Citation
Social information processing in child molesters : an examination of decoding skill and cognitive factors

Material Information

Title:
Social information processing in child molesters : an examination of decoding skill and cognitive factors
Alternate title:
Examination of decoding skill and cognitive factors
Creator:
Medlin, Julie C
Publication Date:
Language:
English
Physical Description:
vii, 133 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Adults ( jstor )
Child abuse ( jstor )
Cognition ( jstor )
Criminals ( jstor )
Decryption ( jstor )
Empathy ( jstor )
Inmates ( jstor )
Repression ( jstor )
Sex offenders ( jstor )
Social desirability ( jstor )
Child Abuse, Sexual ( mesh )
Cognition ( mesh )
Cognition Disorders ( mesh )
Department of Clinical and Health Psychology thesis Ph.D ( mesh )
Dissertations, Academic -- College of Health Related Professions -- Department of Clinical and Health Psychology -- UF ( mesh )
Mental Processes ( mesh )
Sexuality ( mesh )
Social Perception ( mesh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1995.
Additional Physical Form:
Also available online.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Julie C. Medlin.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Julie C. Medlin. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
027964082 ( ALEPH )
49016021 ( OCLC )

Downloads

This item has the following downloads:


Full Text














SOCIAL INFORMATION PROCESSING IN CHILD MOLESTERS: AN EXAMINATION OF DECODING SKILL AND COGNITIVE FACTORS









By

JULIE C. MEDLIN














A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1995












ACKNOWLEDGEMENTS


I would like to thank Jacquelin Goldman, Ph.D., ABPP, my dissertation chair, for supporting me in pursuing my interest in research with a particularly difficult forensic population. She encouraged me to follow my true research interests rather than design a dull but easy, research study. I also thank her for helping me make some difficult design decisions during the data collection process, and then later in the statistical analysis process. Overall, she allowed me to develop into an independent researcher, as she gave me the freedom to be creative and encouraged me to take risks. As a result, I have truly enjoyed and have learned from my research pursuits in graduate school.

I would also like to thank my dissertation committee: Michael Peri, Ph.D., James Johnson, Ph.D., James Algina, Ph.D., and Otto von Mering. I offer my thanks for their patience during my proposal meeting when I explained entirely too many hypotheses and a complex design. I appreciate their willingness to allow me to venture into such a difficult project potentially fraught with many practical problems, given the subject population and prison environment.

The Florida Department of Corrections also deserves enormous gratitude for allowing me to conduct research in their prison system. In particular, I would like to thank William Bales, Ph.D., Chief of the Bureau of Planning, Research, and Statistics for the Florida Department of Corrections. He granted me permission to conduct my research at various


ii







institutions within the DOC system. I would also like to thank Nancy Haynes, Ph.D, Senior Psychologist; William Cook, Human Services Program Director, Joe Thornton, M.D., Medical Executive Director, and Dennis O'Neill, Superintendent at Union Correctional Institution. They helped make my research a reality by offering to let me conduct research at their institution. Without their help, I simply would not have been able to conduct this research!! In particular, Nancy Haynes, Ph.D., and William Cook were incredibly generous and giving in welcoming me to UCI and helping me get my data collection started. I cannot expressjust how thankful I am for their help! I would also like to thank Don Kimbrell, Ph.D., at the New River Correctional Institution for facilitating my entrance into the New River East and West institutions In addition, I send my thanks to the staff at the Marion Correctional Institution.

I also extend my thanks to my dedicated research assistants: Jennie Shenick, Nicole Mahoney, Barry Kays, Laura McCoy, and Mike Britt. They were priceless assistants as they were flexible in their schedules and their attitudes as I bounced them from prison to prison, assigning difficult and varying tasks. I was especially grateful for their patience as they had to keep up with my crazy, breakneck schedule. I would not have been able to finish on time without their hours of tireless work!

Last, but certainly not least, I would like to thank my loving husband who has been put through the wringer while I have been driven to finish my dissertation. He has been patient and understanding while I have maintained an unforgiving schedule and have been completely obsessed with my research. I would not have been able to get through this difficult time were it not for his reassurance and unwavering support!


iii













TABLE OF CONTENTS



ACK N OW LED GEM ENTS ............................................................................................

A B STR A C T ................................................................................... ........ ......... v

CHAPTERS

1 INTRODUCTION AND REVIEW OF LITERATURE ................................ 1

Social Skills of Child M olesters ........................................ .... ................... 1
Cognitions of Child Molesters .................................... .......................6
Subtypes of Child M olesters .................................................................... 10
Other Variables Related to Decoding Skill and Cognitions ................... : .........16
Rationale for the Proposed Study.........................................20
H ypotheses ..........................................................23
Alternative Hypotheses............................. ....................24

2 M ETHODS ......................................................26

Subjects .......................... .......................... ...26
Research Design .......... ...... ....................... 27
Dependent Variables....................... ........ .. .........28
Procedures ...................................................................................................39

3 RESULTS ......................................................41

Background V ariables ........................................ ........................ 41
M ethod of A nalysis .................. .. .............................................. ......... 44
Analysis of Decoding Skill ......................... .......................46
Analysis of Cognitive Distortions .............................. .......... ..........47
Comparison of Child Versus AdultVignettes..............................................66
Analysis of Em pathy................................................................................67
Analysis of Repression...............................................68

4 DISCUSSION ..................................................110

Summary of Findings ............................... ..... 125


iv







C onclusions ................................................................ 125

REFEREN CE S .............................................................. 127

BIOGRAPH ICAL SKETCH ................................................................................. 133












Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy



SOCIAL INFORMATION PROCESSING IN CHILD MOLESTERS: AN EXAMINATION
OF DECODING SKILL AND COGNITIVE FACTORS By

Julie C. Medlin
December 1995


Chairperson: Jacquelin Goldman
Major Department: Clinical and Health Psychology


This study examined social information processing in heterosexual child molesters.

More specifically, this study examined child molesters' decoding skill and cognitions regarding sexual contact with children. The literature has indicated that child molesters are socially deficient; however, the nature ofthis deficiency is unclear. That is, this deficiency may result from any number of problems in their cognitive operations, including deficiencies in decoding skill. Thus, the present study examined child molesters' decoding skill, or ability to "read" others' cues. Contrary to expectations, child molesters did not show a deficit in decoding skill or social perception when compared with control groups. Consistent with expectations, child molesters showed cognitive distortions regarding sexual contact with children. All four sex offender groups (nonfamilial admitter, nonfamilial denier, incestuous admitter, incestuous


vi







denier) prescribed significantly less punishment and treatment for a child molester than did the control groups. Most of the other cognitive distortions, however, were found among nonfamilial admitters rather than among the other sex offender groups. When reading vignettes portraying sexual interaction with a female child, nonfamilial admitters attributed significantly more benefit, responsibility, desire, and enjoyment to the child than the other sex offender groups and control groups. As expected, incestuous offenders differed significantly from nonfamilial offenders in that they scored more similarly to controls. In addition, sex offenders who denied having committed their sexual crimes generally endorsed fewer cognitive distortions than admitters and showed more social desirability than admitters. Contrary to expectations, child molesters did not differ from controls in their ratings of anxiety in a sexual situation with an adult female.

Other variables examined included empathy, repression, and abuse history. Contrary to predictions, child molesters did not score significantly lower on empathy than controls. There was also the unexpected finding that nonfamilial offenders scored higher than incestuous offenders on empathy. There were no differences among groups on repression. Consistent with expectations, more child molesters than controls reported having been physically or sexually abused as a child, as well as more admitters than deniers. In sum, significant differences were found between incestuous and nonfamilial offenders, as well as between admitters and deniers.








vii












CHAPTER 1
INTRODUCTION AND REVIEW OF LITERATURE


Child molestation is a major social problem which has only recently begun to receive adequate research attention over the past decade (Knight, Rosenberg, & Schneider, 1985; Abel, Rouleau, & Cunningham-Rathner, 1985; Marshall, Laws, & Barbaree, 1990). Despite these research efforts, few theories have emerged to explain this phenomenon. Two areas of potential relevance to theory construction include child molesters' social perception and their cognitions about sexual interaction with children Research in these areas suggests a deficiency in child molester's social information processing, and highlights the need for a more focused investigation of these variables and their potential relationship.

Social Skills of Child Molesters

There have been numerous studies focusing on the social competence of child

molesters (Segal & Stermac, 1989). The social deficit model of child molestation suggests that child molesters lack social skills and, as a result, have unsatisfactory heterosocial relationships (Groth & Birnbaum,1979). The clinical literature has frequently characterized child molesters as having impaired social relationships and discomfort with adult sexuality (Hobson, Boland, Jamieson, 1985). More specifically, Finkelhor (1984, 1986) proposed a four-factor model of child molestation postulating that child molesters turn to children partly because alternative sources of sexual gratification have been blocked or inhibited.




2


Segal and Marshall (1985) conducted one ofthe first studies documenting a social skill deficit in child molesters The authors compared rapists, child molesters, non-sexual offenders, and community control subjects on social skills measures of videotaped conversations with females. Although all incarcerated groups were less socially skilled than the community control group, the child molesters also demonstrated a skill deficit relative to the rapists and inmate control group.

In a similar study using behavioral assessment of controlled social interaction, both rapists and child molesters were significantly less skilled than community control subjects on measures of form of conversation and general level of social skill (Overholser & Beck, 1986). Although child molesters did not show a social skills deficit relative to the rapists in this study, they did show a trend toward lower assertiveness than rapists and controls on a questionnaire assessment of assertiveness. In addition, the child molesters displayed significantly more fear of being negatively evaluated than did rapists and controls.

Segal and Marshall (1985b) also found that child molesters showed more heterosocial inadequacy than rapists. Compared to the rapist group, child molesters rated themselves as less skilled and as more anxious in a heterosexual interaction with a confederate. They also scored significantly higher than rapists on the Social Avoidance and Distress Scale (SADS). The authors note that this social inadequacy appears to be largely cognitive, given that the child molesters were rated as equally competent in the social situation as the other groups by both the confederate and independent judges. Consistent with the results of the Overholser and Beck (1986) study, child molesters rated themselves as less assertive in accepting positive and negative feedback from others and scored significantly lower than rapists and controls on the




3

general assertion subscale on the Callner-Ross Assertiveness Scale. In a follow-up study, Segal and Marshall (1986) investigated social perception in child molesters in heterosexual situations by having them interact with a female confederate. Compared with other sex offenders and community controls, child molesters were found to be significantly worse at prediction and evaluation of their own performance. Taken together, these findings suggest that child molesters perceive themselves as less socially skilled and assertive than others, and are more anxious than others about their social interaction. Given the inconsistent results, it is unclear whether or not child molesters are actually less socially skilled than rapists and non-sexoffending inmate controls.

Although the literature generally indicates lower social competence among child molesters, there is currently no specific theory or clear pattern of deficits regarding child molesters' social information processing. Their interpersonal functioning may be affected by a number of factors including accuracy of processing social cues and information. In addition to appropriate response skills, "ability to accurately read the social environment," or decoding skill, is necessary for social performance (Morrison & Bellack, 1981, p.70). Decoding skills refer to "the afferent processes involved in the accurate reception, perception, and interpretation of incoming sensory information (McFall, 1990, p. 313)." Ifa man does not receive a woman's cues or misperceives or misinterprets these cues, then his social behavior toward the woman is more likely to be inappropriate. Based on his social information processing model, McFall (1990) suggests that the logical place to begin exploring social information processing is at the beginning of the input process with decoding skills. This is the point where the information input is more or less equivalent for everyone.




4

Sexual assaulters in general have been hypothesized to have difficulty processing social information from women as well as show a tendency to misinterpret social messages. Lipton, McDonel, and McFall (1987) investigated this hypothesis, comparing rapists, violent nonrapists, and nonviolent nonrapists on assessments of heterosocial cue-reading accuracy in intimate and first-date situations. Rapists were found to miscontrue women's cues, particularly cues involving negative moods in first-date situations, suggesting a deficit in their social information processing There appears to have been no such study conducted with child molesters examining their ability to read others' cues. In fact, there have been few studies of cognitive operations in sexual offenders in general. Cognitive operations refers to the various ways in which the information processing system operates, including attention allocation, encoding, decoding or control processes.

Such studies are needed since child molesters' social skill deficiencies and cognitions may be related to a difficulty in reading others' cues. It has been noted clinically that child molesters misidentify emotions (Hobson, Boland, & Jamieson, 1985). Another reason to expect that child molesters might be generally deficient in this skill is that low social competence (Feldman & Rime, 1991) and an abuse history (Camras, Grow, & Ribordy, 1983) are often associated with deficiencies in decoding. The latter variable is important to account for since many child molesters (estimates range from 25% to 75%) report a history of abuse (Mayer, 1988; Hamilton & Ondrovik, 1993) and those who report abuse may comprise a distinct clinical subgroup (Becker, Kaplan, & Tenke, 1992).

In addition, child molesters are often characterized as lacking empathy (Hobson,

Boland, & Jamieson, 1985), and ability to read others' cues is believed to be a key component







of empathy (Hall, 1979; Riggio, Tucker, & Coffaro, 1989; N.D. Feshbach, 1978). That is, one can experience true empathic concern only if he or she is nonverbally sensitive in order to vicariously experience the other's emotional state (Mehrabian & Epstein, 1972; Riggio et al., 1989). Thus, child molesters' clinically noted lack of empathy may result from a deficiency in decoding and may also contribute to the commission of the offense by serving as a disinhibitor. In addition, a decoding deficiency might also be related to cognitive distortions found among child molesters (Stermac & Segal, 1989).

According to McFall's (1990) social information processing model, the next two steps following the decoding process are the decision skill process and the enactment skill process, respectively. Decision skills involve "generating response options, matching these to task demands, selecting the best option, searching for that option in the behavioral repertoire, and evaluating the subjective utility of that option's likely outcomes relative to the likely outcomes of other options (McFall, 1990, p. 313)." Enactment skills then involve carrying out the behavioral option selected. Inappropriate social behavior may result from a problem at any of these three levels.

A study conducted by Barbaree, Marshall, and Connor (1988; as referenced in

Marshall, Laws, & Barbaree, 1990) suggests that child molesters may not have difficulty with decision skills, but may have problems with enactment skills. This study assessed child molesters' social problem-solving skills in social situations involving children and adults. Child molesters were found to be as skilled as nonoffenders in recognizing problems and generating alternative solutions, but tended to choose socially unacceptable solutions. They also failed to recognize the likely negative outcomes of these solutions. The authors suggest that these




6


results may point to a deficit in the cognitive processing that precedes social behavior Clearly, there needs to be more research focusing on the nature of this deficiency in child molesters' social information processing.

