World internet users and population stats, 2013). The current generation is the first that takes the existence of internet for granted. The EU Kids Online Survey (Livingstone & Haddon, 2009) reports that 9-16 years go online for on average 88 minutes per day. However, rather positive labels to this generation like whiz kids disguise the negative site of this increase in internet use. Reports, such as the EU kids online report, also point at the risky side of their internet behaviour (Livingstone & Haddon, 2009). De Moor and colleagues (2008) provide a classification of three different categories of risk that children can encounter when they are on the internet. The first category to be named is content risk. This category includes different kind of risks concerning possible harmful contents, for example websites showing naked or porn images. More than half of the teenagers accidently saw porn web sites when surfing on the internet (De Moor et al. 2008). Although not all children have negative effects when they face sexual pictures 44

46 on the internet, one quarter of the children were extremely upset (Mitchell, Jones, Finkelhor, & Wolak, 2014). Other examples of risks in the content category are violence or racism websites, but also refer to a lack of critical skills of children to judge the reliability of information that they see on the internet (De Moor et al., 2008; for an overview of negative effects also see Valcke, De Wever, Van Keer, & Schellens, 2011). The second named category by De Moor et al. (2008) is contact risk. Risks in this category all refer to activities on the internet towards known or unknown persons. Examples named by De Moor et al. (2008) include the disclosure of personal information such as the home address or phone number, cyber bullying, and chatting in risky settings. An example of the latter is that 16.1% of the children indicated that they were asked to give sexual information about themselves, and 10.6% even received a question to perform a sexual action for another person (De Moor et al., 2008). The third category is commercial risk. This category refers to the acts of commercial organizations that focus on the exploitation of internet users (De Moor et al., 2008). Examples are the abuse of personal information and spam. A mere part of the literature concerning risky behaviour can be classified under content and contact risks. The most common risks named in the EU kids online report (Livingstone & Haddon, 2009) refer to these two categories as well; chatting with online contacts not met before, cyber bullying, pornography, sexting, texting, websites with harmful user-generated content (e.g., hate, pro-anorexia, drug taking or suicide), personal data abuse, and excessive internet use. We will therefore include both content and contact risks when we refer to online risk behaviour in the current study. Although most surveys tend to focus on young adolescents, research on primary school internet users are demanded since children go online at ever younger ages. The average age of first internet use is seven years in Denmark and Sweden and eight in several Northern European countries (Livingstone & Haddon, 2009). Although the age at which children are going online is decreasing, teacher s engagement with children s internet use is least among 9-10 year olds (Livingstone & Haddon, 2009). Children in their early teens tend to gradually switch from compliance and commitment to their parents to behave like their peers (Fuligni & Eccles, 1993). They begin to assert their autonomy and are less compliant to parental instructions than young children (Pasquier, 2001). Also, this transition means that children explore the internet by themselves without the presence of their caregivers (Byron, 2008). Knowing this, it is necessary to intervene at an early stage of internet use to protect children from harm on the internet. It thus seems important that daily caregivers and coaches such as parents and teachers provide guidelines for the safe use of internet for young children. Supporting children in their internet use is also stressed in the Byron review (2008). Here it is stated that these children rely on others in making their choices since they are still immature in self-regulation, and their ability to inhibit and control impulses is still low. However, this is also the time that they start exploring website boundaries originally set for them by their parents. Byron (2008) makes a useful analogy between young children on the internet and young children in public swimming pools. Before we let children go to swim, we first teach them how to swim; there are lifeguards who watch them, and there are swimming aids available for the younger swimmers (Byron, 2008). Although risk taking is a developmental imperative in childhood, children should be taught how to use the internet safely when they explore it by themselves. The first policy alternative for the protection of minors in cyberspace that comes to mind would be parental control, logically because the ability of parents to restrict the amount of time that children can spend on the internet. However, when it comes to safety on the internet, parents came to rely on what Livingstone (2002, p.250) calls the European context of strict broadcasting regulation for protection of minors. Parents do seem to talk about the risks on the internet with their children (70%; Livingstone & Haddon, 2009). On the contrary, most children who have encountered an online risk did not tell their parents about this (Livingstone & Haddon, 2009). This might explain why only nine percent of the parents state that they do not need more information about the safety of their child on the internet. The majority of the parents state that they want more information from public organizations such as schools and local authorities (Livingstone 45

47 & Haddon, 2009). However, relying on parents solely to promote safer internet use in children would be problematic since not all parents are equally capable to carry out this task because of the current generation gap in internet skills between them and their children (Livingstone & Bober, 2006). There thus seems to be an increasing demand and responsibility for primary schools to develop educational and preventive programs concerning risky behaviour on the internet. This seems even more evident since school work is the main online activity reported by children (Livingstone & Haddon, 2009). The idea of giving the responsibility of informing about the internet to schools is in accordance with the policy recommendations made in the EU Kids Online report (EUKO; Lobe, Livingstone, Ólafsson, & Vodeb, 2011). In this report, it is stated that schools have the resources to reach all children and should therefore attempt to reach all children. However, as said before, the youngest group of children questioned in the EUKO survey (9-10 years old; Lobe et al., 2011) report the least mediation of teachers. This is an important finding when we take into account that this age group tends to lack skills and confidence to cope with risks they encounter on the internet (Haddon & Livingstone, 2012) and that children go online at an ever younger age (Livingstone & Haddon, 2009).More mediation of teachers in the young group of children could possibly reduce the risks that young children may encounter when on the internet. Although internet safety of young children is a regular topic in media, there is a scarcity of scientific studies testing the effectiveness of interventions to reduce online risk behaviour in (primary) schools. Informing primary school children about online risks could be an effective way to raise risk awareness. This idea is backed up by research and traditional models of behavioural change. The classic Theory of Planned Behaviour by Ajzen (1991) suggests that behaviour is dependent on one s intention to perform the behaviour. Behavioural intention, in its turn, is influenced by the attitude of a person with certain beliefs and values about the outcome of the behaviour, subjective norms based on normative beliefs, and his perceived behavioural control which depends on control beliefs. This behavioural theory suggests that it may be important to present information to change attitudes towards the wanted behaviour. The importance of information provision is also argued in the classical trans-theoretical stages of the change model by Prochaska (1998). This model suggests that people follow several steps toward behavioural change. It is argued that people will remain stuck in early stages due to a lack of motivation to move through the stages. Therefore it is essential to have a planned intervention to travel through the first steps of the model towards behavioural change. An indication made by Prochaska (1998) is to increase awareness of the risk, for example by educational materials and informing about the behaviour. Both classic models thus stress the importance of informing and awareness raising interventions to promote behavioural change, in this case to reduce online risk behaviour. The effectiveness of awareness raising campaigns for children, such as the well-known VERB campaign, is embedded in scientific literature (Huhman et al., 2010). VERB was a health marketing campaign from , targeting 9-13 year olds to be physically active every day. The awareness campaign entailed television advertising, school directed promotions, and community based activities. Children who were aware of the campaign were more likely to be physically active compared to the children who noted that they did not hear from the campaign. Another example of a successful awareness raising campaign was a school based anti-violence intervention for middle schools. The intervention included use of media and classroom presentations showing other ways to solve a conflict (Swaim & Kelly, 2008). The intervention had a positive effect on both the cognitive and behavioural aspects of the behaviour of the youngsters. These outcomes show that awareness raising interventions can have a positive effect on children s behaviour. Awareness campaigns are therefore widely used by local governments and the European Union to promote positive behaviour in children. Specifically focusing on risky behaviour on the internet there have been several awareness campaigns in the last decade (Valcke, Schellens, Van Keer, & Gerarts, 2007). However, most articles concerning these campaigns are descriptive in nature and there tend to be a lack of evaluative research concerning the effect of such interventions (for an overview see Valcke et al., 2007). To the best of 46

48 our knowledge, ours is the first study to compare two experimental conditions and its effect on online risk awareness and behaviour of primary school children. In line with the tradition of governmental and European activities to promote health behaviour by raising awareness, we will experimentally explore the effectiveness of an awareness raising intervention on online risk behaviour in primary school children. Our research will address whether and how such an intervention influences both young children s risk awareness and actual risk behaviour. Based on the discussed literature, we assume that a school-based educational intervention will raise the awareness of children and as a result reduce online risk behaviour. Furthermore, we will examine gender and grade difference in the effect of the intervention on both awareness and risk behaviour. Literature suggests that boys and older children tend to be more risk taking than their female peers and younger children (see for example Morrongiello & Rennie, 1998). The outcomes of the current study will hopefully provide a valuable starting point for further research concerning the reduction of online risk. Methods Procedure and sample This intervention study was conducted in the five provinces of Flanders (Limburg, Antwerp, East Flanders, West Flanders, and Flemish Brabant). Fifteen Flemish primary schools were selected from the three official Flemish Educational Networks, namely Community education, Subsidized publicly run education (e.g., schools of municipals) and Subsidized privately run education (e.g., Catholic education). 812 children from 22 classes and 15 schools were selected to participate in the intervention study. Our final sample consisted of 812 children from fourth (n = 350) and sixth (n = 462) grade. Only those children who participated in both measurements were included in the analyses. The sample consists of 51.2% girls (n = 391) and 48.2% boys (n = 391); for 5 children, gender information was missing. The majority of children (47.8%) are born in 2001 (n = 388) and 2003 (37.6%, n = 305). The average age (an approximation based on year of birth) in fourth graders was M = 10.10, SD =.37 and in sixth graders M = 12.13, SD =.39. The data of the current intervention study were collected from November to December 2012 (Time 1) and again in May 2013 (Time 2). The written questionnaires were assessed anonymously, classically and in the presence of a researcher. Prior to the first measurement a 10-minutes intervention was carried out per class by one researcher. Per province, classes were randomly assigned to the intervention or control group. Nine classes (n = 355) did not receive the intervention and 13 classes (n = 457) did. Children in the intervention group were shown a slide show presentation with oral explanations about risks associated with specific online behaviours. The control group received an informative presentation about two internet applications (Wikipedia and OpenStreetMap), without any emphasis on online risks. Immediately after the presentation (either intervention or control), children were asked to fill in a paper-and-pencil questionnaire. During the second measurement four months later in the same school year and the same classes, children who participated in one of the two conditions (intervention or control), were asked to fill in a questionnaire identical to that of the first measurement four months earlier. Because of a four months time interval, no memory bias was expected because it is assumed that over a period of four months, extinction occurs. Due to the request of almost all participating schools to guarantee total anonymity of the pupils, no individual numbers or scores could be matched over time what means that repeated measures were excluded because pupils in both measurements could not be linked to each other due to the absence of a unique identification number. 47

49 Measurements Risk behaviour. The dependent variable Risk behaviour was measured by 15 questions, all referring to online behaviour carried out in the last six months. This scale covered several online behaviours which can be considered as risky. The following topics of risk behaviour were questioned: talking, gaming, webcamming, meeting, or chatting with a stranger met online, getting to know someone online, providing the own home address or phone number on a public profile, having an address parents do not know of, using the internet without parents knowing, pretending to be older or someone else. An example question is: Have you got to know someone online you didn t meet in real life? Answer categories were dichotomous with 1 = never done this and 0 = I did do this at least once. The reliability of the scale was good with Cronbach s alpha =.77 at time 1 and alpha =.81 at time 2. Awareness. The dependent variable Awareness was measured with nine statements concerning a potentially risky activity on the internet. Children were asked to answer to each statement on a scale from 1 to 5 with 1= very safe, to 5= very unsafe. An example statement of a risky activity is: Sarah is a 10-year old girl. After school, she regularly chats on a website with girls and boys she does not know. That way, she gets to know new friends. The following subjects were included: chatting with strangers, using non- concealing chat alias, sharing cellphone number with strangers, opening an unknown attachment, webcamming with strangers, using public social network sites, clicking on fraudulent web links and pretending to be older. The reliability of the scale was good (Cronbach s alpha =.79 at both time 1 and time 2). Intervention and non-intervention. The independent variable in this study was whether or not the child received the intervention at time 1. The intervention addressed five topics concerning online risk behaviour. Included topics were: textual contact over the internet (e.g., chatting, grooming, bullying), audiovisual contact (e.g., bullying, extortion, strangers), social network services, online games, and offline meetings with people met online. For a detailed description of the intervention presentation, see the Appendix. The non-intervention group received a presentation about two internet applications (Wikipedia and OpenStreetMap), without any emphasis on online risks. Both presentations took approximately 10 minutes and were presented by a research associate. Covariates. Both grades and gender were entered as covariates in our analyses since both are expected to be related to the outcome variables in this study, namely online risk awareness and online risk behaviour. Both were dummy coded: grade with 0 = 4 th grade and 1 = 6 th grade and gender with 0 = girl and 1 = boy. Statistical analysis All data were analyzed using SPSS 19 for Windows and A p value of.05 was used to determine the significance of the effects. First, correlations between the covariates, independent and dependent variables were calculated to explore the data. To further analyse whether the condition (control/intervention) was associated with the two outcome variables (i.e., online risk behaviour and online risk awareness), hierarchical linear regressions were used. To identify the unique variance contributed by our main predictors, we first entered covariates into the linear regression models. Both gender and grade were used as control variables in all analyses to test our research question. The dichotomous variable condition (i.e. control/intervention) was entered second for both outcome variables (online risk awareness and online risk behaviour). These analyses were computed for both Time 1 and Time 2 in order to make a comparison between both measurements. To test whether online risk awareness modulates the effect of the condition on online risk behaviour, risk awareness was entered in the third step and a risk awareness x condition interaction term was added in a fourth and last step with online risk behaviour as outcome variable. 48

50 Results Correlations As expected, online risk awareness was positively correlated with participation in the intervention group and children in a higher grade were associated with more online risk behaviour. Boys were less online risk aware at Time 1, and showed more online risk behaviour at time 2 compared to girls. Surprisingly, online risk awareness correlates negatively with online risk behaviour at both measurements which means that consciousness and behaviour are not related to each other automatically (see Table 1). Online Risk Awareness and the Intervention Time 1. In the hierarchical linear regression analyses predicting online risk awareness, the covariates gender and grade entered in step 1 explained 1.9% of the variance at Time 1. Gender significantly predicted online risk awareness at Time 1. As expected, boys were less likely to be aware of online risks than girls did (β = -.160, t(806)= -3.77, p= <.001). Grade did not significantly predict online risk awareness. In the second step, condition added 9.8% to the variance. In total, the predictor and the covariates explained 11.7% of the variance in the online risk awareness scale, R 2 =.12, F (3, 803) = 36.70, p <.001. The experimental condition significantly predicted risk awareness (β =.39, t(806)= 9.54, p= <.001). As expected, participation in the intervention group was related to more online risk awareness while participation in the control group was associated with less online risk awareness. Time 2. In the hierarchical linear regression analyses predicting online risk awareness the covariates entered in step one explained.10% of the variance at Time 2. Both grade and gender did not significantly predict online risk awareness. In the second step, condition added 1.1% to the variance. In total the predictor and the covariates explained 1.2 % of the variance in the online risk awareness scale, R 2 =.001, F (3, 787) = 3.06, p =.03. The experimental condition significantly predicted risk awareness at Time 2 (β =.134, t(790)= 2.93, p=.004). Similar to the findings at Time 1, participation in the intervention group was related to more online risk awareness whereas participation in the control group was associated with less online risk awareness. Online Risk Behaviour, Online Risk Awareness and the Intervention Time 1. In the hierarchical linear regression analyses predicting online risk behaviour, the covariates entered in step one explained 4.1 % of the variance at Time 1. In the second step, condition added.01 % to the variance. Both gender and grade significantly predicted online risk behaviour at Time 1. As expected, boys were more likely to engage in online risk behaviour than girls did (β =.05, t(803)= 3.54, p= <.001). Also, children in the sixth grade were more likely to report online risk behaviours than children in the fourth grade (β =.06, t(803)= 4.54, p= <.001). In total, the predictor and the covariates explained 4.2 % of the variance in the online risk behaviour scale, R 2 =.04, F (3, 800) = 11.74, p <.001. The effect of the intervention was not significantly associated with online risk behaviour at Time 1 (β =.013, t(803)= 1.00, p=.317). Time 2. In the hierarchical linear regression analyses predicting online risk behaviour, the covariates entered in step one explained 4.4 % of the variance at Time 2. Gender was not related to online risk behaviour, however grade was. Children in the sixth grade were more likely to report online risk behaviours compared to children in the fourth grade β =.088, t(790)= 5.98, p=.026). In the second step, condition added.06 % to the variance. In total, the predictor and the covariates explained 5 % of the variance in the online risk behaviour scale, R 2 =.05, F (3, 787) = 13.68, p <.001. Whereas participating in the control group was related to less online risk behaviour, participating in the intervention group was associated with more online risk behaviour (β =.033, t(790)= 2.23, p=.026). Risk Awareness. In the third and fourth step of the linear hierarchical regression model predicting online risk behaviour, we examined for both Time 1 and Time 2 whether online risk awareness was associated 49

51 with online risk behaviour. Additionally we examined whether online risk awareness moderated the association of the condition with online risk behaviour. At Time 1, online risk awareness added another 8% of the variance (R 2 =.13, F (4, 799) = 26.68, p <.001) and 20% to the variance at Time 2 (R 2 =.25, F (4, 786) = 65.24, p <.001). Against our expectations, online risk awareness was significantly associated with more online risk behaviour at both times (T1: β = -.096, t(803)= -8.73, p = <.001; T2: β = -.146, t(790)= , p= <.001). The interaction term condition x risk awareness was not significantly associated with online risk behaviour at Time 2, but did at Time 1 (β = -.051, t(803)= -2.36, p =.019) showing that a higher level of risk awareness moderates the association between the intervention and online risk behaviour. Discussion Despite several policies to reduce the risks that children encounter on the internet, there is only a limited amount of studies focusing on the impact of school-based internet safety interventions (Valcke et al., 2007). To the best of our knowledge, ours is the first study to compare two experimental conditions and its effect on online risk awareness and behaviour of primary school children over time. We hypothesized that children who received the awareness raising presentation show an increase in online risk awareness and as a result of that a decrease in online risk behaviour. Furthermore, we assumed that boys and older children were more likely to behave more risky when online. The current study revealed that a school-based intervention has a positive effect on the online risk awareness of primary school children. Although diminishing, this effect on online risk awareness was still visible four months after the intervention. However, against our expectations the intervention did not seem to diminish the amount of online risk behaviour. Four months after the intervention children who received the intervention were even more likely to report online risk behaviour. Also a grade difference was observed, girls were more likely to be aware of the risks directly after the intervention than boys were. Additionally, girls and children in a lower grade were less likely to behave in a risky manner when online. As expected, the intervention had a positive effect on online risk awareness. Although the association between the intervention and more online risk awareness was stronger at the first measurement, it was still noticeable after four months. It thus seems that a relatively short intervention already raises the awareness of children about dangers on the internet. Our findings confirm previous literature concerning evidence based school interventions, which shows that children are more aware of the risks and wanted behaviour after they have received an intervention (Huhman et al., 2010; Swaim & Kelly, 2008). Compared to the control group, children who received the intervention showed no difference in online risk behaviour at the first measurement compared to children in the control group. This can be subscribed to the used measurement in which we asked for their behaviour in the past six months, so before the intervention. However, surprisingly is that the children who received the intervention engaged in more online risk behaviour four months after. The lack, or either the opposite effect, of an intervention effect on actual risk behaviour aligns with an overview concerning school-based interventions by Valcke et al. (2007). They stated that school-based safe internet interventions for fourth to sixth graders did not influence their actual reported internet behaviour. The intervention provided in the current study increased online risk awareness but was insufficient to actually change online risk behaviour in primary school children. A study by Luna and Finkelhor (1998) offers an explanation for the absence of behavioural change. Based on various prevention campaigns to promote positive behaviour, they suggest that a successful intervention program should repeatedly invite people to focus on skill development, include interactive instructional strategies and consider individual differences (Luna & Finkelhor, 1998). Since the intervention used in the current study lacks these characteristics, it is plausible that no decrease in online risk behaviour was induced. Moreover, the fact that the intervention was limited to transmission of knowledge induction of fear, a limited focus on the influence of attitudes and a mostly general approach (Luna & Finkelhor, 1998). 50

52 Contrary to our hypothesis, the current study indicates that (an increase in) awareness of online risk does not necessarily lead to a reduction of online risk behaviour. The model of behavioural change by Prochaska et al.(1998), which describes several steps toward behavioural change, can clarify this finding. Awareness raising is a key element to be able to travel through the first steps of the model toward behavioural change. However, it might be essential to increase self-efficacy to move from awareness to actual behavioural change (Ronda, Van Assema, & Brug 2001). The inclusion of more comprehensive programs which are not limited to the transmission of knowledge, as in the current study, seems necessary (Luna & Finkelhor, 1998). The current study revealed that gender differences are important to note in online risk studies. Boys were more likely to show online risk behaviour compared to girls. This difference in risk behaviour could be explained by the type of internet use and parental involvement. According to a Portuguese study in 8-17 year olds, most boys can be identified as intensive internet users without parental mediation, whereas most girls are classified as moderate users with parental mediation (Almeida, Delicado, & Alves, 2010). Because of their higher level of internet use and lack of parental involvement, it is plausible that boys stand a higher chance to (un)consciously engage in online risk behaviour. Despite careful consideration of the research model, our study shows some limitations. At first, due to strict anonymity regulations of the participating primary schools, using a prospective design was not possible. Individual scores of each pupil over time could not be matched, which made the use of repeated measures to examine individual change between two time periods impossible. Therefore, the design was cross-sectional in nature, making it difficult to draw any causal inference. Future research should focus on longitudinal studies to further explore the effect of a school based intervention to reduce both online risk awareness and behaviour. Second, the intervention in the current study was related to the cognitive ability of primary school children and therefore relatively short. A future study should invest in a longer term intervention embedded in the education program, teaching children technical skills to prevent them from harm when online (Luna & Finkelhor, 1998). In this context, the successful school-based program Net- Detectives to reduce online risk behaviour (Wishart, Oades, & Morris, 2007), is worth mentioning. Through computer-based role play, children became detectives who investigated misuse of school computers and practiced their computer skills with peers. The success of this study reveals that an intervention which includes a more active role of children as well as education about technical skills can achieve safer internet behaviour (Wishart et al., 2007). Strengths of the current study include the exploration of a relatively unexplored field of research. There is only a limited amount of evaluative studies focusing on the impact of school based internet safety interventions and even less include experimental conditions. The current study will hopefully provide a valuable starting point for further research concerning this topic. Furthermore, the current study confirms earlier research, showing that awareness and actual behaviour are separate components and one does not necessarily leading to another. Although many international campaigns about safe internet focus on the increase of awareness (Valcke et al., 2011), this thus not necessarily has the desired effect on behaviour itself. Additional research concerning the link between online risk awareness and online risk awareness is needed. Concluding, the current study sheds light on a relatively unexplored research field concerning the effect of a school-based intervention on both online risk awareness and behaviour. Our research reveals that a relatively short school-based intervention positively affects online risk awareness up to four months after the intervention. However, this did not automatically result in a reduction in online risk behaviour. In fact online risk behaviour even seems to increase four months after the intervention. Since most research concerning internet risk behaviour and interventions are descriptive in nature, more experimental 51

53 studies are necessary to explore the effects of school-based interventions on the actual risk behaviour of young children. Acknowledgment. The authors would like to thank Karel Demeyer for the description of the intervention as presented in the appendix. 52

54 Table 1. Correlations of the covariates (gender/grade), dependent (condition), and independent variables (online risk awareness/behaviour) for both Time 1 and Time 2. Condition Awareness T1 Awareness T2 Risk Behaviour T1 Risk Behaviour T2 Grade Gender Condition (1= intervention) -.322**.107** Awareness T1.322** ** ** Awareness T2.107** ** Risk Behaviour T ** *.161**.128** Risk Behaviour T **.074* -.210** Grade (1= 6 th grade) **.210** Gender (1= boy) ** ** * p= <.05, **p= <

55 References Almeida, A. N., Delicado, A., & Alves, N. A. (2010). Children and the internet in Portugal: A diversified portrait. In J. Mucha & K. leszczynska (Eds.), Society, culture and technology at the dawn of the 21 st century (pp ). Cambridge: Cambridge Scholars Publishing. Ajzen, I. (1991). The theory of planned behaviour. Organizational behaviour and human decision processes, 50, Byron, T. (2008). Safer children in a digital world: the report of the Byron Review: be safe, be aware, have fun. De Moor, S., Dock, M., Gallez, S., Lenaerts, S., Scholler, C., & Vleugels, C. (2008). Teens and ICT: Risks and opportunities. Retrieved July 6, 2010 from. Fuligni, A. J., & Eccles, J. S. (1993). Perceived parent-child relationships and early adolescents' orientation toward peers. Developmental psychology, 29, 622. Haddon, L., & Livingstone, S. (2012). EU Kids Online: national perspectives. Huhman, M. E., Potter, L. D., Nolin, M. J., Piesse, A., Judkins, D. R., Banspach, S. W., & Wong, F. L. (2010). The Influence of the VERB Campaign on Children's Physical Activity in 2002 to American Journal of Public Health, 100, 638. Livingstone, S. (2002). Young people and new media: Childhood and the changing media environment. Sage. Livingstone, S., & Haddon, L. (2009). EU Kids Online. Zeitschrift Für Psychologie/Journal of Psychology, 217, 236. Livingstone, S., & Bober, M. (2006). Regulating the internet at home: contrasting the perspectives of children and parents. Digital generations: Children, young people, and new media, Lobe, B., Livingstone, S., Ólafsson, K., & Vodeb, H. (2011). Cross-National Comparison of Risks and Safety on the Internet: Initial analysis from the EU Kids Online survey of European children. Luna, R., & Finkelhor, D. (1998). School based prevention programs: Lessons for child victimization prevention. Retrieved October 21, 2014 from. Mitchell, K. J., Jones, L., Finkelhor, D., & Wolak, J. (2014). Trends in unwanted online experiences and sexting final report. Crimes against children research center, University of New Hampshire: Durham. Morrongiello, B. A., & Rennie, H. (1998). Why do boys engage in more risk taking than girls? The role of attributions, beliefs, and risk appraisals. Journal of Pediatric Psychology, 23, Pasquier, D. (2001). Media at home: Domestic interactions and regulation. Children and their changing media environment: A European comparative study, Prochaska, J. O., Johnson, S. S., & Lee, P. (1998). The Transtheoretical Model of behaviour change. In E. Schron, J. Ockene, S. Schumaker, & Exum, W.M. (Eds.), The Handbook of Behavioural Change (pp ), 2 nd ed. New York: Springer. Ronda, G., Van Assema, P. and Brug, J. (2001b) Stages of change, psychological factors and awareness of physical activity levels in the Netherlands. Health Promotion International, 16, Swaim, R. C., & Kelly, K. (2008). Efficacy of a randomized trial of a community and school-based anti-violence media intervention among small-town middle school youth. Prevention Science, 9,

