Sexual Interests of Child Sexual Exploitation Material (CSEM) Consumers: Four Patterns of Severity Over Time

Abstract

This study analyzed the evolution over time of the activity of consumers of child sexual exploitation material (CSEM). To this end, images and metadata were extracted from the hard drives of 40 individuals convicted of possession of child pornography and analyzed. A sample of these images (N = 61,244) was categorized by the age of the subjects depicted and—using the Combating Paedophile Information Networks in Europe (COPINE) scale—by severity of the acts depicted. Collecting activity was observed to follow four patterns. The most prevalent pattern was a progressive decrease in the age of the person depicted and a progressive increase in the severity of the sexual acts. In light of the results, we propose four explanations of the nature of, and variations in, child-pornography collections.

Keywords

child sexual exploitation material (CSEM)child pornography severity illegal images sexual violence

Introduction

The advent of the Internet has increased the availability of child sexual exploitation material (CSEM). As recently as the early 1980s, obtaining CSEM still depended on making contact with specific individuals or visiting specific locations (sex shops, video clubs). These contacts were very risky affairs, and the preservation of anonymity was a preoccupation of all concerned (Sellier, 2003). The difficulty in making contact with other consumers of CSEM, the rarity of CSEM, and the risks inherent in exchanging child pornography allowed law enforcement agencies to limit this type of illegal trafficking (Wortley & Smallbone, 2006). Hence, the technical specifications of an online world have transformed how CSEM is exchanged. It has been estimated that 2% to 4% of all men have consumed this kind of material using various virtual tools (Seto, 2013). As early as 1997, researchers for the Combating Paedophile Information Networks in Europe (COPINE) project studied chat rooms (Taylor & Quayle, 2003), and it has been reported that two thirds of individuals arrested for the possession and distribution of CSEM had exchanged their material in chat rooms (Carr, 2004; Roy, 2004). Newsgroups have also been reported to be important exchange hubs for this material (Carr, 2001; Fortin, 2013; Sellier, 2003; Taylor, 1999). More recently, researchers have demonstrated the importance of peer to peer (p2p) exchange networks (Wolak, Liberatore, & Levine, 2014).

Studies that have compared adult male CSEM consumers with adult male contact sex offenders have reported the former to be younger, more frequently in a marital relation, more educated, more intelligent, and less likely to be unemployed (Babchishin, Hanson, & Hermann, 2011; Babchishin, Hanson, & VanZuylen, 2015). Furthermore, CSEM consumers present fewer mental health and substance abuse problems. In addition, they reported having had fewer difficulties in childhood and had committed fewer criminal offences. In general, CSEM consumers seem to have more psychological constraints to committing sexual contact offences than hands-on offenders; they tend to have less pronounced antisocial tendencies and more self-control (Babchishin et al., 2015). A better understanding of the characteristics of CSEM collectors is important as they now account for a significant portion of the caseload of professionals working with sex offenders (Middleton, Mandeville-Norden, & Hayes, 2009).

Researchers have adopted two methodological strategies to assess the type of content exchanged by CSEM consumers on the Internet. Analyses of CSEM images have either included (a) collectors’ hard disk content following arrest or (b) samples of the images available on the Internet in a given period, obtained by searching directly in virtual spaces. In the collections of CSEM that have been analyzed, images of female subjects have been predominant (A. Carr, 2004; Roy, 2004; Wolak, Finkelhor, & Mitchell, 2005, 2011); for example, the children depicted in police image databases was following a girl/boy ratio of 4 to 1 (Quayle & Jones, 2011). The images in the collections generally fall into three categories: (a) sexually explicit images (found in 90%-94% of the collection; A. Carr, 2004; Roy, 2004; Wolak et al., 2005, 2011); (b) images of sexual contact between adults and children (found in 71%-90%; A. Carr, 2004; Wolak et al., 2005, 2011); or (c) nonexplicit images of nude or seminude subjects (found in 79%-82%; Wolak et al., 2005, 2011). It appears that most child-pornography collections include a wide range of content. In general, collections are primarily composed of prepubescent or pubescent girls (Quayle & Jones, 2011), with children aged 6 to 12 years generally forming the largest age category (83%-86%; Wolak et al., 2005, 2011). By comparing their results with those from a previous study (Wolak et al., 2005), Wolak et al. (2011) observed a significant increase in the number of collectors possessing images depicting children younger than 3 years old and of collectors possessing only images of children younger than 12 years old.

Several explanations may be suggested to clarify the nature of images men are collecting. One of them is that the content of the collection may reflect the offender’s sexual interest. In this regard, Glasgow (2010) has suggested that the analysis of pornography collectors’ hard drives would significantly improve the identification of their sexual interests, and that, specifically, “The pattern of images downloaded over time reflects evolving sexual interests, escalation of instrumental behaviour, growing compulsivity and possibly psychopathology of personality traits” (Glasgow, 2010). However, he notes that there is no quantitative evidence to support this hypothesis.

The testimony of arrested pornography collectors suggests that the content of collections may become more extreme over time, which indicates a progressive transformation of sexual interest (Quayle & Taylor, 2002; Roy, 2004). This transformation is thought to reflect excitatory habituation (Reifler, Howard, Lipton, Liptzin, & Widmann, 1971), or what some authors have termed the fantasy excitation effect (Sheehan & Sullivan, 2010). The reported changes in the nature of the collections are due to habituation over time; collectors may find the same type of content boring, and seek out more extreme types. Some authors have suggested that “this conditioned response might shift in a more deviant direction as the subject is looking for example at more violent pictures” (Laws & Marshall, 1990). Thus, some researchers have suggested that habituation may leads consumers to seek pornographic images depicting younger children and more brutal acts (Beech, Elliott, Birgden, & Findlater, 2008; Oosterbaan, 2009; Palmer, 2005).

In addition to the explanations for the increase in the severity of CSEM collections, there is also an explanation for a decrease in severity. In fact, for some consumers, the initiation of the CSEM collection may be motivated by curiosity (see Fortin & Corriveau, 2015; Krone, 2004; Taylor & Quayle, 2003). After an exploration phase, this type of offender may realize that he has no interest in sexual content involving children and, then, switch to adult content.