Coganitions of Child Molesters

Possible deficiencies in social information processing may be related to distorted cognitions reported among child molesters (Stermac & Segal, 1989). In contrast to the plethora of research conducted on child molesters' social skills, very little research has been conducted on child molesters' cognitions, although cognition is believed to play an important role in any form of human sexual arousal (Rook & Hammer, 1977). Most of the studies focusing on child molesters' cognitions have focused on cognitive products that are thoughts or images that form as a result of input of information and the interaction of cognitive structures, propositions, and operations. Examples of cognitive products include self-statements, attn'butions, and inferences.

Investigation of child molesters' cognitions is important as cognitions may play a

number of roles in the commission of a sexual offense. For example, specific cognitions may contribute to the development of adults' sexual interest in children or they may be post hoc rationalizations that faciliate continued offending It is also possible that these cognitions are an epiphenomena of deviant sexual arousal which would disappear once arousal was eliminated (Stermac & Segal, 1989).

One of the first studies to examine cognitive factors among child molesters was

conducted by Frisbie, Vanasek, and Dingman (1967). Using the semantic differential, these authors investigated child molesters' self view using ratings from institutionalized and




7

community-based child molesters on their actual and ideal views of themselves. The measure consisted of pairs of bipolar variables which reflect aspects of an individual's personality traits, physical characteristics, and value orientations. Subjects located themselves on these dimensions for their actual versus ideal selves. The largest discrepancies between the two ratings were found on adjectives related to interactions with others. Although the results suggest that child molesters feel less socially competent than they desire, comparisons with nonoffender controls cannot be made since this group was not included in the study.

A later study on cognitive factors (Howells, 1978) investigated heterosexual child molesters' perceptions of adult females, adult males, and children using the repertory grid technique. The results indicated that child molesters employed more constructs related to dominance in their descriptions of both children and adults. Children were perceived as being at the nondominant pole, as "undemanding" and "passive" while adults were viewed as more "domineering" and "demanding." The results also suggested that child molesters may be attracted to children for more than sexual reasons, such as for their innocence and capacity for unconditional love. Unfortunately, these results are only suggestive given the small sample size of ten child molesters, problems in the statistical analysis, and the failure of a follow-up study (Horley, 1988) to replicate these findings.

A more recent study (Abel et al., 1989) has focused on child molesters' cognitions concerning child molestation. This study used a 29-item Likert scale and found 6 reliable factors covering the general area of harm caused to the child by the molestation. The factors successfully separated the child molesters from the control group, suggesting that child molesters show cognitive distortions minimizing the harm of the molestation.




8

In a similar study, Stermac and Segal (1989) tested the hypothesis that child molesters' cognitions are more permissive and accepting of sexual contact with children. They used written vignettes describing sexual contact between an adult and child. The vignettes contained varying degrees of sexual contact (touching, fondling, no clothes, ejaculation) and varying responses from the child (smiling, passive, crying). After reading each vignette, child molesters and controls responded to questions assessing their view of the adult's and child's behavior.

The results indicate that child molesters ascribed more benefits of the molestation to the child than did the controls. Unlike the controls, the child molesters did not significantly decrease their amount of ascribed bnefit to the child as the amount of sexual contact increased. The child molesters also differed significantly from the controls in their perceptions of benefit based on the child's response, although their response was similar to that of the controls when the child's response was negative. Compared with the controls, the child molesters attributed more responsibility to the child and less overall responsibility to the adult. In general, the child molesters' ratings were less affected by the amount of sexual contact or the child's response.

The child molester group also showed significant differences from controls on the

Cognition Questionnaire (Abel, Becker, & Cunningham-Rathner, 1984), endorsing more items indicative of permissive beliefs and attitudes towards adult sexual contact with children. These results are consistent with clinical reports of child molesters' distorted cognitions regarding sexual contact with children (Abel, Becker, & Cunningham-Rathner, 1984; Abel, Mittleman, & Becker, 1985; Finkelhor, 1984).




9

Although the results suggest a difference in cognition, it is unclear if these cognitions or schemata develop as post hoc rationalizations for their behavior or if they were formed before the behavior. Abel et al. (1984) suggest that some offenders who find themselves sexually attracted to children respond to this dilemma by modifying their cognitions in order to justify their deviant behavior. These cognitive distortions may also be viewed in terms of Fmkelhor's (1984, 1986) theoretical model of child molestation which describes various disinhibitors such as alcohol use or psychosis which enable child molesters to overcome their internal inhibitions. Thus, cognitive distortions may be viewed as another type of internal disinhibitor (Hollin & Howells, 1991).

The results from the Stennac and Segal (1989) study suggest that the child molesters were less sensitive or responsive to the child's cues than were the controls, suggesting a lack of empathy. As the authors indicated, child molesters' perceptions may be modified only in response to a clearly negative and unambiguous response such as the child's crying It is important to note that the vignettes in this study were written and did not require the subjects to "read" or interpret others' cues. It is likely that in real life situations, the child's cues would probably not be as clearly portrayed as they are in a written vignette. Since child molesters tend to interpret ambiguous responses on the child's part as positive within the structure of a simple written vignette, they may also have difficulty reading others' ambiguous nonverbal cues in real life situations. It may be that child molesters have difficulty accurately reading others' cues and that this deficiency in information processing contributes to permissive cognitions.




10

Subtypes of Child Molesters

Nonfamilial Versus Incestuous Child Molesters

Unfortunately, many of the studies cited above did not subdivide their samples of child molesters based on gender of the victim or relationship to the victim. This tendency to lump many types of sex offenders into one general category has been cited as one of the potential reasons for inconsistent results (Howells, 1981). A variety of studies over the years have demonstrated the need to differentiate between heterosexual and homosexual offenders, and between nonfamilial and familial offenders. Howells (1981) offers the conceptualization that child molesters fall on a continuum, ranging from pedophiles with a primary sexual preference for children, to persons with a primary sexual preference for adults. Most reseachers in the area agree that not all child molesters are pedophiles. Thus, some theorists propose a distinction between molestation based on situational factors versus molestation based on sexual preference. For example, Pawlak, Boulet, and Bradford (1991) suggest that some child molestation may be related to situational factors interacting with various types of psychopathology.

Many theorists view incestuous offenders as situational offenders and nonfamilial offenders as more sexual preference offenders (Pawlak, Boulet, & Bradford, 1991). Thus, incestuous offenders' deviant sexual behavior is thought to be less related to a primary sexual interest in children and more related to taking advantage of easy sexual access. Consistent with this view, incestuous offenders showed more age appropriate preferences and recidivated at lower rates than nonfamilial child molesters (Marshall, Barbaree, &







Christophe, 1986). Abel et al. (1981) found, however, that incestuous offenders showed sexual arousal to stimuli depicting sexual contact with children.

Pawlak, Boulet, and Bradford (1991) found differences between incestuous and

nonfamilial offenders on three of the subscales on the Derogatis Sexual Functioning Inventory. Incestuous offenders scored higher than nonfamilial offenders on sexual experience and sexual satisfaction. The authors note, however, that this difference is difficult to interpret since these subscale scores are based on factors which may be largely affected by involvement with a regular sexual partner, and unfortunately, marital status differed greatly across the two groups. As expected, incestuous offenders tended to be married while the majority of nonfamilial offenders were single. In addition, nonfamilial offenders scored higher than incestuous offenders on sexual fantasy, meaning that the nonfamilial offenders endorsed more items depicting common sexual themes. This higher score on sexual fantasy may reflect a greater need among nonfamilial offenders to resort to fantasy since their sexual experience and sexual satisfaction are lacking. Although the two groups differed on these three subscales, the subtest scores overall did not effectively discriminate between the two groups.

Marshall, Barbaree, and Christophe (1986) found differences between incestuous and nonfamilial offenders in their erectile responses to sexual stimuli. Compared to incestuous offenders and nonoffenders, the nonfamilial offenders showed greater arousal to the stimuli involving children. Incestuous offenders, on the other hand, showed less arousal to adult female stimuli than the other two groups. In a later study, Barbaree and Marshall (1989) found that incestuous offenders showed either a non-discriminating sexual arousal profile or an adult preference pattern while the nonfamilial group was more




12


heterogeneous, displaying a variety of profiles ranging from nondiscriminating, to child or adult preferences. Consistent with these results, a recent study found that incestuous offenders reported less sexual interest in children than did nonfamilial offenders on the Pictorial Sexual Interest Card Sort (Haywood and Grossman, 1994).

Differences have also been found between the MMPI profiles of incestuous versus nonfamilial offenders (Erickson et al., 1987). In general, the incestuous offenders showed a lesser degree of disturbance than the nonfamilial offenders. Among offenders who were the biological fathers, 4-3 was the most common profile; however, this code type was equally prevalent among the control groups. The authors suggest that the 4-3/3-4 profile describes "chronic anger, overcontrolled hostility, marital discord, and passive-aggressive personalities," with the 4-3 characterized by more overt acting-out while the 3-4 is characterized by more passivity (p.568).

Among incestuous stepfathers 4-7/7-4 was the most common profile. The authors associate this profile with "insensitivity to others, brooding interposed with episodic acting-out, and alcoholism" (p.568). For the nonfamuilial offenders, the 4-2/2-4 and 4-8/8-4 profile was more common than among the incestous offenders. The 2-point peak code for the mean profile of the child molesters was a 4-2. This same code has been found in several other studies and is considered to be consistent with the clinical profile portraying child molesters as passive, dependent, socially uncomfortable, and impulsive (Erickson et al., 1987).

Although these studies have yielded somewhat inconsistent results, the majority of the studies have found that incestuous offenders appear more similar to controls than do




13


nonfamilial offenders. Thus, one might expect incestuous offenders to show fewer cognitive distortions than their nonfamilial counterparts regarding sexual contact with children.

Admitters Versus Deniers

Another distinction among child molesters which is often overlooked in the research is whether or not the alleged offender admits to having committed the crime. Many studies have focused on those who admit to their crime, however, the majority of sex offenders deny culpability (Salter, 1988; French, 1988; Rogers & Dickey, 1991). Thus, the sample of offenders being studied is not representative of the majority of offenders. In addition, studies which combine admitters and deniers may yield confusing results since several recent studies have shown that these two groups differ in significant ways (Becker, Kaplan, & Tenke, 1992; Haywood, Grossman, & Hardy, 1993; Haywood & Grossman, 1994).

Becker, Kaplan, and Tenke (1992) found significant differences in the erectile

response profiles of admitters and deniers. They found that a disproportionate number of deniers failed to show a significant sexual response to cues depicting any age category (nonresponders). While only 18% of the admitters were classified as nonresponders, 58% of the deniers fell in this category. Although the deniers tended to be less responsive in general, they were maximally responsive to child cues. Admitters, in contrast, were more responsive to peer cues, but also responded to child and adult cues. Forty-two percent of the admitters showed peer or adult profiles while only 11% of the deniers were in these categories. Deniers were significantly less responsive to adult cues than admitters, but




14


were not less responsive to child cues. Thus, deniers tend to show a different pattern of sexual responsivity, suggesting that they may constitute a clinically distinct subgroup. Given these differences, deniers and admitters may also show different response patterns on other variables such as on cognitions and social perception. It is interesting to note that the adolescent sex offenders in the Becker et al. (1992) study showed the greatest sexual response to age-appropriate female cues, supporting the hypothesis that child molesters would prefer to have a sexual relationship with a peer, but may turn to younger females because of difficulty with initiating or maintaining a sexual relationship with an adult female. This finding is consistent with social skills research and anecdotal information on child molesters (Marshall, Laws, & Barbaree, 1990) which suggests that sexual assaulters have difficulty forming and maintaining healthy intimate relationships with women because of their lack of social competence.

Haywood, Grossman, and Hardy (1993) also found significant differences between admitters and deniers. Deniers showed more minimization than admitters on the FakeGood Scale on the Sixteen Personality Factors Inventory while admitters showed more exaggeration of problems on the Fake-Bad scale. Both scales were able to correctly discriminate between admitters and deniers. Deniers were more likely to be defensive about undesirable personality traits than were admitters. Interestingly, the personality factor which was most highly correlated with the Fake-Good scale was emotional sensitivity. This scale is associated with "affective maturity, lack of anxiety, and ability to




15

deal with frustration (p.187)." Based on this finding, one might expect deniers to score higher on empathy in the current study, in an effort to present themselves as emotionally sensitive.

In a more recent study, Haywood and Grossman (1994) found that deniers

reported significantly less sexual interest in children than did admitters or normal control subjects. Deniers also reported more sexual interest in women than did admitters. However, both admitters and deniers reported significantly less sexual interest in adult women than the controls. Among the child molesters, low self-reported sexual interest in children was significantly correlated with minimization ofpsychopathology on the F-K index and the positive malingering scale. The authors conclude that, for the sex offender population, subjective measures of sexual interest are "highly susceptible to distortion or minimization of deviant sexual arousal (p.338)." Abused Versus Nonabused Subiects

In addition to differences between admitters and deniers, Becker et al. (1992) also found differences between abused and nonabused groups. Subjects who reported a history of physical or sexual abuse were more sexually responsive in general than subjects with no reported history. Abused subjects also showed a greater proportion ofnondiscriminator profiles. These findings suggest the need to differentiate child molesters based on abuse history as well as by denial.

Overall, there has been little research in the area of cognitive factors in child

molestation, with only one empirically sound study to date (Stermac & Segal, 1989). Given




16


this lack of research, the role of these cognitions remains unclear, particularly in regards to how they interact with other variables to facilitate the commission of the offense (Quinsey, 1986).

Other Variables Related to Decoding Skill and Cognitions



Prior to the past decade, child molesters' denial and defensiveness was viewed as related to massive repression and unconscious denial (Rogers & Dickey, 1991). Rogers & Dickey (1991) suggest that the pathogenic model was largely discarded in the past decade, however, as a result of growing disenchanment with psychoanalytic theory and its problems with clinical efficacy rather than on the basis of research findings. The results of the study by Becker, Kaplan, and Tenke (1992), however, raise the possibility that repression may play a role in child molesters' denial since the deniers were sexually nonresponsive to all sexual cues presented. Although it is possible that the deniers were able to consciously control their sexual arousal, repression might better account for these deniers' ability to suppress their sexual impulses to prevent them from physiologically responding to sexual cues. That is, the denial may involve more than a conscious attempt to deny the crime in order to avoid perceived negative legal consequences; instead, or in addition, it may involve an unconscious defense mechanism employed-as a means of controlling anxiety and guilt.

People who use repression as a primary defense mechanism often "distort their own self-evaluative responses as a way to make themselves more comfortable, to block perceptions of behavioral outcomes, and to avoid disturbing or uncomfortable implications in a variety of situations (Dahstrom, p. 104)." Thus, repression might help the child molester deny the





17


seriousness of his offense, and may facilitate the development and maintenance of cognitive distortions to rationalize or justify the behavior. Assuming that deniers would show more repression than admitters, one might expect them to show more cognitive distortions regarding sexual contact with children, however, as several studies suggest (Becker et al., 1992; Haywood et al., 1993; Haywood & Grossman, 1994), deniers appear to be effective at suppressing both deviant sexual responses as well as deviant thoughts. Thus, deniers would be expected to have more cognitive distortions, but would be less likely to show these distortions, especially on a face valid measure.