56 Valcke, M., Schellens, T., Van Keer, H., & Gerarts, M. (2007). Primary school children s safe and unsafe use of the Internet at home and at school: An exploratory study. Computers in Human Behaviour, 23, Valcke, M., De Wever, B., Van Keer, H., & Schellens, T. (2011). Long-term study of safe Internet use of young children. Computers & Education, 57, Wishart, J. M., Oades, C. E., & Morris, M. (2007). Using online role play to teach Internet safety awareness. Computers and Education, 48, World Internet Users and Population Stats (2013, December 31). Retrieved on October 28, 2014 from. 55

57 Appendix Here we describe how the intervention was conducted. The slides are built up using very easy wordings and graphics. The title The intervention presentation was titled Op de computer en het internet zitten kan soms gevaarlijk zijn which can be translated as Using computers and the internet can possibly be dangerous. When the title slide was shown, the children were explained what was meant with can possibly be dangerous. It was explained that using a computer isn t dangerous in se and they shouldn t be afraid of doing so but that it depends highly on how you behave and how aware you are of possible risks and ways to avoid them or cope with them. First topic: textual contact This first topic focusses broadly on risks associated with textual contact with others over the internet. The wordings and icons on the slides of this topic refer to chatting and ing and the children are given examples orally to make clear that this embraces all kinds of messaging on different platforms. The first risk of this activity that was discussed is the fact that you cannot be sure of the true identity of your chat-partner if you haven t met him or her before. A first slide depicts a chat/ partner who says Hi, I m Laura, I m 10 years old as a girl silhouette with a question mark. In the next slide, this girl is exchanged for an evil looking man. The slide now also features 2 danger -signs: one with an exclamation mark and the word strangers and one with a question mark and the word trust. The message given to the children is that, because you cannot know for sure the identity of strangers you have textual conversations with, one should be very careful not to trust strangers too much or too quickly. They are told be very cautious with passing on personal data to conversational partners (because of the risk of for example grooming, identity theft, etc.). A next slide points to the risks of being bullied in a textual conversation and getting computer virus infections; 2 danger sign warn for bullies and viruses. The children are given the message that bullying might happen not only on the playground but using on textual internet media too with some examples. It is also explained how a computer can be contaminated with a computer virus (e.g. by clicking on fraudulent links or receiving data transfers). In the following slide, this is overlaid with a big danger-sign and the word undesirable. Now the children are given more examples of possible risks that can occur via textual media: e.g. unwanted solicitation, grooming ). 56

58 Next topic: audiovisual contact The second part of the presentation starts off with a slide showing a webcam and a headset icon and the words webcam and talking. This part embraces all kinds of audiovisual contact: where contacts can see or hear the contact partner. The following slide shows the webcam and headset icon and a silhouette with a question mark on it. The children are now told that even they can hear and/or see their contact, they can still not be sure about the real identity of the conversational partner: they are taught that voice or even video pre-recordings could be used by malign people to adopt a false identity. In a next slide, a picture of a camera operator and a recording picture are added while the children are explained that everything they say or do on webcam or during an audiovisual call could be recorded by the contact and used against them (e.g. for bullying, extortion ) and that once something is posted online somewhere it can be very difficult to remove. The last slide on this topic adds four danger signs with the words bullies, undesirable, strangers and trust ; several examples of risks are given to the children with inter alia repetitions of the advices already given with the former topic. Third topic: social network services The first slide on this topic shows a number of logos of wellknown social network services. The children are asked if they recognize these logos and to give more examples they know of. The slide also depicts a silhouette of a boy with a question mark on it. Again the children are taught that, also using these media, once cannot be sure of the real identity of the people they have contact with. On the next slide, the boy silhouette is replaced with the evil man picture that was already used before. The virus picture (that was used in a former topic) reappears as well as 5 danger sign with the words bullying, undesirable, strangers, trust and viruses. The children are repeatedly warned about the different risks of bullying, computer virus infections and risks associated with trusting strangers too much or too quickly. Topic four: online games and contests This topic is started with a slide depicting several images luring people to enroll in online games and contests. The children are asked if they recognize these and what they are. In a next slide, these pictures are stained with the virus image used before and a danger sign with the word viruses. The children are taught that lots of these games or contests are fraudulent and clicking on these pictures might cause risks like for instance a computer virus infection. The children are told to be vigilant and sceptical about these contests and to be aware that they might be made by people with bad intentions. A next slide adds 4 more danger signs with the words trust, strangers, undesirable and one with a euro-sign ( ). The message to the children is to be very cautious with clicking on links and images 57

59 that might try to lure them into downloading malicious software or filling in personal data (e.g. filling in a cell phone number or address and unwillingly being subscribed to a costly SMS service or spamming plague). Last topic: offline meetings with people met online The last part of the presentation is about setting up a meeting with a stranger met online. After the title-slide, the image of 2 people handshaking is overlaid with a keyboard picture and a speech bubble with the text Hello, I m Laura, I m 10 years old, the same salutation used in the first slide of the topic on textual contact. Using the same text should ring a bell in the children s mind on what to expect. In the next slide, one of the handshaking figures is overlaid with 4 silhouettes: one of a girl, a boy, a woman and a man. The message being: you cannot know for sure who you are going to meet in real life when you set up a meeting with an online contact. In a next slide these silhouettes are replaced with the (by know well-known) image of the evil man. The last slide adds the already familiar danger signs with the words trust, strangers and undesirable. The children are told about some of the risks they might encounter when setting up a meeting with a malign contact and asked to share their thoughts. 58

60 The influence of parental supervision on online risk behaviour among young children: a gender perspective M.A. Missler & S. Bogaerts ABSTRACT Children use the internet from their first years of primary school onwards. Therefore, there is a need to study online risk behaviour at this early age, as well as the ways in which parents can guide the online behaviour of their younger children. The aim of the current study was to detect the influence of a specific form of parental mediation, parental supervision, on online risk behaviour among 1437 Flemish children in the 3 rd and 5 th grade of primary school. We were particularly interested in possible gender differences. Results of the regression and path analyses showed interesting gender patterns. For girls, online risk behaviour was negatively influenced by parental supervision and by school grade. For boys, family structure and school grade were significant predictors. This implies that when studying online risk behaviour among younger children, gender differences should be taken into account. While an active and communicative approach by parents seems to be effective in lowering online risk behaviour among girls, for boys it appears necessary to explore other strategies. Key words: online risk behaviour, children, parental mediation, gender differences Introduction Across Europe, many children are using the internet long before their tenth birthday (Livingstone, Haddon, Vincent, Mascheroni, & Ólafsson, 2014). While the European mean age of first use is about eight years old, British children use the internet for the first time at around six years. With the rise of mobile devices such as tablets and smartphones, which are easily accessible even for very young children, it is likely that the age that children go online for the first time will continue to decline. In light of these developments, there is an increasing need for parents to be aware of the advantages and of the risks of internet use and social media by young children (Livingstone, Haddon, Görzig, & Ólafsson, 2011). Previous research into online risks however focused on teenagers, starting from 12 years (e.g., Dowdell, 2013; Dowell, Burgess, & Kavanaugh, 2009; Liau, Khoo, & Peng, 2005; Livingstone & Helsper, 2008), assuming that the transition to secondary school makes these children more vulnerable to a range of online risks. However, since children make use of the internet before their 12 th birthday, it seems warranted to study risk behaviour on the internet among children considerably younger than 12 years, since, given their younger age, this group is probably less aware of what constitutes risky behaviour and less able to assess the impact of the internet. Leung and Lee (2011) indeed concluded that research into online risk behaviour among children should recognize the different developmental stages and the potential differential effects of risk-lowering practices for younger and older children. Also, at a younger age, parents could lay the foundation for supporting, controlling and advising their children in their online activities and behaviour with positive effects throughout the teenage years (Kerr & Stattin, 2000; Greenfield, 2004). Therefore, the aim of the current study is to detect the influence of a specific form of parental monitoring, namely parental control, on online risk behaviour among primary school children. We aim to study to what extent parents engage in these behaviours to support their children, what the effects of providing supervision are on the online risk behaviour of their children, and whether these effects depend on the child s gender. Below, we will first attend to online risk behaviour among children, before turning to the different strategies parents can pursue to guide their children. 59

61 Online risk behaviour by children As stated above, most research on online risks experienced by children focused on children from 12 years up. The most commonly reported worries by parents of children in this age group are their child coming across sexually explicit materials and their child meeting online strangers (Dowdell, 2013). Several studies showed that parental supervision strategies (such as checking the child s behaviour and joining the child when it is online), were not consistent with the risk of the child engaging in such face-to-face meetings with strangers in real life (Mitchell, Finkelhor, & Wolak, 2001; Liau, et al., 2005). Also, if their child shared negative experiences with them, the majority of the parents did not know where to report this (Dowdell, 2013). This is alarming, given that the risks are not negligible; Mitchell et al., (2001), in their telephone survey of 1501 children aged years, found that 19 percent of children had been a victim of unwanted online sexual solicitation during the past year. They reported several groups who were at special risk: girls, older children, children who used the internet more frequently, chat room visitors and children with psychological problems, such as depressive symptoms or children who had experienced a negative life event. Moreover, one fourth of the sexually solicited children reported high levels of distress after the incident. The authors concluded that those children that avoided chat rooms and were prudent in contacting strangers online were solicited less. Liau et al. (2005) examined 1124 adolescents between 12 and 17 years old. They reported several predictors of adolescents meeting someone in person whom they first encountered online. Significant predictors of this behaviour were: frequency of chatting and gaming behaviour (higher frequency heightened the risk), communication with parents (lower risk for adolescents who informed their parents about negative online experiences), disclosure of personal information (such as phone number and photo), amount of inappropriate messages received (such as pornographic spam), whether inappropriate websites had been visited, and type of internet advice heard (lower risk for children who were taught to never download anything and to never agree meeting anyone encountered online). Whether parents have the same worries when their children are younger than 12 years old and whether risk and protective factors differ between these age groups is still largely unknown. Some studies focused on online risk behaviour by children younger than ten years old. However, in most cases, a broader age range was measured and separate analyses for the youngest children were not explicitly performed. For example, Dowell et al. (2009) studied 404 children ranging from 9-15 years old, with a mean age of 12 years. Importantly, the clustering of several risk behaviours (such as posting personal information, corresponding online with an unknown person, online-initiated harassment, visiting sex sites and overriding internet filters or blocks) posed a child at serious risk rather than a single risky behaviour by itself (such as posting a personal picture). Furthermore, Leung and Lee (2011) studied 718 children between 9 and 19 years old. The stricter and the more involved parents were, the less internet risks their children were involved in. Older (and male) adolescents were more likely to be the target of online harassment, while older females were more likely to receive unwelcome invitations to share personal information. Also, adolescents who visited websites with pornographic or violent content tended to be older, male and less educated. A few studies reported specifically on a sample of younger children. Valcke, Schellens, Van Keer and Gerarts (2006) studied 1700 children attending the 4 th, 5 th and 6 th grade of primary school, with a mean age of 11 years. They concluded that risk behaviour was very high among these groups and only influenced by parental control over the online behaviour of the children. Furthermore, using an elegant longitudinal crosssectional design, Valcke, De Wever, Van Keer, and Schellens (2011), studied 10,000 pupils of the 4 th, 5 th and 6 th grade of 78 schools in Flanders, Belgium (mean age of the children was 11.6 years). The authors concluded that, on average, overall online risk behaviour was not very high. However, some specific risk behaviours warranted special attention, such as, notably, agreeing to a meeting with someone first encountered online. Thus, in line with parents worries, this is likely an important risk factor for the younger age groups too. Also, Livingstone, Bober and Helsper (2005) reported on their subsample of 9-11 year olds and found that the youngest children from their sample were least likely to have received information about internet risks. This is 60

62 remarkable, given that younger children are most in need of information and education about safe internet use, simply because they have little experience in dealing with online risks. Parental behaviour, which we will discuss next, seems to be crucial in guiding the internet behaviour of younger children. Parental monitoring of the child s online behaviour Most parents report general rules about the internet usage of their children (Dowdell, 2013; Duerager & Livingstone, 2012; Wang, Bianchi, & Raley, 2005), such as restrictions on the amount of time children may spend online. Parental mediation theory proposes that, beyond these general rules, a variety of interpersonal communication strategies may be used by parents to mediate potential negative effects of media on their children (Schofield- Clark, 2011). A second assumption of the theory is that the interactions between parents and children about media use facilitate children s socialization into society. Research based on parental mediation theory originally identified three mediation strategies used by parents, concerning the use of television: active mediation, restrictive mediation and co-viewing (Nathanson, 1999; Valkenburg, Krcmar, Peeters, & Marseille, 1999). Active mediation means communicating with children about the content they have seen; restrictive mediation refers to rules established by parents regarding the medium; co-viewing consists of watching along with children. Several studies concluded that these three traditional strategies were also applicable to internet use (Warren & Bluma, 2002); however, more recent research (e.g., Livingstone & Helsper, 2008) showed that there are some differences between how parents mediate the use of television and the online behaviour of their children. New strategies specifically applicable to the internet were identified, such as active co-use; interactive restrictions; technical restrictions and monitoring. These new strategies were thought of as necessary because of the greater interaction between the child and the device when using the internet, compared to watching television (Schofield- Clark, 2011; Nikken & Jansz, 2014). Moreover, using the internet is much more a solitary activity than watching television, which more or less forces parents to seek alternative strategies (Nikken & Jansz, 2014) for guiding and supporting their child. Active co-use acknowledges the active component which is more likely to be present when co-using the internet than when watching television together. Interactive restrictions refer to restrictions on instant messaging, downloading and playing online games. Using technical restrictions, such as blocking websites with potential harmful content, is the third internet-specific factor. Finally, monitoring refers to checking the online activities of the child after it has used the internet (Livingstone & Helsper, 2008). Importantly, Livingstone and Helsper (2008) concluded that neither active co-use, nor software-based strategies were effective in reducing online risk behaviour among children. This structure has been replicated by Sonck, Nikken, and De Haan (2013), except for active co-use. Nikken and Jansz (2014) found a slightly different structure (co-use, active mediation and restrictive mediation), and added a new style, namely supervision by the parents. Monitoring as described by Livingstone and Helsper (2008) and Sonck et al. (2013) could not be identified as a distinct strategy. Possibly, parental monitoring consists of many different behaviours initiated by parents to keep track of their children s behaviour. Liau, Khoo, and Ang (2008) for example, measured four different aspects of parental monitoring, namely parental supervision, parental communication, parental tracking and adolescent disclosure. They concluded that adolescents differ considerably from their parents in the degree of supervision (sitting with the child while he/she uses the internet); communication (talking about safety); and tracking (checking the browser history) that is reported. That is, parents reported to engage in these behaviours more than was perceived by their children. Low caregiver monitoring has been related to child engagement in online harassment (Ybarra & Mitchell, 2004), while higher levels of parental monitoring were found to be protective against internet addiction (Lin, Lin, & Wu, 2009), and to online risk behaviour (Valcke et al., 2006). Notably, Valcke et al. (2011) found no effect of parental supervision on the online risk behaviour of children. However, their measure is mainly focused on parents controlling the online behaviour of their 61

63 children at home, while in this study, we focus on parents being actively and constructively involved in guiding their children s online behaviour. Parental supervision (Nikken & Jansz, 2014) includes the parental supervision component as described by Sonck et al. (2013). It refers to the parent being physically available (in the same room) when the child uses the internet, while actively and regularly monitoring what the child is doing (for example by watching along). We focus on this strategy as Nikken and Jansz (2014) showed that this factor is especially important in the context of younger children. Warren and Bluma (2002) reported that one fifth of their sample of parents always used co-surfing and discussion of online content. Moreover, simply prohibiting certain online behaviours without much involvement of the parent s side does not seem to work effectively (Lee & Chae, 2007). Active involvement of the parent through open communication with the child about the internet and online activities seems to be much more effective in guiding children compared to prohibition only (Greenfield, 2004; Lee & Chae, 2007; Liau et al., 2005; Livingstone et al., 2005; Kerr & Stattin, 2000). In this way, an environment can be created in which the child feels free to discuss his or her online activities. For example, in her study of monitored and unmonitored child and adolescent chat rooms, Greenfield (2004) concluded that a close interaction between parents and their children about the use of the internet potentially protects children from risks on unwanted exposure to sexual and aggressive materials. Also, Kerr and Stattin (2000) concluded that the disclosure of information by the child predicted parental knowledge about their child s online behaviour much better than controlling and tracking of the same behaviour. Therefore, we explicitly focus on the active component of parental supervision. Given that most young teenagers use the internet from home (Dowell et al., 2009), parental behaviour has a great potential to lower online risk taking among this group. Since many studies (e.g., Dowdell, 2013; Livingstone, Kirwil, Ponte, & Staksrud, 2014; Livingstone et al., 2014) gave a descriptive overview of online risk taking behaviour among children and adolescents (i.e., measuring the amount and the character of the risk-taking behaviour), our aim is to relate both parent- and childrelated variables to the risk-taking behaviour of children, with an explicit focus on gender differences. In this way, a first step into developing more understanding of online risk-taking behaviour among young children will be presented. As little is known about the effects of parental supervision on risk behaviour among primary school children and its interaction with demographical variables such as gender, our research is largely exploratory and we will thus not develop specific hypotheses here. However, based on the ubiquitous support for an active approach in guiding children s internet behaviour, we do expect that parental supervision will be negatively related to online risk behaviour. This means that online risk behaviour increases when parental supervision decreases. We will focus on possible gender differences in the role of parental supervision and its potential interactions with other demographical variables. Method Participants Participants were 1,437 primary school children from the region of Flanders in Belgium. The sample consisted of 430 (29.9 %) third graders and 1007 (70,1%) fifth graders. Gender was equally distributed among the sample with 720 boys (50.1%) and 711 girls (49.5%). For 0.4 % of the sample, gender info is missing. The mean age of the respondents is 10.5 years (SD=1.05), with a range of 8-13 years. Procedure Sampling and recruitment was done in December A list including all primary schools in Flanders was obtained from the Ministry of Education. Each school was assigned a random number; the schools were subsequently sorted from small to large. To ensure that our sample would be representative for all provinces 62

64 and educational networks3, a formula was used: 1540 * <PctPupilsInProvince> * <PctPupilsInNetwork> children. Power analysis (effect size = 0.30; confidence interval = 95%, alpha = 0.5) showed that at least 1400 participants would be needed to ensure statistical analyses with enough power. To account for possible dropout of 10% (for example due to absence during the time of measurement), we aimed to recruit 1540 respondents. From the list of schools, for each category (a combination of province and educational network) we selected the ones appearing first on the list (i.e., with the smallest random numbers). These schools were contacted both by and telephone to ask for their participation. Also, we asked for the actual number of students in the 3rd and 5th grades. Note that we thus recruited whole grades and not individual children, which means that the actual number of pupils from each combination of province and network is slightly different from the calculated number for all combinations of provinces and networks. Following this procedure, we contacted 87 primary schools, of which 45% responded negative and 55% positive. From the 47 schools that were positive about participation, we selected 39 schools, quasirepresentatively distributed among the provinces and the educational networks. Participating schools distributed a passive informed consent form among the parents. The children were giving the opportunity to fill in the questionnaire during class. This questionnaire was pilot-tested among 78 3 rd and 5 th graders to check for comprehensibility and usability. Some changes were made to the questionnaire based on this pilot testing. One of the researchers was always available for answering questions during the actual administration. Measures Demographics The questionnaire started with several demographical items, namely age, gender, birth country of the parents and of the child (Belgium or another country), school grade (3 rd grade or 5 th grade) and whether the child lived in a single-parent or two-parent family. Parental supervision The amount of supervision of the parents was measured using two items, indicating whether parents were in the immediate vicinity when children accessed the internet and whether parents viewed what the children were doing online. Children could indicate their response on a 3-point scale for the first aspect (Never, Sometimes, Most of the time) and on a 4-point scale for the other. Response options for the latter varied from (Almost) never to Always. Risk behaviour Risk behaviour was measured by summing up a list of eight potential risky behaviours. Participants could indicate whether they had engaged in the risky behaviour by responding with Yes or No. Example items are: showing yourself with a webcam to somebody you have not seen before and making an appointment with someone you only know from the internet. Statistical Analyses Analyses were performed using Stata 13. Data were analyzed using ordinary least squares (OLS) regression analysis with heteroskedasticity-consistent standard errors and path analysis. First, we performed a regression analysis for the full sample to detect the influence of all independent variables (parental supervision, family 3 In Flanders, there are three educational networks: community education, subsidized publicly run education, and subsidised privately run education. 63

65 status, country of birth of the child and of the parents, school grade, age and gender) on the dependent variable, online risk behaviour. Next, a robustness check was performed to confirm the status of parental supervision as an exogenous variable; not being determined by the other predictors. After all, parental supervision may be argued to be determined by the same demographic variables that explain risk behaviour. Therefore, we performed a path analysis using the pathreg command in Stata ( Finally, to analyze possible gender differences, we performed a regression analysis in which all predictors were interacted with gender. This results in gender-specific estimates of all variables on risk behaviour. The analysis allows for tests of the equality of effects in the subgroups of boys and girls. Results The regression model for the full sample was significant (F (8 ; 1188) = 3.79; p < Two percent of the variance in online risk behaviour can be explained from the variance in the included predictors (R 2 = 0.02). Parental supervision (b*= -.32; t = -2.53; p < 0.01; 95 % CI [ ] and school grade (b*= -1.29; t = -3.77; p < 0.00; 95% CI [ ) turned out to be significant predictors of online risk behaviour (Table 1). The subsequent path analysis showed that the effect of parental supervision is not determined by the other predictors. None of the predictors reached significance when predicting parental supervision, while parental supervision (β = -.072; t = -2.49; p=0.01) and school grade (β = -.13; t = -3.64; p <0.00) were again significant in explaining online risk behaviour. Thus, the robustness check showed that the effect of parental supervision is independent of the effect of the controls. The model testing for gender differences was significant (F (14 ; 1182) = 2.65; p < 0.00). Three percent of the variance in online risk behaviour can be explained by this model (R 2 = 0.03) (Table 1). Results showed important differences between girls and boys: for girls, but not for boys, parental supervision was a significant predictor of online risk behaviour (b*= -.34; t = -2.09; p < 0.04; 95 % CI [ ]). The effect reported for the whole sample is thus based on the effects for girls only. Also, school grade showed a significant effect for girls (b*= ; t = -3.40; p = 0.00; 95 % CI [ ]), indicating that girls from the 5 th grade displayed less risk behaviour, compared to girls from the 3 rd grade. For boys, living in a single-parent family (b*= ; t = -2.22; p < 0.03; 95 % CI [ ]) was a significant predictor. Also, for boys, school grade was a significant predictor (b*= ; t = -2.18; p = 0.03; 95 % CI [ ]). However, this effect is somewhat less strong than for girls. 64

66 Table 1 Regression Model with all Interaction Terms for Online Risk Behaviour b* all children b* Girls b* Boys Parental supervision -.32* -.34* -.31 Family status * Country of birth Country of birth father Country of birth mother School grade -1.29** -1.56** -1.00* Age Gender -.08 Constant 7.98** 8.14** 8.14** Notes: * p <.05. ** p < p <.10. G refers to girls, while B refers to boys. Discussion The aim of the current study was to detect the influence of parental supervision on online risk behaviour in a sample of young children attending the 3 rd and 5 th grade of primary school. Results show that a specific form of parental mediation, parental supervision, indeed predicts online risk behaviour of the child. Interestingly, different patterns of results were found for boys and girls: for girls, parental supervision appears to be more important in preventing online risk behaviour than for boys. For boys, demographical factors, such as school grade and country of birth appear to have a stronger influence on their online risk behaviour. Predictors of online risk behaviour among children thus seem to be gender-specific. Below, we will discuss the patterns of relationships, starting with the model for girls. For girls, parental supervision seems to buffer against online risk behaviour of pre-pubertal girls. These girls benefit from parents who are physically available when they are using the internet and from active and regular monitoring of their online activities. This is consistent with research showing that an active approach of the parents is most helpful in guiding the child s behaviour (Greenfield, 2004; Lee & Chae, 2007; Liau et al., 2005; Kerr & Stattin, 2000). It is possible that this approach is more effective for girls compared to boys because they are more in need of discussing their online activities with their parents, and therefore they benefit more from a climate where this is facilitated (because the caregiver is available for discussion and questions). 65

