Bartels 14 PHD

UNDERSTANDING THE COGNITIVE PROCESSES
ASSOCIATED WITH SEXUAL FANTASIES:

TOWARDS A DUAL-PROCESS MODEL
By
Ross Matthew Bartels
A thesis submitted to the College of Life and Environmental Sciences of the

University of Birmingham for the degree of
DOCTOR OF PHILOSOPHY
School of Psychology
College of Life and Environmental Sciences
The University of Birmingham
August, 2013
University of Birmingham Research Archive

e-theses repository
This unpublished thesis/dissertation is copyright of the author and/or third
parties. The intellectual property rights of the author or third parties in respect
of this work are as defined by The Copyright Designs and Patents Act 1988 or
as modified by any successor legislation.
Any use made of information contained in this thesis/dissertation must be in
accordance with that legislation and must be properly acknowledged. Further
distribution or reproduction in any format is prohibited without the permission
of the copyright holder.
ABSTRACT
Deviant sexual fantasies are regarded as an important factor in terms of understanding and
treating sexual offenders. However, while there is a large amount of research on deviant
sexual fantasies, little has been done to understand the cognitive processes underlying and
associated with this phenomenon. Thus, in this thesis, a novel theoretical model of deviant
sexual thoughts and fantasies is proposed and subjected to empirical testing. In Chapter 1, the
literature is examined to ascertain what is currently known about deviant sexual fantasies and
what needs to be done. This provides the rationale for the thesis and also contributed to the
development of the theoretical model, which is termed the Dual-Process of Sexual Thinking
(DPM-ST). Chapter 2 describes the DPM-ST. In brief, the DPM-ST states that associative
processes give rise to spontaneous deviant sexual thoughts, whereas controlled processes are
responsible for the act of using a deviant sexual fantasy. The subsequent five chapters aimed
to test the DPM-ST's main assumptions. In Chapter 3, the results showed some support for
the assumption that child abusers hold sex-related associations (i.e., child-sex associations)
that share a relationship with sexual fantasies about children. The structural equation
modelling study in Chapter 4 provides support for the idea that deviant sexual fantasies are
more likely to be used if the individual has a greater proclivity to fantasise generally and
holds explicit attitudes that are supportive of the deviant sexual fantasy. Using a dual-task
paradigm, Chapter 5 provides support for the hypothesis that sexual fantasising is a controlled
process that requires cognitive resources. Chapter 6 outlines data that does not support the
hypothesis that sexually fantasising about dominance activates 'self-powerful' associations. In
Chapter 7, the results indicate that abusers who repeatedly use sexual fantasies about children
strongly associate the concepts of 'children' and 'sexual fantasy'. Chapter 8 concludes the
thesis by summarising the findings and discussing their theoretical and clinical implications.
Also, the limitations of the studies are outlined, as are some ideas for future research.
i
ACKNOWLEDGEMENTS
First and foremost I would like to thank my two supervisors, Prof. Tony Beech and Dr. Leigh
Harkins. Throughout this project, both have been unwavering in their support and guidance
and, as a result, I have achieved more than I thought possible. I am especially thankful for the
encouragement they gave me, as it provided me with the confidence I needed to develop my
ideas and produce the present thesis. I am also truly appreciative of the patience and trust
they showed while I pedantically developed these ideas. In addition, their feedback and
insights have benefited me so much, both as a researcher and as an academic writer. For this I
am both grateful and honoured.
ii
I would also like to thank the PhD students (both past and present) within the Centre for
Forensic and Criminological Psychology; Shannon Vettor, Juliane Kloess, Teresa da Silva,
Chelsea Slater, Steven Gillespie, Shelley Price, and Ian Elliot. You have each been
supportive in your own way and I am glad to be able call you all friends. A big thank is also
owed to Sue Hanson. Not only were you always there to help me if I needed it.
To all my friends that I have met and who have been there throughout my PhD, I thank you
all immensely. I am specifically grateful to one of my closest friends, Shannon Vettor. Your
support and friendship kept me sane during this process and I continue to miss you. I also
want to thank one of my oldest friends - Chris Bolger - for being supportive even before I
started the PhD. Our nights out were always a welcome break to the grind of the thesis.
Finally, I want to thank my family for their love and support. Mum, thank you for having
total confidence in me and for being there whenever I needed a good chat. Dad, thank you for
always supporting me and for listening to me go on and on about my ideas. Without the
support of both of you, this would have been a much, much harder process. Also, thank you
to my brother Scott (as well as Robert Daggers who is essentially a brother) for our climbing
adventures. These were some of the best times Ive had and they helped take my mind off of
work for a while so that I could return to it refreshed. Thank you also to Will Dunn for your
help and support, and for giving me the room at the top of the house to stay and write up.
I am proud of what I achieved in this thesis, but it would not have been possible without the
help of all those I have mentioned. So thank you once again. Now on to the next endeavour...
Table of Contents
Abstract
Acknowledgements
ii
Table of Contents
iv
List of Figures
vii
List of Tables
viii
Introduction
Statement of Authorship
Chapter 1: Deviant sexual fantasies: What is currently known
Chapter Rationale
1.1 Introduction
1.2 Definition of deviant sexual fantasy
1.3 Prevalence of deviant sexual fantasies
10
1.4 Effects and functions of deviant sexual fantasy
14
1.5 Theories of deviant sexual fantasy
18
1.6 Conclusions
29
Chapter 2: The Dual-Process Model of Deviant Sexual Thinking (DPM-ST)
30
Chapter Rationale
30
2.1 Introduction
31
2.2 A dual-process approach
31
2.3 Remembering and imagining episodic events
37
2.4 Outlining the DPM-ST
39
2.5 Conclusion
44
Chapter 3: Investigating Sex-related Associations in Child Abusers
46
Chapter Rationale
46
3.1 Introduction
47
3.2 Method
51
3.3 Results
61
3.4 Discussion
74
Chapter 4: Investigating the Factors that Influence the Use of Deviant

Sexual Fantasies
81
Chapter Rationale
81
4.1 Introduction
82
4.2 Method
85
4.3 Results
93
4.4 Discussion
98
Chapter 5: Testing Whether Sexual Fantasies Require Working Memory

Resources
103
Chapter Rationale
103
5.1 Introduction
104
5.2 Method
106
5.3 Results
111
5.4 Discussion
118
Chapter 6: Investigating Whether Sexual Fantasies Activate Non-sexual

Associations
124
Chapter Rationale
124
6.1 Introduction
123
6.2 Method
127
6.3 Results
133
6.4 Discussion
138
Chapter 7: Investigating Whether Child Abusers Associate Sexual Fantasies

with Children
142
Chapter Rationale
142
7.1 Introduction
143
7.2 Method
145
7.3 Results
154
7.4 Discussion
162
Chapter 8: General Discussion
169
8.1 Summary of findings
169
8.2 Implications of the findings
172
8.3 Limitations and directions for future research
183
8.4 Conclusions
187
References
189
Appendices
209
List of Figures
Chapter
Figure
Title
Page
2.1
40
3.1
Schematic representation of the Dual-Process Model of Sexual

Thinking.
Example screen from the Sorting Paired Features task
3.2
56
4.1
Mean D-associations from the experimental SPF task for

extrafamilial abusers, non-extrafamilial abusers, and nonoffenders.
Mean D-associations from the control SPF task for extrafamilial
abusers, non-extrafamilial abusers, and non-offenders.
Mean and SEs for the child-sex and adult-sex amended
D-associations for extrafamilial abusers, non-extrafamilial abusers,
and non-offenders
Mean D-associations (and SEs) for each of the non-extrafamilial
subgroups, extrafamilial abusers, and non-offenders
Means for the amended D-association (and SEs) for each of the
non-extrafamilial subgroups, extrafamilial abusers, and nonoffenders
Structural component of the hypothesised SEM model.
4.2
Hypothesised structural equation model
91
4.3
Final structural equation model.
95
5.1
111
5.1
Scores on each WSFQ subscale as a percentage of the overall

score
Mean difference scores for vividness for all four conditions
5.2
Mean difference scores for emotionality for all four conditions
115
5.3
Mean difference scores for arousability for all four conditions
117
6.1
134
6.2
7.1
7.2
Mean scores (and SEs) for each WSFQ subscale across both
groups
Mean response latencies (treated data) for each association,
across both groups
A graph showing the mean scores (and SEs) on each WSFQ
subscale across all groups
Mean D-scores for each group using the experimental GNAT
7.3
Mean D-scores for each group using the control GNAT
159
7.4
A graph showing the mean GNAT D-score (and SEs) for the
extrafamilial group and the three non-extrafamilial subgroups
162
3.3
3.4
3.5
3.6
62
65
71
73
74
85
113
136
154
157
vii
List of Tables
Chapter
3
3
3
3
3
Table
Title
Demographic information for the extrafamilial abusers, non3.1
extrafamilial abusers, and non-offenders.
Mean scores for static and dynamic risk, psychopathy, and IQ in
3.2
both offender groups.
The proportion of participants in treatment and pre-treatment for
3.3
Offence history information for both offender groups.
3.4
3.5
3.6
3.7
3.8
3.9
3.10
4.1
4.2
4.3
4.4
5.1
5.2
6.1
6.2
Mean raw response latencies (milliseconds) from the Critical

SPF for extrafamilial abusers, non-extrafamilial abusers, and
non-offenders.
Mean error rates (%) and SDs from the experimental SPF for
extrafamilial abusers, non-extrafamilial abusers, and nonoffenders
Mean error rates (%) and SDs from the control SPF for
Significant correlations between the four D-associations and
WSFQ sexual fantasies for extrafamilial abusers
WSFQ sexual fantasies for non-extrafamilial abusers
WSFQ sexual fantasies for non-offenders
Means, standard deviations, and internal consistencies for the
indicator (observed) variables.
Correlations (using Spearmen's Rho) for all indicator (observed)
variables.
Comparisons between the original sample and mean bootstrap
sample of 1,000 (using unstandardised regression weights)
Comparisons between the original sample and mean bootstrap
sample of 1,000 (using standardised regression weights).
Means and SDs for vividness, emotionality, and arousability
ratings before and after eye-movements for memory-based
sexual fantasies.
Means and SDs for vividness, emotionality, and arousability
ratings before and after eye-movements for imagination-based
sexual fantasies.
Implicit Association Task (IAT) for Powerfulness: Task
Sequence
Mean error rates (%) and SDs for each block of the IAT for each
condition
Page
53
54
54
55
60
62
65
68
69
69
93
94
97
98
112
112
128
133
8
7.1
7
7
7.2
7.3
7.4
7.5
7.6
7.7
Demographic information for the extrafamilial abusers, nonextrafamilial abusers, and non-offenders.
Mean scores for static and dynamic risk, psychopathy, and IQ in
The proportion of participants in treatment and pre-treatment for
Offence history information for both offender groups.
Mean raw reaction times (SDs) for all three groups on each block
of each GNAT
Mean error rates (%) and SDs from the experimental GNAT for
Mean error rates (%) and SDs from the control GNAT for
146
147
147
148
153
158
159
INTRODUCTION
Sexual offending is a universal problem that can have a hugely damaging effect on the
individuals involved, as well as the community in general. In the United Kingdom, the latest
statistics (i.e., 2011/2012) indicate that 53,665 sexual offences were recorded by the police
(Ministry of Justice, 2013). Of these, 82% can be regarded as serious sexual offences (i.e.,
sexual assault, rape, and activity with a minor). Over the past three to four decades, major
advances have been made - in both academia and practice - to understand sexual offenders
(Ward, Polaschek, & Beech, 2006). This work has identified a number of factors that
contribute to the commission of a sexual offence and treatment programmes have been
developed to address the factors that are most associated with sexual (re)offending.
One factor that was of central importance during the early approaches to treatment was
deviant sexual interests (Marshall, O'Brien, & Marshall, 2009). That is, it was long presumed
that deviant sexual behaviour was the direct product of a deviant sexual interest (McGuire,
Carlisle, & Young, 1965). For example, child abusers were thought to sexually abuse children
because they have a deviant sexual interest in children. As a result, deviant sexual fantasies
became a core target for clinicians during treatment, mainly because behaviourists had started
conceptualising deviant sexual fantasy as an independent variable that played a role in causing
deviant sexual interests (Abel & Blanchard, 1974).
Nowadays, many different factors are of clinical importance when treating sexual
offenders (Mann, Hanson, & Thornton, 2010). However, deviant sexual interests are still a core
concern as studies show it to be a strong predictor of sexual recidivism (Hanson & MortonBourgon, 2005). As such, it follows that deviant sexual fantasies should also remain a core
concern (Bartels & Gannon, 2011). Moreover, deviant sexual fantasies are not just sexual in
nature but also include issues of power and control, conveying aggression and a desire to
humiliate (Marshall, 1996). Thus, their influence on sexual offenders can often extend beyond
just creating and/or strengthening a deviant sexual interest (Gee, Ward, & Eccleston, 2003).
1
Despite the clinical importance of deviant sexual fantasies and the advances that have
been made in psychology, little has been done to investigate or understand deviant sexual
fantasies more deeply. As Sheldon and Howitt (2008) state, at the theoretical level, the role
of sexual fantasy needs to be better understood" (p. 140). Moreover, the construct itself seems
to require a better understanding. This includes explaining the cognitive processes that
influence and lead to the use of a deviant sexual fantasy; the processes that underlie the actual
act of deviant sexual fantasising; and any other cognitive processes related the use of deviant
sexual fantasies.
The field of psychology, and its various approaches (e.g., cognitive psychology), has
provided valuable insights into the human mind that can be drawn upon to make predictions
about sexual fantasies. It has also developed a number of paradigms and techniques that can be
useful in investigating sexual fantasies more fully. Thus, the first overall aim of this thesis was
to take what is currently known about deviant sexual fantasies and knit it with contemporary
findings from relevant areas of cognitive psychology so that a novel theoretical account of
deviant sexual fantasy could be constructed. The second overall aim was to test a number of
hypotheses generated by the theory. Not only would this provide empirical support for the
theory but it would provide new insights into the cognitive processes underlying and related to
deviant sexual fantasies. Such insights may be useful in devising new clinical strategies for
treating deviant sexual fantasies in sexual offenders.
Main aims of the thesis

- To examine and evaluate the empirical and theoretical literature on deviant sexual fantasies
- To outline a new theoretical model of deviant sexual thoughts and fantasies
- To empirically test some of the core assumptions of the theoretical model through via five
different studies
Structure of the thesis

This thesis aims to understand the cognitive processes associated with deviant sexual fantasies
by developing and testing a novel theoretical model. Chapter 1 presents a contemporary
review of the literature that examines: 1) the definitional issues related to deviant sexual
fantasy; 2) research related to the prevalence of deviant sexual fantasies; 3) research on the
effects and functions of deviant sexual fantasies; and 4) the existing theoretical accounts of
deviant sexual fantasy.
Chapter 2 presents the Dual-Process of Sexual Thinking (DPM-ST); a theoretical model that
accounts for the cognitive processes associated with deviant sexual thoughts and fantasies.
This model was constructed by combining the results of the literature review in Chapter 1
with principles and findings from relevant areas of cognitive psychology. These areas are
outlined before the theory itself is described, with the aid of a schematic representation.
Chapter 3 describes a study that investigates - using a recently developed indirect measure whether extrafamilial child abusers hold stronger distorted sex-related associations in their
long-term memory relative to other child abusers and non-offenders. Also, this study
investigates whether distorted sex-related associations are correlated with deviant sexual
fantasies.
Chapter 4 reports a structural equation modelling study that aimed to test whether certain
variables influence the use of deviant sexual fantasies in male non-offenders. These variables
include fantasy proneness, dissociation, vividness of visual imagery, and explicit distorted
attitudes about women.
Chapter 5 describes a study that investigates the hypothesis that sexual fantasising is a
controlled process requiring the resources of working memory. The study employs a dual-task
paradigm, whereby a task known to require working memory is performed at the same time as
mentally envisioning a sexual fantasy. Ratings on vividness, emotional valence, and
3
arousability are measured before and after the dual-task, as well as before and after a control
task.
Chapter 6 reports a study designed to show that sexually fantasising about a deviant theme
(i.e., dominance) can prime other, non-sexual associations (i.e., self-powerful associations)
and increase one's explicit self-concept (i.e., self-reported powerfulness). Also, the study
investigates whether self-esteem and the use of dominance-related sexual fantasies moderate
the priming effect.
Chapter 7 describes a study that investigates whether child abusers who regularly use deviant
sexual fantasies hold strong associations between the concepts of 'sexual fantasy' and
'children'. The study employs an indirect measure to detect and assess the strength of this
presumed association.
In concluding the thesis, Chapter 8 provides a summary of the results, which are then pulled
together to discuss the theoretical implications of the research (in terms of the DPM-ST). The
clinical implications of the findings are also discussed. Finally, the limitations of each study
are addressed, along with some considerations for future research.
STATEMENT OF AUTHORSHIP
The work in this thesis has been used for a number of forthcoming publications (see
Appendix A). Specifically, the contents of Chapter 1 and Chapter 2 are being used in a
chapter for a forthcoming book and a journal article. The results described in Chapter 5 have
been used for a manuscript that has been submitted for publication. Similarly, the results in
Chapter 7 have been used for a manuscript that is currently under review. An earlier version
of this manuscript was also submitted and awarded the Graduate Student Award from the
Association for the Treatment of Abusers (ATSA). I am the lead author on each of these
manuscripts, although my supervisors (Anthony Beech and Leigh Harkins) are also included
as authors due to their collaborative contribution. In addition, David Thornton is listed as an
author on the Chapter 7-based manuscript due to his help in gaining access to forensic
participants.
CHAPTER 1
DEVIANT SEXUAL FANTASIES: WHAT IS CURRENTLY KNOWN
Chapter Rationale
The purpose of this chapter is to introduce the concept of deviant sexual fantasy by providing a
contemporary review of the empirical and theoretical literature. By examining what is currently
understood about deviant sexual fantasies - in terms of definition, prevalence, function, and
theory - this chapter aims to underscore the importance of understanding deviant sexual
fantasies more fully, particularly from a cognitive perspective. Thus, this chapter can
be seen as providing the rationale for the overall thesis.
1.1 Introduction
Deviant sexual fantasies have long played an important in role in the understanding and
treatment of sexual offenders. For the most part, this role has been related to the development
and maintenance of deviant sexual interests (Abel & Blanchard, 1974), due mainly to the
influence of conditioning theories of sexual deviation (e.g., McGuire, Carlisle, & Young,
1965). However, others have held the view that deviant fantasy can influence offending
behaviour more directly. These influences include providing offenders with an offence plan
(Pithers, 1990; Ward & Hudson, 2000); refining an offenders offence-specific expertise or
organisation (Bourke, Ward, & Rose, 2012; Deu & Edelmann, 1997); disinhibiting offenders
to their fantasised deviant behaviour (Ward, Hudson, & Keenan, 1998); and providing
offenders with a particular modus operandi (Leclerc, Carpentier, & Proulx, 2006).
Given their assumed importance, it is no surprise that deviant sexual fantasies are of clinical
interest and have been included within relapse prevention models (Pithers, Kashima,
Cummings, & Beal, 1988) and treatment programs (Stinson & Becker, 2012). Similarly,
deviant sexual fantasies are typically included within macro-theories of sexual offending (e.g.,
Marshall & Barbaree, 1990; Ward & Beech, 2006) and have been the focus of many empirical
studies over the past few decades. Thus, in order to gain an understanding of what is currently
understood about deviant sexual fantasies, this chapter examines the literature with regards to
four core domains; namely, definition, prevalence, function, and theories of deviant sexual
fantasies.
1.2 Definition of deviant sexual fantasy

Sexual fantasy has been defined as almost any form of mental imagery that is sexually
arousing to the individual (Leitenberg & Henning, 1995). However, despite its frequent use
within the literature, there is no formal definition for deviant sexual fantasy (Bartels &
Gannon, 2011). This seems to be due to the contentious nature of the term. For example,
7
Leitenberg and Henning (1995) have debated whether deviant fantasies are simply those that
are statistically the most unusual or those linked to unacceptable behaviour even if they are
not that unusual. The former of these possibilities is a less convincing definition because
many non-offenders report using deviant fantasies (Becker-Blease, Friend, & Freyd, 2006;
Crpault & Couture, 1980; Williams, Cooper, Howell, Yuille, & Paulhus, 2009). In other
words, the sexual fantasies that are often referred to as deviant (i.e., involving rape or
children) appear to be within the normal realm of human experience (Gee, Devilly, &
Ward, 2004; p.51). However, defining deviant fantasy as a fantasy linked with unacceptable
behaviour is also problematic because many non-offenders fantasise about deviant
behaviours and do not carry out the behaviour (Leitenberg & Henning, 1995).
Given this, perhaps the referent for the word deviant is simply the behaviour depicted
within the fantasy; in other words, deviant sexual fantasies may best be defined as "fantasies
that depict deviant sexual behaviour" (Bartels & Gannon, 2011, p. 552). However, this raises
the question of what can be regarded as a deviant sexual behaviour. Although formal
definitions of sexual deviance remain inconsistent (Ward, Laws, & Hudson, 2003), a
number of authors agree that sexually deviant behaviours are largely infrequent in the
population; involve intentionally inflicting harm in a sadistic or sexually aggressive way;
involve illegal acts (e.g., rape, child abuse, exhibitionism); and involve socially marginal or
unacceptable behaviours (Gee and Belofastov; 2006; Gee et al., 2004; Leitenberg & Henning,
1995; Williams et al., 2009; Ward et al., 2003). From this general consensus, it can be seen
that deviant does not necessarily mean illegal (i.e., consensual sadistic behaviour). It also
means that infrequent yet harmless and non-illegal behaviours - such as the transvestic
fetishism (DSM-IV; American Psychiatric Association, 2000) - may be considered deviant.
Given that sexual deviance tends to be discussed within the context of sexual offending
(Ward et al., 2003), many may view this as problematic.
Noting the contention associated with the term deviant sexual fantasy, other
researchers have proposed alternate terms and definitions (e.g., Bartels & Gannon, 2011; Gee
et al., 2004). For example, Gee et al. (2004) introduced the term offence-focused fantasy,
defined as fantasies involving deviant material that if acted out would constitute a sexual
offence. According to Gee and Belofastov (2006), this term can be used to differentiate the
fantasies used by sex offenders from those evident in non-offender populations. However, as
Bartels and Gannon (2011) note, it is unclear how the term helps such a differentiation, since a
fantasy involving rape would always constitute an offence if acted out, irrespective of whether
an offender or a non-offender created and used it. Also, the term offence-focused fantasy
does not allow for the inclusion of fantasies involving force with a consenting partner. While
these fantasises would not constitute a sexual offence if acted out, they may increase the risk
of an offence occurring if the fantasiser has certain traits or is in a certain state that does not
enable him to realise or care that a woman is not consenting (Bartels & Gannon, 2011). Thus,
while the fantasy was not offence-focused, it still played a part in influencing an offence.
Based on this, Bartels and Gannon (2011) propose focusing on the
risk-increasing function of deviant sexual fantasies when defining deviant sexual fantasies, as
opposed to the content of the fantasy.
Another important definitional issue concerns the term fantasy, as many researchers
consider sexual fantasies to be anything from fleeting sexual thoughts to elaborated sexual
scenarios (Leitenberg & Henning, 1995; Wilson, 1978). This is an issue because treating
sexual fantasy as an umbrella term is likely to obscure important differences between the
different forms of sexual cognition (Bartels & Gannon, 2011; Renaud & Byers, 2001). Over
the past few decades, many prominent researchers have shown support for the idea that there
is a difference of kind between thoughts and fantasies. For example, Singer (1966) defined
fantasy as a shift of attention away from a physical or ongoing task towards an internal
stimulus. Thus, by Singer's definition, sexual thoughts triggered by an external stimulus are
9
not equivalent to a sexual fantasy. Barlow, Hayes, and Nelson (1984) made a distinction
between externally triggered sexual thoughts (which they termed 'urges') and internally
generated sexual fantasies. Moreover, Jones and Barlow (1990) found support for Barlow et
al.'s (1984) distinction by showing that the former occurs at a higher rate than the latter.
Finally, Byers, Purdon, and Clark (1998) furthered emphasised that sexual thoughts are
distinct from sexual fantasies because they are triggered spontaneously by external cues,
whereas the sexual fantasies are generated purposefully and internally.
This distinction corresponds with Kavanagh, Andrade, and Mays (2005) work on
desire. According to their research, intrusive appetitive thoughts are triggered automatically
by external and internal cues, whereas elaborated thought is a more deliberate and controlled
process involving greater cognitive effort. Thus, if we draw upon this work and define sexual
fantasy as a deliberate and elaborated thought process, it becomes more apparent how sexual
fantasies differ from fleeting sexual thoughts. This allows for a more refined conceptualisation
to be developed; one that may have important implications for understanding the factors
related to deviant sexual fantasies, including how they are assessed and treated.
1.3 Prevalence of deviant sexual fantasies

Research has shown that sex offenders in general tend to use deviant fantasies more than nonsex offender controls (Curnoe & Langevin, 2002; Langevin, Lang, & Curnoe, 1998; TurnerMoore, Waterman, & Smith, 2010). Particularly high prevalence rates have been found in
exhibitionists (88%; Dandescu & Wolfe, 2003); serial sexual murderers (86%; Prentky et al.,
1989); child molesters (81%; Dandescu & Wolfe, 2003). It was also recently found that highrisk sex offenders (i.e., irrespective of offence type) show a high prevalence rate for deviant
fantasies (82%; Woodworth et al., 2013).
10
As Williams et al. (2009) note, the content of sex offenders sexual fantasies is
typically reflective of their sexual offending behaviour (Gee et al., 2004; Looman, 1995). For
example, child abusers tend to fantasise about children and rapists about violence
(Woodworth et al., 2013). However, other research shows that the content of sexual fantasies
can be rather varied for some offenders (Goldstein & Kant, 1973; Langevin et al., 1998;
Prentky et al., 1989). For example, Prentky et al. (1989) found that serial sexual murderers (n
= 25), compared to non-serial sexual murderers (n = 17), reported more voyeuristic (75% vs.
43%), fetishistic (71% vs. 33%), and transvestic (25% vs. 0%) sexual fantasies. Similarly,
Langevin et al. observed that 33% of 51 exhibitionists and 17% of 24 rapists reported fantasies
of children under 16 years old. It is also important to note that many sex offenders also use
non-deviant sexual fantasies (Cortoni & Marshall, 2001; Langevin et al. 1998; Looman, 1995;
Sheldon & Howitt, 2008; Turner-Moore et al., 2010). For instance, Langevin et al. found that
83% of heterosexual paedophiles reported using fantasies about females aged
21 and over, and 88% of sexual rapists admitted to using fantasies involving non-deviant
behaviours.
As previously mentioned, the prevalence of deviant sexual fantasies within nonoffender samples has also been reported. Crpault and Couture (1980) observed that, in a
sample of 94 non-offending men, 33% admitted to fantasies about raping a woman. Similarly,
Goldstein and Kant (1973) noted that 31% of 53 non-offending men reported fantasies of
force and sadistic acts. More recently, Williams et al. (2009) noted that, in a sample of 103
students, 68% reported at least one fantasy about sexual assault, 62% about sadism, and 62%
about bondage. These rates for rape-related fantasies are one of the highest within the
literature. Additionally, child-related sexual fantasies have also been reported in nonoffenders. For example, Briere and Runtz (1989) found that 9% of 193 students reported
sexual fantasies involving children. Similarly, Williams et al. (2009) report 13% of their 103
students reported fantasies about paedophilia. Using a larger sample of 531 students,
11
Becker-Blease et al. (2006) found that 18% reported sexual fantasies about children. Although
not as high as sex offenders, these prevalence rates are rather substantial and, thus, warrant
further research.
Caution should be taken, however, when interpreting the prevalence rates of deviant
fantasies as many of the studies suffer from a number of important problems and differences.
First, definitions of sexual fantasy are often missing and inconsistent across studies (Byers et
al., 1998). Second, ambiguity regarding the term deviant sexual fantasy can lead to
misleading interpretations. For example, in their abstract, Langevin et al. (1998) state that
only a third (33.3%) of their sex offender sample (n = 130) reported using deviant sexual
fantasies. From this figure, they suggest that deviant fantasies are of little aetiological
significance. However, on inspection of Langevin et al.s results, the 33.3% actually refers to
the prevalence of sexual fantasies about children (i.e., <16 years old), rather than deviant
fantasies in general. This makes for a rather misleading interpretation of their study.
Moreover, in addition to heterosexual (n = 14) and homosexual paedophiles (n = 23), their
overall sample was also comprised of rapists (n = 24), exhibitionists (n = 51), and incest
offenders (n =17), all of whom are less likely to use child-specific fantasies. Thus, it is not
surprising that the prevalence rate for child-related sexual fantasies was relatively low for the
overall sample. Indeed, when examined individually, the homosexual and heterosexual
paedophiles reported the highest rates for child-specific fantasies (43% and 57%,
respectively).
Langevin et al. did note the prevalence rates for sexual fantasies of deviant activities
for each of the five offender types (i.e., rapists, exhibitionists, homosexual paedophiles,
heterosexual paedophiles, & incest offenders). These deviant activities consisted of a number
of behaviours, such as exposing, forcing sexual contact, and beating a partner. However,
Langevin et al. did not provide an overall prevalence rate for the sex offenders as a combined
group in the way they did for the child-related fantasies. Thus, for the present chapter, the five
12
individual percentages were deconstructed. It was found that approximately 85 of the 130 sex
offenders (65%) reported using sexual fantasies involving deviant sexual activities. This figure
is of a higher level of concern than the 33.3% reported by Langevin et al. (1998).
Furthermore, if there are any sex offenders within the sample who report using only childspecific fantasies, they would be added to this prevalence figure making it even higher.
Unfortunately, there is no way to ascertain whether or how many sex offenders reported using
just child-specific fantasies. Thus, a prevalence rate for all of the deviant sexual fantasies
combined (i.e., sexual fantasies about deviant activities + sexual fantasies about children)
cannot be determined.
Nevertheless, providing prevalence rates for sex offenders as a homogenous group is
rather uninformative and can lead to divergent conclusions, as sex offenders show a marked
degree of heterogeneity (Robertiello & Terry, 2007). Similarly, sexual fantasy itself is a
heterogeneous phenomenon (Gee, Ward, & Eccleston, 2003). Thus, a single prevalence rate
for a combination of deviant fantasy themes is also likely to be misleading. For instance,
Langevin et al. found that 71% of heterosexual paedophiles reported fantasies involving
deviant activities. However, it is indiscernible which deviant activities they scored most
highly on (i.e., forcing sex, exposing). Thus, researchers should instead try to establish the
prevalence rates for various deviant fantasy themes (similar to Williams et al., 2009) for each
type of sex offender. This approach would allow for more appropriate comparisons and
provide more informative insights.
Another concern is that many studies differ in terms of how they measure sexual
fantasy. Most studies typically use questionnaires that ask how frequently certain fantasies are
imagined and/or used during masturbation. However, some questionnaires have more items
for a fantasy theme than others do. For instance, there are just two items on the
Wilson Sex Fantasy Questionnaire (Wilson, 1978) that are associated with child molestation
(Baumgartner, Scalora, & Huss, 2002), whereas the Paraphilic Sexual Fantasy Questionnaire
13
(ODonohue, Letourneau, & Dowling, 1997) has 14 items directly related to sexual fantasies
involving children. Thus, assuming respondents answer honestly, the latter questionnaire has
the capability of eliciting more information. Also, the period of time within which participants
are asked to state having used certain fantasies is often not provided or inconsistent across
studies. For example, some studies measure across ones lifetime (e.g., Sheldon & Howitt,
2008), whereas others specify a certain time period, such as within the last 12 months
(Becker-Blease et al., 2006) or the last 24 hours (Lussier, Proulx, & McKibben, 2001).
Arguably, the prevalence rates will vary between studies that differ with regards to the
specified time-period.
Finally, prevalence rates are likely to vary due to dishonest responding, in that, sex
offenders may simply be untruthful about the presence and use of deviant fantasies,
particularly since they are measured through self-report. Nevertheless, the prevalence rates
reported within the existing literature indicate that deviant fantasies are a concern.
1.4 Effects and functions of deviant sexual fantasy

Deviant sexual fantasy has been assumed to serve a number of functions. From examining the
literature, the two most commonly discussed functions are its capability to elicit and strengthen
sexual arousal towards deviant themes (Kalmus & Beech, 2005; Laws & Marshall,
1990; McGuire et al., 1965) and act as a form of mental rehearsal for deviant sexual behaviour
(Deu & Edelmann, 1997; MacCulloch, Snowden, Wood, & Mills, 1983; Pithers et al., 1988;
Prenkty et al., 1989; Ward & Hudson, 2000). The former function clearly provides the basis
for Leitenberg and Hennings (1995) arousal-focused definition of sexual fantasy, as well as
theories (McGuire et al., 1965) and treatment strategies (Laws & Marshall, 1991) for deviant
interests. The latter function has been useful for understanding why certain offenders use
certain modus operandi (Leclarc et al., 2006; Bourke et al., 2012) and, as such, has been
14
of interest to those in the offender profiling field (e.g., Gee & Belofastov, 2006; Hazelwood &
Michaud, 2001).
However, other research indicates that sex offenders fantasies can also serve other
functions. For example, in a qualitative study, using a sample of 24 sex offenders, Gee et al.
(2003) identified four functions that sex offenders fantasies can serve. First, they
corroborated the link between deviant sexual fantasies and sexual arousal. Specifically, they
found that such fantasies can both induce and enhance pre-existing arousal levels in sex
offenders. This finding is supported by previous research. For example, in a study designed to
recondition the sexual interests of 12 patients (which included 5 paedophiles and 2 rapists),
Marshall (1973) showed that engaging in deviant fantasies (pre-treatment) induced high levels
of sexual arousal. Similarly, Palk and O'Gorman (2004) found that when sex offenders
listened to audio content that matched their deviant sexual fantasies, their sexual arousal
levels increased.
Second, Gee et al. (2003) found that sexual fantasies allow sex offenders to model
future behaviour, either implicitly (i.e., by re-living or rehearsing past experiences) or
explicitly (i.e., through the simulation of future scenarios). This function - which they labelled
modelling - corroborates the presumed link between fantasy and offence rehearsal/planning.
For example, MacCulloch et al. (1983) found that sadistic sex offenders would engage in
behavioural try-outs of their fantasies, before committing the eventual index offence.
Leclerc, Beauregard, and Proulx (2008) found that, in a sample of adolescents who offended
against children, the use of deviant fantasies prior to the offence was associated with the use
of a specific modus operandi. Most recently, in a qualitative study, Bourke et al. (2012) found
that experienced (or expert) sex offenders used deviant fantasies to aid in the planning of
their offences. More specifically, the offenders would envision practiced and rehearsed
scenarios (which included specific dialogue and sexual actions) in order to visualise and
simulate successfully achieving their desired goal.
15
A third function Gee et al. (2003) identified was termed affect regulation and
involved offenders using sexual fantasies as a way of eliminating negative affect; elevating an
ambivalent affective state (i.e., boredom); and enhancing a pre-existing positive mood/affect.
This supports previous research which has shown that deviant sexual fantasies (in both child
abusers and rapists) are associated with negative affect (Cortoni & Marshall, 2001; Looman,
1995; McKibben, Proulx, & Lusignan, 1994; Proulx, McKibben, & Lusignan, 1996). In other
words, many sex offenders have been found to report experiencing negative affective states,
such as loneliness, humiliation, and inadequacy, either prior to or in conjunction with the use
of deviant sexual fantasies.
The fourth and final function reported by Gee et al. was termed coping, as offenders
were found to use sexual fantasies to deal with real or perceived external and/or internal
threats. For example, by distorting and manipulating reality within their mind, offenders were
able to justify their offending behaviour and gain control over their internal inhibitors. In
addition, sexual fantasies also helped some offenders escape or dissociate from a situation or
the reality of a situation. This function is supported by a previous study showing that some
offenders use deviant (and non-deviant) fantasies to cope with stressful and problematic
events (Cortoni & Marshall, 2001).
Gee et al.s study clearly provides some influential findings. Moreover, there has since
been some further corroboration. For example, preliminary results from Turner-Moore and
Waterman (2011), using a sample of 93 sex offenders, showed that sexual arousal, affect
regulation, and modelling were reported as functions for sexual fantasies about children and
coercion. For both types of fantasies, sexual arousal was the most common function. Also,
some offenders reported substitution for action to be a function of coercive fantasies, which
was also found in Gee et al.s study, albeit as a sub-feature of modelling.
It should be noted, however, that Gee et al.s sample (n = 24) was comprised of 20
child abusers, with only four rapists making up the rest of the sample. Thus, it is not clear
16
how well the findings generalise to other types of sexual offenders, such as exhibitionists or
sexual murderers. Also, it would have been useful if the authors had provided information
about the motives of the offenders, as offenders with different motives use sexual fantasies for
different purposes. For example, based on Ward, Hudson, Keenans (1998) Self-Regulation
Model, sex offenders with offence-avoidant goals are likely to use sexual fantasies for
different purposes (i.e., action substitution) to those harbouring offence-approach goals (e.g.,
offence preparation). Also, the deviant fantasies of intrafamilial abusers have been found to
serve different purposes to those of extrafamilial abusers1 (Cortoni, Proulx, Paquette,
Longpr, & Coutre, 2009). Another related point is that Gee et al. (2003) do not indicate
whether more than one function can be experienced by a single offender, as has been
suggested by Langevin et al. (1998, p. 323). For example, a recent case report describes how
the deviant fantasies of a serial rapist served more than one function (Carabellese, Maniglio,
Greceo, & Catanesi, 2011).
Finally, it is possible that other potential functions may also exist. For example, it has
been argued that sexual fantasies are often an offenders preferred way of attaining certain
non-sexual needs (Ward, Mann, & Gannon, 2007). For example, Marshall and colleagues
(e.g., Marshall & Barbaree, 1990; Marshall & Marshall, 2000) proposed that deviant fantasies
provide some offenders with a sense of mastery and power that cannot be attained in real-life.
Indeed, a few studies offer some support for these assertions. For example, Baumgartner et
al., (2002) found that, relative to non-sex offenders, child abusers reported more fantasies
involving ...a position of control rather than a position of submission (p. 26). In the
aforementioned case report involving a serial rapist (Carabellese et al., 2011), the offender
was reported to experience feelings of grandiosity and omnipotence as a result of using sexual
fantasies about forced sex, coercion, and dominance. Finally, Cortoni et al. (2009) found that
Extrafamilial abusers offend against children outside of their family, whereas intrafamilial abusers offend
against their own relatives.
17
intrafamilial child abusers mainly use deviant fantasies as a way of compensating for a low
sense of self-worth. Thus, from these findings, it seems that deviant fantasies can also serve a
compensatory function.
Other functions have also been suggested. For example, Langevin et al. (1998)
proposed that deviant fantasies may function as wish fulfilment for some offenders.
Interestingly, there is some limited evidence for this. For instance, Tidefors and Kordon
(2009) thematically analysed group therapy transcripts of four rapist-murderers and found that
a common theme within their fantasies was a wish to be understood (p. 8). This involved
wanting their victim to see them for who they are and providing the offender with what they
need. The offenders also held a wish for the perfect rape as evidenced in their fantasies and
offence planning (p. 9). This limited support suggests that wish fulfilment may be a function
worth investigating further.
In sum, deviant sexual fantasies serve a multitude of functions. More research now
needs to be done to understand them further, especially given the heterogeneity among sex
offenders and the ways in which the functions may influence offending behaviour. Given their
multifaceted nature, it is important to have a sound theoretical understanding of the
mechanisms that underlie and drive deviant sexual fantasies. Thus, the following section
examines a number of theoretical accounts in order to establish what is conceptually
understood about deviant sexual fantasies.
1.5 Theories of deviant sexual fantasy

A thorough search of the literature indicated that there is no existing theory that explains
deviant sexual fantasies or fantasising in general, although MacCulloch, Gray, and Watt
(2000) propose a theory for the aetiology of sadistic fantasies (see Section 1.5.3). Instead,
deviant sexual fantasy tends to be conceptually or theoretically discussed within the context of
explaining deviant sexual preferences, with a predominant focus on the content of the fantasy.
18
Thus, to gain any insight into what has been theoretically proposed about deviant sexual
fantasies, one must consult existing theories of deviant sexual preferences and piece together
the various propositions. The two main sexual preference theories that place deviant fantasy at
its core include McGuire, Carlisle, and Youngs (1965) sexual deviation hypothesis and Laws
and Marshalls (1990) conditioning theory. Thus, both of these theories will be examined to
see what they tell us about deviant fantasies, along with MacCulloch, Gray, and Watts (2000)
account of how sadistic sexual fantasies emerge.
1.5.1 McGuire, Carlisle, and Youngs (1965) account of deviant sexual fantasy
McGuire et al.s (1965) theory of deviant sexual interest is inspired by the early work of Binet
(1888), who postulated that deviant interests are learned during a single sexual experience.
However, McGuire et al.s account differs in an important way. By examining the case
histories of 45 men deemed to be sexually deviant, McGuire et al. theorised that the learning
process occurs after the initial sexual experience and in a more gradual manner. This is
because, for 75% of their sample, the initial sexual experience was reported to be important
only as far as providing a sexual fantasy for later masturbation. The authors argue that, because
the fantasy tends to be a memory of their initial sexual experience, it is likely to have a greater
stimulus value than other sources of fantasy (e.g., photos, stories). Drawing upon conditioning
principles, the authors theorised that by repeatedly masturbating to this deviant fantasy (before
ejaculation), the fantasy becomes even more sexually exciting. As a result, the individual ends
up deliberately choosing the deviant sexual memory as a masturbatory stimulus over other less
stimulating (non-deviant) sexual fantasies. McGuire et al. state that it is this process that
eventually leads to the development of a deviant sexual preference. Thus,
sexual fantasy is seen to be a central feature of McGuire et al.s theory; that is, an intervening
variable between the first deviant experience and the resulting deviant sexual interest (Abel
& Blanchard, 1974).
19
By explaining the role sexual fantasy plays within their theory of sexual deviance,
McGuire et al. indirectly provide some theoretical assumptions about sexual fantasies. First,
McGuire et al. regard sexual fantasies as a stimulus that can be conditioned to elicit sexual
arousal in the same way an external stimulus can. They further this point by suggesting that,
through conditioning processes, particular cues within the fantasy can become a dominant
focus if given emphasis during masturbation. Second, McGuire et al. regard sexual fantasy as a
sexual memory (p. 188), as deviant fantasies mainly depicted participants first sexual
experience. Third, McGuire et al. appear to regard sexual fantasies as something one
deliberately chooses to envision during masturbation, as opposed to a more intrusive form of
cognition. On this point, McGuire et al. provide two reasons for why deviant fantasies are
chosen over fantasies that are more readily available... and socially acceptable (p. 186). The
first is that the deviant fantasy is more sexually exciting. This is because, according to
McGuire et al., it is a memory of a real experience and so has a stronger initial stimulus
value as a masturbatory fantasy (p. 186). This stimulus value is increased through
masturbatory reinforcement until it gains dominance as the preferred fantasy. McGuire et al.s
second reason is that many sexual deviants (i.e., 50% of their sample) often believed they
could not engage in a normal sex life for a particular reason (e.g., physical or social
inadequacy). Thus, they turned to using deviant fantasies.
1.5.1.1 Evaluation of McGuire et al.s account on deviant fantasies

