Delays in attentional processing when viewing sexual imagery: The development and comparison of two measures
by
Carmen L. Z. Gress
B.A.(hons), University of Victoria, 1997 M.A., University of Victoria, 2001
A Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree of
DOCTOR OF PHILOSOPHY
in the Department of Educational Psychology & Leadership Studies
© Carmen L. Z. Gress, 2007 University of Victoria
All rights reserved. This thesis may not be reproduced in whole or in part, by photocopy or other means, without the permission of the author.
Supervisory Committee
Delays in attentional processing when viewing sexual imagery: The development and comparison of two measures.
by
Carmen L. Z. Gress
B.A.(hons), University of Victoria, 1997 M.A., University of Victoria, 2001
Supervisory Committee
Dr. John O. Anderson, Department of Educational Psychology & Leadership Studies Supervisor
Dr. C. Brian Harvey, Department of Educational Psychology & Leadership Studies Departmental Member
Dr. Joan M. Martin, Department of Educational Psychology & Leadership Studies Departmental Member
Dr. D. Richard Laws, Department of Psychology Outside Member
Dr. Stephen Hart, Department of Psychology, Simon Fraser University External Examiner
Abstract
The purpose of this study was threefold: (a) develop, validate, and compare two measures, viewing time and choice reaction time, that sexual content induced delay (SCID; Geer & Bellard, 1996) among youth non-sexual offenders, university students, and adults who had sexually offended, (b) address some of the methodological
weaknesses in prior research, and (c) examine the measures’ clinical utility by
investigating their predictive validity via estimates of sensitivity and specificity. Viewing time (VT) assesses how long an individual takes to view an image of a single person while completing a task, and choice reaction time (CRT) measures how quickly and accurately an individual indicates to which category (there must be two or more from which the participant can choose) the presented stimulus belongs. I administered the two measures plus questionnaires on sexual orientation (Friedman et al., 2004) and social desirability (BIDR-6; Paulhus, 1991) to three samples: youth non-sexual offenders, university students, and adult sex offender. I examined the clinical utility of the measures by investigating their predictive validity via ROC estimates of sensitivity and specificity. Each measure consisted of a preset randomized presentation of computer-modified clothed male and female images of various ages. There are five central results from this study. First, both the VT and CRT measures produced subtest scores with high reliability, via item and scale analysis, with all three samples, and there appears to be one dominant underlying construct for both measures. Second, there were significant differences between the adult sexual offenders and the youth non-sexual offenders when assessed with the VT measures, but not between the youth non-sexual offenders and the university
were significant differences between youth non-sexual offenders and the university sample when assessed with the CRT measure, but not between the adult sex offenders and either the youth non-sexual offenders or university students. Fourth, as evidenced by point two and three, the VT and CRT measures provided significantly different results. Finally, the VT measure demonstrated excellent clinical utility in its ability to
differentiate adult heterosexual sexual offenders from non-sexual offenders (for example, AUC = 0.87 female mature images, 0.88 male child images).
Examiners:
______________________________________________________________________ Dr. John O. Anderson, Supervisor (Department of Educational Psychology & Leadership Studies)
______________________________________________________________________ Dr. C. Brian Harvey, Departmental Member (Department of Educational Psychology & Leadership Studies)
______________________________________________________________________ Dr. Joan M. Martin, Departmental Member (Department of Educational Psychology & Leadership Studies)
______________________________________________________________________ Dr. D. Richard Laws, Outside Member (Department of Educational Psychology & Leadership Studies)
______________________________________________________________________ Dr. Stephen Hart, External Examiner (Department of Psychology, Simon Fraser
Table of Contents
Supervisory Committee ... ii
Abstract ... iii
Table of Contents ... v
List of Tables ... vii
List of Figures ... viii
Acknowledgments... ix
Dedication ... x
Chapter 1 ... 1
Introduction to this study ... 1
Introduction to self-regulation ... 4
Dysfunctions in self-regulation ... 6
Self-regulation and sexual offending ... 8
This Study ... 11
Chapter 2 ... 15
Cognitive modeling of sexual arousal and interest ... 15
Introduction ... 15
Information Processing ... 17
IP and cognitive attention-based models of sexual arousal and sexual interest ... 19
Summary ... 34
Matters of Measurement ... 35
Validity and reliability ... 35
Response latencies and sex research: Related issues and potential confounds ... 37
Summary ... 45 Chapter 3 ... 46 Method ... 46 Participants ... 46 Response measures ... 48 Procedure ... 53 Chapter 4 ... 57 Results ... 57 Preliminary analyses ... 58 Viewing time ... 61
Choice Reaction Time ... 76
Overview of main findings ... 85
Chapter 5 ... 87
Discussion & Conclusions ... 87
Overview ... 87
Can we reliably identify SCID in youth? ... 88
Do differences exist in SCID patterns between youth and adults? ... 90
Do differences exist in SCID patterns between individuals who have committed a sexual offence and those who have not? ... 92
of these measures? Are the measures valid? ... 95 Limitations ... 97 Future studies ... 98 Conclusions ... 100 References ... 104 Appendices ... 118
Appendix 1 – Informed Consent Forms ... 118
Youth in Custody ... 118
University Students ... 121
Adult Sexual Offenders ... 124
Appendix 2 - Questionnaires ... 127
Background Information ... 127
Information on Sexual Orientation and Behaviour (Friedman et al., 2004) ... 129
Balanced Inventory of Desirable Responding (BIRD-6; Paulhus, 1991) ... 132
Additional Questions for Individuals convicted of a Sexual Offence. ... 134
Questions to accompany images for viewing time assessment ... 134
Appendix 3 – Software & Research Protocol for Research Assistants ... 135
Appendix 4 - Univariate Outlier Analyses Viewing Time ... 151
Appendix 5 - Univariate Outlier Analyses Choice Reaction Time ... 155
Appendix 6 - Histograms for Viewing Time Items ... 159
Appendix 7 - Histograms for Choice Reaction Time Items ... 170
Appendix 8 - Descriptive Statistics for Viewing Time Items ... 181
Appendix 9 - Descriptive Statistics for Choice Reaction Items ... 192
List of Tables
Table 1: Means (standard deviations) of Age Variables by Sample ... 59 Table 2: Frequencies (percent of sample) of Self-Reported Education by Sample ... 60 Table 3: Pearson R correlations of age and average VT and CRT image categories ... 60 Table 4: Viewing time tests means, standard deviations, Cronbach’s α, and sample
sizes ... 65 Table 5: Viewing time tests means, standard deviations, Cronbach’s α, and sample
sizes ... 69 Table 6: Choice reaction time tests means, standard deviations, Cronbach’s α, and
sample sizes ... 79 Table 7: Choice reaction time tests means, standard deviations, Cronbach’s α, and
List of Figures
Figure 1: ROC Curve predicting adult sexual offenders from university students on three VT test scores. ... 68 Figure 2: Estimated marginal means of viewing time test scores collapsed across all
participants ... 70 Figure 3: Estimated marginal means of viewing time test scores for youth non-sexual
offenders ... 71 Figure 4: Estimated marginal means of viewing time test scores for university
students ... 72 Figure 5: Estimated marginal means of viewing time test scores for adult sex
offenders ... 72 Figure 6: ROC Curve predicting adult sexual offenders from university students on
the four VT test scores. ... 74 Figure 7: Estimated marginal means of CRT test scores for each age category
according to sample and collapsed across image gender ... 82 Figure 8: Estimated marginal means of CRT test scores for youth non-sexual
offenders ... 83 Figure 9: Estimated marginal means of CRT test scores for university students ... 83 Figure 10: Estimated marginal means of CRT test scores for adult sex offenders ... 84
Acknowledgments
Each person listed here played an essential role in my education. Throughout my graduate schooling, they took the time outside of class and regular hours and challenged
me to learn, shake off my preconceptions and assumptions, push past external expectations, and develop my own standards.