Aim of the Study

Although many studies of CSEM have analyzed the characteristics of collectors and the content of their collections (Babchishin et al., 2011; Babchishin et al., 2015; Elliot, Beech, & Mandeville-Norden, 2013; Seto, Wood, Babchishin, & Flynn, 2012), our study focused on the evolution of collecting behaviour over time, as reflected in the material collected. Because most studies have simply characterized the collection at the time of arrest, it appeared necessary to examine the way child pornography is collected. The patterns that will be identified will be a prerequisite to carrying out studies aiming at identifying the psychosocial factors associated with their evolution (increase and decrease in severity of the content).

This exploratory and descriptive study used innovative data sources: under the terms of an agreement with a Canadian police force, we had access to collections of CSEM. To our knowledge, no previous study has had access to comparable content. This type of data is generally unavailable to researchers, because in most industrialized countries, including Canada, accessing child-pornography sites and possessing child pornography is illegal. The difficulty gaining access to collections of child pornography is undoubtedly a major obstacle to research on this material. However, the data collected in this study allowed us to analyze the manner in which child-pornography collections are compiled, and how the severity of the collections’ images evolves. Analysis of hard disks has the advantage of eliminating biases related to narratives, perceptions, and avoidance strategies that would occur during a survey or an interview, as the disk acts like an airplane’s “black box,” recording all system activity. Thus, the aim of the current study was to analyze the patterns of evolution of child-pornography collectors’ collections.

Method

Cases

The initial sample was composed of collectors of CSEM who had been arrested and convicted for child-pornography offences in Quebec between 2004 and 2012 (N = 59). All offenders whose hard disk was available to the Sûreté du Québec (the provincial police force of Quebec) were selected. Only offenders whose case had been definitively disposed of (all the judicial procedures were completed) were selected. Sociodemographic data on the offenders at the time of arrest was obtained from police database records. As presented in Table 1, the offenders were all male, and their mean age at the time of arrest was 35.8 years (SD = 13.47). They came from a wide range of occupational backgrounds, of which the most common were trade/construction/manual labor (17.9%), business (8.9%), and computer-related (8.9%). As suggested in other studies (e.g., Babchishin et al., 2015), a majority of collectors (83.3%) had no prior offences. In our sample, 3.7% had prior sexual offences while 13.0% had nonsexual prior offences.

Table 1.

Offender Demographics.

	n	%
Age (N = 56)
18-24	14	25.0
25-29	12	21.4
30-34	5	8.9
35-44	10	17.9
45-54	11	19.6
55-64	3	5.4
65 and over	1	1.8
Employment by type (at time of arrest; N = 56)
Trade/Construction/Manual Labor	10	17.9
Business	5	8.9
Computer-related	5	8.9
Bar/restaurant	4	7.1
Student	4	7.1
Government/Government Contractor	4	7.1
Retirement	3	5.4
Mental Health/Counseling	2	3.6
Sales	2	3.6
Unemployed	2	3.6
Engineering/Medical	2	3.6
Education/Arts	2	3.6
Undeclared	11	19.6
Prior offences (N = 54)	n	%
Nonsexual	7	13.0
Sexual	2	3.7
No prior offences	45	83.3

Note. All cases but one were Caucasian (Other—but unknown).

Measures

To evaluate the content of the images, we use three measures. First, the severity of the images was rated according to the COPINE scale (Taylor, Holland, & Quayle, 2001). This scale ranges from 1 = nonerotic and nonsexualized pictures to 10 = sadistic/bestiality (the details on coding criteria are presented in the appendix). The typology was created by analyzing more than 80,000 images of CSEM collected from the material publicly available and that “represents a very large sample of the total amount of material in public circulation” (Taylor, Holland, & Quayle, 2001, p. 102). The COPINE project used both the descriptive analysis of the collection of images as well as the experiences of the project team to categorize material. The categorising system was based on the Platform for Internet Content Selection (PICS) and the Recreational Software Advisory Council (RSACi) rating system (Akdeniz, 1997), to focus on adult sexual interest in children (Taylor et al., 2001).

To favour consistency in the coding, each rater received training that included explanations of the criteria for each of the 10 levels of the COPINE scale. Results of intraclass correlation coefficients (ICC) using Two-Way Random-Effects Model ICC (2,1; McGraw & Wong, 1996) for COPINE scale indicate that the level of agreement between raters was good (Portney & Watkins, 2000) with an intraclass correlation of .81 for a single measure and of .93 for the average (sample of three raters with a sample of 200 images). A study has demonstrated that the scale has a high level of validity: results showed that participants showed high interrater agreement in their rankings (Spearman r = .951; Merdian, Thakker, Wilson, & Boer, 2013). In Canada, it can be estimated that pictures are illegal from level 3 to 4, as the Canadian criminal code states that the child has to be in a sexual context. It can be argued, as did Taylor et al. (2001), that our wider approach allowed us to consider all pictures that adults with a sexual interest in children found attractive, and not simply those legally defined as obscene.

Second, the age of the youngest child or children depicted was coded from 0 to 17, and individuals at least 18 years old were coded as adult. Infants younger than 1 year were coded as 0 for age. It is important to mention that the measurement of the age of the children depicted in the pictures represented an important challenge in the study. All sampled images were evaluated by raters who were trained by the first author. To favour consistency in the coding of the age, each rater received training that included Tanner’s (1990) description of the five stages of physical development of the breasts, genitals, and pubic hair. However, because Tanner’s stages are quite broad, the training also involved the presentation of pictures of children of known ages from 0 to 17 years. Following the training, each rater coded the first 100 images in the presence of the first author. To ensure consistency of the ratings, there was discussion between the first author and each rater. ICCs were calculated to measure interrater reliability. ICCs were calculated for the age scored by a subsample of raters (n = 3), using two-way random effects model ICC (2,1; McGraw & Wong, 1996). Results indicate that the level of agreement between raters was good (Portney & Watkins, 2000), with an ICC of .79 for a single measure and .92 for the average.

Third, the sex of the youngest child or children depicted was coded (boy = 0, girl = 1, both = 2). The sex of the child depicted was essentially determined by the visual evaluation of the genitals.