Despite being developed over 40 years ago, McGuire et al.s (1965) theory was, and still is,
regarded as highly influential in explaining deviant sexual preferences (Ward, Polaschek, &
Beech, 2006). The following section will evaluate the assumptions that the theory makes
about deviant sexual fantasy.
First, the idea that sexual fantasy is a stimulus that can be conditioned to produce a
response (i.e., sexual arousal) has some indirect support, both conceptually and empirically.
20
On a conceptual level, many behaviourists are in agreement that private (covert) events, such
as mental imagery, are forms of behaviour (Cautela & Baron, 1977; Roche & Barnes, 1998;
Skinner, 1953). This is based on the idea that private events share a functional equivalence
with public (overt) events (Day, 1969), meaning that they can undergo the same conditioning
processes. The only difference is that private events are inaccessible to other observers.
There is some research that supports this view. For example, Dadds, Bovbjerg, Redd,
and Cutmore (1997) examined the literature on mental imagery and classical conditioning and
found that that mental imagery can influence the outcome of classical conditioning and, in
some cases, act as a substitute for an actual unconditioned stimulus. Since sexual fantasy is a
form of mental imagery, this provides indirect support for McGuire et al.s assertions about
the conditioning of deviant sexual fantasies. Some tenuous support can be gleaned from case
studies involving conditioning-based techniques for treating deviant preferences. For
example, Kremsdorf, Holmen, and Laws (1980) found that having a paedophilic client
masturbate solely to non-deviant fantasies (i.e., directed masturbation) daily for eight weeks
led to an increase in penile arousal towards non-deviant stimuli. It could be argued here that
the non-deviant fantasy became conditioned to elicit sexual arousal. Indeed, the client
reported post-treatment that adult females were now his main source of arousal in fantasy.
Moreover, since the client showed increased arousal towards actual non-deviant stimuli, it can
be argued that the fantasy acted as a substitutive unconditioned stimulus. In sum, the available
literature provides some limited, indirect support for the idea that deviant sexual fantasy is a
stimulus that can be conditioned. More research is needed, however, particularly in the form
of well-controlled experimental studies.
McGuire et al.s assumption that sexual fantasy is a sexual memory is important as it
acknowledges the cognitive nature of sexual fantasies. For example, McGuire et al. state that,
since they are memories, sexual fantasies are subject to the usual psychological processes of
recall with the result that distortion and selection of cues take place (p. 186). This
21
assumption has some conceptual and empirical support. Conceptually speaking, Prentky and
Burgess (1991) argue that early childhood abuse can become encoded in memory and
retrieved in the form of deviant sexual fantasies. More recently, Ward and Beech (2006)
proposed that deviant sexual fantasies are the product of problems occurring in the
memory/perception neuropsychological system of the brain. On an empirical level, it has been
found that, for some offenders, early abuse and/or past sexual activities were the origin of
their deviant fantasies (Gee, Ward, Belofastov, & Beech, 2006), suggesting that memory is
playing a role. Also, Gee et al. (2003) found that many sex offenders sexual fantasies served
as way to recall and re-live their previous offences.
However, in spite of this support, other research would suggest that McGuire et al. are
only partly correct. For example, in their study on 16 sadistic sex offenders, MacCulloch et al.
(1983) found that 12 offenders first experienced non-aggressive sexual fantasies that changed
to sadistic content some years after the onset of masturbation (M = 5 years). This suggests that
their deviant sadistic fantasies may have originated from imagination. Moreover, in Gee et
al.s (2006) study, some sex offenders reported that their sexual fantasies originated purely
from imagination, as well as other external sources (e.g., pornography). These findings
indicate that deviant sexual fantasies are not always memories of previous experiences.
McGuire et al.s suggestion that deviant fantasising is a deliberate act has some
conceptual support. For example, Byers et al. (1998) state that sexual fantasies are
distinguishable from sexual thoughts on the basis of being deliberately and internally
generated, whereas sexual thoughts are triggered spontaneously by an external event. Some
have also alluded to the distinction within the sex offending literature. For example, Stinson
and Becker (2012) state that many sex offenders often confuse sexual fantasies with general
sexual thoughts. However, most of the work within the sex offending literature does not make
the distinction between deliberate sexual fantasies and intrusive, fleeting sexual thoughts, and
so there is little in way of empirical support.
22
In summary, McGuire et al. provide some theoretical suggestions about the nature of
deviant sexual fantasies. Although these suggestions have received little in the way of
empirical testing, research from other areas (both within and outside the sex offending field)
offer some corroboration. Therefore, although the theory is not fully comprehensive, it does
provide some insights into deviant fantasies that may prove useful for developing a more
comprehensive understanding.
1.5.2 Laws and Marshalls (1990) account of deviant sexual fantasy

In their theory of deviant sexual preferences, Laws and Marshall (1990) propose that sexual
deviance is acquired through the same mechanisms that are thought to produce more socially
accepted forms of sexual behaviour; that is, conditioning and social learning processes. They
also explain how these two modes of learning can result in the formation of deviant sexual
fantasies. These propositions will now be described.
With regards to conditioning, Laws and Marshall essentially elaborate upon McGuire et
al.s (1965) main assumption. That is, that sexual fantasy is an unconditioned stimulus that can
be classically conditioned to produce sexual arousal. They elaborate by arguing that any
stimulus associated with the original conditioned stimulus (including symbolic stimuli such as
fantasies) can elicit sexual arousal; a process known as second-order Pavlovian conditioning.
Thus, if an individual becomes conditioned to experience sexual arousal towards a rape scene,
he may also come to find other related stimuli sexually arousing, such as fantasies about the
rape scene. Laws and Marshall also add that operant conditioning reinforces the Pavlovian
associations that are formed.
Laws and Marshall also extend McGuire et al.s idea of cue selection by introducing
the concept of autoerotic influence. This involves deliberately selecting and pairing specific
sexual or nonsexual elements within a fantasy with masturbation so that they become
eroticised. Laws and Marshall state that this can lead to the creation of new deviant fantasies
23
or some variation of the original. For example, a man may deliberately focus on or introduce
the concept of control into his fantasies. Through conditioning processes, this element can be
erotised and become a predominant focus, perhaps even escalating to sadism.
A unique feature of Laws and Marshalls theory is their introduction of social learning
theory (Bandura, 1977), particularly the processes of: 1) participant modelling, where an
individual experiences and then copies the behaviour of another; 2) vicarious learning of a
non-participant behaviour, whether in vivo, print, or visual media; and 3) symbolic modelling,
where the behaviour and its effects are envisioned and elaborated within ones imagination.
Laws and Marshall state that the content for symbolic modelling can originate from any
deviant contact, random deviant fantasy, or exposure to textual or visual representations of
deviant behaviour (p. 221). However, regardless of its origin, the theory argues that symbolic
modelling is important in the development of deviant fantasies as it can lead to changes in the
original stimulus or to a completely new scenario.
On examination, Laws and Marshalls theory appears to hold some of the same
assumptions about deviant sexual fantasy as does McGuire et al.'s (1965) account. For
example, they view sexual fantasy as a stimulus that can be conditioned and also appear to
hold the view that fantasising is a deliberate act (as suggested by the process of autoerotic
influence). However, the theory differs in terms of how the deviant imagery originates.
Unlike McGuire et al., Laws and Marshall hold the view that deviant fantasies can reflect
more than just the memory of ones first sexual experience (e.g., deviant textual or visual
representations).
1.5.2.1 Evaluation of Laws and Marshalls account of deviant sexual fantasy

Laws and Marshalls conditioning theory clearly provides a more thorough description on
how deviant sexual fantasies may develop. Much of the same research that supports the
suggestions offered by McGuire et al. (1965) is also applicable here. For instance, the
24
application of second-order Pavlovian conditioning is supported by research showing that

mental imagery can work as a substitute for the actual stimulus (Dadds et al., 1997). However,
their claim that operant conditioning reinforces the deviant associations formed through
Pavlovian conditioning is harder to corroborate since, as Laws and Marshall note, the two
forms of conditioning are difficult to separate. For example, in the aforementioned case study
where Kremsdorf et al. (1980) used directed masturbation with a paedophilic client, it is
possible that operant conditioning also played a role. That is, the rate of non-deviant sexual
fantasy use may have increased because non-deviant fantasies had become classically
conditioned to produce sexual arousal, which itself became a reinforcer.
Laws and Marshalls idea of autoerotic influence, which involves deliberately
altering small aspects within a fantasy so that it changes or becomes more focused, is very
useful. For instance, it can account for the idiosyncratic differences observed in the content of
offenders deviant fantasies. This idea is supported by MacCulluch et al. (1983) who found
that nine sadistic sex offenders (out of 16) deliberately changed an element of their sexual
fantasy so that it continued to elicit sexual arousal. Although Laws and Marshall mention
reasons why a fantasy might be changed (e.g., to avoid boredom or ensure their fantasy
remains sexually arousing), they do not explain why certain themes (e.g., youthfulness,
control) or directly deviant elements (e.g., children) are introduced, changed, or focused upon.
This is something that needs to be researched further since a study failed to show that sex
offenders habituate to their deviant fantasies (Palk & OGorman, 2004). Moreover, recent data
indicates that far fewer sex offenders change the content of their fantasies compared to nonoffenders (Turner-Moore & Waterman, 2011).
Laws and Marshall also place a lot of emphasis upon the social learning process of
symbolic modelling when referring to deviant fantasies. However, this process of change
appears similar to the process of autoerotic influence and it is not clear whether they are
distinct. Nevertheless, there is some support for the assumptions Laws and Marshall (1990)
25
make about the source of symbolic modelling (i.e., any deviant contact, random deviant
fantasy, or exposure to textual or visual representations of deviant behaviour, p. 221). For
example, Gee et al. (2006) found that sex offenders sexual fantasies can originate from their
own abusive and non-abusive sexual experiences (cf. any deviant contact); from
pornography or the media (cf. exposure to textual and visual representations); and from
imaginative processes (cf. a random deviant fantasy).
In summary, Laws and Marshall provide more detailed insights into deviant sexual
fantasies. The assumptions made provide direct insight into how some of the processes related
to deviant fantasy operate, particularly those that shape fantasy content. They also introduce the
idea that sexual fantasies can originate from imagination, as opposed to just memory. However,
some of their assumptions lack a thorough explanation (e.g., why a certain cue is changed or
focused upon) and there is some overlap among concepts (e.g., change occurring via autoerotic
influence and change occurring during symbolic modelling). Thus, much like McGuire et al.
(1965), the theory does not provide a complete understanding of deviant sexual fantasy.
Moreover, while there is some indirect support for the assumptions made, on the
whole there is a need for more empirical support.
1.5.3 MacCulloch, Gray, and Watts (2000) account of (sadistic) sexual fantasies
A theory developed specifically to explain how sadistic fantasies develop has been proposed
by MacCulloch et al. (2000). The authors recognised that, while some authors found sadistic
fantasies are shaped over time (MacCulloch et al., 1983), there was no explanation for how
sadistic fantasies are initiated in the first place. Thus, drawing upon the notion of sensory
preconditioning (Brogden, 1939), they developed a theoretical account to tackle this question.
Sensory preconditioning (SP) refers to a learning process where associations between two
representations are formed, in the absence of reinforcement, by simply being presented
together or in close succession. Using the example provided by MacCulloch et al. (2000), if a
26
click is paired with a light, representations of the two stimuli will form (sensory
preconditioning). If the click is then paired with a shock, the click will be able to activate a
representation of the shock and produce a fear response (Pavlovian conditioning). Due to the
previous SP, the light will now appear to produce a fear response. This is because the light
activates a representation of the click, which activates a representation of the shock, which
elicits a fear response.
Drawing upon sensory preconditioning theory and research, MacCulloch et al. (2000)
propose that during and immediately after an episode of sexual abuse, a child may
concurrently experience high levels of sexual arousal and aggression. As per the assumptions
of SP, it is during this moment that an association will be formed between the internal
representation of aggressive feelings and sexual arousal. This association will be strengthened
with repeated abuse. As a result, whenever the individual feels aggressive, it will activate
feelings of sexual arousal, and vice versa. According to MacCulloch et al., this creates a
pathway to generating and envisioning sadistic sexual fantasies. Thus, according to this
account, if such an individual ruminates on wanting to hurt someone, they will also feel
sexually aroused as a result. This may cause the imagery to become used as a sexual fantasy.
1.5.3.1 Evaluation of MacCulloch et al.s (2000) account

MacCulloch et al. (2000) provide a novel and well-considered theory on how sadistic
fantasies arise. The main strength of the account is that it provides a novel explanation for
why an individual may chose to envision a certain sexual fantasy theme (i.e., aggression),
especially before they have developed a related sexual interest (i.e., sadism). However, the
theory also appears to suffer from a number of limitations. First, the authors base the
assumption that sexual arousal is experienced during childhood abuse on a single non-human
study. The same goes for the assumption that aggression is experienced during sexual abuse.
Second, despite being a central tenet of their theory, there is little to suggest that sexual sadists
27
experience repeated and chronic childhood sexual abuse. Instead, the literature refers more to
physical abuse (Brittain, 1970; Gratzer & Bradford, 1995). Third, the theory does not account
for the target of the behaviour, such as why one develops sadistic fantasies involving rape (i.e.,
Beech, Ward, & Fisher, 2006) when another generates sadistic fantasies involving children
(i.e., Firestone, Bradford, Greenberg, Larose, & Curry, 1998). Fourth, the theory has received
no direct testing of its assumptions since its publication. For example, the theory would
predict that a sample of sadistic fantasisers would report more incidences of repeated
childhood sexual abuse relative to a control group or that sadistic offenders are also likely to
feel sexual arousal when primed to experience aggression. Finally, while it was not
MacCulloch et al.s intention, it is not known whether the notion of SP can be applied to
explain other types of deviant fantasy content. This may be something worth considering,
since there is support that the sexual fantasies of other sex offenders can originate from their
own abuse (Gee et al., 2006).
From examining these theoretical accounts, it is clear that a number of important
insights regarding the nature and processes underlying deviant sexual fantasies have been
uncovered and alluded to over the past few decades. These insights have helped in
understanding deviant sexual fantasies and, thus, should continue to influence future
theoretical accounts of deviant sexual fantasies. While each theoretical account suffers from a
number of shortcomings in terms of explaining deviant sexual fantasies, it should be
emphasised that these result from their theoretical perspective and the topic upon which they
focus. For example, the accounts are based primarily on behaviourist principles and so ignore
many of the cognitive features associated with sexual fantasies. As Leitenberg and Henning
(1995) stated, although conditioning plays an important role, the development of a sexual
fantasy also involves other factors. Indeed, Marshall (1996) suggests that existing
conditioning theories of sexual preferences may be strengthened by translating them into
cognitive terms (p.190).
28
In terms of focus, MacCulloch et al.s (2000) account is constrained to sadistic sexual

fantasies. More importantly, McGuire et al.'s (1965) and Laws and Marshall's (1990) theories
are designed to explain deviant sexual preferences. As a result, deviant sexual fantasy is not
explicitly defined and does not receive the level of attention it would otherwise receive had
the theories been focused exclusively on deviant sexual fantasies. In addition, the focus of
these two theories means that the assumptions made about deviant sexual fantasy are largely
confined to its relationship with sexual arousal. This is a problem when considering the
multitude of functions deviant sexual fantasy can serve (see Section 1.4).
1.6 Conclusions
In this chapter, a comprehensive examination of the literature relating to deviant sexual
fantasies has been presented. Over the past few decades, a great deal has been learned about the
content, function, and correlates of deviant sexual fantasies. This work demonstrates that
sexual fantasies are a multifaceted phenomenon and one that is important in understanding the
psychology and offending behaviour of many sexual offenders. As such, the existing research
has helped inform clinicians about how to best assess and treat the deviant sexual fantasies in
sexual offenders.
In spite of this, there are many questions that have been ignored and/or left unanswered.
More importantly, these questions lie at the centre of some of the most important factors
associated with understanding deviant sexual fantasies, such as definition, prevalence, and
theoretical insight. Thus, by highlighting that a distinction should be made between
sexual thoughts and fantasies; that deviant sexual fantasies may be more prevalent than is
otherwise indicated; and that a cognitive understanding of deviant sexual fantasy is needed,
there is a strong case for developing a contemporary understanding of deviant sexual
fantasies.
29
CHAPTER 2
THE DUAL-PROCESS MODEL OF SEXUAL THINKING
Chapter Rationale
The literature reviewed in Chapter 1 indicates that deviant sexual fantasy is not a well
conceptualised construct. Moreover, an account of the cognitive mechanisms and processes
that give rise to deviant sexual fantasies is clearly lacking. Thus, by drawing upon relevant
research from cognitive psychology (e.g., associative/controlled processes; mental imagery;
working memory; episodic remembering/imagining) and tying it together with what is known
about sexual fantasies, a new a priori theoretical model has been developed; the Dual-Process
Model of Sexual Thinking (DPM-ST). In the present chapter, the cognitive principles
underlying the DPM-ST are first outlined, followed by a description of the theory itself. This
chapter provides the basis for the rest of the thesis, as the subsequent chapters focus on testing
a number of the DPM-STs core assumptions.
30
2.1 Introduction
According to the literature, the understanding of deviant sexual fantasies is lacking in
theoretical scope (Gee, Ward, & Eccleston, 2003; Sheldon & Howitt, 2008). In light of this, it
is argued that a cognitive account of deviant sexual fantasies and their associated processes is a
useful addition to the field. In this chapter, a theoretical model designed to explain the chief
cognitive processes and mechanisms associated with deviant sexual fantasies will be
described. This theory is termed the Dual-Process Model of Sexual Thinking (DPM-ST) and
was constructed by knitting together the insights of earlier theories (see Chapter 1) with
insights from relevant areas of cognitive psychology. These areas include: dual-process
accounts of cognition (i.e., associative and controlled processes); mental imagery; working
memory; and episodic memory and episodic future thought. Therefore, before outlining the
DPM-ST, the basis of the theory will first be discussed as it will help elucidate the rationale
behind the main assumptions of the theory.
2.2 A dual-process approach

As discussed in Chapter 1 (Section 1.2), researchers of sexual cognition have distinguished
externally triggered sexual thoughts from internally generated sexual fantasies (Jones &
Barlow, 1990). Thus, the DPM-ST seeks to encapsulate this distinction by adopting a dualprocess approach to explaining 'sexual thinking'. Human cognition has long been categorised
into two processing modes (Evans, 2008; Kahneman, 2011; Schneider & Shiffrin, 1977). One
mode is associative (also called automatic, implicit, impulsive) and involves the automatic,
rapid, and cognitively effortless processing of stimuli. As Smith and DeCoster (1999) state,
associative processing is not consciously experienced but the results of it are (i.e., a thought
popping into mind or an immediate gut feeling). The other mode is controlled processing
(also called rule-based, explicit, and reflective) and involves deliberate, slower, and more
effortful processing. This dual-processing conceptualisation of human cognition has become
31
widely embraced, extensively researched, and used to construct dual-process models for an
array of phenomena, including attitudes (Gawronski & Bodenhausen, 2006), sexual arousal
(Janssen, Everaerd, Spiering, & Janssen, 2000), desire (Kavanagh, Andrade, & May, 2005),
and decision making (Kahneman, 2011).
With regards to sexual thinking (i.e., sexual thoughts and fantasies), Kavanagh et al.s
(2005) theory of desire is of particular interest as it distinguishes between automatic intrusive
thoughts and a controlled process they term cognitive elaboration' (defined as mentally
embellishing or 'going beyond' the stimulus). Kavanagh et al. propose that intrusive thoughts
about an appetitive target arise when associations are activated in response to an external or
internal stimulus. Such thoughts are fleeting unless they are awarded further attention (e.g., if
they elicit a strong affective reaction). If a thought is attended to further, Kavanagh et al. argue
that the controlled process of cognitive elaboration will occur, which involves holding
information about the target in working memory, typically in the form of mental imagery.
Byers et al. (1998) state that sexual thoughts and sexual fantasies can be distinguished on the
basis of being externally triggered and spontaneous (i.e., associative) and internally generated
and purposeful (i.e., controlled). This has some parallels with the propositions made by
Kavanagh et al. (2005). Thus, by knitting Byers et al.s (1998) proposition with Kavanagh et
al.'s theory, the DPM-ST views sexual thoughts and fantasies as distinct (yet interrelated)
phenomena that operate on the basis of associative and controlled processes, respectively.
These two processes will now be discussed in a little more depth.
2.2.1 Associative processes

In their Associative-Propositional Evaluation (APE) model of attitudes, Gawronski and
Bodenhausen (2006) state that associative processes are responsible for immediate affective
reactions towards an object or experience. Hence, encountering a stimulus automatically
elicits an affective reaction due to the activation of particular associative links in ones
32
cognitive network. According to Gawronski and Bodenhausen, the pattern of activation is

dependent upon: (a) the pre-existing structure of associations in memory; and (b) the
particular context in which the stimulus is encountered. For example, for a child abuser in a
certain context, the stimulus child may activate an associative pattern that includes concepts
such as daughter but not sex. As a result, this may elicit a nurturing affective response.
However, in another context, the associative pattern activated by children may include
activating the concept of sex but not the concept of daughter, which may lead to a
sexualised affective response (i.e., arousal). Thus, the same stimulus (i.e., children) has the
capability of eliciting different associative processes and, therefore, different affective
reactions (Gawronski & Bodenhausen, 2006).
There is some support for the view that people hold sex-related associations in longterm memory. For example, a growing body of evidence using cognitive tasks suggests that
sex-related associations are held by both normophilic populations (Snowden, Wichter, & Gray,
2008) and sex offender populations (Babchishin, Nunes, & Hermann, 2013). Moreover, some
have alluded to the idea that an individual's distorted associative network (often called a
'schema') can give rise to deviant sexual thoughts and fantasies. For example, Johnston, Ward,
and Hudson (1997) argued that intrusive deviant sexual thoughts are more chronically
accessible for preferential sex offenders as their offending begins earlier in life, compared to
situational offenders who begin offending in adulthood. In other words, the associations that
give rise to unwanted deviant sexual thoughts in preferential offenders are stronger due to
having been formed earlier in life and, therefore, activated more often. In addition, Mann and
Beech (2003) stated that distorted schemas interact with environmental variables to produce
thoughts supportive of sexual offending (e.g., I am attracted towards this little girl).
Similarly, Beech and Ward (2004) proposed that deviant fantasies are an acute dynamic
(changeable) risk factor that arises when an offenders core set of schema interact with
specific triggers.
33
Another important feature of associative processes is that they are independent of truth
values (Gawronski & Bodenhausen, 2006). In other words, affective reactions can be
activated regardless of whether they are endorsed by explicit (conscious) attitudes. However,
Gawronski and Bodenhausen state that automatic reactions typically provide people with the
basis for their explicit attitudes. In other words, the default mode of evaluative judgment is
affirming a proposition based on an affective reaction (e.g., affirming I dislike X based
upon a negative affective reaction towards X). However, in cases where the default
proposition is inconsistent with other momentarily considered propositions and explicit
attitudes (e.g., It is not right to judge someone before knowing them. X may be a nice
person), the default proposition is liable to be rejected. As Gawronski and Bodenhausen
note, the idea that affective reactions and endorsed judgments can differ explains the
discrepancies often found between implicit attitudes (i.e., those implied by scores on an
indirect measure) and explicit attitudes (i.e., those measured using self-reports).
There is some evidence to support this validation process in the context of sexual
cognition. For example, 'sexually intrusive thoughts' (SITs) have been defined as "unwanted,
personally unacceptable thoughts that are definitely uncharacteristic of one's usual habits and
beliefs" (Renaud & Byers, 2005; p. 254). Moreover, a number of studies have documented
that SITs can be appraised as negative or distressing (Byers et al., 1998; Clark, Purdon, &
Byers, 2000, Renaud & Byers, 1999, 2005; Wetterneck, Smith, Burgess, & Hart, 2011). There
is also some indication that some sexual offenders negatively appraise their sexual
thoughts. For example, in a case study, an intrafamilial child abuser reportedly complained of
"being troubled by pedophilic thoughts every day or two" (Edwards, 1972; p. 57). Also,
Johnston et al. (1997) stated that situational sex offenders (who do not harbour a deviant
sexual preference) are likely to experience deviant sexual thoughts as unwanted. As a
consequence, they are more likely to be motivated to try and control or suppress the thought.
34
Knitting together ideas from Gawronski and Bodenhausens (2006) APE model and
Kavanagh et al.s (2005) theory of desire, the DPM-ST proposes that a spontaneous deviant
sexual thought (e.g., I am sexually attracted to this child) is the result of a default
proposition (i.e., immediate affective response), ensuing from an association that has been
activated by a relevant stimulus (e.g., a smiling child). Since they are based upon associative
processes, spontaneous thoughts are fleeting unless they are attended to or retained in working
memory; a controlled process termed elaboration (Kavanagh et al., 2005).
2.2.2 Controlled processes

Controlled processing involves the conscious, deliberate, effortful, and goal-directed
manipulation of information in working memory (Baddeley, 1996; Bargh, 1994). Working
memory is a limited capacity system where task-relevant information is retained and
manipulated during the performance of a cognitive task (Baddeley & Hitch, 1974). According
to Kavanagh et al. (2005), the cognitive elaboration of a spontaneous appetitive thought is an
effortful controlled process requiring working memory resources. For example, elaboration
can involve the search and retrieval of episodic information from memory. Indeed, previous
research indicates that this specific cognitive operation requires working memory resources
(Moscovitch, 1994; Rosen & Engle, 1997). For example, Moscovitch (1994) found that the
effortful retrieval of information from long-term memory can be impaired by increasing the
cognitive load placed on working memory. Also, recent research has shown that traumatic
episodic memories can also be impaired by taxing working memory (Gunter & Bodner, 2008;
van den Hout, Muris, Salemink, & Kindt, 2001).
According to Kavanagh et al. (2005), a central tenet of the elaborative process is the
involvement of mental imagery. Mental imagery is defined as a perceptual experience that
occurs in the absence of a physical stimulus (Finke, 1989). However, mental imagery is not a
single, undifferentiated activity but instead involves an array of cognitive processes working
35
together (Kosslyn, 1995). Moreover, many of these processes, such as the generation and
retention of imagery-related information, require the working memory resources (Baddeley &
Andrade, 2000; Keller, 2012; Moulton & Kosslyn, 2009). This is demonstrated by studies
showing that visual mental imagery (or tasks reliant on visual mental imagery) becomes
impaired when the component of working memory responsible for processing visual
information (i.e., the visuospatial sketchpad) is concurrently taxed. For example, Logie (1986)
showed that looking at visual displays can impair memory performance based on visual
imagery. Similarly, Baddeley and Andrade (2000) found that when a visual task is performed
concurrently with visual mental imagery, the visual mental imagery becomes impaired (i.e.,
rated as less vivid). Thus, research suggests that elaborative forms of mental imagery can be
regarded as a top-down controlled process (Blair, Ma, & Lenton, 2001; Keller, 2012).
While no research has been conducted to directly demonstrate that sexual fantasising
is a deliberate, controlled process, there exists some indirect support. First, sexual fantasies
are defined as a form of mental imagery (Leitenberg & Henning, 1995), suggesting that
factors necessary for other forms of mental imagery are also at play when sexually
fantasising. Also, there are some who hold the view that deviant sexual fantasising is a
conscious and deliberate act (Russell, Sturgeon, Miner, & Nelson, 1989). Moreover, Gee et al.
(2006) found that the frequency and duration of a sexual fantasy can be controlled by a sex
offender, depending on the extent to which they actively or passively regulate the process.
This may be suggestive of a controlled process.
A final point to note is the bidirectional nature of associative and controlled processes.
As described, activated associations affect controlled processing by providing the basis for a
default proposition, which can lead to cognitive elaboration. However, controlled processes
can also influence associative processes (Gawronski & Bodenhausen, 2006). For instance,
Blair et al. (2001) state that mental imagery has the capability of increasing the accessibility
of related mental representations, describing it as a powerful method of priming (p.829).
36
Thus, if a rapist envisions himself forcing sex on a woman, concepts such as women, sex,
power, and the self may become activated. As a consequence, associations between these
concepts have the potential to form, such as the self becoming associated with power.
These associative processes may explain the non-sexual effects that sex offenders experience
through fantasising, such as a sense of mastery and control (Cortoni & Marshall, 2001; Gee et
al., 2003; Carabellese, Maniglio, Greco, & Catanesi, 2011 Marshall & Barbaree, 1990).
2.3 Remembering and imagining episodic events

The literature suggests that deviant sexual fantasies can be constructed from memories of
previous sexual experiences as well as from purely imagined sexual scenarios (Beckett, 1994;
Gee et al., 2006). This indicates that understanding the processes that underpin remembering
and imagining episodic events may be useful in understanding sexual fantasising. The former
relies upon episodic memory, whereas the latter is reliant upon future episodic thought; that
is, the ability to envision a personal future event (Atance & ONeil, 2001).
A growing body of research indicates that future episodic thought is related to our
ability to recall episodic memories about our past (Schacter, Addis, & Buckner, 2008;
Szpunar, 2010). For example, Szpunar and McDermott (2008) found that imagining a future
event in a familiar setting (requiring information sampled from episodic memory) was rated
as more vivid than future events in a novel setting. Also, studies using functional magnetic
resonance imaging (fMRI) have shown that imagining future events shares many of the same
neural correlates as does remembering past episodic events (Addis, Wong, & Schacter, 2007).
According to Schacter, Addis, and Buckner (2007), these brain regions comprise a core
network that is responsible for integrating information from past experiences and using it to
construct mental simulations about possible future events. Some refer to this capability as
mental time travel (Corballis, 2011).
37
The work described above has yet to be applied to topic of remembering and
imagining sexual episodic events. However, there are a few studies that suggest sexual
offenders draw upon and recombine their episodic memories to create imagery for sexual
fantasies. For example, in a case study, Marshall (2006) describes an exhibitionist who,
during masturbation, typically imagined one of his exposing victims approaching him and
engaging in consenting intercourse (p. 17). Arguably, this demonstrates episodic information
(i.e., about an offence) being recombined to create a novel scenario. Also, Cortoni, Proulx,
Paquette, Longpr, and Coutre (2009) found that some child abusers create deviant sexual
fantasies that reflect their early sexual experiences suggesting that past memories are being
used as imagery for sexual fantasies. In their qualitative study, Gee et al. (2003) found that
some sexual offenders report using sexual fantasies to relive past offences, whereas others
simulate future behavior in their sexual fantasies. Arguably, the former involves retrieving
episodic information, whereas the latter could involve retrieving and recombining episodic
information to form a novel future scenario. This idea is supported by a recent study, which
found that 'expert' (or competent) sexual offenders draw upon their past offences when
constructing sexual fantasies that function as a mental simulation for a future offence
(Bourke, Ward, & Rose, 2012). From this, the authors propose that deviant sexual fantasies
can be viewed as a form of mental time travel that can aid offenders in devising the best
possible offence plan.
2.4 Outlining the DPM-ST

Having outlined the various cognitive theories, principles, and empirical findings relevant to
understanding sexual fantasies, the following section will describe how they were knitted
together to construct the Dual-Process Model of Sexual Thinking (DPM-ST). A graphical
representation of the theory can be seen below in Figure 2.1. It should be noted that, since a
large proportion of the theory is based upon literature pertaining to normative human
38
cognition, the DPM-ST is likely to apply to all sexual thoughts and sexual fantasies regardless
of their content. However, for the purposes of this thesis, the model will be discussed within
the context of deviant sexual thoughts and fantasies.
39
External or internal cue/trigger
Sex-related
associations
Associative
component
Immediate affective
reaction
Mental imagery
activates and
strengthens other
associations in
memory
SPONTANEOUS
SEXUAL THOUGHT
Validation
Explicit
attitudes
Strong affective response;

individual differences;
SEXUAL FANTASY
(in working memory)
Episodic
information
about sexual
experiences
Controlled
(elaborative)
component
Retrieval
Retrieval and
recombination
Script readily
accessible for
retrieval
during future
sexual fantasising
Remembered
sexual event
Imagined
sexual event
Incorporation
External
stimuli
Conditioning and
encoding processes
Specific fantasy
script/theme
stored in memory
Figure 2.1: Schematic representation of the Dual-Process Model of Sexual Thinking
40
2.4.1 Associative processes and sexual thoughts

According to the DPM-ST, most people associate sex with a number of possible concepts.
These concepts can include categories of people (i.e., men/women; adults/children;
behaviours (i.e., oral sex, spanking); objects (i.e., lingerie, fetishes); and emotions (i.e.,
intimacy, enjoyment). These associations are likely to guide how, when, and what stimuli
people regard as sexually relevant and, as such, are likely to form the basis of sexual schema
(Andersen, Cyranowski, & Espindle, 1999; Beech, Bartels, & Dixon, 2013). The DPM-ST
proposes that when a certain external stimulus is encountered (e.g., a smiling child), sexrelated associations held in memory (e.g., a child-sex association) are likely to become
automatically activated. This associative process will be dependent upon the cue, context, and
pre-existing pattern of associations (Gawronski & Bodenhausen, 2006). Once the association
has been activated, it will create an immediate affective response (Gawronski &
Bodenhausen, 2006; Janssen et al., 2000), which will be immediately translated into a
spontaneous sexual thought (e.g., I am sexually attracted to this child). Thus, the thought
will feel as though it just popped into mind (Beevers, 2005) and will only be fleeting unless
given more deliberate attention (Kavanagh et al., 2005).
2.4.2 From thought to fantasy

The DPM-ST proposes that the likelihood of a sexual thought being elaborated upon (in the
form of a sexual fantasy) will be dependent upon a number of factors. First, in line with
Kavanagh et al. (2005), the sexual thought will have to elicit a strong affective response, such
as sexual arousal or excitement. Indeed, Johnston et al. (1997) note that sexual thoughts are
less likely to be suppressed if strong feelings of sexual arousal are experienced. Second, a
sexual thought is more likely to be elaborated upon if it is congruent with explicit attitudes
and/or other momentarily considered propositions (Gawronski & Bodenhausen, 2006). For
example, a deviant sexual thought (i.e., This child wants sex with me) resulting from a
41
childs ambiguous gesture will be regarded as congruent or 'true' if it is consistent with

distorted explicit attitudes (i.e., Children enjoy sex with adults). On the other hand, the
same thought will be rejected and negatively appraised if it is inconsistent with other endorsed
attitudes or considered propositions (i.e., Children are seriously harmed by sex with adults).
This is more likely to occur if a sex offender has successfully undergone cognitive- behavioural
therapy to correct their distorted attitudes.
Third, a sexual offender who has to abide by a relapse prevention plan may be less
likely to elaborate upon a fleeting sexual thought, as engaging in a deviant sexual fantasy
(referred to as a 'lapse') can cause a negative feeling referred to as the 'abstinence violation
effect' (Russell et al., 1989). Fourth, the elaboration of a sexual thought may simply be
inappropriate in certain situations (e.g., at work). Finally, elaborating upon a sexual thought in
the form of a sexual fantasy is likely to be affected by various individual differences. For
example, an offender is more likely to use a deviant sexual fantasy if they are more sexually
preoccupied or if they have a greater working memory capacity. They are also more likely to
elaborate upon a sexual thought if they are able to effectively form vivid mental imagery
(Smith and Over, 1987). Also, offenders who are more fantasy prone in general (Wilson &
Barber, 1981) and who have more dissociative experiences are likely to engage in deviant
sexual fantasies more often than those who are less fantasy prone and who do not regularly
experience dissociative states.
2.4.3 The controlled process of sexual fantasising

According to the DPM-ST, sexual fantasising is an elaborative controlled process that is
deliberate, effortful, and goal-oriented2. During the process, information is retrieved from
episodic memory and is either: 1) reconstructed as a remembered sexual event; or 2)
2
There are occasions when sexual fantasising can occur in the absence of a preceding sexual thought, such as
when asked to describe or write a sexual fantasy on instruction (i.e., in research or clinical practice).
42
recombined to create a novel imagined sexual event. This aspect demonstrates that sexual
fantasising can incite mental time travel, which can lead some offenders to cognitively
reflect on a previous offence or possible future offences (Bourke et al., 2012).
Since it involves the retrieval, manipulation, and retention of target-related
information (i.e., children, deviant behaviours) in working memory, the act of sexual
fantasising is likely to compete with other cognitive tasks for working memory resources.
This means that, during times of cognitive load (i.e., when stressed), sexual fantasising will be
a difficult cognitive operation to perform. As a result, information from external sources (e.g.,
pornography) may be incorporated into the imagined fantasies (Gee et al., 2006; Goldstein &
Kant, 1973). Indeed, evidence would suggest that some sex offenders use the Internet to fuel
their deviant sexual fantasies and interests (Kloess, 2013; Long, Alison, & McManus, 2013).
This use of the external world to perform a task that requires working memory is known as
off-loading in the domain of embodied cognition (Wilson, 2002). Furthermore, the use of
external world has also been highlighted as an important point to address when dealing with
sexual offenders cognition (Ward & Casey, 2010).
Another proposition of the DPM-ST refers to the cognitive-based consequences of
sexual fantasising. One example is that sexual fantasising will cause the initially activated
associations to strengthen further, essentially creating a feedback loop. Also, other associations
may also become activated, since mental imagery has been shown to be a powerful form of
priming that can cause other associations to be activated (Blair et al., 2001). For example,
sexual fantasising may lead other non-sexual associations to become activated, such as those
between the self and power or the the self and children. This would provide a cognitive
explanation for how sexual fantasies provide sexual offenders with non- sexual benefits, such
as a sense of mastery/control or a sense of emotional congruence with children, respectively.
43
A second cognitive effect relates to the encoding of fantasy-related information. For

example, the repeated use of a deviant sexual fantasy may lead to the formation of an
association between the concept of 'sexual fantasy' and the deviant theme that is depicted in the
fantasy. That is, some child abusers who frequently use sexual fantasies about children may
come to associate the concept of 'sexual fantasy' with 'children'. Also, the specific pattern of
behaviour within a repeatedly used deviant sexual fantasy is likely to become encoded in
memory as a cognitive script (Huesmann, 1998; Ward & Hudson, 2000). This will increase
the likelihood that it will be retrieved during future instances of elaboration (Huesmann,
1998). In other words, due to being more cognitively accessible, an encoded cognitive script
is more likely to be brought to mind when an offender elaborates upon a sexual thought, as it
requires less cognitive effort than generating a new scenario.
2.5 Conclusion
The aim of this chapter has been to describe the basis of the DPM-ST as well as the DPM-ST
itself. The literature that was reviewed in Chapter 1 helped guide what aspects of cognitive
psychology would be most useful in devising the DPM-ST. It also highlighted the areas within
the literature that are in need of clarification, such as distinguishing fleeting sexual thoughts
from elaborate sexual fantasies. These aspects and insights were incorporated into the DPMST so that, not only does the theory draw together what has already been theorised and
empirically uncovered about deviant sexual fantasies, but it also provides new assumptions
that have never been applied to the construct before. It is because of this that the DPM-ST has
the potential to open up new and exciting avenues for future research and treatment. However,
the core assumptions of the DPM-ST need to be tested and corroborated in order for these
implications to hold. Fortunately, the DPM-ST is able to generate a number of testable
hypotheses; a number of which are tested in the following chapters of this thesis. These
include testing whether:
44
Child abusers hold sex-related associations that show a relationship with

corresponding deviant sexual fantasies (Chapter 3).
Individual differences (i.e., in the ability to fantasise generally) and congruent explicit
attitudes influence the use of deviant sexual fantasies (Chapter 4).
Memory and imagination-based sexual fantasies require the resources of working

memory (Chapter 5).
Deviant sexual fantasies prime other associations related to the content of the fantasy
(Chapter 6).
Child abusers who repeatedly use deviant fantasises associate the concept of sexual
fantasy with the theme of their deviant sexual fantasy (i.e., children) (Chapter 7).
45
CHAPTER 3
INVESTIGATING SEX-RELATED ASSOCIATIONS IN CHILD ABUSERS
Chapter Rationale
As discussed in Chapter 2, the Dual-Process Model of Sexual Thinking states that when a
relevant stimulus is encountered, sex-related associations are activated leading to the
automatic generation of a fleeting deviant thought and possible sexual fantasy use. Therefore,
the DPM-ST rests upon the idea that sex-related associations are stored within individuals'
associative memory. This chapter describes a study aimed at corroborating the presence of
distorted associations in child abusers and investigating whether such associations show a
relationship with self-reported deviant fantasies.
46
3.1 Introduction
A core aspect of the Dual-Process Model of Sexual Thinking (DPM-ST) is that people hold
sex-related associations within their associative networks. Since the DPM-ST proposes that
these associations - when activated - give rise to spontaneous sexual thoughts and subsequent
sexual fantasies, the assumption can be viewed as a foundational component of the DPM-ST.
As such, examining sex-related associations is a logical starting point for corroborating the
theory.
One way to infer the strength of a particular association is to use a reaction-time task
that measures the speed in which an individual responds to stimuli congruent with the
association. This is because stronger associations are thought to influence a person's reaction
time to related stimuli (Mihailides, Devilly, & Ward, 2004). The most widely used reactiontime (or indirect) measure is the Implicit Association Task (IAT; Greenwald, McGhee, &
Schwartz, 1998). In brief, the IAT is a blocked reaction-time task designed to measure how fast
words and/or pictures (i.e., orgasm and school) are categorised into the correct superordinate
category (i.e., Sex and Child, respectively). One of the blocks involves pairing two
congruent categories (i.e., Sex and Adult) on the same key, while another block involves
pairing two incongruent categories (i.e., Sex and Child) on the same key. To assess association
strength, researchers measure how fast participants categorise stimuli into the congruentlypaired category relative to the incongruently-paired category. In the example above, child
abusers would be expected to be faster than non-offenders during the incongruent block (with
child and sex paired) relative to the congruent block.
Over the past decade, research using the IAT has provided support for the assertion
that people hold automatic sex-related associations. For example, heterosexual and
homosexual participants have been found to hold associations that correspond to their sexual
orientation ( Ciardha & Gormley, 2012; Snowden, Wichter, & Gray, 2008; Snowden &
Gray, 2013). Moreover, child abusers have been found to demonstrate stronger associations
47
between children and sex than non-child abusers (Banse, Schmidt, & Clarbour, 2010; Brown,
Gray, & Snowden, 2009; Gray, Brown, MacCulloch, Smith, & Snowden, 2005; Mihailides et
al., 2004; Nunes, Firestone, & Baldwin, 2007; Steffens, Yundina, & Panning, 2008). Although
these findings are highly encouraging, they cannot be used to make entirely conclusive
statements about sex-related associations. This is because the IAT is based on a
relative, rather than absolute, paradigm. Thus, researchers are unable to rule out the possibility
that the results on child-sex IATs are due to reduced adult-sex associations or stronger adultnot sex associations (Snowden, Craig, & Gray, 2011).
In an attempt to corroborate the assumption that child abusers hold stronger child-sex
associations than adult-sex associations, Hempel, Buck, Goethals, and van Marle (2012) used
the Single-Category IAT (SC-IAT; Karpinski & Steinman, 2006). The SC-IAT is similar to
the original IAT except it has no opposing category for one of the concepts being tested. In
Hempel et al.s study, the opposing category of not sex was omitted. The results showed that
child abusers responded faster when child and sex-related words were categorised using the
same response key. Conversely, the non-abusers and adult sex offenders (i.e., 7 sexual
assaulters, 3 rapists, 2 exhibitionists, and a frotteur) were faster when adult and sex shared
the same key. While informative, these results do not conclusively indicate that child-sex
associations were stronger than adult-sex associations, as it is possible that the results were
due to decreased adult-sex associations; a point acknowledged by Hempel et al., (2012).
Unfortunately, like the original IAT, the SC-IAT does not offer the capability of reliably
comparing the strength of each association.
An indirect measure that may prove fruitful is the Sorting Paired Features task (SPF;
Bar-Anan, Nosek, & Vianello, 2009). This is because the SPF assesses the strength of all four
associations in a single block. The task involves sorting two stimuli (e.g., Husky and
Happy) into its corresponding paired category (e.g., dog-pleasant, dog-unpleasant, catpleasant, and cat-unpleasant), which are located in the four corners of the screen. Thus, in the
48
example above, the correct category would be dog-pleasant. In the original SPF paper, BarAnan et al. (2009) present findings showing that Caucasian people hold stronger white-good
and black-bad associations than white-bad and black-good associations, relative to Black
people. Similarly, female participants were found to hold stronger female-good and male-bad
associations than female-bad and male-good associations compared to male participants.
Thus, the SPF shows good known-groups validity.
Moreover, Bar-Anan et al.'s SPF results show that opposing categories not measured
by an IAT (e.g., black-bad, white-bad) may be important in understanding an individuals
socio-cognitive processing. Applied to child abusers, these associations would be adult-not
sex and child-not sex. In addition to assessing and comparing all possible associations, the
SPF has other benefits. For example, the fact that the SPF involves a single block means that
the problems related to block switching (e.g., order effects) are absent. Also, the developers
suggest that, because the SPF displays two stimuli concurrently, it is more sensitive to
stimulus meanings than other indirect measures.
Given these unique properties, the SPF seems an ideal tool to use with child abusers.
First, all four associations can be assessed and compared. This will provide insights into
whether abusers hold increased child-sex associations or decreased adult-sex associations and
allows not-sex associations to be examined. Second, due to being more sensitive to the
meaning of each stimulus, the SPF should be better able to tap the associations of interest
(i.e., child-sex), as opposed to other child-related associations (e.g., child-parenting).
Furthering this point, child abusers with more chronically accessible associations are likely to
be more responsive to this heightened sensitivity than abusers with less accessible
associations. According to the literature, exclusively extrafamilial child abusers (i.e., who only
abuse children unrelated to them) hold deeply entrenched distorted schemas (or associative
networks) compared to non-extrafamilial abusers (Gannon, Keown, & Polaschek,
2007; Gannon & Polaschek, 2006; Ward, 2000). Also, extrafamilial abusers tend to report
49
using more deviant fantasies than other abusers (Marshall, Barbaree, & Eccles, 1991). Thus, it
can be argued that extrafamilial abusers will attend to the meaning of the child-sex stimuli
more readily (i.e., faster) than other abusers on the SPF.
In light of this, the present study had two main aims: 1) to apply the SPF to the
assessment of sex-related associations in exclusively extrafamilial abusers, non-extrafamilial
abusers, and non-offenders, and: 2) to examine whether child-sex scores on the SPF are
correlated with self-reported child fantasies. Four specific hypotheses were tested.
Hypothesis 1: Due to having more chronically accessible child-sex associations, extrafamilial