They taught me to strive for and expect more.
My supervisor Dr. John O. Anderson My committee members
Dr. C. Brian Harvey, Dr. Joan M. Martin, Dr. D. Richard Laws, and Dr. Stephen D. Hart Additional contributors
Dr. Dan Bachor, Dr. Allyson F. Hadwin, Dr. John Walsh, & Dr. Phillip H. Winne
I give my sincere thanks to the following individuals for their time and effort.
My research assistants Kathleen Koenig, Lorraine Bates, Catherine McLaren, and Phoebe Lin.
Dr. Steve Gray in Arizona and the adult clients who participated in this study. The many mental health professionals and participants at the BC Burnaby & Victoria
Youth Custody Centres.
Dr. Michael Masson in the Department of Psychology at the University of Victoria and the psychology students who participated in this study.
Tyler Buttle and Jamie Brown at Limestone Technologies for designing my software tool and providing tech support at no cost.
Dedication
To Mum and Pops, for instigating and encouraging my love of learning, fine-tuning the stubborn qualities necessary to conduct research and finish a Ph.D., and for always being an open and ready source of love and support along the way.
To my husband, for all of his support and encouragement, as well as his
unwavering belief in my abilities to accomplish this goal. The small brewery at home may have helped …
Chapter 1 Introduction to this study
Sexual offenders, as defined by their criminal convictions, have failed to regulate their sexual behaviour in ways deemed appropriate by contemporary societal standards. In modern Western law, this means that their sexual behaviour(s) have harmed others. To inform prevention and treatment of sexual offending, it seems logical that research should focus on the measurement of sexual self-regulation; for example, measurement of sexual thoughts and strategies, measurement of factors underlying successful self-regulation, measures that allow us to explore why sexual self-regulation fails, and measures that can direct our attempts to strengthen sexual self-regulation skills (Ward & Hudson, 2000; Wiederman, 2004). A review of the literature demonstrates, however, that few published studies have developed and evaluated measures that address these four areas. The
measures that do exist are mainly self-report.
Self-report questionnaires measure a person’s awareness, perceptions,
recollections, and biases about their sexual interests and sexual behaviours. In addition, questionnaires assess a general ability to regulate or control those behaviours, impulses, and desires, rather than describe a particular situation. To-date, self-report measures for sexual offenders focus on products of ineffective self-regulatory; for example, “When I am aroused I will do anything with anyone” (Exner, Meyer-Bahlburg, & Ehrhardt, 1992, p.384). This type of information informs treatment plans by identifying treatment needs and risk concerns, but these items do not measure the process or achievement of sexual self-regulation. This type of information is important for treatment plans and risk management, but these items do not measure the process or achievement of sexual
self-regulation. For example, agreement with the item above does not explain why the person is fixated on their arousal or what processes underlying self-regulation, such as attention, are facilitating that fixation. The field requires attention to, therefore, and measures that assess self-regulatory processes and dysfunctions in those processes.
One potential measureable characteristic that indicates a failure or dysfunction in self-regulation among sexual offenders is sexual content-induced delay (SCID), a specific form of attentional bias associated with preferred sexual content (images or text), first proposed by Geer and colleagues (Geer & Bellard, 1996; Geer & Melton, 1997). SCID occurs when a salient sexual stimulus triggers attentional processes interfering with or limiting attentional processes to other tasks. This can cause delays in task processing and may be, therefore, a characteristic of a failing or dysfunctional self-regulatory system. The purpose of this study was to design, evaluate, and compare two measures to examine SCID. To provide an appropriate context for this study, I begin by briefly introducing theories of regulation, dysfunctions in regulation, and the application of self-regulation to research on sexual offending. I then introduce the hypothesis that some of the key processes underlying self-regulation, such as attentional processes, may be inhibited or over stimulated by particular events or objects, such as those matching an individual’s sexual interest. Upon review of the various sexual behaviour theories, sexual interest, most likely a psychological construct, appears to be comprised of initial interest due to hormonal and/or other biological factors, which motivates or leads to sexual arousal and arousability to certain age(s) and gender(s), which then leads to a more established sexual interest. Since the main focus of this study is to develop and validate an attention-based measure of SCID, the literature section focuses on the information
processing approach to research on attention, the literature on attentional-based response latency measures of SCID, and key methodological issues associated with using response latencies to assess attentional processes.
Introduction to self-regulation
Self-regulation of thoughts and behaviours is a goal-guided and iterative process. It incorporates internal and external information, such as memories of previous
experience, feedback, and the current environment, to inform the implementation, evaluation, and adaptation of thoughts and behaviours over an extended period of time and varying contexts (Boekaerts, Maes, & Karoly, 2005; Karoly, 1993; Maes & Karoly, 2005; Raffaelli, Crockett, & Shen, 2005; Winne, 2001; Zimmerman, 2000). Theoretical models of self-regulation suggest that the responsibility for regulation and change lies with each individual and the effort exerted by individuals to self-regulate and adapt is considered a key component of development (Cicchetti & Tucker, 1994).
Successful self-regulators engage cyclically in (a) perceiving their situation; (b) setting goals and planning steps to reach them; (c) enacting plans and monitoring thoughts and behaviours; (d) evaluating their needs and the products of their strategies, adjusting thoughts and behaviours when necessary to accomplish their goals; and (e) metacognitively examining their personal and external resources (Winne, 2001;
Zimmerman, 2000). Interacting mechanisms underlying each of these phases most likely include automatic or unconscious filters (e.g., long term memories and/or schemas), impulses, attentional processes, and conscious strategies to enact, monitor, and control thoughts and behaviours (Banfield, Wyland, Macrae, Münte, & Heatherton, 2004; Posner & Rothbart, 2000). Understanding the nuances and individual differences of
self-regulation in normal and atypical individuals will lead to advances in learning and instruction as well as diagnosis, prevention, and treatment of developmental problems (Posner & Rothbart, 2000; Winne, 2001; Zimmerman, 2000).