67 Indeed, Kerr and Stattin (2000) found that, spontaneously, girls are disclosing more to parents than boys do. Also, since girls seek more social support than boys do (e.g., Eschenbeck, Kohlmann, & Lohaus, 2007), girls are possibly better in co-facilitating an open climate in which there is room for discussion. That is, because they actively seek the support of the parent, they might strengthen the active role of the parent by stimulating parental self-efficacy, in that the parent recognizes that his or her availability is valued and effective which has subsequent positive effects on the future fulfillment of the active role by the parent (cf. Coleman & Karraker, 1998). In this way, the parent and the child stimulate each other to create an environment in which the child can optimally benefit from the many advantages of the internet, while being cautious for its risks. Differences in development and maturation could also play a role. As Jacobson and Crockett (2000) suggested, since girls mature earlier than boys, it could be that parental monitoring is especially important for them at the end of primary school already, while it grows in importance for boys when they reach puberty in later years. Notably, Valcke et al. (2011) found no effect of parental supervision on online risk behaviour. This could be due to the measure of Valcke et al. (2011) being more focused on parental control (rules and regulations), while our parental supervision measure is explicitly focused on parental availability and communication with the aim of offering support and guidance. For boys, parental supervision did not predict online risk behaviour. Future research is necessary to determine what strategies are helpful in supporting boys to use the internet safely. This is important since boys seem to experience more risks and also a broader range of risks than girls (Livingstone et al., 2005). Country of birth influenced risk behaviour among boys, in that boys that were born outside of Belgium were slightly more likely to display a variety of online risk behaviours. This is consistent with earlier research finding that there is a large between-country variation across Europe in the involvement of parents with the online behaviour of their children, with children from Eastern and Southern European countries explicitly wishing that their parents would be more involved (Livingstone et al., 2011). While we could not make a fine-grained between-country analysis here, it is likely that parents who are originally born in those countries display less oversight than their West-European counterparts. It is interesting that this effect turns out to be relevant only for boys. It could be that boys are more susceptible to a knowledge gap between them and their parents, less willing to try to explain to their parents what they are doing online, even when parents make serious efforts to discuss these issues. Maybe girls, due to their greater need for active guidance and relatively more matured brains, succeed much more in bridging the gap between them and their parents. However, these are still speculative suggestions and future research among different ethnic groups should shed more light on these issues. Remarkably, single parenthood was a buffer against online risk behaviour for boys. Earlier research pointed to the opposite effect: that single parents are less able (due to time constraints and additional burden) to supervise their children, which heightens their chances on online risk behaviour (Notten, 2013). However, it could be, that boys, since they seem less in need of an environment in which active and open communication is co-facilitated, benefit from a situation in which a parent has to deal with many different demands and in which they are given more freedom and responsibilities. Moreover, not all single parents are overloaded, therefore, for the less matured boys, having clear rules (because only one parent is involved) could be beneficial for them to refrain from online risk behaviour. For both girls and boys, school grade was a significant predictor of online risk behaviour. For girls, this effect was even stronger than for boys. Children from the higher grades were less likely to display online risk behaviour. This is consistent with Valcke et al. (2011), who found that children from lower grades displayed significantly more risk behaviour than children from higher grades. Consequently, it seems that as children age, they are more aware of potential risks and therefore display less online risk behaviour. This again points to the importance of research into online risk behaviour among younger children since they seem to face the most risks. As in all research, also this study has its limitations. First, we were only able to measure the perspective of the children. Triangulation of the data, in which also parents and teachers would participate, has the 66

68 potential to shed new light on the development of online risk behaviour among young children, and the role parents and schools could play to prevent this behaviour from occurring. Second, while we found significant effects of different predictors for the full sample and the subsamples, the models did explain relatively little variance in risk behaviour. Other variables we were not able to measure, such as, school-related variables or peer characteristics, will likely add to the predictive power of the models. Third, while we focused on a group of young primary school children, we were not able to assess risk behaviour among children younger than eight years. It seems plausible, that, just as has happened in the UK, the mean age when children go online for the first time, will get lower also for other European countries. Researchers should follow this trend and incorporate these younger age groups (6-8 years) into their research designs. As mentioned by Valcke et al. (2011), this has important consequences for research instruments, because children younger than eight years have less developed writing and reading skills. This was an issue we faced also in the current research project. Qualitative designs incorporating interviews could offer a solution to this problem. Moreover, with in-depth interviews, more insight into the mechanisms underlying online risk behaviour and parental supervision can be obtained. Also, designs in which researchers read the questions out loud to the children and let them pick an answer by using for example pictures, could be a promising alternative. Either way, designs in which young children are studied ask for more resources, but have the potential to offer new insights which are heavily needed in this rapidly developing field. In conclusion, the current study showed that when studying online risk behaviour among children, at least among primary school children, gender differences should be taken into account. When predicting online risk behaviour, we found that for girls, parental supervision was an important buffer against online risk taking, while for boys, demographic factors were much more important, such as country of birth. Future research should assess in what way parents can guide young boys in their online behaviour. For practice, these results have important consequences. Policy makers should realize that girls and boys could have differential needs regarding guidance behaviour from their parents. For girls, caregivers offering active oversight while facilitating an environment in which there is room for discussion and learning can support them in using the internet in a responsible way, taking full advantage of its possibilities. For boys, it is much less clear what parents can do to reach the same effect. It could be that boys benefit more from being guided on a group level, since peers influence the behaviour of boys more than that of girls (Rose & Rudolph, 2006). Future research, incorporating a variety of parental mediation strategies, could shed more light on this issue. Most important, parents do influence the online risk behaviour of their children, and should be educated about the ways in which they can guide and support them, starting early in primary school. References Coleman, P.K., & Karraker, K.H. (1998). Self-efficacy and parenting quality: Findings and future applications. Developmental Review, 18, Eschenbeck, H., Kohlmann, C., & Lohaus, A. (2007). Gender differences in coping in children and adolescents. Journal of Individual Differences, 28, Dowdell, E.B. (2013). Use of the internet by parents of middle school students: Internet rules, risky behaviours, and online concerns. Journal of Psychiatric and Mental Health Nursing, 20, Dowell, E.B., Burgess, A.W., & Cavanaugh, D.J. (2009). Clustering of internet risk behaviours in a middle school student population. Journal of School Health, 79, Duerager, A., & Livingstone, S. (2012). How can parents support children s internet safety? EU Kids Online, London, UK. Greenfield, P. M. (2004). Developmental considerations for determining appropriate Internet use guidelines for children and adolescents. Applied Developmental Psychology, 25,

69 Jacobson, K.C., & Crockett, L.J. (2000). Parental monitoring and adolescent adjustment: An ecological perspective. Journal of Research on Adolescence, 10, Kerr, M., & Stattin, H. (2000). What parents know, how they know it, and several forms of adolescent adjustment: Further support for a reinterpretation of monitoring. Developmental Psychology, 36, Lee, S., & Chae, Y. (2007). Children s internet use in a family context: Influence on family relationships and parental mediation. Cyberpsychology & Behaviour, 10, Leung, L., & Lee, P.S.N. (2011). The influences of information literacy, internet addiction and parenting styles on internet risks. New Media & Society, Liau, A.K., Khoo, A., & Peng, H.A. (2005). Factors influencing adolescents engagement in risky internet behaviour. CyberPsychology & Behaviour, 8, Liau, A.K., Khoo, A., & Ang, P.H. (2008). Parental awareness and monitoring of adolescent internet use. Current Psychology, 27, Lin, C., Lin, S., & Wu, C. (2009). The effects of parental monitoring and leisure boredom on adolescents internet addiction. Adolesecence, 44, Livingstone, S., Bober, M., & Helsper, E. (2005). Internet literacy among children and young people: findings from the UK Children Go Online project. Londen: LSE Research Online. Retrieved October 29, 2014, from Livingstone, S., & Helsper, E. (2008). Parental mediation and children s Internet use. Journal of Broadcasting & Electronic Media, 52, Livingstone, S., Haddon, L., Görzig, A., and Ólafsson, K. (2011). Risks and safety on the internet: the UK report. LSE, London: EU Kids Online. Livingstone, S., Haddon, L., Vincent, J., Mascheroni, G., & Ólafsson, K. (2014). Net Children Go Mobile: The UK Report. London: London School of Economics and Political Science. Livingstine, S., Kirwil, L., Ponte, C., & Staksrud, E. (2014). In their own words: what bothers children online? European Journal of Communication, 29, Mitchell, K.J., Finkelhor, D., & Wolak, J. (2001). Risk factors for and impact of online sexual solicitation of youth. JAMA, 285, Nathanson, A. (1999). Identifying and explaining the relationship between parental mediation and children s aggression. Communication Research, 26, Nikken, P., & Jansz, J. (2014). Developing scales to measure parental mediation of young children s internet use. Learning, Media, & Technology, 39, Notten, N. (2013). Risicogedrag en het wereldwijde web. De invloed van gezin en samenleving op het online risicogedrag van adolescenten vanuit een Europees perspectief. [Risk behaviour and the world wide web. The influence of family and society on online risk behaviour of adolescents from a European perspective]. Mens en Maatschappij, 88, Path analysis. UCLA: Statistical Consulting Group. Retrieved September 16, 2014, from 68

70 Rose, A.J., & Rudolph, K.D. (2006). A review of sex differences in peer relationship processes: Potential tradeoffs for the emotional and behavioural development of girls and boys. Psychological Bulletin, 132, Schofield Clark, L. (2011). Parental mediation theory for the digital age. Communication Theory, 21, Sonck, N., P. Nikken, and J. De Haan Determinants of Internet Mediation: A Comparison of the Reports by Parents and Children. Journal of Children and Media, 7, Stattin, H., & Kerr, M. (2000). Parental monitoring: a reinterpretation. Child Development, 71, Valcke, M., Schellens, T., Van Keer, H., & Gerarts, M. (2006). Primary school children s safe and unsafe use of the internet at home and at school: An exploratory study. Computers in Human Behaviour, 23, Valcke, M., De Wever, B., Van Keer, H., & Schellens, T. (2011). Long-term study of safe internet use of young children. Computers & Education, 57, Valkenburg, P. M., Krcmar, M., Peeters, A. L., & Marseille, N. M. (1999). Developing a scale to assess three styles of television mediation: Instructive mediation, restrictive mediation, and social coviewing. Journal of Broadcasting and Electronic Media, 43, Wang, R., Bianchi, S.M., & Raley, S.B. (2005). Teenager s internet use and family rules: a research note. Journal of Marriage & Family, 67, Warren, R., and A. Bluma. (2002). Parental mediation of children s internet use: The influence of established media. Communication Research Reports, 19, Ybarra, M.L. & Mitchell, K.J. (2005) Exposure to Internet pornography among children and adolescents: a national survey. 69

71 Privacy, Risk, Information protection and Social Network Site-Using Behaviour in a Sample of Flemish University Students Zsuzsanna Horváth, Stefan Bogaerts and Karel Demeijer Abstract Hundreds of millions of people from all age categories use social network sites and share a wide range and enormous amount of information of themselves and others, such as personal details and connections with friends, often without realising the potential privacy, security and profiling risks they take. Objectives: The current paper intends to examine influential factors contributing to the degree to which university students are concerned about their privacy and risk, and the extent to which they engage in information protecting behaviour. More specifically, we intend to shed light on predictors, such as unpleasant experience when using social network sites, experience of privacy invasion, type of education (law, informatics and criminology), gender, age, and the personality trait extraversion in a sample of university students. Additionally, a description of social network site-using behaviour of the participants is also provided. Based on the relevant literature, the following hypotheses are addressed: (1) Age, gender, type of education, unpleasant experience, extraversion, and privacy invasion are expected to significantly contribute to privacy concerns. (2) Age, gender, type of education, unpleasant experience, extraversion, and privacy invasion are expected to significantly contribute to risk concerns. (3) Age, gender, type of education, unpleasant experience, extraversion, and privacy invasion are expected to significantly contribute to information protection. Extraverted students have a larger online social network size. Results: The results suggest education type and privacy invasion to be significant predictors of privacy concerns, with privacy invasion having the largest contribution. The findings also imply that education type, unpleasant experience and privacy invasion are significant predictors of risk concerns, with unpleasant experience having the largest contribution. The role of education type, age, extraversion in information protection was also found to be significant, with extraversion having the largest contribution. Results regarding students social network site-using behaviour reveal that extraverted students have a larger online social network size, the majority of the participants have a Facebook profile, are most active on Facebook, use social network sites more than twice per day, spend between 1 and 5 hours a week with checking their profiles, have their total number of friend between 201 and 300, and update their profile more than twice a week. Most of the participants have read the privacy policy but not carefully. Most of the students who have read the privacy policy understood it as a whole. And finally, the lack of interest was generally the reason for not reading the privacy policy. Conclusion: Further studies are required to investigate the relationship among these variables to raise people s awareness of privacy, risks, and information protection. Furthermore, the identified influential factors, such as experience of privacy invasion and education type should also be taken into account in future research. 70

72 Introduction Hundreds of millions of people of all ages use social network sites and share a wide range and enormous amount of information of themselves and others, such as personal details and connections with friends often without realizing the potential privacy, security and profiling risks they take. Trading personal information without permission and identity theft are common violations of consumers privacy. Therefore, in order to prevent misuse of personal data on social media networks, it is essential for internet users to be aware of potential danger of their privacy and the possibility of personal data protection. The current paper intends to examine influential factors contributing to the degree to which university students are concerned about their privacy and risk, and the extent to which they engage in information protecting behaviour. More specifically, we intend to shed light on predictors, such as unpleasant experiences when using social network sites, experiences of privacy invasion, type of education (law, informatics and criminology), gender, age, and the personality trait extraversion in a sample of university students. Additionally, the relationship between the personality trait extraversion and online social network size and a description of social network site-using behaviour of the participants is also provided. Privacy, risk, information protection, influential factors, and Social Network Site- Using behaviour Facebook, MySpace and other social network sites store an enormous amount of information about their users and such amount of personal information requires high confidentiality. People especially students are getting increasingly concerned about privacy and cyber- security of their data (Bryer & Chen, 2010). Social network sites provide the users with the option to hide personal details, such as birth date, address or phone number. Furthermore, it is also possible to make such details visible only for those accepted as friends. Social network sites constantly work on improving their privacy settings but full protection against privacy violations is not possible (Chahal, 2011). With the help of certain technology companies, people can be easily tracked down and monitored (Acohido, 2011; Murphy, 2012). The other side of the coin is that using social network sites can be essential for students as they use social network sites on a daily bases (Smith & Caruso, 2010) not only to keep contact with friends but also for education purposes (Roblyer, McDaniel, Webb, Herman, & Witty, 2010). Thus, the question arises: Is using social network sites beneficial or rather harmful for the individual? Govani and Pashley (2005) found that even though the majority of the students are aware of the risks, they take by providing or restricting private information, they do not take actions in order to protect their personal data. In another study Newk-Fon Hey Tow, Dell and Venable (2010) reported that the reason for this lack of actions might be that students are either not aware of the risks or they perceive them as too low to take protective actions. The literature suggests the existence of several factors influencing the degree to which individuals are concerned about their privacy and risk, and the extent to which they engage in information protecting behaviour, such as age, gender, extraversion and unpleasant experience when using social network sites (Byrnes, Miller, & Schafer, 1999; Christofides, Muise, & Desmarais, 2009; Petronio, 2002; Skeel, Neudecker, Pilarski, & Pytlak, 2007). One of these factors is age. According to Christofides et al. (2009), it seems to be that adults use their privacy settings more often than youngsters. The authors also found that it is common to know how to change the privacy settings both among youth and adults; however, the actual usage seems to be rather limited in both groups. On the other hand, the Pew Internet & American Life Project (Lenhart & Madden, 2007) showed a decreased use of the privacy settings as age increases. Several studies have implied the presence of gender differences in privacy and risk concerns, and reported that women/girls are more concerned about these issues than men/boys (e.g., Lenhart & Madden, 2007; Moscardelli & Divine, 2007; Moscardelli & Liston-Heyes 2004; Fogel & Nehmad, 2009). An unpleasant experience in the past also plays an essential role in extent to which students engage in information protecting behaviour. Christofides, Muise, and Desmarais (2012) reported that adolescents who have had unpleasant experiences with using the internet are more concerned about the potential risks and tend to restrict their profile, with knowledge of privacy settings serving as a mediator between the two variables. Sociability refers to the individual s social presence in other words, one s ability to socialize which is considered as the most important aspect of 71

73 extraversion. Individuals who meet the extraversion trait can be described as active and assertive. They are generally positive in life and get inspiration from others. Extroverts are especially outgoing and energetic while introverts are solitary and reserved (Costa & McCrae, 1992). Extraverted individuals like having a lot of people around them and enjoy getting to know new people; thus, based on the fact that social network sites are the online representation of one s real social life, respondents with a high extraversion trait, measured by the relevant subscales of the NEO-Five Factor Inventory (NEO-FFI) have been reported to have a larger online social network size (Ong, Ang, Ho, Lim, Goh, & Lee et al. 2011). Extraversion has also been found to be relevant in the context of privacy concerns, given that it reflects one s sociability, social activity, flexibility, and interpersonal activity (Digman, 1990), thus, more extraverted people disclose more information about themselves and are less likely to protect their personal data. Additionally, Skeel et al. (2007) found that extraverted individuals tend to have high risk taking scores in general. Taking into account the fact that the popularity of social network sites among students is increasingly growing Golder, Wilkinson, Bernardo and Huberman (2007) reported that 90% of American undergraduate students use at least one social network site regularly. Similar results have been found in the United Kingdom among young adults using social network sites on a daily bases (Comscore, 2011). Furthermore, it has been shown that the majority of university students spend tremendous quantity of time using various social network sites (Budden, Anthony, Budden, & Jones, 2007). Despite the increasing popularity of social network sites among university students, the literature addressing this issue is rather limited (Boyd & Ellison, 2007; Ross Orr, Mia Sisic, Arseneault, Simmering, & Orr, 2009). As the literature lacks information about the influence of experience of privacy invasion and the participants study programme on the degree of privacy and risk concerns and the extent to which students engage in information protecting behaviour, the current study intends to shed light on these issues as well. We assume that students who have previously experienced that their personal information was used against them or made them feel uncomfortable by for instance, an embarrassing photo or a video has been posted on a social network site without their permission, are more concerned about their privacy and the potential risks, therefore, protect their personal data more. Furthermore, we also expect to notice differences across education type (or current study program). As informatics students have the best computer skills, it is expected that they engage in information protecting behaviour to a greater extent than law or criminology students, therefore, assumed to be less concerned about privacy and potential risk factors. Moreover, law and criminology students are expected to be the most concerned about their privacy due to the fact that they learn a lot about privacy issues in general, however, might lack the necessary computer skills to protect their data. Based on the relevant literature, the following hypotheses are addressed: (1) Age, gender, type of education, unpleasant experience when using social network sites, extraversion, and experience of privacy invasion are assumed to significantly contribute to privacy concerns. (2) Age, gender, type of education, unpleasant experience when using social network sites, extraversion, and experience of privacy invasion are expected to significantly contribute to risk concerns. (3) Age, gender, type of education, unpleasant experience when using social network sites, extraversion, and experience of privacy invasion are anticipated to significantly contribute to information protection. (4) The number of online friends is assumed to significantly correlate with extraversion. Methods and Materials Data have been collected from 455 Flemish university students. 67% of the participants were females and 33% were males. The mean age of the patients during the research was 22.9 years (SD=2.14; range years). 47.3% were law students, 10.9% were informatics students, and 41.8% were criminology students. All participants were officially enrolled to their full-time study program at the University of Leuven in Belgium. The participants have been contacted via their university , and those who agreed to participate were presented with a link to access the standardised online questionnaire. The program Limesurvey was used 72

74 to conduct the survey, which was also available via facebook groups created and used by law, informatics and criminology students of the same university. All questions and answers were provided in Dutch. The dependent variable privacy concerns was assessed by the item: To what extent are you concerned about your privacy on the social network sites? The answer categories range from very concerned to very unconcerned, and were recorded on a five point Likert scale. Risk concerns were studied based on the composite scored created from the 6 items assessing potential sources of risk on social network sites, such as: Are you concerned about identity theft? The answer categories range from very concerned to very unconcerned, and were recorded on a five point Likert scale. Information protection was examined based on the composite scores made from the 14 items examining information protecting behaviour, such as: I access my privacy settings. The answer categories range from I never do it to I do it very often on a five point Likert scale. The independent variable unpleasant experience was calculated based on the composite score of the 9 items describing the different kinds of unpleasant experiences that can happen while using social network sites, such as: Have you ever experienced disturbing commercials and to what extent did it influence you?. The answer categories range from no to yes, and I was very shocked on a four point Likert scale. Privacy invasion was assessed by the item: Has anybody in the last 6 months shared a photo or video of you on a social network site on which you were obviously recognisable and that you did not want to see online? The answer categories were yes and no. A composite score of extraversion was computed from the 12 items of the Dutch version of the NEO-Five Factor Inventory (NEO-FFI) (Costa & McCrae, 1992). The NEO-FFI consists of 60 items and is a shorter version of the NEO-RI-R measuring five personality dimensions, namely: neuroticism, extraversion, openness, agreeableness, and conscientiousness. In the current study we focused on the personality trait extraversion measured by three subscales: sociability, positive affectivity and activity. The internal consistency of this dimension is.79 and the test-retest reliability is.82. An example question is: I like having a lot of people around me. The answer categories range from strongly disagree to strongly agree. The variable number of online friends was assessed by the question: How many online friends do you have? Demographical details, such as age, gender, and education type were documented. Furthermore, the following questions were asked to acquire information about the social network site-using behaviour of the participants: On which social network sites do you have a profile? The following options were addressed: Facebook, MySpace, Hi5, Orkut, Friendster, Google+, Hyves, Badoo, other:... and I do not have any (the participants were provided with the possibility to select more than one option); On which social network site are you the most active? (the answer categories here were the same as in the previous question); In general how often do you use these social network sites? ; (more than twice a day, oncetwice a day, more than twice a week, once-twice a week, more than once-twice a month, once-twice a month, less than once a month); How many hours a week do you spend on these social network sites? (less than an hour, 1-5 hours, 5-10 hours, hours, more than 20 hours); Approximately many friends do you have on these social network sites? (0-1, 11-50, , , , , , more than 600); How often do you update your profile? (More than twice a day, once-twice a day, more than twice a week, once-twice a week, more than once-twice a month, once-twice a month, less than once a month); Have you ever read the privacy policy of these social network sites? (Yes, I have read it carefully; Yes, but not carefully; No, I know that it exists but I ve never read it; I didn t know about the existence of it); In case if you have read the privacy policy, did you understand it? (I understood the terminology; I understood it as a whole; I understood the details, but didn t understand it as a whole); In case if you did not read the privacy policy, why not? (I m not interested in it; I assume that my data will be handled confidentially; I couldn t find the privacy policy anywhere; I find the privacy policy too long and difficult to understand; other reason). 73

75 Statistical analyses Ordinal logistic regression was used to determine the role of age, gender, education type, extraversion, unpleasant experience and experience of privacy invasion by a video or photo in predicting the degree of privacy and risk concerns, and information protecting behaviour. Descriptive statistics were computed to analyse the demographical variables and to provide information regarding the students internet-using behaviour. Results Ordinal logistic regression (N= 474) analyses were carried out with privacy concerns, risk concerns and information protection as dependent variables, and gender, age, education type, extraversion, unpleasant experience, and privacy invasion as predictor variables. The first ordinal logistic regression analysis tested the presence of an effect of the six independent variables on the odds of one step increase on the privacy concerns scale. The model fit indices ( 2 LL = ; χ2 (7) = , p < 0.000) demonstrate that the model fits the data well, the explained variance by the model is 11% (Cox & Snell R2 =.103, Nagelkerke R2 =.112). Education type and privacy invasion have been confirmed to be significant predictors of privacy concerns (p <.05). More specifically, informatics students have been shown to significantly differ from law students in the sense that they have.814 times lower chances to be in a higher category on the privacy concern scale, while keeping all other variables constant. Criminology students have been found to differ from informatics students in the sense that they have.737 times lower chances to be in a higher category on the privacy concern scale, while keeping all other variables constant. Furthermore, moving to a higher category on the privacy concerns scale leads to a.556 increase on the privacy invasion scale while keeping all other variables constant. No significant contribution of Age, gender, extraversion and unpleasant experience to privacy concerns have been found. With 95% certainty we can say that when moving from the law student category to the informatics student category on the education type scale, one step decrease in odds on the privacy concerns scale is times higher than in one category below; and moving from the informatics student category to the criminology student category on the education type scale, one step decrease in odds on the privacy concerns scale is times higher than in one category below. With 95% certainty we can say that with an increase of one step on the privacy invasion scale, the odds of one step increase on the privacy concerns scale is times higher than in one category below (See Table 1). 74

76 Table 1: Parameter estimates of the effect of extraversion, unpleasant experience, privacy invasion, gender, age, and education type on privacy concerns Parameter B Std. Error 95% Confidence Interval 95% Confidence Interval for Exp(B) Lower Upper Sig. Exp(B) Lower Upper threshold 1-6,967 1,5415-9,988-3,946,000,001 4,594E-,019 5 threshold 2-4,319 1,5191-7,296-1,341,004,013,001,261 threshold 3-2,381 1,5109-5,343,580,115,092,005 1,786 threshold 4 -,419 1,5288-3,415 2,578,784,224,658,033 13,170 extraversion -,383,3152-1,001,235,224,682,367 1,264 unpleasant experience privacy invasion females males age law students informatics students criminology students -,285,556,208 0 a -,145 -,814 -,737,2347,2084,2021.,0399,1896, ,745,147 -,188. -,223-1,186-1,402.,175,964,604. -,067 -,442 -,071.,008,303.,000,000,030.,752 1,743 1,231 1,865,443,479 1,475 1,159,829.,800,306,246. 1,191 2,622 1,830.,936,642, a The second ordinal logistic regression analysis tested the presence of an effect of the six independent variables on the odds of one step increase on the risk concerns scale. The model fit indices ( 2 LL = ; χ2 (7) = , p < 0.000) show that the model fits the data well, the explained variance by the model is approximately 9% (Cox & Snell R2 =.087, Nagelkerke R2 =.097). Education type, unpleasant experience and privacy invasion (marginally, p=.053) have been confirmed to be significant predictors of risk concerns (p <.05). More specifically, informatics students have been found to significantly differ from law students in the sense that they have.698 times lower chances to be in a higher category on the risk concerns scale while keeping all other variables constant. However, no significant difference has been found between criminology and informatics students. Moving to a higher category on the risk concerns scale leads to a increase on the unpleasant experience scale while keeping all other variables constant. No significant contribution of Age, gender, privacy invasion, and extraversion to risk concerns has been found. With 95% certainty we can say that when moving from the law student category to the informatics student category on the education type scale, one step decrease in odds on the risk concerns scale is times higher than in one category below. With 95% certainty we can say that with an increase of one step on the unpleasant experience scale, the odds of one step increase on the risk concerns scale is times higher than in one category below (See Table 2). 75