abusers will show stronger child-sex associations on the experimental SPF than the nonextrafamilial abusers and the non-offenders.
Hypothesis 2: Within the extrafamilial group, child-sex associations will be significantly

greater than adult-sex associations. For the other two groups, adult-sex associations will be
significantly greater than child-sex associations.
Hypothesis 3: On the control SPF, there will be no group differences in how strongly flowergood associations are held. Support for this hypothesis will help rule out the possibility that
variables associated with general SPF performance account for the results on the experimental
SPF.
Hypothesis 4: For the extrafamilial abusers, the child-sex association will positively correlate
with the use of deviant sexual fantasies involving children.
50
3.2 Method
3.2.1 Participants
3.2.1.1 Offender participants
The child abusers in this study were patients from a secure treatment facility in Wisconsin,
USA, who were detained under Wisconsins Sexually Violent Persons law. The researcher
visited each wing of the facility to notify patients that the study was running. They were
informed that the study was looking at the speed in which people respond to pictures of
people as well as sexual and non-sexual words. A sign-up sheet was left on the notice-board
of each wing as a way for offenders to indicate their interest in volunteering. This form was
collected a week later. Participants were included in the study if their case file described a
current or previous conviction for a sexual offence against a child. A staff member notified
the researcher about which of the signed-up patients met this criterion. Given the restricted
timeframe in which to recruit and test participants, as well as the possibility of excluding
certain participants (e.g., due to low cognitive functioning, unusually high error rates), as
many eligible participants were recruited as possible for this study. Thus, the final sample size
was a result of convenience sampling.
Offenders who only abused post-pubescent children (i.e., hebephiles) were removed
from the sample, as previous research suggests that they do not hold indirectly assessed childsex associations (Brown et al., 2009). Thus, their inclusion in the sample would have
contaminated the findings. In addition, participants that had very low cognitive functioning
were excluded because they may have found the SPF task too difficult to understand and
perform. Participants who were taking neuroleptic medication were also excluded because the
side-effects of the drugs affected their motor skills, which greatly slowed their reaction time.
The resulting number of male child abusers used for this study was 39. Using case file
information to review sexual offence histories, the subtype of each child abuser was
51
ascertained. Twenty-six were identified as exclusively extrafamilial (i.e., only abused children
outside the family unit), while the remaining 13 were a mix of non-extrafamilial abusers (e.g.,
5 intrafamilial child abusers; 3 predominantly intrafamilial abusers; 5 individuals who
assaulted both children and adults). Coincidentally, 84.6% of the extrafamilial group (n=22)
and 84.6% of the non-extrafamilial group (n=11) had a diagnosis of paedophilia, as defined
by the Diagnostic and Statistical Manual for Mental Disorders (DSM-IV-TR; American
Psychiatric Association, 2000).
Table 3.1 below displays demographic information for all participant groups in the
study. As shown, the mean age of the two offender groups differed by only 2 years. Also,
both groups were mainly comprised of Caucasian participants. Other ethnicities in the
extrafamilial group included Black (n=1), Asian (n=1), and Native American (n=1). For the
non-extrafamilial group, the other ethnicities included Native American (n=2) and Black
(n=1). Further, all of the offenders reported not being in a relationship, except for one
offender in the extrafamilial group. This is not surprising, however, given the circumstances
of their life at the time of the study (i.e., incarcerated indefinitely in a secure mental health
facility).
52
Table 3.1: Demographic information for the extrafamilial abusers, nonextrafamilial abusers, and non-offenders.
Extrafamilial
Non-extrafamilial Non-offender
abusers
abusers
controls
(n = 26)
(n = 13)
(n = 28)
Mean age:
46 (SD=11.5)
44 (SD=11.3)
20 (SD=2.6)
Ethnicity:
Caucasian
Black
N. American
Asian
Other
88.6%
3.8%
3.8%
3.8%
0
76.9%
7.7%
15.4%
0
0
85.7%
0
0
10.7%
3.6%
Relationship
status:
In a relationship
Single
Divorced/separated
Widowed
3.8%
76.9%
15.4%
3.8%
0
84.6%
15.4%
0
21.4%
78.6%
0
0
Table 3.2 presents mean scores for a number of clinical factors for each offender group. This
information was drawn from the participants files and was always based on the most recently
recorded score/assessment. The factors include static risk (as measured by the Static-99R;
Helmus, Thornton, Hanson, & Babchishin, 2012), dynamic risk (as measured by the
Structured Risk Assessment - Forensic Version, Thornton, 2002), psychopathy (as measured
by the Psychopathy Checklist-Revised; Hare, 1991), and IQ (as measured by the Wechsler
Adult Intelligence Scale, 3rd and 4th Editions; Wechsler, 1997, 2008). As shown, both groups
exhibited very similar static risk, dynamic risk, and IQ scores. Indeed, independent t-tests
indicated no significant group differences on these factors. Also, while the psychopathy
scores differ by 3 points on the PCL-R, the difference was not significant.
53
Table 3.2: Mean scores (SDs) for static and dynamic risk,
psychopathy, and IQ in both offender groups
Extrafamilial
Non-extrafamilial
abusers
abusers
(n = 26)
(n = 13)
5.6 (1.3)
5.8 (1.9)
Static risk:
3.8 (0.5)
3.6 (0.7)
Dynamic risk:
Psychopathy:
23.4 (4.0)
26.3 (4.5)
IQ:
97.2 (12.9)
96.3 (14.1)
Many of the participants were undergoing cognitive-behavioural treatment at the time of

study, which involved moving through two phases depending on their progress. Relative to
pre-treatment participants, most of the offenders in each group were in treatment, particularly
Phase 2 (see Table 3.3). As shown, there were a greater proportion of extrafamilial abusers in
treatment than non-extrafamilial abusers, although the difference was not significant.
Table 3.3: The proportion of participants in treatment and

pre-treatment for both groups
Extrafamilial
Non-extrafamilial
abusers
abusers
(n = 26)
(n = 13)
15.4%
38.5%
Pre-treatment
In-treatment:
Phase 1
Phase 2
84.6%
26.9%
57.7%
61.5%
15.4%
46.2%
Table 3.4 shows information about the offenders offence history, which was also derived
from their case files. Both offender groups appeared to have started sexually offending around
the same age and, thus, showed no group differences. Extrafamilial abusers offended against
just male children at a higher rate than non-extrafamilial abusers, although this was not
significantly different. Extrafamilial also offended against both genders at a higher rate but,
again, the difference was not significant. Non-extrafamilial abusers mainly offended against
female children and at a higher rate than extrafamilial abusers, 2 (1) = 7.8, p < .01.
54
A final observation is that most of the offenders in both groups had prior convictions
for sexual offences, with no significant differences found. However, non-extrafamilial abusers
had significantly more prior convictions for both violent offences, 2 (1) = 5.85, p <.05, and
nonsexual, nonviolent offences, 2 (1) = 4.18, p < .05.
Table 3.4: Offence history information for both offender groups

Extrafamilial
Non-extrafamilial
abusers
abusers
(n = 26)
(n = 13)
Mean age at
18 (SD= 5.9)
19 (SD = 7.8)
first sexual
offence
Victim gender:
- Male
- Female
- Both
34.6%
23.1%
42.3%
15.4%
69.2%
15.4%
Prior sexual
convictions
88.5%
92.3%
Prior violent
convictions
11.5%
46.2%
Prior nonsexual
nonviolent
convictions
34.6%
69.2%
3.2.1.2 Non-offender participants

A comparison sample of 28 non-offending males was recruited from the University of
Birmingham, UK via the online Research Participation Scheme (RPS). This sample size was
also a function of convenience sampling. As shown in Table 3.1, the mean age of the nonoffenders was 20 years old (SD = 2.6, range = 18-31). Like the offender groups, the majority
of non-offenders were Caucasian, with three participants having an Asian background and one
responding with Other. The majority of non-offenders were single, although six stated to be
'in a relationship'.
55
Ethical approval to conduct this study with sex offenders was granted from: 1) the
Institutional Review Board in Wisconsin (Appendix B) and; 2) the Science, Technology,
Engineering and Mathematics Ethical Review Committee at the University of Birmingham
(see Appendix C). Ethical approval to conduct this research with non-offending males was
granted from the University of Birminghams Science, Technology, Engineering and
Mathematics Ethical Review Committee (see Appendix C).
3.2.2 Materials
3.2.2.1 Experimental SPF
In order to assess sex-related associations, a SPF task was constructed using information
outlined by Bar-Anan et al. (2009). The category pairings were Child-Sex, Child-Not Sex,
Adult-Sex, and Adult-Not Sex. As shown in Figure 3.1, each category corresponded with a
particular response key according to its position on the screen. The SPF task involved
participants categorising two different stimuli simultaneously. Thus, if the words school and
orgasm were presented together (as shown in Figure 3.1), a participant would have to
quickly press the key for the Child-Sex' category (i.e., 'P' in Figure 3.1). Note also that the
positions of the category labels were counterbalanced between-participants, resulting in 8
different screen arrangements.
Figure 3.1: Example screen from the Sorting Paired Features task.
56
The stimuli for the task were 8 sex, 8 not sex, 8 adult, and 8 child words (see Appendix D). All
words were drawn from Gray et al.s (2005) IAT study, as, in their study, the words were
judged by sex offenders as being valid exemplars of their respective category. Using E-Prime,
adult and child-related words were set to randomly pair-up with a sex or not sex word. These
stimulus-pairings were presented in random order throughout the task. The total number of
trials was conservatively set at 80 (i.e., 20 trials per category). This is because any more
would have run the risk of causing mental fatigue (Vianello, personal communication, 2010),
which would have negatively affected the participants performance. However, before the 80
critical trials began, a 16-trial practice phase was first completed to ensure the task was
understood.
3.2.2.2 Control SPF

A control SPF was also constructed in order for the results of the experimental SPF to be
more reliably interpreted. Previous IAT studies have commonly used the Flower-Insect IAT
as a control task (e.g., Brown et al., 2009) because research shows that most people hold
positive associations towards flowers relative to insects (Greenwald et al., 1998). The set-up
of the task was identical to the experimental SPF except that the category pairs were
Flowers-Good, Flowers-Bad, Insects-Good, and Insects-Bad. The stimuli were derived
from previous studies using a Flower-Insect IAT (e.g., Gray et al. 2005; Greenwald et al.,
1998). However, it was thought that many of the offenders would not be familiar with the
names of certain flowers (e.g., snowdrop, crocus). Thus, only the most common flower names
were used, which resulted in five flower stimuli. To match this, five Insect, five Good, and five
Bad words were used (see Appendix D). Finally, as with the experimental SPF, the position of
the category labels was counterbalanced between-subjects.
57
3.2.2.3 Wilson Sexual Fantasy Questionnaire (WSFQ; Wilson, 1978).

The WSFQ is a 40-item questionnaire that uses a 6-point Likert Scale (0-5) to measures how
frequently 40 specific sexual fantasies are used (see Appendix E). This measure is comprised
of four factor-analytically derived subscales labelled: Exploratory, Impersonal,
Sadomasochistic, and Intimate. For the subscales and overall scale, scores are summed to
provide a total score. Baumgartner, Scalora, and Huss (2002) found that the four subscales
have good internal consistencies when used with child abusers ('s ranging from .83 to .92),
as well as the overall scale ( = .95). The WSFQ is argued to have two items that may refer to
children (Baumgartner et al., 2002). These are: Having sex with someone much younger than
yourself and Seducing an innocent. Baumgartner et al. found that child abusers scored
higher on these two items compared to non-sex offenders. Thus, only these two items of the
WSFQ were focused upon in this study.
3.2.2.4 Thoughts and Fantasies Questionnaire (Thornton, unpublished).

This is an open-ended questionnaire designed to assess whether sex offenders have recently
used a deviant sexual fantasy during their time in treatment. The measure asks about six
specific deviant fantasies; that is, abduction, force, children <13 years, children between 13
and 17 years old, sadism, and past victims. For each fantasy an offender reports using, the
questionnaire asks a number of further open-ended questions, such as how often it is used in a
month; when it was last used; and how long does it last for? It also asks the respondent to
describe the fantasy in detail. The measure was adapted slightly so as to not be confined to
fantasies used during treatment. Also, for the purposes of this study (i.e., which included nonoffenders), the sexual fantasies about past victims was changed so that it asked about any
other frequently used sexual fantasy (see Appendix F).
58
3.2.3 Procedure
The offender participants and non-offenders completed the study in a quiet meeting room and
lab, respectively. Each participant was informed (verbally and in writing) about the nature of
the study. Specifically, it was explained that the study aimed to investigate the speed in which
people respond to words about adults, children, and sex. All participants were informed that
some of the words would be of a sexual nature. Participants were not informed about why
these responses were being recorded until after the study was completed. Participants were
also told that they would be asked to complete a couple of follow-up questionnaires after the
computer task. Offender participants were also asked to provide consent for the researcher to
use data within their case files and psychometric reports. All participants were informed that
they had to right to withdraw at any point in the study if they felt they wanted to.
After consenting to take part, each participant completed the control SPF, then the
experimental SPF, and finally the questionnaires. Both SPF tasks were presented on a PC
using the experiment software package, E-Prime. The control SPF was administered before
the experimental SPF so that participants would be familiar with the procedure for when they
completed the experimental version. The fantasy questionnaires were administered last to
reduce any priming effects that they may have had on the experimental SPF.
Identical instructions on how to complete the SPF were presented onscreen to all
participants when they started. The instructions explained that the task involved presenting a
picture and a word together, and that they had to categorise them into the correct paired
category using one of four keys responses. The instructions also stated that mistakes were to
be expected and that, whenever a mistake is made, a red cross would appear (i.e., error
feedback). It was stated that the correct response must be made in order for the next trial to
begin. Finally, they were told to go as fast as possible but also to be as accurate as possible.
After reading the instructions, participants began the practice phase. After the practice
phase, participants completed the critical phase. This exact procedure was used for both SPF
59
tasks. Finally, all participants completed the two sexual fantasy questionnaires in a counterbalanced order. After the study was completed, each participant was fully debriefed and
thanked for their time.
3.2.4 Data treatment

Before any statistical analyses were performed, the raw data had to be converted into a
standardised form. This is important because there is likely to be unwanted variance from
procedural variations and individual differences (e.g., natural reaction-time, cognitive ability,
age). Indeed, from looking at the raw reaction times (RTs) in Table 3.5, it can be seen that
non-offenders are generally faster on all four categories when compared to the sex offenders.
Thus, in order for the response latencies to be comparable between groups, a standardised
score needed to be calculated.
Table 3.5: Mean raw response latencies (milliseconds) and SDs from the experimental
SPF for extrafamilial abusers, non-extrafamilial abusers, and non-offenders.
Extrafamilial
abusers
abusers
(n = 26)
(n = 13)
(n = 28)
Adult-Sex
1838.2 (519.8)
1791.1 (527.6)
1327.0 (372.6)
Child-Sex
1776.9 (401.6)
2023.4 (708.5)
1406.2 (313.1)
Adult-Not Sex
1992.5 (470.2)
2024.9 (528.02)
1434.1 (292.6)
Child-Not Sex
1822.4 (457.9)
1892.1 (475.6)
1391.9 (326.3)
For the IAT, Greenwald, Nosek, and Banaji (2003) devised a scoring algorithm that produces
a D-score (see Chapter 6) that helps reduce the confounding effects mentioned above. This
algorithm has been used in almost all subsequent IAT studies since its publication. On this
basis, the SPF developers devised a similar scoring algorithm to be used for each of the four
associations. This scoring procedure results in a D-association and is thought to be comparable to
the IATs D-score (Bar-Anan et al., 2009).
60
The procedure first involves removing any RTs below 400ms or above 5000ms, as well
as removing any participant with more than 10% of their responses outside these cut- offs.
This is because, as with the IAT, RTs below or above these cut-off points are reflective of
anticipated and distracted responding, respectively. According to the SPF developers, these RT
cut-offs were based on informed judgment. That is, when the SPF was being developed, the
RT cut-offs for the IAT were 300ms and 3000ms. However, the SPF is a harder task to
perform than the IAT. Therefore, the SPF developers decided to increase the cut-offs for the
SPF in order to account for this increased difficulty (Nosek, personal communication, 2012).
The next step involved generating a D-association for each of the four associations. Each
D-association represents the performance of sorting stimuli into a specific paired-category
compared to the overall performance. Thus, the equation for the D-association is as follows:
Dassociation = (M overall M association)/SD overall
That is, the mean RT for each category is subtracted from the mean overall RT. The resulting
figure is then divided by the overall standard deviation. This is done for each category,
thereby, producing four individual D-associations. Each D-association represents the strength of a
particular association. This scoring algorithm was used for both SPF tasks in this study.
3.3 Results
3.3.1 Hypothesis 1: Extrafamilial abusers will show stronger child-sex associations than
non-extrafamilial abusers and non-offenders.
Looking at Figure 3.2, it can be observed that only the extrafamilial abusers showed a positive
child-sex D-association (black bar) indicating a strong child-sex association. For the other two
groups, adult-sex was the strongest D-association (bars with diagonal lines). Interestingly, the
non-extrafamilial abusers showed stronger adult-sex associations than non-offenders. It can
also be seen that the D-association for adult-not sex (dotted bar) was very weak in all three
61
groups. While this may indicate that all groups hold weak adult-not sex associations, it is
more likely to have been an artefact of that particular stimulus set (see Section 3.3.5).
Figure 3.2: D-association means and SEs from the experimental SPF task for
extrafamilial abusers, non-extrafamilial abusers, and non-offenders.
Table 3.6: Mean error rates (%) and SDs from the experimental SPF for extrafamilial
abusers, non-extrafamilial abusers, and non-offenders
Extrafamilial
abusers
abusers
(n = 26)
(n = 13)
(n = 28)
Adult-Sex
8.5 (11.8)
11.9 (12.8)
7.5 (8.2)
Child-Sex
7.9 (8.9)
14.2 (14.3)
8.2 (6.3)
Adult-Not Sex
9.2 (15.2)
15.4 (21.5)
13.4 (10.8)
Child-Not Sex
6.5 (9.2)
13.1 (23.4)
7.7 (9.9)
To test for statistical differences, the four D-associations were subjected to a 3 (Group) x 4
(Association) mixed factorial ANOVA, with Group as the between-subjects factor and
Association as the within-subjects factor. Since Hypothesis 1 was directional, the ANOVA
62
was run with between-subjects planned contrasts (with regards to the child-sex D-association). The
results showed a significant main effect of Association, F(3, 192) = 9.83, p <.001, but not
Group, F(2, 64) = 1.55, p>.05. The interaction effect between Association and Group was
approaching significance F(6, 192) = 2.09, p = 0.056, and so the planned contrasts were
examined to determine whether Hypothesis 1 was supported.
The planned contrasts of this interaction were non-orthogonal (as the data for the
extrafamilial group was being contrasted twice). Therefore, to control the risk of increased
familywise error associated with multiple comparisons, a sequential Dunn-Sidk correction3
was applied. This adjusted the alpha () levels to 0.025 for the comparison between
extrafamilial abusers and non-offenders, and 0.05 for the comparison between extrafamilial
abusers and non-extrafamilial abusers. As such, the planned contrasts showed the child-sex Dassociation
to be significantly stronger in extrafamilial abusers than in non-extrafamilial abusers
(p <.05, one-tailed d =.75) and non-offenders (p <.01, one-tailed, d =.74). These findings seem
to support Hypothesis 1. Note also that, although the non-extrafamilial abusers and nonoffenders showed stronger adult-sex D-associations than extrafamilial abusers (as shown in Figure
3.2), post-hoc comparisons (with Bonferroni corrections) indicated that the differences were
not significant.
As shown in Table 3.6, the mean error rates show a pattern of responding that is
reflective of the hypothesised biases. For example, the extrafamilial abusers made fewer
mistakes for the child-sex category than the adult-sex category. Also, while the nonextrafamilial abusers and the non-offenders made more mistakes overall, they both produced
more incorrect responses for the child-sex category than did the extrafamilial abusers.
However, error analyses using between-subject ANOVAs on arcsine-transformed error rates
(i.e., to normalise distributions) found no significant differences between groups.
3
1/k,
To calculate sequential Dunn-Sidk corrections, the smallest p-value must be less than 1-(1-alpha) where k
equals the number of comparisons and alpha is 0.05. If it is less, the next smallest p-value must then be less
1/(k-1)
than 1-(1-alpha)
(McDonald, 2009).
63
3.3.2 Hypothesis 2: For extrafamilial abusers, child-sex associations will be significantly

greater than adult-sex associations. For the other two groups, adult-sex associations will
be significantly greater than child-sex associations.
As shown in Figure 3.2, the extrafamilial abusers were the only group whose child-sex Dassociation
was greater than the adult-sex D-association. In the two comparison groups, the adult-sex
D-association was much greater than the child-sex D-associations. Thus, the findings are in the desired
direction.
To test for statistical differences, within-subject planned contrasts were also run with
the 3 x 4 mixed factorial ANOVA described above. Syntax had to be used for this since SPSS
does not have a specific function (i.e., from the drop-down menus) for doing planned
contrasts with repeated-measure ANOVAs. Results indicated that the child-sex D-association did
not significantly differ from the adult-sex D-association in extrafamilial abusers. However, the
adult-sex D-association was found to be significantly greater than the child-sex D-association in nonoffenders (p <.01, one-tailed, d = 2.2) and the non-extrafamilial abusers (p <.01, one-tailed, d
= 1.3). Thus, on the basis of these results, Hypothesis 2 seems to be partially supported.
3.3.3 Hypothesis 3: On the control SPF, there will be no group differences in how
strongly flower-good associations are held.
Examining whether all participants hold an innocuous flower-good association is a common
strategy undertaken in many IAT studies, including those with sex offenders (Gray &
Snowden, 2009). This is because it helps support the assumption that any group differences
on the experimental task are due to differences in association strength rather than factors that
affect performance on the task. As can be seen in Figure 3.3, all three groups showed very
similar flower-good D-associations (dotted bar).
64
Figure 3.3: D-association means and SEs from the control SPF task for extrafamilial
abusers, non-extrafamilial abusers, and non-offenders.
Table 3.7: Mean error rates (%) and SDs from the control SPF for extrafamilial
Extrafamilial
abusers
abusers
(n = 26)
(n = 13)
(n = 28)
Flower-Good
4.4 (6.4)
8.5 (12.5)
6.9 (6.1)
Insect-Good
10.4 (12.8)
13.9 (17.2)
12.3 (8.1)
Flower-Bad
10.2 (10.9)
12.7 (15.4)
10.5 (9.1)
Insect-Bad
5.6 (6.1)
9.6 (10.5)
8.8 (8.0)
To ensure that the groups showed no differences on the flower-good D-association, a (Group) x 4
(Association) mixed factorial ANOVA was conducted, with Group as the between-subjects
factor and Association as the within-subjects factor. Analyses showed that the main effect of
Association was significant, F(3, 192) = 12.81, p < .001. However, as predicted, the main
65
effect for Group was not significant, nor was the interaction effect between Association and
Group. These results provide support for Hypothesis 3.
In addition, the pattern of mean error rates is also reflective of Hypothesis 3 (see Table
3.7). For example, all three groups appeared to make fewer mistakes when responding to the
congruent associations (i.e., Flowers-Good and Insect-Bad) and more mistakes on the
incongruent associations (i.e., Flowers-Good and Insects-Good). Also, it can be seen that the
non-extrafamilial abusers made slightly more mistakes across all associations. However, an
error analysis using between-subject ANOVAs on arcsine-transformed error rates found no
significant differences between groups.
3.3.4 Hypothesis 4: For extrafamilial abusers, the child-sex association will positively
correlate with the use of deviant sexual fantasies about children
According to Hypothesis 4, the extrafamilial abusers child-sex D-association would correlate
with the use of deviant sexual fantasies involving children. For the Thoughts and Fantasies
Questionnaire, a bi-serial correlation was conducted as the data was categorical due to the
dichotomous coding (i.e., Deviant fantasiser = 'Yes' versus 'No'). The only significant result
was found in the extrafamilial group, which was a negative relationship between a childrelated sexual fantasy and the child-not sex D-association. As indicated by a Shapiro-Wilk test,
the data for the WSFQ were non-normal for all groups. Therefore, a non-parametric
correlation test (Spearmans Rho) was used to analyse the relationship between each of the 40
WSFQ items and the four D-associations.
As shown in Table 3.8, the child-sex D-association did not correlate with any of the childrelated fantasies for the extrafamilial group. Instead, the adult-not sex D-association showed a
positive correlation with the two WSFQ items related to children. The adult-sex D-association also
correlated with these two WSFQ items but in a negative direction. The child-not sex Dassociations
showed negative correlations with Seducing an innocent. It can also be seen that the
66
child-sex D-association was negatively correlated with some other WSFQ items, (e.g.,
Participating in an orgy, Tying someone up). The adult-sex D-association also negatively
correlated with a number of other WSFQ items (e.g., Taking someones clothes off, Kissing
passionately). As shown in Tables 3.9, the child-not sex D-association was negatively correlated
with a number of WSFQ items (including the two child-related fantasies) for nonextrafamilial abusers. Finally, Table 3.10 shows that there were only three positive
correlations for the non-offenders, which were between the adult-sex D-association and
Participating in an orgy, Homosexual activity, and Giving oral sex.
Although these correlational results suggest that Hypothesis 4 is not supported, there is
a possibility that the markedly slow adult-not sex RTs were having a negative effect on the Dassociations, thereby,
masking any existing relationships (see Section 3.3.5).
67
Table 3.8: Significant correlations between the four D-associations and WSFQ sexual fantasies
for extrafamililal abusers
Child
Adult
AdultChildGroup
Measure
-Sex
-Sex
Not Sex
Not Sex
.21
-.46*
.45*
-.36
Seducing an innocent
.15
-.58**
.69***
-.52**
Participating in an orgy
-.46*
.11
.04
.27
Watching others have sex
- .44*
- .09
.25
.19
Mate-swapping
-.47*
.15
-.06
.29
Tying someone up
-.42*
-.12
.25
.29
Transvestism
-.39*
.14
-.06
.22
Receiving oral sex
.12
-.44*
.17
.07
Giving oral sex
.04
-.43*
.17
.14
Taking someones clothes off
.15
-.58**
.29
.12
Making love elsewhere than

bedroom
.31
-.40*
.15
-.15
Having sex with someone of a

different race
-.14
-.52**
.38
.09
Looking at obscene pictures or

films
-.17
-.48*
.41*
.01
Kissing passionately
.30
-.45*
.12
.04
Intimate subscale
.15
-.50**
.18
.09
Exploratory subscale
-.17
-.45*
.33
.11
Impersonal subscale
-.21
-.54**
.38
.16
Extrafamilial Having sex with someone much

younger than yourself
* p < .05
** p <.01
***p <.001
68
Table 3.9: Significant correlations between the four D-associations and WSFQ sexual
fantasies for non-extrafamilial abusers
Child
-Sex
Adult
-Sex
AdultNot Sex
NonMaking love outdoors in a

Extrafamilial romantic setting
.35
-.03
.08
-.72*
Sex with two other people
.19
-.21
.28
-.66*
Forcing someone to do
something
.14
.23
-.08
-.61*
Homosexual activity
.08
-.28
.39
-.59*
Giving oral sex
.15
.07
.09
-.60*
Taking someones clothes off
.32
-.16
.11
-.56*
Being excited by material or

clothing
.40
.12
-.14
-.75**
Transvestism
.27
-.06
.09
.70**
Being promiscuous
.53
.24
-.33
-.81**
Being much sought after by the

opposite sex
-.06
.39
.23
-.59*
Having sex with someone of a

different race
.49
-.17
.06
-.65*
Mate-swapping
-.09
-.22
.57*
-.43
Intimate subscale
.16
.11
.09
-.66*
Exploratory subscale
.15
-.09
.23
-63*
Group
Measure
ChildNot Sex
* p < .05
** p <.01
Table 3.10: Significant correlations between the four D-associations and WSFQ sexual
fantasies for non-offenders
Child
-Sex
Adult
-Sex
AdultNot Sex
ChildNot Sex
Participating in an orgy
-.29
.47*
-.03
-.09
Homosexual activity
-.12
.39*
-.01
.03
Giving oral sex
-.12
.40*
.03
.04
Group
Measure
Nonoffender
* p < .05
69
3.3.5 Exploring the possible biasing effects of Not Sex stimuli

The results from the SPF (as shown in Figure 3.2) indicate that all three groups sorted adultnot sex stimuli at a much slower rate than the other three categories. An examination of the
error rates across all groups indicated that the trials for the Adult-Not sex category contained
a higher number of initial wrong responses (see Table 3.6). This may account for the apparent
slower responding for the adult- not sex category in all three groups. If this is the case, it
could highlight a problem with the stimulus set for this category (see section 3.4.1 of the
Discussion).
To explore whether the adult-not sex RTs may have been biasing the results of this
study, two 'amended D-associations' were computed for child-sex and adult-sex associations. This
followed the same formula as the original D-association but only the data for adult-sex and childsex categories were used. The results still appeared to support the hypothesis that
extrafamilial abusers hold stronger child-sex associations than the comparison, and vice versa
(see Figure 3.4). T-tests with Bonferroni corrections ( = .025) showed that extrafamilial
abusers produced significantly greater amended child-sex D-associations than non-extrafamilal
abusers (p = .008) and non-offenders (p = .02). In contrast, non-extrafamilial abusers
demonstrated marginally greater amended adult-sex D-associations than extrafamilial abusers (p
= .029). No significant differences were found between non-offenders and extrafamilial
abusers.
Also similar to the original analysis, the amended child-sex and adult-sex D-associations
did not differ within the extrafamilial group. However, within both the non-extrafamilial
abusers and non-offenders, the adult-sex D-association was significantly stronger than the childsex D-association (both ps = .02).
70
Figure 3.4: Mean and SEs for the child-sex and adult-sex amended
D-associations for extrafamilial abusers, non-extrafamilial abusers, and non-offenders
Regardless of these similar results, there was an indication that a biasing effect was present in
the original analyses. For example, the extrafamilial abusers now appeared to hold weaker
adult-sex associations, as shown by the negative amended D-association (see Figure 3.4).
Furthermore, correlational analyses revealed that, for the extrafamilial abusers only, the
amended child-sex D-association was now positively correlated with the WSFQ item 'Seducing
an Innocent' (Spearman's Rho = .41, p < .05). The relationship with the item 'Having sex with
someone much younger than yourself' was just shy of significance (Spearman's Rho = .38, p =
.055). These results are in clear contrast to the original results and fall more in line with
Hypothesis 4. Other significant correlations also emerged (all ps <.05) between the amended
child-sex D-association and the following WSFQ items: Taking someones clothes off
(Spearman's Rho = .39), Having sex someone other than the bedroom (Spearman's Rho =
.46), and Kissing passionately (Spearman's Rho = .43). No significant correlations were found
in the non-extrafamilial group. However, one significant correlation emerged for the non71
offenders between the amended child-sex D-association and the WSFQ item Being seduced as
an innocent (Spearman's Rho = .43, p <. 05).
3.3.6 Exploring the heterogeneity within the non-extrafamilial group
While the results have so far provided some support for the hypotheses, it should be noted that
the comparison offender group (i.e., non-extrafamilial abusers) was not a pure subtype.
Rather, it was comprised of various subtypes of abusers; namely, intrafamilial,
predominantly intrafamilial, and those who have offended against both children and adults.
This point needs to be addressed as important differences may be disguised as a consequence
of treating the subtypes as one overall group. For example, it could be the case that
intrafamilial abusers actually respond just as fast on the child-sex category as extrafamilial
abusers, but due to slow responding in the other subgroups, it is not being made apparent.
Thus, in this section, the results for each subtype will be explored.
As shown in Figure 3.5, each of the three subtypes show a similar negative child-sex
D-association, suggesting that they are not showing a child-sex bias. Rather, they show a pattern
of responding indicative of an adult-sex bias, with those who offended against children and
adults showing the greatest adult-sex D-association.
72
Figure 3.5: Mean D-associations (and SEs) for each of the non-extrafamilial subgroups,
extrafamilial abusers, and non-offenders
In spite of these patterns, t-test analyses showed that none of the subgroups significantly
differed from extrafamilial abusers on the child-sex D-association. However, given that the data
may be skewed by the non-sex words (as suggested by the analysis in section 3.5.5), the
amended D-associations were also examined. As shown in Figure 3.6, the offenders against
children and adults had the weakest child-sex amended D-association, which was significantly
different from the extrafamilial abusers (p < .05). The intrafamilial abusers also showed a
lower child-sex D-association than the extrafamilial abusers (p = .05), while the predominantly
intrafamilial abusers did not significantly differ. No significant differences were found
between the three subgroups. Thus, while the offenders who offended against children and
adults showed the weakest child-sex association, the intrafamilial abusers also showed weak
child-sex associations, which is in support of the Hypothesis 1. Furthermore, in support of
Hypothesis 2, the intrafamilial abusers showed a significant difference between the adult-sex
73
and child-sex amended D-associations. Finally, no significant associations were found between the
amended child-sex D-association and the measures of child-related sexual fantasies.
Figure 3.6: Means for the amended D-association (and SEs) for each of the
non-extrafamilial subgroups, extrafamilial abusers, and non-offenders
3.4 Discussion
Previous research using the IAT has provided evidence that child abusers hold stronger childsex associations than controls. However, the evidence it is not wholly conclusive due to the
limitations of the IAT. For example, the IAT does not measure all four associations. Also,
previous research has failed to take into account the heterogeneity evident in child abusers.
Thus, using the Sorting Paired Features task (SPF) to assess all four possible associations, the
present study aimed to show that extrafamilial child abusers hold stronger child-sex
associations than non-extrafamilial abusers and non-offenders, and that the child-sex
associations would be stronger than adult-sex associations. In line with the DPM-ST, it was
74
hypothesised that stronger child-sex associations would be correlated with child-related

sexual fantasies.
Results showed that extrafamilial abusers held significantly stronger child-sex
associations than non-extrafamilial abusers and non-offenders, thus, supporting Hypothesis 1.
In addition, the error rates also corresponded with this pattern of results, in that, the
extrafamilial abusers made fewer mistakes on the child-sex category (and more on the adultsex category) relative to the other two groups. However, there were no significant group
differences. Results also showed that, although the extrafamilial abusers held stronger childsex associations compared to adult-sex (as shown in Figure 3.2), the difference was not
significant. However, the non-extrafamilial abusers and the non-offenders were found to hold
significantly stronger adult-sex associations compared to child-sex associations. Thus,
Hypothesis 2 was partially supported.
An analysis of the reaction times and error rates also showed that there were no group
differences on the Flowers-Good association of the control SPF. This supported Hypothesis 3.
It also supports the assumption that the group differences on the experimental SPF were due
to differing association strengths as opposed to differences in general SPF performance.
Curiously, Figure 3.3 shows that both of the abuser groups hold weaker insect-bad
associations compared to the non-offenders. As to why this may be the case is beyond the
scope of this study. However, it does demonstrate the benefits of being able to assess the
strength of all four associations.
Finally, the correlational analyses revealed that the child-sex D-associations did not
correlate with deviant sexual fantasies about children in extrafamilial abusers. However, it did
show negative correlations with some other fantasy items (e.g., Watching others having sex,
Tying someone up). These results suggest that Hypothesis 4 was not supported. However, the
other three D-associations showed significant correlations in this group. For example, the adultsex D-associations were negatively correlated with the child-related WSFQ items. In the non75
extrafamilial group, the child- not sex D-association negatively correlated with a number of
WSFQ items from the Exploratory and Intimate subscales (and with the total score for these
two subscales).
It is important to emphasise that the results of the experimental SPF in this study
appeared to have been affected by the slow RTs of the adult-not sex category. As shown by
the equation in the Method section, each D-association is based on the difference between the
single association and the overall performance on the SPF. This means that each D-association is
algebraically related to the other three D-associations (Bar-Anan et al., 2009). Thus, although
each D-association has its own distinct validity allowing for each association to be compared with
the others (Bar-Anan et al., 2009), if one category suffers from a methodological issue that
results in unusually slow RTs, the overall D-association will be distorted. As Bar-Anan et al.
state, this highlights an important limitation of the SPF.
To test if this was the case, two 'amended D-associations' based on just the adult-sex and
child-sex categories were calculated. Results showed that extrafamilial abusers still displayed
stronger child-sex associations than the other two groups. However, the adult-sex associations
had become weaker, indicated by the adult-sex D-associations turning from positive to negative.
Despite this change, the adult-sex D-association still did not significantly differ from the child-sex
D-association within the extrafamilial group. Importantly, however, the amended child-sex Dassociations now
showed significant positive correlations with the two child-related WSFQ items
for the extrafamilial abusers. It also showed significant correlations with some other items,
such as Taking someones clothes off, Having sex somewhere other than the bedroom, and
Kissing passionately. Interestingly, the amended child-sex D-association was positively
correlated with the item Being seduced as an innocent in the non-offender group. This
suggests that non-offenders who hold a child-sex association may be more likely to engage in
fantasies involving themselves as a child. As to whether a child-sex association leads to this
type of fantasy (or vice versa) cannot be ascertained from these correlational results.
76
However, it highlights a course for future research. It also demonstrates that using D-associations
can be problematic if methodological issues arise. As such, this point emphasises the
importance of using well-chosen stimuli in the SPF (as well as other indirect measures).
Taken altogether, these results suggest that extrafamilial abusers hold stronger childsex associations than other abusers and non-offenders, and that the child-sex associations may
have a positive relationship with child-related sexual fantasies. While this conclusion would
support the DPM-ST, great caution needs to be made when interpreting these results as: 1) the
slow adult-not sex RTs appear to have distorted the D-associations of the original analysis; and 2)
the amended D-associations involved removing the not-sex categories, which means the amended
scores may have less validity. Thus, it is more prudent to take these results as grounds for
further investigation using an improved SPF task.
3.4.1 Limitations of the study

Although this study has uncovered some interesting findings, there are a number of
limitations that need highlighting. First, the SPF involved a single session of 80, which was
advised by the SPF developed authors via personal communication. However, this only
allows for 20 trials per association and so it may have been advantageous to have had more
trials. Indeed, other SPF studies have involved administering an 80-trial SPF (Ranganath,
Smith, & Nosek, 2008) and a 40-trial SPF (Bar-Anan & Nosek, 2013) three times, resulting in
240 and 120 trials, respectively.
A second limitation concerns the stimuli used for the SPF. The words were derived
from an IAT used in a previous study (Gray et al., 2005). In hindsight, this stimulus set has
two problems specifically associated with the not-sex words. The first issue is that some of
the words are ambiguous and could arguably be seen as relating to sex rather than not-sex
(i.e. smile, eyes, toe). Indeed, one offender participant stated that "eyes can be sexy", while
another mentioned that people can have a toe fetish. An examination of the error rates did
77
indicate that the trials for not-sex categories contained a high number of initial wrong
responses, which explains the slower responding for the adult-not sex category in all of the
groups. The second issue is that many of the words are neutral in nature (e.g., run,). Thus, it
may have been better to have used words that denote a lack of sexual appeal, such as: ugly,
cold, dull, bland, and boring (O Ciardha & Gormley, 2012). Given that the slower adult-not
sex RTs appeared to have an effect on the SPF results, this recommendation is strongly
advised for future researchers, including those using other indirect measures such as the IAT.
Third, there were some issues regarding the samples used in this study. For example,
considering that the SPF only involved 80 trials (20 per category), the sample size of each
group (particularly the non-extrafamilial group) is likely to have been inadequate. Moreover,
the non-extrafamilial group was comprised of a number of subgroups (e.g., intrafamilial
abusers, abusers of children and adults, predominantly intrafamilial). By disaggregating this
group and analysing the data, the results showed that all three subgroups showed similar
child-sex D-associations. An analysis of the amended child-sex D-association showed that the
intrafamilial abusers and the abusers of children and adults were significantly different from
the extrafamilial abusers. However, the predominantly intrafamilial abusers were not. These
results indicate that, while the non-extrafamilial subgroups showed similar patterns of childsex associations on the SPF, future research should aim to use more homogenous comparison
groups (e.g., a group of pure intrafamilial abusers).
In addition, many of the abusers were in treatment and had progressed to Phase 2.
Thus, it is possible that the strength of the child-sex associations and adult-sex associations
were influenced by the treatment these participants were receiving (i.e., weakened and
strengthened, respectively).
Finally, positive correlations were found between a number WSFQ items and the
amended child-sex D-association. From this, it is tempting to conclude that child-sex associations
can be associated with sexual fantasies that are unrelated to children. However, these
78
correlations may highlight the limitations of the WSFQ, particularly in terms of assessing
child-related sexual fantasies. For example, it is possible that the items Taking someones
clothes off and Having sex somewhere other than the bedroom positively correlate with the
amended child-sex D-association because the respondent was thinking of a child within those
fantasies (e.g., taking a childs clothes off). In other words, the WSFQ does not allow one to
ascertain who the person is when referring to sexual fantasies about specific behaviours or
locations. It should also be noted that the individual significant correlations reported in this
study should be treated with caution, as they are based upon multiple correlation analyses. In
other words, there is an increased risk of committing a Type 1 error. Thus, future research in
this area should account for this issue and aim to use a more detailed means of assessing
sexual fantasies.
3.4.2 Future directions for research