Procedure

The contents of the offenders’ hard disk were inventoried; as some offenders possessed more than one hard disk, a total of 112 hard disks were analyzed. Other media (compact disc read-only memory [CD-ROMs], universal serial bus [USB] keys, diskettes, printed photographs, etc.) were excluded, as police officers mainly use data from hard drives and may neglect data from other types of media available (CD-ROMs, USB keys, diskettes, printed photographs, etc.). Police officers consider the former strong evidence for trial. We also used the data from hard drives to ensure comparability between subjects. To extract these data from the hard disks, the Forensic Tool Kit (FTK) analytical software package was used. This software guarantees that the collection of data meets generally accepted standards of legal proceedings. These practices include ensuring the integrity of these data (by precluding voluntary or accidental modification of the digital evidence), the authenticity of the evidence, and the reproducibility of all procedures related to evidence collection and analysis (Mocas, 2004). Before starting the collecting phase of the study, we obtained approval (#CERAS-2011-12-255-A) from the ethical committee of the University of Montreal. The first author also signed an agreement that formalized the collaboration with the law enforcement agency.

Two key elements were extracted from every disk: the list of image files, which contains metadata for each file (name, size, creation date, modification date, etc.), and the image files themselves. Using the FTK program, a hash value was calculated for each file. Hash values are a form of digital signature, analogous to a fingerprint,¹ and are used by investigators to determine whether images are identical to one another (Roussev, Chen, Bourg, & Richard, 2006).² In this study, hash values were used to determine whether any hard-disk images were identical in all respects to images previously identified as relevant material (child pornography, but no other information on images such as COPINE scores) by the Sûreté du Québec. Once the files were inventoried and the content collected, FTK identified CSEM images whose digital signatures were recorded in reference databases. This comparison of new and known images is also practised by law enforcement agencies, to maximize efficiency and automatically categorize images.

Given the astronomical number of images, it was necessary to analyze samples of the 59 collections, rather than the entire collections. There were approximately 5 million images of all kinds in the hard drives analyzed, of which 175,014 were previously identified child-pornography images. To ensure the representativeness of the images sampled, sample size calculations with a correction for finite populations were performed. For example, a collection of 3,101 images required a sample of 794 images to obtain a 5% margin of error of 19 times out of 20.

A minimal collection period of 6 months was considered adequate for the identification of changes in the content of a collection. In fact, this period was selected to strike a balance between ensuring a sufficient number of measurements (one per month) to calculate a slope and the loss of cases associated with each month added. The application of this criterion reduced the number of collections from 59 to 40, with the number of months for the excluded collections distributed as follows: 1 to 2 months: N = 4; 3 to 4 months: N = 9; 5 months: N = 6. Possible explanations for the short duration of the excluded collections are that collectors had only a recent interest in CSEM, or preferred to use external storage, or were using a computer that was too recently acquired to reflect the temporal evolution of their collecting activities. We conducted t tests (independent samples test) to compare the excluded group (N = 19) with the sampled group (N = 40). Results showed that there were no significant differences between the groups with respect to images possessed by the cases (age and sex of children depicted in images, and COPINE scale). We also compared the groups for included sociodemographic variables, and there was no significant difference for age, employment type, or prior offences.

Hard-disk analysis is a recognized method of reconstructing computer-based crimes (Bang, Yoo, & Lee, 2011; G. Palmer, 2001). Many parameters are to be considered when assessing the inner workings of changes of dates recorded by the operating system (Bang et al., 2011). In our study, a monthly calendar of new images files was constructed for each case. To achieve this goal, we used the valid creation dates extracted by the FTK™ software.

More than 61,244 images were coded, for a mean of 1,038 images per individual (Mdn = 552, SD = 1,354.48). The images were coded by four graduate students and two police officers.³ The variables were recorded in a custom software package. Images were presented randomly to coders, who filled out the coding grid. Results showed that the mean on COPINE scales on all images was of 5.3 (student#1 = 5.1, student#2 = 4.9, student#3 = 6.1, student#4 = 5.7, officer#1 = 5.1, officer#2 = 5.0). We considered that mean COPINE scores differences were reasonably similar, and no adjustments were made.

Results

The mean duration of image collection for the selected group with at least 6 months of collection data (n = 40) was 23.1 months (Mdn = 19.5 months, range = 6-70 months, SD = 16.5). Overall, the mean age of the children depicted in the images was 9.9 years, with a median of 10 years and a mode of 7 years (see Table 2). The mean COPINE score was 5. Almost all the images (90%) were of girls.

Table 2.

Sample Characteristics.

	M	Median	Mode
Age	9.9	10	7
COPINE score	5.3	5	5
Sex	0.9	1	1

Note. N = 61,244; mixed-sex images were excluded for the calculation of the mean (0 = boys, 1 = girls). COPINE = Combating Paedophile Information Networks in Europe.

The general distribution of the frequency of the age of the subjects in the collections of CSEM followed a bell curve. Specifically, as presented in Table 3, the 6 to 12 years category was predominant in 64.4% of cases, and the adult category was predominant in 16.9%. The mean age of victims in collections in which images of boys predominated was greater than in collections in which girls predominated (13.06 years vs. 10.58). However, no statistical analyses were performed for mainly homosexual collections, as only six cases had such collections.

Table 3.

Category of Images, by Type of Subject Depicted (N = 59).

	Number of cases	% of cases
Sex of subject
Heterosexual (>66% girls)	49	83.1
Homosexual (>66% boys)	6	10.2
Mostly heterosexual (approx. 60% girls, 30% boys, 10% mixed-sex)	4	6.8
Predominant Age Group
0-5 years	2	3.4
6-12 years	38	64.4
13-15 years	8	13.6
16-17 years	1	1.7
Adult	10	16.9
Sexual Interest in Collection
Exclusive (>66% in one age category)	23	39.0
Nonexclusive (<66% in one age category)	36	61.0
Mean COPINE score, all images
2. Nudity	3	5.1
3. Eroticism	2	3.4
4. Posing	8	13.6
5. Erotic posing	23	39.0
6. Explicit erotic posing	17	28.8
7. Sexual acts involving children	6	10.2

Note. COPINE = Combating Paedophile Information Networks in Europe.

Cases had more frequent (defined as at least 66% of the collection) female images in 83.1% of cases and for male images in 10.2% of cases. A threshold of two thirds (66%) was used to increase the probability of correctly assessing cases’ interest for images of a specific sex; a lower threshold (e.g., 51%) may have caused classification error. For example, an individual who downloaded pictures of children of a nonpreferred sex out of curiosity, for trading purposes, or accidentally (in the case of large transfers of pornography files; see Seto & Eke, 2015), could have had his interest coded in a way that did not reflect his genuine sexual interest. This recoding strategy also highlighted the fact that for 6.8% (n = 4) of cases, the interest was unclear (e.g., 60% girls, 30% boys, 10% both sexes).