The majority of research on self-regulation examines the products of effective versus ineffective self-regulatory skills. For example, Raffaelli, Crockett, and Shen (2005) investigated the development of self-regulation among boys and girls, with initial assessments at ages 4 and 5, and follow ups at ages 8 and 9, then 12 and 13. They
measured aspects of affective, behavioural, and attentional self-regulation via a selection of items from the Behavior Problems Index (BPI; Peterson & Zill, 1986) such as “he/she is stubborn, sullen, or irritable” and “he/she has difficulty concentrating, cannot pay attention for long” (as cited in Raffaelli et al., 2005, p.75). Raffaelli et al. (2005) determined that behaviours associated with self-regulation of affect, behaviour, and attention, as measured by the BPI, formed a single integrated construct. In addition, the authors found that the level of self-regulatory skills increased with age, with girls
exhibiting better self-regulation than boys in each age group. Buckner, Mezzacappa, and Beardslee (2003) investigated self-regulation and resiliency among 115 youth from very low income families: 45 classified as resilient and 70 as non-resilient. The authors compared behavioural products of executive functioning and emotional reactivity, such as attentiveness, concentration, persistence, planfulness, reaction to minor frustrations, and inappropriate emotional responses. After controlling for negative life experiences and chronic strains, Buckner et al. found the resilient youths exhibited significantly higher levels of successful self-regulatory behaviours than non-resilient youths.
A second approach to researching self-regulation examines self-regulatory processes while attempting to preserve temporal integrity of the targeted construct. In other words, rather than focusing on the products of self-regulatory skills, researchers record in real time how and with what skills a person self-regulates as well as why some
strategies fail. For example, Hadwin, Winne, Nesbit, and Kumar (2005) detailed how advances in computer technology permit researchers to trace cognitive and metacognitive processes associated with real-time self-regulation in a learning environment. The authors traced the activity of 188 undergraduate students in education during two study sessions. This exploratory analysis demonstrated that trace information can inform learners and researchers about self-regulatory strategy use, compare self-perceptions of self-regulation and actual behaviours, as well as investigate real-time monitoring and adaptation of self-regulatory skills. In addition to investigating successful (or unsuccessful) regulation, the measurement of real-time processes permits examinations of constructs that potentially frustrate productive self-regulation.
Dysfunctions in self-regulation
Zimmerman (2000) notes that self-regulation models of change and management will not succeed without addressing variables that may unhinge or thwart productive self-regulation. Baumeister and Heatherton (1996) describe three patterns of dysfunctional self-regulation: under-regulation, mis-regulation, and effective but inappropriate
regulation. Under-regulation occurs when there is a lack of ingrained societal standards, inadequate monitoring of those standards, or a “lack of the cognitive resources necessary to achieve desired goals” (Keenan & Ward, 2003, p. 123). Examples are substance abuse, excessive eating, and/or inappropriate sexual behaviour. Mis-regulation occurs when strategies to self-regulate towards specific goals are ineffective or counterproductive, such as drinking alcohol to relax and improve mental state or when sex offenders who have been convicted of child molestation ‘test’ themselves by walking past a children’s park. Effective (but inappropriate) regulation with a dysfunction set of goals is
exemplified in the actions of a child molester who does not attempt to inhibit his sexual desire for children because he or she believes children need to be taught about sexual activity.
Zimmerman (2000) suggests four categories of self-regulatory dysfunction or ‘personal limitations’ that have potentially severe consequences: ineffective forethought and performance control, apathy or disinterest, learning disabilities, and mood disorders. While the first two categories suggest personal limitations amenable to training programs on self-regulatory strategies, the latter two categories, similar to the category of under-regulation above, hint towards shortcomings or a weak point in the processes or mechanisms underlying active self-regulation, such as working memory and/or attentional processes. This study focuses on the mechanics of attention and individual differences associated with these processes.
Attentional processes are considered key functions in self-regulation because these functions “maintain information in the mind for the execution and sequencing of mental operations” (Banfield et al., 2004, p. 71). It stands to reason, therefore, that if this process is not working or is inhibited, self-regulation would also be inhibited or fail (MacCoon, Wallace, & Newman, 2004). This pattern is evident in studies demonstrating that increases in attentional processes to a particular stimulus or stimuli result in a decrease in attentional processes to other stimuli and in the ability to suppress or restrain inappropriate responses (Engle, Conway, Tuholski, & Shisler, 1995; Roberts, Hager, & Heron, 1994; Williams, Mathews, & MacLeod, 1996). Consequently the identification and examination of the variables that ‘draw on’ or bias attention are essential.
Research indicates individuals with emotional disorders, such as posttraumatic stress disorder, are preoccupied with and demonstrate increased attention to stimuli that represent their area of concern or perceived threat (Williams et al., 1996). This attentional bias interferes with cognitive processing for other tasks. Williams et al. (1996) asserted that in general, cognitive models assume attentional bias is a by-product of an emotional state and is important to the creation and maintenance of that state. As body arousal increases, so does attentional bias, maintaining arousal and forming a cyclical
relationship. The extent of the interference is an indicator of the strength of the attention bias. Researchers have investigated the relationship between attentional bias and
increased arousal due to fear and extreme stress (Doost, Moradi, Taghavi, Yule, & Dagliesh, 1999), anxiety (Beck, Emery, Greenberg, & Lindemann, 1996), panic disorder (McNally, 2002), depression (Nolen-Hoeksema, Morrow, & Fredrickson, 1993), and posttraumatic stress disorder (Yule, 2001).
Self-regulation and sexual offending
An area interested in the application of self-regulation models is research, assessment, and treatment of sexual offending. This interest in self-regulatory processes is a consequence of dissatisfaction with research and treatment based on existing
cognitive-behavioural relapse prevention strategies (Ward et al., 2004; Ward, Yates, & Long, 2006). For example, researchers and treatment professionals allude to problems in the conceptual models, which may not apply to sexual offenders as there is a lack of empirical evidence and significant long term treatment effects (Barbaree, 2006; Laws, Hudson, & Ward, 2000; Marshall, 1996; Ward & Hudson, 2000; Ward & Sorbello, 2003; Yates, Kingston, & Hall, 2003).
Self-regulation model of the offence, treatment, and relapse process focuses on individuals’ choice of goals, and whether the goals are to approach inappropriate sexual thoughts and behaviours or avoid them (Ward et al., 2004; Ward & Hudson, 2000; Ward et al., 2006). The model consists of nine phases, which include goal setting, enacting, evaluations, and attitudes, which an individual can enter or exit any time.
Like research in other areas of self-regulation, the identification of variables that thwart self-regulation of sexual thoughts and behaviours is integral to comprehensive models of self-regulation as applied to sexual offending. Again, the study of attentional processes can provide insight into the underlying dysfunction. For example, sexual content-induced delay can provide a real-time indicator of sexual bias (SCID; Conaglen, 2004; Geer & Bellard, 1996; Geer & Melton, 1997). SCID is hypothesized to occur when a salient sexual stimulus diverts attention from another relatively simple task towards it. This can cause delays in task processing.