In light of the present findings, the SPF has some potential as a research tool for assessing
sex-related associations in child abusers. However, since this is the first time the SPF has
been applied to this area of cognition, it is necessary for future studies to be conducted. For
example, in addition to addressing the issues outlined above, researchers should think about
using pictures as the stimuli for adults and children categories. First, pictures have been found
to be just as good as words in IAT studies (Banse et al., 2010; Brown et al., 2009). Second,
the SPF allows non-focal associations, such as gender of the person in the image, to be
assessed (Bar-Anan et al., 2009). In other words, in addition to comparing child-sex and
adult-sex associations, the SPF could be used to examine whether child abusers hold
stronger associations between male children and sex or stronger associations between
'female children and sex'. If successful, such a feature would be useful in distinguishing
heterosexual abusers from homosexual abusers.
79
The correlational results between the D-associations and the child-related sexual fantasies
cannot be used to draw any causal conclusions. Thus, experimental research is needed to
support the hypothesis that deviant thoughts and fantasies arise from activated sex-related
associations. For example, in a recent, albeit non-forensic study, Birmbaum, Simpson,
Weisberg, Barnea, and Assulin-Simhon (2012) found that the content of a sexual fantasy can
be affected by explicitly and implicitly priming an attachment orientation. For example, after
explicitly priming insecurity, participants wrote fantasies that involved themes of interpersonal
distance and hostility, especially in men who were anxiously attached. In light
of this, it would be valuable to conduct a similar study in which child abusers are asked to
write a sexual fantasy after having a child-sex association primed.
80
CHAPTER 4
INVESTIGATING THE FACTORS THAT INFLUENCE
THE USE OF DEVIANT SEXUAL FANTASIES
Chapter Rationale
Having found some support for the assumption that sex-related associations are held in longterm memory in Chapter 3, the purpose of this chapter was to examine the factors that may
influence the use of deviant sexual fantasies once such associations are (presumably)
activated. According to the DPM-ST, these factors can include individual differences related
to the ability to fantasise in general as well as the presence of explicit cognitions that are
congruent with the activated associations and prospective sexual fantasy content. The
structural equation modelling study described in this chapter aimed to test these assumptions
in the context of aggressive sexual fantasies used by men in the general population.
81
4.1 Introduction
According to the Dual-Process Model of Sexual Thinking (DPM-ST), a spontaneous sexual
thought will be fleeting unless attended to and elaborated upon in the form of a sexual fantasy.
As discussed in Chapter 2, the transition of a spontaneous sexual thought to a sexual fantasy
is more likely to occur if the thought elicits a strong affective reaction (e.g., arousal,
excitement). Also, the transition is more likely if the sexual thought is congruent with an
individuals explicit attitudes. For example, if a sexual thought about forcing sex on a woman
intrusively enters the consciousness of a man, it is more likely to be elaborated upon if the
man holds an explicit attitude that women enjoy forced sex. Indeed, a number of studies have
demonstrated that non-offending males who hold negative or hostile attitudes about women, or
who endorse rape-myths, are more likely to report using dominance, forceful, or coercive
sexual fantasies (Bartels & Gannon, 2009; Greendlinger & Byrne, 1987; Malamuth, 1981;
Plaud & Bigwood, 1997; Renaud & Byers, 2005; Smeaton & Byrne, 1987; Zurbriggen &
Yost, 2004). From the perspective of the DPM-ST, these attitudes have an influential role in
whether someone generates and uses aggressive sexual fantasies.
The DPM-ST also proposes that certain individual differences will influence whether
someone elaborates upon a sexual thought via sexual fantasy, some of which relate to the act
of fantasising. These include: 1) how prone someone is to fantasising in general; 2) the ability
to form vivid visual mental imagery; and 3) whether someone often has dissociative
experiences in daily life.
'Fantasy proneness' - a constellation of personality traits and experiences related to
deep involvement in fantasy (Wilson & Barber, 1981) - has rarely been investigated in
relation to sexual fantasies. However, from interviews that led to the discovery of the fantasyproneness construct, Wilson and Barber (1981) found that, out of 27 fantasy-prone women,
75% reported using vivid sexual fantasies that resulted in orgasm in the absence of additional
stimulation. However, no such information was provided about sexual fantasies in fantasy82
prone men. A thorough search of the literature failed to find any empirical studies
investigating fantasy proneness in relation to deviant sexual fantasies. However, Curnoe and
Langevin (2002) suggest that sexual offenders who engage in deviant sexual fantasies may be
fantasy-prone in general. Curnoe et al. (2002) based this on previous research showing that
people either have a tendency to sexually fantasise or not (Langevin, Handy, Paitich, &
Russon, 1985).
There is also some indication that sexual fantasies are related to vividness. For
example, in non-offenders, research had shown that the more sexual experience an individual
has, the more vivid their sexual fantasies will be (Brown & Hart, 1977; Gold & Gold, 1991).
Importantly, it has been found that sexual fantasies are affected by an individual's ability to
produce vivid mental imagery. For example, Smith and Over (1987) found that men who are
able to generate vivid mental imagery were more aroused by their sexual fantasies, relative to
men who are less able to produce vivid mental imagery. There is a dearth of published
research examining vividness and sexual fantasies within the sex offender literature. However,
Prentky and Burgess (1991) have proposed that the vividness of sexual fantasies will vary
among sexual offenders.
Dissociation is a state characterised by a lack of integrating various aspects of an
experience (i.e., thoughts, feelings, images) into consciousness (Carlson, Putnam, Ross,
Torem, Coons, & Dill, 1993). While very little has been done to investigate the relationship
between dissociation and sexual fantasising, some have stated that sexual fantasies provide
sex offenders with a step towards dissociating from both their victim's distress and their own
offensive behaviour (Giannangelo, 1996). Similarly, Becker-Blease and Freyd (2007) have
argued that dissociation may contribute to sexual abuse by influencing intense deviant sexual
fantasies, which, as a result, enables the offenders to dehumanise their victims and behave
'normally' in their daily life. However, no empirical research has been conducted to support
these ideas.
83
It is important to highlight that there are some interrelations among these three
constructs. For example, fantasy proneness has been shown to have a strong relationship with
dissociation (Merckelbach, Campo, Hardy, & Giesbrecht, 2005; Pekala, Angelini, & Kumar,
2001). It has also been found that fantasy prone individuals have greater vividness of visual
mental imagery than those who are less fantasy-prone (Aleman & de Haan, 2004). Thus, by
combining the assumptions of DPM-ST with the findings and suggestions discussed in this
Introduction, the present study used structural equation modelling (SEM) to test the
hypothesis that:
1) The use of aggressive sexual fantasies will be influenced by distorted attitudes about
women.
2) Higher fantasy proneness, dissociation, and vividness of visual mental imagery will
increase the likelihood that aggressive sexual fantasies will be used.
In the following subsection, the hypothesised SEM will be outlined along with the empirical
and theoretical rationale that underlies it.
4.1.1 The hypothesised structural equation model

Fantasy proneness, dissociation, and vividness of visual imagery were subsumed under a
single, overarching latent factor, which was termed 'Rich Fantasy Life'. Thus, this variable
represents the proclivity to engage in deep and vivid fantasies without being distracted by the
external world. 'Distorted Attitudes about Women' was also defined as a latent factor and was
made up of variables such as calloused attitudes towards women and the belief that women are
deceitful. Finally, 'Aggressive Sexual Fantasies' was conceptualised as a latent factor
comprised of rape-related and sadistic sexual fantasies (Knight & Sims-Knight, 2003). The
84
relationship between these three latent factors is represented in the structural component of
the hypothesised SEM model in this study (see Figure 4.1 below).
Figure 4.1: Structural component of the hypothesised SEM model.
This a priori model was based upon theoretical and empirical insights. For example,
research indicates that fantasy proneness is a constellation of personality traits that develop in
childhood (Wilson & Barber, 1981; Rhue & Lynn, 1987) and dissociation (as defined in this
study) is a trait construct (Hagenaars & Krans, 2011). Therefore, Rich Fantasy Life was
considered to be a more stable latent variable. In contrast, Distorted Attitudes about Women
was regarded as less stable because research indicates that explicit attitudes can be changed
more quickly and easily than implicit attitudes (e.g., Rydell & McConnell, 2006). Indeed, in
the sexual offending field, distorted attitudes are regarded as a risk factor amenable to change;
that is, a dynamic risk factor (Hanson & Harris, 2001; Thornton, 2002).
Taken together, having the proclivity and ability to engage in deep fantasies (i.e., Rich
Fantasy Life) was hypothesised to be the initial factor influencing aggressive sexual fantasies,
followed by the presence of congruent explicit distorted attitudes about women. Put
differently, the SEM model proposed that a Rich Fantasy Life, via Distorted Attitudes about
Women, influences the use of Aggressive Sexual Fantasies.
4.2 Method
4.2.1 Participants
85
To test the proposed SEM model, 159 male non-offenders from the general community were
recruited via a convenience sampling methodology. The sample had a mean age of 23.5 years
(SD = 6.5, range 18-60). Initially, 20 participants completed the study in a laboratory setting
using hard copies of the materials. However, to increase the response rate, the study was
transposed to an online format. Thus, the majority of participants (n = 139) completed the
study online. Out of the 139 participants, 88 were recruited through the University of
Birmingham's Research Participation Scheme. The study was advertised as an online
psychology study on sexual fantasies that male participants could anonymously take part in.
The remaining 51 participants completed the study online via online survey software (Survey
Monkey). These particular participants were recruited in number of ways. For example, the
link to the study on Survey Monkey was distributed through social networking sites (e.g.,
Facebook); advertised on websites that host online studies (e.g.,
www.onlinepsychresearch.co.uk); and passed on by colleagues at other universities and work
places. Ethical approval was granted from the Science, Technology, Engineering and
Mathematics Ethical Review Committee at the University of Birmingham (see Appendix G).
4.2.2. Materials
4.2.2.1 Creative Experiences Questionnaire
The Creative Experiences Questionnaire (CEQ; Merckelbach, Horselenberg, & Muris, 2001)
is a 25-item self-report measure designed to assess fantasy proneness (see Appendix H). Each
item requires a 'Yes' or 'No' response, with the 'Yes' responses being summed to produce a
total score (ranging from 0 to 25). The higher the score, the greater the level of fantasy
proneness. The CEQ has been found to have good internal consistency ( = .79; Merckelbach
et al., 2001). It should be noted that eight of the items specifically refer to indications of
fantasy-proneness during childhood. For the purposes of the present study, only fantasy
proneness in adulthood was of interest. Therefore, the eight childhood items were omitted in
86
the analysis. According to Saucier and Skrzypiska (2006), this 17-item measure of 'current
fantasy-proneness' has been found to have good internal consistency ( = .77).
4.2.2.2 Vividness of Visual Imagery Questionnaire

The Vividness of Visual Imagery Questionnaire (VVIQ; Marks, 1973) is a 16-item measure
designed to assess individual differences in the vividness and clarity of visual mental imagery
(see Appendix I). The VVIQ is comprised of four general scenes that the respondent must
visualise (e.g., the rising sun). For each scene, four additional images (e.g., "A rainbow
appears") must then be visualised and rated in terms of vividness. Each item is scored on a
scale from 1 ("No image at all") to 5 ("Perfectly clear and as vivid as normal vision"). For the
present study, the scores were reversed so that when summed to create a total score, higher
scores indicated higher imagery vividness. Previous research shows the VVIQ has a
Cronbach's alpha of .85 (D'Argembeau & van der Linden, 2006).
4.2.2.3 Dissociative Experiences Scale-II

The Dissociative Experiences Scale II (DES II: Carlson & Putman, 1993) is a 28-item measure
designed to assess dissociation in non-clinical and clinical populations (see Appendix J). Each
item describes a different dissociative experience and then asks the respondent to indicate the
percentage of time spent (on a scale ranging from 0% to 100%) experiencing the same
phenomenon. Internal consistencies have been shown to be very good, with 's ranging from
.83 to .93 (Bruce, Bruce, Hancock, & Lynch, 2009).
4.2.2.4 Women are Deceitful scale

The Women are Deceitful Scale (WDS; Offending Behaviour Programmes Unit, unpublished)
is a short five-item questionnaire that measures the extent to which women are believed to be
devious and manipulative (see Appendix K). Reponses range from 0 (strongly disagree) to 4
87
(strongly agree) and the items are summed to produce a total score. Higher scores indicate
greater endorsement of the belief that women are deceitful. The scale has a Cronbach's alpha
of .79 (Webster, Bowers, Mann, & Marshall, 2005).
4.2.2.5 Calloused Sex Attitudes Towards Women subscale

The Calloused Sex Attitude Towards Women (CSATW) is one of three subscales that make up
the Hypermasculinity Inventory (Mosher & Sirkin, 1984). The Hypermasculinity Inventory is
an assessment of 'macho personality' and the CSATW subscale specifically examines the extent
to which a man believes sex with women establishes masculine power and female submission,
with little concern for the women or her experience. The CSATW is comprised of ten items that
require one of two responses; one of which is supportive of a callous sex attitude (see Appendix
L). Responses that support a callous sex attitude are given a score of 2, while
the alternative answer is given a 0. A total subscale score is calculated by summing the items.
This specific subscale was chosen as it is used within treatment programmes within
HM Prison Service and has been shown to account for most of the variance related to forceful
sex in college students (Mosher & Anderson, 1986). According to Mosher and Sirkin (1984),
the subscale has a Cronbach's alpha of .79, indicating good internal consistency.
4.2.2.6 Rape-related sexual fantasies

Rape-related sexual fantasies were measured using four items from the sexual fantasy
questionnaire designed by Gray, Watt, Hassan, and MacCulloch (2003). The four items were
identified as relating specifically to rape-related acts (e.g. "Forcing somebody to have sex
against their will"). This particular questionnaire was chosen because: 1) it was designed by
Gray et al. (2003) to specifically address aggressive (or sadistic) sexual fantasies; 2) the items
are a lot less ambiguous than those of the more popular Wilson Sex Fantasy Questionnaire
(Wilson, 1978); 3) while they focus on deviant themes, the items are not as graphic as those in
88
O'Donohue, Letourneau, and Dowling's (1997) Paraphilic Sexual Fantasy Questionnaire; and
4) it has been previously used in studies with non-offenders (Gray et al., 2003; Maile & Jeglic,
2009). Each item requires stating how often each fantasy theme is used via a 4-point Likerttype scale, ranging from 0 (No sexual interest) to 3 (I cant get it out of my mind). All items
are summed to produce a total score for the subscale, with higher scores indicating more
frequent use of rape-related fantasies (see Appendix M; bolded items indicate those that are
rape-related).
4.2.2.7 Sadistic-related sexual fantasies

Using items from the same sexual fantasy questionnaire (Gray et al., 2003), ten items were
deemed to reflect sadistic acts/themes (e.g., "physically hurting the person you are having sex
with"). As described above (Section 4.2.2.6), each item required stating how often the fantasy
theme is used on a 4-point scale ranging from 0 (No sexual interest) to 3 (I cant get it out of
my mind). To produce a total score, all responses are summed. Higher scores reflect more
frequent use of the sadistic-related sexual fantasies (see Appendix M).
4.2.3 Procedure
All male participants were asked to take part in a study looking at the attitudes, thoughts, and
experiences that relate to sexual fantasies. Prior to starting the study, participants were made
aware that their responses would remain completely confidential and as anonymous as
possible (i.e., using an ID code). As such, the participants were asked to be as honest as
possible on each questionnaire. They were also reminded of their rights to withdraw before
providing consent to take part. Each participant completed all of the questionnaires at their
own pace in a randomised order. As stated in the 'Participants' section, most of the participants
(n = 139) completed the study online, while 20 completed it by hand. Following the study, the
participants were debriefed and thanked for their help.
89
4.2.4 Analytic plan

Initially, descriptive (i.e., means and standard deviations) and internal consistency (i.e.,
Cronbach's alpha) tests were conducted for each of the measures. Following this, correlational
analyses were conducted between all of the indicator (observed) variables. This helped
confirm whether or not the multiple indicator variables (e.g., WDS, CSATW) that were related
to a specific latent variable (i.e., distorted sexual attitudes) were associated in the predicted
direction. It was also a way to check for multicollinearity, which occurs when two or more
indicator variables are highly correlated (> .80). This is important to check as multicollinearity
can have negative effects on the regression analyses that occur in SEM, rendering results
uninterpretable (Pedhazur & Schmelkin, 1991).
SEM was used to test how well the data in this study fit the hypothesised model. SEM
is a combination of multiple regression and confirmatory factor analysis (Schreiber, Nora,
Stage, Barlow, & King, 2006) and is regarded as a highly useful analytic method. This is
because it allows for: the analysis of latent (i.e., unobserved) variables; a priori hypothesised
pathways to be tested; and adjustments to be made for measurement error (Byrne, 2010).
As shown in Figure 4.2, the full hypothesised SEM model was comprised of a
structural component (indicated by etched arrows) and a measurement component (indicated
by solid arrows). The structural component of the model had three latent variables (i.e., Rich
Fantasy Life, Distorted Sexual Beliefs, and Aggressive Sexual Fantasies), one of which was
exogenous (Rich Fantasy Life) and two of which were endogenous (Distorted Sexual Beliefs,
and Aggressive Sexual Fantasies). The measurement component relates to the links between
indicator (observed) variables and their associated latent variable, as well as the error
variances. As shown in Figure 4.2, the indicator variables associated with Rich Fantasy Life
were fantasy proneness (CEQ), vividness of visual imagery (VVIQ), and dissociation (DESII). The Distorted Sexual Beliefs latent variable was comprised of calloused sex attitudes
90
(CSATW) and women are deceitful (WDS). Finally, Aggressive Sexual Fantasies was made
up of Rape-related and Sadistic sexual fantasies.
Figure 4.2: Hypothesised structural equation model. Note: Etched arrows between the latent
variables (ellipses) signify the structural model. The indicator variables (rectangles) and errors
(circles with an 'e' inside) represent the measurement model.
Having specified the model based upon theoretical assumptions derived from the DPM-ST, the
model then needed to be identified. Model identification is the extent to which the parameters
of the model (i.e., the paths being estimated) correspond to the available data points. The aim
is to specify a model that has more data points than parameters so that it has positive degrees
of freedom (i.e., known as an 'over-identified' model). This makes the model scientifically
useful as it can be rejected (Byrne, 2010). The present model has 16 parameters; that is, seven
error variances; four indicator paths (the other three are fixed to the value of one and so are not
estimated); two structural paths (between latent variables); two disturbances (i.e., error term
associated with the two endogenous latent variables); and one factor variance (for the
exogenous variable). As there were seven measured variables, the number of data points is:
7(7+1)/2 = 28. Therefore, the model is over-identified as there are more data points than
parameters (28 versus 16, respectively).
91
With regards to sample size, a common 'rule of thumb' in SEM is to have 10

participants per parameter (Schreiber et al., 2006). With 16 parameters, 160 participants were
required. Therefore, 159 participants in this study was just about enough to conduct the
analysis. Using data from all 159 participants (with no missing data), the next step involved
evaluating the fit of the hypothesised model. This was done with the aid AMOS 21; a software
package that can be used to run SEM analyses. Based on recommendations in the SEM
literature (e.g., Businelle, Kendzor, & Wetter, 2010; Byrne, 2010; Hu & Bentler, 1999;
Schreiber et al. 2006), various goodness-of-indices were consulted to determine how well the
model fitted with the observed data.
Specifically, a Chi-square goodness-of-fit index (2) was used to test the hypothesis
that the model is significantly different to the data. Thus, a good model will not differ, as
indicated by a p-value >.05. In addition, the Comparative Fit Index (CFI) and Tucker-Lewis
Index (TLI) were used, both of which compare the hypothesised model with a null (i.e., worstfitting) model. According to Hu and Bentler (1999), scores >.95 indicate an excellent fit. Also,
Root Means Square Error of Approximation (RMSEA) was also used. The RMSEA,
essentially, measures the error of approximation in the population, expressed per degree of
freedom (Byrne, 2010). Since degrees of freedom are related to the number of parameters in the
model, RMSEA is sensitive to model complexity and favours models with less parameters. An
RMSEA of 0.8 of more indicates greater error of approximation and, therefore, poor fit, while
an RMSEA below 0.05 is considered excellent fit. However, a cut-off value of 0.6 (Hu
& Bentler, 1999), or a stringent upper limit of 0.7 (Steiger, 2007), can be suggestive of a
reasonable or good-fitting model.
Finally, due to the small sample size, it was predicted that the data would be not be
normally distributed (Ievers-Landis, Burant, & Hazen, 2011). Therefore, a nonparametric
"bootstrapping" procedure was carried out after the initial model was tested. Essentially,
bootstrapping is a resampling method that involves treating the original sample as the
92
population so that numerous subsamples (the same size as the original) can be randomly drawn
from it. As a result, the same parameter estimates and tests of fitness are able to be tested on
each of the subsamples so that a distribution of the result is produced (Ievers-Landis et al.,
2011). In this study, a bias-corrected bootstrap procedure was used with 95% confidence
intervals. Using AMOS 21, parameter estimates were tested across 1000 subsamples (Cheung
& Lau, 2008). This was followed by the Bollen-Stine bootstrap; a procedure that allows
researchers to ascertain whether the original model fits with the model derived from the
multiple bootstrap samples (Ievers-Linden et al., 2011).
4.3 Results
4.3.1 Means, internal consistencies, and correlations among indicators
Table 4.1 displays the means and standard deviations for each of the observed variables, as
well as the internal consistency () associated with each measure. As can be seen, most of the
measures had very good internal consistencies as indicated by high Cronbach's alphas, with
the exception of the CSATW scale ( =.63).
Table 4.1:
Means, standard deviations, and internal consistencies for the indicator (observed) variables
Indicator
N
Mean
SD
Cronbach's
Variable
Alpha ()
159
4.9
3.2
.75
CEQ
VVIQ
159
57.3
10.7
.92
DES-II
159
17.5
13.7
.95
CSATW
159
4.1
4.3
.63
WDS
159
8.5
3.7
.85
Rape SF
159
.74
1.8
.89
Sadistic SF
159
2.7
3.8
.88
Note: SF = Sexual Fantasies

93
Spearman's Rho correlations are shown in Table 4.2. Spearmans Rho was used because, as
predicted, the data in this study was not normally distributed (see Section 4.3.2). Looking at
Table 4.2, it can be seen that the strongest correlations were between indicators associated
with a specific latent variable, with the exception of VVIQ and DES-II. For example, in
relation to Rich Fantasy Life, the CEQ showed a positive relationship with the VVIQ and the
DES-II; in relation to Distorted Attitudes about Women, the CSATW and WDS were
correlated; and in relation to Aggressive Sexual Fantasies, the two sexual fantasy subscales
(i.e., rape-related and sadistic) were strongly associated. None of the correlations exceeded
0.8, thereby, ruling out multicollinearity.
Table 4.2:
Correlations (using Spearmen's Rho) for all indicator (observed) variables
Indicator
1
2
3
4
5
Variable
1. CEQ
--
--
--
--
--
--
2. VVIQ
.23**
--
--
--
--
--
3. DES-II
.49***
.09
--
--
--
--
4. CSATW
.08
-.01
.15
--
--
--
5. WDS
.23**
-.09
.29**
.40***
--
--
6. Rape SF
.16*
.09
.20*
.33***
.25**
--
7. Sadistic SF
.20*
.09
.21**
.23**
.18*
.61***
*** <.001, ** < .01, * < .05 Note: SF = Sexual Fantasies
4.3.2 Testing the SEM model

SEM was run using AMOS 21. Maximum likelihood (ML) was the chosen method of
parameter estimation because, according to Schreiber et al. (2011), ML can be appropriately
used with sample sizes ranging from 100-150. Figure 4.3 shows the final SEM model with the
standardised coefficients derived from the SEM analysis added. As can be seen, the paths
between the three latent variables were significant. The specific details of these two paths
94
coefficients are as follows: Rich Fantasy Life Distorted Sexual Beliefs (standardised =
.50; unstandardised = .52; standard error [SE] = .15, p = <.001); and Distorted Sexual
Beliefs Aggressive Sexual Fantasies (standardised = .58; unstandardised = .40, SE =
.09; p = <.001). Furthermore, all of the non-fixed indicators were significant at the p <.001
level, except for the path between Rich Fantasy Life to VVIQ (p <.05).
Figure 4.3: Final structural equation model.

** <.001, * <.05.
In evaluating the model, the goodness-of-fit indices indicated that the model had a reasonable
to good fit with the observed data; 2 (12) = 20.43, p >05; CFI = .97; TLI = .95; RMSEA =
.067. As indicated, the Chi-square statistic was not significant, suggesting that the
hypothesised model did not differ from the observed data. The CFI and TLI indices were
above and equal to the cut-off points for a good fitting model, respectively. Finally, the
RMSEA suggested a reasonably fitting model. Taken together, no further adjustments or
model modifications were deemed necessary to improve the model (Byrne, 2010). While these
results were encouraging, these indices are susceptible to inflation by small samples (IeversLandis et al., 2011). Moreover, as predicted (see Section 4.2.4), the data was found to be nonnormally distributed, indicated by a high multivariate kurtosis value (z-statistic = 16.76). For
normality to be established, this value should be < 5.00 (Bentler, 2005). Thus, the following
95
section outlines the results of the bootstrapping procedures that were applied to help validate
the model.
4.3.3 Bootstrapping procedures

Bootstrapping was used to examine the stability of the original hypothesised model and
determine whether it was unbiased. This was done via two steps (Ievers-Landis et al., 2011).
The first involved comparing the bootstrap regression weight means (unstandardised and
standardised) derived from the 1000 bootstrap samples with those produced from the original
sample. If they did not considerably differ, it would provide evidence of an unbiased model.
As shown in the fourth column of Tables 4.3 and 4.4, the difference between the weights
(unstandardised and standardised) of the mean bootstrap sample did not greatly differ from
those of the original sample. Taking the path between Rich Fantasy Life and Distorted Sexual
Attitudes in Table 4.3 as an example, it can be seen that the difference between the original
and bootstrap unstandardised regression weights was only -.01. These findings provided the
first indication that the model was unbiased.
The second step to further determining whether the model was unbiased involved
comparing the SE of the mean bootstrap with the SE-Bias (which is the difference between the
SE of the original sample and the bootstrap sample) for all paths. An unbiased model is
indicated if the SE-Bias is less than the SE of the mean bootstrap. As shown in the last two
columns of Tables 4.3 and 4.4, all of the SE-Biases were less than SE's derived from the mean
bootstrap. Thus, the overall bootstrapping procedure suggested that all of the paths were
unbiased.
Table 4.3: Comparisons Between the Original Sample and Mean Bootstrap Sample of 1,000
(using unstandardised regression weights)
Original
Mean
Difference
SE of
SE-Bias
Path
Sample Bootstrap
(Bias)
Mean
96
.52
Sample
.51
-.01
.40
.42
.02
.17
.005
1.8
1.85
.02
.38
.012
.97
1.03
.06
.34
.011
Fixed to 1
Fixed to 1
7. 'Rich Fantasy Life' to 'VIVQ'
.92
.85
-.08
.39
.012
8. 'Rich Fantasy Life' to

'DES-II'
9. 'Aggressive SF' to
'Rape-related SF'
4.2
4.4
.09
2.1
.065
Fixed to 1

'Distorted Sexual Attitudes'
2. 'Distorted Sexual Attitudes'
to 'Aggressive SF
'Sadistic SF'
to 'CSATW'
to 'WDS'
6. 'Rich Fantasy Life' to 'CEQ'
Bootstrap
.27
.008
SE = Standard Error, SF= Sexual Fantasies
Table 4.4: Comparisons Between the Original Sample and Mean Bootstrap Sample of 1,000
(using standardised regression weights)
Original
Mean
Difference
SE of
SE-Bias
Path
Sample Bootstrap
(Bias)
Mean
97

'Distorted Sexual Attitudes'
to 'Aggressive SF'
'Sadistic SF'
To 'CSATW'
to 'WDS'
6. 'Rich Fantasy Life' to 'CEQ'
.50
Sample
.48
-.03
Bootstrap
0.17
0.005
.58
.57
-.02
0.14
0.004
.82
.81
-.002
0.09
0.003
.65
.65
.004
0.12
0.004
.59
.58
-.003
0.11
0.003
.74
.78
.04
0.18
0.006
7. 'Rich Fantasy Life' to 'VIVQ'
.21
.21
.00
0.10
0.003

'DES-II'
'Rape-related SF'
.74
.73
-.01
0.16
0.005
.95
.96
.01
0.09
0.003
SE = Standard Error, SF= Sexual Fantasies
Finally, a final test of the model's validity was conducted using the Bollen-Stine method. This
bootstrapping procedure involves transforming the data so that it perfectly matches the model.
Then multiple samples are drawn from the transformed sample. Essentially, it uses the 2
statistic to assess whether the bootstrap-derived model significantly differs from the originally
hypothesised model. Thus, a Bollen-Stine 2 with a p value >.05 indicates that the two models
do not significantly differ. In this study, the Bollen-Stine statistic and the p-value was found to
be .126. Thus, this provided further validation of the original SEM model.
4.4 Discussion
The main aim of this study was to test the hypothesis that individual differences related to the
ability to fantasise in general (i.e., high fantasy proneness, dissociation, and vividness of visual
imagery), in conjunction with congruent explicit cognitions, influences the use of aggressive
sexual fantasies in non-offending males. This hypothesis was represented as a
conceptual single-path model where Rich Fantasy Life, via Distorted Attitudes about Women,
98
influenced Aggressive Sexual Fantasies. Using structural equation modelling with data from
159 male non-offenders, the model was found to be a good fit with the data. As a result,
support for the model (and, therefore, the main hypotheses) was obtained.
In line with the DPM-ST, the results suggest that aggressive sexual fantasies are more
likely to be used by male non-offenders if they have: 1), a greater ability to dissociate from
reality and engage in fantasies involving vivid mental imagery; and 2), distorted explicit
cognitions that are congruent with the aggressive themes within the sexual fantasy.
Furthermore, the findings would appear to add to the literature showing a relationship between
distorted attitudes about women and deviant sexual fantasies about women (i.e., those involving
coercion, dominance, force) (Greendlinger & Byrne, 1987; Zurbriggen & Yost,
2004). While these previous studies have been purely correlational, the present study provides
support for a possible causal direction. This causal direction has previously been articulated in
the literature. For example, in their model of risk for sexual offending, Beech and Ward (2004)
proposed that offence-supportive cognitions produce deviant thought/fantasies.
The results of this chapter can also be contrasted with the Knight and Sims-Knight's
(2003) three-path model of sexual coercion against women. While their model is more
complex in nature, given that it attempts to explain a far more complex phenomenon (i.e.,
sexual coercion), they did propose a path between two latent variables called
'Callousness/Unemotional' (CU) and 'Aggressive Sexual Fantasy'. Using SEM to test the
model with adult sexual offenders (n = 275), Knight and Sims-Knight (2003) found support
for this path ( = .39). However, when they tested the model with community males, CU
appeared to serve as a suppressor variable ( = -.25). It should be noted that, while the
Aggressive Sexual Fantasy construct was made up of aggressive and sadistic sexual fantasies
subscales (similar to the current study), the CU factor was comprised of measures of general
callousness/unemotionality (i.e., negative masculinity, superficial charm, and dyscontrol).
Thus, given the results of the present study, perhaps distorted attitudes about women are more
99
pertinent to explaining aggressive sexual fantasies in non-offending males than is a general

callous/unemotional trait. However, that is not to say that such a trait does not further increase
the likelihood of using aggressive sexual fantasies. For example, Williams, Cooper, Howell,
Yuille, and Paulhus (2009) found that psychopathy - which is comprised of callous/emotional
traits (Knight & Sims-Knight, 2003) - was associated with sexual fantasies of bondage and
sadism in a sample of non-offenders.
There are a number of strengths associated with this study. First, the hypotheses, latent
variables, and a priori SEM model were all empirically and theoretically informed, allowing
for stronger conclusions about the model to be drawn. Second, rather than examining
observed variables (i.e., fantasy proneness, dissociation, and vividness of visual imagery) in
relation to aggressive sexual fantasies, the use of SEM allowed for the testing of latent
variables defined by multiple observed variables. Third, in other multivariate analyses (e.g.,
regression), the error of observed variables is treated as having disappeared. This is
problematic because it is essentially analogous to ignoring error, which can lead to major
inaccuracies (Byrne, 2010). SEM, however, is able to avoid these potential problems as it
takes into account and makes adjustments for measurement error.
In addition, this study is one of the first to directly examine the theorised link between
sexual fantasies and the constructs of fantasy proneness and dissociation. Moreover, as the
results suggest (see Table 4.1), a positive relationship appears to exist between sexual fantasies
(i.e., involving rape and sadistic acts) and both fantasy proneness and dissociation. Future
research should extend these findings by examining the relationship between these variables
and other types of sexual fantasy. Finally, and pertinent to this thesis, the study supports some
of the unique assumptions of the DPM-ST.
4.4.1 Limitations and suggestions for future research
100
The present study is not without limitations. First, the sample size was not quite adequate for a
SEM study. The 'rule of thumb' for estimating sample size is that there should be 10
participants per estimated parameter. With 16 parameters, 160 participants were required. This
study involved 159 participants, resulting in 9.9 participants per parameter. Thus, the sample
size was borderline appropriate. As such, the data was predicted and found to be nonnormally distributed. In these cases, Ievers-Landis et al. (2011) have advised using
bootstrapping procedures to test whether the model is unbiased. In this study, results from two
bootstrapping procedures verified that the model was valid or unbiased. Nevertheless, "SEM
is still a large sample analysis technique" (Schreiber et al., 2011, p. 334) and so the model
should certainly be tested with a much larger sample.
Second, it is strongly advocated that latent variables be defined by multiple indicators
as opposed to one aspect of measurement. This is because a single indicator is unlikely to
account for all of the variance in the latent variable, causing the latent variable to be reflective
of biased or unreliable information (Raykov & Marcoulides, 2000). While all of the latent
variables in the present study had more than one indicator, two of them were made up of only
two indicators. This can sometimes be problematic because two indicators per latent variable
can sometimes cause estimation problems (Bollen, 1989; Kline, 2004). However, as noted by
Violato and Hecker (2007), most SEM models often only use two indicators per latent
variable. Nevertheless, it would be beneficial to run a model with three indicators for the
Distorted Attitudes about Women and Aggressive Sexual Fantasies factors. Note that this
would mean adding more parameters to the model, which would require a larger sample.
Third, no information was gained as to the sexual orientation of the sample. This
highlights a potential issue because if there were any homosexual participants present in the
sample, it is possible that their attitudes about women would have no bearing on their
aggressive sexual fantasies, as the content would likely involve men. Future research would
benefit from gaining information on sexual orientation or having homosexuality as an
101
exclusion criterion. Alternatively, testing the current model in relation to homosexual attitudes
and homosexual fantasies would be of benefit, both as way of addressing this limitation and
further validating the core premise of the SEM model.
Finally, although the model was a reasonably good fit to the data, other models may
have been an identical or better fit. In light of this, it is important to keep in mind that the
present results suggest only that the model is plausible; not conclusively correct. However, the
model is a first step into examining the impact of fantasy proneness, dissociation, and imagery
vividness on sexual fantasies. Thus, validation of the model with larger samples and offender
samples is required. Also, applying the model to include different explicit cognitions and
sexual fantasies (e.g., child-related, non-deviant) are also warranted.
CHAPTER 5
102
TESTING WHETHER SEXUAL FANTASIES

REQUIRE WORKING MEMORY RESOURCES
Chapter Rationale
The Dual-Process Model of Deviant Sexual Thinking proffers some novel assumptions about
the cognitive mechanisms associated with envisioning a sexual fantasy. The first and primary
assumption is that sexual fantasies are a controlled cognitive process that one deliberately
engages in. As such, the DPM-ST states that sexual fantasies rely on the processing resources
of working memory, since visual information drawn from the external world and the internal
world (i.e., long-term memory) have to be activated, retained, and manipulated in the minds
eye. This assumption is integral to the model and will have a number of important implications
if it is corroborated. Therefore, this next chapter describes a study aimed at establishing
whether working memory underpins the act of sexually fantasising.
5.1 Introduction
103
A core assumption of the Dual-Process Model of Sexual Thinking (DPM-ST) is that sexual
fantasising is a controlled cognitive process that requires the resources of working memory
(WM). WM is conceptualised as a limited-capacity storage system for information that is
necessary to complete an immediate or current cognitive operation (Baddeley & Hitch, 1974).
According to Baddeley and Hitch, WM is comprised of a main control system (the central
executive) that controls attention and the allocation of resources. It is also comprised of two
subsidiary slave systems; the visuospatial sketchpad and the phonological loop. The former
subsystem is necessary for the retention and manipulation of visual information, while the
latter does the same for auditory information. It has been assumed that the maintenance of
information in the slave systems is essential for the operations underpinning mental imagery
(Baddeley & Andrade, 2000).
One way to test if a cognitive operation requires WM is to employ a dual-task. This
involves having an individual perform the cognitive operation of interest at the same time as
performing another task known to require WM. If this cognitive operation requires the
resources of WM then an individuals performance will be impaired relative to when it is
performed with no dual task. This is because the dual task leaves less WM capacity/resources
for the cognitive operation. With respect to mental imagery (as the cognitive operation),
bilateral eye-movements (EMs) have been increasingly used as a dual task (van den Hout &
Engelhard, 2012) as evidence suggests that bilateral EMs tax working memory (Andrade,
Kavanagh, & Baddeley, 1997). For example, it has been found that responses on a reaction
time task (known to require WM) are significantly slower when EMs are performed
simultaneously, relative to when no EMs are made (Engelhard, van Uijen, & van den Hout,
2010; van den Hout et al., 2011).
There is a growing body of research showing that the vividness and emotionality of
negative memories are reduced after one undergoes a session of bilateral EMs at the same
time as envisioning the negative memory (Andrade et al., 1997; Gunter & Bodner, 2008;
104
Maxfield, Melnyk, & Hayman, 2008; van den Hout, Muris, Salemink, & Kindt, 2001; van den
Hout et al., 2011). Moreover, Gunter and Bodner (2008) found that these effects are greater
for individuals with a naturally lower WM capacity, further supporting the account that EMs
tax WM. A recent meta-analysis of these studies revealed a large effect size for reductions in
vividness of negative memories (Lee & Cuijpers, 2013). Given that EMs are shown to tax
WM, these findings suggest that mental imagery requires WM resources.
This conclusion is reinforced by research showing that the effects of EMs are not
confined to negative memories. For example, EMs have been found to impair positive
memories (Hornsveld et al., 2011) and negative imagery related to the future (Engelhard, van
den Hout, Janssen, & van der Beek, 2010; Engelhard et al., 2011). Furthermore, EMs have
also been found to reduce the vividness of food imagery, as well as the level of craving
related to the food (Kemps, Tiggemann, Woods, & Soekov, 2004; McClelland, Kemps, &
Tiggemann, 2006).
In light of this research, it can be argued that sexual fantasies as a form of mental
imagery require the resources of WM. Thus, the present study aimed to test this hypothesis
by using EMs as a dual-task. As research of this kind of research has never been conducted
with sexual fantasies in general, it was decided that deviant sexual fantasies should not be the
sole focus and that non-offenders should be tested before sex offenders.
Given that future-based imagery is affected by EMs in the same way as memory-based
imagery (Engelhard et al., 2010b), this study tested both memory and imagination-based
fantasies separately with the prediction that they will not significantly differ in terms of how
they are affected by EMs. Thus, drawing upon the methods used in previous research, this
study comprised two Experimental conditions (one where memory-based sexual fantasies
were envisioned while making EMs; and one where imagination-based sexual fantasies were
envisioned while making EMs), and two Control conditions (one where a memory-based
105
sexual fantasy was envisioned with no EMs; and one where an imagination-based sexual
fantasy was envisioned with no EMs).
For each condition, sexual fantasies were rated on three variables (emotionality,
vividness, and arousability) both before and after the task. Arousability was a unique addition
to this study as it is an effect of sexual fantasies (Gee, Ward, & Eccleston, 2003). This is
similar to the study by Kemp et al. (2004) who added a measure of food craving to their study
since food craving is an effect of that particular imagery. Since EMs were expected to affect
memory and imagination-based fantasies to the same extent (i.e., not significantly differ),
fantasy type was not considered as a factor in the design of this study. Thus, the following
hypotheses were tested:
Hypothesis 1: The vividness of memory and imagination-based sexual fantasies will reduce
for the EM condition relative to the control condition.
Hypothesis 2: The emotionality of memory and imagination-based sexual fantasies will

reduce for the EM condition relative to the control condition.
Hypothesis 3: The arousability of memory and imagination-based sexual fantasies will

5.2 Method
5.2.1 Participants
In this study, a within-subjects design was used to reduce the effect of individual differences
(e.g., habituation rates, levels of imagery vividness, emotionality, and arousability). A sample
of 47 students was recruited from the University of Birmingham to take part in a study looking
at the cognitive and structural properties of sexual fantasies. The sample size was a
106
result of convenience sampling via the online Research Participation Scheme (RPS). The
sample had a mean age of 19.4 years old (SD=1.1, range 19-21 years) and was comprised of
38 female participants and 9 male participants. The sample was comprised of mainly
Caucasian participants (83%), although individuals from other ethnic backgrounds also took
part, including Asian (11%), Black (2%), and Mixed Race (4%). Also, the sample was
comprised of mainly heterosexual individuals (94%), although three homosexual individuals
(6%) took part. Approximately half of the sample identified themselves as Single (51%),
while the other half identified themselves as being In a relationship (49%). Ethical approval
for this study was granted from the Science, Technology, Engineering and Mathematics
Ethical Review Committee at the University of Birmingham (see Appendix N.)
5.2.2 Materials
5.2.2.1 Adapted Wilson Sexual Fantasy Questionnaire (WSFQ; Wilson, 1978)
Sexual fantasies were measured using the Wilson Sex Fantasy Questionnaire (WSFQ; Wilson,
1978). A more detailed description can be found in Chapter 3 (Section 3.2.2.3). However, for
this study, the WSFQ was adapted to include six extra questions. This was done to elicit other
necessary information about each fantasy. These six questions were: 1) Does the fantasy
involve an elaborate scenario, or is it just a passing thought? 2) Is it based on the memory of a
previous experience, or pure imagination? 3) In general, does it automatically pop in your mind
or do you deliberately bring it to mind? 4) Do you expect it to occur in the future, or is it an
unrealistic, fanciful idea? 5) How arousing do you find this fantasy? 6) How do you evaluate
this fantasy? 7) Do you watch or read pornography related to this fantasy? All questions
required one of two categorical-type responses (e.g., Memory versus
Imagination for Question 2), with the exception of questions 5 and 6, as these were scored
on a Likert-type scale (see Appendix O). To aid participants in answering the additional
questions, guidelines were provided (see Appendix P).
107
5.2.2.2 Visual analogue scales

Following many of the previous studies in EMs (e.g., van den Hout et al., 2011), emotionality,
vividness, and arousability levels for each sexual fantasy were measured using visual
analogue scales (VAS; see Appendix Q). Each VAS was comprised of a single 10cm line that
had descriptive anchors at each end. The emotionality VAS ran from extremely unpleasant
(left) to extremely pleasant (right); the vividness VAS ran from not clear at all (left) to
very clear (extreme right); and the arousability VAS ran from extremely non-arousing
(left) to extremely arousing (right). For each variable, participants had to mark a vertical line
through the horizontal line of the VAS wherever it felt appropriate. The VASs were scored
by manually measuring (i.e., using a ruler) the position of each response.
5.2.2.5 Eye-movement task