In sum, since some form of denial is common among most child molesters, it might be expected that child molesters as a group would show more repression than non-child molesters. The subgroup of deniers might also be expected to show more repression than child molesters who admit to the offense.

Empath

Like repression, empathy is another variable which might be related to decoding skill and cognitive distortions among child molesters. Clinicians have long noted that sex offenders show little empathy for their victims. Many authors have also suggested, however, that this deficit in empathy may be generalized in the sense that it applies toward all people, rather than applying solely to the victim (Marshall, Jones, Hudson, & McDonald, 1993). The few studies that have measured empathy among sex offenders have focused on this concept of generalized empathy, and have yielded inconsistent results.




s18


Two studies reported a generalized deficit in empathy among child molesters. One study reported that 35.8% of child molesters lacked empathy, however, the method for assessing empathy was unclear (Abel, Mittleman, & Becker, 1985). Another study reported significant differences between sex offenders and community subjects using Hogan's Empathy Scale (Rice, Chaplin, Harris, & Coutts, 1990; as referenced in Marshall et al., 1993). Marshall et al. (1993) suggest that the samples used in these studies may not be representative of the general sex offender population, as they consisted of a more deviant and chronic group of sex offenders than is typically found in a general prison population or outpatient clinic. Marshall et al. (1993) also point out the importance of the method used to assess empathy. A variety of assessment instruments measure different and limited aspects of empathy. They cite a study conducted by Seto (1992) which reported that incarcerated rapists showed no empathy deficiency on the Mehrabian and Epstein (1972) measure of empathy, but did show a deficit on Hogan's scale.

In their 1993 study, Marshall et al. used the Interpersonal Reactivity Index (Davis, 1983), citing it as the best available measure for assessing the complex nature of empathy. They found no deficits in empathy among incarcerated child molesters when compared with controls. In contrast, outpatient child molesters did show a deficiency in empathy. More specifically, child molesters scored significantly lower on the total IRI score and on the Fantasy subscale. Unfortunately, the study did not differentiate between familial and nonfamilial offenders or between homosexual and heterosexual offenders, thus introducing a potentially confounding factor. Nonetheless, the authors suggest that the differences among the child molesters may have resulted from motivation among the incarcerated




19


offenders to present themselves in a desirable light in order to facilitate early release from jail. The authors conclude, however, that nonspecific empathy does not appear to be child molesters' main problem. Rather, they recommend focusing on child molesters' empathy toward specific targets such as their victims. Finkelhor & Lewis (1987) also suggest exploration of this more specific type of empathy since "otherwise-empathic men sometimes sexually exploit children. (p.76)"

Given the inconsistent results in the literature, the present study examined generalized empathy among child molesters. Specific empathy for children was not assessed since an assessment instrument of this type could not be found. Child molesters' hypothesized repression was thought to be related to a lack of empathy since repression might allow the offender to block out the victim's response, thus resulting in impaired empathy. Lack of empathy might then facilitate the commission of the offense since empathy is thought to play an important function in the inhibition of aggressive or antisocial acts toward others, with vicarious experience of the victim's pain serving as a deterrent (Miller & Eisenberg, 1988). Miller and Eisenberg (1988) conducted a metaanalysis of studies addressing this issue and found empathy to be negatively related to aggression and antisocial, externalizing behaviors.

Empathy has been described in the clinical literature as both a cognitive and affective response (Miller & Eisenberg, 1988). Empathy has been defined as the ability to take the viewpoint of the other person and to respond affectively to the other's perceived emotional experience (Lorr, Youniss, & Stefic, 1991). Since empathy involves achieving a cognitive understanding of the other person's feelings, a lack of empathy might manifest itself in cognitive





20


distortions about the other person's feelings. Thus, the cognitive distortions seen in child molesters could be related to a lack of empathy as well as repression, both of which could be related to poor decoding skills.

Child molesters' perceived lack of empathy could also be related to their reported

histories of abuse. Miller and Eisenberg (1988) concluded from their meta-analysis that abused subjects tend to have lower levels of empathy than nonabused subjects. Abused children are thought to have little opportunity to learn to identify others' affective states or experience empathic responding themselves since their needs and feelings are not recognized or appropriately responded to in interactions with their abusive parents. Thus, abused children tend to lack models of appropriate behavioral and emotional responding to others in need, and as a result, show deficiencies in empathy which remain throughout adulthood.

In sum, child molesters' perceived lack of empathy was thought to be related to their history of childhood abuse, repression, and poor decoding ability. In addition, lack of empathy may play a role in their cognitive distortions regarding sexual contact with children.

Rationale for the Proposed Study

This study focused on two aspects of child molesters' social information processing: decoding skill and cognitive distortions. The literature has indicated that child molesters are socially deficient, however, the nature of this deficiency is unclear. That is, this deficiency may result from any number of problems in their cognitive operations, including deficiencies in their decoding skill. Since the first step in exploring deficiencies in social information processing





21


begins with an examination of decoding skill (McFall, 1990), the present study examined this variable among child molesters.

Given child molesters' low social competence and the recent finding of a decoding deficiency found among other sex offenders (Lipton, McDonel, & McFall, 1987), it was expected that child molesters would also show a deficiency in decoding skill when compared to non-sex-offending control groups. In addition to differences between child molesters and controls, differences between admitters and deniers were examined since a preliminary finding suggested that deniers may represent a distinct clinical subgroup (Becker et al., 1992). In addition to examining nonverbal decoding, this study examined child molesters' performance on a more complex social perception task which involves the next step in McFall's (1990) social information processing model decision skills. Child molesters were expected to perform as poorly on this measure as on the decoding task since the social perception task involves decoding as well as decision skills. If the child molesters did not show a deficiency in decoding skills, then this social perception task would allow for an examination of child molesters at the next level of social information processing decision skills.

Since child molesters were expected to show deficits in decoding skill, they were also expected to show deficits in empathy since empathy requires good decoding skills. This notion of a deficit in empathy is consistent with clinical observation of child molesters, and has received some empirical support, although the research in this area has yielded inconsistent findings. Thus, another reason for examining empathy among this group stems from the inconsistent findings in the literature. Unlike previous studies, the current investigation used a more appropriate measure of empathy which assesses a multifaceted notion of empathy, and in





22


addition, examined this variable among subgroups of child molesters familial versus nonfamilial and admitters versus deniers.

The second aspect of the social information processing examined was child molesters' cognitions regarding sexual contact. Part of this investigation involved an attempt to replicate Stermac and Segal's (1989) results, with the modification that child molesters were subdivided into appropriate subgroups. Based on Stermac & Segal's (1989) results, it was expected that child molesters as a group would ascribe significantly more benefit to the child than control groups when the child's response was positive or ambiguous, and would show no difference from controls when the child's response was clearly negative. The child molesters were also expected to attribute more responsibility and desire to the child and less overall responsibility to the adult.

In addition, child molesters were expected to endorse more items on the Cognition Questionnaire which are indicative of permissive beliefs and attitudes toward sexual contact with children. Within the child molester group, deniers were expected to endorse fewer permissive cognitions than admitters on both the vignettes and the Cognition Questionnaire. This was expected since deniers show the most extreme form of denial and thus are not likely to admit to cognitive distortions involving some acknowledgement of a sexual interest in children. In addition, these measures of cognitive distortions tend to be face valid and thus responses to these measures were expected to be highly influenced by social desirability.

In addition to replicating Stermac and Segal's (1989) results, this study investigated child molesters' cognitions regarding sexual contact with adults. In contrast to their expected distorted cognitions regarding sexual contact with children, child molesters were expected to






23

show no differences on these same variables relating to sexual contact with an adult. That is, child molesters' cognitive distortions relating to sexual contact were hypothesized to be largely limited to interactions with children, however, child molesters were expected to rate more anxiety than controls in a sexual situation with an adult female. This rating of anxiety was expected to reflect child molesters' lack ofconfidence and difficulty with adult heterosexual relationships. No differences were expected between admitters and deniers on the adult vignettes.

Social desirability was also examined since denial among sex offenders is thought to be at least partially related to a desire to present oneself in a positive light. Deniers were expected to show more social desirability than admitters, while child molesters as a group were not expected to differ significantly from controls. Repression was also examined since denial was also thought to be partly related to unconscious defense mechanisms. Thus, deniers were expected to show greater repression than admitters. Repression was thought to contribute to permissive cognitions and was expected to negatively correlate with level of empathy since repression might help molesters distance themselves from their sexual offending behavior, thus reducing their ability to empathize with their victims' emotional reaction.



This study examined decoding skill and cognitive factors among incarcerated heterosexual child molesters (admitters and deniers), non-sex offending inmates, and community controls. It was expected that:

(1) Child molesters would score lower than the two control groups on the decoding and social perception tasks, with no differences expected between admitters and deniers.





24


(2) Child molesters would show distorted cognitions regarding sexual contact with a female child as measured by the vignettes and Cognition Questionnaire. More specifically, child molesters would ascribe more benefit, desire, and responsibility to the child, and less responsibility and punishment for the man. Nonfamilial sex offenders would also show more cognitive distortions than incestuous offenders.

(3) Deniers would show fewer cognitive distortions than admitters on both measures regarding sexual contact with children.

(4) On the adult vignettes, child molesters would not show cognitive distortions on the same variables as they would on the child vignettes. However, child molesters would rate more anxiety than controls regarding adult sexual interaction with an adult female.

(5) Child molesters would demonstrate less empathy than controls; deniers would show less empathy than admitters. Empathy scores would correlate positively with decoding score and negatively with repression.

(6) Child molesters would endorse more symptoms of repression than controls. Deniers would show more repression than admitters.

Alternative Hypotheses

Child molesters' cognitive distortions could be related to sexual contact in general, rather than specific to sexual contact with children. In that case, child molesters would be expected to show the same cognitive distortions on the vignettes presenting sexual contact with an adult female as in the vignettes presenting sexual contact with a female child.

Deniers may not show greater repression, but may show more social desirability. Thus, denial might be more related to perceived demands for social desirability and less related to





25

unconscious defense mechanisms. In this case, deniers would be expected to score significantly higher on social desirability than the other groups, but not higher on repression. Another possible alternative is that the child molesters may not be deficient on general nonverbal decoding skill, but may show a deficiency on the more complex social perception task. Thus, their deficiency may be more related to difficulties in decision skills than in decoding skills.












CHAPTER 2
METHODS

Subjects

There was a total of 214 subjects that completed the study. Subjects were divided into six groups: nonfamilial heterosexual male child molesters [admitters (n=31) and deniers (n=32)], incestuous heterosexual male child molesters (admitters (n=26) and deniers (n=34)], nonsex offending male inmates (n=60), and male community controls (n=31). All subjects were at least 18 years of age. Subjects were recruited from the following correctional facilities in Florida: Union Correctional Institution, New River Correctional Institution, and Marion Correctional Institution.

In order to qualify for the study, subjects in the child molester groups must have been convicted of a sexual offense against a female who was under the age of 12. They could also not report current or prior participation in an inpatient sex offender treatment program since differential treatment histories could serve as a confounding variable. The non-sex offending inmate group was selected on the basis of having no prior record of sex offenses according to self-report and their prison files. The community control group was recruited through local advertisements posted in areas designed to attract a low socioeconomic group. In addition, subjects were solicited at local construction companies and various volunteer organizations.

Child molesters were divided into admitters and deniers based on their written selfreport to the question "Did you commit the crime(s) that you are charged with? If not, please



26





27


explain why you disagree with the charge(s)." Their admission or denial was also assessed during a semi-structured interview in which they were asked to describe their charge and whether or not they committed the crime.

Child molesters were also subdivided into incestuous and nonfamilial offender groups based on information in their prison file describing their relationship to the victim. Since this information was not always available in the prison records, this information was also gathered during an individual interview in which the inmate was asked to describe the nature of his relationship to the victim. Sex offenders who molested their daughters or stepdaughters were considered to be incestuous offenders, as well as those molesting any female child living in their home under their custody or authority. Thus, if the molester was involved in some kind of parental relationship with the victim, as in the case of a live-in boyfriend who molests his girlfriend's children, then he was considered to be an incestuous or familial offender. This broad definition of incest was used since it is the dynamics of the relationship that are considered to be most important, and since most studies on incest have used this broader definition (Marshall et al., 1986).

Research Design

In order to test the hypotheses, the six subject groups were administered a battery of tests measuring the following variables: cognitive distortions regarding sexual contact with children, decoding ability, empathy, repression, and social desirability. Background variables included age, socioeconomic status, intelligence, and history of abuse. Group served as the main independent variable with six levels (nonfamilial admitters, nonfamilial deniers, incestuous admitters, incestuous deniers, community controls, and inmate





28


controls). Within the six levels, comparisons were made between child molesters and controls, nonfamilial and incestuous offenders, and admitters and deniers.

Group served as the sole independent variable in the analysis of all of the

dependent variables except for the cognitive distortions. This dependent variable was assessed using clinical vignettes which contained two independent variables which varied across the vignettes Level of Sexual Contact (4 levels) and Female's Response (3 levels). Thus, a 6x4x3 design was used in the analysis of the vignette data, with Group as a between subjects variable and Level of Sexual Contact and Female's Response as within subjects variables. The following section lists each of the dependent variables and the specific measures used.

Dependent variables

Measures of Cognition

Clinical vignettes

Cognitions regarding sexual contact were measured using the clinical vignettes from Stermac and Segal's (1989) study. There were twelve vignettes describing sexual contact between a male adult and seven year old child. The vignettes were designed by Stermac and Segal (1989) based on documented reports of actual incidents between children and adults. There are two independent factors in each vignette the degree of sexual contact and the child's response to the contact. The levels of sexual contact include: 1) touching only, 2) rubbing genital area over clothing (fondling), 3) fondling genitals under clothing (no clothes), and 4) genital contact with ejaculation. The levels of the child's response include: 1) smiling, 2) passive/no response, and 3) crying with resistance.




29


The subjects were asked to read each vignette and respond to six questions using a five point Likert-type rating scale ranging from 1 (not at all) to 5 (definitely). The questions were designed to reflect child molesters' central cognitive beliefs about adult sexual contact with children, as described in previous clinical investigations. Subjects were asked the following questions (the variable name for each question is in parentheses following the question): 1) Do you think the child enjoyed what happened? (Child Enjoyment) 2) Do you think the child wanted this to happen? (Child Desire) 3) Do you think the child could benefit from this experience? (Child Benefit) 4) Do you think the child was responsible for what happened? (Child Responsibility) 5) Do you think the child could be harmed by this experience? (Child Harm) 6) Do you think the man was responsible for what happened? (Man Responsibility)

In order to assure face validity, Stermac and Segal (1989) asked subjects in a pilot study to rank order the vignettes in terms of degree of sexual contact between the child and adult. They were also given a multiple choice format and were asked to determine the child's response. The authors reported complete agreement among subjects on these tasks.