As a special case of attentional bias in general, SCID is subject to the same theoretical assumptions attention bias: in other words, SCID is a by-product of an emotional state and is important to the creation and maintenance of that state and the strength of attentional bias, and therefore SCID increases as body arousal increases (Williams et al., 1996). In addition, the extent of the interference is an indicator of the strength of SCID. When assessed in males, study results suggest the strength or length of SDIC provides a reliable estimate of sexual interest (age and gender). This is
(potentially)1 evidenced by studies presenting a range of visual sexual information, such as images and/or text, and using an assortment of attention-based measures including
1 Potentially, because no studies exist that compare these measures or examine the measures ability to assess
choice reaction time (CRT; Giotakis, 2005; Wright & Adams, 1994, 1999), the Emotional Stroop (Price, 2006; Smith & Waterman, 2004), eye startle probe reflex (Giargiari,
Mahaffey, Craighead, & Hutchison, 2005; Hecker, King, & Scoular, 2006), rapid serial visual presentation (RSVP; Beech, Kalmus, Tipper, Baudouin, & Humphreys, 2006), and viewing time (VT; Abel, Huffman, Warberg, & Holland, 1998; Beech et al., 2006; Gress, 2005; Harris, Rice, Quinsey, & Chaplin, 1996; Letourneau, 2002).
Identifying or substantiating a sexual offender’s sexual interest is important in clinical forensic settings as it can lead to enhanced court decisions regarding risk assessment, sentencing, civil commitment decisions, and dangerous offender designations. This is because certain sexual interests are associated with a higher recidivism risk. For example, Hanson and Bussière (1998) state “the risk for sexual offense recidivism was increased for those who had…selected male victims” (p. 351). More often than not, a mental health professional can determine a client’s sexual interest via the offender’s self-report information and/or sexual history. On occasion, however, a documented history does not exist, information is not forthcoming, or an offender denies his sexual offence history and/or index offence. A reliable, objective, and standardized sexual preference instrument would augment legal and treatment assessment.
In summary, the results of these studies and others suggest that individuals experience delays in attentional processing when viewing stimuli that match their sexual interest. It is unknown, however, if these delays are a characteristic, among some
individuals, of dysfunctional processes in self-regulatory processes. If so, individuals who have demonstrated previous failures in self-regulation associated with sexual thoughts and behaviours, such as sexual offenders, should demonstrate significant
differences in their processing of sexual information than non-sexual offenders. In other words, patterns of SCID should reliably identify sexual interest for all the participants and sexual offenders should produce significantly longer SCID than non-sexual offenders.
This Study
The crux of this study was a methodological focus on the development, examination, and comparison of two response latency measures: choice reaction time (CRT) and viewing time (VT) to examine sexual content-induced delay (SCID). VT assesses how long an individual takes to view an image of a single person while
completing a task and CRT measures how quickly and accurately an individual indicates to which category (there must be two or more from which the participant can choose) the presented stimulus belongs. The purpose of this focus was to address a number of
methodological issues evident in the current literature. For example, studies utilizing any of the above measures to assess SCID cite many of the other methods as evidence of applicability and appropriateness. Although the measures as a group provide strong evidence for the utility of attention-based measures in sex offender research and applied settings, the measures may assess different cognitive processes and one or more
underlying constructs. In this study, the same group of participants completed both the CRT and VT measures. In addition, I conducted factor analyses on each measure to determine if there is one or more underlying constructs (i.e., is sexual interest a single construct or a group of constructs?).
A second issue is the lack of evidence that SCID, as assessed by our current measurement approaches such as VT or CRT, occurs in youth in a similar manner as
adults. Findings on adult males dominate this research area. SCID investigations that include adolescents are scarce. Two recent studies examined the ability of attention-based measures of sexual interest (specifically viewing time) in youth sexual offenders in a clinical setting (Abel et al., 2004; Worling, 2006). Neither study provided strong evidence for the applied use of viewing time measures as an identifier of sexual interest among youth. The lack of results may be due to the particular design of the tools or to developmental processes. There may be substantial differences in cognitive processing of sexual material between youth and adults, potentially invalidating current methods designed on adult males and applied to youth. This study steps back, therefore, from clinical application to examine potential differences in cognitive processing of sexual material, as evidenced by attentional process measures, between youth and adult males.
Finally, this study addressed a few measurement issues associated with response latency measures, which have yet to receive attention in the sex offender literature. For example, very few studies state whether or how they conducted outlier analysis. Outliers are common to response latency research and can cause a variety of problems, such as disproportionate increases or decreases in the mean and variance (Ratcliff, 1993). Although the use of non-parametric methods can address this problem, these corrections reduce statistical power and effect size. This study acknowledges and evaluates the potential outlier problems, provides a brief review of outlier analysis, and a clear description of the outlier methods used in this study. Second, known confounds of response latency measures, such as age and education levels (Bugg, Zook, DeLosh, Davalos, & Davis, 2006), are rarely addressed in this literature. This study includes a
variety of age groups and provides statistical analyses that examine the relationship between age and response latency.
To summarize, for this study I developed two attention-based response latency instruments based on viewing time (VT) and choice reaction time (CRT) methods to elicit and assess SCID, and I administered these measures to three distinct samples: youth non-sexual offenders2, university students, and adults who had sexually offended. In addition to examining the validity and reliability of these measures, I examined the clinical utility of each measure by investigating its predictive validity. To accomplish this I estimated the measures sensitivity (the probability that a test is positive given the person is classified as positive via other means, or the true positive rate) and specificity (the probability that the test is negative given the person is classified as negative via other means, also known as the true negative rate). Specifically, I examined the following:
(a) Can we reliably identify SCID in non-sexually offending adolescents? (b) Do differences exist in SCID patterns between those adolescents and
non-offending adults as well as adult sexual offenders?
(c) Do differences exist in SCID patterns between individuals who have committed a sexual offence and those who have not?
(d) Do differences exist in the SCID patterns as produced by the VT and CRT measures?
(e) Are the differences substantial enough to warrant clinical application of one or both of these measures?
To provide further context for the study, chapter two reviews attention-based measures that assess individual differences in response to sexual stimuli and issues
2 The initial research design included youth sexual offenders and non-sexual offenders to permit comparison
analyses. Youth sexual offenders turned out to be an extremely difficult group to recruit, therefore only youth non-sexual offenders are included in this study. Recruitment efforts are ongoing.
related to test validation in general and issues specific to response latency measurement and research on sexuality.
Chapter 2
Cognitive modeling of sexual arousal and interest Introduction
Researchers model and measure attention, specifically the performance associated with different aspects or attention processes such as attention span, interference, selective attention, or attentional bias, to investigate complex cognitive and behavioural patterns ranging from memory and emotion to the recent examinations of sexual arousal and interest (Geer, Estupinan, & Manguno-Mire, 2000). The impetus for cognitive attention-based models of sexual arousal and interest arose from studies investigating the effect of distraction on male sexual arousal (Barlow, 1986). For example, Laws and Rubin (1969) as well as Henson and Rubin (1971) demonstrated that prompting a person to distract themselves with non-sexual thoughts during a physiological assessment of sexual arousal can lead to decreases in penile circumference or blood flow. A study by Rook and
Hannen (1977) demonstrated that those participants who identified non-sexual
physiological arousal (for example, blood flow in the genital region can increase during fearful events) as sexual arousal consistently interpreted nonsexual situations as sexual if genital sensations occurred simultaneously.