77 Table 2: Parameter estimates of the effect of extraversion, unpleasant experience, privacy invasion, gender, age, and education type on risk concerns Parameter B Std. Error 95% Confidence Interval 95% Confidence Interval for Exp(B) Lower Upper Sig. Exp(B) Lower Upper threshold 1,582 1,4943-2,347 3,510,697,697 1,789,096 threshold 2 2,879 1,5003 -,061 5,820,055,055 17,798,940 threshold 3 5,292 1,5208 2,311 8,273,001, ,751 10,087 extraversion,178,3171 -,443,800,574,574 1,195,642 unpleasant experience 1,006,2411,534 1,479,000,000 2,735 1,705 privacy invasion -,403,2083 -,811,005,053,053,668,444 females,283,2007 -,110,677,158,158 1,328,896 males 0 a age -,002,0410 -,082,079,971,971,998,921 law students,698,1899,326 1,070,000,000 2,009 1,385 informatics students,583,3387 -,081 1,247,085,085 1,792,923 criminology students 0 a The third ordinal logistic regression analysis tested the presence of an effect of the six independent variables on the odds of one step increase on the information protection scale. The model fit indices ( 2 LL = ; χ2 (7) = , p = 0.002) indicate that the model fits the data well, the explained variance by the model is approximately 8% (Cox & Snell R2 =.074, Nagelkerke R2 =.092). Education type, age and extraversion have been confirmed to be significant predictors of information protection (p <.05). More specifically, informatics students have been shown to significantly differ from law students in the sense that they have.794 times higher chances to be in a higher category on the information protection scale, while keeping all other variables constant. However, no significant difference has been found between criminology and informatics students. Moving to a higher category on the information protection scale leads to a.128 decrease on the age scale, while keeping all other variables constant. Additionally, moving to a higher category on the information protection scale leads to a.955 increase on the extraversion scale, while keeping all other variables constant. No significant contribution of gender, privacy invasion, and unpleasant experience to information protection has been found. With 95% certainty we can say that when moving from the law student category to the informatics student category on the education type scale, one step decrease in odds on the information protection scale is times higher than in one category below. With 95% certainty we can say that with an increase of one step on the age scale, the odds of one step increase on the information protection scale is times higher than in one 76

78 category below. With 95% certainty we can say that with an increase of one step on the extraversion scale, the odds of one step increase on the information protection scale is times higher than in one category below (See Table 3). Table 3: Parameter estimates of the effect of extraversion, unpleasant experience, privacy invasion, gender, age, and education type on information protection Parameter B Std. Error 95% Confidence Interval 95% Confidence Interval for Exp(B) Lower Upper Sig. Exp(B) Lower Upper threshold 1-2,317 2,0980-6,429 1,795,269,099,002 6,020 threshold 2 1,847 2,0775-2,225 5,919,374 6,341, ,993 threshold 3 6,006 2,1612 1,770 10,242, ,992 5, ,252 extraversion,955,4425,087 1,822,031 2,598 1,091 6,184 unpleasant experience,434,3215 -,196 1,065,177 1,544,822 2,900 privacy invasion -,082,2740 -,619,455,765,921,539 1,576 females -,146,2870 -,708,417,612,864,493 1,517 males 0a Age -,128,0617 -,249 -,007,038,880,780,993 law students,794,2666,272 1,317,003 2,213 1,312 3,731 informatics students,471,4320 -,376 1,317,276 1,601,687 3,733 criminology students 0a Pearson correlation (N= 488) indicated a significant relationship between extraversion and the number of online friends, r =.214, p (two-tailed) <.01. Descriptive statistics revealed that 99% of the participants indicated to have a Facebook profile, 15.4% to have Google+ profile, 6.6% to have a Myspace profile, and 0.4% or less to have other profiles. 98.8% of the participants rated to be most active on Facebook,.8% on Google+,.2% on Myspace, and.2% on Orkut (See Table 4). 60.9% of the participants specified their social network site use as more than twice per day, 27% as once or twice per day, 6.5% as more than twice a week, and less than 3.8% as once or twice per week or less frequently. 51.2% of the participants spend between 1 and 5 hours per week with checking their social network profiles, 20.1% between 5 and 10 hours, 20.1% less than an hour, 5.9% between 10 and 20 hours and less than 2.8% more than 20 hours. 26.2% of the participants indicated their total number of friends to be between 201 and 300, 20.7% to be between 101 and 200, 20.1% to be between 301 and 400, 20.1% to be between 401 and 600, 5.5% to be more than 600, and less than 5% to be 100 or less. 22.1% of the participants updates their profile more than twice a week, 20.1% once or twice a week, 13.7% more than once or twice a month, 13.7% once or twice a month, 12.3% less than once a month, 12.3% once or twice a day, and 5.7% more than twice a day. 44.7% of the participants have read the privacy policy, but not carefully, 37.3% have not read the privacy policy but knew about the existence of it, 77

79 only 16.2% have read the privacy policy carefully, and 1.8% did not know about the existence of it. 64% of those who have read the privacy policy understood it as a whole, 31.3% understood the terminology and 4.7% understood the details but did not understand it as a whole. And finally, 42.4% of those who did not read it indicated that they were not interested in it, 29.4% indicated that they found the privacy policy too long and difficult to understand, 23.9% indicated that they assumed that their data will be handled confidentially, and 4.3% reported that they could not find the privacy policy anywhere (See Table 5). Table 4: Answers to Items Assessing the Preference for Social Network Sites in Percentage Yes Not selected Do you have a Facebook profile? 99% 1% Do you have a Myspace profile? 6.6% 93.4% Do you have a Hi5 profile?.4% 99.6% Do you have an Orkut profile?.2% 99.8 Do you have a Friendster profile? 0% 100% Do you have a Google+ profile? 15.4% 84.6% Do you have a Hyves profile? 1.2% 98.8% Do you have a Badoo profile?.6% 99.4% I don t have any social network site profile 0% 100% 78

80 Table 5: Answers to Items Assessing Social-Network-Site-Using Behaviour in Percentage Question Percentage On which social network site are you the most active? I am most active on Facebook 98.8% I am most active on Myspace.2% I am most active on Orkut.2% I am most active on Google+.8% How often do you use these social network sites? More than twice a day 60.9% Once or twice a day 27.0% More than twice a week 6.4% Once or twice a week 3.7% More than once or twice a month 1.0% Once or twice a month.4% In general how many hours do you spend on checking your profile per week? Less than an hour 20.1% 1-5 hours 51.2% 5-10 hours 20,1% hours 5.9% >20 hours 2.7% Approximately, how many friends do you have on these social network sites? % % % % % % % 79

81 > % In general, how often do you update your social network site profile? More than twice a day 5.7% Once or twice a day 12.3% More than twice a week 22.1% Once or twice a week 20.1% More than once or twice a month 13.7% Once or twice a month 13.7% <Once a month 12.3% Have you ever read the privacy policy of these social network sites? Yes. I have read it carefully 16.2% Yes. But not carefully 44.7% No. I know that it exists but I ve never read it 37.3% No. I didn t know about the existence of it 1.8% In case if you have read the privacy policy, did you understand it? I understood the terminology 31.3% I understood it as a whole 64.0% I understood the details but didn t understand it as a whole 4.7% In case if you have not read the privacy policy, why not? I m not interested in it 42.3% I assume that my data will be handled confidentially 23.9% I couldn t find the privacy policy anywhere 4.3% I find the privacy policy too long and difficult to understand 29.4% Discussion The aim of the current study was to investigate the degree to which university students are concerned about their privacy, the potential risks they take when using social network sites, and the extent to which they engage in information protecting behaviour as a function of their gender, age, education type, extraversion, unpleasant experience while using social network sites, and experience of privacy invasion. Furthermore, the relationship between the personality trait extraversion and online social network size and the social network site- using behaviour of the participants was also examined. The results show that the study program and 80

82 experience of privacy invasion in the form of a photo or video significantly contribute to the extent to which these students are concerned about their privacy. The findings also suggest that experience of privacy invasion plays the greatest role in this context. Moreover, the results imply that informatics students in the current sample are less concerned than law students, and criminology students are less concerned than informatics students about their privacy. These findings are in line with our expectations. The current order might be due to the fact that law and criminology students are aware of the legal issues with regard to privacy, but might lack the necessary informatics skills to protect their personal data. On the other hand, informatics students know more about the available options to protect their privacy; therefore, they are less concerned. Education type, unpleasant experience while using social network sites and experience of privacy invasion play a significant role in the degree to which these students are concerned about risk. The findings also suggest that an unpleasant experience while using social network sites in the past has the greatest influence on the degree to which students are concerned about potential risk. In terms of risk concerns difference has only been found between informatics and law students in the sense that informatics students seem to be less concerned about the risks than law students. Interestingly, informatics and criminology students seem to be less concerned about risks compared with law students, which might be due to the extensive lexical knowledge in combination with a potential lack of technical skills of law students. Having unpleasant experience with using social network sites appears to be another contributing factor to risk concerns, which is in line with our hypothesis and previous finding of Christofides et al. (2012). The authors concluded that adolescents with a history of unpleasant experiences when using the internet are more concerned about the potential risks thus; tend to restrict their profile to a greater extent. The results showed that education type, age, extraversion play a significant role in information protecting behaviour among the participating students and that extraversion has the largest impact on performing such behaviour. Informatics students have been found to significantly differ from law students but not from criminology students- in terms of their information protecting behaviour, in a way that they protect their personal data to a greater extent. An interesting finding is that information protection leads to a decreased level of privacy and risk concerns, and these behaviours are also education type-specific. Age has been shown to be a significant predictor of information protection in the manner that younger students have been found to take more actions in order to protect their privacy, which is in line with our assumptions and with the findings of the Pew Internet & American Life Project (Lenhart & Madden, 2007). In this survey, authors revealed a decrease of using the privacy settings as age increases. In spite of our expectations and contrary to the literature it seems to be that more extraverted participating students protect their data more than less extraverted people (e.g., Digman, 1990). Age and extraversion seem to have an influence only on information protection in our sample, which may be due to the fact that the sample consisted of university students from approximately the same age group and with similar levels of extraversion, given that most of the participants were law students, and in this field extraversion is an essential personality trait. Or it might be the case that even those students with a high level of extraversion are more aware of the danger of using such social network sites therefore, they pay more attention to protect their personal data than students following different study programs, such as art or architecture students. Gender, privacy invasion, and unpleasant experience have found to be non-significant with regard to the contribution to information protection. In the current study a significant correlation has been found between extraversion and the number of online friends. The findings concerning students social network site-using behaviour imply that the majority of the participants have a Facebook profile, are most active on Facebook, use social network sites more than twice per day, spend between 1 and 5 hours a week with checking their profiles, have their total number of friend to be between 201 and 300, and update their profile more than twice a week. Most of the participants have read the privacy policy but not carefully. The majority of the students who have read the privacy policy understood it as a whole. And finally, the lack of interest was mostly the reason for not reading the privacy policy, which is in line with the findings of Govani and Pashley (2005). 81

83 Conclusion It can be concluded that in the researched sample, experience of privacy invasion appeared to be the most influential factor in predicting the degree to which university students are concerned about their privacy. Furthermore, unpleasant experience when using social network sites seems to play the greatest role in predicting the level of risk concerns. Extraversion has been found to have the greatest impact on information protecting behaviour. Moreover, education type appears to have an impact on all three dependent variables, suggesting that education type plays an essential role in studying privacy, and needs to be included in future research. Additionally, the current study revealed a correlation between the personality trait extraversion and the number of online friends. Finally, the results imply that most of the participants did not read the privacy policy carefully, the majority of the students who have read the privacy policy reported that they understood it as a whole, and those who did not read the privacy policy indicated that they were not interested in it. Limitations and future research As the research was carried out among university students, the results cannot be extrapolated to a wider population. Remarks need to be mentioned with regard to gender differences and extraversion. In contrast to the majority of previous research, in the current study no gender differences have been found and extraversion seems to increase the chances for information protection. These findings might be biased due to homogeneity problems in the sample, given that the majority of the participants were either law or criminology students, which is an important limitation of this study. Furthermore, the variables privacy concerns and privacy invasion were measured by one item; therefore the findings concerning these two variables need to be taken with caution. Moreover, the contribution of the variables was rather low (11 %, 9%, and 8%), which implies that other skills, such as computer skills or personality-related variables, such as conscientiousness and self-esteem might also have an impact on the degree to which students are concerned about their privacy and risk, and on the extent to which they engage in information protecting behaviour. Despite the limitations we believe that our research contributes to the scientific development of a comprehensive framework for privacy protection by strengthening the findings of previous research and by giving prominence to influential factors that have not been involved in privacy research in the past. Further studies are required to investigate the relationship among these variables to raise people s awareness of privacy, risks, and information protection. Finally, the identified influential factors, such as experience of privacy invasion and education type should also be taken into account in future research. References Acohido, B. (2011). Facebook privacy: Social network tracks users after they leave site. USA Today. Boyd, D. and Ellison, N. (2007). Social network sites: definition, history, and scholarship. Journal of Computer- Mediated Communication, 13, 1, Budden, C., Anthony, J., Budden, M., & Jones, M. (2007). Managing the evolution of a revolution: marketing implications of Internet media usage among college students. College Teaching Methods & Styles Journal, 3, 1, Bryer, T. A., and Chen, B. (2010). Using social networks in teaching public administration. In C. Wankel, (Ed.), Cutting-edge social media approaches to business education: Teaching with LinkedIn, Facebook, Twitter, Second Life, and Blogs (pp ). Charlotte, NC: Information Age Publishing. Byrnes, J. P., Miller, D. C., and Schafer, W. D. (1999). Gender differences in risk taking: A meta-analysis. Psychological Bulletin, 125, 3,

84 Chahal Shirin (2011). Balancing the scales of justice: undercover investigations on social networking sites. The Journal on Telecommunications and High Technology Law, 9, 1, Christofides, E., Muise, A., and Desmarais, S. (2009). Information disclosure and control on Facebook: Are they two sides of the same coin or two different processes? CyberPsychology & Behaviour, 12, 3, Christofides, E., Muise, A. and Desmarais, S. (2012) Risky Disclosures on Facebook: The Effect of Having a Bad Experience on Online Behaviour." Journal of Adolescent Research, 27, 6, Clark, L. A., and Roberts, S. J. (2010). Employer s use of social networking sites: A socially irresponsible practice. Journal of Business Ethics, 95, 4, ComScore (2011). It's a social world. Top 10 need-to-knows about social networking and where it s headed. Costa. P. T. Jr. and McCrae. R. R. (1992). Revised NEO Personality Inventory (NEO-PI-R) and NEO Fiva Factor Inventory (NEO-FFI) professional manual, Odessa. FL: Psychological Assessment Resources. Digman, J. M. (1990). Personality structure: Emergence of the five-factor model. Annual review of psychology, 41, Fogel, J. and Nehmad, E., Internet social network communities: Risk taking, trust, and privacy concerns. Computers in Human Behaviour, 25, 1, Golder, S., Wilkinson, D., & Huberman, B. (2007). Rhythms of social interaction: messaging within a massive online network. Communities and Technologies, Govani, T., and Pashley, H. (2005). Student Awareness of the Privacy Implications while Using Facebook. Unpublished manuscript. Lenhart, A. and Madden, M., (2007). Teens, Privacy and Online Social Networks. Longitudinal Internet Studies for the Social Sciences (LISS) (2006). Online panel survey, administered by CentERdata. Madden, M. and Smith, A. (2010). Reputation management and social media, Pew Internet & American Life Project. Moscardelli, D.M. and Divine, R., (2007). Adolescents' Concern for Privacy When Using the Internet: An Empirical Analysis of Predictors and Relationships with Privacy-Protecting Behaviours. Family and Consumer Sciences Research Journal, 35, 3, Moscardelli, D. M. and Liston-Heyes, C. (2004) Teens Surfing The Net: How Do They Learn To Protect Their Privacy? Journal of Business and Economics Research, 2, 9, Murphy, D. (2012). Google sued over safari privacy problem. PC Magazine. Newk-Fon Hey Tow, W., Dell, P., and Venable, J. R. (2010), Understanding Information Disclosure Behaviour in Australian Facebook Users. Journal of Information Technology, 25,

85 Ong, E. Y. L., Ang, R. P., Ho, J. C. M., Lim, J. C. Y., Goh, D. H., and Lee, C. S., et al. (2011). Narcissism, extraversion and adolescents self-presentation on Facebook. Personality and Individual Differences, 50, 2, Petronio, S. (2002). Boundaries of privacy: Dialectics of disclosure. Albany, NY: State University of New York Press. Roblyer, M. D., McDaniel, M., Webb, M., Herman, J., and Witty, J. V. (2010) Findings on Facebook in higher education: a comparison of college faculty and student uses and perceptions of social networking sites. The Internet and Higher Education, 13, 3, Ross, C., Orr, E., Mia Sisic, B., Arseneault, J., Simmering, M., & Orr, R. (2009). Personality and motivations associated with Facebook use. Computers in Human Behaviour, 25, 1, Skeel, R. L., Neudecker, J., Pilarski, C., and Pytlak, K. (2007). The utility of personality variables and behaviourally-based measures in the prediction of risk-taking behaviour. Personality and Individual Differences, 43, 1, Smith, S. D., & Caruso, J. B. (2010). ECAR study of undergraduate students and information technology. Boulder, CO: EDUCAUSE Center for Applied Research. 84

86 From child pornography offending to child sexual abuse: A review of child pornography offender characteristics and risks for cross-over Jenny A.B.M. Houtepen (1), Jelle J. Sijtsema (2), & Stefan Bogaerts (3) 1. Department of Developmental Psychology, Tilburg University, the Netherlands; [email protected] 2. Department of Developmental Psychology, Tilburg University, the Netherlands; [email protected] 3. Department of Development Psychology, Tilburg University, the Netherlands; [email protected] *Correspondence concerning this article should be addressed to J.A.B.M. Houtepen, Tilburg University, School of Social and Behavioural Sciences, Developmental Psychology, P.O. Box 90153, 5000 LE Tilburg, the Netherlands. Phone: ; [email protected] Highlights We examine risk factors for child pornography offending and cross-over behaviour. We draw attention to current gaps and inconsistencies in this research area. Heterogeneity among child pornography offenders is highlighted. Specifically, we propose a new way of mapping offender specific risk factors. Future research should focus on risk factor integration and pedophilia in general. Keywords: child pornography offending, offender characteristics, offender typology, antisocial behaviour, pedophilia, risk for cross-over 1. Introduction Child pornography, which includes visual depictions of sexual conduct involving a minor, has been a problem for many decades. However, the advent of the internet has made it a bigger and more widespread societal problem (Webb, Craissati, & Keen, 2007). Despite its high relevance, it still remains unclear what factors are associated with engaging in child pornography offending or predispose one to cross-over to committing child sexual abuse. For example, for some offenders, child pornography may help controlling their deviant sexual interests in children and prevents them from committing sexual hands-on offenses. However, for others viewing this material stimulates existing fantasies and lowers inhibitions to act on them which may result in cross-over behaviour (Quayle & Taylor, 2002). Furthermore, research on the characteristics of child pornography offenders is scarce, inconsistent, and still in development. Only recently, researchers have begun to develop typologies that characterize different child pornography offending motivations and behaviours (e.g., Krone, 2004; Lanning, 2010; Seto, Cantor, & Blanchard, 2006; Seto, Wood, Babchishin, & Flynn, 2012). Moreover, despite efforts to explain child pornography offending within more general frameworks of sex offending (Elliott & Beech, 2009; Middleton, Elliott, Mandeville-Norden, & Beech, 2006), there is only one model to date that provides an explanation for cross-over behaviour directly related to internet offending: the model of potential problematic Internet use (Quayle & Taylor, 2003). This model, based on the etiology of Pathological Internet Use (Davis, 2001), states that cognitive distortions not only promote problematic behaviour, such as downloading of child pornography, but also partly cause the progression of this behaviour into sexual contact offending. According to this model, other factors that contribute to the process of engagement in the internet and problematic internet behaviour include individual risk factors such as early sexual experience and sexual preference for children (i.e., setting 85

87 events ), internet characteristics, such as anonymity, and cognitive-social factors such as less contact with people in the offline world. However, despite a growing interest in child pornography offending, an overview of the state of knowledge and a clear focus for future research is lacking. This study aims to address these limitations, by providing a selective overview of the contemporary literature on child pornography offending. Specifically, current gaps in the literature are identified and promising directions for future research are suggested. We will discuss four domains of child pornography offender characteristics, including demographics, psychological problems related to socio-affective issues, distorted cognitions, and psychosexual difficulties. Furthermore, we highlight the heterogeneity among child pornography offenders and the importance of focusing on offender specific risk factors in order to enhance risk assessment and treatment planning. It is proposed that offender specific risk factors may be mapped onto two continua: (1) features related to criminal behaviour, and (2) sexual deviance/fantasy. Moreover, we discuss risk factors for cross-over, including individual characteristics, factors provided through engagement in the internet, and through contact with the offline environment. Finally, we provide concrete directions for future research. In order to better understand child pornography offending, we will first touch upon the issues related to legally defining child pornography and the diverse working definitions in practice and research. 2. Legal definition and research problems It is complicated to provide a global legal definition of child pornography because views about children and child pornography are bound to moral, cultural, social, and religious beliefs. Hence, definitions differ across countries and even among legal jurisdictions within the same country (Healy, 1996). This has consequences for law enforcement because different legal definitions make it difficult to combat child pornography offending internationally (Burke, Sowerbutts, Blundell, & Sherry, 2002). United States federal law, for example, defines child pornography as any visual depiction, including any photograph, film, video, picture, or computer-generated image or picture, that is made or produced by electronic, mechanical, or other means, of sexually explicit conduct that involves or seemingly involves a minor (18 U.S.C. 2256(8), 2006). However, whether or not clear definitions of child pornography are at hand, it can be difficult for law enforcement to make decisions on the basis of computer data and images to determine which individuals should be prosecuted. For example, it may be hard to estimate the age of the victims depicted on the images in order to determine if a minor is involved (Wells, Finkelhor, Wolak, & Mitchell, 2007). What makes research on child pornography even more complex is that definitions of child pornography may differ between legal and academic contexts (Beech, Elliott, Birgden, & Findlater, 2008). For example, although images depicting clothed children may be considered legal by law, they may serve as child pornography for some individuals because the erotic nature of the images can not only be viewed in light of the objective qualities of the material itself, but is also determined through the person that is watching. Hence, such images may be used to relieve sexual arousal by someone with pedophilic interests (Taylor, Holland, & Quayle, 2001). Furthermore, views about what constitutes child pornography and child pornography offending may differ across studies because inclusion criteria for child pornography offending vary greatly. That is, some studies include those who have sexually abused a child in the past (e.g., Bates & Metcalf, 2007; Laulik, Allam & Sheridan, 2007; Niveau, 2010; Seto, Cantor & Blanchard, 2006), whereas other studies focus on community samples that include individuals who have admitted that they have engaged in child pornography, but who did not come into contact with the legal justice system (Seigfried, Lovely, & Rogers, 2008). Moreover, insight into the type of material is often not specified; yet, this may give insight into offender specific risk factors because the type of material that offenders use varies greatly. For example, pictures range from depicting relatively innocent nudity through to pictures of children being sexually abused (Taylor et al., 2001). Furthermore, pseudo imagery exists in which non-sexual photographs are digitally adjusted into child pornographic images (a phenomenon that is called morphing) next to virtual images of 86

88 digitally created children, films, stories, and even live child pornography via webcams (Krone, 2004; van Wijk, Nieuwenhuis, & Smeltink, 2009). As a consequence, different definitions of child pornography and different research samples have led to problems with the generalization of the findings and the comparability of results across studies. This should be kept in mind while reading this article. In this study, the term child pornography offender refers to those who accessed and/or distributed and/or produced child pornographic material without having committed a (known) child sexual hands-on offense. When studies are described that also included child pornography offenders that have committed a contact offense, this will be clearly specified. 3. Offender characteristics 3.1. Demographics There is some consensus about the demographic characteristics of child pornography offenders. For example, studies report offenders to be primarily white, aged between 25 and 50 years, and compared to child sexual abusers, more likely to be employed (Burke et al., 2002). With regard to intellectual functioning a considerable amount of child pornography offenders has above average intelligence; in several studies about 30% had completed higher education (McCarthy, 2010; Seto et al., 2006; Surjadi, Bullens, van Horn, & Bogaerts, 2010). With regard to employment, law enforcement and mental health professionals argue that child pornography offenders often have jobs in sectors with little or no social interaction or where contact with children is a daily activity (van Wijk et al., 2009). However, a typical profile of child pornography offenders is missing. Research on marital status shows mixed results with studies showing that more than 50% of child pornography offenders were single, either at the time of the offense or at the moment of the conducted study (e.g., Henry, Mandeville-Norden, Hayes, & Egan, 2010; Neutze, Seto, Schaefer, Mundt, & Beier, 2011; Reijnen, Bulten, & Nijman, 2009). Some professionals argue that these offenders are often characterized by marital problems and inequality between romantic partners (van Wijk et al., 2009). In short, although the literature identified several demographic risk factors of child pornography offending, studies often do not include control groups of the general population. Hence, it is unclear to what extent these child pornography offenders differ from the normal population with regard to demographic characteristics. As research on criminal history indicates, child pornography offending occurs in all social groups with much heterogeneity in offenders (Wortley & Smallbone, 2006). Although it is often assumed that child pornography offenders have no history of offenses (Burke et al., 2002; Sullivan, 2009) and a low risk to commit future offenses (Seto, Hanson, & Babchishin, 2011), empirically little is known of child pornography offenders criminal history. Research shows that before the emergence of the internet, between one-fifth and one-third of the offenders arrested for possession of child pornography were also involved in sexual contact abuse. Because the internet has made accessing child pornography much easier (Wortley & Smallbone, 2006), the number of noncontact child pornography offenders may well have increased since. In self-referred pedophilic and hebephilic males (i.e., males who have admitted to have a paraphilic sexual interest in either prepubescent or pubescent children), 50% claimed to have committed child pornography offenses in the past, only 14.3% acknowledged to have sexually abused a child in the past, and 35.7% reported a history of both child pornography offending and child sexual abuse (Neutze et al., 2011). Yet, whether these numbers reflect the true criminal history of these offenders remains unclear because most data depend on police records or selfreports which may have limited reliability and often does not provide information about undetected offenders (i.e., the dark number) (Bogaerts, Buschman, Kunst & Winkel, 2010; Seto et al., 2011). An understudied area of child pornography offenders are early life risk factors and experienced difficulties in the past. However, a few notable exceptions showed that child pornography offenders had a history of child sexual abuse, with estimates ranging from 11 to 26% (McCarthy, 2010; Webb et al., 2007). In 87