Stermac and Segal (1989) conducted Pearson correlation matrices for each of the

twelve vignettes in order to examine correlations among the variables. They found that child benefit, harm, and enjoyment were highly correlated across the levels of sexual contact and child's response. Child benefit and enjoyment were negatively correlated with child harm, with correlation coefficients ranging from .32 to .65. Child responsibility and child's desire were also found to be highly correlated, with correlation coefficients ranging from .37 to .71. For analysis purposes, the authors combined child benefit, enjoyment, and harm into a composite variable called "benefit to child," and combined child responsibility and desire into "child





30


complicity." Test-retest reliability data was not available for this measure. Thus, the present study included this analysis (see the Results section).

In addition to the twelve vignettes with a child, twelve vignettes with an adult were designed for the current study. This set of vignettes was designed to assess attitudes toward aggressive sexual contact with an adult female. The same vignette scenes and questions were used as in the child vignettes, although the female portrayed was a 30-year old female. In addition to the above questions asked regarding the woman (instead of the child), the following questions were added (the variable names are in parentheses): 1) Do you think the woman was sexually attracted to the man? (Woman's Sexual Attraction) 2) Do you think the man felt anxious? (Man's Anxiety) 3) Do you think the man felt confident that the woman was attracted to him? (Man's Confidence) These questions were added in order to assess the subject's perception of the man's anxiety and confidence, as well as the subject's perception of the woman's degree of sexual attraction to the man. Since child molesters are thought to have difficulty with relationships with adult women, they were expected to score differently from controls on these variables.

Cognition questionnaire

The Cognition Questionnaire (Abel, Becker, Cunningham-Rathner, Rouleau, et al. 1984) was also administered as a second measure of cognitive distortions regarding sexual contact with children. This instrument contains 29 items and was developed for use with adult child molesters. The subject reads statements which reflect values about adult sexual contact with children and selects a response ranging from 1-5 (strongly agree strongly disagree). Sample items on the Cognition Questionnaire include: 1) A child 13 or younger can make





31


her(his) own decision as to whether she (he) wants to have sex with an adult or not, and 2) Sex between a 13-year old (or younger) child and an adult causes the child no emotional problems. Test-retest reliability for this questionnaire has been reported as .76 over a three week interval (Abel, Becker, Cunningham-Rathner, Rouleau, et al. 1984). Six subscales were identified with coefficient alphas ranging from .59 to .71. Scores on the questionnaire distinguished child molesters from normals, demonstrating discriminant validity. Measures of Decoding/Social Perception

Profile of nonverbal sensitivity

The Profile of Nonverbal Sensitivity (PONS) (Rosenthal et al., 1979) was used as a general measure of nonverbal decoding ability. This widely used measure has the benefit of distinguishing between several channels of nonverbal communication and various types of cues. The PONS is a 47-minute videotape which is comprised of 220 two second film clips in which eleven nonverbal channels are isolated: three "pure" visual (face, body, figure) and two "pure" auditory channels (random spliced, content free), with six additional paired combinations. In each film segment, one of twenty affectively charged interpersonal situations is portrayed by a woman who is not an actress. The twenty situations can be categorized into four affective quadrants: positive-dominant, positive-submissive, negative-dominant, and negativesubmissive. Subjects are asked to choose which of two statements best describes the behavior in each film clip. For example, on the first item the subject must decide whether the woman in the film clip is "expressing jealous anger" or "talking to a lost child." In another item the respondent must choose between "leaving on a trip" or "saying a prayer." The test was designed to measure one's ability to "read" and interpret nonverbal cues.





32

The PONS was standardized on 492 subjects. Test-retest reliabilities average .81 with an interval of 6 to 8 weeks between testing, although they vary over the three dimensions of voice tone (lowest reliability), facial expressions, and body movements (Rosenthal et al., 1979). The internal consistency coefficients were reported as .86 for total score and as follows for the affective quadrants: positive-submissive (.78), positive-dominant (.84), negative-submissive (.87), and negative-dominant (.87).

Principal components analysis on the eleven channel accuracy scores yielded four factors measuring: 1)decoding of facial cues, 2) decoding of randomized spliced cues, 3) decoding of content-filtered cues, and 4) decoding of cues from the body without the face. Principal components analysis was also conducted for the negative and positive affect scenes, dominant and submissive channels, and the twenty scenes. The factor analysis results of the twenty scenes suggests that decoding skill on the PONS can be subdivided into six skills involving decoding cues of interpersonal control, cues of contrition, cues of strong but controlled affects, cues of separation, cues of dependence, and cues associated with recitation. The results of these factor analyses provide support for the various channels and affective quadrants assessed by the PONS.

Interpersonal erception task

General social perception was measured using the Interpersonal Perception Task (IPT) (Costanzo & Archer, 1989) which "is designed to assess accuracy in interpreting the expressive behavior of others" (p.231). The IPT is a 38-minute videotape depicting naturalistic fullchannel sequences of behavior. There are 30 scenes which show one to four people. A question appears on the screen before each scene, requiring the viewer to answer the question





33

based on the scene that follows. During the six second interval following the scene, the viewer chooses the correct answer from a multiple choice format. For example, the first scene shows an adult male and female having a conversation with two seven year-old children. The viewer is asked "Who is the child of the two adults?" Since the questions are based on actual, naturalistic scenes there is an objective criterion of accurate judgment for each item.

The IPT was standardized on 438 college students. The accuracy rates for the scenes are reported as comparable to those for other similar measures. Accuracy rates are listed for each scene. Test-retest reliability using 46 subjects after a five week break was .70. Internal consistency was .52 using the KR-20, which the author suggests may reflect the diversity and relatively small number of scenes. Subjects who scored in the high social sensitivity group based on 17 peer ratings scored significantly higher in the IPT than subjects in the low social sensitivity group, demonstrating construct validity. In addition, the correlation between IPT scores and social sensitivity scores based on peer ratings was r=.48. The authors also report significant positive correlations between the IPT and two other standardized measures ofability to interpret the expressive behavior of others.

Measure of Empathy

Empathy was measured using the Interpersonal Reactivity Index (IRI, Davis, 1979) which Marshall et al. (1993) cited as the best available measure for assessing the complex nature of empathy. The IRI is a 28-item self-report questionnaire with four 7-item subscales measuring different aspects of empathy. The four subscales are 1) perspective-taking, 2) fantasy 3) empathic concern, and 4) personal distress. The Perspective-Taking (PT) scale assesses the tendency to adopt others' psychological points of view. The Fantasy scale (FS)





34

assesses the subject's tendency to "transpose themselves imaginatively into the feelings and actions offictitious characters in books, movies, and plays (Davis, p. 114)." The Empathic Concern (EC) scale measures feelings of sympathy and concern for others, and the Personal Distress (PD) scale measures the subject's feelings of personal anxiety in tense interpersonal situations.

The IRI was developed from a pool of 50 items some of which were adapted from other measures of empathy such as the Mehrabian & Epstein emotional empathy scale (1972). New items were added to measure cognitive aspects of empathy such as the ability to adopt different perspectives. Factor analysis revealed four factors which now form the four subscales. This factor structure has been shown to be stable when repeatedly administered to different samples (Davis, 1983). The final 28-item version of the IRI was standardized on 579 male and 582 female college students. Internal reliabilities range from .71 to .77 and test-retest reliabilities range from .62 to .71 (Davis, 1980).

Davis (1983) cites data supporting the validity of the four IRI subscales and his

multidimensional view of empathy. The four subscales which represent different aspects of empathy showed distinctive patterns of relationships with measures of social functioning, selfesteem, emotionality, sensitivity to others, and unidimensional measures of empathy. The Hogan Empathy Scale which is more cognitive in its approach showed highest correlations (r=.40) with the cognitive PT scale, as expected. The PT scale also showed the least association (r=.20) of the four IRI scales with the Mehrabian and Epstein Emotional Empathy Scale which taps emotional rather than cognitive empathy. In addition, the FS and EC scales showed greater associations with the Emotional and Empathy Scale (r=.52 and .60,





35

respectively). Davis (personal communication, 1994) recommends examining each of the subscale scores separately rather than using the total IRI score since each of the subscales appears to measure a different aspect of empathy. Other researchers, however, argue that the IRI total score may serve as an overall measure of empathy. Measure ofRepression

The Repression-Sensitization Scale (Bryne, 1961) was used to measure repression. The R-S scale is composed of 127 items from six Minnesota Multiphasic Personality Inventory (MMPI) scales, three measuring sensitization and three measuring repression. The range of possible scores is 0 to 127; high scores indicate sensitization while low scores indicate repression. The behavioral dimension ofrepression-sensitization is viewed as a continnum, with the repressive extreme involving avoidance defenses such as denial and the sensitizing extreme involving approach defenses such as intellectualization.

The R-S scale was standardized on 733 male and 571 female college students (Byrne, Barry, & Nelson, 1963). Test-retest reliability over three months was .82 and split-half reliability was .94. In terms of face validity, seven of 10 judges showed agreement on 90%/6 of the items when asked to respond as they thought a repressor would respond. R-S scores have been found to be consistently related to self-report measures. More specifically, scores on the R-S scale were related to self-ideal discrepancy (r=.63) and negative self-description (r=.68), supporting the notion that individuals who use repressive defense mechanisms are less likely to verbalize negative self-descriptions than individuals with sensitizing defense mechanisms.

Correlations between the R-S scale and Marlowe-Crowne Social Desirability Scale range from -.37 to -.49, suggesting that repression-sensitization is partially, but not entirely a





36

function of a socially desirable response style (Byme, 1961). There is some evidence to suggest some independence between repression-sensitization and social desirability (Byrne, 1961). In a study by Merbaum (1972), extreme sensitizers and repressers were asked to retake the test three months after having initially completed the R-S scale. In this second administration they were asked to respond in a way that would make them look as good as possible. The scores of both groups shifted in the repressor direction, however, the sensitizers still scored higher in the direction of sensitization than did repressers. Thus, differences between on repression-sensitization remained despite the instructions to adopt a socially desirable response style.

Measure of Social Desirability

The Marlowe-Crowne Social Desirability Scale (MCSDS) (Crowne & Marlowe, 1960) was used to control for socially desirable response sets. The MCSDS is a 33-item scale which was originally developed to discriminate the effects of item content from the subject's need to present him or herself in a socially desirable light. As a result, the MCSDS focuses on culturally sanctioned and approved behaviors that do not occur frequently. Sample items from this questionnaire include: 1) I never hesitate to go out of my way to help someone in trouble, and 2)1 am always careful about my manner of dress. The internal consistency coefficient is reported at .88, and the test-retest reliability is .89 for a sample of normal 39 college students. Normative data is presented for 28 males and 27 females (Crowne & Marlowe, 1960). The MCSDS is highly correlated with the Edwards Social Desirability Scale (.35) and the following MMPI subscales: L (.54), K (.40), and F (.36). The MCSDS has been used in numerous studies, including the Stermac and Segal (1989) study.





37

Self-Report of Abuse History

The Life Experiences Questionnaire (LEQ) developed by Bryer et al. (1987) was used to assess abuse history. Bryer et al. (1987) administered the LEQ to 66 females upon admission to a private psychiatric hospital. They found that the more severely disturbed patients were more likely to report having been abused in childhood. A recent study showed that female inpatients were more likely to report a history of abuse on the LEQ than during an intake interview (Dill, D., Chu, J., Grob, M, & Eisen, S., 1991). There was low agreement between the LEQ and intake format in regard to reported histories of physical abuse, k=.32. The authors note that the low agreement was largely due to subjects who reported abuse on the LEQ but not during the intake interview. 59% of subjects reported physical abuse on the LEQ while only 26% reported this type of abuse in the interview. Ninety-two percent of subjects reporting abuse during the interview also reported abuse on the LEQ, however, 47% of those who did not report abuse during the interview later reported abuse on the LEQ. Agreement between interview and LEQ reports of sexual abuse were also fairly low, k=.44. 52% of subjects reported a history of sexual abuse on the LEQ while only 35% reported this history in the intake interview. These results strongly support the need to assess abuse history in a written format since many subjects may be reluctant to report abuse history during an interview.

In the present study, physical abuse was defined as an affirmative response to the LEQ question "were you ever hit really hard, kicked, punched, stabbed, or thrown down?" Similarly, sexual abuse was defined as an affirmative response to the question "were you ever pressured against your will into forced contact with the sexual parts of your body or [his/her]





38

body?" A supplement to the questionnaire asked subjects for specifics concerning the frequency and period of time during which the abuse occurred. In order to facilitate comparisons with the Becker (1992) study, the variable of abuse encompassed both physical and sexual abuse. Thus, within this study, the term "a history of abuse" refers to a selfreported history of physical and/or sexual abuse. Measure of Social Status

The Hollingshead Four Factor Index of Social Status (Hollingshead, 1975) was used to estimate subjects' socioeconomic status. This index assesses social status based on education, occupation, sex, and marital status. The occupation factor is based on the approximately 450 occupational titles and codes of the 1970 United States census, and is graded on a 9-point scale. The education factor is based on the number of years of schooling and is scored on a 7-point scale. Scores from the occupation, marital status, and education factors are combined to yield a social status score ranging from 8 to 66.

The Four Factor Index was found to be reliable and valid based on data derived from the 1970 Census and the National Opinion Research Council. Social strata categories were validated using factor analysis and criterion validity was established by comparing occupation index scores with prestige scores developed by the National Opinion Research Center (r=.93). Significant differences were found among these categories when mass communication data were used as criteria of social behavior (Hollingshead, 1975). The Four Factor Index also correlates highly with other measures of socioeconomic status: the Revised Duncan Socioeconomic Index (r=.79) and the Siegel 1965 NORC Prestige Scale (r=.73). Measure of Intelligence





39

The Vocabulary Scale of the Wechsler Adult Intelligence Scale-Revised was used as a rough estimate of level of intellectual functioning. The Vocabulary Scale contains 35 items or vocabulary words which the subject is asked to define. Responses are scored on a 0-2 point scale. Split-half reliability for the Vocabulary Scale ranges from .94 to .96 across the 16-74 age range and the test-retest reliability is .93 for ages 25-34, and .91 for ages 45-54 (Wechsler, 1981). The Vocabulary score is highly correlated with the WAIS-R Full Scale score (r=.85) and the Verbal Scale score (r=.90). The WAIS-R was standardized on a sample of 1,880 subjects stratified according to age, sex, race, geographic region, education, and occupation. Summary ofthe Variables.