Barlow (1986), informed by his and colleague’s work on sexual dysfunction, proposed a cognitive-affective model of sexual arousal focusing on the perception of physiological arousal and the cognitive processing of erotic cues. In his 1986 article, Barlow noted that sexually dysfunctional men (specifically, erectile dysfunction) reacted negatively to sexual situations, suggesting that their sexual experiences begin with awareness and identification of sexual cues and move to negative perceptions and
emotions. Geer and Janssen (2000) proposed that after appraisal and recognition, emotional saliency dictates the degree to which regulatory processes focus additional attention on sexual behaviours or sensations and whether that attention focuses on positive or negative aspects. Additional attention to sexual information is evidenced by studies demonstrating that anxiety and/or sexual distractors can increase sexual arousal in sexually functional men (due to additional attention to positive sexual aspects) but
hampers it in sexually dysfunctional men due to additional attention to negative sexual aspects (Barlow, 1986). For example, Barlow, Sakheim, and Beck (1983) examined the effects of performance anxiety and generalized anxiety by comparing three conditions: telling participants either that (a) there was a 60% chance of electric shock if they did not demonstrate adequate levels of physiological (penile) arousal to the presented stimuli, (b) shock may occur and it is unrelated to arousal, or (c) no shock will occur. The authors found that both shock conditions increased physiological arousal. Barlow (1986) found non-sexual distractors can cause the opposite patterns, and that sexually dysfunctional men under-report levels of sexual arousal.
Barlow’s and other models of sexual arousal have made considerable contributions to the field of sexual research. These models treat sexual arousal as a unified construct, however, which fails to explain individual differences found in many studies on sexual arousal and interest (Janssen, Everaerd, Spiering, & Janssen, 2000). The interest in cognitive models of sexual arousal has grown, prompted perhaps by the
prolific research on cognition in applied psychology (Geer et al., 2000; Geer &
Manguno-Mire, 1996). Information processing approaches to cognition inform many of the current studies on sexual arousal.
Information Processing
The information processing approach (IP) to cognition consists of various models that focus on quantitative cognitive changes, that is, changes over time in thought
patterns, memory recall, and attention, rather than quantitative shifts in patterns of
thoughts and behaviour characterized by developmental stages (Flavell, Miller, & Miller, 2002). IP models view information as cognitive input to be processed, encoded, stored, and retrieved. Input is of various types, such as declarative knowledge (knowing ‘what’), procedural knowledge (knowing ‘how), and conditional knowledge (knowing ‘when’), and can be organized into a variety of forms, sizes, levels of complexity and abstraction (Flavell et al., 2002). The overarching goal of IP models is to operationally define elements of the cognitive system, such that the mechanisms of cognitive function and change are detailed in real time and minute detail (Flavell et al., 2002; Geer, Lapour, & Jackson, 1993). An area that has received a great deal of consideration in IP is attention or attentional variables because it is through the attentional processes that stimuli first enter the cognitive system (Geer & Janssen, 2000).
IP models of attention focus on how stimulus information are perceived, filtered, and then translated into a response. IP models concentrate on three approximate
processing phases (with numerous variations in the proposed properties of each stage); first perception and identification, then decision making and response selection, and finally response execution (Geer et al., 1993; Johnson & Proctor, 2004). Factors known to affect processing within or across each of these phases include (but are not limited to) information previously processed and encoded into implicit or explicit memory,
attention to tasks participants are overtly asked to perform, evidenced by a change in speed and/or accuracy in task performance or amount of short-term memory recall. For example, individuals typically take less time to perceive, make decisions and complete a task for which they have established a high level of skill (Flavell et al., 2002).
Alternatively, individuals will take more time to complete a task and may make a higher percentage of errors if presented with stimuli that represent emotional concerns or states (Williams et al., 1996). The literature suggests this occurs because the stimuli occupy attentional resources leaving the task with less attention.
The ability of certain stimuli to distract an individual, which result in delays when performing a particular task, highlights a key assumption of all IP models: the human cognitive system can experience overload. In other words, input and processing can exceed response capacity, which indicates that a maximum capacity exists (Flavell et al., 2002). Maximum capacity is manifested behaviourally when increases in attention to one stimulus or task results in decreases in attention to another, whether or not the individual is aware of the shift in direction of their attention. The notion of system overload does not suggest, however, that all information processing is serial, that is, one input is processed at a time. On the contrary, research on IP models of emotion and emotion regulation demonstrate that the brain uses both serial and parallel processing, depending on the information to be processed and the nature of the task (Geer et al., 1993; Lang, 1994). A variety of studies suggest that sexual arousal and interest are, in part, due to emotional reactions and filters. Cognitive models of sexual arousal, therefore, must include both serial and parallel interactions between memory, regulatory, and attentional processes (Spiering, Everaerd, & Laan, 2004).
IP and cognitive attention-based models of sexual arousal and sexual interest
The information processing approach to sexual arousal and interest is an attempt to model cognitive components associated with sexual interest and sexual deviance. The majority of research in this area is built on two suppositions: (a) sexual arousal and interest, in addition to physiological and behavioural aspects, includes cognitive and affective components; and (b) stimuli that originate in the environment produces results similar to stimuli that are generated internally, (e.g., when visualizing images or engaging in sexual fantasies) (Geer et al., 1993; Janssen et al., 2000).
The appraisal of a stimulus as sexual is the first phase of cognitive models of sexual arousal and interest (Geer et al., 1993). A stimulus requires appraisal and then encoding and/or decoding to determine if the stimulus matches explicit and/or implicit long-term memories for what is ‘sexual.’ Explicit memories of sexual information include recollections of previous sexual encounters, attitudes towards sex, and fantasies. (Spiering et al., 2004). This is the type of information gathered by sexual interest
questionnaires and interviews (Geer et al., 1993; Janssen et al., 2000). Implicit memories of sexual information include innate sexual reflexes, automatized scripts, and classically conditioned sensations. If via implicit memories a stimulus is appraised as ‘sexual’ the system responds by stimulating genital arousal (Janssen et al., 2000; Morris, 2002). Genital arousal is assessed in research and clinical settings with penile plethysmography (aka phallometry), a measure of penile blood flow through circumferential measurement (or volumetric pressure) of the penis (Freund, 1963; Freund & Watson, 1991; Laws, 2003). If explicit memories corroborate the stimulus as ‘sexual’ then the system responds by triggering additional attentional and regulation processes.
The trigger of conscious and unconscious regulatory processes is the next step in cognitive models of sexual arousal (Spiering et al., 2004). Conscious regulation processes encode or filter a potentially arousing sexual stimulus through explicit and short-term memories of current societal expectations and the current environment. For example, North American society expects individuals to inhibit their expressions of physical sexual arousal in public places. If conscious information regulation indicates further attention to the sexual stimulus is appropriate, then the system responds by providing more attention. This moves a person closer to a “complete sexually emotional experience” (Damasio, 2003, p.86). If the information gathered indicates further attention and processing of the sexual stimulus is inappropriate, then the system responds with conscious inhibitory control (Baars, 1998; Gross, 1998).