Using E-Prime, a 1cm wide white dot was programmed to move back and forth across a
computer screen (black background) at a rate of about one movement per second (i.e., back and
forth the screen in one second). The horizontal movement of the white dot is designed to
facilitate bilateral eye-movements in participants and has been successfully used by previous
researchers (van den Hout et al., 2011). One of these researchers (Daniel Beetsma) kindly
shared the basic E-Prime script for the moving dot, which was adapted to suit the present study.
For example, following the procedure used by the previous researchers, the circle was
programmed to make 24 full right-to-left movements over four separate blocks (24 per block).
Each block was separated by a 10 second break (where the word Rest was presented on
screen). Thus, the task comprised bilateral 96 movements in total (van den Hout et al., 2001;
Engelhard et al., 2010b). Ninety-six eye movements have been used in previous research (and
the present study) in order to avoid the issues associated with dose-exposure research. For
instance, any more EMs may induce mere-exposure effects that overshadow the effects that
108
EMs can have, whereas any less may be too little to observe any EM effects (van den Hout,
personal communication, March, 2013).
A similar task was also programmed for the control (no eye-movement) condition. This
simply involved a blank black screen being displayed for four 24-second blocks, each
separated by a 10 second break. During the presentation of the blank screen, participants
would envision a sexual fantasy. When the screen presented the word Rest, they would have
to rest for 10 seconds.
5.2.3 Procedure
Each participant entered the lab and sat in front of a computer. Each participant was informed
about what they had to. That is, to envision a series of sexual fantasies that they regularly use
while doing a secondary task at the same time. After providing informed consent, all
participants first completed the Adapted WSFQ. The researcher then identified four sexual
fantasies from the completed Adapted WSFQ; two based upon an experience and two based
upon pure imagination. The fantasies that were most frequently used (i.e., a score of 3 or
more) were chosen. These four sexual fantasies were used for each of the four conditions.
Following completion of the Adapted WSFQ, all participants were first told to
envisage a sexual fantasy for 20 seconds (nb: The participants were informed about which of
their four selected fantasies they had to envision. That is, a number corresponding to the
selected WSFQ fantasy theme was written by the researcher on the top of the VAS sheet).
After envisioning the fantasy for 20 seconds, participants then completed a pre-task rating of
the fantasys emotionality, vividness, and arousability using the VASs.
Next, all participants took part in the four conditions (i.e., EM-Memory, EMImagination, No EM-Memory, and No EM-Imagination), the order of which was
counterbalanced by Experimental and Control Condition. For each condition, the instructions
were presented to them on a computer screen (via E-prime). During the Experimental
109
conditions (EM-Memory, EM-Imagination), participants were instructed to visualise a sexual

fantasy with their eyes open for a period of 96 seconds (i.e., four 24-second blocks separated
by a 10 second rest). At the same time as envisioning the fantasy, participants were told to
make 96 EMs by following a white dot that moved across the computer screen. After the EM
task ended, a screen appeared informing participants to re-visualise the same sexual fantasy
and re-rate it using a set of new VASs. This procedure was completed twice: (1) for a
memory-based fantasy; and (2) for an imagination-based fantasy. The order of which fantasy
was visualised first (i.e., memory versus imagination) was also counterbalanced between
participants within the Experimental condition to ensure that order effects were reduced.
In the two Control conditions, participants were asked to simply visualise their
memory and imagination-based fantasies for a period of 96 seconds with their eyes open,
again in a counterbalanced order. No eye-movements were made. After each control task,
participants re-visualised and re-rated the sexual fantasy using new VASs. After the study
ended, each participant was debriefed and thanked for their time.
5.2.4 Statistical Analyses

Memory and imagination-based sexual fantasies were not predicted to differ with regards to
the effects of EMs. Thus, fantasy type was not included as an independent variable in the
analyses. In order to highlight the similar pattern of results for each fantasy type, the analysis
of the three dependent variables (vividness, emotionality, and arousability) were analysed
separately for memory-based and imagination-based fantasies.
Thus, for each type of sexual fantasy (Memory and Imagination), a two (pre-test, posttest) x two (EM condition, control condition) repeated measures ANOVA was conducted on
each dependent variable to examine any main and interaction effects. Also, for each
dependent variable, pre-test scores were subtracted from post-test scores to produce a
difference score for each condition (i.e., a negative mean difference signifies a decrease in the
110
dependent variable). Whenever a significant interaction effect was found, paired t- tests were
conducted on the mean differences to compare the conditions. Effect sizes (Cohens d) were
calculated using Morris and DeShons (2002) equation to correct for any dependence among
the means. According to Cohen (1988), a d of 0.2 is regarded as a small effect size; 0.5 is
regarded as a medium effect size, and 0.8 indicates a large effect size.
5.3 Results
Before describing the experimental results, it is useful to examine the participants scores on
the WSFQ so as to gain an idea of their sexual fantasies they most often engage in.
Figure 5.1: Scores on each WSFQ subscale as a percentage of

the overall score
As Figure 5.1 shows, Intimate fantasies were the most frequently used (M = 30.2, SD = 7.9).
The next most frequently used sexual fantasies were Exploratory (M = 9.5, SD = 6.9), with
Impersonal (M = 6.5, SD = 4.6) and Sadomasochistic fantasies (M = 6.4, SD = 7.1) both
being used the least. This pattern of results has also been found in both males and females in
111
previous WSFQ studies (Wilson & Lang, 1981). Paired-samples t-tests (with Bonferroni
corrections) indicated that the scores on each subscale were significantly different from each
other (all ps <.001), except between the Impersonal and Sadomasochistic subscales. In
addition, independents t-test showed no significant differences between males and females on
any of the subscales in the present study.
Table 5.1 and Table 5.2 show the mean pre-ratings and post-ratings for vividness,
emotionality, and arousability for memory-based fantasies and imagination-based fantasies,
respectively. It can be seen, in both tables, that there is a trend for the ratings to decrease in
the EM condition and increase in the control condition.
Table 5.1: Means and SDs of vividness, emotionality, and arousability ratings before and
after eye-movements for memory-based sexual fantasies.
Eye-movement condition
Control condition (Fantasy only)
Pre
Post
Pre
Post
Vividness
7.8 (1.7)
6.9 (2.3)
7.8 (1.5)
8.5 (1.5)
Emotionality
7.6 (1.6)
7.2 (1.8)
7.6 (1.8)
8.1 (1.9)
Arousability
7.5 (1.5)
6.8 (2.0)
7.4 (1.9)
7.9 (1.9)
Table 5.2: Means and SDs of vividness, emotionality, and arousability ratings before and
after eye-movements for imagination-based sexual fantasies.
Eye-movement condition
Control condition (Fantasy only)
Pre
Post
Pre
Post
Vividness
6.2 (2.4)
5.8 (2.4)
5.8 (2.3)
7.0 (2.2)
Emotionality
7.0 (2.2)
6.5 (2.0)
6.6 (2.2)
7.0 (2.5)
Arousability
6.6 (2.5)
5.8 (2.5)
6.3 (2.4)
6.7 (2.4)
112
5.3.1 Hypothesis 1: The vividness of memory and imagination-based sexual fantasies will
5.3.1.1 Memory-based sexual fantasies

First, paired t-tests showed that baseline (or pre-test) vividness ratings did not significantly
differ between the EM and Control condition. Pre- and post-test ratings for fantasy vividness
were subjected to a two-way ANOVA, with Time (pre-test, post-test) and Condition (EMs
with memory-based fantasy, no EMs with memory-based fantasy) as within-subjects factors.
Results revealed a significant main effect of Condition F(1, 46) = 14.15, p= <.001, as well as
for the crucial Time x Condition interaction F(1, 46) = 28.67, p = <.001.
Figure 5.2: Mean difference scores and SEs for vividness for all four conditions.
In order to meaningfully decompose the interaction, a paired t-test was conducted on the mean
difference scores (i.e., pre-test minus post-test ratings) for each condition (Engelhard et al.,
2010b; Hornsveld et al., 2011). Figure 5.2 depicts these mean differences. As can be seen,
vividness decreased in the EM condition and increased in the control condition (no EMs).
113
Results revealed that the decrease in vividness after EMs was significantly different to the
condition involving fantasising only, t(46) = 5.4, p <.001, d =.80.
5.3.1.2 Imagination-based sexual fantasies

Again, paired t-tests showed no significant differences on baseline vividness between the EM
and control conditions. A two-way ANOVA was conducted, with Time (pre-test/post-test)
and Condition (EMs with imagination-based fantasy/no EMs with imagination-based fantasy)
as within-subjects factors. The two main effects were not significant but the crucial Time x
Condition interaction was significant, F(1, 46) = 18.18, p = <.001. To meaningfully
decompose this interaction, a paired t-test was conducted on the mean difference scores (i.e.,
pre-test minus post-test ratings) for each condition. As can be seen in Figure 5.2, vividness
decreased in the EM condition and increased in the control condition (no EMs). The paired ttest showed these two mean differences were significantly different, t(46) = 4.3, p <.001, d
=.64.
5.3.1.3 Comparing memory and imagination-based fantasies

As depicted in Figure 5.2, the decrease in vividness for the memory-based fantasies after EMs
appears greater than that for imagination-based fantasies. The reverse is also the case when
looking at the increase in vividness after making no EMs. However, as stated earlier, no
differences were expected between the two fantasy types. Thus, to ensure these differences
were not significant, two post-hoc paired t-tests (with Bonferroni corrections; = .025) were
conducted on the mean differences for the two fantasy types (i.e., one t-test per condition). As
expected, the results showed no significant differences (ps both >.05).
5.3.2 Hypothesis 2: The emotionality of memory and imagination-based sexual fantasies

will reduce for the EM condition relative to the control condition.
114

As with vividness, paired t-tests showed that the baseline ratings for emotionality did not
differ between conditions. A two-way ANOVA conducted on the pre-test and post-test
emotionality ratings revealed a significant main effect of Condition, F(1, 46) = 5.44, p <.05,
and a significant Time x Condition interaction effect, F(1, 46) = 5.91, p <.05. To decompose
this interaction, the mean difference scores for the EM and control condition (see Figure 5.3)
were analysed using a paired t-test. It was found that the decrease in emotionality after EMs
was significantly different to the change in emotionality found in the control condition, t(46)
= 2.43, p <.05, d =.40.
Figure 5.3: Mean difference scores and SEs for emotionality for all four
conditions.

For both conditions, no significant differences were found between the baseline emotionality
ratings. Running a two-way ANOVA on the pre- and post-test ratings for emotionality
revealed no significant main effects. However, the crucial Time x Condition interaction was
found to be significant, F(1, 46) = 6.1, p <.05. A paired t-test showed that the mean difference
115
for the EM condition was significantly different to that of the control condition, t(46) = 2.5, <
.05, d =.40.

Small observable differences between memory and imagination-based fantasies can be seen for
each condition (see Figure 5.3). Thus, to ensure these differences were not significant, two
paired t-tests (with Bonferroni corrections) were conducted. The results revealed no
significant differences between the two fantasy types for each condition (ps both >.05).
5.3.3 Hypothesis 3: The arousability of memory and imagination-based sexual fantasies

will reduce for the EM condition relative to the control condition.

In line with the other ratings for memory-based fantasies, no significant differences were
found for baseline ratings of arousability. A two-way ANOVA was conducted on the pre-test
and post-test scores, which revealed a significant main effect of Condition, F(1, 46) = 5.5,
<.05, as well as a significant Time x Condition interaction effect, F(1, 46) = 13.81, p <.01.
Figure 5.4 shows the mean differences for each condition. Using a paired t-test on these mean
difference scores, it was found that the decrease in arousability in the EM condition was
significantly different to the increase in arousability in the control condition, t(46) = 3.71, p
<.01, d =.54.
116
Figure 5.4: Mean difference scores and SEs for arousability for all four
conditions.

A paired t-test showed that baseline arousability scores for imagination-based fantasies were
not significantly different between conditions. A two-way repeated measures ANOVA was
conducted on the pre- and post-test VAS scores for arousability. The only significant result
was for the crucial Time x Condition interaction, F(1, 46) = 21.9, p <.001. A follow-up paired
t-test on the mean difference scores (see Figure 5.4) showed that the large decrease in
arousability for the EM condition was significantly different to the mean difference of the
control condition, t(46) = 4.7, p < .01, d =.71.

Comparing the mean difference scores (as shown in Figure 5.4), it can be seen that memory
and imagination-based fantasies hardly differ. Indeed, two paired t-tests (with Bonferroni
corrections) showed no significant differences between the two fantasy types for either
condition (both ps >.05).
117
5.4 Discussion
This study is the one of the first attempts to investigate the cognitive processes underlying
sexual fantasies. The DPM-ST posits that sexual fantasies are a form of mental imagery
(either memory or imagination-based) that people deliberately engage in. In other words,
sexual fantasy is a controlled cognitive operation that requires working memory (WM). To
test this, bilateral eye-movements (EMs) were used as a dual-task competing for the resources
of WM. Drawing upon previous research on other forms of mental imagery, it was
hypothesised that the vividness, emotionality and arousability of sexual fantasies (regardless of
whether they are memory or imagination-based) would reduce after EMs are performed while
sexual fantasising, relative to the control condition (i.e., no EMs).
The findings supported all three hypotheses. Specifically, the level of vividness,
emotionality, and arousability were found to decrease for the EM condition. Since EMs tax
working memory, these findings support the DPM-STs assumption that sexual fantasies are a
controlled cognitive operation that require the resources of WM. The results are also a novel
addition to the existing research on how EMs impair mental imagery. First, the fact that EMs
impaired both memory and imagination-based fantasies to the same extent supports previous
studies showing that EMs impair both memories (van den Hout et al., 2001) and
flashforwards (based on imaginative processes) (Engelhard et al., 2010b). Second, sexual
fantasies are predominantly positive, enjoyable experiences (Ellis & Symons., 1990). Thus,
the results support previous findings that EMs can impair positive mental imagery (Hornsveld
et al., 2011; van den Hout et al., 2001).
To explain the present findings in more depth, it is worth referring to previous authors
who have researched in this domain. For example, Gunter and Bodner (2008) postulated that
EMs deplete WM resources causing a reduction in the quality of the mental imagery (i.e.,
lowered vividness). This lowered vividness causes the imagery to be appraised as less
emotional. Put differently, individuals create and appraise low-quality mental imagery. A
118
recent study investigating time-course changes in vividness and emotionality provides support
for this view. Specifically, Smeets, Dijs, Pervan, Engelhard, and van den Hout (2012) found
that vividness dropped after only two seconds of making EMs, whereas emotionality
decreased after 74 seconds. In support of Gunter and Bodner, these findings suggest that
vividness decreases before emotionality reduces.
Drawing upon visual imagery theory (Kosslyn, 1994), Smeets et al. (2012) state that
visual information held in the visuospatial sketchpad of WM needs to be refreshed in order to
maintain the vividness of the mental imagery. They argue that EMs interfere with this
refreshment process leading to lowered imagery vividness and, in turn, lowered emotionality
when the imagery is appraised. In the present study, it can be argued that the EMs interfered
with the refreshment of the sexual visual information held in WM. This led to reduced
vividness and, in turn, a lowered rating of emotionality and arousability.
Interestingly, the findings from the control condition showed that vividness,
emotionality, and arousability levels (for both memory and imagination-based fantasies)
increased after the fantasies were simply envisioned. This supports previous research showing
that sexual fantasies can enhance sexual arousal levels (Gee et al., 2003). Furthermore, these
results can be explained by reversing Smeets et al.s (2012) explanation of how EMs impair
imagery. That is, since the imagery is not impaired, it is able to be refreshed in working
memory. This increases the vividness of the imagery, which is how it is re-encoded into longterm memory. Note that this may explain why some sex offenders report their sexual fantasies
to be more concrete and vivid with each use (Gee, Ward, Belofastov, & Beech, 2006). As a
consequence of the improved imagery quality, sexual fantasies are likely to have been
appraised as more pleasant and more arousing than when first rated. Indeed, Smith and Over
(1987) found that men were more sexually aroused by sexual fantasies that were deemed
vivid, relative to men with non-vivid imagery. If this interpretation of the control condition
119
findings results is correct, it suggests a cognitive explanation for how sexual fantasies can
become more arousing with repeated use.

As this is presently the first study of its kind with regards to sexual fantasy, it is important to
address the limitations. First, the research relied upon self-reported ratings of vividness,
emotionality, and arousability. As with any self-report procedure, there is a risk of socially
desirable responding. This risk may have been even greater in this study as it dealt with
mental imagery of a sensitive (i.e., sexual) nature. Also, since the study follows a basic doseresponse design, there is an increased risk that demand characteristics may have been an
issue, in that, participants may have tried to respond in a manner they thought was expected of
them. Considering that participants were told to engage in a sexual fantasy while concurrently
making EMs and while making no EMs, participants may have guessed the aim of the study
and responded in a manner that confirms the hypotheses.
Gunter and Bodner (2008) found that the vividness of aversive memories reduced after
they were recalled at the same time as making EMs, but not when EMs were performed with
no imagery in mind. They argued that this challenges a demand characteristic account of the
findings as the demand would have also been present in the latter. A more recent EM study
directly addressed this issue by using an objective behavioural (reaction-time) measure to
assess reductions in memory vividness after making concurrent EMs (van den Hout,
Bartelski, & Engelhard, 2012). The results demonstrated that a reduction in the vividness of a
memory during the EM-recall condition (relative to a no-EM condition) can be observed from
objective, non-self report data. These results corresponded with their self-report data. These
results further challenge a demand characteristic account of the effects of EM (at least in
relation to vividness). However, while these findings are promising, they were not employed in
the present study. Thus, before any solid conclusions can be made about the effects of EMs
120
on sexual fantasies, future research should aim to circumvent the issue of demand
characteristics. This can include using a between-subjects design where one group of
participants are told a hypothesis that is the opposite of that actually being tested. If no
increase in the indices occurs, but reductions in the indices are observed in the other groups
who are not told the hypothesis, then demand characteristics could be ruled out. Also, an
objective measure could also be used, as well as a physiological measure that assesses
changes in the autonomic nervous system. For example, reductions in heart rate variability
could provide further evidence that opposes a demand characteristic account of EM effects.
Second, as with most research involving mental imagery, there is no way to ascertain
whether the participants were actually envisioning the sexual fantasies during this study. It is
possible that participants did not envision certain (or any) sexual fantasies during the
conditions. It is also possible that participants tried to envision the sexual fantasies but had
trouble keeping them in mind due to natural mindwandering. This has always been an issue
with mental imagery research and needs to be addressed in the future. For example, one way to
address this issue is to include a behavioural measure that provides some indication of sexual
arousal (e.g., penile plethysmography, heart rate variability) while the sexual fantasy is being
envisioned.
Third, the sexual fantasies that the participants had to envision were chosen by the
researcher on the basis of how frequently they were used by each participant. Thus, it would
be beneficial if a follow-up study used a more rigorous method of selecting fantasies. For
example, participants could complete the Adapted WSFQ before the day of the EM study. This
would allow the fantasy themes to be broken down into separate units (perhaps using a
computer program such as E-Prime) so that a more randomised selection/presentation process
can be implemented. Note also that this procedure would reduce any priming or carry-over
effects that may occur when completing the questionnaire only minutes before envisioning and
rating the sexual fantasies. Alternatively, it may have been beneficial for participants to
121
haven chosen which fantasy they wanted to envision, as this would be more ecologically valid.
It could have made them feel more comfortable with the task. Also, the fantasy they chose may
have been the one they most often engage in outside the lab, meaning that it would (arguably)
be the most vivid, positive, and arousing fantasy in their repertoire of fantasies.
Another limitation is that no measures were made to confirm that WM was actually
being taxed by the EMs. However, as discussed in the Introduction, previous researchers have
found support for this assumption. For example, responses on a reaction time task known to
tax WM are slower when making EMs (Engelhard et al., 2010a; van den Hout et al., 2011).
Also, the effects of EMs on memory impairment are stronger for those with lower WM
capacity (Gunter & Bodner, 2008). Thus, there is sufficient evidence for assuming that WM is
being taxed by the EM procedure in this study. However, to evaluate the scientific integrity of
these results, future research on EMs and sexual fantasies needs to include a measure that
indicates WM is being taxed (e.g., including a measure of WM capacity; engaging in
progressively more taxing tasks and comparing results; or including a third condition that
involves doing EMs at the same as a task known to require WM).
A final limitation is that the sample was comprised of predominantly female
participants. No gender differences were found with regards to the WSFQ subscales.
However, with only 9 males taking part, this comparison must be interpreted cautiously.
Although WM resources are likely to underpin sexual imagery in the same way for both males
and females (as indicated by previous EM research), there may be other important gender
differences at play when investigating sexual fantasies (e.g., in terms of how vividly they
were picturing certain fantasies; their willingness to engage in the task; their subjective
arousal to certain fantasies). Also, the high number of female participants makes the
generalisation of these results to male sex offenders less reliable. Indeed, female sexual
offenders differ from male sex offenders in a number of important ways (Gannon & Rose,
122
2008). Future research should aim to use a sample of male participants only, as well as a
sample of male sex offenders who predominantly use deviant sexual fantasies.
In addition to addressing these limitations, future research should also investigate
whether the changes in vividness, emotionality, and arousability remain at a follow-up, since
previous studies using negative memories have found that the effects can last for at least a
week (van den Hout et al., 2001). It would also be advantageous to investigate whether the
effects of EMs crossover to other sexual fantasies that are different yet share similarities in
their content. Also, in light of the present thesis, steps need to made to further these results by
focusing on deviant sexual fantasies and testing offender populations.
This chapter describes the first investigation into whether WM underpins sexual
fantasising. The results support all of the hypotheses and show a pattern that is similar to the
findings of previous studies using eye-movements to impair mental imagery. Moreover, the
results corroborate the DPM-ST by indicating that sexual fantasies require working memory
resources. Although deviant sexual fantasies were not focused upon, the idea that WM
underlies sexual fantasies applies to all sexual fantasies irrespective of their content and so
can be viewed as being applicable to deviant fantasies. Nevertheless, as stated above, it is
important to test the same hypotheses with a sex offender population, particularly those who
use deviant fantasies. Not only will this corroborate the DPM-ST further, but it may pave the
way for a new treatment strategy for reducing deviant sexual fantasies in sexual offenders (see
Section 8.2.2 of Chapter 8).
123
CHAPTER 6
INVESTIGATING WHETHER SEXUAL FANTASIES
ACTIVATE NON-SEXUAL ASSOCIATIONS
Chapter Rationale
Chapter 5 supported the assumption that sexual fantasising is a controlled process involving
the manipulation of relevant information in working memory. The manipulation of this
information results in mental imagery that can be observed in the mind's eye. Following on
from this, the present study aimed to test whether the information present in sexual mental
imagery (i.e., fantasies) primes other, non-sexual associations held in long-term memory.
Since it is theorised that a sense of power can be acquired by deviantly fantasising, sexual
fantasies about dominance and associations between the concepts of 'self' and 'powerful' were
focused upon.
124
6.1 Introduction
The Dual-Process Model of Sexual Thinking (DPM-ST) proposes that sexual fantasising can
have a number of cognitive effects. One of these proposed effects is the priming of other
associations. In other words, it is argued that sexual fantasising can activate associations held
in long-term memory. This is based on research showing that mental imagery increases the
accessibility of related cognitive associations and representations (Blair, Ma, & Lenton, 2001;
Carroll, 1978). For example, across a number of experiments, Blair et al. (2001) found that
mental imagery about a counter-stereotypic (i.e., strong) woman led to increased 'strongwoman' associations, relative to neutral imagery and no imagery. Thus, as Blair et al. (2001)
state, mental imagery can function as a powerful method of priming. The objective of the
present study was to test this assumption in relation to sexual mental imagery (i.e., sexual
fantasy).
As Ciardha (2011) argues, sexual fantasising is likely to strengthen sex-related
associations (or schemas) via a feedback loop. In addition, it is possible that other associations
become activated or primed by sexually fantasising. For example, sexual fantasies typically
involve the fantasiser being present within the imagery. Thus, it can be argued that certain
self-related associations will also be activated through sexual fantasies. According to previous
studies, the concept of 'self' is often found to be associated with other concepts within memory.
For example, using the Implicit Association Task (IAT), researchers have studied self-esteem
by measuring 'self-positive' associations (Greenwald & Farnham, 2000);
aggression via 'self-aggressive' associations (Uhlmann & Swanson, 2004); suicide ideation via
'self-death' associations (Nock et al., 2010); and emotional congruence with children via
'child-self' associations (McPhail, Sewell, Nunes, & Hermann, 2010).
Of particular note are IAT studies showing that people hold 'self-powerful'
associations. For instance, using an IAT, Nunes, Firestone, and Baldwin (2007) found that
child abusers and non-sex offenders held self-powerful associations to a similar extent. Also
125
using an IAT, Haines and Kray (2005) found women held weaker self-power associations
compared to men. However, they also found that when women were primed with a sense of
power (i.e., told they have superior trading skills on a laboratory-based 'bargaining game'),
their self-power associations strengthened. Based on these findings, Haines and Kray
proposed that activities and situations associated with high power can activate self-power
associations.
The self-power association is of interest because a number of theories suggest that a
sense of power - acquired through deviant sexual fantasy - is an important factor in the
aetiology of sexual offending. For example, Marshall and Marshall (2000) state that, due to
their experiences of rejection and neglect as a child, sex offenders are more likely to
incorporate non-sexual elements, such as power and control, into their sexual fantasies.
Similarly, in their theory on paraphilia and lust murder, Arrigo and Purcell (2001) argue that
sexual fantasies are a major influential factor. They propose that feelings of inadequacy (due
to social and sexual difficulties at a young age) can lead individuals to develop sexual
fantasies about power and domination. There is also some empirical support for these
assertions. For example, Cortoni, Proulx, Paquette, Longpr, and Coutre (2009) found that a
subset of child abusers reported using sexual fantasies to compensate for a low sense of selfworth. In a case study, Carabellese, Maniglio, Greco, and Catanesi (2011) described a serial
rapist who gained a sense of omnipotence and grandiosity from his sexual fantasies of
dominance.
Based on the DPM-ST, it can be argued that the reason individuals gain a sense of
power from their dominance sexual fantasies is because the content of the fantasy activates
self-power associations. Thus, by combining the results of Blair et al.'s (2001) study, which
showed that mental imagery can prime associations, with those of Haines and Kray (2005)
who found that power-related activities/situations can prime self-power associations, the
present study aimed to test whether sexual fantasies about dominance (where the fantasiser is
126
in a position of power) activate self-powerful associations. In addition to this main aim, it was
decided that self-reported powerfulness should also be measured to ascertain whether changes
in powerfulness were confined to participants' automatic self-concept, their explicit selfconcept, or both. Finally, based on existing theories, individuals with low self-esteem and
dominant-related sexual fantasies are likely to be most affected by sexual fantasies about
dominance. Thus, the following hypotheses were proposed:
Hypothesis 1: Compared to a control group, participants who create and envision a sexual
fantasy involving dominance will demonstrate a stronger association between 'self' and
'powerful', as measured by a subsequent IAT.
Hypothesis 2: Compared to the control group, participants who create and envision a sexual
fantasy involving dominance will evidence stronger self-reported feelings of powerfulness.
Hypothesis 3: An increase in self-powerful associations in the sexual fantasy group will be

moderated by self-esteem and the use of dominance-related sexual fantasies.
6.2 Method
6.2.1 Participants
A convenience sample of fifty-one participants (40 females and 11 males) was recruited from
the University of Birmingham via the online Research Participation Scheme (RPS).The mean
age of the sample was 22 years old (SD=3.8, range = 18-36) and it comprised of Caucasian
(92%), Asian (4%), and Mixed Race (4%) participants. Also, 63% of the sample identified
themselves as Single, with 37% stating they were In a relationship. Ethical approval for
this study was granted from the Science, Technology, Engineering and Mathematics Ethical
Review Committee at the University of Birmingham (see Appendix R).
127
6.2.2 Materials
6.2.2.1 Self-powerful Implicit Association Task (IAT)
To measure self-powerful associations, Greenwald et al.s (1998) original seven-stage
Implicit Association Task (IAT) was constructed using E-Prime (see the Introduction in
Chapter 3 for a brief description of the IAT). This particular version of the IAT was chosen
because it has been used in previous studies investigating self-powerful associations (Haines
& Kray, 2005; Nunes et al., 2007).
Table 6.1
Implicit Association Task (IAT) for Powerfulness: Task Sequence
Response Key Assignment
Task
Stage
No. of
Left
Right
trials
Key (Q)
Key (P)
1
20
Target discrimination
Other
Self
2
20
Attribute discrimination
Weak
Powerful
3
4
20
48
Combined task (practice)

Combined task (critical)
Other/Weak
Other/Weak
Self/ Powerful
Self/ Powerful
20
Reversed target
discrimination
Self
Other
20
Reversed combined
Other/Powerful
Self/Weak
task (practice)
7
48
Reversed combined
Other/Powerful
Self/Weak
task (critical)
Note: The order of Stages 2-4 and Stages 5-7 were counterbalanced between participants to
reduce any order effects that may arise from presenting the Self+Powerful combination first.
The self-powerful IAT involved assigning randomly presented words (via a left or right
keyboard response) into one of two categories. As Table 6.1 shows, Block 1 involved
categorising words into an appropriate target concept (i.e., 'Other' or 'Self'). Block 2 involved
categorising words into an appropriate attribute category (i.e., 'Weak' or 'Powerful'). Blocks 3
and 4 involved combining target and attribute concepts so as to form two opposing paired
categories (i.e., 'Weak/Other' versus 'Powerful/Self'). Thus, following the examples shown in
Table 6.1, participants would press the right key (i.e., P) if either a powerful or self-related
128
word was presented. Block 5 involved categorising words into the same attribute concepts as
Block 2. However, the position of the categories was now the opposite of those in Block 2
(i.e., 'Powerful' or 'Weak'). In Blocks 6 and 7, these reversed attribute concepts were
combined with the target concepts to form a reversed combination task. Thus, continuing the
example used above, participants would now press the right key when responding to either a
weak or self-related word.
Five 'Self' and five 'Other' words were used for the present IAT. These stimuli were
derived from previous IAT studies that focused on these two concepts (Greenwald &
Farnham, 2000). The words for the attribute concepts were derived from a small pilot study
involving individuals from the community that did not take part in the main study (n = 26).
That is, six words most strongly related to 'Powerful' and six words most strongly related to
'Weak' were used (see Appendix S).
6.2.2.2 Self-reported measure of powerfulness

Two 'feeling thermometers' were created to measure self-reported powerfulness. The first
required participants to rate themselves on a scale from 1 ("Not powerful at all") to 8 ("Very
powerful"). The second was the exact same scale but in relation to "Other people" (see
Appendix T).
6.2.2.3 Short Self-esteem Scale (SSES; Thornton, 1989)

The SSES is a 12 -item questionnaire (four of which are distracter items) designed to measure
self-esteem (see Appendix U). It was initially used for both research and clinical purposes
within HM Prison Service. It was later incorporated into the Sex Offender Treatment
Programmes operated by HM Prison Service. For each item, participants must respond with
either a 'Yes' or a 'No'. A response of 'Yes' is scored as 2 and a response of 'No' is scored as 0.
A score of 1 is given for a missing response. The scale has been used with non-offender
129
samples which, as expected, showed higher scores than sex offenders (Webster, Mann,
Thornton, & Wakeling, 2007). Moreover, the scale has been shown to have a Cronbach alpha
coefficient of .84 (Webster et al., 2007), thus, demonstrating an excellent internal consistency
(Kline, 2000).
6.2.2.4 Wilson Sexual Fantasy Questionnaire (WSFQ; Wilson, 1978)

The WSFQ is a 40-item questionnaire designed to measures how often people use 40 specific
sexual fantasies (see Appendix E). It is scored on a Likert-type scale from 0-5 and has been
used with both non-offenders and sex offenders (Baumgartner, Scalora, & Huss, 2002).
Moreover, Baumgartner et al. (2002) found that the scale had very good internal consistency
(Cronbachs = .95). For the purposes of the present study, only dominance-related sexual
fantasies were of interest. Thus, the following items were focused upon: "Tying someone up",
"Forcing someone to do something", and "Hurting a partner".
6.2.3 Procedure
Upon entering the lab, each participant was reminded of the sexual/sensitive nature of the
study. They were informed that the study aimed to investigate the non-sexual effects of sexual
mental imagery. After each participant provided their consent to take part, they were assigned
to either the Fantasy Condition or the Control Condition. Participants assigned to the Fantasy
Condition were asked to create and imagine a consensual sexual scenario where they are in
total control (see Appendix V). They were told to imagine, from a first-person perspective (i.e.,
through their own eyes), that they were engaging in a number of controlling/dominant
behaviours. This could include tying up their partner; whipping and/or spanking them;
holding them down, and so on. A first-person perspective was chosen as research shows that
it has a stronger emotional effect than an observer-oriented perspective (Holmes, Coughtrey,
& Connor, 2008).
130
Participants in the Control Condition were simply told to create and imagine (from a
first-person perspective) doing their grocery shopping (see Appendix V). This particular
scenario was chosen for the Control Condition because it was regarded as a neutral event that
participants could easily create and envisage within their mind's eye. Indeed, previous
research indicates that "doing some shopping" is representative of a neutral event (D'
Argembeau, Comblain, & van der Linden, 2003, p.287).
In addition to these instructions, participants in both conditions were told to imagine
the scenario for a total of three minutes, which was measured by the researcher using a
stopwatch. After this time, all participants were required to describe the scenario in writing.
This allowed the researcher to gain some insight into how well the participants engaged in the
task. It also provided an approximate indication of how dominant each participant was being in
the scenario (Fantasy Condition only). Finally, participants had to indicate how vivid the
imagery was using a scale ranging from 1 (Not clear at all) to 7 (Extremely vivid). This was
done to provide further indication of how well the scenarios were being envisioned.
Following the mental imagery task, participants were required to complete the selfpowerful IAT. For each group, half of the participants completed the IAT that presented the
'Self+Powerful' block first, while the other half completed the IAT with the 'Self+Weak' block
presented first. For all participants, full instructions for how to perform the IAT were
presented on-screen via E-Prime. The instructions were also repeated verbally by the
researcher. Participants would only begin the IAT after indicating that they understood the
instructions.
Following the IAT, all participants completed the various self-reported measures in a
random order. These included the two feeling thermometers (i.e., self-reported powerfulness),
the SSES, and the WSFQ. The self-report measures were administered after the IAT so as to
reduce any priming effects they may elicit. It is also worth noting that, according to previous
research, completing an IAT before self-report measures does not induce any biasing effects
131
in the subsequent self-reports (Nosek, Greenwald, & Banaji, 2005). After the study had
ended, participants were debriefed and thanked for their time.

Before any meaningful statistical analyses could be performed on the IAT data, the raw data
needed to be scored according to the algorithm recommended by Greenwald, Nosek, and
Banaji (2003). This scoring algorithm has been used in most IAT studies since its publication
and only involves the combined blocks (i.e., 3, 4, 6, and 7). There are a number of steps
associated with this scoring method. First, RTs above 10,000ms are removed. This is because
anything above this cut-off is likely to be due to inattention or distraction (Greenwald et al.,
2003). Also, participants with more than 10% of RTs below 300ms are removed, as it
indicates they are engaging in anticipated responding (Greenwald et al., 1998). However,
none of the participants in this study had more than 10% of scores below 300ms.
Next, a single pooled standard deviation (SD) is calculated for all RTs in Blocks 3 and
6, and another for all RTs in Blocks 4 and 7. Then, the mean RT for correct trials is calculated
for each of the four blocks. Any errors are replaced with the newly calculated mean RT plus a
penalty of 600ms (Greenwald et al., 2003). Following this, all RTs in each block are averaged
producing four means (i.e., for Blocks 3, 4, 6, and7). The mean difference between Blocks 6
and 3 is then computed, as is the mean difference between Blocks 7 and 4. The resulting mean
difference scores are then divided by their associated pooled SD. Finally, these two quotients
are averaged. The resulting score is commonly referred to as the D-score and is shown to be
less vulnerable to extraneous influences such as cognitive ability and IQ (Greenwald et al.,
2003). In this study, D-scores above zero indicated stronger self-powerful associations,
whereas those below zero indicated stronger self-weak associations.
132
6.3 Results
6.3.1 Hypothesis 1: Compared to a control group, participants who create and envision a
sexual fantasy involving dominance will demonstrate a stronger association between
'self' and 'powerful'.
As shown in Figure 6.1, the mean response latencies (using treated data) depict a trend in this
predicted direction. In other words, participants in the Fantasy Condition appear to categorise
words into the 'Powerful + Self' category faster (M = 831ms, SD = 224.8) than those in the
Control Condition (M = 978ms, SD = 298.4). Furthermore, results from a t-test show this
difference to be marginally significant; t(49) = 2.0, p = .051. The error rates indicate that
participants in the Fantasy condition made more incorrect responses on the Powerful-Self
block (see Table 6.2). This is at odds with Hypothesis 1 as fewer mistakes would be expected
in this block for this condition. However, t-test analyses on the arcsine-transformed error rates
found no significant differences between the conditions for both IAT stages.
Table 6.2: Mean error rates (%) and SDs for each block of the IAT for each condition
Powerful-Self
Powerful-Other
Fantasy condition
8.6 (4.5)
9.2 (4.9)
Control condition
6.9 (5.9)
10.2 (8.3)
133
Figure 6.1: A graph showing mean response latencies (treated) and SEs for
each association, across both groups
It should be noted that these decomposed results are not indicative of an absolute association
between 'powerful and 'self' because the IAT operates on a relative basis. Thus, analytically
separating the data into a single 'self-powerful' association does not provide conclusive
results. This is because the isolated association reflects some aspect of the relative comparison
between 'Powerful + Self' and 'Weak + Self'. As a result, D-scores (which embody the relative
nature of the IAT's procedure) were focused upon in the main analysis.
When looking at the D-scores, the same pattern of results shown in Figure 6.1 was still
observed. For example, while both D-score were positive (indicating that both groups hold
stronger self-powerful associations than weak-self), participants in Fantasy Condition had
slightly greater D-scores (M = .22, SD = .42) than those in the Control Condition (M = .15,
SD = .52). However, the results of a two (Group: 'Fantasy' versus 'Control') two (Block Order:
'Powerful/Self' versus 'Weak/Self') ANOVA were less supportive. Specifically, there was no
134
significant main effect for Group. In other words, the observed difference in D-scores
between the two groups was not significantly different.
A significant main effect was found for Block Order; F(1, 47) = 18.23, p < .001.
Follow-up analyses indicated that those who completed the 'Powerful + Self' block first had a
higher D-score (M = .39, SD = .39) than those who completed 'Weak + Self' block first (M =
-.04, SD = .43). This difference was highly significant, t(49) = 3.8, p <.001. However, this
result was expected as it is a well-documented extraneous effect associated with the IAT
procedure (Nosek et al., 2005). Moreover, by counterbalancing the block order in the present
study, these effects were likely to been cancelled out or reduced. Finally, there was no
significant interaction between Group and Block Order. It should also be noted that the
vividness levels for the mentally imagined scenarios was greater in the Control Condition (M
= 5.3, SD = 1.5) than in the Fantasy Condition (M = 4.4; SD = 1.2). Moreover, this difference
was found to be statistically significant, t(49) = 2.4, p < .05.
In summary, while the results are in the predicted direction, the results of the statistical
analyses did not support Hypothesis 1.
6.3.2 Hypothesis 2: Compared to the control group, participants who create and
envision a sexual fantasy involving dominance will evidence stronger self-reported
feelings of powerfulness.
To test this hypothesis, the ratings of the self-powerful feeling thermometer were subtracted
from the ratings of the other people-powerful feeling thermometer. This produced a mean
difference that is comparable to the scores of the IAT (Uhlmann & Swanson, 2004). In other
words, a positive score would suggest an explicit attitude that oneself is more powerful than
other people.
Contrary to the hypothesis, the results indicated that participants in the Fantasy
Condition saw themselves as less powerful than other people (Mdifference = - 0.7, SD = 1.87).
Participants in the Control Condition also rated themselves as less powerful than other people
135
(Mdifference = - 0.52, SD = 1.63). Results from a paired t-test indicated that there was no
significant difference between the two groups. In summary, these findings do not provide
support for Hypothesis 2.
6.3.3 Hypothesis 3: An increase in self-powerful associations in the sexual fantasy group

will be moderated by self-esteem and the use of dominance-related sexual fantasies.
Examining the WSFQ scores in Figure 6.2, it can be seen that the pattern of results are similar
to those found in Chapter 5.
Figure 6.2: Mean scores (and SEs) for each WSFQ subscale across both
groups
Specifically, scores on the Intimate subscale were the highest for both the Fantasy group (M =
27, SD = 7.1) and the Control group (M = 23.2, SD = 10.5). The next most frequent fantasies
were Exploratory, with the Fantasy group showing a mean of 6.7 (SD =4.8) and the Control
group showing a mean of 8.5 (SD = 8.0). For both groups, the least used fantasies were
Sadomasochistic (Fantasy group: M = 4.7, SD = 4.3; Control group: M = 5.8, SD = 7.7) and
136
Impersonal (Fantasy group: M = 4.2, SD = 2.8; Control group: M = 6.1, SD = 6.4). From
these results, it can be seen that the Control group reported less Intimate fantasies than the
Fantasy group, but more Exploratory, Sadomasochistic, and Impersonal fantasies. However,
in spite these trends, t-tests analyses revealed there were no significant differences between
the two groups on any of the subscales (all ps >.05).
Since Hypothesis 1 was not supported, Hypothesis 3 was unable to be directly tested.
Exploratory analyses of the self-report data showed no group differences on the SSES or any
of the three dominanc-related sexual fantasies. Furthermore, for both groups, D-scores did not
significantly correlate with any of the self-report measures of interest. However, in the
Control group, the D-score was found to positively correlate with the items Intercourse with
someone you know but have not had sex with (r = .75, p <. 001), Being seduced as an
innocent (r = .49, p < .05), and Having sex with someone of a different race (r = .46, p <
.05). In addition, when the two groups were collapsed together, D-scores were found to
correlate with the sexual fantasy theme labelled "Forcing someone to do something" (r = .29,
p < .05).
To explore the data further, the two main groups were split by gender. In the Fantasy
Condition, the D-scores of male participants were found to be significantly (and positively)
associated with the sexual fantasy themes of "Forcing someone to do something" (r = .89) and
"Hurting a partner" (r = .89). These findings were not found in males in the Control
Condition or in females in either condition. However, these results should be interpreted with
caution given their ad-hoc nature and the fact that only 6 men were present in the Fantasy
Condition. At most, they provide a possible direction for further research (i.e., comparing the
self-powerful associations of males who report already using dominance-related sexual
fantasies with those who do not).
137
6.4 Discussion
The main aim of the current study was to test whether sexual fantasies involving dominance
activated and strengthened self-powerful associations, relative to neutral imagery. The results
indicated that, although the means of the raw RTs were in the predicted direction (i.e.,
participants who imagined a dominance-related sexual fantasy tended to assign stimuli into a
'Powerful + Self' category faster than they assorted stimuli in a 'Weak + Self' category), the
critical D-scores derived from the IAT did not differ between the two groups. Thus, the main
hypothesis of the study was not supported.
It was also hypothesised that explicit powerfulness would increase after sexually
fantasising about dominance, as indicated by higher self-reported powerfulness in the Fantasy
Condition. However, both groups were found to report feeling less powerful than other
people. Moreover, the two groups did not significantly differ on this measure. Thus,
Hypothesis 2 was not supported. Finally, it was hypothesised that an increase in self-powerful
associations would be moderated by self-esteem and the use of dominant-related sexual
fantasies. However, given that the experimental manipulation appeared to have no effect on
participants D-scores, this hypothesis could not be tested. When comparing groups on their
level of self-esteem and use of dominant-related sexual fantasies, no significant differences
were found.
However, it was found that, for the Control group, powerful-self associations were
positively associated with fantasies about having sex with someone you know but have not
had sex with, being seduced as an innocent, and having sex with someone of a different
race. Since this occurred after imagining a neutral scenario and not after imagining a
dominant fantasy, one interpretation may be that engaging in the latter has an inhibitory effect
on reporting sexual fantasies. Also, it was found that, when the two groups were collapsed
into one, a positive relationship emerged between the D-scores and the item of the WSFQ
related to forceful sex. This may indicate that, regardless of their group assignment,
138
participants who had previously used force-related fantasies were more likely to associate the
concepts of powerful and self. On the other hand, the results could simply reflect a priming
effect of the self-powerful IAT. However, as mentioned earlier (Section 6.2.3), Nosek et al.
(2005) report that performing the IAT before self-report measures does not have an
extraneous effect on subsequent self-report measures. There was also some tentative
indication that males who reported using sexual fantasies about forcing and hurting someone
were more likely to have a higher D-score on the self-powerful IAT. However, these results
are open to criticism due to the fact that they are based on two combined groups, each of
which was subjected to different experimental manipulations.
At face value, this series of findings suggests that sexual fantasising does not activate
and/or strengthen associations related to the content of the fantasy. To be more specific,
dominance-related sexual fantasies do not appear to prime self-powerful associations,
regardless of whether they are indirectly or directly assessed. However, to the author's
knowledge, this is the first study of this kind with respect to sexual fantasy. Thus, before any
solid conclusions are made, it is worth considering other reasons for the null effects. Note that
these reasons also highlight the various limitations associated with the study.