The background variables included age, socioeconomic status score on the

Hollingshead Four Factor Index of Social Status, vocabulary score on the WAIS-R, and selfreport of physical or sexual abuse as assessed by the Life Experiences Questionnaire. The dependent variables used to assess cognitive distortions included the total score on the Cognition Questionnaire, and the six variables on the child and adult versions of the written vignettes: 1) Female's Enjoyment of the sexual interaction, 2) Female's Desire, 3) Female's Benefit, 4) Female's Responsibility, 5) Female's Harm, and 6) the Man's Responsibility. There were two additional variables on the adult female vignettes including the Woman's Sexual Attraction to the man and the Man's Anxiety regarding the sexual interaction.

Decoding ability was assessed using the total scores from the Profile of Nonverbal Sensitivity Test and the Interpersonal Perception Task. Other dependent variables included the total scores on the Repression-Sensitization Scale and the Marlowe-Crowne Social Desirability





40

Scale, the empathy total score on the Interpersonal Reactivity Index, and the four subscale scores measuring perspective-taking, fantasy, empathic concern, and personal distress.

Procedures

Prison files were reviewed in order to select subjects who met the criteria for the study. All records of inmates with a child sex offense were reviewed. In order to generate a random list for the control group, every fifth name was selected from the prison roster. These files were reviewed and those inmates with no history of a sex offense were assigned to the nonsex offending control group. Subjects selected for both groups were sent a letter which described the nature of the study and listed the date and time of an informational meeting they could attend to learn more about the study. Approximately ten subjects were scheduled for each initial group meeting.

At the informational meeting, subjects were provided with a complete description of the project. Those who volunteered were asked to complete the Introductory Questionnaire, the Social Desirability Scale, and the Interpersonal Reactivity Index. Subjects were then administered the IPT, and asked to complete one of the two sets of vignettes based on random assignment. At the end of the first session, subjects were scheduled to return for an individual session and the second group session within the next seven days.

In the individual session, subjects were administered a semistructured interview and the Vocabulary Scale from the WAIS-R In the second group session, subjects were administered the R-S scale and the LEQ, followed by the PONS test. Subjects were then asked to complete the second set of vignettes.













CHAPTER 3
RESULTS


Based on the literature, it was predicted that: 1) child molesters would score lower than controls on decoding tasks; 2) child molesters would show more cognitive distortions than controls regarding sexual contact with children, and nonfamilial offenders would show more cognitive distortions than incestuous offenders; 3) deniers would show fewer cognitive distortions than admitters; 4) child molesters would attribute more anxiety to the man in the adult vignettes than would controls; 5) child molesters would show less empathy than controls, and deniers would show less empathy than admitters; and 6) child molesters would show more repression than controls, and deniers would show more repression than admitters.

Background Variables
Background variables included age, socioeconomic status, vocabulary scores, social desirability, and self-reported abuse history. In order to examine among-group differences on these variables, several one-way ANOVAs were conducted with Group (incestuous admitters, incestuous deniers, nonfamilial admitters, nonfamilial deniers, nonsex offending inmate controls, community controls) as the between subjects factor and age, socioeconomic status, and vocabulary scores as the dependent measures. Several ANOVAs were conducted rather than a MANOVA since the primary interest was to examine group differences on each variable. There were no significant differences on


41




42


socioeconomic status, F(5,208)=.576, p=.718 or vocabulary score F(5,208)=1.942, p=.089. There was a sigificant difference, however, on age, F(5,208)=8.518, p=.000. The mean ages of the groups were 29.86 for the inmate control group, 34.74 for the community control group, 38.89 for the incestuous admitters, 38.32 for the incestuous deniers, 40.45 for the nonfamilial admitters, and 39.72 for the nonfamilial deniers. A Scheffe post-hoc analysis showed that the inmate control group was significantly younger than each of the four child molester groups. There were no significant differences among the other groups. A contrast between child molesters and controls revealed that the child molesters were significantly older than controls, F(1,208)=27.421, p=.000. Other studies have also found that child molesters were significantly older than controls (Haywood & Grossman, 1994). This age difference is not surprising since the general-inmate population tends to be younger than the sex offender population, by virtue of their shorter prison sentences in Florida and sex offenders' older age at arrest.

It was expected that a greater proportion of child molesters than controls would report a history of physical and/or sexual abuse. In order to examine reported abuse history across groups, a chi-square analysis was performed. There was a significant difference among the six groups, X2(6)=13.626, p=.018. The following percentage of each group reported a history of abuse: Incestuous Admitters (61.54%), Nonfamilial Admitters (54.84%), Incestuous Deniers (47.06%), Nonfamilial Deniers (31.25%), Controls (30%), and Community Controls (29%) (see Figure 1). One-tailed planned comparisons showed that the child molesters reported significantly more abuse than controls as expected, Z=2.862, p=.002, and admitters reported significantly more abuse





43

than deniers, Z=2.158, p=.015. The differences between incestuous and nonfamilial sex offenders was not significant, Z=1.276, p=.202, two-tailed.

In order to examine differences between abused and nonabused subjects, several one-way ANOVAs were conducted with abuse history as the between subjects factor and total scores from the IPT, PONS, empathy measure, Cognition Questionnaire, Social Desirability Scale, and Repression Scale as the dependent measures. Several ANOVAs were used because group differences on each variables were of interest. There was a significant difference between abused and nonabused subjects on repression, F(1,208)=4.952, p=.027, indicating that those who denied a history of abuse scored higher on repression than those reporting a history of abuse (see Table 1). When Group was added as a second between subjects factor, however, the main effect of abuse history became nonsignificant, F(1,198)=2.972, p=.086. There was also a significant main effect of abuse history on social desirability, F(1,210)=8.491, p=.004, with those reporting a history of abuse scoring significantly lower than nonabused subjects on social desirability (see Table 1). When Group was added as the second between subjects factor, there was a significant main effect of Group, F(5,200)=4.238, p=.001, and a significant Abuse X Group interaction, F(5,200)=3.032, p=.012. Figure 2 shows that there was no consistent pattern between abused and nonabused subjects across the six groups.

To examine the relationship between abuse history and admission/denial, an ANOVA was conducted with Abuse and Admit as the two between subjects variables. The community control subjects were not included in this analysis since the variable of admission/denial was not relevant to them. There was a significant main effect of Abuse,





44


F(1,179)=5.266, p=.023, and a significant main effect of Admit, F(1,179)=7.905, p=.005, with a nonsignificant Abuse X Admit interaction, F(1,179)=.447, p=.505 (see Table 2). Figure 3 shows that subjects reporting no history of abuse scored significantly higher on social desirability than those reporting abuse. In addition, inmates who denied having committed their crimes scored significantly higher on social desirability than those who admitted to committing their crimes. Thus, deniers who did not report abuse showed the highest social desirability while admitters who reported a history of abuse showed the lowest social desirability.

No significant differences were expected between child molesters and controls on social desirability. One outlier was identified by the studentized residual value and was dropped from the analysis. As expected, a planned comparison revealed no significant differences in social desirability between child molesters and controls, F(1,205)-.337, p=.562. An ANOVA with the six groups, however, yielded a significant main effect for Group F(5,205)=4.258, p=.001. As expected, deniers scored significantly higher than admitters on social desirability, F(1,205)=16.769, p=.000 (see Table 2). There were no significant differences between incestuous and nonfamilial sex offenders, F(1,205)= 2.899, p=.090. All possible pairwise comparisons using the Bonferroni correction showed that incestuous deniers and nonfamilial deniers scored significantly higher than incestuous admitters on social desirability, p=.003, p=.002, respectively.

Method of Analysis

The dependent variables were analyzed using ANCOVAs or MANCOVAs. Decoding skill, empathy, repression, and cognitive distortions on the Cognition





45

Questionnaire were analyzed using separate analyses of variance. Group served as the between subjects variable across analyses, and social desirability served as the covariate in the analysis of the Cognition Questionnaire, empathy, and repression scores. The assumptions for ANOVA were met for each of the dependent variables.

The vignette data was analyzed using principal components analysis followed by a multivariate analysis of covariance. For the univariate and multivariate analyses, planned comparisons were conducted between child molesters and controls, admitters and deniers, and between incestuous and nonfamilial sex offenders. Each of these comparisons was considered to be its own separate family since each was thought to involve a conceptually distinct comparison. Thus, the Bonferroni correction was not applied to these three comparisons. The Bonferroni correction was used, however, when all possible pairwise comparisons were conducted. All possible pairwise comparisons were examined since it was expected that there might be differences among the six groups other than those among the three families. More specific hypotheses than those regarding the three families were not proposed since there has been little research on the differences among these particular subgroups of child molesters. Thus, for example, differences between incestuous deniers and nonfamilial admitters might be of interest, however, there was no specific hypothesis regarding these two groups since there has been no research to date in this area.

In addition to the planned comparisons and Bonferroni pairwise comparisons, some specific post hoc comparisons were conducted between a particular group and the five other groups combined. This contrast was conducted when one group seemed to





46

score significantly higher than all of the other groups, which was frequently the case with the nonfamilial admitters. The Scheffe test was used to conduct these post hoc comparisons.

Analysis of Decoding Skill

Child molesters were predicted to score lower on decoding measures than control subjects. Contrary to expectations, there were no significant differences between child molesters and controls on the total score of the Profile of Nonverbal Sensitivity Test (PONS), F(2,203)=.018, p=.892. There were also no significant differences between admitters and deniers, or between incestuous and nonfamilial sex offenders. An ANOVA conducted with the six groups also showed no significant difference among the groups F(5,203)=.737, p=.597 on PONS total score (see Table 3). An ANOVA with the four PONS subscale scores as the repeated measures revealed no significant differences among groups, F(5,203)=.789, p=.559. As expected, subjects showed significant differences across the four subscales, F(3,609)=122.449, p=.000. It was predicted that the decoding scores would be significantly correlated with the empathy scores. Contrary to expectations, the correlation of the PONS total score with the IRI total score was not significant, r=.039.

Consistent with the results on the PONS, there were no significant differences between child molesters and controls, admitters and deniers, or between incestuous and nonfamilial sex offenders on the Interpersonal Perception Task (IPT). An ANOVA also showed no significant differences among the six groups on the IPT score, F(5,207)=.203, p=.961 (see Table 3) or on the subjects' estimation of their IPT score F(5,184)=.549,





47

p=.739. It was expected that empathy scores would be significantly positively correlated with decoding scores. Consistent with the results from the PONS, the correlation of the IPT score with the IRI total score was not significant, r=. 110.



Analysis of Cognitive Distortions

Cognition Ouestionnaire

It was predicted that child molesters would show more cognitive distortions than controls on the Cognition Questionnaire which assesses one's attitude toward sexual contact with children. Seven outliers were identified based on studentized residual values and these cases were dropped from the analysis. A planned comparison revealed that child molesters did endorse significantly more cognitive distortions about sexual contact with children than did the control groups, F(1,188)=7.478, p=.007 (see Table 4). Lower scores on this measure indicate more cognitive distortion. An ANOVA with the six groups revealed a significant main effect for Group F(5,188)=2.862, p-.023 and social desirability, F(1,188)=6.929, p=.009 (see Figure 4). It was also predicted that deniers would show fewer cognitive distortions than admitters, however, planned comparisons showed no significant differences between these two groups F(1,188)=.210, p=.648. Nonfamilial offenders were expected to show more cognitive distortions than incestuous offenders. Consistent with this prediction, the difference between incestuous and nonfamilial sex offenders was significant, F(1,188)=4.041, p=.046, with nonfamilial sex offenders endorsing more cognitive distortions than incestuous offenders. Pairwise





48

comparisons using the Bonferroni correction showed that the nonfamilial deniers endorsed significantly more cognitive distortions than the community control group. Clinical Vignettes

Child vignettes

Child molesters were expected to show more cognitive distortions on the child vignettes than the controls, and admitters were expected to show more cognitive distortions than the deniers. Based on Stermac & Segal's (1989) results, child molesters were expected to ascribe more benefit to the child than control groups when the child's response was positive or ambiguous, but to show no difference when the child's response was clearly negative. The variable of"benefit" was a composite ofthe variables benefit to child, harm to child, and child's enjoyment. Compared to controls, child molesters were also expected to attribute more responsibility and desire to the child and less responsibility to the adult.

Test-retest reliability. The child vignettes were readministered to subjects after a 3 to 4 week interval. Scores on each of the nine variables were summed across the twelve vignettes for each administration and Pearson correlations were calculated to measure test-retest reliability from Trial 1 to Trial 2. The test-retest reliability for the variables is as follows: Child Benefit (r=.83), Child Harm (r=.82), Child Enjoyment (r=.83), Child Desire (r=.66), Man's Responsibility (r-.71), Treatment (r=.80), Other Men (r=.81), and Punishment (r=.72). Analysis of the specific variables on each vignette also showed strong correlations between the first and second administrations, demonstrating that subjects tended to respond in a consistent manner on this measure over time.





49

Reduction of variables. In order to facilitate comparison with the results of Stermac & Segal (1989), Pearson correlation matrices were conducted for each of the twelve child vignettes to see which variables were highly correlated across levels of sexual contact and child's response (see Table 5). This matrix shows that the nine variables are highly correlated with each other. Stermac & Segal (1989) addressed this issue of multicollinearity in their study by combining highly correlated variables into two composite variables. They found that Benefit to Child, Harm to Child, and the Child's Enjoyment were highly correlated. Child Benefit and Enjoyment were negatively correlated with Child Harm, with correlation coefficients ranging from .32 to .65. They combined these variables into a composite variable called "Benefit to the Child."

In the present study, the negative correlation coefficients between Child Benefit and Enjoyment and Child Harm ranged from .09 to .34. Child Benefit and Child Enjoyment showed correlation coefficients ranging from .18 to .59. Since these correlations were low and spanned a large range, these three variables were not combined into a composite variable. When the variables were summed across levels ofchild's response and sexual contact, Child Benefit correlated with Enjoyment at .60 and with Harm at -.30. The latter two variables correlated -.35.

Stermac & Segal (1989) also formed a second composite variable called "Child Complicity" from the variables of Child Responsibility and Child Desire. Across the levels of sexual contact and child's response, these two variables showed correlations ranging from .37 to .71. In the present study, these correlations ranged from .40 to .75, closely replicating Stermac & Segal's (1989) findings. Although we could have formed a





50

composite variable as did Stermac & Segal (1989), these two variables were only two of the nine variables that were all highly correlated with each other. Thus, specifying these two particular variables for combination did not seem to be statistically warranted. Instead, a principal components analysis was conducted in order to reduce the number of variables in a more statistically appropriate manner.