89 general, both child pornography offenders and child sexual abusers are more often victims of sexual abuse in childhood than people in the normal population (Babchishin, Hanson, & Hermann, 2011). However, child pornography offenders suffer less from sexual victimization than child sexual abusers (Sheldon & Howitt, 2007; Webb et al., 2007). Finally, there is an increasing focus on child and adolescent sexual behaviour in order to understand the etiology of child pornography offending. Early sexual activity seems to be common in both internet and contact sex offenders (Sheldon & Howitt, 2007), and may be a risk factor for pedophilia in general. In a study on pedophiles, many of those without a criminal history claimed to be sexually active as a child, often from a very young age. Sixteen of the 36 individuals reported that they had engaged in peer sexual play before the age of 10. These activities did not only include childish games of exploratory sexuality such as playing doctor or discovering masturbation, but some claimed to be engaged in oral sex and/or penetration (Goode, 2010). Yet, these results should be treated with caution as there is little knowledge of normal levels of childhood sexual play in the general population (Elliott & Beech, 2009) Socio-affective difficulties Studies on socio-affective difficulties in child pornography offenders mostly center on psychological differences between child pornography offenders and child sexual abusers. There is consensus that although some child pornography offenders do not demonstrate high levels of psychopathology (Henry et al., 2010; Laulik et al., 2007; both included some contact offenders), many have socio-affective deficits. More specifically, mood disorders, such as depression, and anxiety-related problems (McCarthy, 2010), such as obsessive compulsive disorder (Bourke & Hernandez, 2009), are commonly reported. Child pornography offenders may also suffer from low self-esteem, emotional loneliness, and personal distress (Henry et al., 2010). Although child pornography offenders are more often characterized by personality traits typically related to cluster C personality disorders than to cluster A and B personality disorders (Niveau, 2010; included two contact offenders), some offenders also show traits related to cluster A disorders, such as feeling misunderstood by others and being isolated and somewhat withdrawn (Laulik et al., 2007). Interpersonal difficulties include under-assertiveness (Henry et al., 2010), low dominance and warmth, and little empathy in social relationships. Finally, there is support for antisocial related tendencies in a subgroup of child pornography offenders. These include personality traits related to exploitation and manipulation, low moral choice internal values (Seigfried et al., 2008), addictive tendencies (Niveau, 2010), impulsivity (Middleton et al., 2006), and psychopathic deviation (Reijnen et al., 2009). However, there is debate about the severity of psychological problems of child pornography offenders in comparison to those of child sexual abusers. On the one hand, several studies have suggested that child pornography offenders have fewer socio-emotional problems than child sexual abusers because child pornography offenders seems to have higher self-esteem, less personal distress, a lower external locus of control, and less features related to criminal and antisocial behaviour (Bates & Metcalf, 2007; Elliott, Beech, Mandeville-Norden, & Hayes, 2009; Magaletta, Faust, Bickart, & McLearen, 2012). On the other hand, child pornography offenders seem to have more problematic scores on impression management, underassertiveness and emotional-loneliness compared to child sexual abusers (Bates & Metcalf, 2007; Elliott et al., 2009), suggesting more troublesome interpersonal functioning. Yet, other studies found no differences on emotional deficits (including intimacy and loneliness problems) and sociality between child pornography offenders and child sexual abusers. Also, these groups did not differ with regard to socio-emotional problems such as impulsivity, thrill-seeking, extraversion, and frustration-tolerance (Neutze et al., 2011; Reijnen et al., 2009) Cognitive distortions Cognitive distortions are highly relevant in understanding sex offending. With regard to child pornography offending, it is argued that some of the cognitive distortions that offenders possess are more 88

90 related to sex offending in general, than specifically to engaging in child pornography (Howitt & Sheldon, 2007; Neutze et al., 2011). One study compared cognitive distortions displayed by child sexual abusers without a history of child pornography offenses, child pornography offenders without sexual contact offenses, and mixed contact-child pornography offenders (Howitt & Sheldon, 2007). Results indicated few differences in cognitive distortions between these groups. Many offenders agreed with cognitions about the perceived uncontrollability of the offense behaviour, cognitions that portrayed the sexual abuse of children as reasonable under certain circumstances, and cognitions that stated that the world is a hostile place. Yet, there was a significant difference between the child pornography offending group and the child sexual abusers group in the evaluation of cognitions concerning children as sexual subjects. These cognitions included statements as: Sometimes children don t say no to sexual activity with an adult because they are curious about sex or enjoy it. Results showed that child pornography offenders supported more of these items than child sexual abusers did (Howitt & Sheldon, 2007). In contrast, Bates and Metcalf (2007) found that child sexual abusers agreed more with this type of cognitions. An explanation for these conflicting results may be the use of different instruments to measure cognitive distortions in these two studies. Particularly, Howitt and Sheldon (2007) developed measures specifically for internet sex-offenders which may have been more sensitive in studying child pornography offenders (Henry et al., 2010). In line with their findings, child sexual abusers may have learned through their experience with contact abuse that these statements about the sexuality of children are not true (Elliott & Beech, 2009). Child pornography offenders instead may justify their behaviour by the fact that these images depict laughing children who look like they are willing and eager to engage in sexual activities (Quayle & Taylor, 2002). Indeed, Quayle and Taylor (2002) showed that child pornography offenders have cognitive distortions that are directly related to the content of the images. They argued that these offenders often use some sort of moral code to decide which images are acceptable to use for their sexual gratification. These are based on victim characteristics as age and sex of the victim, but also on the sexual activity depicted. Furthermore, they may engage in justification of their offense behaviour by comparing the seriousness of downloading child pornography with the actual act of abusing a child and argue that the latter is not something that they would ever engage in. Also, many child pornography offenders minimized their behaviour by engaging in depersonalization of the pictures, e.g., seeing them as objects that could be collected and traded, like base-ball cards (Quayle & Taylor, 2002). However, it should be noted that 4 out of the 13 child pornography offenders in this sample had also committed child sexual abuse. Therefore, these distortions are not only exclusive for noncontact child pornography offenders. Finally, compared to child sexual abusers, child pornography offenders agreed less with statements measuring emotional congruence with children, had higher victim empathy (Bates & Metcalf, 2007; Elliott et al., 2009), and had healthier attitudes toward general sexual assault (Webb et al., 2007). Although research has provided insight into the type of cognitive distortions child pornography offenders may display, it should be kept in mind that these distortions are not stable characteristics but susceptible to change. For example, Quayle and Taylor (2003) showed that cognitive distortions can change due to contact with other offenders that may justify and reinforce sexually deviant behaviour. More specifically, child pornography offenders may display two types of cognitions: offense-level cognitions that are related to the appropriateness and consequences of viewing child pornography images, and sexual abuse-level cognitions that are related to the appropriateness and consequences of having sex with children. Although child pornography offenders are more likely to make use of the first set of cognitions, they may develop sexual abuse-level distorted thinking by repeated engagement in child pornography (Elliott & Beech, 2009). Finally, because of the ongoing debate on cognitive distortions, different views about the onsets and functions of these distortions should be considered (Howitt & Sheldon, 2007). First, cognitive distortions are a set of beliefs that offenders generate and use in order to overcome inhibitions against offending and are part of the preparation process. Second, cognitive distortions are rationalizations that are used by offenders 89

91 to justify their behaviour after committing an offense. Third, cognitive distortions develop prior to the offense and reflect the offenders distorted experiences and hence contribute to the initiation of behaviour. Likely, in explaining cognitive processes all these views may be important, assuming that the term cognitive distortions reflects different mechanisms and functions. Offenders may thus not only differ in the content and the development of their distortions but also in the onset of these cognitions and the functions these distortions have at a given time. For example, it may be argued that child sexual abusers more often develop cognitive distortions early in life due to traumatizing experiences such as being sexually abused. In contrast, child pornography offenders who accesses child pornography accidently while searching for regular pornography may develop these distortions post-offending to justify and minimize the fact of watching children being abused. Later on, offenders may use other excuses to overcome inhibitions to engage in this behaviour again, or worse: cross-over to contact offenses. Hence, getting more insight into the function and development of these cognitions in offenders may give more insight into the initiation of child pornography offending and risks for cross-over Psychosexual difficulties Regarding use of pornographic material, people look at first glance for material that corresponds to their sexual preference (Seto, 2010). The specificity of pornographic material seems to be a stronger diagnostic indicator of sexual preference than actual sexual contact. For example, child pornography offenders have a stronger pedophilic pattern of sexual arousal than child sexual abusers (Seto et al., 2006; including some child pornography offenders with a history of sexual offending). This is also supported by a meta-analysis indicating that compared to child sexual abusers, child pornography offenders possessed more sexual deviance (Babchishin et al., 2011). However, not all child pornography offenders are pedophiles. In order to discriminate between child pornography offenders with and without pedophilia it is important to include characteristics of the child pornographic material such as age of the victims depicted on the images. That is, it is expected that child pornography offenders without a paraphilic preference for minors are attracted to images of teenage boys and/or girls in a way that non-pedophilic men may commit sexual offenses against girls who did not yet reached adulthood, but not show a preference for images that depict very young children (Seto et al., 2006). To better understand the relationship between child pornography offending and pedophilia, it is important to examine how many pedophiles engage in child pornography offending. One study with selfidentified pedophiles showed that a small minority had no history of abusing a child or watching child pornography (Neutze et al., 2011). However, more insight into pedophilic child pornography offenders and non-offending pedophiles is needed in order to determine risk- and protective factors specifically for pedophilic individuals. Research on pedophiles outside a clinical or juridical setting is scarce but one study indicated that pedophiles who do not want to offend may actually be troubled by their desire to have sexual contact with children and are often scared of the inability to control their arousal (Hossack, Playle, Spencer, & Carey, 2004). In order to refrain from offending, they may seek help if they are able to overcome their fear of the possible consequences of disclosure. It should thus be kept in mind that there are no direct relationships between pedophilia and child pornography offending and that stigmatization and self-regulatory issues are important risk factors for pedophilic offending. Moreover, inhibitory control to refrain from offending may be more important for child pornography downloaders than for child sexual abusers because it is relatively easy to access child pornography online and, therefore, it may be harder to resist. Some researchers have touched upon this issue in comparing child pornography offenders with child sexual abusers on self-regulation. Although Neutze et al. (2011) found no differences between these groups, Webb, Craissati, and Keen (2007) found that child pornography offenders had more sexual self-regulation problems than child sexual abusers. They argued that a part of sexual self-regulation is sexual preoccupation, which is reflected in routine pornography use. Regarding sexual preoccupation, more than half of child pornography offenders had an indication for 90

92 compulsive sexual behaviour (Niveau, 2010). Therefore, compulsive or addictive sexual problems seem to be part of the psychosexual make-up of child pornography offenders. 4. Child pornography offender typology In this section, to highlight the heterogeneity in child pornography offender and their different motivations, different typologies that have been identified in literature will be described and integrated. Starting with the classification of Elliot and Beech (2009), based on work done by other researchers (e.g., Krone, 2004), it is argued that earlier typologies can roughly be divided into four groups of child pornography offenders. Two of these groups, the direct victimization and commercial exploitations offenders, include child sexual abusers and will, therefore, not be discussed here. The other two groups comprise: (1) periodically prurient offenders who access child pornographic material impulsively or out of curiosity as part of a larger interest in pornography; and (2) fantasy-only offenders who access and or trade material in order to find gratification for their sexual interest in children but do not have a known history of child sexual abuse. In addition to Elliott and Beech (2009), who described the major problem areas of these groups, we argue that risk factors of these offender types can broadly be described along the lines of two continua: (1) their features related to criminal behaviour (e.g., self-control issues, impulsivity, and cognitive distortions), and (2) their sexual deviance/fantasy. This can give insight into the different underlying motivations of these two offender types and offers an explanation of the variation in severity of psychological problems between child pornography offenders of the same offender type. As such, mapping individuals on these continua may result in better risk assessment and more insight into specific treatment needs for child pornography offenders. In the following, we will support the validity of these continua with empirical work on child pornography offender clustering (Henry et al., 2010; Middleton et al., 2006; Surjadi et al., 2010). However, more in-depth research is needed to get more insight into the developmental pathways of different offending types Periodically prurient offenders In periodically prurient offenders, socio-affective deficits (e.g., low self-regulation and inhibitory control) may be the primary cause of accessing material beyond the scope of legal pornography. However, as noted at the beginning of section 4, offenders within a particular offender type may differ significantly in the severity of their psychological problems. Henry et al. (2010) found support for three groups of child pornography offenders, i.e., a normal, an emotionally inadequate and a deviant offender group. The normal group did not differ from the normal population with regard to pro-offending and/or socio-affective features. Also, the total impulsivity scores measured with the Barratt Impulsiveness Scale (BIS-11) (Patton, Stanford, & Barratt, 1995) were lower than in the emotionally inadequate and deviant offender groups. However, scores on the subscale of non-planning impulsivity, which assesses self-control, were similarly high across offender groups. Further, normal offenders scored more problematic on self-control than on cognitive and motor impulsivity. Therefore, it seems that these offenders have some difficulties with appropriately controlling behaviour and thus fall within the periodically prurient offender type, although they do not have many impulsivity problems. Research also indicates that some periodically prurient offenders have severe interpersonal problems. This is reflected in their more severe impulsivity issues, interpersonal difficulties and personal distress compared to both the normal population and other sexual offenders (Henry et al., 2010; Middleton et al., 2006). Middleton et al. (2006) argued that these offenders fell within the Emotional dysregulation pathway of the Pathways Model of Sexual Offending (Ward & Siegert, 2002): a model consisting of pathways describing engagement in child sexual abuse. The Emotional Dysregulation pathway describes offenders who mainly have problems related to successfully regulating their negative emotions and use sex as a coping strategy (Ward & Siegert, 2002), which is partly the result of antisocial personality structures that are often present in these offenders (Bogaerts, Buschman, Kunst, & Winkel, 2010). They may claim that they are primarily drawn to child pornography out of non-sexual motivations such as to avoid real life confrontations or to gain enjoyment from collecting child pornographic material (Surjadi et al., 2010). With regard to cognitive 91

93 distortions, emotionally deregulated child pornography offenders do not report many distortions about children and sex or emotional congruence with children but some may lack empathy for victims of sexual abuse (Middleton et al., 2006). Differences in the severity of psychological problems between these periodically prurient offenders indicate that even within this particular offending type, offenders may differ in their need for treatment. For example, according to the risk-need-responsivity (RNR) model (Andrews & Bonta, 2007), interventions should fit offenders learning styles, be focused on dynamic risk factors and be linked to the risk of re-offending. The first subgroup of periodically prurient offenders does not have many psychological difficulties and therefore probably does not need intensive therapy in order to refrain from engaging in child pornography offending in the future. In fact, according to the RNR model, placing these individuals in intensive treatment programs when this is not indicated can even be counterproductive. In contrast, the second periodically prurient group has multiple psychological issues related to criminal behaviour and, therefore, seems at higher risk for reoffending. These child pornography offenders may need intensive therapy, such as cognitive behavioural therapy, to enhance impulse control and to provide them with better coping skills in order to deal with their negative emotions in more appropriate ways. In turn, this may lower their risk for reoffending Fantasy-only offenders Pedophilic interest (Elliott & Beech, 2009) is the main motivation for fantasy-only offenders to engage in child pornography offending. Surjadi et al. (2010) showed empirical support for the existence of such a group. They found that offenders admitted that their primary motivation to access child pornography was their fixated sexual interest in children. This was also supported by the fact that this group was more likely than others to have started masturbating immediately during their first time accessing this material. However, pedophilic orientation alone does not explain why some engage in deviant sexual behaviour. Although it is clear that the primary motivation to engage in child pornography for these offenders is to find an outlet for their sexual interest in children, it does not explain why they break the law and put their own feelings before those of the children depicted on the images. Therefore, fantasy-only offenders are likely to also have some features related to criminal behaviour. One important risk factor for sex offending is adverse experiences of intimate relationships in early life. These can lead to intimacy problems and loneliness and have been found in pedophiles (Sawle & Kear-Colwell, 2001). Ward and Siegert (2002) termed this the Intimacy deficits pathway that describes individuals who experience problems in establishing healthy relationships with adults due to insecure attachment (see also Bogaerts, Vanheule, & Desmet, 2006; Marshall, 1993). These individuals may express distorted cognitions about sex with children, have low social skills and low self-esteem, and are more attracted to children as sexual partners in times of loneliness or dissatisfaction with more appropriate adult relationships. In child pornography offender research, Middleton et al. (2006) found that the Intimacy deficits pathway was the most common in which 35% of the studied child pornography offenders fitted. These offenders displayed high levels of emotional loneliness and low levels of self-esteem. Unfortunately, it is impossible to state that this group also had a sexual preference for children because their sexual orientation and fixation was not measured. Furthermore, they did not report distorted cognitions with regard to views about sex with children. In treatment, these offenders would probably benefit from cognitive therapy that gives insight into their problem behaviour, together with a more direct approach of enhancing their circles of support. In line with what we argued in section 4, it is likely that the number of psychological difficulties differs across groups of pedophiles and that some offenders may experience issues in multiple areas of functioning. For example, the deviant child pornography offender group of Henry et al. (2010) included a subgroup who had issues primary related to the Intimacy deficits pathway and also had features related to non-sexual offending. Compared to the other groups these offenders scored high on all risk factors for offending, including emotional congruence, cognitive distortions concerning children and sex, and distorted victim 92

94 empathy. Hence, it is likely that these offenders are at high risk for re-offending and are in need of more intensive treatment to address all relevant problem areas. Finally, it should be noted that the Intimacy deficits pathway is probably not the single pathway to child pornography offending the fantasy-only group. However, this pathway does shed light on the importance of social behaviour and (early) social relationships and suggest that social isolation and exclusion are important risk factors for offending. We will elaborate on this in the next section, where we discuss crossover risks (i.e., the shift to sexual child abuse). 5. Risk factors for cross-over In this last section, we distinguish between three domains of risk factors for cross-over: individual characteristics, engagement with the internet, and the offline environment. These factors partly overlap with those identified by Quayle and Taylor (2003) in their earlier described model of potential problematic internet use. However, to include all relevant risk factors, broader definitions of these domains are necessary Individual characteristics The most evident risk factors for cross-over are likely those psychological characteristics known to be related to antisocial behaviour in general, such as low victim empathy and cognitive distortions, because these lower behavioural inhibitions to act on impulses (Babchishin et al., 2011). Long, Allison, and McManus (2013) found that child pornography offenders who committed sexual hands-on offenses were more likely to be convicted for prior criminal behaviour, specifically for nonsexual offenses like theft. These antisocial characteristics are likely risks for cross-over in periodically prurient offenders and not in fantasy-only offenders who often do not possess these characteristics. Furthermore, low inhibitory control is exactly what puts periodically prurient offenders at risk for engaging in child pornography in the first place. In turn, more selfcontrol in fantasy-only offenders may be a protective factor against cross-over. The presence of a fantasyonly group could thus also explain findings of low rates of child sexual abuse among child pornography offenders despite the high rates of sexual deviance (Babchishin et al., 2011). In contrast, fantasy-only offenders may be more at risk for cross-over because of feelings of loneliness and low self-esteem that may be remedied by factors provided by the online environment. For example, online contact with like-minded others may encourage committing child sexual abuse. We will discuss this in more detail when describing risk factors provided by engagement with the internet. Finally, apart from the sexual interest, the willingness to engage in behaviours that may cause direct harm to children is a prerequisite for crossing over to contact offenses. This is illustrated by the cognitive distortions that child pornography offenders use in order to minimize the moral impact child pornographic images have on victims (e.g., masturbation to child pornography is a substitute for abuse ) (Quayle & Taylor, 2002). Some offenders have more difficulties with watching children being hurt and victimized than child sexual abusers. Some pedophiles are interested in having romantic relationships with children, and not only sexual contact (Seto, 2012), which indicates that besides sexual attraction, feelings of intimacy and love for children are also present in a subgroup of pedophilic fantasy-only offenders. In these offenders, risk of committing a contact sexual offense where they are personally responsible for the abuse seems less likely Engagement with the internet There is a positive association between time spent online and number of child pornography images that child pornography offenders have collected (Quayle & Taylor, 2003). In turn, this may be associated with cross-over behaviour because child pornography offenders who also committed child sexual abuse had larger collections of child pornographic material than child pornography offenders without a history of sexual abuse (McCarthy, 2010). However, according to Long et al. (2013), not the number of images per se is critical in discriminating child sexual abusers from child pornography offenders, but rather the type and severity of 93

95 the material possessed and the qualitative variation among these types. They found that although the severity of the material possessed varied greatly, many collections showed signs of anchor points that indicated the primary interest of the collector. For example, if this collection mainly concerns mildly erotic images of children the risk of engaging in child sexual abuse may be lower compared to collections mainly concerning images of explicit sexual conduct. Finally, both child pornography offenders who had committed child sexual abuse and internet-only offenders possessed more material depicting severe child sexual abuse when they had a longer history of downloading (Quayle & Taylor, 2002). This suggests that prolonged online engagement leads to habituation and heightens the need for more severe material to reach satisfaction. Eventually, imagery may not be enough for some offenders and results in acting out fantasies in real life. Indeed, some child pornography offenders who committed sexual contact offenses stated that their contact offense was an extension of their online behaviour (Quayle & Taylor, 2003). In addition, engagement with the internet may lead to online contact with other child pornography offenders. According to Krone (2004), the severity of offending increases when an offenders network becomes more intense. For example, other child pornography offenders might provide tips and tricks about how to engage in the internet and how to avoid detection. Also, other offenders behaviours may justify their own offenses which may result in the development of offense-supporting cognitive distortions or may reinforce already existing distorted values and ideas. Moreover, social support from others with similar interests and beliefs seems very important in explaining cross-over behaviour in child pornography offenders (Quayle & Taylor, 2002). It was found that in producers of child pornography, the motivation for taking images was often that it provided online popularity and led to increases in self-esteem (Sheehan & Sullivan, 2010). This may be especially dangerous for fantasy-only offenders because they are more easily tempted to engage in cross-over offenses in order to feel better and less isolated. Furthermore, as offenders may need to prove to others online that they are trustworthy in order to obtain material or to become a member of a certain group, the exchange of new or rare material may be undeniable (Quayle & Taylor, 2003). Some offenders may feel obligated to start producing child pornography themselves and may be stimulated into fulfilling particular requests of others (Quayle & Taylor, 2002). However, although the influences of internet social contacts may contribute to the likeliness of crossing-over to contact offenses, an already existing sexual interest in children seems to be a pre-condition to engage in child sexual abuse (Sheehan & Sullivan, 2010). Finally, engagement with the internet may give access to online contact with children with whom offenders may engage in cybersex (Quayle & Taylor, 2003). Although this does not necessarily lead to sexual contact abuse in the real world, it is a form of cyber sexploitation because it leads children to engage in varying degrees of sexually explicit conversation or activities. It also provides offenders with potential victims who may be seduced into actual sexual contact by grooming (O Connell, 2003). The transmission of online contacts to the offline environment, in turn, may result in cross-over (Quayle & Taylor, 2003) The offline environment Next to access to potential online victims, research shows that offline access to children may also be a risk factor for cross-over in child pornography offenders. Child pornography producers often choose victims who are easily available (Sheehan & Sullivan, 2010). Furthermore, Long et al. (2013) found that child pornography offenders who also committed child sexual abuse were more likely to have access to children than those who did not have committed a contact offense. Victims were most often unrelated children whom offenders knew from the neighborhood (Buschman, Wilcox, Krapohl, Oelrich, & Hackett, 2010). Finally, loneliness seems an important risk for child pornography offending. In turn, having no offline contact with others or with people who do not share pedophilic interests can be a risk factor for cross-over. Quayle and Taylor (2003) found that prolonged engagement with the internet was associated with a decline in offline contact with people. In turn, this can limit healthy evaluation of the appropriateness of offense 94