Unconscious or automatic regulatory processes use emotional salience to determine the amount of attention given to a potential sexual stimulus (Spiering et al., 2004). Matches between the stimulus and positive explicit and/or implicit long term memories trigger additional attention to the positive aspects of the stimulus, leading to increases in sexual arousal, while negative implicit and explicit memories indicate a focus on negative information leading to avoidance or a discontinuation of attentional processes (Barlow, 1986; Spiering, Everaerd, Karsdorp, Both, & Brauer, 2006; Spiering et al., 2004). A by-product of the increase in attentional processes for a positive match is sexual content-induced delay (SCID), an attentional bias associated with preferred sexual content (images or text), first proposed by Geer and colleagues (Geer & Bellard, 1996; Geer & Melton, 1997). SCID is hypothesized to occur when a salient sexual stimulus triggers attentional processes that then interfere with or limit attentional processes to
other tasks causing a delay in task processing. As a special case of attentional bias in general, SCID is subject to the same theoretical assumptions as attentional bias. The assertion that attention bias is a by-product of an emotional state and is important to the creation and maintenance of that state applies to SCID. In addition, the strength of attentional bias, and therefore SCID, increases as body arousal increases, maintaining arousal and forming a cyclical relationship (Williams et al., 1996). Geer and Melton (1997) examined the effect of priming on lexical decision making and found erotic text in either the prime or target sentences, when compared to neutral text in both prime and target sentences, caused significant delays in decision making.
Researchers have investigated the interaction between attention processes and the emotional salience of explicit and implicit sexual memories by examining the
relationship between various sexual behaviours and varying lengths of SCID. These studies present a range of visual sexual information, such as images and/or text, to participants and measure SCID using an assortment of attention-based measures including choice reaction time (CRT; Giotakis, 2005; Wright & Adams, 1994; CRT; Wright & Adams, 1999), the Emotional Stroop (Price, 2006; Smith & Waterman, 2004), eye startle probe reflex (Giargiari et al., 2005; Hecker et al., 2006), rapid serial visual presentation (RSVP; Beech et al., 2006), and viewing time (VT; Abel et al., 1998; Beech et al., 2006; Gress, 2005; Harris et al., 1996; Letourneau, 2002). The implicit association test is another cognitive attention-based measure that is under investigation for its potential to inform research and clinical application to sex offender treatment (Gray, Brown, MacCulloch, Smith, & Snowden, 2005; Mihailides, Devilly, & Ward, 2004; Nunes, Firestone, & Baldwin, in press), but it assesses the assess the strength of unspoken or hidden relations between two separate concepts such as child and adult and two contrary
attributions such as pleasant and unpleasant. I will not include it, therefore, in the review below.
Some researchers approach the investigation from an IP viewpoint and therefore investigate attentional bias or SCID, while others approach the investigation from an applied or clinical point of view within the context of sexual offender assessment and treatment, and state they are investigating a measure’s ability to quantify and identify sexual interests. Most studies have focused on determining if a particular measure can reliably differentiate adult participants by their sexual orientation and/or sexual age preference (for example, if individuals who have offended against children demonstrate significantly longer delays to images of children than adults) (Abel, Jordan, Hand,
Holland, & Phipps, 2001; Gress, 2005; Harris et al., 1996; Worling, 2006). A few studies compared SCIDs of sexual offenders to SCIDs of non-sexual offenders (Beech et al., 2006; Smith & Waterman, 2004), and one study evaluated the difference in SCIDs between participants who self-reported high sexual desire and those self-reporting low sexual desire (Giargiari et al., 2005).
Choice reaction time (CRT). Wright and Adams (1994) used choice reaction
time (CRT) to investigate whether a favourable cognitive appraisal and sexual response towards an attractive individual would interfere or compete with other cognitive activity. CRT is a simple measure of how quickly and accurately an individual indicates the category a presented stimulus belongs (there must be two or more from which the participant can choose). In general, sexual preference studies using CRT instruct a participant to use specific keys on a keyboard to indicate the position of a white dot superimposed on the image (there can be multiple positions). When participant makes a
choice, the software program removes the two stimuli (image and dot) and presents the participant with another image and dot. The location of the white dot, the image content, and the apparatus used to indicate location vary from study to study.
Wright and Adam evaluated 80 university undergraduate and local community participants, evenly categorized into four groups by sexual orientation (heterosexual and homosexual) and gender. Their CRT measure contained 60 slides, comprised equally of commercially available nude males, nude females, and neutral images from travel magazines and a white dot in any one of five locations (each corner and the middle) counterbalanced by location. The authors found that (a) homosexual males were significant slower at locating the dot on male than female or neutral slides, (b)
heterosexual males were significant slower on female than male and neutral slides, (c) heterosexual women were significant slower on male than female slides (but not neutral slides), and (d) homosexual women were significant slower on female and neutral slides. In addition, that male participants made significantly more errors on the CRT task than female participants and both groups made more errors on male images.
Wright and Adams (1999) repeated their experiment using two series of 60 slides (a clothed set with pictures from fashion magazines and a nude set using the same images as the 1994 study) and an additional neutral set consisting of a solid blue background. The results demonstrated that homosexual males had significantly longer latencies when indicating dot location on male nude images than homosexual women (as expected), but not longer than heterosexual men. No significant results were found between groups on clothed imagery.
Heterosexual men demonstrated significantly longer latencies on female nude images than heterosexual women and homosexual men (as expected). When examining within group latencies, the results were mixed. Each group took significantly longer to complete the task on their respective sexual interest image set than their non-interest group and neutral image set, except for heterosexual women who took longer on male and female images when compared to time taken on neutral images but their latencies on male images were not significantly longer than the latencies on female images. In
addition, heterosexual males had significantly longer latencies on female clothed images than male images but not neutral images. In the learning task, more participants identified novel images and each group demonstrated higher levels of recognition for their preferred category when the images were nude. In summary, these results demonstrate that
individuals are distracted from their task when they see images that represent their sexual interest, indicating that a relationship exists between sexual interests and attentional processes.
Giotakis (2005) conducted a similar study of CRT using commercially available clothed images. The author administered 90 photos, 10 photos from nine categories (violence [boxing, war, football], neutral [scenic photos], adult males, adolescent males, boys, adult females, adolescent females, girls, and blank) to 135 participants, 58 of whom were sexual offenders (31 convicted for rape, 8 for intra-familial child molestation, 19 for extra-familial child molestation) and 53 males and 24 females as a comparison group (staff from a nearby hospital). The author simplified the CRT task by reducing the number of choices to two possible white dot locations, in either the right or left corner, with a two-second interval between slides. Two main between-group results emerged:
groups convicted for sexual offenses demonstrated significantly longer overall reaction times than the comparison group of males and females, and images of boys and of girls were viewed significantly less in relation to the other stimuli. There were no significant differences in error analyses. This suggests there is a relationship between SCID and sexual offender status.