Principal components analysis of the child vignettes. In addition to reduction of the number of variables, principal component analyses with varimax rotation were conducted in order to examine the relationships among the variables. A principal components analysis was conducted on the nine variables collapsed across levels of sexual contact and child's response. The analysis yielded two factors, accounting for 38.16% and 31.96% of the total variance. The variables with high loadings on Factor 1 included child desire, child responsibility, child enjoyment, child benefit, and other men (see Table 6). Thus, this factor appears to measure the child's reaction and complicity in the sexual contact. In contrast, Factor 2 measures the man's responsibility for the sexual contact as well as the seriousness of the behavior. The variables with high loadings on this factor included the man's responsibility, the harm to child, the man's need for treatment, and the amount of punishment prescribed for the man.

Although the analysis of the nine variables summed across vignettes yielded two meaningful factors, it was important to conduct principal components analyses for each of the twelve vignettes to see if these two factors were stable within each of the vignettes. As expected, the same factor structure was found in each of the twelve vignettes, indicating that these same variables tend to cluster together within each vignette.





51

The principal components analysis was conducted to reduce the number of

variables in an effort to minimize the redundancy associated with results involving multiple highly correlated dependent measures. Thus, the two factors were used as dependent measures in analyses of variance. An ANOVA conducted with Group as the between subjects variable and Factor 1, Child Complicity, as the dependent measure yielded a significant main effect of group, F(5,202)= 3.91, p=.002. Planned comparisons showed that child molesters did not attribute more complicity to the child than controls, F(1,202)= .369, p=.544. A post-hoc comparison, however, showed that nonfamilial admitters, one of the child molester groups, attributed significantly more complicity than all of the other groups combined, F(1,202)= 16.835, which was significant using the Scheffe test. As expected, nonfamilial sex offenders as a group attributed more complicity than incestuous offenders, F(1,202)= 6.050, p=.015, however, much of this difference may be attributed to the nonfamilial admitters. Similarly, the admitters scored higher on this factor than the deniers, F(1,202)=3.638, p=.058, although much of this difference is also probably attributable to the nonfamilial admitters.

An ANOVA with Factor 2 as the dependent measure also yielded a significant main effect of group, F(5,202)= 2.480, p=.033, indicating a significant difference among groups in their attitudes regarding the man's responsibility and the seriousness of the sexual behavior. Planned comparisons showed that child molesters rated the sexual behavior as less serious and less deserving of severe punishment than controls, F(1,202)=8.040, p=.005. There were no significant differences between admitters and deniers or incestuous and nonfamilial offenders.





52

Multivariate analysis of variance. The principal components analysis reduced the variables into two sets one relating to the Child's Complicity and another relating to the Seriousness of the Offense. A doubly multivariate analysis of variance was conducted on each of the two sets of variables. The Child Complicity variable set consisted of Child Benefit, Child Enjoyment, Child Desire, Child Responsibility, and Other Men. The variable set relating to the Seriousness of the Offense consisted of Child Harm, Man's Responsibility, Man's Need for Treatment, and Punishment. Group served as the between subjects factor with Child's Response (Child) and Level of Sexual Contact (Sex) as the within subjects factors. The dependent measures were the variables in each variable set. Social desirability was used as a covariate in the initial analysis, however, it was not significant in the analysis of Child Complicity, F(5,194)=.422, p=.833, or Seriousness of the Offense, F(4,195)=.539, p=.707, using the Wilk's criterion, so it was dropped from further analysis.

The doubly multivariate analysis for Child Complicity yielded a significant main effect for Group, F(25, 733.32)=1.54, p=.046, using Wilk's criterion. There was also a significant Group X Child interaction, F(50, 1818.52)= 1.47, p=.018, using the averaged multivariate tests of significance. Planned comparisons were conducted between child molesters and controls, incestuous and nonfamilial offenders, and admitters and deniers. Incestuous offenders scored significantly different from nonfamilial offenders on Child Complicity, F(5,197)=2.921, p=.014, using the Wilk's criterion. There was no significant difference between child molesters and controls, F(5,197)=1.410, p=.222, or between admitters and deniers, F(5,197)-.8372, p=.528.





53

Child molesters and controls differed significantly on the Group X Child

interaction, F(10, 796)=2.134, p=.020, using the averaged multivariate test of significance. The difference between incestuous and nonfamilial offenders on this interaction approached significance, F(10, 796)=1.817, p=.054, on the averaged tests. There were no significant differences among groups on the Group X Sex interaction. The difference between child molesters and controls approached significance on the Group X Child X Sex interaction, F(30, 4810)=1.428, p=.061, using the averaged multivariate tests.

The doubly MANOVA on the Seriousness of the Offense yielded a significant main effect of Group, F(20, 657.64)=2.34, p=.001, and a significant Group X Sex interaction, F(60, 893.47)=1.522, p=.008. Planned comparisons revealed that the child molesters differed significantly from controls on the Seriousness of the Offense, F(4,198)=6.015, p=.000. There were no significant differences among the other groups.

Child molesters differed from controls on the Group X Sex interaction,

F(12,190)=2.505, p=.004. The difference between admitters and deniers on the Group X Sex interaction approached significance, F(12,1587.74)=1.744, p= .052, using the averaged multivariate test of significance. The Group X Child X Sex interaction approached significance for the planned comparison between admitters and deniers, F(24, 178)=1.516, p=.067.

Analysis of variance. The significant MANOVA results were followed by univariate analyses of variance on each of the nine variables. Group effects were examined on each of the variables, with the Group X Child interaction examined for the





54

Child Complicity variables and the Group X Sex interaction examined for the Seriousness of the Offense variables.

Analysis of"Child Complicitv" variables. As expected, there was a significant difference between groups on the variable Benefit to Child, F(5,202)=3.380, p=.006 (see Figure 5). Contrary to expectations, there were no significant differences between child molesters and controls in their overall benefit ratings, however, there was a significant finding regarding the nonfamilial admitters. A post-hoc comparison showed that the nonfamilial admitters ascribed significantly more benefit to the child than the other five groups, F(1,202)=14.974, using the Scheffe test. Nonfamilial sex offenders attributed more benefit to the child than did incestuous offenders, F(1,202)=9.214, p=.003, as predicted. Contrary to expectations, there were no significant differences between admitters and deniers, F(1,202)=1.292, p=.257.

There was a significant difference between child molesters and controls on the Group X Child's Response interaction, F(2,404)=4.070, p=.018. There were no significant differences between admitters and deniers or nonfamilial and incestuous offenders on this interaction.

Groups differed significantly on vignettes in which the child smiled,

F(5,202)=4.106, p=.001, and gave a passive response, F(5,202)=4.403, p=.001. There were no significant differences among groups, however, on the vignettes in which the child cried, F(5,202)=1.422, p=.218. This finding is consistent with Stermac and Segal's (1989) results. Planned comparisons showed that the child molesters did not significantly differ from controls at any level of the child's response. A post-hoc comparison, however,





55

showed that nonfamilial admitters ascribed significantly more benefit to the child than the other groups when the child smiled, F(1,202)=19.907, p=.000. This post-hoc contrast was not significant when the child was passive or cried. Bonferroni adjusted comparisons, however, showed that the nonfamilial admitters gave significantly higher ratings on the "passive" vignettes than every group except for the nonfamilial deniers.

As expected, groups differed in their ratings of the child's enjoyment, however, the difference only approached significance, F(5,202)=2.192, p=.057. Planned comparisons showed no significant differences between child molesters and controls, F(1,202)=1.603, p=.207, or between admitters and deniers, F(1,202)=1.484, p=.225, however, nonfamilial offenders gave significantly higher ratings of child enjoyment than incestuous offenders, F(1,202)=3.984, p=.047, as predicted.

There was also a significant Group X Child's Response interaction

F(10,404)=4.00, p-.000, indicating that the groups' ratings of child enjoyment differed across the levels of the child's response. More specifically, the groups differed significantly in their ratings of child enjoyment on the vignettes in which the child smiled, F(5,202)=3.997, p=.002, but did not show significant differences when the child cried or was passive.

There was a significant Group X Child interaction for the planned comparison between child molesters and controls, F(2,404)=8.443, p=.000, and for the comparison between nonfamilial and incestuous offenders, F(2,404)=.019 (see Figure 6). When the child was smiling, child molesters ascribed more enjoyment to the child than did controls, F(1,202)=5.919, p=.016. Child molesters did not significantly differ from controls when





56

the child was passive, F(1,202)=1.507, p=.221, but did differ when the child cried, F(1,202)=4.873, p=.028. Child molesters gave higher child enjoyment ratings than controls when the child smiled, but gave lower ratings than controls when the child cried. Nonfamilial offenders ascribed more enjoyment to the child than did incestuous offenders when the child smiled, F(1,202)=4.798, p=.030, and was passive, F(1,202)=4.153, p=.043, but did not differ from incestuous offenders when the child cried, F(1,202)=.065, p=.799.

The post-hoc comparison between nonfamilial admitters and the other five groups approached significance on the Group X Child interaction, F(2,404)=15.706, using the Scheffe test. Nonfamilial admitters attributed significantly more enjoyment to the child than the other five groups when the child smiled, F(1,202)=17.561, p=.000, but not when the child was passive, F(1,202)=7.839, p=.006, or when the child cried, F(1,202)=1.444, p=.231 (see Table 7, Figure 7).

Planned comparisons showed no significant differences between child molesters and controls on their ratings of the child's desire for the sexual interaction, F(1,202)=.713, p=.399. There were also no significant differences between admitters and deniers, F(1,202)=1.438, p=.232, or between incestuous and nonfamilial offenders, F(1,202)=1.780, p=.184.

Groups differed significantly, however, in their ratings ofthe child's desire depending on the child's response. This was indicated by a significant Group X Child interaction on Child Desire, F(10,404)=4.101, p=.000. Groups differed significantly when the child smiled, F(5,202)=3.425, p=.005, but not when the child was passive or cried (see Figure 8). Planned comparisons on the Group X Child interaction showed that child





57

molesters scored significantly different from controls in their ratings across the levels of the child's response, F(2,404)=7.186, p=.001. The difference between between admitters and deniers was also significant, F(2,404)=4.327, p=.014 (see Figure 9), as was the difference between incestuous and nonfamilial offenders, F(2,404)=4.750, p=.009 (see Figure 10).

Child molesters did not differ from controls in their desire ratings when the child smiled, F(1,202)=3.440, p=.065, or when the child was passive, F(1,202)=1.378, p=.242. The child molesters' ratings were significantly lower than those of controls, however, when the child cried, F(1,202)=4.449, p=.036 (see Figure 11). Nonfamilial offenders did not score significantly different from incestuous offenders when the child smiled, F(1,202)=2.665, p=. 104, was passive, F(1,202)=3.679, p=.057, or cried; F(1,202)=1.081, p=.300.

The post-hoc comparison between the nonfamilial admitters and the five groups approached significance using the Scheffe test, F(2,404)=16.234. Nonfamilial admitters ascribed more desire to the child than the other groups when the child smiled, F(1,202)=15.262, p=.000, but not when the child was passive, F(1,202)= 7.883, p=.005 or cried (Table 10). Figure 11 illustrates that nonfamilial admitters showed the same pattern on this variable as on'Child Benefit and Enjoyment. Nonfamilial admitters attributed significantly more desire to the child than did the other groups when the child smiled, but scored the same as others when the child's response was ambiguous or clearly negative.

Groups differed significantly in their ratings of the child's responsibility, F(5,205)=2.703, p=.022. There was also a significant Group X Child interaction,





58

F(10,404)=3.739, p=.000 (see Figure 12). Planned comparisons showed significant differences between child molesters and controls, F(2,404)=3.751, p=.024 and between incestuous and nonfamnilial offenders, F(2,404)=6.822, p=.001 on the Group X Child interaction. Admitters and deniers did not significantly differ on this interaction, F(2,404)=1.479, p=.229.

On the smile vignettes, the difference between child molesters and controls

approached significance, F(1,202)=3.726, p=.055. There was no significant difference between child molesters and controls when the child was passive, F(1,202)=2.764, p=.098, or when the child cried, F(1,202)=.012, p=.911. Nonfamilial offenders ascribed more responsibility to the child than did incestuous offenders when the child smiled, F(1,202)=6.616, p=.011 and was passive, F(1,202)=5.585, p=.019, but not when the child cried, F(1,202)=.022, p=.882.

The post-hoc comparison between nonfamilial admitters and the other five groups approached significance using the Scheffe test, F(2,404)=16.365. Nonfamilial admitters showed a similar pattern on this variable as on Child Benefit, Enjoyment and Desire. Using the Scheffe test, the nonfamilial admitters ascribed significantly more responsibility to the child than the other groups when the child smiled, F(1,202)=14.129, p=.000, but not when the child was passive or cried.

As expected, groups differed significantly in their ratings of the likelihood of other men engaging in sexual contact with children, F(5,202)=2.837, p=.017 (see Figure 13). Child molesters did not differ significantly from controls as expected, F(1,202)=3.0311, p=.083. Nonfamilial sex offenders, however, did differ from incestuous offenders as





59

expected. Nonfamilial offenders rated other men as more likely to molest children than did incestuous offenders, although this difference only approached significance, F(1,202)=3.423, p=.066.

As predicted, admitters were more likely than deniers to rate other men as willing to engage in sexual contact with a child, F(1,202)=3.853, p=.051, although this difference only approached significance. A post-hoc comparison showed that, compared to the other five groups, nonfamilial admitters rated other men as significantly more likely to molest children, F(1,202)=12.625, p=.000 (see Table 11). There were no significant differences between child molesters and controls, nonfamilial and incestuous offenders, and admitters and deniers on the Group X Child interaction.

Analysis of"Seriousness of Offense" variables. As predicted, the groups differed significantly in their ratings of the amount of punishment the molester should receive, F(5,202)=6.623, p=.000. The Group X Sex Contact interaction was also significant, F(15,606)=2.954, p=.003. Planned comparisons showed that the child molesters prescribed significantly less punishment than the controls, (1,202)=28.90, p=.000. In contrast to results on the other eight variables, this significant result did not appear to be largely affected by the pattern of the nonfamilial admitters. Child molesters and controls also differed significantly on the Group X Sex interaction, F(3,606)=9.905, p=.000. Figure 14 shows that the child molesters prescribed significantly less punishment than the controls across all four levels of sexual contact: touching F(1,202)=33.470, p=.000, rubbing F(1,202)=28.924, p=.000, fondling F(1,202)=18.428, p=.000, and ejaculation F(1,202)=11.647, p=.001. The difference between the child molesters and controls was





60

largest when the the sexual contact involved touching only, and this difference decreased as the level of sexual contact increased. There were no significant differences between nonfamilial and incestuous offenders, F(3,606)=1.968, p=. 118, or between admitters and deniers, F(3,606)=1.736, p=. 158 on the Group X Sex interaction.

There were no significant differences among groups on Child Harm,

F(5,202)=.662, p=.653 and there were no significant interactions with Group. There were also no significant differences among groups on Man Responsibility F(5,202)=1.227, p=.298 and no significant interactions with Group.