96 behaviour because newly developed cognitions remain unchallenged by the offline environment. Importantly, research indicates that when pedophiles also receive support from non-pedophiles, they more often have appropriate views about children s sexuality. In contrast, individuals who only have contact with other pedophiles more often acknowledge being supportive of child sexual abuse (Goode, 2010). Hence, child pornography offenders who only engage in contact with other offenders and are isolated from others with more healthy views about this type of offending, may be at greater risk for committing child sexual abuse than those who are also able to discuss their feelings with other non-pedophilic individuals in the offline environment. 6. Summary and discussion We aimed to provide an overview of the contemporary literature on child pornography offenders in order to structure and guide future research. First, we presented the difficulties in defining child pornography and child pornography offending both legally and academically. Second, we identified child pornography offending risk factors pertaining to demographic characteristics, socio-affective difficulties, cognitive distortions, and psychosexual issues. We showed that there are several understudied risk domains, such as offenders negative childhood experiences, early sexual behaviour, and the development and onset of cognitive distortions. However, despite individual differences, it is apparent that many offenders have difficulties in one or multiple risk domains. Third, we highlighted the heterogeneity in child pornography offenders and integrated the risk domains within particular child pornography offenders and offending types. In addition to the typology developed by Elliott and Beech (2009), we argued that individual offenders can be categorized into either periodically prurient or fantasy-only offenders by describing their risk behaviour along the lines of two continua: (1) their features related to criminal behaviour, and (2) their sexual deviance/fantasy. We hypothesized that periodically prurient offenders have more risk factors related to the first continuum, whereas fantasy-only offenders have risk factors mainly related to psychosexual development and hence the sexual deviance continuum. Yet, we have showed that the severity of psychological problems also differs between child pornography offenders within a particular offending type, which may have implications for treatment. Furthermore, in order to achieve a clear treatment plan for individual child pornography offenders, it is recommended to map psychological difficulties on the proposed continua instead of only ascribing offenders to a particular offending type. Fourth, studies indicated that the heterogeneity in child pornography offenders is further reflected in risks for cross-over. Characteristics related to both antisocial behaviour in general (Babchishin et al., 2011; Long et al., 2013) and social withdrawal, such as loneliness, are risk factors for cross-over to child sexual abuse. Whereas periodically prurient offenders may be vulnerable for cross-over because of their impulsive behaviour, fantasy-only offenders are at risk due to feelings of loneliness and low self-esteem. With regard to environmental factors such as engagement with the internet, child pornography offenders may experience habituation through regular exposure, which results in the need for more severe child pornographic material (Quayle & Taylor, 2002). Ultimately, for some offenders more severe material may not be stimulating enough and may turn to hands-on offenses. Furthermore, child pornography offenders may be at risk for negative influences from like-minded others online who provide them with tips and tricks to offend and who justify sexual offenses. Also, they can reinforce distorted values and beliefs, and provide social status and support (Quayle & Taylor, 2002). This may be especially dangerous for those offenders who have less contact with non-pedophiles who could provide them with healthier views about children and sexuality (Quayle & Taylor, 2003). However, because of the stigmatization associated with pedophilia, individuals with a sexual interest in children often seek little support outside the pedophile community (Goode, 2010). Offenders with a sexual interest in children may feel safer or more at ease in disclosing their feelings to like-minded others than to non-pedophilic individuals and may therefore choose to isolate themselves from others in the real world. Finally, other environmental factors 95

97 such as access to children (both online and offline) are also considered risk factors for crossing over to child sexual abuse (O Connell, 2003; Quayle & Taylor, 2003). This review has several limitations. First, the discussed studies differed in using forensic and nonforensic child pornography offender samples. As such, these child pornography offenders may differ in the degree to which they have received treatment and this could have affected their disclosure of offense behaviours (Bourke & Hernandez, 2009). Moreover, several studies included child pornography offenders who had also committed child sexual abuse. Although we have specified this for studies in which this was the case, some results cannot be generalized to child pornography offending in non-sentenced offenders. Finally, almost all studies were solely based on self-reports. Research indicates that in studying child pornography offenders this is more reliable than using official police records (Seto et al., 2011), but still limits their disclosure of risk behaviour (Bogaerts et al., 2010). We suggest two lines of future research that follow from this literature overview. The first line needs to focus more in depth on the developmental pathways to a particular type of child pornography offending. Specifically, the integration of the hypothesized risk factors for the different child pornography offending types and their specific risk factors for cross-over deserve more attention. To this end, qualitative research such as in-depth interviews with child pornography offenders can provide valuable information. Hereby researchers should in particular focus on factors such as the type of child pornography offending, type and severity of child pornographic material, offenders childhood characteristics, such as own experienced abuse, development and onset of cognitive distortions, and pedophilic interest (specifically the age group that offenders prefer). Currently, these factors are often overlooked in research on child pornography offending. Also, from a quantitative viewpoint, cluster analyses seem promising to assess the validity of the proposed continua to differentiate between periodically prurient and fantasy-only offenders. This differentiation can provide the field with more standardized measures and help clinicians to improve treatment planning. The second line of future research may want to focus on non-forensic pedophiles in order to study their coping strategies to deal with their sexual preferences. With regard to sexual deviance, we showed that there is currently little information about factors that explain the relationship between pedophilia and child pornography offending. Moreover, very little is known about the etiology of sexual interest in children (Fagan, Wise, Schmidt, & Berlin, 2002). Insight into pedophilia and its relationship with child pornography offending may offer important knowledge about possible protective factors against this type of offending. Furthermore, insight into the underlying processes of pedophilia and child pornography offending may be helpful in treatment to counteract and prevent reoffending. Finally, it may eventually help pedophiles who are struggling with their feelings and desires to refrain from engaging in (non-)contact offenses. References 18 U.S.C. 2256(8) (2006). Andrews, D., & Bonta, J. (2007). The psychology of criminal conduct (4th ed.). Cincinnati, OH: Anderson. Babchishin, K. M., Hanson, R. K., & Hermann, C. A. (2011). The characteristics of online sex offenders: A metaanalysis. Sexual Abuse, 23(1), doi: / Bates, A., & Metcalf, C. (2007). A psychometric comparison of internet and non-internet sex offenders from a community treatment sample. Journal of Sexual Aggression, 13(1), doi: / Beech, A. R., Elliott, I. A., Birgden, A., & Findlater, D. (2008). The internet and child sexual offending: A criminological review. Aggression and Violent Behaviour, 13(3), doi: /j.avb

98 Bogaerts, S., Buschman, J., Kunst, M., & Winkel, F. (2010). Intra-and extra-familial child molestation as pathways building on parental and relational deficits and personality disorders. International Journal of Offender Therapy and Comparative Criminology, 54(4), doi: / X Bogaerts, S., Vanheule, S., & Desmet, M. (2006). Feelings of subjective emotional loneliness: An exploration of attachment. Social Behaviour and Personality, 34(7), doi: /sbp 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(3), doi: /s y Burke, A., Sowerbutts, S., Blundell, B., & Sherry, M. (2002). Child pornography and the Internet: Policing and treatment issues. Psychiatry, Psychology and Law, 9(1), doi: /pplt Buschman, J., Wilcox, D., Krapohl, D., Oelrich, M., & Hackett, S. (2010). Cybersex offender risk assessment. An explorative study. Journal of Sexual Aggression, 16(2), doi: / Davis, R. A. (2001). A cognitive-behavioural model of pathological Internet use. Computers in Human Behaviour, 17(2), doi: /S (00) Elliott, I. A., & Beech, A. R. (2009). Understanding online child pornography use: Applying sexual offense theory to internet offenders. Aggression and Violent Behaviour, 14(3), doi: /j.avb Elliott, I. A., Beech, A. R., Mandeville-Norden, R., & Hayes, E. (2009). Psychological profiles of Internet sexual offenders: Comparisons with contact sexual offenders. Sexual Abuse, 21(1), doi: / Fagan, P. J., Wise, T. N., Schmidt, C. W. Jr., & Berlin, F. S. (2002). Pedophilia. Journal of the American Medical Academy, 288(19), doi: /jama Goode, S. D. (2010). Understanding and addressing adult sexual attraction to children: A study of paedophiles in contemporary society. Abingdon: Routledge. Healy, M. A. (1996). Child pornography: An international perspective. Paper presented at the World Congress against Commercial Sexual Exploitation of Children. Stockholm, Sweden. Retrieved from Henry, O., Mandeville-Norden, R., Hayes, E., & Egan, V. (2010). Do internet-based sexual offenders reduce to normal, inadequate and deviant groups? Journal of Sexual Aggression, 16(1), doi: / Hossack, A., Playle, S., Spencer, A., & Carey, A. (2004). Helpline: Accessible help inviting active or potential paedophiles. Journal of Sexual Aggression, 10(1), doi: / Howitt, D., & Sheldon, K. (2007). The role of cognitive distortions in paedophilic offending: Internet and contact offenders compared. Psychology, Crime & Law, 13(5), doi: / Krone, T. (2004). A typology of online child pornography offending. Trends and Issues in Crime and Criminal Justice, 279, 1-6. Retrieved from Lanning, K. V. (2010). Child Molesters: A behavioural analysis for professionals investigating the sexual exploitation of children (5th ed.). Alexandria, VA: National Center for Missing & Exploited Children. Retrieved from 97

99 Laulik, S., Allam, J., & Sheridan, L. (2007). An investigation into maladaptive personality functioning in Internet sex offenders. Psychology, Crime & Law, 13(5), doi: / Long, M. L., Alison, L. A., & McManus, M. A. (2013). Child pornography and likelihood of contact abuse: A comparison between contact child sexual offenders and noncontact offenders. Sexual Abuse, 25(4), doi: / Magaletta, P. R., Faust, E., Bickart, W., & McLearen, A. M. (2012). Exploring clinical and personality characteristics of adult male internet-only child pornography offenders. International Journal of Offender Therapy and Comparative Criminology, 58(2), doi: / X Marshall, W. L. (1993). The role of attachments, intimacy, and lonliness in the etiology and maintenance of sexual offending. Sexual and Marital Therapy, 8(2), doi: / McCarthy, J. A. (2010). Internet sexual activity: A comparison between contact and non-contact child pornography offenders. Journal of Sexual Aggression, 16(2), doi: / Middleton, D., Elliott, I. A., Mandeville-Norden, R., & Beech, A. R. (2006). An investigation into the applicability of the Ward and Siegert Pathways Model of child sexual abuse with Internet offenders. Psychology, Crime & Law, 12(6), doi: / Neutze, J., Seto, M. C., Schaefer, G. A., Mundt, I. A., & Beier, K. M. (2011). Predictors of child pornography offenses and child sexual abuse in a community sample of pedophiles and hebephiles. Sexual Abuse, 23(2), doi: / Niveau, G. (2010). Cyber-pedocriminality: Characteristics of a sample of internet child pornography offenders. Child Abuse & Neglect, 34(8), doi: /j.chiabu O Connell, R. (2003). A typology of cybersexploitation and on-line grooming practices. Preston: University of Central Lancashire. Retrieved from Patton, J. H., Stanford, M. S., & Barratt, E. S. (1995). Factor structure of the Barratt impulsiveness scale. Journal of Clinical Psychology, 51(6), doi: / (199511)51:6<768::AID- JCLP >3.0.CO;2-1 Quayle, E., & Taylor, M. (2002). Child pornography and the Internet: Perpetuating a cycle of abuse. Deviant Behaviour, 23(4), doi: / Quayle, E., & Taylor, M. (2003). Model of problematic Internet use in people with a sexual interest in children. CyberPsychology & Behaviour, 6(1), doi: / Reijnen, L., Bulten, E., & Nijman, H. (2009). Demographic and personality characteristics of Internet child pornography downloaders in comparison to other offenders. Journal of Child Sexual Abuse, 18(6), doi: / Sawle, G. A., & Kear-Colwell, J. (2001). Adult attachment style and pedophilia: A developmental perspective. International Journal of Offender Therapy and Comparative Criminology, 45(1), doi: / X Seigfried, K. C., Lovely, R. W., & Rogers, M. K. (2008). Self-reported online child pornography behaviour: A psychological analysis. International Journal of Cyber Criminology, 2(1), Retrieved from 98

100 Seto, M. C. (2010). Child pornography use and internet solicitation in the diagnosis of pedophilia. Archives of Sexual Behaviour, 39(3), doi: /s Seto, M. C. (2012). Is pedophilia a sexual orientation? Archives of Sexual Behaviour, 41(1), doi: /s Seto, M. C., Cantor, J. M., & Blanchard, R. (2006). Child pornography offenses are a valid diagnostic indicator of pedophilia. Journal of Abnormal Psychology, 115(3), 610. doi: / X Seto, M. C., Hanson, R. K., & Babchishin, K. M. (2011). Contact sexual offending by men with online sexual offenses. Sexual Abuse, 23(1), doi: / Seto, M. C., Wood, J. M., Babchishin, K. M., & Flynn, S. (2012). Online solicitation offenders are different from child pornography offenders and lower risk contact sexual offenders. Law and Human Behaviour, 36(4), 320. doi: /h Sheehan, V., & Sullivan, J. (2010). A qualitative analysis of child sex offenders involved in the manufacture of indecent images of children. Journal of Sexual Aggression, 16(2), doi: / Sheldon, K., & Howitt, D. (2007). Sex offenders and the Internet. Chichester, U.K: John Wiley & Sons. Sullivan, C. (2009). Internet traders of child pornography and other censorship offenders in New Zealand: Profiling research-update. Auckland, New Zealand: Department of Internal Affairs. Retrieved from: December2009.pdf Surjadi, B., Bullens, R., van Horn, J., & Bogaerts, S. (2010). Internet offending: Sexual and non-sexual functions within a Dutch sample. Journal of Sexual Aggression, 16(1), doi: / Taylor, M., Holland, G., & Quayle, E. (2001). Typology of paedophile picture collections. The Police Journal, 74, 97. Retrieved from van Wijk, A., Nieuwenhuis, A., & Smeltink, A. (2009). Een verkennend onderzoek naar downloaders van kinderporno [An exploratory investigation on child pornography offenders]. Arnhem: Bekereeks. Ward, T., & Siegert, R. J. (2002). Toward a comprehensive theory of child sexual abuse: A theory knitting perspective. Psychology, Crime and Law, 8(4), doi: / Webb, L., Craissati, J., & Keen, S. (2007). Characteristics of Internet child pornography offenders: A comparison with child molesters. Sexual Abuse, 19(4), doi: /s Wells, M., Finkelhor, D., Wolak, J., & Mitchell, K. J. (2007). Defining child pornography: Law enforcement dilemmas in investigations of Internet child pornography possession 1. Police Practice and Research, 8(3), doi: / Wortley, R. K., & Smallbone, S. (2006). Child pornography on the internet (Problem-Specific Guides Series, Problem-Oriented Guides for Police, No. 41). Retrieved from 99

101 Being sexually attracted to minors: Sexual development, coping, and risk factors for offending in self-identified pedophiles Jenny A.B.M. Houtepen MSc 456, & Prof. Dr. Stefan Bogaerts 123 * Correspondence concerning this article should be addressed to J.A.B.M. Houtepen, Tilburg University, School of Social and Behavioural Sciences, Developmental Psychology, P.O. Box 90153, 5000 LE Tilburg, the Netherlands. Phone: ; [email protected] Being sexually attracted to minors: Sexual development, coping, and risk factors for offending in selfidentified pedophiles Pedophilia (or pedophilic disorder in the DSM-5) is defined as the sexual attraction to prepubescent children (American Psychiatric Association, 2013). This sexual interest is indicated by recurrent, intense, sexually arousing fantasies, sexual urges, or behaviours involving sexual activity with children generally aged 13 years or younger, and must be present for at least six months. There is ongoing debate about the etiology of pedophilic attraction. Currently, pedophilia is viewed as a psychological disorder with early life risk factors such as early sexual behaviour (Goode, 2010), history of sexual abuse (Freund, Watson & Dickey, 1990), and prenatal or early childhood neurodevelopmental problems (Blanchard, 2002). However, pedophilic interest seems more common in contemporary society than one would expect. Estimates range from 3 to 9% of males in community samples who are in some way sexually attracted to children (Wurtele, Simons & Moreno, 2013). Furthermore, Seto (2012) showed recently that pedophilia has many similarities with views about sexual gender orientation (i.e., hetero- and homosexual feelings), such as the early age of onset, the fantasy or desire to engage in romantic behaviour, and the stability of this sexual preference across time. Therefore, Seto suggested that pedophilia may be better considered as a sexual age orientation, instead of a mental health disorder or a sexual preference that is chosen or somehow learned. This view is controversial given that earlier research on child sexual abusers showed that these individuals may be impulsive (Neutze, Seto, Schaefer, Mundt & Beier, 2011), have antisocial, narcissistic, or passive-aggressive personality structures (Bogaerts, Buschman, Kunst, & Winkel, 2010), possess little victim empathy, and have distorted cognitions supportive of having sex with children (Cohen & Galynker, 2012; Marshall, Hamilton & Fernandez, 2001). Yet, not all child sexual abusers are pedophiles. Conversely, someone who experiences distress because of intense sexual fantasies involving children but has never had any physical contact with a child, can receive the clinical diagnosis of pedophilia without having engaged in child sexual abusive behaviour. Furthermore, even pedophiles who engage in offense behaviour show great variation in psychological make-up and hence in the type and severity of psychological risk factors they possess. More specifically, in a recent review on child pornography offending (Houtepen, Sijtsema & Bogaerts, 2014), it was proposed that risk factors related to child pornography offending can be viewed along the lines of two continua: (1) features related to criminal behaviour, and (2) sexual deviance or fantasy. This suggests that some offenders do not engage in offending out of antisocial tendencies, but rather because of psychopathology related to deviant sexual interest, such as own history of abuse in childhood (i.e., child sexual and/or physical abuse or neglect) (Webb, Craissati, & Keen, 2007), intimacy deficits (Bogaerts, Buschman, Kunst, & Winkel, 2010), low self-esteem, loneliness (Henry, Mandeville-Norden, Hayes & 4 Department of Developmental Psychology, Tilburg University, the Netherlands 5 Belgian Cybercrime Centre of Excellence for Training, Research and Education, Leuven, Belgium 6 FPC de Kijvelanden, Poortugaal, the Netherlands 100

102 Egan, 2010), or other general interpersonal difficulties. For example, for these individuals, sexual activity with children may evoke less anxiety compared to sexual activity with same-aged partners (Cohen & Galynker, 2009; Cohen, Grebchenko, Steinfeld, Frenda & Galynker, 2008). Some pedophiles may engage in child pornography offending in order to relief sexual arousal, but have distorted cognitions concerning both the images and the act of watching. These cognitions minimize the severity of watching of child pornographic material to the extent that some pedophiles believe that children do not have to be harmed in order to produce this material (Quayle & Taylor, 2002). Moreover, pedophiles often acknowledge having romantic and loving feelings for children next to desires to engage in sexual contact (Seto, 2012). Therefore, it has been suggested that some pedophilic offenders are less likely to engage in child sexual contact offenses when they are directly responsible for the abuse (Houtepen et al., 2014). Because of the great stigma associated with pedophilia, some pedophiles may be troubled by their sexual interest (Freimond, 2013), fear losing control over their feelings (Hossack, Playle, Spencer & Carey, 2004), or experience difficulties in dealing with feelings that are forbidden in contemporary society. In order to refrain from offending these pedophiles need sufficient self-control and some need professional help in order to learn better coping strategies to relief sexual arousal. However, disclosure of sexual interest is a highly frightening activity for many pedophiles. Even when they do find the courage to seek help, resources to provide pedophilic individuals with sufficient help are currently lacking (Seto, 2012). This is in part because previous research on pedophilia has focused mainly on pedophilic offenders who were sentenced to imprisonment or compulsory treatment because of sex offenses (Capra, Forresi & Caffo, 2014; Hall & Hall, 2009). As a consequence, information about pedophiles outside forensic and criminal settings is lacking. Although it has been suggested that many undetected pedophiles have engaged in child pornography offending or sexual contact with children at some point in their lives (Neutze et al., 2011), it is not clear what other (possible preventive) strategies these pedophiles use in order to cope with their sexual interest and secretive nature of these feelings. Hence, because there is little empirical evidence to support that these non-forensic and non-incarcerated pedophiles show the same risk factors as convicted pedophiles, it is not clear what factors other than pedophilic interest, result in actual child sexual abusive behaviour. In this study we aim to contribute to the understanding of pedophilia by providing more insight into pedophilic attraction and risk and protective factors in non-clinical and non-incarcerated pedophiles. To this end, we conducted two studies. In the first study we interviewed 15 self-identified pedophiles about their sexual attraction. First, in order to provide more insight into pedophilic sexual development, we examined the onset of pedophilic interest, the content of this attraction (i.e., the extent to which there were sexual and/or romantic fantasies and desires present), and risk factors of pedophilia such as early sexual behaviour and history of child sexual abuse. Second, we investigated experiences with acknowledging being pedophile and coping mechanisms that are used in dealing with forbidden feelings and stigma. Third, we examined behaviours to relief sexual arousal, including self-reported offense behaviour such as child pornography offending and sexual child abuse. In the second study we examined some of the earlier described risk factors to engage in sexual offending in pedophiles by the use of standardized questionnaires in order to provide a more comprehensive image of the psychological make-up of these non-forensic and non-incarcerated pedophiles. More specifically, we compared pedophiles with males in the general population on risk factors including history of abuse, number of experienced negative life events, intimacy deficits, self-esteem, and selfcontrol/impulsivity. 101

103 Study 1: Pedophilic sexual development, coping with forbidden feelings and relieving sexual arousal Method Participants The sample comprised 15 self-identified males with pedophilic interests. All males were Dutch, except for one male who came from Belgium. It is important to note that we use the term pedophile to refer to individuals that admit to having pedophilic feelings, but who do not necessarily have an official diagnosis of pedophilia according to the DSM. During the interviews, pedophiles were asked specific questions about the age group they were most attracted to in order to be able to discriminate between individuals who qualify for the official diagnosis and those who are sexually attracted to children, but are not considered having pedophilic disorder in clinical terms. We found that 2 individuals did not fit the clinical description of pedophilia. Instead, 2 fitted the term hebephile which denotes the erotic preference for pubescent children (generally aged 12 to 14 years) (Blanchard et al., 2009), but is not recognized as an official diagnosis by the DSM. After sending an online request to participate in the study, 18 pedophiles responded. After we provided respondents with further information, 3 pedophiles dropped out and did not respond again. Of the final sample of 15 pedophiles, there were no missing data on any of the questionnaires used. Furthermore, most pedophiles were open about their sexual preference, behaviour, and past during the interview. Procedure Pedophiles were approached through an online request that was placed on the forums of three Dutch pedophile sites on the Internet in November Furthermore, these requests were sent by to pedophiles who were known to be active in self-help groups in the Netherlands at that time. Pedophiles who were interested in participating in this study could contact one of the researchers by or telephone, or they could fill out a contact form that was placed on the website of Tilburg University, so they could be contacted to provide further information. After pedophiles agreed to participate, personal appointments were made. Most of the interviews took place at Tilburg University. We chose this location because we felt that this was a safe environment in which the anonymity of individuals was guaranteed. However, some of the pedophiles preferred a home interview or were not able to travel to Tilburg University for personal reasons. In two cases, pedophiles were visited by the first and second author of this study. In one case, the second author met the pedophile at another university in the Netherlands. All other interviews were conducted by the first author at Tilburg University. Interviews took place between December 2013 and January Pedophiles were first asked to sign an informed consent form in which the purpose of this study was explained. Furthermore, pedophiles were informed that any disclosed information about previous or current offenses that were not yet known to the police but potentially involved the safety of a minor, were reported to the authorities. However, to guarantee the anonymity of individuals, we did not ask for any personal information during the study. In one case we suspected that a pedophile was currently having sexual contact with minors, but he did not disclose any specific information about possible victims or abusive acts. After having contacted organizations dedicated to support victims of abuse, we found that he had not disclosed enough personal information for identification. All other responses were strictly confidential and pedophiles were able to withdraw from the study at any time. The in-depth interviews took approximately 45 minutes and were recorded after approval by the individual. Audio recordings were deleted after the interviews were transcribed. We strongly emphasized the importance of answering honestly to the questions asked. One pedophile did not want us to record the part of the interview were we asked questions about child pornography offending. We interviewed pedophiles about their sexual preference, demographic characteristics, sexual development, coping mechanisms in 102

104 dealing with pedophilic feelings, quality of life, and their ways of relieving sexual arousal, including engagement in child pornography offending and child sexual abuse. We also assessed background information such as mental health issues. In order to provide a complete overview of the interview data, we firstly summarized the answers pedophiles gave on the interview questions. Secondly, we compared their responses and rated the number of pedophiles who provided us with a somewhat similar response for every interview question. Pedophiles received a refund of all travel expenses and a 20 euro gift voucher for their participation. Results Descriptive Statistics More than half of the pedophiles (9) were highly educated (i.e., either a bachelor or master s degree), 2 received intermediate vocational education, 2 only finished high school, and 2 had less than a high school education. Only a few (3) were married or in a relationship at the time of the interviews, and 2 were divorced. Three pedophiles reported having children. Of the pedophiles who were involved in a romantic relationship with an adult partner, most admitted not wanting to stay single for the rest of their lives and preferred to be in an intimate relationship even though they did not love their partner in a romantic way. Five pedophiles had never been involved in a serious relationship. As noted earlier, 2 of the 15 pedophiles better fitted the term hebephile. Aside from this distinction, there was still large variation in preferred sexual age orientation between individuals. For example, some pedophiles with an interest in prepubescent minors acknowledged to be sexually attracted to children from 5 years of age, whereas others preferred minors that were almost starting puberty. Furthermore, not all pedophiles had an exclusive sexual interest in children, but 9 were also attracted to adult males and/or females. Finally, 8 pedophiles were solely attracted to boys, 3 to girls, and 4 were attracted to both sexes. Seven pedophiles had experienced mental health problems. They were often troubled with feelings of anxiety, depression, and inferiority. Also, one pedophile was diagnosed with Post Traumatic Stress Disorder (PTSD), and one reported to have once tried to commit suicide because of severe psychological difficulties, including drug abuse, and psychotic episodes. Of the 7 pedophiles with mental health issues, one argued that although a professional had declared him unfit for work due to psychological illness, he did not consider himself as having major mental health problems. Yet, he later disclosed having another (non-diagnosed) paraphilia (i.e., fetishism), being diagnosed with Attention Deficit Hyperactivity Disorder (ADHD), and having a history of aggression-related issues in childhood, indicating multiple risk factors related to criminal behaviour. In addition, one other pedophile reported being diagnosed as unfit for work due to psychological difficulties, yet he could not specify his psychopathology and he also did not consider himself mentally ill, therefore we decided not to label him as such. For 3 pedophiles, mental health issues started after the age of 16. The remaining 4 had already experienced mental health issues early in life. Sexual development Onset of pedophilia Eleven pedophiles described becoming aware of their sexual attraction to minors as a gradual process. Their sexual development started relatively normal at puberty, while they were still interested in their same-aged peers. However, as they became older, most (11) recognized becoming different from others because their sexual age preference did not mature with them. Others (4) became aware of their sexual attraction quite sudden. During contacts with minors either in real life, or by seeing a picture, they became aware of their sexual preference for youngsters. From that moment on, all except one (see section coping with forbidden feelings) had felt this sexual attraction being present during their lives, indicating a stable sexual preference. 103