First, perhaps participants should have been told exactly what to fantasise about (i.e., via
guided imagery), rather than being allowed to create a dominant sexual fantasy themselves.
For example, on inspecting the written descriptions, participants appeared to create
predominantly playful and romantic fantasies. Thus, while most participants engaged in
dominant and controlling behaviours in their fantasy, they did so in the context of playfulness
and intimacy. Perhaps this was not adequate for priming self-powerful associations. This also
highlights the issue of having participants envision fantasies that they themselves are not likely
to hold naturally. In hindsight, it may have been more beneficial to have only used
139
participants who report actively using dominant fantasies and examining whether their selfpowerful associations are increased after using such fantasies.
Second, it has been found that people's cognitive load can be increased by performing
an irrelevant task (Gilbert & Hixon, 1991). Thus, as noted by Blair et al. (2001), it is possible
that creating and imagining neutral mental imagery increased the IAT results from baseline
levels, thereby, obscuring the actual difference between the two conditions. To rule-out of this
possibility, it would have beneficial to have included and compared a 'no imagery' condition.
Third, Haines and Kray (2005) found that the strength of women's self-powerful
associations (as measured by an IAT) can change in response to the situational context. Thus, it
is possible that the participants in the present study- the majority of which were female - felt a
lack of power in the context of the study (i.e., having to imagine a sexual fantasy that they may
not normally use). Put differently, the context of the experiment would not have primed selfpowerful associations. In fact, the context may have primed weak-self associations. It would
have been useful to have tested the participants on both the self-powerful IAT and selfreported measures before the imagery task, as well as after. This would have provided pretask (or baseline) scores that the post-task scores could have been compared to.
A fourth point is that the design of the IAT could have been at fault. For example, in
Haines and Kray's (2005) study, the actual category labels for the attribute concept were
'Dominant' and 'Subordinate'. Perhaps sexual fantasies about dominance increase the strength
of the 'dominant-self' associations. This is a subtle point but there may be an important
conceptual difference between the two associations.
Finally, the sample in the present study was comprised mostly of female participants.
Research shows that women are more likely to report sexual fantasies about submission (Knafo
& Jaffe, 1984; Pelletier & Herold, 1988; Zurbriggen & Yost, 2004), which may explain why
the vividness of the mental imagery was lower for those in the Fantasy Condition compared to
those in the Control Condition. Moreover, it has been argued that fantasies of
140
submission are more about having the power to overwhelm a man with lust, rather than about
being overpowered by a man (Hariton, 1973). In other words, submission-related sexual
fantasies are actually about female power, suggesting that submission fantasies may have been
better for priming self-powerful associations in female participants. This may explain why
tentative correlations between D-scores and fantasies of forcing and hurting someone were
found only for male participants in the Fantasy Condition. Perhaps successful results would
have been acquired in female participants if they had been asked to create and imagine sexual
fantasies where they have the power to seduce a man.
Given these post-hoc observations, this study may have generated successful results
had just males been used. Indeed, the results of the structural equation modelling study in
Chapter 4 would suggest that men who harbour distorted attitudes about women may be a
population to target. Moreover, as mentioned in Chapter 5, results from a study using solely
male participants would allow for more appropriate generalisations with regards to male sex
offenders.
Thus, before any conclusions about whether sexual fantasies prime other associations,
it is clear that more research is needed to address the various issues and limitations associated
with this study. Therefore, at this stage, the priming aspect of the DPM-ST is not supported. It
should also be noted that, from these results, no insights can be made as to how sex offenders
acquire a sense of power and mastery from their sexual fantasies (Marshall & Barbaree,
1990). To test this, a study similar to this one needs to be conducted with a sample of sexual
offenders who report using deviant sexual fantasies.
141
CHAPTER 7
INVESTIGATING WHETHER SEXUAL ABUSERS
ASSOCIATE SEXUAL FANTASIES WITH CHILDREN
Chapter Rationale
The aim of this chapter was to assess another proposed cognitive effect of sexual fantasising;
namely, that an association will form between the concept of 'sexual fantasy' and the theme
within the fantasy. The study in this chapter tested this assumption using an indirect measure
the Go/No-Go Association Task with a sample of child abusers. It was hypothesised that
extrafamilial child abusers (who abuse children outside of their family and are known to
regularly use sexual fantasies about children; Marshall, Barbaree, & Eccles, 1991) will
demonstrate a stronger association between 'children' and 'sexual fantasy' than nonextrafamilial child abusers and non-offenders.
142
7.1 Introduction
Following on from Chapter 6, another cognitive effect of repeatedly using a deviant sexual
fantasy, proposed by the Dual-Process Model of Sexual Thinking (DPM-ST), is that
information within a deviant sexual fantasy will become encoded in long-term memory. One
manner in which the information can become encoded is in the form of a cognitive script.
This is based on previous assertions that the repeated use of aggressive fantasies can lead to
the encoding of an aggressive script (Huesmann, 1988) and that the repeated use of deviant
sexual fantasies can lead to the generation of an implicit offence script (Gee, Ward, &
Eccleston, 2003; Ward & Hudson, 2000). These scripts are thought to be more readily
accessible for retrieval, thereby, increasing the likelihood that they will be used again
(Huesmann, 1998).
In addition to this, the DPM-ST proposes that a simpler form of encoding occurs from
repeated deviant fantasising. That is, the concept of sexual fantasy will become associated
with the concept it depicts. Put differently, if an individual repeatedly fantasises about a
certain person, object, or behaviour, the concept of sexual fantasy is likely to form an
association with that same person, object, or behaviour in memory.
Given this reasoning, it can be hypothesised that child abusers who have a history of
frequently using child-related sexual fantasies are likely to hold a strong association between
the concepts of sexual fantasy and children, which may be detectable using an indirect
measure. It can be argued that extrafamilial child abusers are more likely to evidence a history
of deviant sexual fantasising and, therefore, a child-sexual fantasy association. This is because
research shows that extrafamilial child molesters tend to show higher levels of sexual deviance,
as measured phallometrically (Firestone, Bradford, Greenberg, & Serran, 2000; Quinsey,
Chapline, & Carrigan, 1979) and psychometrically (Beech, Friendship, Erikson, & Hanson,
2002). Also, Pawlak, Boulet, and Bradford (1991 found that, relative to incestuous abusers,
extrafamilial abusers endorsed more Sexual Fantasy items on the Derogatis Sexual
143
Functioning Inventory (Derogatis, 1976). Moreover, Marshall, Barbaree, and Eccles (1991)
found that extrafamilial molesters used more deviant fantasies (after their first offence) than
did incest offenders. Langevin et al. (1998) also found that heterosexual and homosexual
paedophiles4 reported more child fantasies than incest offenders. Finally, extrafamilial
molesters have been found to form deviant fantasies earlier in life than do intrafamilial
molesters who use deviant fantasies (Cortoni et al., 2009; Marshall et al., 1991).
To indirectly measure child-sexual fantasy associations, a Go/No-Go Association Task
(GNAT; Nosek & Banaji, 2001) was used. This is because the concept of sexual fantasy - as a
form of mental imagery - has no clear opposite (i.e., as male versus female; child versus
adult). Most of the other measures, such as the Implicit Association Task (IAT), require the use
of an opposite concept whereas the GNAT can be (and has been) used to measure associations
towards a single concept. For example, Teachman (2007) previously used the GNAT in this
manner to assess the strength of a fear-spider association relative to a calm- spider association
(i.e., 'spider' was the single category). It should be noted that the GNAT has previously been
used to assess child-sex associations in child abusers with some success (Snowden, Gray,
Brown, & Power, 2007). However, opposing categories were used (Adult versus Child; Sex
versus Not Sex).
The main aim of the present study was to investigate whether child-related sexual
fantasies in exclusively extrafamilial abusers could be indirectly assessed using the GNAT.
There were three main hypotheses:
Hypothesis 1: Extrafamilial abusers will report using more child-related sexual fantasies than
non-extrafamilial abusers and non-offenders.
However , Langevin et al. (1998) did not explicitly state that these offenders were extrafamilial
144
Hypothesis 2: Due to engaging in elaborate deviant sexual fantasies more frequently,

exclusive extrafamilial child abusers will demonstrate stronger associations between 'child'
and 'sexual fantasy on the GNAT than non-extrafamilial abusers and non-offenders.
Hypothesis 3: On a control GNAT, there will be no group differences. Support for this
hypothesis will help rule out the possibility that task-specific confounding variables account
for the results on the experimental GNAT.
Hypothesis 4: The scores on the experimental GNAT will predict deviant sexual fantasisers
regardless of subtype.
7.2 Method
7.2.1 Participants
7.2.1.1 Offender sample
As in Chapter 3, offender participants were patients held under Wisconsins Sexually Violent
Persons law at a secure treatment facility in the State of Wisconsin (USA). All patients in each
wing were informed by the researcher that a study measuring the speed in which people
respond to pictures of people and words related and unrelated to sex was taking place. The
patients were informed that they could voluntarily take part and to indicate their interest by
writing their name on a sign-up sheet, which was left in each wing.
A week later, the sign-up sheets were collected. Patients with a current or previous
sexual offence against a child (as indicated in their case file) were eligible to take part in the
study. Staff members informed the researcher on which of the signed-up patients met this
criterion. Information from the case files also led to some patients being excluded. For
example, due to the fact that the indirect task used in the study presented pictures of only prepubescent children (see Method), child abusers who offended against post-pubescent children
145
(13 years and above) were excluded so that clearer results were obtained. Also, as in Chapter
3, patients with very low cognitive functioning and taking antipsychotic medication affecting
their motor skills were excluded.
Due to these criteria and the limited time-frame, as many patients were tested as was
possible. Thus, the final sample was based on convenience and consisted of 48 child abusers.
Using information from case files, 29 participants were identified as being exclusively
extrafamilial, whereas 19 were identified as non-extrafamilial offenders (i.e., 6 intrafamilial
child molesters, 5 predominantly intrafamilial child molesters, and 8 individuals who
assaulted children and adults). All of the participants had a diagnosis of paedophilia except for
four extrafamilial offenders and three non-extrafamilial offenders. Table 7.1 displays the
demographic information for all three groups in the study. As shown, the mean age of the
extrafamilial offenders only differed by two years. It can also be seen that most of the
offender participants were Caucasian, although the groups were also comprised of ethnicities,
including Black, Asian, and Native American participants.
Table 7.1: Demographic information for the for extrafamilial abusers, nonextrafamilial abusers, and non-offenders.
Extrafamilial
abusers
abusers
controls
(n = 29)
(n = 19)
(n = 26)
Mean age:
47 (SD=10.6)
45 (SD=9.9)
20 (SD=1.9)
Ethnicity:
Caucasian
Black
N. American
Asian
Other
86.2%
3.4%
3.4%
3.4%
0%
68.4%
21.1%
10.5%
0%
0%
80.8%
0%
0%
7.7%
11.5%
Relationship
status:
In a relationship
Single
Divorced/separated
Widowed
3.4%
69%
24.1%
3.4%
0%
73.7%
21.1%
5.3%
34.6%
65.4%
0%
0%
146
Table 7.2 shows the mean scores for a number of clinical factors for each offender group;
namely, static risk (measured using the Static 99R, Helmus, Thornton, Hanson, & Babchishin,
2012), dynamic risk (measured using the Structured Risk Assessment - Forensic Version,
Thornton, 2002), psychopathy (measured using the Psychopathy Checklist-Revised; Hare,
1991), and IQ (measured using the Wechsler Adult Intelligence Scale, 3rd and 4th Editions;
Wechsler, 1997, 2008). This information was derived from the participants' case files and was
based on the most recently recorded score/assessment. As shown in the Table 7.2, both groups
showed similar scores on all four variables (all ps > .05).
Table 7.2: Mean scores (SDs) for static and dynamic risk,
psychopathy, and IQ in both offender groups
Extrafamilial
Non-extrafamilial
abusers
abusers
(n = 29)
(n = 19)
5.6 (1.2)
5.5 (1.93)
Static risk:
3.8 (.57)
3.5 (.69)
Dynamic risk:
Psychopathy:
24.2 (4.9)
26.6 (4.03)
IQ:
93.4 (14.9)
94.8 (12.6)
At the time of the study, some of the participants were undergoing cognitive-behavioural
treatment. As shown in Table 7.3, most of the extrafamilial abusers were in treatment and a
high proportion of non-extrafamilial abusers were in treatment. A comparison of the groups
showed that significantly more extrafamilial abusers were in-treatment relative to nonextrafamilial abusers, 2 (1) = 4.89, p <.05. It is also worth noting that, in both groups, a
higher proportion of abusers that were in-treatment had progressed to Phase 2.
Table 7.3: The proportion of participants in treatment and
pre-treatment for both groups
Extrafamilial
Non-extrafamilial
abusers
abusers
(n = 29)
(n = 19)
10.3%
36.8%
Pre-treatment
In-treatment:
Phase 1
Phase 2
89.7%
34.5%
55.2%
63.2%
15.8%
47.4%
147
Finally, Table 7.4 shows the offence information for the two offender groups. As indicated,
the extrafamilial offenders committed their first offence only slightly earlier than nonextrafamilial offenders, t(45) = .91, p < .05. It can also be seen that extrafamilial abusers
appear to abuse male children at a higher rate than non-extrafamilial abusers, although it was
not significantly different. Extrafamilial abusers also abused both genders and at a higher
frequency than non-extrafamilial abusers, 2 (1) = 6.3, p <.05. Conversely, the nonextrafamilial offenders abused female children at a much higher than extrafamilial abusers, 2
(1) = 13.89, p <.001. Finally, it can be seen that a similar proportion of offenders in each
group had prior convictions for a sexual offence. While the non-extrafamilial abusers appeared
to have more prior convictions for violent and nonsexual, nonviolent offences, the difference
was not significant (both ps > .05).
Table 7.4: Offence history information for both offender groups

Extrafamilial
Non-extrafamilial
abusers
abusers
(n = 29)
(n = 19)
Mean age at
first sexual
18 (SD= 5.7)
20 (SD = 6.9)
offence
Victim gender:
- Male
- Female
- Both
34.5%
20.7%
44.8%
15.8%
63.2%
21.1%
Prior sexual
offences
93.1%
94.7%
Prior violent
offences
20.7%
44.4%
Prior nonsexual
offences
44.8%
68.4%
7.2.1.2 Non-offender sample

A comparison sample consisting of 26 non-offending students was recruited from the
University of Birmingham in the United Kingdom. The mean age of the comparison sample
148
was 20 years old (SD = 1.87, range = 18-25). As shown in Table 7.1 above, most of the nonoffenders were Caucasian, with a small proportion of sample having a Black and Other ethnic
background. Most of the non-offenders were also single at the time of the study.
Ethical approval to conduct this research with offenders was granted
7.2.2 Materials
7.2.2.1 Experimental GNAT
To indirectly measure child-sexual fantasy associations, the Go/No-Go Association Task
(GNAT; Nosek & Banaji, 2001) was used. The GNAT involves categorising stimuli presented
on a computer screen as belonging to a particular concept (e.g., a child, an adult, or a sexual
fantasy) as quickly and as accurately as possible. However, only stimuli related to the two
concepts displayed on screen are responded to (the Go trials). Stimuli that are unrelated to
the concepts are not responded to (the No-Go trials). In this study, a GNAT consisting of
two blocks was used. In one block, the concept of Sexual Fantasy was paired with 'Child'. In
the other block, Sexual Fantasy was paired with the opposing category of Adult. The
concept labels were displayed in the top two corners of the screen with each stimulus being
presented in the centre in a randomised order (see Appendix W).
Participants were required to respond to the stimuli from one category pairing while
ignoring unrelated stimuli. For example, when Child and Sexual Fantasy were paired and
displayed on screen, participants had to respond (by pressing the spacebar) to a total of 15
images of children and 15 words related to sexual fantasy. Based upon prior GNAT studies
149
(Teachman, 2007), each block consisted of 16 practice trials and 60 critical trials. Stimuli for
Go trials remained on screen for 1400ms, while No-Go trials remained on screen for
1000ms (Teachman, 2007). If participants made a mistake or did not respond in time, error
feedback was given.
Stimuli for the concepts of Child and Adult were computer modified images of 15
adults and 15 children. All of the children images were taken from the commercially available
Not Real People (NRP) set devised by Laws and Gress (2004) for use in assessing and
researching sexual offenders (see Appendix X for examples). A set of computer modified
adult images developed by Ciardha (2010) were also used as there were not enough adult
images in the NRP set (see Appendix X for examples). The words used for the concept of
Sexual Fantasy were derived from commonly used sexual fantasy questionnaires, such as the
Wilson Sexual Fantasy Questionnaire (Wilson, 1978). Fifteen words judged as being most
strongly related to sexual fantasy by a sample of students (n = 27) were used for the final
stimuli set (see Appendix X). Finally, 15 distracter words (i.e., words unrelated to Sexual
Fantasy) were chosen in order to form half of the No-Go' trials (see Appendix X). Each
distracter word was matched with a Sexual Fantasy word on the initial letter, character length,
and syllable length (e.g., Orgy versus Oven).
7.2.2.2 Control GNAT

In order to corroborate that the GNAT has the capability to capture child-related associations
more generally, a control GNAT was also designed. This was similar to the Sexual FantasyGNAT except the target concept was 'Toy. Fifteen words describing a toy (e.g., yoyo,
Barbie), which were judged by the same student sample as being most strongly related to the
concept of Toy, were used (see Appendix X). There were also 15 different distracter words,
matched to the Toy words on initial letter, syllable length, and character length (see Appendix
150
X). Children and adult-related stimuli were taken from the same two sets of computer
modified images, although they were different to those in the experimental GNAT.
7.2.2.3 Wilson Sexual Fantasy Questionnaire (WSFQ; Wilson, 1978).

The WSFQ is a 40-item tool designed to measure the frequency of fantasy use (see Appendix
E). See Chapter 3 (Section 3.2.2.3) for a more detailed description. As in Chapter 3, only two
items that have reference to children were used (i.e., Having sex with someone much younger
than yourself and Seducing an innocent).
7.2.2.4 Thoughts and Fantasies Questionnaire (Thornton, unpublished).

The Thoughts and Fantasies Questionnaire is an open-ended self-report measure designed to
establish how often and for how long an individual uses six particular deviant fantasies
(Appendix F). See Chapter 3 (Section 3.2.2.4) for a more detailed description. Only the
section related to children under 13 years old was used in this study.
7.2.3 Procedure
All offender participants completed the study in a quiet meeting room in one of the facilitys
wings. Each of the non-offender participants completed the study in a lab within the
University of Birmingham. All participants were informed - both verbally and in writing
that the study required them to complete a task designed to investigate the speed in which
people respond to certain pictures and words. As with Chapter 3, all participants were made
aware that some words would be of a sexual nature. However, to help reduce the likelihood of
socially desirable responding, participants were only told about the full purpose of the study
after they had participated. In addition, participants were made aware that there would be a few
questionnaires to complete after the computer task. The offender participants were also asked
to give consent for the researcher to access information within their case files and
151
psychometric reports. All participants were informed that they had to right to withdraw at any
point in the study without consequence.
After providing informed consent, each participant completed the GNATs and then the
questionnaires. With regards to the GNATs, the control GNAT was always administered first
so that the GNAT procedure was familiar when completing the experimental GNAT. Both
GNAT tasks were run using the experiment software package, E-Prime. For both tasks,
identical instructions were presented to all participants on the computer screen before they
started. The instructions stated that a picture of an adult or child would be presented as would
words related and unrelated to sexual fantasies. For the first block, the instructions stated that
they had to respond to whichever labels were present on the screen. Sexual fantasy would
always be present but in one block child was a label and in the other adult was a label.
Thus, for the block with child as the label, participants were told to respond only to pictures
of children by pressing the Spacebar. If they made an incorrect response or were too slow to
respond, they would be presented with error feedback on the screen. The instructions also told
the participants to respond as quickly and as accurately as possible, and that mistakes were to
be expected.
All participants read these instructions before completing the practice phase. Once it
was deemed that the participant was comfortable with what they had to do, the participant
completed the critical phase. The order of the two blocks was counterbalanced so that half of
the participants completed the Child + Sexual Fantasy block first and other half completed
the Adult + Sexual Fantasy block first. This was done to reduce any possible order effects.

Prior to conducting any statistical analyses with the GNAT data, the raw reaction times (see
Table 7.5) were converted into a standardised form. As discussed in Chapter 3 (with the SPF)
and Chapter 6 (with the IAT), standardising reaction-time data helps reduce the effects of
152
unwanted variance related to procedural variations and individual differences (e.g., reaction
speed, cognitive ability, age).
Table 7.5: Mean raw reaction times (SDs) for all three groups on each block of each GNAT
Child Abusers
Non-offenders
Extrafamilial
Non-extrafamilial
Child + Toy
593.1 (65.1)
652.0 (126.5)
624.4 (93.7)
Adult + Toy
641.9 (83.4)
655.9 (115.8)
645.6 (78.6)
Child + Sexual Fantasy 616.7 (69.4)
670.5 (117.4)
659.7 (112.7)
Adult + Sexual Fantasy 637.7 (88.6)
629.5 (103.3)
618.8 (78.5)
Reduction of the data was based on the procedure used by Teachman (2007), who
successfully used a single-concept GNAT. Teachman's procedure followed the
recommendations of the GNAT developers (i.e., Nosek & Banaji, 2001), as well as
Greenwald, Nosek, and Banajis (2003) algorithm for the IAT. First, all trials <300ms were
removed. This is because research indicates that responses to stimuli that occur within 300ms
are not consciously controlled (Bargh & Chartrand, 2000) and categorisation tasks, such as the
GNAT, require the execution of consciously intended responses. Also, Greenwald et al.
(1998) state that some of these outlying responses indicate anticipated responding (i.e.,
response made before the stimulus has been consciously perceived). Errors produced a RT of
zero because, in E-Prime, they reflect no response for a stimulus that should have been
responded to. Using unaltered errors in the scoring of the GNAT would have been a flawed
strategy as the errors would mistakenly suggest that participants were responding to the
stimulus at the exact moment it was presented. Therefore, following the scoring procedure of
other indirect measures, such as the IAT, an error penalty was applied. Nosek (personal
communication, June 6, 2012) recommended replacing the errors with the stimuli duration
deadline (i.e., 1400ms) as it represents the maximum possible RT a participant could produce.
Second, RTs for distracter trials were removed as they are considered noise and, thus,
are not included in the analysis. Third, for all participants, a D-score was calculated for both
the experimental and control GNAT. The D-scores were calculated by subtracting the mean of
153
the trials in Child + Sexual Fantasy block from the mean of the trials in the Adult + Sexual
Fantasy block and then dividing the difference by the standard deviation of all the trials across
both blocks. D-scores that were positive indicated a stronger association between child and
sexual fantasy, whereas negative D-scores indicated a stronger adult-sexual fantasy
association.
7.3 Results
7.3.1 Hypothesis 1: Extrafamilial abusers will report using more child-related sexual
fantasies than non-extrafamilial abusers and non-offenders.
Before testing this hypothesis, the four WSFQ subscales (Intimate, Exploratory, Impersonal,
Sadomasochistic) were examined for each group to determine the types of fantasies the
participants most regularly use.
Figure 7.1: A graph showing the mean scores (and SEs) on each WSFQ subscale across
all groups
154
As shown in Figure 7.1, all three groups showed a similar pattern of results in relation to the
sexual fantasy subscales. Across the three groups, the sexual fantasies that were most
frequently used were Intimate. The least common were Sadomasochistic, with Extrafamilial
abusers using them the least. Both offender groups appeared to use more Exploratory fantasies
than the non-offenders. However, a series of ANOVA tests (where Group was the betweensubjects factor and the total subscale score was the dependent variable) produced no
significant main effects for Group (all ps <.05). Thus, although there appears to be a trend
suggesting that child abusers in this study used more exploratory fantasies and that the
extrafamilial abusers were less sadomasochistic, the differences did not reach significance.
To test the hypothesis that extrafamilial abusers used more sexual fantasies related to
children relative to the two comparison groups, all three groups were compared on the two
child-related WSFQ items using separate one-way ANOVAs. Only one significant main
effect was found, which was for the WSFQ item Having sex with someone much younger
than yourself, F(2, 70) = 18.9, p <.001. Follow-up analyses showed that the mean score for
extrafamilial child abusers (M = 2.35, SD = 1.72) and the other child abusers (M = 1.68, SD =
1.42) were both significantly greater than that of the non-offenders (M = 0.16, SD = 0.37).
The two offender groups did not significantly differ. This does not support Hypothesis 1, but
does support previous findings that child abusers report using child-related fantasies more
frequently than non-offenders (Baumgartner et al., 2002).
On the Thoughts and Fantasies Questionnaire, only sexual fantasies about children
under the age of 13 were of interest. Since this questionnaire does not provide continuous
data, a dichotomous categorical variable labelled deviant fantasiser (Yes/No) was created.
An individual was deemed a deviant fantasy user if they reported using sexual fantasies of a
pre-pubescent child more than once a month; that lasted longer than a few seconds; and that
were clearly elaborate (i.e., involved vivid imagery, had a clear script/narrative). None of the
non-offenders reported such fantasies. Of the 29 extrafamilial offenders, 72% (n = 21)
155
reported using child-related fantasies and 60% (n = 8) of the 19 other offenders reported using
such fantasies. Using chi-square analysis, a significant association was found between offender
group and whether or not a child-related fantasy was used, 2 (1) = 4.41, p <.05. This indicates
that the odds that extrafamilial child abusers will use an elaborate child-related sexual fantasy
are 3.60 times more than the odds of non-extrafamilial abusers. Thus, the results from this
questionnaire support Hypothesis 1.
7.3.2 Hypothesis 2: Extrafamilial abusers will demonstrate stronger associations

between 'child' and 'sexual fantasy on the GNAT than non-extrafamilial abusers and
non-offenders.
It can be seen from Figure 7.2 below that only the extrafamilial abusers had a positive Dscore, indicating a stronger association between child and sexual fantasy. To compare
groups, the D-scores of the experimental GNAT were analysed using a one-way ANOVA. A
significant main effect was found, F(2, 71) = 5.46, p < .05. In order to test the hypothesis that
extrafamilial abusers would hold a significantly stronger association between children and
sexual fantasy, three planned comparisons were performed. Results showed that extrafamilial
abusers had a significantly higher D-scores than both the other child abusers, t(71) = 2.78, p
<.01, (one-tailed), d =.97, and the non-offenders, t(71) = 2.81, p < .01, (one-tailed), d =.71.
According to Cohen (1992), these effects sizes are reflective of a large and a medium effect,
respectively. The D-scores for the other child abusers and the non-offenders did not differ.
Thus, these results are support of Hypothesis 2.
156
Figure 7.2: Mean D-scores and SEs for each group on the experimental GNAT
To corroborate this conclusion further, the error rates for Go and No-Go trials across both
blocks were calculated (see Table 7.6). It can be seen that the non-offenders made the fewest
mistakes overall, while the non-extrafamilial abusers made the most mistakes. After arcsinetransforming the error rates, one-way ANOVAs were performed and showed there was a
significant main effect of Group for the No-Go trials in the Adult-Sexual Fantasy block
[F(2, 71) = 5.9, p <.01] and the Child-Sexual Fantasy block [F(2, 71) = 6.9, p <.01]. Post-hoc
analyses indicated that the non-extrafamilial abusers made more No-Go mistakes (i.e., pressed
the Spacebar when they should not have) than both the extrafamilial abusers and non- offenders
on both blocks.
Given that the extrafamilial abusers made fewer mistakes than the non-extrafamilial
abusers, these data can be seen to provide some corroboration for Hypothesis 2. However, it
should be noted that the within-group differences for the extrafamilial abusers show a bias
that is the opposite of what would be expected. For example, the extrafamilial abusers would
157
be expected to make more No-Go mistakes in the Adult block (i.e., pressing the response key
when they saw an image of a child), relative to the Child block, but this was not found.
Table 7.6: Mean error rates (%) and SDs from the experimental GNAT for extrafamilial
Child Abusers
Non-offenders
Extrafamilial
Non-extrafamilial
Go trials
Child + Sexual Fantasy
1.8 (2.8)
3.9 (5.9)
1.2 (1.6)
Adult + Sexual Fantasy
1.9 (2.9)
2.8 (5.2)
2.3 (2.6)
No-Go trials
Child + Sexual Fantasy
Adult + Sexual Fantasy
4.7 (4.3)
2.9 (4.2)
8.4 (6.7)
7.2 (6.0)
3.1 (3.9)
2.4 (2.6)
7.3.3 Hypothesis 3: Hypothesis 3: On the control GNAT, there will be no group

differences.
As shown in Figure 7.3, all three groups show a positive D-score on the control GNAT. This
suggests that all three groups hold an association between children and toys. Moreover, a oneway ANOVA showed no significant main effect and follow-up analyses confirmed there were
no significant differences between the groups (all ps >.05). This provides corroborating
evidence that the GNAT can be used to assess child-related associations. Also, it indicates
that the group differences on the experimental GNAT are most likely to be due to differing
associations strengths as opposed to factors related to general performance on the GNAT.
158
Figure 7.3: Mean D-scores and SEs for each group using the control GNAT
Table 7.7: Mean error rates (%) and SDs from the control GNAT for extrafamilial
Child Abusers
Non-offenders
Extrafamilial
Non-extrafamilial
Go trials
Child + Toys
1.8 (2.5)
3.9 (5.8)
1.5 (2.7)
Adult + Toys
2.1 (3.3)
3.3 (6.4)
2.2 (3.4)
No-Go trials
Child + Toys
Adult + Toys
6.9 (6.2)
8.3 (8.8)
10.7 (8.3)
13.2 (12.4)
3.7 (3.8)
3.6 (4.9)
Table 7.7 shows the error rates for the Go and No-Go trials for both blocks. It can be seen that,
as with the experimental GNAT, the non-offenders made the least number of incorrect
responses, while the non-extrafamilial abusers made the most errors. One-way ANOVAs on
the arcsine-transformed error rates found a significant main effect of Group for the No-Go
trials in the Adult-Sexual Fantasy block [F(2, 71) = 9.2, p <.001] and the Child-Sexual Fantasy
block [F(2, 71) = 6.7, p <.01]. Post-hoc analyses showed that the non-offenders made
fewer No-Go errors than both the extrafamilial abusers and non-extrafamilial abusers across
159
both blocks. However, there were no group differences between the extrafamilial abusers and
the non-extrafamilial abusers. This lack of difference between the two offender groups
provides some further support for the reaction time results.
7.3.4 Hypothesis 4: Scores on the experimental GNAT will predict deviant sexual
fantasisers regardless of offender subtype
To test this hypothesis, the two offender groups first had to be collapsed into a single sample
(n=48). Then, Receiver Operating Characteristics (ROC) was then used to determine how
well the D-scores on the experimental GNAT predicted deviant fantasisers, as measured by
the Thoughts and Fantasies Questionnaire. ROC curve analysis produces an Area Under the
Curve (AUC), which can be used to determine the predictive effectiveness of a tool. In this
case, the AUC indicates how likely an offender randomly drawn from the group of deviant
fantasisers would have a higher D-score than an offender drawn from the group of nondeviant fantasisers.
An AUC of .50 can be produced by chance alone and indicates that the tool has no
predictive value, while an AUC of .80 is regarded as good (Santtila et al., 2009). The ROC
curve analysis in this study yielded an AUC of .74 (95% CI: .59-.89; p <.01). Thus, this
demonstrates that the experimental GNAT has reasonable predictive validity, which is
supportive of Hypothesis 4.
7.3.5 Exploring the heterogeneity within the non-extrafamilial group

As in Chapter 3, the comparison group (i.e., non-extrafamilial abusers) was made up of three
subgroups (i.e., intrafamilial abusers, predominantly intrafamilial abusers, and abusers of
children and adults). Thus, to strengthen the conclusions that can be drawn, it is important to
determine that the pattern of differences found in this study still hold when the three
160
subgroups are examined individually, particularly those who only abused children (i.e., the
intrafamilial abusers and predominantly intrafamilial).
When examining the two child-related WSFQ items, no significant differences were
found except for between the extrafamilial abusers and the predominantly intrafamilial
abusers. For the item Having sex with someone much younger than yourself, results of a ttest (using a Bonferroni adjustment) indicated that extrafamilial abusers had a significantly
higher score (M = 2.4, SD = 1.72) than the predominantly intrafamilial abusers (M = 0.8, SD
= 0.84); [t(11) = 3.1, p <.01]. Similarly, on the item Seducing an innocent, extrafamilial
abusers showed significantly higher scores (M = 1.5, SD =1.66) than the predominantly
intrafamilial abusers (M = 0.2, SD = 0.45); [t(25) = 3.5, p <.01].
Looking at the GNAT D-scores in Figure 7.4, it can be seen that all three subgroups
showed a pattern of responding that supports the Hypothesis 2. That is, they show a trend for
associating adults with sexual fantasy (relative to the extrafamilial abusers who show the
opposite).
161
Figure 7.4: A graph showing the mean GNAT D-score (and SEs) for the extrafamilial group
and the three non-extrafamilial subgroups
T-test analyses indicated that the predominantly intrafamilial abusers did not significantly
differ from the extrafamilial abusers. However, the intrafamilial abusers were found to
significantly differ from the extrafamilial abusers (p <.05), as did the abusers of children and
adults (p <.05). Given the very small sample sizes of the subgroups, these results are to be
interpreted with caution. However, these results, in conjunction with the trend shown in
Figure 7.4, are supportive of the overall hypothesis that extrafamilial hold stronger childsexual fantasy associations.
7.4 Discussion
The main aim of the present study was to investigate whether child abusers who regularly use
child-related sexual fantasies (i.e., extrafamilial abusers) hold a child-sexual fantasy
162
association in their long-term memory. This was done using an indirect measure called the
GNAT and the results appear to support this assumption. First, in support of Hypothesis 1,
extrafamilial abusers did report more child-related sexual fantasies on the Thoughts and
Fantasies Questionnaire than the two comparison groups. Second, the extrafamilial child
abusers evidenced positive D-scores on the GNAT indicating an association between sexual
fantasy and children. More importantly, this score was significantly different to the negative D-scores produced by both the non-extrafamilial abusers and non-offenders. Thus, this
supported Hypothesis 2.
These results were partially strengthened by the pattern of error rates shown by the
offenders. That is, the extrafamilial abusers made fewer No-Go errors in the Adult block than
the non-extrafamilial abusers. Moreover, the support for Hypothesis 2 was also strengthened
by the corroboration of Hypothesis 3 and 4. With regards to Hypothesis 3, a control GNAT
that measures a non-deviant child-related association (i.e., toy-child association) was
administered to all participants. The fact that the D-scores on the control GNAT was positive
for all three groups and did not significantly differ between the groups indicates that the
GNAT is able to detect child-related associations. Thus, differences between groups on the
experimental-GNAT can be more confidently concluded as being due to group-related
differences (i.e., that the extrafamilial offenders associate children with sexual fantasy). With
regards to Hypothesis 4, it was found that child abusers who use deviant fantasies could be
predicted reasonably well by their D-scores on the experimental GNAT. This further
supported the idea that child-sexual fantasy associations are held by abusers who regularly use
deviant fantasies about children.
The DPM-ST proposes that frequently fantasising about a deviant theme can have a
number of cognitive-based consequences. One consequence is that the deviant theme will
become associated with the concept of sexual fantasy in memory. Similar to research that
assesses belief-related associations in long-term memory (as in Chapter 3), it was
163
hypothesised that this assumption of the DPM-ST could be tested using an indirect measure.
As discussed, the corroborated hypotheses in this study collectively support this assumption
and, therefore, provide some support for this aspect of the DPM-ST.
This study also had a number of other elements that helped strengthen the reliability
of the results. First, the stimuli were words known to be related to sexual fantasy, as they were
derived from existing sexual fantasy measures. Furthermore, each stimulus had a
corresponding distracter word matched in terms of initial letter, syllable length, and character
length. Thus, participants had to pay extra attention to the stimuli and not assume, for
example, that a word starting with Wh was the fantasy word whipping as it could have just
as likely been whirling (a distracter word). Second, the Thoughts and Fantasies Questionnaire
was used as a way of detecting elaborate deviant sexual fantasies as opposed to fleeting sexual
thoughts. The DPM-ST stipulates that it is the controlled process of sexual fantasising that will
lead to the formulation of child-sexual fantasy associations. Thus, the fact that extrafamilial
abusers used more elaborate child-related sexual fantasies and had a
stronger child-sexual fantasy association than the other child abusers, it can be argued that this
specific point of the DPM-ST has some empirical support.
As mentioned in the Introduction, the GNAT has previously been used to assess childsex associations in child abusers in an unpublished study (Snowden et al., 2007). Using
sensitivity as the dependent variable, the researchers found that child abusers evidenced
stronger associations between children and sex than did rapists. However, child abusers did not
significantly differ from non-offenders on this outcome. Moreover, no significant differences
were found when RTs were used as the dependent variable. From this it could be argued that
the GNAT is unable to adequately detect child-related associations using RTs. However, as
discussed, the results on the control GNAT in the present study indicate that this is not case.
The reason is more likely to be due to the authors using a relatively short response window for
participants to categorise stimuli. Nosek and Banaji (2001) recommend using a
164
short response window when the dependent variable is sensitivity, but to extend this window
when using RTs as the dependent variable. In the present study, this recommendation was
followed. It should be noted, however, that the non-extrafamilial group showed the greatest
amount of incorrect responses on the No-Go trials on both GNAT tasks. This is unlikely to be
due to unfamiliarity with the task or with computers as they had a similar mean age to the
extrafamilial abusers. One possible interpretation of this finding could be related to inhibition,
in that, the non-extrafamilial group may have general inhibitory issues, which the GNAT
tasks were detecting in the No-Go trials for both stages. This could be a novel avenue for
future research in this field.
7.4.1 Limitations of study

There are a number of limitations of this study. First, the study would have benefited from
larger sample sizes, particularly in terms of the non-extrafamilial comparison group.
Moreover, the non-extrafamilial group was made up of intrafamilial abusers, abusers of
children and adults, and predominantly intrafamilial abusers. Thus, the results may have been
affected by the heterogeneity within this group. However, an examination of the subgroups
indicated that the three subgroups responded similarly on the experimental GNAT; that is,
they showed stronger adult-sexual fantasy associations. Moreover, when compared to the
extrafamilial abusers, the intrafamilial and abusers of children and adults showed significantly
different GNAT D-scores. Although these results corroborate Hypothesis 2, stronger
conclusions could have been drawn had a pure comparison group (e.g., a group comprised of
only intrafamilial abusers) been used. This is a crucial aim for future research.
Second, most of the child abusers were in treatment. It may be possible that, due to
talking regularly about deviant sexual fantasies in treatment, some offenders have formed
stronger associations between children and sexual fantasy. However, it is unlikely that
165
extrafamilial abusers would talk about sexual fantasies any more than other abusers would
have to do in therapy.
Third, sexual fantasy may be conceptualised differently in the minds of some of the
offenders. For example, for one participant, sexual fantasy may be regarded as something they
would never do (as opposed to reality), whereas for another it may be anything they think
about to induce sexual arousal, which is typically how the term is used in forensic research
and practice (e.g., Langevin et al., 1998). This is important because participants
conceptualising it as the opposite of reality may not regard certain stimuli presented to them
as being a fantasy (e.g., sex). Indeed, one participant raised this point regarding the word
'sex', stating that sex is not a fantasy but a reality for them.
Fourth, the experimental GNAT showed no significant correlations with the two childrelated items of the WSFQ. This may undermine the convergent validity of the GNAT.
However, these particular items have been noted as being rather vague (ODonohue et al.
1997). For example, the sex offenders in this study had a mean age of 45 and 47. Therefore,
someone much younger could include a 25 year old, which is not illegal. Thus, the poor
criterion validity of the WSFQ items may contribute to their lack of correlation with the GNAT D-scores. A fifth limitation is that a non-offender sample was used to evaluate how
strongly the sexual fantasy words were related to the concept of sexual fantasy. Perhaps sex
offenders would have identified different words. Sixth, it is possible that the Sexual FantasyGNAT was simply assessing deviant interest or sexual schemas as opposed to sexual fantasy
per se. For example, the GNATs positive correlation with being a deviant fantasiser may
reflect an association between deviant interest (measured indirectly) and deviant fantasy use,
as the two concepts are associated (Bartels & Gannon, 2011). Also, the word 'sexual' in the
term sexual fantasy may have had greater valence, meaning that the experimental GNAT was
assessing stronger child-sexual associations. Finally, the study could have been strengthened
by using a non-offender group that was not comprised of undergraduate psychology students,
166
as they are likely to be more familiar with using computers and participating in experimental
(reaction-time) research. This is supported by the fact that the non-offenders error rates were
the lowest on both GNATs. Despite these limitations, this study provides evidence that childsexual associations can be assessed indirectly using reaction-time data from a GNAT.
7.4.2 Suggestions of future research

To corroborate the current findings, future work should look to confirm the construct validity
of the experimental GNAT. For example, researchers should investigate whether similar
results are found when the category is changed from Sexual fantasy to simply Fantasy.
This would eliminate the possibility that the experimental GNAT is simply assessing
associations between child and sexual. Since some child abusers have been found to use
deviant fantasies for nonsexual reasons, such as compensating for a lack of self-esteem
(Cortoni et al., 2009), it would be beneficial to ascertain whether child abusers who show no
deviant sexual arousal (i.e., measured using penile plethysmography), but who report using
deviant fantasies, show a positive D-score. This would aid in verifying that that GNAT is
measuring child-sexual fantasy associations and not the construct of deviant sexual
arousal/interest.
Sexual fantasy in the present study was operationalised as imagery depicting deviant
sexual behaviour and themes. Future work should aim to test only those offenders who regard
deviant sexual fantasy in this way, rather than as the opposite to sexual reality. It would also
be useful to investigate whether other indirect measures designed to measure associations
towards single concepts, such as the Single-Category IAT find similar results. Finally, it is also
worth investigating whether the GNAT can be used to assess the deviant fantasies of other
types of sex offenders (i.e., rapists, sexual murderers). If such future work was to corroborate
the experimental GNAT, not only would the DPM-ST be more strongly
167
supported, but it would indicate its potentiality as a clinical tool for assessing entrenched
deviant fantasies.
The findings add to the general indirect measures literature, in that they demonstrate
the GNATs potential as a valuable research tool, particularly when using response latency as
the dependent variable. Hopefully this provides researchers with the confidence to use the
GNAT for future research on sex offenders cognition, particularly as it offers a number of
features that are not available to the more popular IAT (i.e. the use of two possible dependent
variables; providing associations towards a single concept; and allowing the measurement of
absolute associations).
168
CHAPTER 8
GENERAL DISCUSSION
The main aims of this thesis were to: 1) develop a theoretical model to account for the
cognitive processes underlying and associated with deviant sexual fantasies; and 2)
empirically test some of the core assumptions proposed by the model. To highlight the
academic and clinical importance of deviant sexual fantasies, as well as to justify the need for
a contemporary theoretical account, the thesis also aimed to provide an overview of the
relevant empirical and theoretical literature (Chapter1). This information was then used, in
conjunction with insights from other domains (e.g., cognitive psychology), to develop the
Dual-Process Model of Sexual Thinking (DPM-ST); a novel theoretical account of the
cognitive processes underpinning and related to sexual fantasising (Chapter 2). A number of
studies were then conducted to provide preliminary support for various key aspects of the
model. In this final chapter, a summary of the main findings from each study is provided,
followed by a discussion of their theoretical and clinical implications. Limitations of the
thesis and suggestions for future research are also examined.
8.1 Summary of the findings