Santtila, Mokros, and Viljanen (2006) replicated Wright and Adams (1994; 1999), with a few modifications, with 25 male university students (15 heterosexual and 10 homosexual). First, instead of using a plain blue background as the neutral stimulus, Santtila et al used the same pictures in both sexual and neutral sets. The images, obtained from erotic internet websites, consisted of 40 nude men and women. The authors
modified the images in the neutral set by covering the torso with a solid rectangle. Second, the authors added a priming slide, sexual or neutral, before the target slide using the same images. For statistical analysis, the authors spit the trials into three equal phases. In phase one both groups took significantly longer to respond to sexual stimuli matching their sexual interest than all other images. In phase two, results were not significant. In phase three the priming slide modified results for heterosexual men. When the prime was sexual, heterosexual men took significantly longer to respond to sexual female images but not when the prime was neutral. Homosexual men responded similarly for all images. An AUC3 analysis predicting group membership from difference scores (latency to female images minus latencies to male images) produced estimates of .82 for phase one, .63 for phase two, and .60 for phase three, demonstrating that the test’s performance diminished over trials.
3 Area under the curve (AUC) analyses produce receiver operator characteristic (ROC) curves, a graphical
Eye-startle probe reflexes. Giargiari, Mahaffey, Craighead, and Hutchinson
(2005) examined startle probe reflexes, pre-pulse inhibition (the inhibition of the startle reflex, caused by a preceding weak stimulus) of the startle probe reflex, and self-reported sexual desire for 36 photos, 24 of which were sexually provocative (equally balanced for males and females) and 12 neutral. The startle eye-blink phenomenon, also known as the eye startle blink, acoustic startle response, and startle probe reflex is an autonomic response to an intense stimulus, resulting from primary neural pathways associated with approach and avoidance responses (Cornwell, Echiverri, & Grillon, 2006). The startle blink reflex measures motivational states (e.g. approach vs. avoidance) via the magnitude of the blink (typically assessed via EMG electrodes placed under the lower eyelid) and its latency. These variables vary as a function of the person’s emotional valence toward the stimulus presented (Koukounas & Over, 2000). A person attending to a pleasant or positive visual or auditory stimulus exhibits attenuated (less magnitude and slower) eye-blink responses when presented an intense and unexpected stimulus (e.g., blast of white noise). On the other hand, they will demonstrate an augmented (increased magnitude and faster) eye-blink response when startled while attending to a stimulus they find aversive or negative (Hecker et al., 2006). Giargiari et al., (2005) found that participants with higher self-reported sexual desire showed greater decreases while viewing opposite-sex stimuli as compared to neutral stimuli than did participants with lower sexual desire. In addition, they found participants with higher self-reported sexual desire showed greater decreases while viewing opposite-sex stimuli as compared to neutral stimuli than did participants with lower sexual desire.
Hecker et al. (2006) examined three research questions using the startle probe reflex: (a) will startle probe reflexes attenuate when viewing sexually attractive images? (b) does the startle probe reflex vary when viewing images portraying a variety of potential sexual partners, including deviant partners? and (c) can the startle probe reflex by voluntarily suppressed? Using various erotic, aversive, and neutral images, the authors found that participants (a) demonstrated significantly increased intensity of eye-blink startle reflexes to aversive stimuli in comparison to neutral or sexual images, (b) demonstrated significantly attenuated eye-blink reflexes while viewing photographs of sexually mature females and imagining sexual activity with them, relative to the magnitude of eye-blink reflexes while imagining the same activity with images of non-preferred sexual targets, and (c) were unable to suppress their sexual interest even when instructed.
The Emotional Stroop. Smith and Waterman (2004) investigated between-group
differences in processing sexual information using the Emotional Stroop. The Emotional Stroop task, a modification of the Stroop colour naming task, requires participants to name the font colour of emotional salient words (Williams et al., 1996). The task relies on the fact that after about three years of literacy, a person cannot look at a word without reading it. The font color-naming latencies of words representing areas of concern, threat, or expertise are consistently longer than the colour-naming latencies of matched control words. This effect is best defined as a response interference effect, not a response conflict effect.
The Emotional Stroop is theorized to assess attentional bias, a focus on one or two aspects of a situation while unsuspectingly ignoring the rest (Williams et al., 1996). A
number of studies have demonstrated that participants with clinical anxiety disorders take longer to identify the colour of words with threatening themes than neutral words when compared to controls (Williams et al., 1996). In addition, individuals demonstrating high trait anger demonstrated the Stroop effect when responding to anger words (Eckhardt & Cohen, 1997) and threat words (Van Honk, Tuiten, de Haan, van den Hout, & Stam, 2001). Smith and Waterman (2004) evaluated 43 participants (10 sexual offenders, 10 violent offenders, 10 non-violent offenders, and 13 university undergraduates).
The authors administered 150 experimental trials consisting of 25 words in each of the following themed categories: aggression, sexuality, positive emotion, negative emotion, neutral words, and color words (always incongruent with font colour). The results indicated that sexually themed bias scores for sexual and violent offenders (bias scores were calculated by subtracting the mean neutral word presentations from each of the stimulus word condition means; sexual, aggression, positive, negative, and colour) were significantly slower than the undergraduate sample. In addition, sexual offenders demonstrated the longest reaction times to colour naming sexual words compared to neutral words.
In a replication study by Price (2006), sexual offenders with rape convictions were significantly slower to colour-name sexually themed words than the community sample, and were slower, but not significantly slower than the child molesters, violent offenders, and non-violent non-sexual offenders. Sexual offenders with convictions for child molestation were significantly slower than the community and violent non-sexual offender groups when naming colour words (regular Stroop test).
Rapid serial visual presentation (RSVP). Beech, Kalmus, Tipper, Baudouin,
and Humphreys (2006) used the rapid serial visual presentation (RSVP) method to determine if sexual offenders with child victims would have significantly more errors identifying Target 2 after correctly identifying Target 1 as a child . Rapid serial visual presentation (RSVP), a brief display of information (generally text) in a limited space in sequential order (Potter & Levy, 1969), is a method used in perceptual and cognitive psychology. It provides an opportunity to observe attentional blink, when an individual can detect one target stimulus in a group of distracter stimuli but fails to or takes longer to detect a second different target presented within 500ms of the first (Raymond, Shapiro, & Arnell, 1992; Shapiro & Raymond, 1994).
Attentional blink, believed to be the time taken to select and process stimuli from visual short-term memory (VSTM; Jolicoeur, 2002), is affected by manipulations in visual similarity between targets and distractors and conceptual similarities, for example, when the first target stimulus is more salient to the viewer (Kyllingsbaek, Schneider, & Bundesen, 2001). The authors presented two conditions each with 216 sequences of 11 images to 35 adult male sexual offenders with offences against children (16 intra-familial and 18 extra-familial) and 20 non-offenders. Results indicated that sexual offenders, both intra-familial and extra-familial, made significantly more errors than non-offenders when reporting Target 2 objects after correctly identifying the Target 1 object as a child. The authors also examined the ability of an outcome score (total Child Category Target 2 score minus total Animal Category Target 2 score in the first condition, contingent upon Target 1 accuracy) to classify sexual offenders from non-offenders. An area under the curve (AUC) analysis of the outcome score produced a significant AUC of .75,
suggesting a value of .5 SD above the mean of the outcome score for the non-offender sample as a cut-off, was sensitive enough to correctly identified 74% of the sexual offender sample while indicating 30% in the non-offender sample as false positives.