105 Content of sexual interest Ten pedophiles acknowledged that their attraction to minors was not solely sexual, but that romantic feelings were also present. They reported falling in love with a child, and/or had fantasies about having a real romantic relationship with a minor. Also, when asked what they particularly found attractive in minors, most (8) acknowledged that it was a combination of both physical characteristics, such as their beauty, or bodily shapes, and behaviour, such as openness, spontaneity, honesty, or naivety, whereas only 4 referred solely to children s appearances. Moreover, many (10) argued that having close relationships with children (e.g., friendship, adult coaching role) was highly satisfying, for example because it provided them with the opportunity to make children happy. Especially for pedophiles who acknowledged being less sexually aroused because of their age (3; including 2 who were in their early to mid-thirties), having close social contacts with minors was more important than the sexual component. Furthermore, some (4) stated when they were less socially involved with children they felt more fixated on and troubled by their sexual interest in minors. For example, they argued that during these times certain needs for intimacy and closeness were not met, or that it was easier to imagine children as sexual beings when there was no real contact to question such distorted cognitions about children s sexuality, indicating that social contact with minors also served as a preventive strategy against risks for offense behaviour. In contrast, others (2) argued that although romantic fantasies were present, the sexual interest was most important. Moreover, 2 admitted having romantic relationships with minors in the past, but mainly focused on their sexual engagement. Therefore, it seems that their image of being in love was somewhat distorted or entangled with their feelings of sexual arousal. Pedophiles who were not interested in having romantic relationships with children stated for example that children were not as mature in their cognitive development as they were, or only had fantasies in which they were still a child who had sexual relationships with minors because they did not find the unequal power balance attractive. One pedophile stated that he did not fantasize about children at all, neither in a sexual nor romantic way. He strongly desired to set aside his pedophilic feelings completely, using suppression as the core strategy to deal with his pedophilic interest. Early sexual behaviour and experience of sexual abuse Eight pedophiles reported having early sexual experiences (ages 6 to 11 years), ranging from discovering masturbation (1), to playing doctor with other children (4) and more severe sexual play with others (1), to actual penetration (2). One acknowledged that this early sexual experience had in some way effected his sexual development, and argued: This certainly has made me think about sexuality too early, and I guess I was fixated on this for quite a while. Moreover, 2 pedophiles who both had engaged in excessive sexual behaviour very early in life (i.e., one acknowledged having performed multiple sexual acts on others in primary school, and one acknowledged engaging in penetration before the age of 8), seemed to project these experiences on the concurrent needs of children. Both evaluated their early sexual experiences very positively and now felt that these behaviours were desired by all children, because they argued that children s sexuality was equal to that of adults. However, one of these pedophiles also reported being sexually abused in childhood. Because his abuse experience seemed incompatible with his positive view about sexual behaviour in childhood, he was asked to explain the difference between early sexual behaviour and child sexual abuse. He argued that abusive sexual behaviour is a coercive and undesired act, hereby making the implicit assumption that under all other conditions (including sexual engagement with an adult), children already have the capacity to voluntarily choose to engage in such behaviour. Five others had their first sexual experience around the age of 12, including one who acknowledged being very sexual active at that age, engaging in sexual experiments with numerous boy- and girlfriends. For the other 4, sexual acts early in life only included discovering masturbation. 104

106 Moreover, not all pedophiles experienced sexual behaviour on a young age. One did not remember his first sexual encounter, and one acknowledged to have engaged in sexual activity in late puberty for the first time. Finally, one admitted never having engaged in any real sexual activity to relief sexual arousal, including masturbation. Hence, although early sexual behaviour may have fixated some pedophiles on sexual activity with children, for others such a conditioned association between children and sexual attraction seems to be non-existent. Coping with forbidden feelings Reactions to the realization of pedophilic feelings differed from very negative to extremely positive. In general, most (12) pedophiles had struggled with their sexual attraction in some way. Some felt quite disturbed when acknowledging that the term pedophile fitted their sexual feelings. Because of negative views by both the media and the general public and the subsequent undifferentiated flow of information concerning pedophilia, these pedophiles were afraid of being stigmatized as monsters or child rapists. Also, some felt ashamed and tried to suppress their sexual feelings altogether by using avoidant coping strategies, such as drug abuse. Others (3) had little trouble with accepting their feelings, but were concerned about opinions of others. Therefore, some pedophiles experienced a hard time in keeping their feelings a secret. Of note, one pedophile argued that when he was discovering his sexual attraction, society was less criminalizing pedophilia as it is today. Because he did not have to be secretive about his sexual feelings, these feelings were not subject to rumination or obsession, thus showing the impact that stigma can have on pedophiles well-being. Moreover, this pedophile also stated that when he was married, he forgot his pedophilic interest for a while suggesting that fixation diminished when he was in a meaningful relationship. Finally, one pedophile was extremely positive about his sexual feelings. Although he knew that his feelings were wrong, he thought of it as something honorable, because he had always liked going against the rules. Indeed, he acknowledged having problems with authority from a young age which had led to problematic behaviour, such as school drop-out and not being able to hold a long-term job. Thirteen pedophiles sought help with a professional, like-minded others or other meaningful individuals in their environment. Six pedophiles had sought help from a physician, psychologist, or sexologist, either in order to get assistance in accepting, and/or coping with sexual attraction to minors directly, or because of symptoms that partly seemed to be the result of pedophilic interest, such as depressive feelings and loneliness. Three argued that this help was not sufficient, because of a lack of clinical knowledge about pedophilia in general. For one pedophile, this was also a reason for not seeking help at all. The other 3 argued either that professional support helped them to accept their sexual attraction, because they learned to separate their sexual feelings from sexual behaviour, and hence, to dissociate themselves from stories about child rapists which had troubled them in the past. Or they acknowledged that by seeking help and talking about their feelings, they were stimulated to disclose their feelings to others. In contrast, 2 pedophiles felt that it was too risky to disclose their sexual feelings to a professional, because of the negative stigma of pedophilia. Yet, one of them desired help because he wanted a professional to estimate his risk for offending objectively. His sexual attraction kept worrying him and he repeatedly felt afraid to lose control over his feelings. This struggle with pedophilia became also evident when we asked pedophiles to report their subjective well-being on a scale of zero to ten, which resulted in an average score of 6.3 (SD = 1.6; range = 3 8.5, N = 14 [one pedophile had troubles expressing his subjective well-being by a specific number, but reported Currently, I am not happy ]). Three pedophiles found support through contact with other pedophiles, for example by visiting selfhelp groups or by engaging in (online) contact with like-minded others. Furthermore, 3 others noted that they just enjoyed having such contacts, because they could really be themselves and talk openly about their feelings. Some pedophiles argued that it was the recognition of shared feelings that comforted them, or that 105

107 others had provided them with the courage to talk to other non-pedophilic individuals in their environment. One found it particularly reassuring to see that many other pedophiles did not engage in sexual behaviours with children: They give me moral support. The media kind of tell you that you re a ticking time bomb, but then when you go online You see many ticking time bombs that do not go off at all. Finally, 11 pedophiles also disclosed their feelings to non-pedophilic individuals to receive social support through others acceptance, or to have someone around with whom they could be themselves. Some pedophiles also found it important that others knew about their sexual preference, so that they could function as safe guards when they were around children. Even though many of them felt sure that they would not cross any legal boundaries, some felt safer knowing that someone else was able to give them feedback on their behaviour. Similarly, one pedophile acknowledged that he rather talked about his sexual feelings with non-pedophilic individuals than with other pedophiles, because he felt that non-pedophiles were better able to evaluate his cognitions on a normal level. In turn, he thought that other pedophiles would just confirm already existing cognitions. Two pedophiles had never disclosed their feelings to others. One argued that he did not expect people to be accepting, and because he was not looking for disapproval he kept his feelings for himself. Furthermore, according to him his feelings never had concerned him that much. The other pedophile admitted somewhat proudly that he had engaged in multiple romantic relationships with minors in the past, and therefore did not want people to know about his pedophilic interest (i.e., he argued: Let sleeping dogs lie! ). Finally, next to social support, 8 pedophiles had experienced some negative reactions to their disclosure. Although most understood that people felt uncomfortable knowing that someone close to them had pedophilic interests, they argued that this negativity contributed to secretive behaviour which in turn may put some at risk for committing sex offenses. Relieving sexual arousal Child pornography offending Eleven pedophiles admitted to have watched child pornography at some point during their life. Four had come into contact with the legal justice system for possession of child pornography, but only 3 were convicted. Most pedophiles searched for child pornographic material on newsgroups, peer-to-peer programs, and through regular search engines. Some also received material from others, and/or watched together with others. However, 8 pedophiles argued that they downloaded years ago, often in puberty when they were still very sexually focused and/or were confused about their sexual attraction to minors. Currently, these pedophiles stated not to engage in child pornography offending anymore, with one exception. One pedophile hesitantly admitted that he sometimes still received some child pornographic material from other pedophiles. The type of material that pedophiles watched differed greatly, ranging from virtual child pornography to material depicting minors engaging in sexual activity with adults. However, 9 had closely considered what type of material they could justify watching. For example, they preferred material that depicted children posing naked, old material, and/or material that depicts sexual activities between minors, because they argued that in order to produce this material, children did not have to be sexually abused (anymore). Interestingly, these justifications seemed to overlap with their views about children s sexuality. According to 11 pedophiles, children often show signs of developing sexuality in an experimental fashion, such as through experimenting with nakedness and playing doctor with other children. Therefore, watching child pornographic material where there are no adults involved was considered as viewing material of children who are enjoying themselves, but not as an act that contributed to child sexual abuse. However, many pedophiles mentioned that it is difficult to perform a targeted search for a special type of child pornography, especially if you do not want to pay (which all did not want, because most 106

108 argued that this could also be stimulating sexual abuse of minors in producers). Therefore, in order to see what sort of child pornography a file contains, it often needs to be downloaded first. As a result, most have also been (unintentionally) confronted with material depicting minors being sexually abused. Furthermore, 2 acknowledged that material depicting sexual activity between children and adults was sexually arousing to them. These pedophiles also had more problematic views concerning children and sex. For example, one stated that if children initiated sexual behaviour it is in some cases acceptable to cooperate and engage in the un-heavy stuff, but not penetration, because the latter would be physically impossible. The other pedophile thought that children can desire and engage in the same sort of sexual activities as adults. In addition to the 11 pedophiles who watched illegal material, 3 either watched legal material of children depicted naked, or movies that were considered legal earlier but that are currently labeled as child pornography. For example, because those films included shots of minors being naked while having an erection. One admitted that although at first he did not feel the need to search for child pornography, searching for naked pictures of minors made him curious to more severe material. Yet, as a barrier to offending he said that he never had been a risk taker and that he could not risk getting caught because of his wife and children. Similarly, others also acknowledged the fear of getting caught as a vital protective factor against offending, or argued that they simply did not feel the need because their sexual fantasy had always been sufficient to relieve arousal. Seven others argued that they needed new or more severe material in order to get sexually aroused at the time they were downloading. Moreover, 7 reported experiences of having mixed feelings while watching child pornography. For example, they stated: I regularly felt pathetic or dirty, and I felt sick and was disgusted with what I saw, but at the same time I was frightened by my own excitement. This discomfort and the subsequent realization that this behaviour is harmful to children were for some sufficient reason to stop watching child pornography. One described that he felt his cognitions getting more sexualized: If I was watching a normal non-pornographic movie involving children for example, I thought to myself: When are they going to take off their clothes?. Moreover, he recognized that although normally he was very open to people, he got more secretive when watching child pornography and often felt agitated. One pedophile stated that with time, he got more excited from searching and collecting the material than from actually watching it. For him, child pornography offending rather served the function of a collective addiction. Finally, one pedophile admitted that at times he used downloading as a substitute for child sexual hands-on abuse or at least felt that he needed to download material in order to refrain from hands-on offending. In contrast, 5 pedophiles reported that their sexual arousal shows stable patterns over time; their preferred child pornography images remained sufficient to relieve sexual arousal. Overall, none of these pedophiles reported cross-over sexual deviant behaviour (hands-off to hands-on). Two pedophiles watched child pornography or photographs of children depicted naked only once. Others had watched this material regularly for a while, ranging from one week to fifteen years. With regard to exchanging material, only one admitted to have consciously shared material with familiar others, because he thought they would enjoy this type of material without the expectation to receive child pornographic material in return. None had engaged in the production of child pornographic material, but one was being charged (but not convicted) for production of child pornography because he had made photographs of naked children who were bathing when he was on holiday. According to him, there was nothing sexual about those pictures and the depicted children were too young for him to be sexually attractive. Finally, one pedophile had watched child pornography together with minors (aged 14 to 15). Compared to others, he was very specific and elaborative about the material that he liked to watch, returning to the subject multiple times in a somewhat maladjusted manner, and specifically describing multiple sexual acts. Only one other pedophile who did not engage in child pornography offending, used such offensive language in order to describe sexual acts or fantasies. 107

109 Child sexual contact abuse During the interview, we asked pedophiles if they had ever engaged in physical sexual contact with minors. Ten stated never to have engaged in such contact. Three of them admitted that they had considered seeking sexual contact with a minor at some point, but never fulfilled these fantasies. One acknowledged that this interest was especially salient when he was engaged in child pornography offending; at that time, he also fantasized about how such contact could be initiated. Finally, this was his main reason to stop downloading. Yet, the other 7 pedophiles did never seriously consider having sexual contact with a minor regardless of their sexual desires. For example, one reported that although the sexual desire was present, he could keep his sexual arousal under control and had never harassed children with these feelings. In contrast, one admitted that he still feared that he would commit a sexual contact offense even though he could not imagine himself hurting a child. Based on their personal experiences, 4 of these pedophiles said that although they never initiated sexual contact with children, children may sometimes try to involve adults into their sexual behaviour. According to them, in such cases adults should take responsibility by disapproving this type of behaviour and/or explain to children the reason for the inappropriateness. These 4 pedophiles mainly considered sex between children and adults as abuse because of the age difference and subsequent imbalance in power, but did not seem to consider children unwilling to engage in such behaviour per se. For some, the inappropriateness was mostly determined by the current taboo on children s sexuality. They argued that sex between an adult and a minor would not per definition be accounted for as sexual abuse. However, because children would not be able to talk to others about their sexual engagement this would always be problematic in contemporary society. So although these pedophiles did not really feel that having sex with children is wrong, they made use of strategies in which they could uphold their own feelings and beliefs of their sexual fantasies, together with respecting the rules of current society against having sex with children (i.e., children are already experiencing sexual feelings, but should not be able to express these with adults because of current stigma ). Finally, one pedophile felt concerned when a minor gave sexual signals to him, because he was not sure if other (non-pedophilic) adults had similar experiences. Therefore, he was frightened that he had done something wrong that provoked this sexual interest. The final 5 pedophiles admitted that they had engaged in sexual behaviour with minors, but none of them defined this behaviour as child sexual abuse. In one case, the sexual contact happened more than 20 years ago with a minor with whom he was in love. According to him, this sexual contact was initiated by the minor who was just as him sexually distorted because of early abusive experiences. Finally, this had led to a conviction, and community service. One other pedophile admitted that he had sometimes crossed legal boundaries when providing minors with sexual information. However, he thought that although he had maybe crossed a line, his victims would probably not blame him for what he had done. This shows that although he knew that his behaviour is considered criminal by law, he did not really feel or understand that this type of behaviour is unacceptable. Similarly, one other pedophile acknowledged that he had sought sexual contact with minors when he was younger, but that he did not understand in that period that what he did was wrong and he believed that it was pleasurable experience for both parties. All stated that they would never engage in sexual contact with children again. Two pedophiles acknowledged to have engaged in multiple romantic relationships with minors, with whom they also had engaged in sexual behaviour. Both seemed to understand at some level that their behaviour was considered abusive, because they actively tried to hide this behaviour from others. That is, one kept this behaviour hidden from his environment and the other was being vague about his sexual engagement with minors during the interview. When asked about any engagement in current sexual handson behaviour, the latter answered: Unfortunately, I cannot tell you that. The other pedophile argued that 108

110 he was not looking for contact with minors since a long time, so that he was not a risk for committing child sexual abuse anymore. Other ways to relieve sexual arousal Seven pedophiles acknowledged that they are (or were) also able to relief sexual arousal by watching adult pornography or by engaging in sexual activity with adults. For most, masturbation and fantasy were enough to deal with their sexual interest, and some acknowledged that sexual experiences became far less important with age. Instead, the social component of pedophilia (i.e., having close contact with minors) was experienced as more meaningful. Finally, one pedophile used special Japanese child sex dolls with which it is possible to engage in sexual behaviour. He argued that because the child doll clearly is an object that is especially made for someone to relief sexual arousal, there is no association with real children. He argued that when fantasizing about engaging in child sexual behaviour the boundary between real and make-belief sexual appropriate behaviour is less clear-cut. Therefore, by using the dolls to relieve sexual arousal, he was not seduced to committing child sexual contact abuse. Conclusion In the first study we aimed to provide more insight into pedophilic attraction by interviewing nonforensic and non-incarcerated pedophiles. In concordance with the study of Seto (2012), we found that pedophiles often recognized the onset of sexual attraction in puberty that remained rather stable over time. Next to sexual attraction, most pedophiles reported having romantic feelings for children and emphasized the importance of being able to engage in social contact with minors. With regard to risk factors for developing pedophilia, most pedophiles experienced early sexual behaviour with other minors, which seemed to have influenced their cognitions concerning children as sexual beings. However, it remains unclear if these early sexual behaviours can be seen as deviant, because there is currently little information on normal levels of childhood sexual behaviour in the general population (Elliott & Beech, 2009). Still, according to conditioning theories of the etiology of pedophilia, these early sexual experiences with other children is a first introduction to interpersonal sexuality but also sexual preferences can become aware because individuals are conditioned to become attracted to particular childlike physical appearances. These characteristics are then never properly adjusted to older individuals. Although it is not clear what mechanism is causing this developmental stagnation, conditioning theories suggest that in many pedophiles, sexual attraction is in some way socialized through experience (Seto, 2004). For some pedophiles, their pedophilic interest was solely sexual (i.e., an expression of a sexual drive) without romantic feelings. Furthermore, in some pedophiles sexual development started after puberty. Finally, with regard to age preference and the subject of attraction, there was great variation among pedophiles. For example, in some, attraction was based on personality characteristics of the child whereas others were attracted to physical appearances or to both inner and outer characteristics. To summarize these findings, there is great differentiation among pedophiles sexual preference, and among the development, content, and etiology of this sexual interest in children. The second aim of this study was to provide insight into how pedophiles cope with sexual attraction to minors. Many pedophiles reported having difficulties with acknowledging these pedophilic attractions. Moreover, on average, these pedophiles rated their lives with a 6.3 on a scale of zero to ten, (the average happiness score in The Netherlands was 7.9 in 2012; CBS, 2013), which indicates relatively low subjective wellbeing. In general, two important issues in learning to cope with pedophilia need to be considered: (1) acceptance of having pedophilic feelings, and (2) finding appropriate ways to relief sexual arousal. To resolve these issues, pedophiles may need professional help or help from their social environment to stimulate disclosure and learn to separate their feelings from abusive behaviour. Especially, disclosing sexual feelings to non-pedophilic individuals can be valuable, because these individuals can provide social control and can challenge inappropriate cognitions. 109

111 In contrast, non-disclosure and engagement in secretive behaviour seems to stimulate child sexual abuse because of a lack of social control and subsequent belief that inappropriate behaviour has little consequences. Indeed, the importance of social control in the inhibition of criminal behaviour is already recognized years ago by Hirschi s Social Control Theory (1969). This theory explains engagement in criminal behaviour due by a lack of commitment to others, institutes and rules and laws what can prevent individuals from doing wrong (Pratt, Gau & Franklin, 2011). Also, in diminishing fixation and stimulation feelings of being able to live a meaningful life, it seems important for pedophiles to have healthy social contact with children in the form of for example coaching roles, or friendships. Yet, although this study suggests that such social contact seems to diminish child sexual offending, there is no objective measure to determine who is able to control his or her sexual feelings and who is not, therefore keeping children safe should be the main focus. Third, we examined how these non-forensic and non-incarcerated pedophiles relieved their sexual arousal and found that most (but not all) pedophiles acknowledged to have engaged in child pornography offending and/or child sexual abuse at some point during their life. This is in line with earlier research on selfidentified pedophiles and hebephiles (Neutze et al., 2011). Yet, it is important to note that most pedophiles engaged in this behaviour around puberty when they were highly sexually fixated suggesting that adolescence is a critical time for offense behaviour in pedophiles. One reason for this may be that many pedophiles are still struggling in puberty with accepting their pedophilic interest and sexual identity and do not yet have the capacity to understand and appropriately control sexual arousal. This is in concordance with findings suggesting that in adolescence, individuals engage more in general impulsive behaviours because: (1) their neurocognitive development is still in progress (for example, because they do not yet have a fully developed prefrontal cortex), and (2) many neurophysiological changes happen around puberty (for example, changes in dopaminergic activity may lead to a higher need to engage in sensation seeking) (Steinberg, 2008). Indeed, many pedophiles stopped offending because they were better able to understand that their behaviour is damaging for victims and/or because they were less fixated on their sexual feelings. For example, they were better able to control their sexual interest in more appropriate ways such as through fantasizing and masturbation. Other ways to relief sexual arousal were for example, watching adult pornography, and engaging in sexual activity with adults. Also, one pedophile engaged in sexual activity with sex dolls looking like children. Notably, 4 pedophiles argued that although they never had engaged in sexual contacts with children, they had experienced situations in which children themselves where the initiator of sexual contact and tried to seduce adults to sexual play. A possible explanation for this may be that pedophiles may be more inclined to view situations involving children as sexual. Non-pedophilic individuals instead will probably not have these sexual association biases involving minors, and therefore may not report such experiences. Finally, 5 pedophiles committed child sexual abuse. These pedophiles seemed to differ in the severity of their sexual abusive acts but also in the degree to which they understood or dealt with the fact that this is considered criminal behaviour. Some used justifications for engaging in this abuse, such as the sexual activity happened out of romantic feelings or because of mutual sexual arousal to indicate that this behaviour was not acted out on the basis of egoistic motivations alone. These defense mechanisms are also recognized in earlier studies on pedophilia (Hall & Hall, 2007). Although initially these pedophiles did not comprehend that this behaviour was inappropriate, their justifications suggest that they seemed to have gained more insight into the lack of social acceptance of this behaviour. In contrast, 2 others engaged in child sexual abuse more secretly probably as a result of the awareness that their behaviour is wrong, and with a stronger emphasis on own sexual feelings. Furthermore, these pedophiles differed from others with regard to the manner in which they described their sexual activities. Both made inappropriate explicit sexual comments on multiple occasions, suggesting that next to pedophilia, they also experienced severe problems with their sexual self-regulation. Given the earlier described taxonomy of mapping risk factors into features related to criminal behaviour (continuum 1) and 110

112 sexual deviance (continuum 2) (Houtepen et al., 2014), it seems that these 2 pedophiles have problematic scores on both continua and have a high risk for re-offending. The current hunt on child sexual abusers by both the media and society can be regarded as more legitimate when harm is caused by antisocial oriented pedophiles. Our findings suggest that one disadvantage of this focus on pedophilic offenders is that it causes psychological difficulties such as feelings of shame, anxiety and depression in pedophiles who do not offend. Study 2: Pedophilia and risk factors for sexual offending In study 2, risk-factors of offending in non-forensic and non-incarcerated pedophiles were more closely examined. As mentioned earlier, numerous risk factors of child sexual offending exist, and offenders differ in the type and severity of risk factors they possess. Given that we explored self-identified pedophiles who are currently not involved in the legal justice system, our main focus was not on risk factors associated to general criminal behaviour but more on risk factors related to sexual deviance (Houtepen et al., 2014). One theory that clearly focuses on such risk factors for committing child sexual abuse is Marshall and Barbaree s (1990) Integrated Theory. This theory suggests that general adverse life events experienced during the life span, and especially those experienced early in life, such as own history of physical and sexual child abuse, predispose individuals to engage in child sexual offending. Amongst others, adverse life events may lead to insecure attachment, a failure to explore the world, inability to master a sense of personal security and trust in others, developing distorted internal working models of interpersonal relationships, and social and self-regulation difficulties. Furthermore, insecure attachment may lead to enduring socio-emotional problems such as poor coping, mood problems, and low self-esteem. Finally, the interaction of adverse life events with situational factors such as stress, strong negative affect, and intoxication may increase chances of committing sex offenses (Ward, 2002; Ward & Beech, 2006). In an attempt to capture the complexity and the interrelatedness of risk factors for child sexual offending, we compared self-identified pedophiles with a matched control group on some of the key variables identified in Marshall and Barbaree s (1990) Integrated Theory. Key factors were history of abuse, general negative life events (both before the age of 16 and throughout their whole lives), attachment/intimacy deficits, self-esteem, and self-control/impulsivity. This study is explorative in nature, but given that we examined self-identified pedophiles, we expected that pedophiles score more problematic on these key factors related to sexual deviance but not on risk factors associated with general criminal behaviour. Hence, we hypothesized that pedophiles score more problematic on history of abuse, life events, intimacy deficits, attachment and self-esteem than the control group. We did not expect any differences in self-control. Method Participants The sample comprised the same 15 self-identified males with pedophilic interests as in Study 1. The control group consisted of 62 Dutch males (M age=40.4, SD=11.7; age range = 22 68) recruited from the general population and matched to the group of pedophiles in age, marital status, and educational level. Pedophiles had no missing data on any of the variables. In the control group, one male had missing data on the life events scale (i.e., divorce of parents or stepparents), and one male had missing data on the selfesteem questionnaire. Also, one male from the control group had missing information on one of the attachment scales (i.e., fearful attachment), one male had missing information on self-control, and one had missing information on educational level. These individuals were excluded from the analyses concerning these variables. Groups did not differ with regard to the matched characteristics, but pedophiles had more mental health difficulties than the control group (i.e., 46.7% compared to 4.8%, χ 2 (2, N=77) = 19.34, p<.001). Procedure 111