Interest for a specific age group (0-5 years, 6-12 years, 13-15 years, 16-17 years, adult) was analyzed, with interest defined as a single age group representing at least 66% of the collection. A clear interest was observed in 39.0% of cases. However, in 61.0% of cases, other age categories were also observed. The most popular combination of interests was for the 6 to 12 year and 13 to 15 year categories (n = 6).

Analysis of the mean COPINE score of each case’s collection revealed that 22.1% of the collections had a mean COPINE score of less than 5 (softcore pornography). The mean COPINE score of the remaining collections (almost 80%) was at least 5 (children in sexual or provocative contexts), with 10.2% of all collections exhibiting a mean COPINE score of at least 7 (children engaged in sexual activities such as masturbation, oral sex, sexual touching).

The evolution of the collections’ severity was analyzed in terms of two measures: age and COPINE score. To this end, the mean age and COPINE score were calculated for each month in a collection, and the slope of the two measures of severity was calculated over all the months in that collection. The possible range of the slope was −1 to +1. For example, if the mean age of the subjects depicted in the images was 17 years in the first month and decreased by 1 year per month, over a period of 10 months, until it reached 7 years at the end of the period, the slope would be −1. A slope of +1, however, indicates a constant increase in the age of the subjects depicted. Finally, a slope of 0 indicates no change in mean severity over time. To identify patterns in the evolution of the collections, the collections were allocated to one of four categories defined by the slopes of age and COPINE score for every case under study (increasing, decreasing; Table 4). Because there are no criteria for determining the slope that defines a pattern as stable, in this exploratory and descriptive study, a positive slope on COPINE and a negative slope on age were both considered to indicate an increase in severity. A negative slope on COPINE and a positive slope on age were both considered to indicate a decrease in severity. The selected cases were chosen to illustrate the typical pattern in each group, taking into account slopes.

Table 4.

Severity Patterns.

Pattern	COPINE trend	Mean COPINE slope	Age trend severity	Mean age slope	Number of images(Mean, range)	Collection duration(Mean, range)	N	%
Degenerating Spiral	Increase	0.11	Increase	−0.23	1,292 (276-4,528)	20.6 (7-70)	15	37.5
Sexualized Adolescent	Increase	0.14	Decrease	0.22	534 (91-1,186)	22.3 (12-45)	8	20
Boy/Girl-Love	Decrease	−0.05	Increase	−0.16	1,448 (609-2,907)	33.6 (6-60)	8	20
De-escalation	Decrease	−0.09	Decrease	0.17	1,057 (96-6,308)	18.4 (7-58)	9	22.5

Note. COPINE = Combating Paedophile Information Networks in Europe.

The longest collection period was 70 months, and the mean collection period was 23.05 months (SD = 16.5). As can be seen from Table 4, 37.5% of the collections exhibited increased severity in terms of both age and COPINE score: The children depicted became younger, and the acts became more extreme. In 22.5% of cases, the collections exhibited the inverse pattern: The children became older and the acts less extreme.

Degenerating Spiral Pattern

Figure 1 illustrates the evolution of Case 74, who belonged to the Degenerating Spiral pattern, which is the most prevalent (37.5% of the sample). The group had a mean of 1,292 sampled images and a mean of 20.6 months of collecting. The collections of these cases exhibited increasing slopes of severity for both age and COPINE score.

Figure 1.

Pattern 1. Degenerating Spiral (Case 74).

Over more than 2 years (29 months), the trend line of COPINE score of this individual’s images increased from 4 (intentionally suggestive pictures) to 6 (explicit erotic poses that emphasize the child’s genitals; Figure 1). This indicates a shift of interest to more explicit content at the time of arrest. The individual also appears to have become interested in younger children over time: The mean age of the children decreased from 12.2 years at the beginning of the analytical period to 6.9 years at the end of the period. However, this last month was atypical, and the mean age in the months preceding arrest was stable at approximately 10 years.

For Case 74, the sample analyzed consisted of 1,740 child-pornography images and 57 adult-pornography images. This individual often visited legal pornography sites; for example, several favourites, including bangbus.com, thebigswallow.com, and allanalmovies.com, had been created in the first few days of the analytical period. A few days after installing Microsoft Windows, Case 74 imported images from another source (CD or other medium), presumably to facilitate access to the collection. These images were imported into a folder titled “/Photos/By type/Teen/”.

There were a few noteworthy points to this collection. From the fifth to the ninth month, the individual added several series, following the same pattern: The initial images were of adolescent girls performing sexual modelling, but subsequent images frequently progressed to the depiction of sexual relations between adolescent girls or, more rarely, an adolescent girl and an adult. In particular, there were many images from the LS-Studio Magazine (LSM) series, a series that is popular with consumers of CSEM, and depicts modelling sessions of preadolescent and adolescent girls. This series is a known series taken from an online subscription service in Ukraine that operated in the beginning of 2000 (see Saytarly, 2004).

A greater interest in sexual activities involving preadolescent and adolescent girls was apparent for the 10th and 11th months. Approximately 20 images were added to the folder /Photos/By type/Toy/ (mean age = 8.5 years), and images were also added to Photos/By type/FF/Sonya and Margarita (mean age = 13.5 years).

Sexualized Adolescent Pattern

The Sexualized Adolescent pattern is exemplified by the collection of Case 102 (Figure 2), which was characterized by an increase in the COPINE score and in the age of the subjects depicted over the 24 months studied. This pattern was present in 20% of the sample, and cases exhibiting this pattern had the smallest number of sampled images (M = 534). These results may be related to the difficulties of assessing pictures from adolescents that would be reflected both in our use of police images databases as well as the coder’s ratings (this issue is discussed below, in the section “Limitations” of our study). We observed a mean of 22.3 months of collecting.

Figure 2.

Pattern 2. Sexualized Adolescent (Case 102).

Although images of children in nonsexual contexts were added by Case 102 in a few months, the mean COPINE score trend line increased from 4 at the beginning of the analytical period to 6 at the end. There were, however, a few peaks exceeding 7 (sexual activity involving minors). Case 102 appears to have had an interest for images representing individuals between 12 and 16 years old, although there was a 3-month period in which he collected images of prepubescent children. The trend line showed that the mean ages are between 13 at the beginning and 14.5 at the end of the period.

For this collection, 1,186 images of CSEM and 427 images of adult pornography were sampled. Boys were depicted in 86.6% of the images, and the collection consisted primarily of images of boys 13 to 15 years old. This case differed from the others in that he was charged not just with the possession of CSEM but also with its production.