There were no significant differences among the groups on Treatment,

F(5,202)=1.652, p=. 148, however, child molesters differed significantly from controls on the Group X Sex interaction, F(3,606)=4.451, p=.004. Figure 15 shows that the child molesters consistently prescribed less treatment than controls across all levels of sexual contact. The difference between groups was significant when the sexual contact involved touching only, F(1,202)=9.068, p=.003, but not when the sexual contact increased to rubbing F(1,202)=2.593, p=.109, fondling F(1,202)=2.966, p=.087, or ejaculation F(1,202)=1.784, p=. 183. This pattern was similar to the child molesters' pattern of ratings on the variable of punishment. These results suggest that the finding of differences between child molesters and controls on Factor 2, or Seriousness of Offense, can be attributed to child molesters' different views on punishment and treatment, rather than to differences in ratings of the man's responsibility or harm to the child.

As expected, there were no significant correlations between each of the nine variables and scores on the Social Desirability Scale. This finding is consistent with the





61

results of the Stermac and Segal (1989) study. Correlation coefficients ranged from .008 to .052.

Adult vignettes

Principal components analysis. The data from the adult vignettes was analyzed in the same manner as the child vignettes. A principal components analysis was conducted on the nine variables collapsed across the levels of sexual contact and the woman's response. The analysis yielded two factors which exceeded the eigenvalue criterion of 1. Factor 1 accounted for 43.37% of the total variance while Factor 2 accounted for 17.38% of the variance. The component loadings listed in Table 12 indicate that these two factors are similar to the two factors derived on the child vignettes. The female's Desire, Enjoyment, Responsibility, and Benefit loaded highly on Factor 1 in both the child and adult vignettes. In addition, Factor 2seems to tap the man's role in both vignettes. On the adult vignettes, Factor 2 taps the Man's Anxiety, Confidence, and Responsibility for the sexual interaction. Although some of these specific variables were not included on the child vignettes, these variables focus on the man's role as do the variables which loaded highly on Factor 2 on the child vignettes.

Principal components analyses were conducted for each of the 12 vignettes. The analyses yielded 2 to 3 factors for each vignette, similar to the factor structure found in the analysis of the nine means. Within each vignette, the first factor accounted for 33.01%42.03% of the variance and the second factor accounted for 13.05%-19.89% of the variance. These results are consistent with the results of the analysis using the nine means.





62

Analysis of variance was conducted with Group as the between subjects factor, Woman's Response (Woman) and Level of Sexual Contact (Sex) as the within subjects factors, and Factor I as the dependent measure. There was a significant main effect of Group, F(5,202)=3.124, p=010. Planned comparisons, however, showed no significant differences between child molesters and controls, incestuous and nonfamilial offenders, or admitters and deniers. The significant group effect appears to be related to the inmate controls who scored higher on this factor than the other five groups, although this difference was not significant using the Scheffe test, F(1,205)= 10.096, p=.002.

Analysis of variance using Factor 2 as the dependent measure also showed a significant main effect of Group, F(5,205)=3.436, p=.005. Deniers scored significantly different from admitters on Factor 2, F(1,202)=10.560, p=.001. Much of this difference appears to be attributable to nonfamilial deniers who scored higher than the other groups on this factor, although not significantly higher using the post-hoc Scheffe test, F(1,205)=8.862, p=.003.

Multivariate analysis of variance. Based on the results of the principal components analysis, the nine variables on the adult vignettes were divided into two variable sets, one relating to the Woman's Complicity and the other relating to the Man's Role. A doubly multivariate analysis of variance was conducted with each variable set. Social desirability was initially used as a covariate in the analysis, however, it was found to be nonsignificant with Woman's Complicity, F(6,197)=1.678, p=. 128, and with Man's Role, F(3,200)=2.351, p=.074, so it was dropped from further analysis.





63

Groups differed significantly on the Woman's Complicity, F(30,802)= 1.634, p=.018, using Wilk's criterion. There were no significant interactions with Group. Planned comparisons revealed significant differences between incestuous and nonfamilial offenders, F(6,200)=2.186, p=.046. There were no significant differences between child molesters and controls, or between admitters and deniers. There were no significant differences on the planned comparisons for the Group X Woman interaction. The difference between nonfamilial and incestuous offenders approached significance on the Group X Sex interaction, F(18,1725.83)=1.555, p=.064, using the averaged multivariate tests. Child molesters and controls differed significantly on the Group X Woman X Sex interaction, F(36,170)=1.919, p=.003.

There was no significant main effect of Group on the Man's Role,

F(15,560.80)=1.363, p=.160. There were also no significant interactions with Group. Planned comparisons also showed no significant differences on the main effect of Group or interactions with Group. These results do not support the hypothesis that child molesters would rate more anxiety and less confidence to a man in a sexual interaction with a woman.

Analysis of "Woman's Complicity" variables. In order to further examine the significant results from the doubly MANOVA, univariate analyses of variance were conducted for each variable comprising Woman's Complicity. Group served as the between subjects factor with Woman's Response and Level of Sexual Contact as the within subjects factors. For each of these variables there were significant main effects on Woman's Response and Level of Sexual Contact, indicating that subjects' ratings differed





64

significantly as the level of sexual contact and the woman's response to the sexual contact changed. There was a significant interaction between these two variables on every variable except for Woman's Harm.

On the adult vignettes, groups were expected to differ only on the variables of Man's Anxiety and Man's Confidence. Contrary to expectations, groups differed significantly in their ratings of the amount of benefit the woman received from the sexual contact, F(5,205)=2.818, p=.017 (see Figure 16 & Table 13). There was the unexpected finding that nonfamilial sex offenders rated significantly more benefit to the woman than did incestuous offenders, F(1,205)=7.569, p=.006, and showed a different pattern in their ratings across the levels of sexual contact, F(3,615)=3.259, p=.021. Figure 17 shows that the incestuous offenders decreased their ratings of the woman's benefit as the degree of sexual contact increased, while the nonfamilial offenders did not show this decrease in their ratings. In contrast, the nonfamilial offenders rated the woman's benefit as increasing as the sexual contact progressed from touching only to fondling and ejaculation.

Groups also differed significantly in their ratings of the degree to which the woman enjoyed the sexual contact, F(5,205)=2.509, p=.031. There was also a significant Group X Sex interaction, F(15,615)=1.831, p=.027. There were no significant differences among the groups on the "touching" or "fondling" vignettes, however, there were significant group differences on the "no clothes" F(5,205)=2.595, p=.027, and "ejaculation" vignettes F(5,205)=3.140, p=.009.

Nonfamilial offenders differed from incestuous offenders on the Group X Sex interaction, F(3,615)=3.976, p=.008, and on the Group X Woman X Sex interaction,





65

F(5,1025)= 3.171, p=.008. Figure 18 illustrates how both types of sex offenders showed the same pattern in their ratings of the woman's enjoyment when she smiled or was passive, but differed when the woman cried. Both groups rated the woman's degree of enjoyment as dependent on the level of sexual contact when she smiled or was passive. When she cried, however, nonfamilial offenders continued to rate her enjoyment as dependent on the level of sexual contact, while the incestuous offenders tended to rate her enjoyment as roughly the same across levels of sexual contact. On the ejaculation vignettes, the inmate controls attributed significantly more enjoyment to the woman than the other five groups, F(1,205)=11.337, p=.001 (see Table 14).

Groups differed significantly in their ratings of the woman's desire for the sexual interaction, F(5,205)=2.283, p=.048 (see Table 15 & Figure 19). The planned comparison between child molesters and controls on the Group X Woman X Sex interaction approached significance, F(5,1025)= 2.202, p=.052. Figure 20 shows that the same pattern was found between child molesters and controls as was found between nonfamilial and incestuous offenders on the variable of Woman's Enjoyment. Figure 20 illustrates that the child molesters and controls showed the same pattern in their ratings of the woman's desire when the woman smiled or was passive, but differed when the woman cried. When she cried, the child molesters' ratings of the woman's desire were more dependent on the level of sexual contact while the controls' ratings were roughly the same across levels of sexual contact.

There was also a significant difference among the groups in their ratings of the woman's sexual attraction to the man, F(5,205)=3.723, p=.003 (see Figure 21).





66


Unexpectedly, the inmate control group rated the woman as more sexually attracted to the man than did the other five groups F(1,205)=16.847, p=.000 (see Table 16 & Figure 21). There were no significant differences between child molesters and controls, F(1,205)=1.360, p=.245.

Groups did not significantly differ in their ratings of the harm done to the woman F(5,205)=1.933, p=.090. Planned comparisons showed that the difference between child molesters and controls on the Group X Sex interaction, F(3,615)=2.609, p=.051 approached significance. Figure 22 shows that child molesters ascribed movie harm to the woman than did controls across all levels of sexual contact, but particularly when the sexual contact involved touching or fondling. Both groups scored similarly when the sexual contact involved genital contact or ejaculation.

There were no significant differences among groups on the ratings of the woman's responsibility for the sexual interaction F(5,205)=1.492, p=. 194. There were also no significant interactions with Group.

Analysis of"Man's Role" variables. Although there were no significant differences among groups on the doubly MANOVA using these variables, further analysis was conducted in order to examine the difference found between admitters and deniers on Factor 2. ANOVAs were conducted for the variables of Man's Responsibility, Man's Anxiety, and Man's Confidence. Deniers attributed significantly less responsibility to the man than did admitters, F(1,205)=18.652, p=.000. Deniers also differed from admitters on the Group X Woman X Sex interaction, F(5,1025)=2.934, p=.012. Figure 23 shows that the deniers consistently attributed less responsibility to the man across vignettes. For





67

the most part, the deniers showed the same pattern of ratings as the deniers across the levels of sexual contact and the woman's response.

A post-hoc comparison showed that the nonfamilial deniers ascribed significantly less responsibility to the man than the other five groups combined, F(1,205)=11.789, p=.001 (see Table 17). Thus, much of the difference between the admitters and deniers may have been a result of the extreme pattern of the nonfamilial deniers. Comparison of child versus adult vignettes

Univariate analyses of variance were conducted for each variable with type of

vignette (child or adult) as the repeated measure. This analysis was conducted in order to examine if the subjects gave different ratings on the same variables in the child versus the adult vignettes. As expected, subjects differed significantly in their ratings on each variable on the child versus adult vignettes. Subjects attributed significantly more benefit to the woman than to the child, F(1,199)=433.662, p=.000, and attributed more harm to the child, F(1,199)=322.172, p=.000. Subjects also attributed more enjoyment and desire to the woman, F(1,199)=377.155, p=.000 and F(1,199)=671.605=.000, respectively. The woman was assigned significantly more responsibility for the sexual contact than the child, F(1,199)=704.403, p=.000, and more responsibility was attributed to the man for sexual contact with the child, F(1, 199)=301.838, p=.000. Groups differed significantly in their ratings across the vignettes on the variable of female's enjoyment, F(5, 199)=3.713, p=.003 (see Figure 24). Compared to the other groups, nonfamilial admitters rated the child as deriving more enjoyment, yet scored similarly to others in their ratings of the woman's level of enjoyment.




68

Analysis of Empathv

Child molesters were expected to score lower on empathy than controls, and deniers were expected to score lower than admitters. Three outliers were identified and removed from the analysis. Contrary to expectations, a planned comparison revealed no significant difference between child molesters and controls on the empathy total score F(1,199)=2.917, p=.089. A planned comparison showed that the nonfamilial sex offenders scored significantly higher on empathy total score than the incestuous sex offenders, F(1,199)=5.889, p=.016 (see Table 18). Contrary to predictions, there was no significant difference between admitters and deniers on total empathy score, F(1,199)-=.623, p=.431. An ANCOVA revealed a significant main effect of group F(5,199)=3.007,p=.012 (see Figure 25). The covariate was not significant, F(1,199)=.724, p=.396. All possible pairwise comparisons using the Bonferroni correction showed that the nonfamilial deniers scored significantly higher than the incestuous deniers on the total empathy score (p=.026).

In order to examine differences among the six groups on the four IRI subscales, an ANOVA was conducted with group as the between subjects variable and the four subscale scores as the repeated measures. The Group X Subscale interaction was not significant, F(5,205)=1.36, p=.162, indicating that differences among the groups did not vary across subscales.

Analysis of Repression

It was expected that child molesters would score higher on repression than controls, and that deniers would score higher on repression than admitters. These




69

hypotheses were not confirmed. Child molesters did not differ significantly from controls, F(1,201)=2.434, p=. 120, and admitters did not differ from deniers, F(1,201)=.362, p=.548 on the Repression-Sensitization Scale. An ANCOVA showed a significant effect for social desirability, F(1, 201)= 10.938, p=.001, but a nonsignificant effect for Group, F(5,201)-.804, p=.548 (see Table 3). The correlation of the repression scores and empathy total score was r=.188, X2=7.391, p=.007, indicating that higher scores on repression (which represent a tendency toward sensitization rather than repression) were positively correlated with higher scores on empathy.





70

















70




60




50




40


C= Inmate Controls 30 CC= Conmmunity Controls IA= Incestuous Admitters ID= Incestuous Deniers NFA= Nonfamilial Admitters 20 NFD= Nonfamilial Deniers
C CC IA ID NFA NFD Group

Figure 1. Percentage of each group reporting a history of abuse.





71
















22



20



18
is \



16

A NFA Nonfamilial Admitters NFD Nonfamilial Deiers 14 ----- IA Incestuous Admitters ID Inacestuous Deniers CC Community Controls 12 I 0 C Inmate Controls NO YES Abuse History

Figure 2. Mean social desirability ratings across groups.






72
















20 19







17 16 15

NONABUSE 14 ABUSE A D

Admitters Deniers


Figure 3. Mean social desirability ratings across groups.





73
















140





135





130





125 C= Inmate Controls CC= Community Controls IA= Incestuous Admitters ID= Incestuous Deiers NFA= Nonfamilial Admitters 120 NFD= Nonfamilial Deniers
C CC IA ID NFA NFD o roup

Figure 4. Mean scores across groups on the Cognition Questionnaire.






74
















2.5







2.0







1.5

C= Imate Controls CC= Commuty Controls IA= Incestuous Admitters ID= Incestuous Demniers NFA= Nonfamilial Admitters
1.0 NFD= Nonfamilial Deniers
C CC IA ID NFA NFD Group

Figure 5. Mean group ratings on Benefit to Child.





75
















3.0





2.5





2.0





1.5



NF Nonfamilial Offenders
1.0 I Incestuous Offenders
CRYING PASSIVE SMILING Child Response Figure 6. Mean Child Enjoyment ratings for levels of child response.






76















4I







3







2 A NFA Nonfamilial Admitters M NFD Nonfamilial Deiers O IA Incestuous Admitters ID IncestuousD Diers CC Community Contmls SIC Inmate Controls CRYING PASSIVE SMILING Child Response Figure 7. Mean Child Enjoyment ratings for levels of child response.