Viewing time (VT). Harris et al (1996) investigated the ability of a VT measure
to assess sexual interest in 26 sexual offenders (child molestation) and 25 community volunteers. Viewing time (VT) measures of sexual interest assess how long an individual takes to view an image of a single person, which may indicate the sexual attractiveness of the stimulus. This methodology, also incorrectly reported as “visual reaction time” (Abel et al., 1998; Abel et al., 2004) uses real or computer-modified images comprised of males and females at various ages and in various poses ranging from non-sexual to erotic. Although VT research is similar to reaction time research, that is, there is a specific task to complete within a structured environment, in VT there is no correct answer as in traditional reaction time research, nor are there instructions regarding speed or accuracy. Participants, therefore, are not "reacting," but rather responding to a task at their own pace (Maletzky, 2003).
Harris et al. randomly presented 70 slides (after 20 practice trials) comprised of seven categories (10 slides each of neutral, children between 5 and 8, pubescent, and adult, the latter three categories were nude males and females). The authors, using the latencies from the first task, calculated a VT deviance index by subtracting the longest mean VT for a child or pubescent image from the longest mean for an adult image. Results demonstrated that the deviance scores significantly discriminated between sexual offenders and the comparison group. Interestingly, the comparison group had
Quinsey, Ketsetzis, Earls, and Karamanoukian (1996) conducted two studies investigating viewing time as a measure of sexual interest with 24 male and 24 female university undergraduates. In Study 1, the authors presented 36 slides, comprised of an equal number of nude male and female models in three age categories: adult, pubescent, and child in two random orders. The images were full frontal view and non-sexual (not flirtatious poses, no erections). The authors calculated the VTs from the second task and computed averages for each of the six categories. Results indicated (a) male participants viewed slides of females significant longer than males, with adult females the longest, times decreasing with image age (similar pattern for male images), (b) females viewed slides of males longer than females, adults the longest and again times decreased with image age, but they viewed all slides of females equally long. In addition, adults viewed their preferred sexual interest categories significantly longer than all other categories of images.
In Study 2, the authors presented18 slides depicting nude males and females in a variety of frontal but non-provocative poses in four age categories (>7, 8-12, 13-17, 18+) to 24 heterosexual men from the community. First, half of the participants viewed the 18 slides and rated them for sexual attractiveness while the other half completed a penile plethysmography (PPG) assessment for the same 18 slides. The groups then switched tasks. The authors computed average VTs for each of the categories. Results indicated that participants viewed the adult female images significantly longer than all other female categories and females in general longer than male slides. The VTs for 12 of the 24 participants significantly and positively correlated with PPG scores, and the VTs for 17
of the 24 participants correlated significantly and positively with their rates of image attractiveness.
Abel, Huffman, Warberg, and Holland (1998) examined VTs and sexual interest in 156 sexual offenders who had offended against children. The authors assessed the VTs of 42 slides; seven slides in each of six categories, including males and females aged 8-10, 14-17, and over 22 years of age. This test set was part of a larger stimulus set, the
Abel Assessment for Sexual Interest (AASI), that includes clothed targets of
exhibitionism, voyeurism, frottage, a suffering female, a suffering male, two males hugging, two females hugging, and a male and female hugging, as well as landscapes. Three categories of mean VTs correlated significantly to corresponding classifications of sexual interest (by the author’s screening instruments): male and female child, and male adolescent. In addition, results indicated VT correctly identified 38.5% of those offenders with male child victims and 67.4% of those with female child victims. False positives ranged from 4.4% (male child) to 20.8% (female adolescent). This represents a standard with which to compare other VT results, such as the ones described later in this study.
Letourneau (2002) examined VT and sexual interest with 59 offenders convicted for hands on sexual offences from a Level III military prison (terms of 5 years or more or status as an officer) using a similar methodology to Abel et al. (1998). Results indicated significant correlations between untrimmed (no outlier manipulation) and trimmed VTs (excluded longest VT in each category) for young males and female adolescents and applicable true sexual interest categories. VTs and PPG scores for young males, young females, and adult females were significant correlated: the latter with untrimmed VTs only. Using the rule-of thirds method, “Clinicians are to isolate the space between the
lowest score and the highest adolescent or adult score. They are to then divide that space into thirds and consider any child category score VT that crosses the first third as a category that represents probable deviant sexual interest” (Fischer & Smith, 1999, p. 199). The author found that portions of the AASI statistically able to identify sexual offenders with young male victims but not other groups.
Abel and colleagues (2004) examined the VTs from 16 categories of slides (from a larger stimulus set), divided equally between clothed Caucasian and African American, males and females, and ages 2-4, 8-10, 14-17, over 21. The authors selected the
participants from a database of adolescent clients who completed the AASI between 1994 and March, 2003. They grouped the adolescent offenders into two categories: those who admitted to sexually assaulting someone five years younger (1,704 participants) and those who did not admit, were not accused, and whose therapist did not believe they sexually assaulted someone five years their junior (534 participants). The authors found that mean viewing times to slides of children differed significantly between those who had at least one victim five years younger than them and those who did not. The authors reported an AUC of .64 when discriminating the two participant samples.
Gress (2005) investigated classification schemes produced by viewing time scores and past sexual behaviour. Gress presented 64 computer-modified images comprised of clothed and unclothed males and females of various ages (5, 9, 13, and adult) to 26 adult male sexual offenders referred to an adult community outpatient clinic for treatment. Gress (2005) found VT correctly classified 16 out of 19 or 84% of sexual offenders with child victims, 4 out of 7 or 57% of sexual offenders with adult victims. Results classified incorrectly three out of seven or 43% as having a sexual interest in children (false
positives). In addition, VT correctly classified 16 out of 17 or 94% as heterosexual and 5 out of 9 as bisexual.
Worling (2006) investigated VT with Affinity 1.0, a VT measure developed to assess sexual interest among learning disabled offenders (Glasgow, Osborne, & Croxen, 2003). The author administered Affinity to 78 adolescent males 52 of whom committed offences against someone younger than 12. Affinity 1.0 displays in a fixed random order clothed images of 28 females and 28 males in four age categories: toddlers,
preadolescents, adolescents, and adults. Raw scores are collapsed into categories,
converted to z scores, and then a deviance index is calculated by dividing the VT z score for child images by the VT z score for the adult images. Cronbach alphas estimating internal reliability of the categories ranged from .62 for female adolescents to .82 for female toddlers and male preadolescents. Similar to Letourneau’s (2002) results,
Affinity’s deviance index was capable of statistically discriminating adolescents with a
male child victim from adolescents with victims from all other age/gender groups.
Summary
The role of cognition, particularly the aspects of appraisal, recognition, attention, and emotional saliency, appears to play a significant role in sexual arousal. Recently, research in this area began to switch from treating sexual arousal as a unified construct to models based on the information processing approach, focusing on cognitive changes, that is, changes over time in thought patterns, memory recall, and attention(Janssen et al., 2000) . This facilitates investigations into the individual differences found in many studies sexual arousal and interest.