There were several noteworthy points to this collection. First, the individual regularly added images of preadolescent boys (mean age = 11 years, COPINE score 3-7) and of adolescent boys to a folder—which he had created at the beginning of the period analyzed—titled “Sunshine boys.” Second, there were periods in which he added images with quite low COPINE scores: For example, the images added in the eighth, 14th, and 19th months had a mean COPINE score of 1. The reasons for these dips in COPINE score are not apparent, as the individual continued to collect adult pornography during these periods (albeit less intensively). Finally, in the middle of the period (the 17th month), he added clandestine images of children who were in underwear or nude; the mean COPINE score of these images was 3, and the children’s mean age was 3 years, both of which are inconsistent with the pattern of interest reflected in the rest of the collection. We interpret this as an experiment on his part.

Boy/Girl-Love Pattern

Case 67 exemplifies the Boy/Girl-Love pattern observed in this sample. This pattern represents 20% of the sample with a mean of 1,448 sampled images and a mean of 33.6 months of collecting. Over the 8-month analytical period, he progressively shifted from images depicting explicit sexual acts to images depicting explicit poses. The variations in COPINE scores and age over the analytical period are presented in Figure 3.

Figure 3.

Pattern 3. Boy/Girl-Love (Case 67).

The mean COPINE score trend line fell from 7 to slightly less than 6 over the analytical period. With regard to age, the individual sought increasingly young subjects, with the mean age falling from 14 years (with a peak of 17 years in the first 3 months) to slightly more than 7 years (in the last month).

For this collection, the sample consisted of 656 images. The distribution of the sex of the subjects in the images is noteworthy: 366 images of boys, 166 images of girls, and 124 mixed-sex images. There were two identifiable stages in the collection process. In the first 3 months, there was interest in adolescents (11 years, 15 years, 17 years) depicted in a sexualized context (COPINE scores of 6, 8, 7). The only additions to the collection during this period were single files (rather than series) depicting very hard pornography, including hentai (some with COPINE scores as high as 10). This was followed by installation of the GigaTribe program (a program for the exchange of photos) and a shift in the last months to images depicting subjects 7 to 10 years old in erotic poses (COPINE score of 6).

Although it is not possible to formulate a clinical diagnosis, it appears that this individual’s sexual interest shifted to younger victims in contexts that were less hardcore. This may mirror the migration to more softcore pornography typically seen in the collections of pedophiles who advocate “loving” adult-child relationships.⁴ However, this hypothesis should be considered speculative at this point, as a full understanding of cases’ criminal and sexual evolution requires clinical interviews.

De-Escalation Pattern

The evolution of the collection of Case 81, who exemplifies the De-escalation pattern (22.5% of our sample is in his pattern, with a mean of 1,057 images collected over a mean of 18.4 months), is illustrated in Figure 4. His collection was characterized by a positive slope for age and a negative slope for COPINE score over the 6-month analytical period. The mean age of the subjects increased from 11 years to 14 years, albeit with a dip to 10 years in the third month. The COPINE trend line score exceeded 7 in the second month but subsequently dropped to between 5 and slightly more than 6.

Figure 4.

Pattern 4. De-Escalation Pattern (Case 81).

For this collection, the sample consisted of 96 images, and the analysis of the collection proceeded quite rapidly. The sample of CSEM in this collection comprised 82 images of girls, 11 images of boys, and three mixed-sex images; in addition, 51 images of adult pornography were found on the individual’s hard disk. This individual’s activity was, thus, neither particularly intense nor particularly focused on children. It is possible that he eventually lost interest in CSEM. Moreover, all the images examined had been downloaded from websites, and their filenames (e.g., freetour4[1].jpg, sample.jpg) suggest that they were samples websites provide to entice prospective registered users. Analysis of the collection suggests that the individual was primarily in a discovery mode and that the severity of the images decreased over time.

Discussion

The results of this study demonstrate that the sample of images analysed is comparable to samples of Internet content studied elsewhere (Taylor et al., 2001). The age of the subjects depicted in the images follows a bell curve: a very small number of images of young children, a peak at 10 year olds, and a gradually tailing off to 17 year olds. The analysis of the cases’ interest age groups revealed that collectors who prefer very young children (younger than 6 years) are atypical (n = 2) and that the most popular age group was 6 to 12 years. These results are consistent with the preferred age group reported for sexual aggressors against children (Barbaree & Marshall, 1989). We also found that approximately 61% of the individuals had exclusive age interests, that is, no second choices. Alternatively, nonexclusive age interests, also reported for sexual aggressors against children (Michaud & Proulx, 2009), were observed for 39% of the cases. While the majority of the cases in this study were heterosexual, a finding consistent with other reports in this area (Freund & Watson, 1992), it should be noted that a few individuals with a primarily heterosexual orientation had collections in which almost a third of the images were of boys. Finally, the group in which there was an increase in the severity (COPINE and age) of the pictures was the largest group. All these elements lead to the conclusion that the sexual interests of child-pornography collectors may vary to some extent over time, like sexual aggressors’ (MacCulloch, Snowden, Wood, & Mills, 1983; Money, 1990).

The main objective of this study was to analyse the evolution of the child-pornography collections of individuals convicted of child-pornography offences. In light of the results, we propose four explanations of the nature of, and variations in, child-pornography collections.

The first explanation is that child-pornography collections are an indicator of the sexual interests of the collector (Seto, 2013). This explanation implies that the collector would focus on content that is sexually arousing for him. Analysis of the collected images may shed light on the evolution of the collector’s sexual interests, as the collector’s online activities reflect their interests with regard to the children’s age and sexual acts. However, some collectors appear to have been in discovery mode, and were attempting to determine whether this new type of content pleased them: Should they continue down this road, or return to their initial sexual interests? The representative cases described above are compatible with this explanation. In other words, while it is true that one cannot desire what one does not know, discovery does allow one to know new things to desire. Some of the cases engaged in discovery for months before returning to images depicting the ages and acts they had previously preferred.