113 Pedophiles. After the interviews in Study 1, pedophiles were asked to fill out a questionnaire to assess risk factors for offending, including history of abuse, number of experienced life events both before the age of 16 and throughout their whole lives, attachment/intimacy deficits, self-esteem, and self-control/impulsivity. Control group. A matched control group of 62 Dutch males was recruited from the general population on the basis of their age, marital status and educational level by under-graduates in Psychology and Health. Participants in the control group were informed that they participated in a study on personality characteristics and social relationships. In-depth interviews were not administered in the control group, but controls received the same questionnaires as the pedophile group. Measures History of abuse. History of abuse was assessed using the Negative Life Events Scale (NLES) (Garnefski & Kraaij, 2001). This is a 28-item self-report checklist that assesses negative life events concerning the individual or close (step)family members, including divorce, physical illness, death, mental health issues, history of selfharm, violence within the family, being a victim of a crime, being a victim of a horrible accident, experienced alcohol or drug misuse, and experienced abuse, in three periods of life (i.e., before the age of 16, between age 16 and 1 year before assessment, and within the last year). To assess history of abuse, we counted whether participants had experienced physical abuse or neglect, or sexual abuse before the age of 16. Negative life events. Negative life events were also assessed using the 28-item Negative Life Events Scale (NLES) (Garnefski & Kraaij, 2001). Specifically, we counted the total number of experienced life events before the age of 16, and the total number of experienced life events throughout participants lives. Attachment/intimacy. Participants attachment styles and subsequent intimacy difficulties were assessed using the Attachment Style Questionnaire (ASQ) (Van Oudenhoven & Hofstra, 2004). This questionnaire consists of 24 items. Participants indicated on a five-point scale (1 = strongly disagree, 2 = partly disagree, 3 = don t disagree/don t agree, 4 = partly agree, 5 = strongly agree) how much they agreed with statements such as: I find it relatively easy to become close to others. By averaging across items, we computed four scales indicating secure (7 items; α = 0.66), avoidant (5 items; α = 0.64), preoccupied (7 items; α = 0.79), and fearful attachment style (5 items; α = 0.83). These four attachment styles are based on two dimensions: the way people think about themselves, and the way people think about others, and were shown to have satisfactory construct validity (Van Oudenhoven & Hostra, 2004). The secure attachment style characterizes individuals who feel secure about both themselves and other people. Individuals with an avoidant attachment style feel secure about themselves but do not trust others and therefore try to avoid intimacy and social contact. In contrast, preoccupied individuals strive for intimacy, but feel insecure about their own competencies and likeableness. Finally, the fearful attachment style characterizes individuals that are both doubtful about self and other people. These individuals avoid being intimate with others because of a fear to get hurt. Self-esteem. The Rosenberg self-esteem scale (Rosenberg, 1965) was used in order to measure selfesteem. This is a 10-item questionnaire (α = 0.86), that includes items such as: I feel that I am a person of worth, at least on an equal plane with others. Participants answered each item on a four-point scale (3 = strongly agree, 2 = agree, 1 = disagree, 0 = strongly disagree) to indicate their level of agreement with the statements. Participants total scores were computed by summing their scores on the 10 items. Higher scores indicate higher levels of self-esteem. Self-control. The subscale measuring behavioural inhibition (BIS) of the Behavioural Inhibition System/Behavioural Activation System (BIS/BAS) Scales (Carver & White, 1994) was used in order to assess self-control/impulsivity. This measurement consists of 7 items (α = 0.73) (Dutch version; Franken, Muris & Rassin, 2005), and includes questions such as: I feel worried when I think I have done poorly at something important. 112

114 Participants answered on a four-point scale (1 = untrue, 2 = somewhat untrue, 3 = somewhat true, 4 = true) to indicate how much they agreed with these statements. Mean total scores were computed by averaging the scores across these 7 items. Higher scores on this measurement indicate inhibition of behaviours with negative outcomes, whereas lower scores indicate lack of self-control/impulsivity. Results and conclusion With regard to differences on key factors for offending, we found some support for our hypothesis that non-forensic and non-incarcerated pedophiles are higher on some factors related to sexual deviance than the control group (Table 1). Specifically, pedophiles were more likely to have a history of abuse (physical, emotional or sexual) compared to the control group (χ 2 (1, N=77) = 10.53, p<.01). Moreover, pedophiles experienced more negative life events in general before the age of 16 (t(75) = -2.27, p<.05, d = 0.76). These early negative life events may have predisposed these pedophiles to the development of a number of personal and interpersonal difficulties, for example having mental health issues related to feelings of anxiety and depression that have contributed to the development of pedophilic feelings and behaviours. In contrast to our hypothesis, pedophiles did not differ on attachment styles. Although they did score more problematic on all three of the insecure attachment styles compared to controls, these effects were not significant. Pedophiles also did not differ in self-esteem from the control group. Furthermore, pedophiles were on average more behaviourally inhibited (i.e., less impulsive) than the control group (t(74) = -3.49, p<.01, d = 0.98). A possible explanation for the finding that self-identified pedophiles were more behaviourally inhibited than the control group could be that these pedophiles need to regulate their (sexual) behaviour to a greater extent to avoid committing an offense. Hence, they may have learned to be better self-controlled in general, hereby lowering their risks to engage in sexual offense behaviour. General discussion Results of both studies suggest large variability in sexual interest and development in pedophiles. For most pedophiles, discovering and accepting their deviant sexual interest was difficult and help and/or support are needed to find appropriate ways to cope with these feelings. Moreover, many pedophiles outside clinical and judicial settings seem to be vulnerable individuals with issues related to pedophilic interest, instead of criminally oriented individuals who engage in child sexual abusive behaviour. This image is also supported by our empirical findings. Pedophiles more often experienced mental health issues (i.e., feelings of anxiety and depression) than matched males from the general population. Moreover, pedophiles were more likely to have a history of child sexual abuse and had experienced more negative life events in general before the age of 16. However, in contrast to Marshall and Barbaree s (1990) Integrated Theory and counter to our expectations, this did not result in problems such as poor attachment and low self-esteem in these self-identified pedophiles. Future research with longitudinal data is warranted in order to examine the link between early childhood trauma, attachment style, and self-esteem more explicitly. Moreover, in contrast to previous studies suggesting that child abusers in the legal justice system are low on behavioural inhibition (e.g., Neutze, Seto, Schaefer, Mundt, & Beier, 2011), pedophiles in our research were more behaviourally inhibited than the controls. This suggests that they may better be capable of refraining from criminal sex offences. However, it should be noted that (extremely) high levels of behavioural inhibition may also be considered as having poor (i.e., rigid) self-regulation skills and can have detrimental consequences for mental health. For example, studies have shown that high behavioural inhibition in childhood has been associated with symptoms of anxiety and depression in adulthood (Gladstone, Parker, Mitchell, Wilhelm & Malhi, 2005; Hirshfeld-Becker et al., 2008). Our findings raise the question to what extent early negative life experiences are related to behavioural inhibition and mental health issues in pedophiles, and in what way expression of such key factors contribute to the development of having pedophilic feelings in general. It has been argued that early 113

115 childhood trauma requires individuals to make use of a considerable amount of behavioural inhibition when individuals are not able to talk to others about pain- or shameful experiences such as child abuse (Pennebaker & Beall, 1986). Even in cases where children disclose their feelings after experiencing abuse, they may still adopt a behaviourally inhibited coping style to protect themselves from negative influences and experiences in their environment. In this view, the pedophiles in our sample may have learned to be more behaviourally inhibited because of their negative experiences early in life. It is tempting to speculate that the association between pedophilia and behavioural inhibition is bidirectional. On the one hand, being behaviourally inhibited may induce a preference to social contact with youngsters over contacts with adults because such contact is perceived as provoking less anxiety and judgment and therefore is evaluated more positively. Indeed, some of the pedophiles acknowledged that they felt more at ease in social relationships with children than with adults because of these reasons. Similarly, the inhibition to engage in adult relationships is also a risk factor for conducting child sexual abuse (Bogaerts et al., 2010; Cohen & Galynker, 2009; Cohen, Grebchenko, Steinfeld, Frenda & Galynker, 2008). On the other hand, the need to keep pedophilic feelings to oneself (and potential efforts to refrain from offending or to remain undetected by police), may in turn strengthen inhibition. These effects together will likely lead to problems such as anxiety and depression in those who have troubles with experiencing pedophilic feelings. Indeed, the association between stressful life events and anxiety and mood related disorders is well supported by previous research (Brown, Juster, Heimberg & Winning, 1998; Newmann & Bland, 1994). Testing the directions of these associations warrants further research and currently leaves room for additional explanations. For example, regarding the link between early negative life events and pedophilia, it can be argued that pedophiles have experienced more negative life events because of their sexual orientation (and deviant excesses because of this), such as having early sexual contact with other children that may provoke negative reactions from the environment. Therefore, it should be kept in mind that these risk factors may also be the consequence of a pedophilic sexual orientation. Finally, during the interviews, most pedophiles reported having engaged in sex offenses with children at some point during their life, either by downloading child pornography or by engaging in child sexual abuse. Hence, in order to prevent pedophiles from committing on- and offline sexual offense behaviours, there is a need for sufficient professional help in order to support individuals in accepting their pedophilic interest and coping with their sexual needs in legal ways. Our study suggests in line with previous research (Seto, 2012) that such help is currently lacking, especially in regular (i.e., non-forensic) mental health sectors. To better educate mental health professionals more attention should be given to pedophilia in general, and the treatment of pedophiles who are struggling with their sexual feelings. Furthermore, risk factors such as experiences of early negative life events and mental health issues related to feelings of anxiety and depression should be identified and treated as early as possible. Specifically, it is important to provide individuals with the necessary confidence in both themselves and other people to actively engage in interpersonal relationships and to be able to talk openly to others. Creating more openness about pedophilia as a sexual preference could be a step in the right direction. Our interviews suggest that openness could prevent child sexual abuse, because it enhances feelings of acceptance in pedophiles and enables social control from the environment. In contrast, secretive behaviour seems to promote sexual offense behaviours. As Goode (2010) notes, society should react better to adult sexual attraction to children by recognizing its existence, and by keeping children safe through allowing for open communication regarding the subject of pedophilia. Brochures discussing pedophilic feelings at offices of physicians, sexologists and clinical (also non-forensic) psychologists can be helpful in reaching those individuals who are currently struggling with pedophilic feelings but who do not have engaged in sexual behaviour with children (yet). Moreover, information on pedophilia could be given in general sex education to youth who are developing a sense of their own sexuality. Providing youth with the nuanced information that is often lacking in the media 114

116 and society can give them the right tools to cope with their sexual feelings in appropriate ways and seek help when needed. Our findings should be interpreted against the backdrop of several limitations. First, our sample is not an adequate representation of the whole population of pedophiles in the Netherlands. Because of stigma and fear of the consequences of disclosing their sexual interest, only 15 pedophiles wanted to participate in our study. Therefore, and because of the explorative nature of this study, we should be careful in generalizing our conclusions to the whole population of non-forensic and non-incarcerated pedophiles. Furthermore, our sample consisted mainly of highly educated individuals, and this may have confounded our results. Finally, we recruited our sample via online requests on forums where pedophiles are active. As a result, participants were already interested in discussing this topic and were able to critically reflect on their own and others thinking, feelings, and behaviours. Another limitation is that we only self-reported information is studied what can cause under-reporting of deviant behaviours (Buschman et al., 2010). In addition, when filling out the questionnaire, many pedophiles stated that if we had asked some of the questions when they were younger or had not received help yet, they would have filled out many questions differently. Therefore, we must keep in mind that the research group already gave much attention to their feelings and behaviours in their life. This could also explain the absence of differences between pedophiles and the control group in attachment style and selfesteem. Future research should invest in studying directions and links between risk factors for offending in nonforensic pedophiles, including negative life events such as abuse in early life, and mental health issues related to anxiety and depression. Also, more research is needed on treatment programs for pedophiles struggling with their feelings that may help to support them in accepting their pedophilic interest and providing them with tools to be able to relieve sexual arousal in non-harmful ways. Finally, a broader debate and acceptance of deviant sexual preferences by society may be warranted. If not, individuals with pedophilic preferences remain standing at the edge of society waiting for self-regulation to fail. References American Psychological Association. (2013). Diagnostic and statistical manual of mental disorders, fifth edition: DSM-5. Washington DC: Author. Blanchard, R., Christensen, B. K., Strong, S. M., Cantor, J. M., Kuban, M. E., Klassen, P.,... Blak, T. (2002). Retrospective self-reports of childhood accidents causing unconsciousness in phallometrically diagnosed pedophiles. Archives of Sexual Behaviour, 31(6), Blanchard, R., Lykins, A. D., Wherrett, D., Kuban, M. E., Cantor, J. M., Blak, T.,... Klassen, P. E. (2009). Pedophilia, hebephilia, and the DSM-V. Archives of Sexual Behaviour, 38(3), doi: /s Bogaerts, S., Buschman, J., Kunst, M., & Winkel, F. (2010). Intra-and extra-familial child molestation as pathways building on parental and relational deficits and personality disorders. International Journal of Offender Therapy and Comparative Criminology, 54(4), doi: / X Brown, E. J., Juster, H. R., Heimberg, R. G., & Winning, C. D. (1998). Stressful life events and personality styles: Relation to impairment and treatment outcome in patients with social phobia. Journal of Anxiety Disorders, 12(3), doi: /S (98) Buschman, J., Bogaerts, S., Foulger, S., Wilcox, D., Sosnowski, D., & Cushman, B. (2010). Sexual history disclosure polygraph examinations with cybercrime offences: A first Dutch explorative study. International Journal of Offender Therapy and Comparative Criminology, 54(3), doi: / X

117 Capra, G., Forresi, B., & Caffo, E. (2014). Current scientific research on paedophilia: A review. Journal of Psychopathology, 20, Retrieved from: Capra.pdf Carver, C. S., & White, T. L. (1994). Behavioural inhibition, behavioural activation, and affective response to impending reward and punishment: The BIS/BAS Scales. Journal of Personality and Social Psychology, 67(2), 319. doi: / Centraal Bureau voor de Statistiek (2013). NL-cijfers: Domeinen van duurzaamheid [Dutch-numbers: Domains of durability]. Retrieved from: Cohen, L.J., & Galynker, I.I. (2009). Psychopathology and personality traits of pedophiles: Issues for diagnosis and treatment. Psychiatric Times, 26, Retrieved from: Cohen, L. J., & Galynker, I. (2012). Identifying psychological traits potentially subserving aberrant motivation or inhibitory failure in pedophilic behaviour. Israel Journal of Psychiatry and Related Sciences, 49(4), Retrieved from: doctorsonly.co.il/wp-content/uploads/2013/03/07_- Identifying-psychol-traits.pdf Cohen, L.J., Grebchenko, Y.F., Steinfeld, M., Frenda, S.J., & Galynker, I. (2008). Comparison of personality traits in pedophiles, abstinent opiate addicts, and healthy controls: considering pedophilia as an addictive behaviour. Journal of Nervous and Mental Disease, 196, doi: /NMD.0b013e31818b4e3d Franken, I. H., Muris, P., & Rassin, E. (2005). Psychometric properties of the Dutch BIS/BAS scales. Journal of Psychopathology and Behavioural Assessment, 27(1), doi: /s Freimond, C. M. (2013). Navigating the Stigma of Pedophilia: The Experiences of Nine Minor-Attracted Men in Canada. Arts & Social Sciences: Department of Sociology and Anthropology. Retrieved from Freund, K., Watson, R., & Dickey, R. (1990). Does sexual abuse in childhood cause pedophilia: An exploratory study. Archives of Sexual Behaviour, 19(6), doi: /BF Garnefski, N., & Kraaij, V. (2001). Negative Life Events Scale (Levensgebeurtenissen Vragenlijst). Retrieved from Gladstone, G. L., Parker, G. B., Mitchell, P. B., Wilhelm, K. A., & Malhi, G. S. (2005). Relationship between selfreported childhood behavioural inhibition and lifetime anxiety disorders in a clinical sample. Depression and Anxiety, 22(3), doi: /da Goode, S. D. (2010). Understanding and addressing adult sexual attraction to children: A study of paedophiles in contemporary society. Abingdon: Routledge. Hall, R. C., & Hall, R. C. (2009). A profile of pedophilia: Definition, characteristics of offenders, recidivism, treatment outcomes, and forensic issues. FOCUS: The Journal of Lifelong Learning in Psychiatry, 7(4), Retrieved from: /article.aspx?articleid=53036 Henry, O., Mandeville-Norden, R., Hayes, E., & Egan, V. (2010). Do internet-based sexual offenders reduce to normal, inadequate and deviant groups? Journal of Sexual Aggression, 16(1), doi: / Hirshfeld Becker, D. R., Micco, J., Henin, A., Bloomfield, A., Biederman, J., & Rosenbaum, J. (2008). Behavioural inhibition. Depression and Anxiety, 25(4), doi: /da

118 Hossack, A., Playle, S., Spencer, A., & Carey, A. (2004). Helpline: Accessible help inviting active or potential paedophiles. Journal of Sexual Aggression, 10(1), doi: / Houtepen, J.A.B.M., Sijtsema, J.J., & Bogaerts, S. (2014). From child pornography offending to child sexual abuse: A review of child pornography offender characteristics and risks for cross-over. Aggression and Violent Behaviour, 19(5), doi: /j.avb Marshall, W. L. and Barbaree, H. E. (1990). An integrated theory of the etiology of sexualoffending. In W. L. Marshall, D. R. Laws and H. E. Barbaree (Eds.), Handbook of sexual assault: Issues, theories, and treatment of the offender (pp ). New York: Plenum. Middleton, D., Elliott, I. A., Mandeville-Norden, R., & Beech, A. R. (2006). An investigation into the applicability of the Ward and Siegert Pathways Model of child sexual abuse with Internet offenders. Psychology, Crime & Law, 12(6), doi: / Neutze, J., Seto, M. C., Schaefer, G. A., Mundt, I. A., & Beier, K. M. (2011). Predictors of child pornography offenses and child sexual abuse in a community sample of pedophiles and hebephiles. Sexual Abuse, 23(2), doi: / Pennebaker, J.W., & Beall, S.K. (1986). Confronting a traumatic event: Toward an understanding inhibition and disease. Journal of Abnormal Psychology, 95(3), doi: / X of Pratt, T.C., Gau, J.M. and Franklin, T.W. (2011). Key idea: Hirschi s Social Bond/Social Control Theory. In T.C. Pratt, J.M. Gau, and T.W. Franklin (Eds.), Key ideas in criminology andcriminal justice (pp ). Thousand Oaks, CA: Sage publications. Quayle, E., & Taylor, M. (2002). Child pornography and the Internet: Perpetuating a cycle of abuse. Deviant Behaviour, 23(4), doi: / Rosenburg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University. Seto, M. C. (2004). Pedophilia and sexual offenses against children. Annual Review of Sex Research, 15(1), doi: / Seto, M. C. (2012). Is pedophilia a sexual orientation? Archives of Sexual Behaviour, 41(1), doi: /s Steinberg, L. (2008). A social neuroscience perspective on adolescent risk-taking. Developmental 28(1), doi: Review, Van Oudenhoven, J. P., & Hofstra, J. (2004). De Hechtingsstijllijst (HSL) Handleiding. [The Attachment Style Questionnaire (ASQ): Manual]: Groningen: Rijksuniversiteit. Ward, T. (2002). Marshall and Barbaree's integrated theory of child sexual abuse: A critique. Psychology, Crime and Law, 8(3), doi: / Ward, T., & Beech, A. (2006). An integrated theory of sexual offending. Aggression and Violent Behaviour, 11(1), doi: Webb, L., Craissati, J., & Keen, S. (2007). Characteristics of Internet child pornography offenders: A comparison with child molesters. Sexual Abuse, 19(4), doi: /s

119 Wurtele, S. K., Simons, D., & Moreno, T. (2013). Sexual interest in children among an online sample of men and women: Prevalence and correlates. Sexual Abuse. Advance online publication. doi: / Table 1. Risk-factor comparisons between pedophiles and control group Variables Pedophiles (n=15) Control group (n= 62) History of abuse n (%) Physical/emotional abuse Sexual abuse 5 (33.3) 5 (33.3) 1 (6.7) 3 (4.8)** 1 (1.6) 2 (3.2) Mean (SD) Mean (SD) Negative life events before (1.9) 0.6 (1.0)* Negative life events 3.2 (3.0) 2.3 (1.8) Attachment Secure Avoidant Preoccupied Fearful style 26.7 (4.5) 17.2 (4.8) 19.7 (6.9) 13.6 (5.3) 26.6 (3.9) 16.9 (3.4) 16.4 (4.7) 11.8 (4.2) Self-esteem 22.1 (6.1) 22.9 (4.3) Self-control 2.7 (0.5) 2.2 (0.5)** SD: standard deviation *p<0.05, **p<0.01 Cyberpsychology & Behaviour, 8,

120 Training Tool for Police and Aid Workers Risks and Vulnerabilities in offences with Pedophilia and Child Pornography and child pornography offending Jenny Houtepen, Jelle Sijtsema & Stefan Bogaerts Tilburg University, B-CCENTRE DSM-IV-TR: Recurrent, intense, sexually arousing fantasies, sexual urges, or behaviors involving sexual activity with prepubescent children, for at least six months 3-9% of men in general population have some sexual attraction to children (Wurtele, Simons, & Moreno, 2013) Pedophilia as a sexual orientation? (Seto, 2012) 119

121 factors for E.g.: Impulsivity Antisocial behavior Empathy Cognitive distortions History own abuse Poor attachment, intimacy problems Self esteem Loneliness Criminal orientation Sexual deviance/ Fantasy non-forensic pedophile? Sexual deviance/ Fantasy Criminal orientation 120

122 Many have sexual contact with minors or download CP (child pornography) (Neutze et al., 2011) Psychological assistance? How does pedophilic attraction develop? What does sexual attraction to children look like? How does it feel to discover pedophilic feelings? How are they dealing with forbidden feelings and stigma? How are they relieving sexual arousal? Risk- and protective factors to engage in sexual offending? Procedure Online requests on pedophile websites, s to pedophiles in self-help groups Inclusion criteria: everyone who recognizes himself in term pedophile and is not incarcerated or hospitalized in forensic institution Anonymity vs. safety Method In-depth interviews 121

123 Sexual development as a gradual process, starting in early puberty (73.3%) Stable sexual preference 66.7% also romantic feelings or fantasies: importance of social component >50% attracted to both physical and behavioral characteristics: e.g., their beauty, bodily shapes, openness, spontaneity, honesty, or naivety 122

124 Conditional theories: association early sexual experiences and pedophilic feelings >50% early sexual experiences (6-11 years old): masturbation (1), playing doctor (4), sexual play with others (1), actual penetration (2) Respondent: This certainly has made me think about sexuality too early, and I guess I was fixated on this for quite a while Association early sexual experiences and cognitions regarding children and sex? Also pedophiles with normal sexual development! Most pedophiles struggle with realization: - Fear of being stigmatized as monster or child rapist - Shame - Denial - Fear of reactions environment Some neutral, e.g., because of different attitude society Respondent (1): I think of it as something honorable I have always liked going against the rules. 123

125 40% help from professional (physician, psychologist of sexologist) separate feelings from behavior & talking about feelings 20% help from other pedophiles (+20% enjoy contact with like-minded others) e.g., recognition, being yourself, other non-forensic pedophiles 73,3% help from non-pedophiles e.g., being yourself, social control, evaluate cognitions on a normal level Risks of secrecy and secretive behavior! 73.3% watched CP (3 convicted): 72.7% stopped 20% watched legal CP 33.3% received CP from others 1 consciously shared with familiar others 1 charged (not convicted) for production 1 watched CP together with minors 124

126 Material in news groups, peer-to-peer networks, regular search engines (Google) Some received material from others or watched together with other people Material ranged from virtual CP to explicit sexual abuse 60% ideas about justified material overlap cognitions regarding children s sexuality? Justified material overlap with cognitions about children s sexuality? Cognitions Children s developing sexuality, experimenting Sexuality equal to adults I don t know, but it doesn t matter anyway (e.g., I would never try to involve children in sexual behavior) Material Sexual activities between children, masturbation Sexual activities children with adults Virtual CP 125

127 30% no noticeable influence 46.7% regularly in need of new or more severe material Respondent: I noticed my cognitions getting more sexualized 1 downloading as collective addiction, not for sexual desire But no cross-over effects! More than 50% no fear of getting caught E.g., illusion that behavior isn t wrong, in the past less focus on CP downloaders & consequences, habituation Little use of safety measures: - None (3) - Door closed, hiding and deleting material (7) - TOR-networks to mask IP-address (2) 126

128 Realization that children need to be abuse in order to produce material Fear that behavior gets out of hand Growing fear and realization that benefits do not outweigh harm/disadvantages Social contact with children in real world Mixed feelings: I felt sick and was disgusted with what I saw, but at the same time I was frightened by my own excitement 5 Pedophiles have had sexual contact with minors None of them use term abuse 1 convicted Difference in egocentric description sexual activities & hiding behavior Generally: little understanding of abuse, 3 justifying behavior 127

129 Sexual fantasies Watching adult pornography Sexual activities with adult partner Sexual feelings less important because of age Sexual activities with Japanese child doll (1) Pedophilic feelings just as varied as sexual feelings hetero- and homosexuals Nature or nurture? Many struggled with 1. accepting and 2. dealing with pedophilic feelings need for counseling Talk about feelings with (non-pedophilic) others seems valuable Puberty critical age for sexual delinquency in pedophiles? 128

130 Prevention of abuse in some minors through: Open discussion of pedophilia Information in general sex education Information brochures at offices of physicians, psychologists and sexologists Specialized assistance for pedophiles to cope with their sexual feelings Questions or comments? 129

131 In collaboration with our academic partners