The first study (Chapter 3) examined four sex-related cognitive associations (child-sex; childnot sex; adult-sex; adult-not sex) in child abusers via the use of a recently developed indirect
measure; the Sorting Paired Features task (SPF). First, exclusively extrafamilial child abusers
were found to hold significantly stronger child-sex associations than the two comparison
groups (i.e., non-extrafamilial abusers and non-offenders). Second, within the extrafamilial
group, the strength of the child-sex associations did not significantly differ from adult-sex
associations. However, both comparison groups showed significantly stronger adult-sex
associations compared to child-sex associations. Third, no group differences were found on a
control version of the SPF task that measured an association most people hold (i.e., flowers169
good). This supports the assumption that the group differences on the main SPF were due to
differing association strengths as opposed to some other factor (i.e., task difficulty, age, motor
functioning). Finally, in the extrafamilal group, adult-sex associations showed a negative
relationship with child-related sexual fantasies, while adult-not sex and child-not sex
associations showed positive relationships. Child-sex associations did not show a significant
relationship with child-related sexual fantasies.
An important point to note is that the results of this study may have been affected by
the strikingly slower adult-not sex RTs. To address this, the data were reanalysed using
amended D-associations that were computed without the two adult-not sex categories. The results
showed that the extrafamilial abusers still demonstrated stronger child-sex associations relative
to the comparison groups. However, their D-association for adult-sex became much weaker (i.e.,
the D-association turned from positive to negative). Moreover, the child-sex D- associations now
showed a significant and positive relationship with the child-related items of the WSFQ. Thus,
the results may be more supportive of the hypotheses than was originally thought. In spite of
this, it would be unwise to make any solid conclusions from either the original or amended
results, as they were biased and omitted some of the SPF data, respectively.
The second study (Chapter 4) tested the hypothesis that the use of aggressive sexual fantasies
by non-offending males is influenced by a greater ability to fantasise (i.e., greater fantasy
proneness, dissociative experiences, and vividness of visual imagery) and the presence of
distorted explicit attitudes about women. Correlational analyses showed that fantasy
proneness and dissociation had a positive relationship with sexual fantasies related to rape and
sadistic acts, as did the questionnaires measuring attitudes about women. Using structural
equation modelling (SEM), a single-path model (i.e., 'Rich Fantasy Life' 'Distorted
Attitudes about Women' 'Aggressive Sexual Fantasies') was specified, identified, and
170
evaluated. The results of the study supported the hypothesis, as the SEM model was found to
show a good fit with the data. However, due to the relatively small sample and non-normal
data, it was possible that the model was biased. Therefore, bootstrapping procedures were
applied and provided statistical support that the model was unbiased.
The third study (Chapter 5) used a within-subjects design to investigate the proposition that
sexual fantasising is a controlled process that draws upon the resources of working memory.
In two experimental conditions, male and female non-offenders envisioned a memory and an
imagination-based sexual fantasy whilst performing another task that requires working
memory (i.e., bilateral eye-movements). The two control conditions involved simply
envisioning a sexual fantasy. The sexual fantasies in each condition were rated on vividness,
emotional valence, and arousability. It was hypothesised that, due to competition for
resources, sexual fantasies would become impaired (i.e., less vivid, less emotional, less
arousing) by eye-movements. The results supported these hypotheses, as sexual fantasies
significantly reduced on all three variables after making eye-movements compared to the
control condition. No differences were found between memory and imagination-based sexual
fantasies in terms of their reduction on the three variables.
The fourth study (Chapter 6) investigated whether sexual fantasies of dominance can prime
(strengthen) self-powerful associations. In the experimental condition, a group of female and
male non-offenders created and imagined a dominance (consensual) sexual fantasy before
completing an Implicit Association Task (IAT) designed to measure self-powerful
associations, as well as a self-report measure of powerfulness, self-esteem, and dominancerelated sexual fantasies. Participants in the control condition created and imagined neutral
imagery (i.e., grocery shopping) before completing the IAT and all three self-report measures.
Results showed no group differences on the IAT or the self-report measures. To aid in
171
explaining the results, exploratory post-hoc analyses were conducted. After collapsing both
groups, scores on the IAT showed a positive relationship with sexual fantasies about forcing.
Also, in the experimental condition, the IAT scores for male participants were found to
positively correlate with sexual fantasies of forcing and hurting a partner.
In the final study (Chapter 7), an indirect measure (i.e., the Go/No-Go Association Task;
GNAT) was used to test the hypothesis that extrafamilial child abusers would associate 'child'
and 'sexual fantasy' more strongly than 'adult' and 'sexual fantasy', relative to nonextrafamilial abusers and non-offenders. The findings from the study were found to support
this hypothesis. No group differences were found on a control GNAT designed to measure
'child-toy' associations, indicating that the GNAT can reliably assess child-related associations.
It also indicated that the group differences found on the child-sexual fantasy GNAT were more
likely due to differences in the strength of child-sexual fantasy associations than other factors
(i.e., those affecting performance on the GNAT). Furthermore, extrafamilial abusers reported
more sexual fantasies about children (<13 years) than did non-extrafamilial abusers and nonoffenders. Finally, across all child abusers, scores on the child-sexual fantasy GNAT were
found to be a good predictor of offenders who reported using sexual fantasies about children.
8.2 Implications of the findings

Deviant sexual fantasies have long been thought of as an important factor in sexual offending.
Research findings on the subject, from over the past few decades, have contributed greatly to
our understanding of their prevalence, content, function, and treatment. While this is the case,
there has been a dearth of empirical and theoretical literature pertaining to the cognitive
processes underlying and associated with deviant sexual fantasising. This is an important
observation to address as a great deal is likely to be learned about deviant sexual fantasies if
172
these processes are investigated. Moreover, insights gained from investigating the cognitive
aspects of deviant sexual fantasy have the potential to open up new avenues for therapeutic
intervention. Thus, in the following two subsections, the results from each of the five studies
will be discussed into terms of their theoretical and clinical implications.
8.2.1 Theoretical implications

The present thesis sought to develop and test an a priori theory of the cognitive processes
associated with deviant sexual fantasies; the Dual-Process Model of Sexual Thinking (DPMST). While the model was originally developed to account for deviant sexual fantasies, many of
the theoretical assumptions are based upon the psychological literature on human cognition in
general. As a consequence, it is argued that the DPM-ST is also applicable to non-deviant
sexual fantasies (as exemplified in Chapter 5). The basic premise of the theory is that there are
two kinds of sexual thinking; 1) spontaneous sexual thoughts that operate on automatic or
associative processes; and 2) sexual fantasies that operate on controlled processes. In brief,
the theory proposes that distorted sex-related associations become activated by relevant
external cues. This activation is experienced as a spontaneous sexual thought entering
conscious awareness, which will be fleeting unless it is attended to further. However, to be
granted further attention, the thought must elicit a strong affective reaction (i.e., sexual
arousal). If attended to, the thought may be elaborated upon. If this elaborative process occurs
and involves mental imagery, it can be regarded as sexual fantasising. The likelihood of the
thought being elaborated upon in the form of a sexual fantasy is increased if the person
exhibits greater levels of sexual preoccupation; possesses a greater ability to fantasise; and
holds explicit cognitions that validate/support the sexual thought and fantasy.
As a controlled cognitive process, sexual fantasising is theorised to involve drawing
upon information in episodic memory, either as a process of recollection or as way of
constructing novel (i.e., imagination-based) imagery. Further, as this information needs to be
173
retained and manipulated in the 'mind's eye', the DPM-ST proposes that sexual fantasising
requires the resources of working memory in order to be performed effectively. The theory
also states that processes involved in sexual fantasising (e.g., constructing and envisioning
sexual scenarios) can prime other associations and strengthen the originally activated
association. Finally, the DPM-ST posits that, in addition to encoding a sexual script, the
repeated use of a specific sexual fantasy theme (i.e., a particular person or behaviour) is likely
to cause an association to form between the concept of 'sexual fantasy' and that particular
theme.
The findings from the five studies in this thesis provide direct and indirect support for
many of these assumptions. First, the model is based on the idea that people hold sex-related
associations, for which there is a growing body of empirical support. For example, a number
of IAT studies suggest that child abusers hold child-sex associations (Babchishin, Nunes, &
Hermann, 2013). However, these results are not conclusive because the IAT scores may be
due to decreased sex-adult associations or increased adult-not sex associations (Snowden,
Craig, & Gray, 2011). The indirect measure used in Chapter 3 (i.e., Sorting Paired Features
task) allowed for the assessment of all four possible associations. The results (based on both
the original and amended analyses) supported the assumption that child abusers (specifically
extrafamilial abusers) hold child-sex associations. Also, in line with Snowden et al.'s (2011)
criticism of the IAT, the results (based on both the original and amended analysis) suggested
that extrafamilal abusers hold decreased adult-sex associations, evidenced by weaker adultsex D-associations relative to the two comparison groups.
While the original analysis found no correlation between child-sex D-associations and
child-related fantasies in the extrafamilial group, the amended analysis showed them to be
positively correlated. This provides some support for the DPM-ST's proposition that child-sex
associations lead to deviant sexual fantasies about children. However, as stated earlier, no
firm conclusions should be drawn from these results. First, the results are correlational and
174
cannot be used to infer causality. Second, the results are based on a post-hoc re-analysis and
so need to be corroborated using an improved SPF task; one that addresses the issue
associated with the adult-not sex category.
Second, the DPM-ST states that once a sexual thought has been triggered, the
likelihood of the thought being elaborated upon in the form of a sexual fantasy is increased by
various factors. Two proposed factors were: 1) individual differences in the ability to fantasise
(i.e., fantasy proneness, vividness of visual imagery, dissociation); and 2) whether the person
holds explicit cognitions that are congruent with the triggered thought and prospective sexual
fantasy. The results of the SEM study in Chapter 4 provide some preliminary support for the
first assumption.
According to Butler (2006), the core variable tying dissociative experiences to other
concomitant constructs, such as fantasy proneness, is 'absorption'. Absorption is defined as the
full commitment of one's cognitive, perceptual, and imaginative resources to a representation of
an attentional object (Tellegen & Atkinson, 1974), and is strongly related to dissociativity and
fantasy proneness (Kihlstrom, Glisky, & Angiulo, 1994; Lynn & Rhue, 1988). According to
Butler (2006), absorption can occur in relation to both external stimuli (i.e., a novel, film) and
internal stimuli (i.e., fantasies, daydreams). Arguably, the quality of a stimulus is likely to
affect the ability to become sufficiently absorbed in it. In the case of internal stimuli, greater
vividness can be reflective of higher quality. This may explain why vividness of visual imagery
was positively associated with fantasy proneness in Chapter 4. Moreover, the
findings from Chapter 4 indicate that sexual fantasies are an internal stimulus that an
individual can become absorbed in, especially if they have a proclivity to do so (i.e., high
fantasy proneness, dissociativity). Further research into how these particular variables
interact with sexual fantasy is highly advocated. For example, it is possible that a reduced
ability to engage in sexual fantasy influences whether an individual uses pornography. This
175
has implications for researchers studying internet offenders who use indecent images of
children.
The findings in Chapter 4 also suggest that a certain theme is more likely to be
sexually fantasised about if the person holds explicit (i.e., consciously accessible) cognitions
that are compatible with the theme. This finding supports Marshall and Marshall's (2000)
argument that sex offenders are likely to introduce expressions of anger and hostility towards
women into their sexual fantasies. In line with the present results, they suggest that this will
often take the form of excessive control, cruelty, humiliation, and degradation. The DPM-ST
furthers these assumptions by stating that, in addition to affecting the content of the
aggressive fantasies, individuals' explicit hostile attitudes towards women are used to validate
whether a particular sexual fantasy should be used (or whether a preceding sexual thought
should be elaborated upon). Arguably, a mismatch between the two would lead to a sense of
cognitive dissonance, reducing the likelihood that the sexual fantasy will be used. However, it
is worth emphasising that, although the findings of Chapter 4 provide support for the idea that
explicit cognitions influence congruent sexual fantasies, they are unable to directly support
the idea that they have a validating function. Thus, future research is encouraged to test and
build upon these ideas.
Third, a core assumption of the DPM-ST is that sexual fantasising is a controlled
process requiring the resources of working memory. The results of Chapter 5 provided direct
support for this hypothesis, as sexual fantasies were found to become impaired after being
performed concurrently with a task known to tax working memory (i.e., bilateral eyemovements). Although previous research has shown that recalled personal memories are rated
as more vivid than imagined future thoughts (D'Argembeau & Van der Linden, 2004), the
baseline ratings between memory-based and imagination-based sexual fantasies in Chapter 5
did not differ. Moreover, both types of fantasies were found to reduce in vividness,
emotionality and arousability to the same degree (statistically speaking). This suggests that
176
the impairment of sexual fantasies (via eye-movements) is unaffected by whether the fantasy
involves mentally "re-experiencing" or "pre-experiencing" a sexual event.
While this paradigm has been used with other forms of mental imagery, it has not been
investigated in relation to sexual fantasies. Thus, this is the first study to support the notion
that sexual fantasies require working memory. The effect observed in this study can be
explained by the idea that, due to reduced resources, the sexual imagery was unable to be
formed as effectively (or as vividly) as when the resources were not being used by a
concurrent task. Smeets, Dijs, Pervan, Engelhard, and van den Hout (2012) argue that eyemovements affect the refreshment of visual information in working memory (Kosslyn, 1994).
As a result, the imagery is reconsolidated (i.e., stored in long-term memory) as less vivid.
When recalled later (i.e., post-task), the imagery still appears less vivid and, thus, is deemed as
having less emotional and erotic valence. In contrast, when no eye-movements are made, the
imagery is able be refreshed and become more vivid. The results of the control condition in
Chapter 5 support this proposition as sexual fantasies become more vivid after simply
fantasising (i.e., making no eye-movements). There is some small indication that this occurs in
sex offenders also. For example, in a qualitative study, sex offender reported that their sexual
fantasies "became more concrete each time, they were extremely vivid in my mind" (Gee,
Ward, Beech, & Belofastov, 2006, p. 220).
The idea that a sexual fantasy is deemed less positive and arousing as a consequence
of being less vivid also lends support to the findings (and hypotheses) in Chapter 4, since
vividness of visual mental imagery (in conjunction with low fantasy proneness and
dissociative tendencies) was found to influence the use of sexual fantasies. While vividness
was experimentally reduced in Chapter 5, both findings taken together suggest that the
phenomenological quality of the imagery is important when appraising and using sexual
fantasies.
177
Another assumption of the DPM-ST is that a feedback loop is formed when a sexual
fantasy is used. This is because the content of the sexual fantasy is likely to strengthen the
association that leads to its use. Related to this, it is also proposed that the content of sexual
fantasies have the potential to activate (and strengthen) other associations. In other words,
sexual fantasies can function as a source of priming (Blair, Ma, & Lenton, 2001). This
hypothesis was tested in Chapter 6 using dominance sexual fantasies and self-powerful
associations. Unfortunately, the hypothesis was not supported. Before solid theoretical
conclusions are made, further studies need to be conducted to ensure the results were not due
to methodological limitations. This is because support for this aspect of the DPM-ST can
offer insights into how sexual fantasies provide sex offenders with a sense of power and
control, as well as other non-sexual gains.
The DPM-ST also attempts to provide a cognitive account for why specific themes
(i.e., types of people, a particular person, whipping) are regularly incorporated into the
content of an individuals sexual fantasy. It is proposed that, in addition, to encoding specific
behavioural scripts, repeatedly fantasising about a particular theme can lead that theme to
becoming associated with the concept of sexual fantasy. If this is the case, indirect measures
should be able to detect the association. The results from Chapter 7 support this hypothesis, as
child abusers who regularly use deviant sexual fantasies (i.e., extrafamilial abusers) were
found to harbour child-sexual fantasy associations. These findings suggest that when the word
'sexual fantasy' is heard, thought of, or read by an offender who has a history of repeatedly
using child-related fantasies, they are likely to automatically think of children or a specific
child. However, before any solid conclusions are made regarding these specific theoretical
prospects, replication studies and further research designed to rule out alternative explanations
are needed.
In summary, the DPM-ST has generated a number of testable hypotheses, many of
which are novel and have not been previously investigated. Moreover, many of the
178
hypotheses were supported, providing the initial steps towards corroborating the DPM-ST. As
such, this thesis has contributed to furthering our theoretical understanding of the cognitive
processes underlying and associated with deviant sexual fantasies/fantasising.
8.2.2 Clinical implications

The DPM-ST offers a number of potential implications for clinicians, both in terms of
assessing and treating deviant sexual fantasies. With regards to assessment, the results of
Chapter 7 suggest that indirect measures (i.e., the GNAT) can be used to assess deviant sexual
fantasies, in terms of finding an association between the concept of sexual fantasy and the
deviant theme depicted in the fantasy (e.g., children). While many indirect measures are
thought to be useful for assessing sexual deviance (Banse, Schmidt, & Clarbour, 2010), not all
sex offenders who have a sexual preference for children report corresponding sexual fantasies
(Langevin, Lang, & Curnoe, 1998). As Beech and Ward (2004) note, many sex offenders with
normative sexual scripts will use deviant fantasies as way of increasing their sense of wellbeing. Thus, an indirect measure specifically designed to assess deviant sexual fantasies may
provide a way of detecting individuals who frequently use deviant fantasies, irrespective of
whether they harbour a deviant sexual preference. The results of Chapter 7 provide some
preliminary support for this, as the GNAT showed reasonably good predictive validity for
predicting deviant sexual fantasisers when the subtypes were collapsed together.
Further research aimed at increasing the effectiveness of an indirect assessment tool for
deviant sexual fantasy is a valuable endeavour given that sexual fantasies are a highly private
and personal form of mental imagery. This is likely to be even more so for individuals who
harbour deviant sexual fantasies, particularly sex offenders who may want to appear as though
they are rehabilitated (Hall, 1996). As such, self-report methods of assessing sexual fantasy
are likely to be susceptible to faking and socially desirable responding. A valid and
179
reliable indirect measure for assessing deviant sexual fantasy will provide a way to avoid this
issue, as indirect measures are thought to be less susceptible to faking (Snowden et al., 2011).
Finally, the results of Chapter 5 suggest that, in line with assessing the content of
sexual fantasies, it might also be worthwhile for clinicians to assess factors associated with
fantasy ability (i.e., fantasy proneness, vividness of visual imagery). This may provide an
indication of how deeply the offender becomes involved/absorbed in their sexual fantasies. It
will also indicate whether vividness is a factor to address in treatment (see below).
With regards to treatment, some of the findings provide support for devising
techniques that do not focus on (but can be used in conjunction with) strategies that aim to
alter sexual arousal levels (i.e., directed masturbation, satiation). For example, the results
from Chapter 4 suggest that it might be beneficial to address an individual's ability to fantasise
in general, particularly if they present as fantasy prone and dissociative. This may involve
identifying and encouraging activities that the offender expresses genuine interest in and
which he is likely to become absorbed in (i.e., writing, drawing, solving puzzles, playing
chess, playing music). The aim of this strategy is to channel the individual's attention away
from sexual fantasies and on to other, less risky stimuli. The strategy could also be tied in
with a strengths-based approach to treatment (i.e., the Good Lives Model; Ward & Stewart,
2003), so that the activity also enables the offender to acquire other needs that will help him
function effectively in the community. However, these suggestions are preliminary and based
on results with non-offenders. Thus, more research is needed before any action is taken to
develop or adopt such strategies.
According to the DPM-ST, weakening and/or changing distorted associations would
ultimately reduce the likelihood that sexual fantasies will be used. Unfortunately, due to the
nature of the results in Chapter 3, this theoretical implication is unable to be supported.
However, according to the SEM model in Chapter 4, distorted explicit cognitions about
women played a role in determining whether aggressive sexual fantasies were used. Thus, it
180
can be argued that changing these explicit cognitions would reduce the use of aggressive
sexual fantasies. There are a few studies showing that offenders' distorted explicit cognitions
(i.e., those measured using self-report tools) can be altered using cognitive restructuring
techniques (Beech et al., 2013). Thus, future research should aim to assess deviant sexual
fantasy use before and after cognitive restructuring to see if this hypothesised technique is
effective.
A final implication for treatment is offered by the results in Chapter 5. These results
showed that the vividness, emotionality, and arousability of non-offenders' sexual fantasies are
reduced by eye-movements. Thus, it can be argued that eye-movements may be an effective
technique to use with sexual offenders who have trouble with deviant sexual fantasies. The use
of eye-movements to clinically impair mental imagery has been researched in other domains
and fall under what Hackmann, Bennett-Levy, and Holmes (2011) referred to as "imagerycompeting" tasks. As the name suggests, these techniques affect the cognitive processes
associated with mental imagery by competing for the same working memory resources
(Gunter & Bodner, 2008; van den Hout, Engelhard, Beetsma... et al., 2011). In previous
research, these techniques have been applied to positive memories (Hornsveld et al.,
2011); negative imagery about the future (Engelhard, van den Hout, Janssen, & van der Beek,
2010); and food-related imagery (McClelland, Kemps, & Tiggemann, 2006). However, eyemovements have mostly been applied to traumatic and negative mental imagery, largely as a
way of providing insights into how Eye-Movement Desensitisation and Reprocessing
(EMDR) operates (Andrade, Kavanagh, & Baddeley, 1997; Gunter & Bodner, 2008; van den
Hout, Muris, Salemink, & Kindt, 2001).
EMDR is an effective, 8-stage therapeutic protocol for treating trauma-related
disorders (Shapiro, 1995). During the fourth stage, clients are required to bring the traumatic
memory to mind and make a series of bilateral eye-movements. A recent meta-analysis has
demonstrated that the use of eye-movements is an effective component of the EMDR process
181
(Lee & Cuijpers, 2013). An EMDR study using child abusers aimed to treat traumatic
imagery associated with childhood sexual abuse (Ricci, Clayton, & Shapiro, 2006). The
results showed a reduction in sexual thoughts and deviant sexual arousal. In response to
these findings, Vanhoeck, Van Daele and Gykiere (2011) argued that EMDR could be used to
reduce deviant sexual fantasies related to an offenders victimisation memories. However the
results of Chapter 5 indicate that EMDR (or at least the eye-movement component of EMDR)
can be applied to treating sexual fantasies per se.
It could be argued that sexual fantasies based on real experiences are likely to be
harder to treat with eye-movements due to their higher erotic valence (McGuire et al., 1965)
and the richer quality associated with emotional autobiographical memories (Schaefer &
Philippot, 2005). However, in line with previous findings (Engelhard et al. 2010), the results
from Chapter 5 show that eye-movements can impair purely imagined sexual fantasies to the
same extent as sexual fantasies based on recalled experiences. It can also be argued that the
non-offender participants used their sexual fantasies because they found them to be
pleasurable and positive. Thus, given their impairment, the results of Chapter 5 also support
previous findings related to impairing positive memories with eye-movements (Hornsveld et
al., 2010). More importantly, this is a clinically useful finding as many sexual offenders are
likely to regard their sexual fantasies as positive. A technique that can affect a sexual fantasy
regardless of this fact may prove highly beneficial.
In summary, the results of Chapter 5 can be seen as the first step towards formulating
a new strategy for treating deviant for sexual fantasies. It is now important to replicate these
results, particularly with sex offenders who have problems with deviant sexual fantasies.
Also, Kavanagh, Freese, Andrade, and May (2001) and Gunter and Bodner (2008) found that
the effects of EMs on negative memories were still present at a one week follow-up. Thus, it
would also be advantageous for researchers to investigate whether the changes in vividness,
182
emotionality, and arousability of sexual fantasies persist at a follow-up period, as this would
also have direct implications for the treatment of deviant fantasies.
8.3 Limitations and future directions

Throughout this chapter, a number of limitations have been mentioned. In this following
section, a number of other noteworthy limitations will be highlighted so that future research
can build upon the results effectively. First, in Chapter 3, there were a few issues regarding
the Sorting Paired Features (SPF) task. The most significant limitation relates to the stimuli
used for the adult-not sex category, as all three groups were noticeably slower at categorising
them. This appeared to be facilitated by making an initial incorrect response and greatly
affected the results. This issue may have occurred because the not-sex stimuli are an
inadequate reflection of the category. For example, in their IAT study, Ciardha and
Gormley (2012) used words that could be used to describe an unattractive or non-sexual
person (e.g., dull, ugly, boring). Although the stimuli used in Chapter 3 were derived from
previous IAT studies (Gray, Brown, MacCulloch, Smith, & Snowden, 2005), adult-not sex
and child-not sex associations were not examined in those studies as they are not
needed/included in the calculation of an IAT score. Thus, this limitation of the SPF study may
highlight a general limitation of this particular stimuli set and calls into question the results of
previous IAT studies that have used them.
Another limitations of the SPF was the use of just 80 trials (20 per association).
Although this was recommended by the developer of the SPF and resulted in supportive
findings, it seems that, in retrospect, an increased number of trials may have been better for
obtaining more reliable results. It is also worth highlighting that the sample sizes were
relatively small. While it is important to delineate offender subtypes so that potentially
important effects are not cancelled out, larger sample sizes are needed to ensure the study is
not underpowered. Also, larger samples will increase the likelihood that there will be enough
183
offenders in each subtype. In the SPF, only 13 non-extrafamilial offenders were identified.
Although reliable SPF results have been found and published with samples of just 14
participants (Bar-Anan, Nosek, & Vianello, 2009), it is highly recommended future
researchers try and gain larger samples.
The SEM study in Chapter 4 also suffered from a sample size issue and so it is
recommended that larger samples be used in the future. Also, the results could be made more
reliable if an extra indicator was added to the Distorted Sexual Beliefs and Aggressive Sexual
Fantasies latent variables. In retrospect, another latent variable that may have been important to
have included in the SEM model is that of sexual preoccupation, as it is regarded as being
closely associated with the use of sexual fantasies (Knight & Knight-Sims, 2003). This should
be taken into consideration when looking to test a SEM model of this kind in the future.
While the DPM-ST states that working memory underlies all sexual fantasies
regardless of their content (i.e., deviant and non-deviant fantasies), the present thesis would
have benefited if the study in Chapter 5 had also used a sex offender sample or focused on
deviant sexual fantasies. As stated earlier, this is certainly warranted for future research. Also,
a more randomised strategy for allocating sexual fantasies to participants would have helped
reduce the likelihood of the results being influenced by demand characteristics. Another point
is that, although previous studies have shown eye-movements slow down reaction times on a
task known to require working memory (e.g., van den Hout, Engelhard, Beetsma et al., 2011),
it would have been worthwhile to have demonstrated and corroborated this in Chapter 5. This
would have helped strengthen the argument that sexual fantasies require working memory.
In retrospect, the null effects of Chapter 6 are likely due to a number of possible
methodological limitations. For example, the sample was comprised of mainly female
participants. Previous research shows that females are more likely to use fantasies about
submission rather than dominance (Zurbriggen & Yost, 2004). As the findings in Chapter 4
show, if someone does not hold explicit cognitions related to a particular sexual fantasy, they
184
are unlikely to use it. In light of these two points, it is possible that dominance fantasies were
unable to sufficiently prime self-powerful associations in many of the female participants. As
a result, future research should aim to focus on male participants, particularly those with a
proclivity for coercive sex (including sex offenders).
Also, it would have been useful to have acquired baseline scores on the self-powerful
IAT (i.e., pre-manipulation) in order to more confidently ascertain whether any changes
occurred after envisioning dominance sexual fantasies. Another important consideration is
whether the inclusion of a 'no-imagery' control group would have been useful to ensure that
the neutral imagery was not increasing cognitive load and, therefore, affecting the IAT results.
Given that previous research has convincingly shown that mental imagery can prime
associations (Blair et al., 2001), similar studies to that in Chapter 6 should be conducted, but
with the above limitations taken into account.
Finally, the study in Chapter 7 had some limitations. The main limitation was that the
GNAT results may have been due to participants responding to the word 'sexual' in the
category 'Sexual Fantasy'. Thus, future research should aim to use a GNAT with the category
label 'Fantasy' as it will help provide insights into whether the present results were due to
responding to the word 'sexual' or the phrase 'sexual fantasy'. If such a GNAT was able to show
group differences and predict deviant sexual fantasisers to the same degree as in this thesis,
stronger conclusions could be made. A further point is that some offenders mentioned that the
stimulus word 'sex' was not regarded as a sexual fantasy. This indicates that perhaps the
concept of fantasy was not viewed as a form of mental imagery but rather as the opposite of
reality. This would need to be addressed in future studies, perhaps by using the GNAT with
sex offenders who define sexual fantasy as a form of mental imagery.
In discussing these limitations, a number of suggestions for future research were
raised. In addition to these, there are some other notable future directions to highlight. First,
the use of pictorial stimuli with the SPF may be useful as it will introduce the option of
185
assessing non-focal associations, such as those related to gender or race. In others words, a
pictorial SPF would enable associations between female child-sex and female adult-sex to be
compared and the same for male related associations. While gender has been taken into
account in IAT studies assessing sexual interest in children (Banse et al., 2010), it has done so
using separate IATs (i.e., one for each gender). Thus, a pictorial SPF would allow genderbased associations to be assessed within one measure.
Another idea would be to investigate whether other 'imagery-competing' tasks (as
opposed to eye-movements) can also be used to impair sexual fantasies. Tasks used in
previous research include attentional breathing (van den Hout, Engelhard, Beetsma et al.,
2011), mental arithmetic (Engelhard, van den Hout, & Smeets, 2011), the computer game
Tetris (Engelhard, van Uijen, & van den Hout, 2010), and clay modelling (Andrade, Pears,
May, & Kavanagh, 2012). Not only would this provide further support for the idea that
working memory underlies sexual fantasising, but it would help ascertain if there are better
imagery-competing tasks than eye-movements.
Finally, the DPM-ST is able to generate other hypotheses that were not examined in
this thesis. For example, to test the idea that deviant sexual fantasies are influenced by
activated sex-related associations, different groups of participants could be asked to write a
sexual fantasy after a certain association has been primed. If participants write a fantasy related
to the primed association, then this aspect of the DPM-ST would have some support. Other
research ideas include testing whether individuals with a low working memory capacity or
who experience a high cognitive load (i.e., stress) produce less vivid sexual fantasies than
those with higher (or unloaded) working memory, and whether a low working memory
capacity influences the use of external imagery (i.e., pornography), as opposed to internal
sexual fantasies. Also, to support the DPM-ST further, it would be advantageous for future
research to find more conclusive support that fleeting sexual thoughts are distinct from
186
elaborate sexual fantasies, as well as whether imagination-based sexual fantasies require

drawing upon episodic information stored in memory.
8.4 Conclusion
This thesis aimed to further our understanding of the cognitive processes associated with
deviant sexual fantasies by knitting together the literature on deviant sexual fantasy, mental
imagery, associative and controlled processes, episodic memory/future thought, and fantasyrelated abilities to produce a novel theoretical model; The Dual-Process Model of Sexual
Thinking (DPM-ST). Taken as a whole, the main empirical findings of this thesis indicate
that: 1) extrafamilial child abusers are more likely to hold distorted sex-related associations;
2) the use of a deviant sexual fantasy is partly dependent upon the individual's ability to
fantasise in general (i.e., higher fantasy proneness, vividness of visual imagery, and
dissociation) as well as on whether they harbour explicit cognitions that are congruent with
the sexual fantasy; 3) if a sexual fantasy is engaged in, it will be at the expense of engaging in
other cognitive tasks as the process requires the resources of working memory; 4) if other
cognitively effortful tasks are performed at the time as sexual fantasising, then the sexual
fantasy is likely to become impaired when brought back to mind at a later time; and 5) the
frequent use of deviant sexual fantasies may cause an association to be formed between the
deviant theme and the concept of sexual fantasy.
On a general level, the results of this thesis demonstrate just how multifaceted sexual
fantasies are (Bartels & Gannon, 2011). In addition, the results provide a number of novel
insights into the nature of deviant (and non-deviant) sexual fantasies. This has potential
implications for both researchers and clinicians working with deviant sexual fantasies. Indeed,
researchers in this field should aim to follow the various recommendations for future research
in this thesis as they will help build upon the present findings and contribute to a fuller
understanding of deviant sexual fantasy. Moreover, corroboration of the findings will provide
187
further justification for actualising the clinical implications outlined in this chapter (e.g., using
eye-movements as a treatment strategy). Given that the DPM-ST has gained some preliminary
support, it would also be beneficial for researchers to test other hypotheses generated by the
model.
In closing, it is important to emphasise that the DPM-ST and the findings of this thesis
only provide insights into the cognitive nature of deviant sexual fantasies. This is only one
piece of a larger puzzle and it is important to keep in mind that other factors (i.e.,
endocrinological, neurological, behavioural, and affective) also play an important role in
influencing sexual fantasies (Bartels & Gannon, 2011; Toates, 2009). Thus, an eventual aim
should be to investigate how these cognitive processes interrelate with these various other
factors, as it will provide a more holistic and sophisticated account of how deviant sexual
fantasies operate and effect offending behaviour.
188
References
Abel, G. G., & Blanchard, E. B. (1974). The role of fantasy in the treatment of sexual
deviation. Archives of General Psychiatry, 30, 467-475.
Addis, D. R., Wong, A. T., & Schacter, D. L. (2007). Remembering the past and imagining
the future: common and distinct neural substrates during event construction and
elaboration. Neuropsychologia, 45, 13631377.
Aleman, A., & de Haan, E. H. (2004). Fantasy proneness, mental imagery and reality
monitoring. Personality and Individual Differences, 36, 1747-1754.
American Psychiatric Association. (2000). Diagnostic and statistical manual of mental
disorder (Fourth edition, Text revision). Washington, DC: American Psychiatric
Association.
Andersen, B. L., Cyranowski, J. M., & Espindle, D. (1999). Men's sexual self
schema. Journal of Personality and Social Psychology, 76, 645-661.
Andrade, J., Kavanagh, D., & Baddeley, A. (1997). Eye-movements and visual imagery: A
working memory approach to the treatment of posttraumatic stress disorder. British
Journal of Clinical Psychology, 36, 209223.
Andrade, J., Pears, S., May, J., & Kavanagh, D. J. (2012). Use of a clay modeling task to
reduce chocolate craving. Appetite, 58, 955-963.
Arrigo, B. A., & Purcell, C. E. (2001). Explaining paraphilias and lust murder: Toward an
integrated model. International Journal of Offender Therapy and Comparative
Criminology, 45, 6-31.
Atance, C. M., & ONeill, D. K. (2001). Episodic future thinking. Trends in Cognitive
Sciences, 5, 533539.
Babchishin, K. M., Nunes, K. L., & Hermann, C. A. (2012). The validity of Implicit
Association Test (IAT) measures of sexual attraction to children: A metaanalysis. Archives of Sexual Behavior, 42, 487-499.
Baddeley, A. D., & Andrade, J. (2000). Working memory and the vividness of imagery.
Journal of Experimental Psychology: General, 129, 126145.
Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. H. Bower (Ed.), The
psychology of learning and motivation, Volume. 8 (pp. 4789). New York: Academic
Press.
Baddeley, A. D. (1996). Exploring the central executive. Quarterly Journal of Experimental
Psychology, 49, 528.
189
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change.

Psychological Review, 84, 191-215.
Banse, R., Schmidt, A. F., & Clarbour, J. (2010). Indirect measures of sexual interest in child
sex offenders: A multi-method approach. Criminal Justice and Behavior, 37, 319-335.
Bar-Anan, Y., & Nosek, B. A. (2013). A comparative investigation of seven indirect
measures of social cognition. Unpublished manuscript. Retrieved from:
papers.ssrn.com/sol3/papers.cfm?abstract_id=2074556
Bar-Anan, Y., Nosek, B.A., & Vianello, M. (2009). The sorting paired features task: A
measure of association strengths. Experimental Psychology, 56, 329-343.
Bargh, J. A., & Chartrand, T. L. (2000). The mind in the middle: A practical guide to priming
and automaticity research. In H. Reis & C. Judd (Eds.), Handbook of research methods
in social and personality psychology (pp. 253-285). New York: Cambridge University
Press
Bargh, J. (1994). The four horsemen of automaticity: Awareness, intention, efficiency, and
control in social cognition. In R. S. Wyer & T. K. Srull (Eds.), Handbook of social
cognition, 2nd ed, (pp. 1-40). Hillsdale, NJ: Erlbaum.
Barlow, D. H., Hayes, S. C., & Nelson, R. O. (1984). The scientist practitioner: Research
and accountability in clinical and educational settings (Vol. 128). New York:
Pergamon Press.
Bartels, R M., & Gannon, T. A. (2009). Rape supportive cognition, sexual fantasies, and
implicit offense -scripts: A comparison between high and low rape-prone men. Sexual
Abuse in Australia and New Zealand, 1, 14-20.
Bartels, R. M., & Gannon, T. A. (2011). Understanding the sexual fantasies of sex offenders
and their correlates. Aggression and Violent Behavior, 16, 551-561.
Baumgartner, J. V., Scalora, M. J., & Huss, M. T. (2002). Assessment of the Wilson Sex
Fantasy Questionnaire among child molesters and nonsexual forensic offenders.
Sexual Abuse: A Journal of Research and Treatment, 14, 1930.
Becker-Blease, K., & Freyd, J. J. (2007). Dissociation and memory for perpetration among
convicted sex offenders. Journal of Trauma and Dissociation, 8, 69-80.
Becker-Blease, K., Friend, D., & Freyd, J. J. (2006). Child sex abuse perpetrators among
male university students. Poster presented at the 22st Annual Meeting of the
International Society for Traumatic Stress Studies, Hollywood CA, November 4-7,
2006.
Beckett, R. C. (1994). Assessment of sex offenders. In T. Morrison, M. Erooga, & R. C.
Beckett (Eds.), Sexual offending against children (pp. 80-101). London: Routledge.
190
Beech, A. R., & Ward, T. (2004). The integration of etiology and risk in sexual offenders: A
theoretical framework. Aggression and Violent Behavior, 10, 31-63.
Beech, A. R., Bartels, R. M., & Dixon, L. (2013). Assessment and treatment of distorted
schemas in sexual offenders. Trauma, Violence, and Abuse, 14, 54-66.
Beech, A. R., Ward, T., & Fisher, D. (2006). The identification of sexual and violent
motivations in men who assault women: Implication for treatment. Journal of
Interpersonal Violence, 21, 1635-1653.
Beech, A., Friendship, C., Erikson, M., & Hanson, R. K. (2002). The relationship between
static and dynamic risk factors and reconviction in a sample of UK child
abusers. Sexual Abuse: A Journal of Research and Treatment, 14, 155-167.
Beevers, C. G. (2005). Cognitive vulnerability to depression: A dual process model. Clinical
Psychology Review, 25, 975-1002.
Bentler, P. M., & Wu, E. J. (2005). EQS 6.1 for Windows. Encino, CA: Multivariate
Software.
Binet, A. (1888). Le Fefichisme duns IAmour. Paris.
Birnbaum, G. E., Simpson, J. A., Weisberg, Y. J., Barnea, E., & Assulin-Simhon, Z. (2012).
Is it my overactive imagination? The effects of contextually activated attachment
insecurity on sexual fantasies. Journal of Social and Personal Relationships, 29,
1131-1152.
Blair, I. V., Ma, J. E., & Lenton, A. P. (2001). Imagining stereotypes away: the moderation of
implicit stereotypes through mental imagery. Journal of Personality and Social
Psychology, 81, 828-841.
Bourke, P., Ward, T., & Rose, C. (2012). Expertise and sexual offending a preliminary
empirical model. Journal of Interpersonal Violence, 27, 2391-2414.
Briere, J., & Runtz, M. (1989). University males sexual interests in children: Predicting
potential indices of pedophilia in a non-forensic population. Child Abuse and
Neglect, 13, 65-75.
Brittain, R. P. (1970). The sadistic murderer. Medicine, Science, and the Law, 10, 198-207.
Brogden, W. J. (1939). Sensory pre-conditioning. Journal of Experimental
Brown, J. J., & Hart, D. H. (1977). Correlates of females sexual fantasies. Perceptual and
Motor Skills, 45, 819-825.
Brown, A. S., Gray, N. S., & Snowden, R. J. (2009). Implicit measurement of
associations in child sex abusers: Role of victim type and denial. Sexual Abuse: A
Journal of Research and Treatment, 21, 166180.
191
Bruce, J. M., Bruce, A. S., Hancock, L., & Lynch, S. (2010). Self-reported memory problems
in multiple sclerosis: Influence of psychiatric status and normative dissociative
experiences. Archives of Clinical Neuropsychology, 25, 39-48.
Businelle, M. S., Kendzor, D. E., Reitzel, L. R., Costello, T. J., Cofta-Woerpel, L., Li, Y., ...
& Wetter, D. W. (2010). Mechanisms linking socioeconomic status to smoking
cessation: A structural equation modeling approach. Health Psychology, 29, 262.
Butler, L. D. (2006). Normative dissociation. Psychiatric Clinics of North America, 29, 4562.
Byers, E. S., Purdon, C., & Clark, D. A. (1998). Sexual intrusive thoughts of college students.
Journal of Sex Research, 35, 359369.
Byrne, B. M. (2010). Structural equation modeling with AMOS: Basic concepts, applications,
and programming (2nd ed.). New York: Taylor and Francis Group.
Carabellese, F., Maniglio, R., Greco, O., & Catanesi, R. (2010). The role of fantasy in a serial
sexual offender: A brief review of the literature and a case report. Journal of Forensic
Sciences, 56, 256260.
Carlson, E. B., & Putnam, F. W. (1993). An update on the Dissociative Experiences Scale.
Dissociation, 6, 1627.
Carlson, E. B., Putnam, F. W., Ross, C. A., Torem, M., Coons, P., & Dill, D. L. (1993).
Validity of the dissociative experience scale in screening for multiple personality
disorder: A multicenter study. American Journal of Psychiatry, 150, 1030-1036.
Cautela, J. R., & Baron, M. G. (1977). Covert conditioning a theoretical analysis. Behavior
Modification, 1, 351-368.
Cheung, G. W., & Lau, R. S. (2008). Testing mediation and suppression effects of latent
variables bootstrapping with structural equation models. Organizational Research
Methods, 11, 296-325.
Clark, D. A., Purdon, C., & Byers, E. S. (2000). Appraisal and control of sexual and nonsexual intrusive thoughts in university students. Behaviour Research and Therapy,
38, 439-455.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale,
NJ: Erlbaum.
Corballis, M. C. (2011). The recursive mind: The origins of human language, thought, and
civilization. Princeton, NJ: Princeton University Press.
Cortoni, F., & Marshall, W. L. (2001). Sex as a coping strategy and its relationship to
juvenile sexual history and intimacy in sexual offenders. Sexual Abuse: A Journal of
Research and Treatment, 13, 27-43.
192
Cortoni, F., Proulx, J., Paquette, S., Longpr, N., & Coutre, M. (2009, October). The
developmental context of deviant sexual fantasies among child molesters. Paper
presented at the 28th Annual Research and Treatment Conference of the Association
for the Treatment of Sexual Abusers, Dallas, TX.
Crpault, E., & Couture, M. (1980). Mens erotic fantasies. Archives of Sexual Behavior, 9,
565-581.
Curnoe, S., & Langevin, R. (2002). Personality and deviant sexual fantasies: An examination
of the MMPIs of sex offenders. Journal of Clinical Psychology, 58, 803815.
D'Argembeau, A., Comblain, C., & Van der Linden, M. (2003). Phenomenal characteristics
of autobiographical memories for positive, negative, and neutral events. Applied
Cognitive Psychology, 17, 281-294.
Dadds, M., Bovbjerg, D., Redd, W., & Cutmore, T. (1997). Imagery in human classical
conditioning. Psychological Bulletin, 122, 89103.
Dandescu, A., & Wolfe, R. (2003). Considerations on fantasy use by child molesters and
exhibitionists. Sexual Abuse: A Journal of Research and Treatment, 15, 297-305.
DArgembeau, A., & Van der Linden, M. (2006). Individual differences in the
phenomenology of mental time travel: The effect of vivid visual imagery and emotion
regulation strategies. Consciousness and Cognition, 15, 342-350.
Derogatis, L.R. (1976). The DSFI: A multidimensional measure of sexual functioning.
Journal of Sex and Marital Therapy, 5, 244281.
Deu, N., & Edelmann, R. J. (1997). The role of criminal fantasy in predatory and opportunist
sex offending. Journal of Interpersonal Violence, 12, 18-29.
Edwards, N. B. (1972). Case conference: Assertive training in a case of homosexual
pedophilia. Journal of Behavior Therapy and Experimental Psychiatry, 3, 55-63.
Ellis, B. J., & Symons, D. (1990). Sex differences in sexual fantasy: An evolutionary
psychological approach. Journal of Sex Research, 27, 527-555.
Engelhard, I. M., van den Hout, M. A., & Smeets, M. A. (2011). Taxing working memory
reduces vividness and emotional intensity of images about the Queens Day
tragedy. Journal of Behavior Therapy and Experimental Psychiatry, 42, 32-37.
Engelhard, I. M., van den Hout, M. A., Dek, E. C., Giele, C. L., van der Wielen, J. W.,
Reijnen, M. J., & van Roij, B. (2011). Reducing vividness and emotional intensity of
recurrent flashforwards by taxing working memory: An analogue study. Journal of
anxiety disorders, 25, 599-603.
193
Engelhard, I. M., van den Hout, M. A., Janssen, W. C., & van der Beek, J. (2010). Eyemovements reduce vividness and emotionality of flashforwards. Behaviour
Research and Therapy, 48, 442-447.
Engelhard, I. M., van Uijen, S. L., & van den Hout, M. A. (2010). The impact of taxing
working memory on negative and positive memories. European Journal of
Psychotraumatology, 1, 5623, doi:10.3402/ejpt.v1i0.5623.
Evans, J. S. B. (2008). Dual-processing accounts of reasoning, judgment, and social
cognition. Annual Review of Psychology, 59, 255-278.
Finke, R. A. (1989). Principles of mental imagery. Cambridge, MA: MIT Press.
Firestone, P., Bradford, J. M., Greenberg, D. M., & Serran, G. A. (2000). The relationship of
deviant sexual arousal and psychopathy in incest offenders, extrafamilial child
molesters, and rapists. Journal-American Academy of Psychiatry and the Law, 28,
303-308.
Firestone, P., Bradford, J. M., Greenberg, D. M., Larose, M. R., & Curry, S. (1998).
Homicidal and nonhomicidal child molesters: psychological, phallometric, and
criminal features. Sexual Abuse: A Journal of Research and Treatment, 10, 305-
323.
Gannon, T. A., & Polaschek, D. L. (2006). Cognitive distortions in child molesters: A reexamination of key theories and research. Clinical Psychology Review, 26, 10001019.
Gannon, T. A., & Rose, M. R. (2008). Female child sexual offenders: Towards integrating
theory and practice. Aggression and Violent Behavior, 13, 442-461.
Gannon, T. A., Keown, K., & Polaschek, D. L. (2007). Increasing honest responding on
cognitive distortions in child molesters: The bogus pipeline revisited. Sexual Abuse: A
Journal of Research and Treatment, 19, 5-22.
Gawronski, B., & Bodenhausen, G. V. (2006) Associative and propositional processes in
evaluation: An integrative review of implicit and explicit attitude change.
Psychological Bulletin, 132, 692-731
Gee, D., & Belofastov, A. (2007). Profiling sexual fantasy: Fantasy in sexual offending and
the implications for criminal profiling. In R.N. Kocsis (Ed.), Criminal
profiling: International theory, research and practice (pp. 49-71). Totowa, NJ:
Humana Press.
Gee, D. G., Devilly, G. J., & Ward, T. (2004). The content of sexual fantasies for sexual
offenders. Sexual Abuse: A Journal of Research and Treatment, 16, 315331.
Gee, D. G., Ward, T., & Eccleston, L. (2003). The function of sexual fantasies for sexual
offenders: A preliminary model. Behavior Change, 20, 4460
194
Gee, D. G., Ward, T., Belofastov, A., & Beech, A. R. (2006). The structural properties of
sexual fantasies for sexual offenders: A preliminary model. Journal of Sexual
Aggression, 12, 213226.
Giannangelo, S. (1996). The psychopathology of serial murder: A theory of violence.
Westport, CT: Praeger
Gilbert, D. T., & Hixon, J. G. (1991). The trouble of thinking: Activation and application of
stereotypic beliefs. Journal of Personality and Social Psychology, 60, 509-517.
Gold, S. R., & Gold, R. G. (1991). Gender differences in first sexual fantasies. Journal of Sex
Education and Therapy, 17, 207-216.
Goldstein, M. J., & Kant, H. S. (1973). Pornography and sexual deviance: A report of the
legal and behavioral institute, Beverly Hills, California. Berkeley: University of
California Press.
Gratzer, T., & Bradford, J. M. W. (1995). Offender and offense characteristics of sexual
sadists: a comparative study. Journal of Forensic Sciences, 40, 450-455
Gray, N. S., & Snowden, R. J. (2009). The Implicit Association Test as a measure of sexual
interest. In D. Thornton & R. Laws (Ed.), Cognitive approaches to the assessment of
sexual interest in sexual offenders. Chichester, UK: Wiley
Gray, N. S., Brown, A. S., MacCulloch, M. J., Smith, J., & Snowden, R. J. (2005). An
implicit test of the association between children and sex in pedophiles. Journal of
Abnormal Psychology, 114, 304308.
Gray, N. S., Watt, A., Hassan, S., & Macculloch, M. J. (2003). Behavioral indicators of
sadistic sexual murder predict the presence of sadistic sexual fantasy in a normative
sample. Journal of Interpersonal Violence, 18, 1018-1034.
Greenwald, A. G., & Farnham, S. D. (2000). Using the Implicit Association Test to measure
self-esteem and self-concept. Journal of Personality and Social Psychology, 79,
10221038.
Greenwald, A. G., McGhee, J. L., & Schwartz, J. L. (1998). Measuring individual differences
in implicit cognition: The Implicit Association Test. Journal of Personality and Social
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the
Implicit Association Test: I. An improved scoring algorithm. Journal of Personality
and Social Psychology, 85, 197-216.
Gunter, R. W., & Bodner, G. E. (2008). How eye movements affect unpleasant memories:
support for a working-memory account. Behaviour Research and Therapy, 46, 913931.
195
Hackmann, A., Bennett-Levy, J., & Holmes, E. (2011). Oxford guide to imagery in cognitive
therapy. Oxford: Oxford University Press.
Hagenaars, M. A., & Krans, J. (2011). Trait and state dissociation in the prediction of
intrusive images. International Journal of Cognitive Therapy, 4, 145-153.
Haines, E. L., & Kray, L. J. (2005). Selfpower associations: the possession of power impacts
women's selfconcepts. European Journal of Social Psychology,35(5), 643-662.
Hall, G. C. N. (1996). Theory-based assessment, treatment, and prevention of sexual
aggression. New York: Oxford University Press
Hanson, R. K., & Harris, A. J. (2001). A structured approach to evaluating change among
sexual offenders. Sexual Abuse: A Journal of Research and Treatment, 13, 105-
122.
Hanson, R. K. & Morton-Bourgon, K. E. (2005). The characteristics of persistent sexual

offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical
Psychology, 73, 1154-1163.
Hare, R. D.(1991). The Hare Psychopathy Checklist-Revised. Toronto: Multi-Health Systems
Hariton, E. B. (1973). The sexual fantasies of women. Psychology Today, 6, 39-44.
Hazelwood, R.R., & Michaud, S.G. (2001). Dark dreams: Sexual violence, homicide,
and the criminal mind. New York: St. Martins Press
Helmus, L., Thornton, D., Hanson, R. K., & Babchishin, K. M. (2012). Improving the
predictive accuracy of Static-99 and Static-2002 with older sex offenders revised age
weights. Sexual Abuse: A Journal of Research and Treatment, 24, 64-101.
Hempel, I. S., Buck, N. M. L., Goethals, K. R., & van Marle, H. J. C. (2012). Unraveling
sexual associations in contact and noncontact child sex offenders using the single
categoryImplicit Association Test. Sexual Abuse: A Journal of Research and
Treatment, 2, doi:10.1177/1079063212464660
Hornsveld, H. K., Houtveen, J. H., Vroomen, M., Aalbers, I. K. D., Aalbers, D., &
van den Hout, M. A. (2011). Evaluating the effect of eye movements on positive
memories such as those used in resource development and installation. Journal of
EMDR Practice and Research, 5, 146-155.
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural Equation
Modeling: A Multidisciplinary Journal, 6, 1-55.
Huesmann, L. R. (1998). The role of social information processing and cognitive schemas in
the acquisition and maintenance of habitual aggressive behavior. In R. G. Geen & E.
Donnerstein (Eds.), Human aggression: Theories, research, and implications for
policy (pp. 73109). New York: Academic Press.
196
Ievers-Landis, C. E., Burant, C. J., & Hazen, R. (2011). The concept of bootstrapping of
structural equation models with smaller samples: an illustration using mealtime rituals
in diabetes management. Journal of Developmental and Behavioral Pediatrics, 32,
619-626.
Janssen, E., Everaerd, W., Spiering, M., & Janssen, J. (2000). Automatic processes and the
appraisal of sexual stimuli: Toward an information processing model of sexual
arousal. Journal of Sex Research, 37, 823.
Johnston, L., Ward, T., & Hudson, S. M. (1997). Deviant sexual thoughts: Mental control and
the treatment of sexual offenders. Journal of Sex Research, 34, 121-30.
Jones, J. C., & Barlow, D. H. (1990). Self-reported frequency of sexual urges, fantasies, and
masturbatory fantasies in heterosexual males and females. Archives of Sexual
Behavior, 19, 269-279.
Kahneman, D. (2011). Thinking, fast and slow. New York: Farrar, Straus and Giroux.
Kalmus, E., & Beech, A. R. (2005). Forensic assessment of sexual interest: A
review. Aggression and Violent Behavior, 10, 193-217.
Karpinski, A., & Steinman, R. B. (2006). The single category implicit association test as a
measure of implicit social cognition. Journal of Personality and Social Psychology,
91, 16-32.
Kavanagh, D. J., Andrade, J., & May, J. (2005). Imaginary relish and exquisite torture: The
elaborated intrusion theory of desire. Psychological Review, 112, 446-467.
Kavanagh, D. J., Freese, S., Andrade, J., & May, J. (2001). Effects of visuospatial tasks on
desensitization to emotive memories. British Journal of Clinical Psychology, 40,
267-280.
Keller, P. E. (2012). Mental imagery in music performance: underlying mechanisms and
potential benefits. Annals of the New York Academy of Sciences, 1252, 206-213.
Kemps, E., Tiggemann, M., Woods, D., & Soekov, B. (2004). Reduction of food cravings
through concurrent visuospatial processing. International Journal of Eating
Disorders, 36, 31-40.
Kihlstrom, J. F., Glisky, M. L., & Angiulo, M. J. (1994). Dissociative tendencies and
dissociative disorders. Journal of Abnormal Psychology, 103, 117-124.
Kline, P. (2000). The handbook of psychological testing (2nd ed.). London: Routledge.
Kline, R. B. (2004). Principles and practice of structural equation modeling. New York:
Guilford.
Kloess, J. (2013, July). Solicitation offenders: Behaviour and motivations in interactions with
197
victims via internet communications. Paper presented at the 28th Psychology Affairs
Group Annual Conference, Lancaster, UK.
Knafo, D., & Jaffe, Y. (1984). Sexual fantasizing in males and females. Journal of Research
in Personality, 18, 451462.
Knight, R. A., & SimsKnight, J. E. (2003). The developmental antecedents of sexual
coercion against women: Testing alternative hypotheses with structural equation
modeling. Annals of the New York Academy of Sciences, 989, 72-85.
Kosslyn S. M. (1994). Image and brain: The resolution of the imagery debate. Cambridge,
MA: MIT Press
Kosslyn, S. M. (1995). Mental imagery. In S. M. Kosslyn & D. N. Osherson (Eds.), Visual
cognition: An invitation to cognitive science (Vol. 2, pp. 267-296). Cambridge, MA:
MIT Press.
Kremsdorf, R. B., Holmen, M. L., & Laws, D. R. (1980). Orgasmic reconditioning without
deviant imagery: A case report with a pedophile. Behaviour Research and
Therapy, 18, 203-207.
Langevin, R., Handy, L., Paitich, D., & Russon, A. (1985). Appendix A: A new version of the
Clarke sex history questionnaire for males. Erotic preference, gender identity and
aggression in men: New research studies (pp. 287-306). Hillsdale, NJ: Erlbaum.
Langevin, R., Lang, R.A., & Curnoe, S. (1998). The prevalence of sex offenders with deviant
fantasies. Journal of Interpersonal Violence, 13, 315-327.
Laws, D. R., & Gress, C. L. Z. (2004). Seeing things differently: The viewing time alternative
to penile plethysmography. Legal and Criminological Psychology, 9, 183-196.
Leclerc, B., Beauregard, E., & Proulx, J. (2008). Modus operandi and situational aspects in
adolescent sexual offenses against children: A further examination. International
Journal of Offender Therapy and Comparative Criminology, 52, 46-61.
Leclerc, B., Carpentier, J., & Proulx, J. (2006). Strategies adopted by sexual offenders to
involve children in sexual activity. In R. Wortley & S. Smallbone (Eds.), Situational
prevention of child sexual abuse (Crime Prevention Studies Vol. 19, pp. 251-270).
Monsey, NY: Criminal Justice Press.
Lee, C. W., & Cuijpers, P. (2013). A meta-analysis of the contribution of eye movements in
processing emotional memories. Journal of Behavior Therapy and Experimental
Psychiatry, 44, 231-239.
Leitenberg, H., & Henning, K. (1995). Sexual fantasy. Psychological Bulletin, 117, 469496.
Logie, R. H. (1986). Visuo-spatial processing in working memory. Quarterly Journal of
Experimental Psychology, 38, 229247.
198
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: A Journal of Research and Treatment, 25, 370-395.

Looman, J. (1995). Sexual fantasies of child molesters. Canadian Journal of Behavioural
Science, 27, 321-332.
Lussier, P., Proulx, J., & McKibben, A. (2001). Personality characteristics and adaptive
strategies to cope with negative emotional states and deviant sexual fantasies in
sexual aggressors. International Journal of Offender Therapy and Comparative
Criminology, 45, 159170.
Lynn, S. J., & Rhue, J. W. (1988). Fantasy proneness: Hypnosis, developmental antecedents
and psychopathology. American Psychologist, 43, 35-44.
MacCulloch, M., Gray, N., & Watt, A. (2000). Brittain's Sadistic Murderer Syndrome
reconsidered: an associative account of the aetiology of sadistic sexual fantasy. The
Journal of Forensic Psychiatry, 11, 401-418.
MacCulloch, M. J., Snowden, P. R., Wood, P. J. W., & Mills, H. E. (1983). Sadistic
fantasy, sadistic behavior, and offending. The British Journal of Psychiatry, 143, 20
29.
Maile, C., & Jeglic, E. (2009, March). Differential prevalence rates of deviant sexual fantasy
in an undergraduate sample. Paper presented at the annual meeting of the American
Psychology - Law Society, TBA, San Antonio, TX.
Malamuth, N. M. (1981). Rape fantasies as a function of exposure to violent sexual
stimuli. Archives of Sexual Behavior, 10, 33-47.
Mann, R. E., & Beech, A. R. (2003). Cognitive distortions, schemas, and implicit theories. In
T. Ward, D. R. Laws, & S. M. Hudson (Eds.), Sexual deviance: Issues and
controversies (pp. 135153). Thousand Oaks, CA: Sage.
Mann, R.E., Hanson, R. K, & Thornton, D. (2010). Assessing risk for sexual recidivism:
Some proposals on the nature of psychologically meaningful risk factors. Sexual
Abuse: A Journal of Research and Treatment, 22, 191-217.
Marks, D. F. (1973). Visual imagery differences in the recall of pictures. British journal of
Marshall, W. L. (1973). The modification of sexual fantasies: A combined treatment
approach to the reduction of deviant sexual arousal. Behaviour Research and Therapy,
11, 557-564.
Marshall, W. L. (1996). Assessment, treatment, and theorizing about sex offenders:
199
Developments over the past 20 years and future directions. Criminal Justice and
Behavior, 23, 162199.
Marshall, W. L. (2006). Ammonia aversion with an exhibitionist: A case study. Clinical
Case Studies, 5, 15-24.
Marshall, W. L., & Barbaree, H. E. (1990). An integrated theory of the etiology of sexual
offending. In W. L. Marshall, D. R. Laws, & H. E. Barbaree (Eds.), Handbook of
sexual assault: Issues, theories, and treatment of the offender (pp. 257275). New
York: Plenum Press.
Marshall, W. L., & Marshall, L. E. (2000). The origins of sexual offending. Trauma,
Violence, and Abuse, 1, 250-263.
Marshall, W. L., Barbaree, H. E., & Eccles, A. (1991). Early onset and deviant sexuality in
child molesters. Journal of Interpersonal Violence, 6, 323-336.
Marshall, W. L., OBrien, M. D., & Marshall, L. E. (2009). Modifying sexual preferences. In
A. R. Beech, L. Craig, & K. D. Browne (Eds), Assessment and treatment of sex
offenders: a handbook (pp.311-327). West Sussex, Chichester: Wiley.
Maxfield, L., Melnyk, W. T., & Hayman, C. A. G. (2008). A working memory explanation
for the effects of eye movements in EMDR. Journal of EMDR Practice and
Research, 2, 247-261.
McClelland, A., Kemps, E., & Tiggemann, M. (2006). Reduction of vividness and associated
craving in personalized food imagery. Journal of Clinical Psychology, 62, 355-365.
McDonald, J.H. (2009). Handbook of biological statistics (2nd ed.). Baltimore, Maryland:
Sparky House Publishing,
McGuire, R. J., Carlisle, J. M., & Young, B. G. (1965). Sexual deviation as conditional
behavior: A hypothesis. Behaviour Research and Therapy, 2, 185-190.
McKibben, A., Proulx, J., & Lusignan, R. (1994). Relationships between conflict, affect
and deviant sexual behaviors in rapists and paedophiles. Behavior Research and
Therapy, 32, 571575.
McPhail, I. V., Sewell, R., Nunes, K. L., & Hermann, C. (2010). Assessing emotional
incongruence with children in sexual offenders against children. Paper presented at
the 29th Association for Treatment of Sexual Abusers Annual Conference, Phoenix,
AZ.
Merckelbach, H., Hardy, S., & Giesbrecht, T. (2005). Dissociation and fantasy proneness in
psychiatric patients: a preliminary study. Comprehensive Psychiatry, 46, 181-185.
Merckelbach, H., Horselenberg, R., & Muris, P. (2001). The Creative Experiences
200
Questionnaire (CEQ): A brief self-report measure of fantasy proneness. Personality

and Individual Differences, 31, 987-995.
Mihailides, S., Devilly, G. J., & Ward, T. (2004). Implicit cognitive distortions and sexual
offending. Sexual Abuse: A Journal of Research and Treatment, 16, 333350.
Morris, S. B., & DeShon, R. P. (2002). Combining effect size estimates in meta-analysis with
repeated measures and independent-groups designs. Psychological Methods, 7, 105
125.
Moscovitch, M. (1994). Cognitive resources and dual-task interference effects at retrieval in
normal people: The role of the frontal lobes and medial temporal cortex.
Neuropsychology, 8, 524-534.
Mosher, D. L., & Anderson, R. D. (1986). Macho personality, sexual aggression, and
reactions to guided imagery of realistic rape. Journal of Research in
Personality, 20, 77-94.
Mosher, D. L., & Sirkin, M. (1984). Measuring a macho personality constellation. Journal of
Research in Personality, 18, 150-163.
Moulton, S. T., & Kosslyn, S. M. (2009). Imagining predictions: mental imagery as mental
emulation. Philosophical Transactions of the Royal Society B: Biological
Sciences, 364, 1273-1280.
Nock, M. K., Park, J. M., Finn, C. T., Deliberto, T. L., Dour, H. J., & Banaji, M. R. (2010).
Measuring the suicidal mind: Implicit cognition predicts suicidal
behavior. Psychological Science, 21, 511-517.
Nosek, B. A., & Banaji, M. R. (2001). The Go/No-go association task. Social Cognition, 19,
625666.
Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2005). Understanding and using the
Implicit Association Test: II. Method variables and construct validity.Personality and
Social Psychology Bulletin, 31, 166-180.
OCiardha, C. (2010). The use of implicit cognitive measures in the assessment of sexual
offenders (Unpublished dissertation).
Ciardha, C. (2011). A theoretical framework for understanding deviant sexual interest and
cognitive distortions as overlapping constructs contributing to sexual offending
against children. Aggression and Violent Behavior, 16, 493-502.
Ciardha, C., & Gormley, M. (2012). Measuring sexual interest using a pictorial modified
Stroop task, a pictorial Implicit Association Test and a Choice Reaction Time
task. Journal of Sexual Aggression, 19, 158-170.
ODonohue, W., Letourneau, E. J., & Dowling, H. (1997). Development and preliminary
201
validation of a paraphilic sexual fantasy questionnaire. Sexual Abuse: A Journal of

Research and Treatment, 9, 167178.
Palk, G. R., & O'Gorman, J. (2004). Habituation of deviant sexual arousal in sex offenders:
The role of cognitive processes. International Journal of Forensic Psychology, 1, 5870.
Pawlak, A. E., Boulet, J. R., & Bradford, J. M. (1991). Discriminant analysis of a sexualfunctioning inventory with intrafamilial and extrafamilial child molesters. Archives of
Sexual Behavior, 20, 27-34.
Pedhazur, E., & Schmelkin. L. P., (1991). Measurement, design, and analysis: An integrated
approach. Hillsdale, NJ: Erlbaum.
Pekala, R. J., Angelini, F., & Kumar, V. K. (2001). The importance of fantasyproneness in
dissociation: a replication. Contemporary Hypnosis, 18, 204-214.
Pelletier, L. A., & Herold, E. S. (1988). The relationship of age, sex guilt, and sexual
experience with female sexual fantasies. Journal of Sex Research, 24, 250-256.
Pithers, W. D. (1990). Relapse prevention with sexual aggressors: A method for maintaining
therapeutic gain and enhancing external supervision. In W. L. Marshall, D. R. Laws,
& H. E. Barbaree (Eds.), Handbook of sexual assault: Issues theories and treatment of
the offender (pp. 343-361). New York: Plenum Press.
Pithers, W. D., Kashima, K. M., Cummings, G. S., & Beal, L. S. (1988). Relapse prevention:
A method of enhancing change in sex offender. In A. C. Salter (Ed.), Treating child
sex offenders and victims: A practical guide (pp. 131170). London: Sage.
Plaud, J. J., & Bigwood, S. J. (1997). The relationship of male selfreport of rape supportive
attitudes, sexual fantasy, social desirability and physiological arousal to sexually
coercive stimuli. Journal of Clinical Psychology, 53, 935-942.
Prentky, R. A., & Burgess, A. W. (1991). Hypothetical biological substrates of a fantasybased drive mechanism for repetitive sexual aggression. Rape and sexual assault III,
235-256.
Prentky, R. A., Burgess, A. W., Rokous, F., Lee, A., Hartman, C., Ressler, R., & Douglas, J.
(1989). The presumptive role of fantasy in serial sexual homicide. American Journal
of Psychiatry, 146, 887891.
Proulx, J., McKibben, A., & Lusignan, R. (1996). Relationships between affective
components and sexual behaviours in sexual aggressors. Sexual Abuse: A Journal of
Quinsey, V. L., Chaplin, T. C., & Carrigan, W. F. (1979). Sexual preferences among
incestuous and nonincestuous child molesters. Behavior Therapy, 10, 562-565.
202
Ranganath, K. A., Smith, C. T., & Nosek, B. A. (2008). Distinguishing automatic and
controlled components of attitudes from direct and indirect measurement
methods. Journal of Experimental Social Psychology, 44, 386-396.
Renaud, C., & Byers, E. S. (1999). Exploring the frequency, diversity and content of
university students positive and negative sexual cognitions. Canadian Journal of
Human Sexuality, 8, 17-30.
Renaud, C. A., & Byers, E. S. (2001). Positive and negative sexual cognitions: Subjective
Experience and relationships to sexual adjustment. Journal of Sex Research, 38, 252
262.
Renaud, C. A., & Byers, E. S. (2005). Relationship between sexual violence and positive and
negative cognitions of sexual dominance. Sex Roles, 53, 253-260.
Rhue, J. W., & Lynn, S. J. (1987). Fantasy proneness: Developmental antecedents. Journal of
Personality, 55, 121-137.
Ricci, R. J., Clayton, C. A., & Shapiro, F. (2006). Some effects of EMDR on previously
abused child molesters: Theoretical reviews and preliminary findings .The Journal of
Forensic Psychiatry and Psychology, 17, 538-562.
Robertiello, G., & Terry, K. J. (2007). Can we profile sex offenders? A review of sex
offender typologies. Aggression and Violent Behavior, 12, 508-518.
Roche, B. & Barnes, D. (1998). The experimental analysis of human sexual arousal: Some
recent developments. The Behavior Analyst, 21, 37-52.
Rosen, V. M., & Engle, R. W. (1998). Working memory capacity and suppression. Journal of
Memory and Language, 39, 418-436.
Russell, K., Sturgeon, V. H., Miner, M. H., & Nelson, C. (1989). Determinants of the
abstinence violation effect in sexual fantasies. In D. Laws (Ed.), Relapse prevention
with sex offenders (pp. 63-72). New York: Guilford.
Rydell, R. J., & McConnell, A. R. (2006). Understanding implicit and explicit attitude
change: A systems of reasoning analysis. Journal of Personality and Social
Saucier, G., & Skrzypiska, K. (2006). Spiritual but not religious? Evidence for two
independent dispositions. Journal of Personality, 74, 1257-1292.
Schacter, D. L., Addis, D. R., & Buckner, R. L. (2007). Remembering the past to imagine the
future: the prospective brain. Nature Reviews Neuroscience, 8, 657661.
Schacter, D. L., Addis, D. R., & Buckner, R. L. (2008). Episodic simulation of future events.
Annals of the New York Academy of Sciences, 1124, 39-60.
Schaefer, A., & Philippot, P. (2005). Selective effects of emotion on the phenomenal
203
characteristics of autobiographical memories. Memory, 13, 148-160.

Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information
processing: I. Detection, search, and attention. Psychological Review, 84, 1-66.
Schreiber, J. B., Nora, A., Stage, F. K., Barlow, E. A., & King, J. (2006). Reporting structural
equation modeling and confirmatory factor analysis results: A review. The Journal of
Educational Research, 99, 323-338.
Shapiro, F. (1995). Eye movement desensitization and reprocessing: Basic principles,
protocols, and procedures. New York: Guildford.
Sheldon, K., & Howitt, D. (2008). Sexual fantasy in paedophile offenders: Can any model
explain satisfactorily new findings from a study of Internet and contact sexual
offenders? Legal and Criminological Psychology, 13, 137-158.
Singer, J. L. (1966). Daydreaming: An introduction to the experimental study of inner
experience. New York: Random House.
Skinner, B. F. (1953). Science and human behavior. New York: Macmillan
Smeaton, G., & Byrne, D. (1987). The effects of R-rated violence and erotica, individual
differences, and victim characteristics on acquaintance rape proclivity. Journal of
Research in Personality, 21, 171-184.
Smeets, M. A., Dijs, M. W., Pervan, I., Engelhard, I. M., & Van den Hout, M. A. (2012).
Time-course of eye movement-related decrease in vividness and emotionality of
unpleasant autobiographical memories. Memory, 20, 346-357.
Smith, E. R., & DeCoster, J. (2000). Dualprocess models in social and cognitive
psychology: Conceptual integration and links to underlying memory systems.
Personality and Social Psychology Review, 4, 108131.
Smith, D., & Over, R. (1987). Male sexual arousal as a function of the content and the
vividness of erotic fantasy. Psychophysiology, 24, 334-339.
Snowden, R. J., & Gray, N. S. (2013). Implicit sexual associations in heterosexual and
homosexual women and men. Archives of Sexual Behavior, 42, 475-485.
Snowden, R. J., Craig, R. L., & Gray, N. S. (2011). Indirect behavioral measures of cognition
among sexual offenders. Journal of Sex Research, 48, 192-217.
Snowden, R. J., Gray, N. S., Brown, A. S., & Power, S. (2007). Measuring implicit
cognitions in sexual offenders using the Go-No Go Task. Unpublished manuscript.
Snowden, R. J., Wichter, J., & Gray, N. S. (2008). Implicit and explicit measurements of
sexual preference in gay and heterosexual men: A comparison of priming techniques
and the implicit association task. Archives of Sexual Behavior, 37, 558-565.
Steffens, M. C., Yundina, E., & Panning, M. (2008). Automatic associations with erotic in
204
child sexual offenders: Identifying those in danger of re-offence. Sexual Offender

Treatment, 3. Retrieved from http://www.sexual-offender-treatment.org/22008_03.98.html?&spnCategory=529&spnDomain=17&spnContent=28&spnID=395
23
Steiger, J. H. (2007). Understanding the limitations of global fit assessment in structural
equation modeling. Personality and Individual Differences, 42, 893-898.
Stinson, J. D., & Becker, J. V. (2012). Treating Sex Offenders: An Evidence-Based Manual.
New York: Guilford.
Szpunar, K. K. (2010). Episodic future thought: An emerging concept. Perspectives on
Psychological Science, 5, 142-162.
Szpunar, K. K., & McDermott, K. B. (2008). Episodic future thought and its relationship to
remembering: Evidence from rating of subjective experience. Consciousness and
Cognition, 17, 330-334.
Teachman, B. A. (2007). Evaluating implicit spider fear associations using the go/no-go
association task. Journal of Behavior Therapy and Experimental Psychiatry, 38, 156
167.
Tellegen, A., & Atkinson, G. (1974). Openness to absorbing and self-altering experiences
(absorption): A trait related to hypnotic susceptibility. Journal of Abnormal
Thornton, D. Thoughts and Fantasies Questionnaire. (Unpublished manuscript).
Thornton, D. (2002). Constructing and testing a framework for dynamic risk

assessment. Sexual Abuse: A Journal of Research and Treatment, 14, 139-153.
Tidefors, I., & Kordon, S. (2009). Rape and murder: Acts of desire or disgust? Narratives by
rapists and sexual murderers during group-psychotherapy. Sexual Offender
Treatment, 4, Retrieved from: http://www.sexual-offender-treatment.org/22009_03.html
Toates, F. (2009). An integrative theoretical framework for understanding sexual motivation,
arousal, and behavior. Journal of Sex Research, 46, 168193.
Turner-Moore, T., Waterman, M.G. (2011, October) Contextualising sexual offenders
sexual fantasies: When, how, why and a little bit of what? Paper presented at The 30th
Annual Research and Treatment Conference of the Association for Treatment of
Sexual Abusers (ATSA). Toronto, Canada.
Turner-Moore, T., Smith, P., & Waterman, M. (2010, October). The sexual thoughts of adult
205
male sexual offenders, non-sexual offenders, and non-offenders. Poster presented at

the 29th Association for Treatment of Sexual Abusers Annual Conference, Phoenix,
AZ, USA.
Uhlmann, E., & Swanson, J. (2004). Exposure to violent video games increases automatic
aggressiveness. Journal of adolescence, 27, 41-52.
van den Hout, M. A., & Engelhard, I. M. (2012). How does EMDR work? Journal of
Experimental Psychopathology, 3, 724-738
van den Hout, M., Muris, P., Salemink, E., & Kindt, M. (2001). Autobiographical memories
become less vivid and emotional after eye movements. British Journal
of Clinical Psychology, 40, 121-130.
van den Hout, M., Engelhard, I. M., Rijkeboer, M. M., Koekebakker, J., Hornsveld, H., Leer,
A.,... Akse, N. (2011). EMDR: Eye-movements superior to beeps
in taxing working memory and reducing vividness of recollections. Behavior
Research and Therapy, 49, 92-98.
van den Hout, M. A., Engelhard, I. M., Beetsma, D., Slofstra, C., Hornsveld, H., Houtveen,
J., & Leer, A. (2011). EMDR and mindfulness. Eye movements and attentional
breathing tax working memory and reduce vividness and emotionality of aversive
ideation. Journal of Behavior Therapy and Experimental Psychiatry, 42, 423-431.
Vanhoeck, K., Van Daele, E., & Gykiere, K. (2011). Fantasy management in sex offender
treatment. Sexual Offender Treatment, 6, Retrieved at: http://www.sexual-offendertreatment.org/94.html
Violato, C., & Hecker, K. G. (2007). How to use Structural Equation Modeling in medical
education research: A brief guide. Teaching and Learning in Medicine, 19, 362-371.
Ward, T. (2000). Sexual offenders' cognitive distortions as implicit theories. Aggression and
Violent Behavior, 5, 491507.
Ward, T., & Beech, A. R. (2006). An integrated theory of sexual offending. Aggression and
Violent Behavior, 11, 4463.
Ward, T., & Casey, A. (2010). Extending the mind into the world: A new theory of cognitive
distortions in sex offenders. Aggression and Violent Behavior, 15, 4958.
Ward, T., & Hudson, S. M. (1998). The construction and development of theory in the sexual
offending area: A metatheoretical framework. Sexual Abuse: A Journal of Research
and Treatment, 10, 47-63.
Ward, T., & Hudson, S. M. (2000). Sexual offenders implicit planning: A conceptual
model. Sexual Abuse: A Journal of Research and Treatment, 12, 189-202.
Ward, T., & Stewart, C. (2003). Criminogenic needs and human needs: A theoretical
206
model. Psychology, Crime, and Law, 9, 125-143.

Ward, T., Hudson, S. M., & Keenan, T. (1998). A self-regulation model of the sexual offense
process. Sexual Abuse: A Journal of Research and Treatment, 10, 141-157.
Ward, T., Laws, D. R., & Hudson, S. M. (2003). Sexual deviance: Issues and controversies.
London: Sage.
Ward, T., Mann, R. E., & Gannon, T. A. (2007). The Good Lives Model of offender
rehabilitation: Clinical implications. Aggression and Violent Behavior, 12, 87 107.
Ward, T., Polaschek., D. L. L., & Beech, A. R. (2006). Theories of sexual offending.
Chichester: Wiley.
Webster, S. D., Bowers, L. E., Mann, R. E., & Marshall, W. L. (2005). Developing empathy
in sexual offenders: The value of offence re-enactments. Sexual Abuse: A Journal of
Webster, S. D., Mann, R. E., Thornton, D., & Wakeling, H. C. (2007). Further validation of
the short selfesteem scale with sexual offenders. Legal and Criminological
Wechsler D. WAIS-III Manual (1997). New York: The Psychological Corporation.
Wechsler D. WAIS-IV Manual (2008). New York: The Psychological Corporation.
Wetterneck, C. T., Smith, A. H., Burgess, A. J., & Hart, J. M. (2011). Distress from sexual
thoughts: Do religiosity, emotions, and thought appraisal matter? Journal of
Cognitive Psychotherapy, 25, 189-202.
Williams, K. M., Cooper, B.S., Howell, T. M., Yuille, J. C., & Paulhus, D. L. (2009).
Inferring sexually deviant behavior from corresponding fantasies: The role of
personality and pornography consumption. Criminal Justice and Behavior, 36, 198222.
Wilson, S. C. and Barber, T. X. (1981). Vivid fantasy and hallucinatory abilities in life
Histories of excellent hypnotic subjects (`somnambules`): Preliminary report with
female subjects. In E. Klinger (Ed.), Imagery: Vol. 2. Concepts, results, and
applications (pp. 133-149). NewYork: Plenum.
Wilson, G. (1978). The secrets of sexual fantasy. London: J.M. Dent & Sons.
Wilson, G. D., & Lang, R. J. (1981). Sex differences in sexual fantasy patterns. Personality
and Individual Differences, 2, 343-346.
Wilson, M. (2002). Six views of embodied cognition. Psychonomic Bulletin and Review, 9,
625-636.
Woodworth, M., Freimuth, T., Hutton, E. L., Carpenter, T., Agar, A. D., & Logan, M.
(2013), High-risk sexual offenders: An examination of sexual fantasy, sexual
207
paraphilia, psychopathy, and offence characteristics. International Journal of Law

and Psychiatry, 35,144-156.
Zurbriggen, E. L., & Yost, M. R. (2004). Power, desire, and pleasure in sexual fantasies.
Journal of Sex Research, 41, 288300.
208
Appendices
Appendix List
Appendix
A
Title
Presentations, awards, and forthcoming publications arising
from the work in this thesis
Page
194
196
197
D
E
F
G
199
200
202
208
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W
X
The Wilson Sex Fantasy Questionnaire

Thoughts and Fantasies Questionnaire
Ethical approval from the University of Birminghams Ethical
Review Committee for the study in Chapter 4
Creative Experiences Questionnaire
Vividness of Visual Imagery Questionnaire
Dissociative Experiences Scale II
Women are Deceitful Scale
Calloused Sex Attitudes Towards Women subscale
Gray et al.'s (2003) Sexual Fantasy Questionnaire
Adapted Wilson Sex Fantasy Questionnaire (WSFQ)
Guide for the Adapted Wilson Sex Fantasy Questionnaire
Visual analogue scales for emotionality, vividness, and
arousability
Stimuli and example screen for the Implicit Association Task
in Chapter 6
Measure of self-reported powerfulness
Short Self-esteem Scale
Instructions for the mental imagery tasks in Chapter 6
Example screens for the experimental GNAT in Chapter 7
Stimuli for the Go/No-Go Association Task
210
212
215
219
220
222
224
226
231
233
234
235
236
237
238
240
241
Appendices B-X are not available in the access copy of this thesis
209
Appendix A: Presentations, awards, and forthcoming publications arising from the

work in this thesis
Conference Presentations
Bartels, R. M. (2013, June). Investigating what underlies the dynamic risk factor of sexual
deviancy. Presented at the at the Division of Forensic Psychology Annual Conference,
Belfast UK.
Bartels, R. M. (2013, February). Untangling the cognitive associations of child abusers.
Presented at the 2103 European Winterschool for Forensic Psychology, Seeon,
Germany.
Bartels, R. M., Beech, A. R., & Harkins, L. (2012, September). Using the Implicit Association
Task with child sex offenders: Current issues and future directions. Presented at the
Annual Conference of the National Organisation for the Treatment of Abusers,
Edinburgh, UK.
Bartels, R. M., Beech, A. R., Harkins, L., & Dixon, L. (2012, June). Investigating offence
supportive cognition in male sex offenders: What we know, what we need to do.
Presented at the Division of Forensic Psychology Annual Conference, Cardiff, UK.
Bartels, R. M., Beech, A. R., & Harkins, L. (2011, November). New directions for indirect
measurement? Application of the Sorting Paired Features task and Go/No-Go
Association Task. Presented at the 30th Annual Conference of the Association for the
Treatment of Sexual Abusers, Toronto, Canada.
Poster Presentations
Bartels, R.M., Beech, A. R., & Harkins, L. (July, 2013). Through the eyes of the beholder:
Using EMDR techniques with sexual fantasies. Poster presented at the 28th Annual
psyPAG Conference, Lancaster, UK.
Bartels, R.M., Beech, A. R., & Harkins, L. (2012, October). Can Child Abusers Deviant
Sexual Fantasies Be Measured Indirectly? Poster presented at the 31st Annual
Conference of the Association for the Treatment of Sexual Abusers, Denver,
Colorado, USA.
210
Bartels, R. M., Beech, A. R., & Harkins, L. (2012, July). Towards a dual-process model of
deviant sexual fantasy. Poster presented at the 27th Annual psyPAG Conference,
Newcastle, UK.
Bartels, R. M., Beech, A. R., & Harkins, L. (2012, April). Applying the Sorting Paired
Features task to the study of sex offender cognition. Poster presented at the Annual
BPS Conference, London, UK.
Awards
Best Student Poster Prize (2012, April) for a poster entitled: Applying the Sorting Paired
Features task to the study of sex offender cognition. Awarded at the Annual BPS
Conference, London, UK.
Graduate Research Award (2012, October) for the paper entitled: The Indirect Assessment of
Deviant Sexual Fantasy in Child Molesters. Awarded at the 31st Annual Conference
of the Association for the Treatment of Sexual Abusers, Denver, Colorado, USA.
Prize for Best Poster (2013, July) for a poster entitled: Through the eyes of the beholder:
Using EMDR techniques with sexual fantasies. Awarded at the 28th Annual
psyPAG Conference, Lancaster, UK.
Publications
Bartels, R. M, & Beech, A. R., Harkins, L., & Thornton, D. (under review). Assessing deviant
sexual fantasy in child molesters using an indirect assessment approach.
Bartels, R. M., Beech, & Harkins, L. (under review). Sexual fantasies become less vivid,
emotional, and arousing after making bilateral eye-movements: Implications for
the treatment of deviant sexual fantasies in sexual offenders
Bartels, R. M. & Beech, A. R. (in preparation). Theories of deviant sexual fantasy. In T. Ward
& A. R. Beech (eds.), Theories of sexual offending.
211

Bartels 14 PHD

Uploaded by

Copyright:

Available Formats

Bartels 14 PHD

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bartels 14 PHD

Uploaded by

Copyright:

Available Formats

UNDERSTANDING THE COGNITIVE PROCESSES

ASSOCIATED WITH SEXUAL FANTASIES:

Ross Matthew Bartels

A thesis submitted to the College of Life and Environmental Sciences of the

University of Birmingham Research Archive

Chapter 1: Deviant sexual fantasies: What is currently known

1.2 Definition of deviant sexual fantasy

1.3 Prevalence of deviant sexual fantasies

1.4 Effects and functions of deviant sexual fantasy

1.5 Theories of deviant sexual fantasy

Chapter 2: The Dual-Process Model of Deviant Sexual Thinking (DPM-ST)

2.2 A dual-process approach

2.3 Remembering and imagining episodic events

2.4 Outlining the DPM-ST

Chapter 3: Investigating Sex-related Associations in Child Abusers

Chapter 4: Investigating the Factors that Influence the Use of Deviant

Chapter 5: Testing Whether Sexual Fantasies Require Working Memory

Chapter 6: Investigating Whether Sexual Fantasies Activate Non-sexual

Chapter 7: Investigating Whether Child Abusers Associate Sexual Fantasies

Chapter 8: General Discussion

8.1 Summary of findings

8.2 Implications of the findings

8.3 Limitations and directions for future research

Schematic representation of the Dual-Process Model of Sexual

Mean D-associations from the experimental SPF task for

Hypothesised structural equation model

Final structural equation model.

Scores on each WSFQ subscale as a percentage of the overall

Mean difference scores for emotionality for all four conditions

Mean difference scores for arousability for all four conditions

Mean D-scores for each group using the control GNAT

Mean raw response latencies (milliseconds) from the Critical

Main aims of the thesis

Structure of the thesis

1.2 Definition of deviant sexual fantasy

1.3 Prevalence of deviant sexual fantasies

1.4 Effects and functions of deviant sexual fantasy

1.5 Theories of deviant sexual fantasy

1.5.1.1 Evaluation of McGuire et al.s account on deviant fantasies

1.5.2 Laws and Marshalls (1990) account of deviant sexual fantasy

1.5.2.1 Evaluation of Laws and Marshalls account of deviant sexual fantasy

application of second-order Pavlovian conditioning is supported by research showing that

1.5.3.1 Evaluation of MacCulloch et al.s (2000) account

In terms of focus, MacCulloch et al.s (2000) account is constrained to sadistic sexual

2.2 A dual-process approach

2.2.1 Associative processes

cognitive network. According to Gawronski and Bodenhausen, the pattern of activation is

2.2.2 Controlled processes

2.3 Remembering and imagining episodic events

2.4 Outlining the DPM-ST

External or internal cue/trigger

Strong affective response;

(in working memory)

Figure 2.1: Schematic representation of the Dual-Process Model of Sexual Thinking

2.4.1 Associative processes and sexual thoughts

2.4.2 From thought to fantasy

childs ambiguous gesture will be regarded as congruent or 'true' if it is consistent with

2.4.3 The controlled process of sexual fantasising

A second cognitive effect relates to the encoding of fantasy-related information. For

* <.001, < .01, * < .05 Note: SF = Sexual Fantasies