A second explanation that is also related to the sexual interest explanation is that collectors become habituated to low-severity pornography, which is congruent with the patterns 1, 2, and 3 of the current study. It has been suggested that habituation to pornographic content leads to boredom, which in turn impels the pornography consumer to seek out new content that is more severe (Reifler et al., 1971; Roy, 2004; Seto, 2013; Taylor & Quayle, 2003). According to Laws and Marshall (1990),

a previously conditioned sexual fantasy (conditional stimulus, CS1) plus masturbatory stimulation (unconditioned stimulus, UCS) can produce high sexual arousal plus orgasm. Minor variations of the original fantasy (CS2) successively substitute for the original one (perhaps to avoid boredom) and paired with masturbation, can elicit the same response. (p. 212)

Thus, to maintain their degree of sexual arousal, child-pornography collectors may be driven to explore other age categories and sexual acts. This discovery process presumably takes the form of trial and error in which they establish how congruent the new content is with their evolving sexual interests.

A third explanation is related to the availability of content: Availability determines collection, in both absolute terms and in terms of temporal pattern. As such, the collection is not a completely accurate measure of sexual interest. First, a collector who has interest in very violent sexuality (COPINE score of 10) may find the supply of this type of content to be limited. It should be noted that all the child-pornography collections included mainstream pornography content. Such “low-hanging fruit” is presumably part of the “routine” content that every collector invariably possesses at one point or another. Second, and alternatively, he may lack the expertise and contacts to locate what he truly desires. Beyond mainstream content, however, the collection is shaped by the collector’s skill and relationships. A skilled collector who frequents virtual spaces is more likely to satisfy his interests than a less skilled one, who must settle for more readily available images. The accessibility of some types of content, thus, influences the evaluation of sexual interests on the basis of collected content.

A corollary of this explanation is that the number of images on a given hard disk is the result of the interaction of accessibility, on one hand, and technical and social skills, on the other. While some content may well be rare, technically sophisticated collectors with a wide range of contacts in clandestine collecting communities can overcome the challenges of rarity. It is, thus, reasonable to assume that it is the interaction of these elements that allows collectors to achieve their objectives.

The final explanation is that collectors proceed through trial and error, discovery, and experimentation. During masturbatory activities, CSEM collectors have the possibility of exploring a wider range of sexual interests than offline sexual offenders, who are limited by the availability of victims. Consequently, they may become motivated to search for new illegal content to nourish their sexual fantasies. This explanation is in agreement with Babchishin et al.’s (2015) meta-analysis, which reveals that online offenders have more deviant sexual interests than offline offenders.

As the data used in the study are limited to the hard-drive content of each offender, it is possible that some reported patterns may also be influenced by personality characteristics. Personality features such as impulsivity, risk-taking, perseverance, and domains of personality functioning such as capacity for regulation, self-observing, and use of moral standards are related to pedophilic and/or ephebophilic sexual interests (see Caretti, Schimmenti, & Bifulco, 2015). As such, CSEM consumers’ behaviours may change as a function of personality features. Further studies might investigate the interaction between CSEM collection patterns and dimension of personality.

The analysis of the child-pornography collections appears to be a promising approach to the evaluation of these collectors’ sexual interests, although the effects of technical skills and virtual sociability must also be taken into consideration. In fact, this approach holds two advantages. First, unlike questionnaires or clinical interviews, the approach is immune to the influences or biases associated with individuals’ attempts to present themselves in the best possible light. Second, the approach does not depend on physiological measures that can be controlled by subjects (e.g., phallometry; Proulx, Côté, & Achille, 1993). Because collectors most probably explored communities hosting images of various types, and were exposed to pro-pedophilic values there (Fortin, 2013; Fortin & Corriveau, 2015; Holt, Blevins, & Burkert, 2010), their collections presumably reflect the content they found attractive.

This study’s sampling and classification methods also have pragmatic implications. Graphing collecting activities may prove useful for the evaluation of sexual interests and—in conjunction with clinical evaluation (Glasgow, 2010)—the severity of the material consumed. It is, however, clear that the results of these evaluations must be considered merely indicative, and must be analysed by professionals. A similar caveat was noted by Freund and Blanchard (1989) in connection with phallometric evaluation, and the same precautions that apply to that method apply to this one.

Limitations

There are some limitations to the innovative method used in this study. First, our sample is relatively small, because our data sources were limited by the rate of police investigations and legal constraints, including those related to the divulgation of evidence. However, our sample does encompass all the cases made available to us at the time of the study. It is nevertheless important to bear in mind that our results may not apply to all CSEM consumers. Second, there are limitations related to the collecting methods. The content of the hard disks only reflects a very narrow portion of the collector’s “career,” and this portion may not be representative of the overall pattern. For example, if an individual buys a computer on Day 1, and on Day 2, installs software, transfers images to the new computer, and is arrested, the possibilities of analysis are greatly limited: The only available material is from a single day of collecting. Third, the resources necessary to analyse a complete collection may be daunting: In the current study, external media such as diskettes, CD-ROMs, and USB keys were excluded, as classification and analysis of the astronomical volume of content contained on these media would have required unreasonable resources. As such, some content has been omitted from the analysis. Fourth, as consumption of child pornography is illegal, this method requires the collaboration of police forces, to provide access to both facilities and reference image databases. Fifth, some changes in the mean age of children depicted may be more significant than others. For example, a difference of 1 year in peripubescent ages may be more significant than the same difference in postpubescent ages. Sixth, in contrast to children 0 to 5 years, it is very difficult to evaluate the age of children 12 to 17 years old. Particularly because of the different stage of biological development of each person in adolescence, it is possible that some form of bias, as well as a tendency to overestimate adolescents’ age, may have occurred. As a result, it is possible that images that should have been allocated to this category were classified as “Adult.” This bias is probably also present in police databases of file signatures. Nevertheless, individuals who collect images may appreciate age in a similar manner as the raters of the study. Finally, variations in size and duration of collection may have had an impact on the evolution of the collection.

Conclusion and Future Research

The results of this study demonstrate that the sample of images analysed is comparable to samples of Internet content studied elsewhere (Taylor et al., 2001). The age of the subjects depicted in the images is normally distributed: a very small number of images of young children, a peak at 10 year olds, and a gradual tailing off to 17 year olds. The analysis of the cases’ preferred age groups revealed that collectors who prefer very young children (younger than 6 years) are atypical (n = 2) and that the most popular age group was 6 to 12 years. These results are consistent with the preferred age group reported for sexual aggressors against children (Barbaree & Marshall, 1989). We also found that approximately 61% of the cases had exclusive age interests, that is, no second choices. Alternatively, nonexclusive age interests, also reported for sexual aggressors against children (Michaud & Proulx, 2009), were observed for 39% of the cases. While the majority of the cases in this study were heterosexual, a finding consistent with other reports in this area (Freund & Watson, 1992), it should be noted that a few cases with a primarily heterosexual orientation had collections in which almost a third of the images were of boys. Finally, the group in which there was an increase in the severity (COPINE and age) of the pictures was the largest group. All these elements lead to the conclusion that the sexual interests of child-pornography collectors vary to some extent over time, much like sexual aggressors’ (McCulloch, 1983; Money, 1990).

The results of this study suggest avenues for future research. First, it would be useful to investigate whether collecting behaviour exhibits a cyclical pattern, as our research only provides a window on the evolution of image severity. Are there, for example, cool-down periods or hot points? Do the various stages of collecting differ depending on the type of material collected?

As Lussier, Bouchard, and Beauregard (2011) noted in connection with offline sexual aggressors, it would be valuable to know more about “successful” offenders. What distinguishes effective collectors from those who collect only the low-hanging fruit? The images that we categorized should allow evaluation of the images’ relative rarity and popularity, which would be a step towards separating neophyte collectors from experienced ones. But doing so requires improving methods for the evaluation of the performance of child-pornography collectors.

Footnotes

Appendix

A Typology of Pedophile Picture Collections (Taylor, Holland, & Quayle, 2001).

Level	Name	Description of picture qualities
1	Indicative	Nonerotic and nonsexualized pictures showing children in their underwear, swimming costumes, etc. from either commercial sources or family albums; pictures of children playing in normal settings, in which the context or organization of pictures by the collector indicates inappropriateness.
2	Nudist	Pictures of naked or seminaked children in appropriate nudist settings, and from legitimate sources.
3	Erotica	Surreptitiously taken photographs of children in play areas or other safe environments showing either underwear or varying degrees of nakedness.
4	Posing	Deliberately posed pictures of children fully, partially clothed, or naked (where the amount, context, and organization suggest sexual interest).
5	Erotic Posing	Deliberately posed pictures of fully, partially clothed, or naked children in sexualized or provocative poses.
6	Explicit Erotic Posing	Emphasising genital areas where the child is either naked, partially, or fully clothed.
7	Explicit Sexual Activity	Involves touching, mutual and self-masturbation, oral sex, and intercourse by child, not involving an adult.
8	Assault	Pictures of children being subject to a sexual assault, involving digital touching, involving an adult.
9	Gross Assault	Grossly obscene pictures of sexual assault, involving penetrative sex, masturbation, or oral sex involving an adult.
10	Sadistic/Bestiality	Pictures showing a child being tied, bound, beaten, whipped, or otherwise subject to something that implies pain. Pictures where an animal is involved in some form of sexual behaviour with a child.

Declaration of Conflicting Interests

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

Funding

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

Notes

ORCID iD

Francis Fortin

References

Akdeniz

(1997). The regulation of pornography and child pornography on the Internet. Journal of Information, Law and Technology. Retrieved from https://dx-doi-org.web.bisu.edu.cn/10.2139/ssrn.41684

Babchishin

K. M.

Hanson

R. K.

Hermann

C. A.

(2011). The characteristics of online sex offenders: A meta-analysis. Sexual Abuse: A Journal of Research and Treatment, 23, 92-123. doi:10.1177/1079063210370708

Babchishin

K. M.

Hanson

R. K.

VanZuylen

(2015). Online child pornography offenders are different: A meta-analysis of the characteristics of online and offline sex offenders against children. Archives of Sexual Behavior, 44, 1-22. doi:10.1007/s10508-014-0270-x

Bang

Yoo

Lee

(2011). Analysis of changes in file time attributes with file manipulation. Digital Investigation, 7, 135-144. doi:10.1016/j.diin.2010.12.001

Barbaree

H. E.

Marshall

W. E.

(1989). Erectile responses among heterosexual child molesters, father-daughter incest offenders, and matched non-offenders: Five distinct age preference profiles. Canadian Journal of Behavioural Science, 21, 70-82.

Beech

A. R.

Elliott

I. A.

Birgden

Findlater

(2008). The Internet and child sexual offending: A criminological review. Aggression and Violent Behavior, 13, 216-228. doi:10.1016/j.avb.2008.03.007

Caretti

Schimmenti

Bifulco

(2015). Psychopathology of online grooming. In Webster

Davidson

Bifulco

(Eds.), Online offending behaviour and child victimization: New findings and policy (pp. 91-119). Basingstoke, UK: Palgrave Macmillan.

Carr

(2004). Internet traders of child pornography and other censorship offenders in New Zealand. Wellington: Department of Internal Affairs, Government of New Zealand. Retrieved from https://www.dia.govt.nz/diawebsite.nsf/wpg_URL/Resource-material-Our-Research-and-Reports-Internet-Traders-of-Child-Pornography-and-other-Censorship-Offenders-in-New-Zealand?OpenDocument

Carr

(2001). Theme paper on child pornography for the 2nd World Congress on Commercial Sexual Exploitation of Children. London, England: Children & Technology Unit NCH.

10.

Durkin

K. F.

Bryant

C. D.

(1999). Propagandizing pederasty: A thematic analysis of the on-line exculpatory accounts of unrepentant pedophiles. Deviant Behavior, 20, 103-127.

11.

Elliott

I. A.

Beech

A. R.

Mandeville-Norden

(2013). The psychological profiles of internet, contact, and mixed internet/contact sex offenders. Sexual Abuse: A Journal of Research and Treatment, 25, 3-20. doi:10.1177/1079063212439426

12.

Fortin

(2013). Usenet newsgroups, child pornography and the role of participants. In Morselli

(Ed.), Illicit networks (pp. 231-148). Oxford, UK: Routledge.

13.

Fortin

Corriveau

(2015). Who is Bob_34? Investigating child cyberpornography. Vancouver, British Columbia, Canada: UBC Press.

14.

Freund

Blanchard

(1989). Phallometric diagnosis of pedophilia. Journal of Consulting and Clinical Psychology, 57, 100-105. Retrieved from http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&id=2925958&retmode=ref&cmd=prlinks

15.

Freund

Watson

R. J.

(1992). The proportions of heterosexual and homosexual pedophiles among sex offenders against children: An exploratory study. Journal of Sex & Marital Therapy, 18, 34-43. doi:10.1080/00926239208404356

16.

Glasgow

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

17.

Holt

T. J.

Blevins

K. R.

Burkert

(2010). Considering the pedophile subculture online. Sexual Abuse: A Journal of Research and Treatment, 22, 3-24. doi:10.1177/1079063209344979

18.

Krone

(2004). A typology of online child pornography offending. Trends & Issues in Crime and Criminal Justice, 279, 1-6.

19.

Laws

R. D.

Marshall

W. L. L.

(1990). A conditioning theory of the etiology and maintenance of deviant sexual preference and behavior. In Marshall

W. L. L.

Laws

R. D.

Barbaree

H. E.

(Eds.), Handbook of sexual assault: Issues, theories, and treatment of the offender (pp. 209-229). New York, NY: Plenum Press.

20.

Lussier

Bouchard

Beauregard

(2011). Patterns of criminal achievement in sexual offending: Unravelling the “successful” sex offender. Journal of Criminal Justice, 39, 433-444. doi:10.1016/j.jcrimjus.2011.08.001

21.

MacCulloch

M. J.

Snowden

P. R.

Wood

P. J. W.

Mills

H. E.

(1983). Sadistic fantasy, sadistic behaviour and offending. British Journal of Psychiatry, 143, 20-29.

22.

McGraw

K. O.

Wong

S. P.

(1996). Forming inferences about some intraclass correlation coefficients. Psychological Methods, 1, 30-46.

23.

Merdian

H. L.

Thakker

Wilson

Boer

(2013). Assessing the internal structure of the COPINE scale. Psychology, Crime & Law, 19, 21-34.

24.

Michaud

Proulx

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

25.

Middleton

Mandeville-Norden

Hayes

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

26.

Mocas

(2004). Building theoretical underpinnings for digital forensics research. Digital Investigation, 1, 61-68. doi:10.1016/j.diin.2003.12.004

27.

Money

(1990). Forensic sexology: Paraphilic serial rape (biastophilia) and lust murder (erotophonophilia). American Journal of Psychotherapy, 44, 26-36.

28.

Oosterbaan

A. G.

(2009). Global symposium for examining the relationship between online and offline offences and preventing the sexual exploitation of children. Washington, DC: U.S. Department of Justice, Criminal Division, Child Exploitation and Obscenity Section, Law Enforcement Projects Sub-Group.

29.

Palmer

(2001, August 7-8). A road map for digital forensic research. Report from the first Digital Forensics Research Workshop (DFRWS), Utica, NY.

30.

Palmer

(2005). Behind the screen: Children who are the subjects of abusive images. In Quayle

Taylor

(Eds.), Viewing child pornography on the Internet: Understanding the offence, managing the offender, helping the victim (pp. 61-74). Dorset, UK: Russell House.

31.

Portney

L. G.

Watkins

M. P.

(2000). Foundations of clinical research: Applications to practice. Upper Saddle River, NJ: Prentice Hall.

32.

Proulx

Côté

Achille

P. A.

(1993). Prevention of voluntary control of penile response in homosexual pedophiles during phallometric testing. Journal of Sex Research, 30, 140-147.

33.

Quayle

Jones

(2011). Sexualized images of children on the Internet. Sexual Abuse: A Journal of Research and Treatment, 23, 7-21. doi:10.1177/1079063210392596

34.

Quayle

Taylor

(2002). Child pornography and the Internet: Perpetuating a cycle of abuse. Deviant Behavior, 23, 331-361. doi:10.1080/01639620290086413

35.

Reifler

C. B.

Howard

Lipton

M. A.

Liptzin

M. B.

Widmann

D. E.

(1971). Pornography: An experimental study of effects. American Journal of Psychiatry, 128, 575-582.

36.

Roussev

Chen

Bourg

Richard

III . (2006). md5bloom: Forensic filesystem hashing revisited. Digital Investigation, 3, 82-90. doi:10.1016/j.diin.2006.06.012

37.

Roy

(2004). Étude exploratoire des événements et des caractéristiques des individus mis en cause dans des cas de possession et de distribution de matériel pornographique juvénile sur Internet [Exploratory Study of the Events and Characteristics of Individuals Involved in the Possession and Distribution of Juvenile Pornography on the Internet]. Montréal, Québec, Canada: École de criminologie.

38.

Saytarly

(2004, August 13). Ukrainian child porn ring. Computer Crime Research Center. Retrieved from http://www.crime-research.org/news/2004.08.13/561/

39.

Sellier

(2003). Innocence-en-danger.com: Internet: le paradis des pédophiles [Innocence-in-danger.com: Internet: the paradise of pedophiles]. Paris, France: Éditions Pion.

40.

Seto

M. C.

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

41.

Seto

M. C.

Eke

A. W.

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

42.

Seto

M. C.

Wood

J. M.

Babchishin

K. M.

Flynn

(2012). Online solicitation offenders are different from child pornography offenders and lower risk contact sexual offenders. Law and Human Behavior, 36, 320-330. doi:10.1037/h0093925

43.

Sheehan

Sullivan

(2010). A qualitative analysis of child sex offenders involved in the manufacture of indecent images of children. Journal of Sexual Aggression, 16, 143-167. doi:10.1080/13552601003698644

44.

Tanner

J. M.

(1990). Foetus into man: Physical growth from conception to maturity. Cambridge, MA: Harvard University Press.

45.

Taylor

(1999). The nature and dimensions of child pornography on the Internet. Paper presented at the International Conference Combating Child Pornography on the Internet, Vienna, Austria, 29 September-1 October 1999.

46.

Taylor

Holland

Quayle

(2001). Typology of paedophile picture collections. Police Journal, 2, 97-107.

47.

Taylor

Quayle

(2003). Child pornography: An Internet crime. New York, NY: Brunner-Routledge.

48.

Wolak

Finkelhor

Mitchell

K. J.

(2005). Child-pornography possessors arrested in Internet-related crimes. Alexandria, VA: National Center for Missing & Exploited Children.

49.

Wolak

Finkelhor

Mitchell

K. J.

(2011). Child pornography possessors: Trends in offender and case characteristics. Sexual Abuse: A Journal of Research and Treatment, 23, 22-42. doi:10.1177/1079063210372143

50.

Wolak

Liberatore

Levine

B. N.

(2014). Measuring a year of child pornography trafficking by U.S. computers on a peer-to-peer network. Child Abuse & Neglect, 38, 347-356. doi:10.1016/j.chiabu.2013.10.018

51.

Wortley

R. K.

Smallbone

(2006). Problem-Oriented Guides for Police Problem-Specific Guides Series. Child pornography on the internet. Albany, NY: Center for Problem-Oriented Policing, University of Albany.