77
















3.0





2.5





2.0



i NFA Nonfamilial Admitters

1.5 NFD Nonfamilial Deniers O IA Incestuous Admitters
ID Incestuous Deniers CC Community Controls
1.0 O C Inmate Controls
Crying Passive Smiling Child Response Figure 8. Mean Child Desire ratings for levels of child response.





78
















2.5







2.0







1.5





NF Nonfamilial Offenders
1.0 OI Incestuous Offenders Crying Passive Smiling Child Response Figure 9. Mean Child Desire ratings for levels of child response.






79
















2.5







2.0







1.5





NF Nonfamili Offenders 1.0 0 1 Incestuous Offenders Crying Passive Smiling Child Response Figure 10. Mean Child Desire ratings for levels of child response.





80

















2.5







2.0







S1.5 1C Controls

1.0 0 CM Child Molesters Crying Passive Smiling Child Response Figure 11. Mean Child Desire ratings for levels of child response.






















2.5






2.0






1.5 A NFA Nonfamilial Admitters M NFD Nonfamilial Deniers O IA Incestuous Admitters
ID Incestuous Deniers CC Community Controls
1.0 O C Inmate Controls
Crying Passive Smiling Child Response

Figure 12. Mean Child Responsibility ratings for levels of child response.





82
















2.6




2.4



S2.2




8 2.0




1.8 CC= Community Controls IA= Incestuous Admitters ID= Incestuous Deniers NFA= Nonfamilial Admitters
1.6 NFD= Nonfamilial Deniers
C CC IA ID NFA NFD Group

Figure 13. Mean group ratings on Other Men.





83
















5.0





4.5





S.0





3.5



U C Controls

3.0 1 I CM ChildMolesters
3.0
1 2 3 4 Level of Sexual Contact Figure 14. Mean punishment ratings for levels of sexual contact.






84

















4.8 4.7


4.6 4.5 4.4 4.3


4.2

0 C Controls

4.1 O CM ChildMolesters
1 2 3 4 Level of Sexual Contact Figure 15. Mean treatment ratings for levels of sexual contact.





85
















3.0


2.9


2.8


2.7


S2.6


2.5
C= Inmate Contrls CC= Community Controls 2.4 -1IA= Incestuous Admitters ID= Incestuous Deniers NFA= Nonfamilial Admitters
2.3 NFD= Nonfamilial Deniers
C CC IA ID NFA NFD Group

Figure 16. Mean group ratings on Benefit to Woman.





86
















3.0 2.9 2.8


2.7 j 2.6


2.5


2.4
SNF
2.3
1 2 3 4 Level of Sexual Contact Figure 17. Mean Woman Benefit ratings for levels of sexual contact.





87



















4.3 4.2


4 40





3.8 3.7
NFA
3.6 o
1 2 3 4 1 2 3 4 I 2 3 4
Woman Smiling Woman Passive Woman Crying Level of Sexual Contact l=Touching 2=Fondling 3=No Clothes 4=Ejaculation Figure 18. Mean Woman Enjoyment ratings for levels of sexual contact and child's response.






















3.0





2.9





2.8





2.7 C= Inmate Controls CC= Community Controls IA= Incestuous Admitters ID= Incestuous Deniers NFA= Nonfamilial Admitters
2.6 NFD= Nonfamilial Deniers
C CC IA ID NFA NFD Group

Figure 19. Mean group ratings on Woman Desire.






89


















4.4


4.2 4.0


3.8


3.6

SCM
3.4 OC
1 2 3 4 1 2 3 4 1 2 3 4
Woman Smiling Woman Passive Woman Crying Level of Sexual Contact l=Touching 2=Fondling 3=No Clothes 4=Ejaculation

Figure 20. Mean Woman Desire ratings for levels of sexual contact and woman's response.





90
















3.3




3.2



3.1




S3.0


C= Inmate Controls
2.9 CC= Community Contols IA= Incestuous Admitters ID= Incestuous Deniers NFA= Nonfamilial Admitters
2.8 NFD= Nonfamilial Deniers
C CC IA ID NFA NFD Group


Figure 21. Mean group ratings on Sexual Attraction.





91
















3.3





3.2





g 3.1




3.0



CM Child Molesters
2.9 I I 0C Controls
1 2 3 4 Level of Sexual Contact Figure 22. Mean Woman Harm ratings for levels of sexual contact.





92


















4.0





3.5



..






2.5 OA
1 2 3 4 1 2 3 4 1 2 3 4
Woman Smiling Woman Passive Woman Crying Level of Sexual Contact l=Touching 2=Fondling 3=NoClothes 4=Ejaculation Figure 23. Mean Man Responsibility ratings for levels of sexual contact and child's response.





93
















4







3







2



Vignettes: ADULT
1 O [CHILD C CC IA ID NFA NFD Group Figure 24. Mean group ratings on Child Enjoyment and Woman Enjoyment.




Full Text
53
Child molesters and controls differed significantly on the Group X Child
interaction, F(10, 796)=2.134, p=.020, using the averaged multivariate test of significance.
The difference between incestuous and nonfamilial offenders on this interaction
approached significance, F(10, 796)=1.817, p=.054, on the averaged tests. There were no
significant differences among groups on the Group X Sex interaction. The difference
between child molesters and controls approached significance on the Group X Child X
Sex interaction, F(30, 4810)=1.428, p=.061, using the averaged multivariate tests.
The doubly MANOVA on the Seriousness of the Offense yielded a significant main
effect of Group, F(20, 657.64)=2.34, p=.001, and a significant Group X Sex interaction,
F(60, 893.47)=1.522, p=.008. Planned comparisons revealed that the child molesters
differed significantly from controls on the Seriousness of the Offense, F(4,198)=6.015,
p=.000. There were no significant differences among the other groups.
Child molesters differed from controls on the Group X Sex interaction,
F(12,190)=2.505, p=.004. The difference between admitters and deniers on the Group X
Sex interaction approached significance, F(12,1587.74)=1.744, p= .052, using the
averaged multivariate test of significance. The Group X Child X Sex interaction
approached significance for the planned comparison between admitters and deniers, F(24,
178)=1.516, p=067.
Analysis of variance. The significant MANOVA results were followed by
univariate analyses of variance on each of the nine variables. Group effects were
examined on each of the variables, with the Group X Child interaction examined for the


31
her(his) own decision as to whether she (he) wants to have sex with an adult or not, and 2) Sex
between a 13-year old (or younger) child and an adult causes the child no emotional problems.
Test-retest reliability for this questionnaire has been reported as .76 over a three week interval
(Abel, Becker, Cunningham-Rathner, Rouleau, et al. 1984). Six subscales were identified with
coefficient alphas ranging from .59 to .71. Scores on the questionnaire distinguished child
molesters from normals, demonstrating discriminant validity.
Measures of Decoding/Social Perception
Profile of nonverbal sensitivity
The Profile of Nonverbal Sensitivity (PONS) (Rosenthal et al., 1979) was used as a
general measure of nonverbal decoding ability. This widely used measure has the benefit of
distinguishing between several channels of nonverbal communication and various types of cues.
The PONS is a 47-minute videotape which is comprised of220 two second film clips in which
eleven nonverbal channels are isolated: three "pure" visual (face, body, figure) and two ''pure"
auditory channels (random spliced, content free), with six additional paired combinations In
each film segment, one of twenty affectively charged interpersonal situations is portrayed by a
woman who is not an actress. The twenty situations can be categorized into four affective
quadrants: positive-dominant, positive-submissive, negative-dominant, and negative-
submissive. Subjects are asked to choose which of two statements best describes the behavior
in each film clip. For example, on the first item the subject must decide whether the woman in
the film clip is expressing jealous anger or talking to a lost child. In another item the
respondent must choose between leaving on a trip or saying a prayer. The test was
designed to measure one's ability to read" and interpret nonverbal cues.


120
Although child molesters as a group did not differ significantly from controls, there
were significant differences among subtypes of child molesters on the adult vignettes.
More specifically, nonfamilial and incestuous offenders differed in their ratings of the
womans enjoyment and benefit from the sexual contact. On the benefit ratings,
incestuous offenders attributed less benefit to the woman as the degree of sexual contact
increased, while the nonfamilial offenders tended to attribute more benefit. In addition,
the incestuous offenders ratings of the womans enjoyment became less dependent on the
degree of sexual contact when the womans response was clearly negative. In contrast,
the nonfamilial offenders tended to vary their ratings of the womans enjoyment based on
the degree of sexual contact, although she was portrayed as crying. These results suggest
that the nonfamilial offenders may focus more on the degree of sexual contact as the most
important variable in a sexual interaction rather than the womans response to the sexual
contact. This might reflect a self-centeredness among nonfamilial offenders which might
help explain their noted lack of success in relationships with adult women.
There were several unexpected findings regarding the inmate control group.
Compared to the other five groups, the inmate controls attributed significantly more
enjoyment to the woman on the ejaculation vignettes which portray nonmutual orgasm
with the man ejaculating after rubbing his penis between the womans legs. In addition,
the inmate controls gave significantly higher ratings than the other groups on the womans
sexual attraction to the man. These findings suggests that nonsex offending inmates tend
to view women as being sexually interested regardless of their outward response (smiling,
passive, or crying) and as enjoying sexual contact focused primarily on the mans


45
Questionnaire were analyzed using separate analyses of variance. Group served as the
between subjects variable across analyses, and social desirability served as the covariate in
the analysis of the Cognition Questionnaire, empathy, and repression scores. The
assumptions for ANOVA were met for each of the dependent variables.
The vignette data was analyzed using principal components analysis followed by a
multivariate analysis of covariance For the univariate and multivariate analyses, planned
comparisons were conducted between child molesters and controls, admitters and deniers,
and between incestuous and nonfamilial sex offenders. Each of these comparisons was
considered to be its own separate family since each was thought to involve a conceptually
distinct comparison. Thus, the Bonferroni correction was not applied to these three
comparisons. The Bonferroni correction was used, however, when all possible pairwise
comparisons were conducted. All possible pairwise comparisons were examined since it
was expected that there might be differences among the six groups other than those
among the three families. More specific hypotheses than those regarding the three families
were not proposed since there has been little research on the differences among these
particular subgroups of child molesters. Thus, for example, differences between
incestuous deniers and nonfamilial admitters might be of interest, however, there was no
specific hypothesis regarding these two groups since there has been no research to date in
this area.
In addition to the planned comparisons and Bonferroni pairwise comparisons,
some specific post hoc comparisons were conducted between a particular group and the
five other groups combined. This contrast was conducted when one group seemed to


121
enjoyment. These results also suggest that cognitive distortions regarding sexual
interaction are not limited to sex offenders, particularly when they relate to adult women.
There were no significant differences among groups on the ratings of the mans
anxiety or the mans confidence, suggesting that child molesters did not view the man as
more anxious or less confident in the situation than did controls. While these questions
were designed to tap into the child molesters own anxiety and lack of confidence
regarding sexual contact with women, it may be that the subjects answered the questions
solely on the information in the vignette, without projecting any of their own feelings or
beliefs onto the man. Thus, the child molesters may have perceived the man in the
vignettes as confident without perceiving themselves as being confident in that same
situation.
Child molesters also did not differ from controls in their ratings of the womans
responsibility for the sexual contact. These results differ from those on the child vignettes
in that the nonfamilial admitters did not attribute significantly more responsibility to the
woman than the other groups as they did with the child. Thus, the nonfamilial admitters
cognitive distortion about the females responsibility was limited to sexual contact with
children and did not generalize to adult women.
Hypothesis #5 proposed that child molesters would show less empathy than
controls. Contrary to expectations, child molesters did not differ from controls in their
level of empathy. This finding is consistent with Marshall et als (1993) finding of no
differences between incarcerated child molesters and controls. It may be that child
molesters do not lack general empathy for others, but that they lack specific empathy for


Mean Child Desire Ratings
77
Crying Passive Smiling
Child Response
NFD
G IA
* ID
* CC
c
Nonfamilial Admitters
Nonfamilial Demers
Incestuous Admitters
Incestuous Deniers
Community Controls
Inmate Controls
Figure 8. Mean Child Desire ratings for levels of child response.


109
Table 18
Group means on empathy total score.
Nonfamilial Sex Offenders
M
SD
N
64 03
2.00
61
Incestuous Sex Offenders
M
SD
N
59.09
2.10
57
p=.016.


29
The subjects were asked to read each vignette and respond to six questions using a five
point Likert-type rating scale ranging from 1 (not at all) to 5 (definitely). The questions were
designed to reflect child molesters' central cognitive beliefs about adult sexual contact with
children, as described in previous clinical investigations. Subjects were asked the following
questions (the variable name for each question is in parentheses following the question): 1) Do
you think the child enjoyed what happened? (Child Enjoyment) 2) Do you think the child
wanted this to happen? (Child Desire) 3) Do you think the child could benefit from this
experience? (Child Benefit) 4) Do you think the child was responsible for what happened?
(Child Responsibility) 5) Do you think the child could be harmed by this experience? (Child
Harm) 6) Do you think the man was responsible for what happened? (Man Responsibility)
In order to assure face validity, Stermac and Segal (1989) asked subjects in a pilot
study to rank order the vignettes in terms of degree of sexual contact between the child and
adult. They were also given a multiple choice format and were asked to determine the child's
response. The authors reported complete agreement among subjects on these tasks.
Stermac and Segal (1989) conducted Pearson correlation matrices for each of the
twelve vignettes in order to examine correlations among the variables. They found that child
benefit, harm, and enjoyment were highly correlated across the levels of sexual contact and
childs response. Child benefit and enjoyment were negatively correlated with child harm, with
correlation coefficients ranging from .32 to .65. Child responsibility and childs desire were
also found to be highly correlated, with correlation coefficients ranging from .37 to .71. For
analysis purposes, the authors combined child benefit, enjoyment, and harm into a composite
variable called benefit to child, and combined child responsibility and desire into child


Mean Child Desire Ratings
80
C Controls
O CM Child Molesters
Figure 11. Mean Child Desire ratings for levels of child response.


70
70
C CC IA ID NFA NFD
Group
Figure 1. Percentage of each group reporting a history of abuse.
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers


Ill
this difference only approached significance. Interestingly, the child molesters gave
significantly lower ratings than controls on child enjoyment and desire when the child
cried. This may reflect child molesters tendency to overcompensate in their ratings when
the childs response is clearly negative and the socially desirable response is more obvious.
It might also suggest that child molesters are more aware of childrens lack of enjoyment
and desire when their responses are clearly negative.
Child molesters also prescribed less punishment for the offender regardless of the
degree of sexual contact, although their ratings became more similar to those of controls
as the degree of sexual contact increased. Thus, the child molesters pattern of responses
suggest that they used the degree of sexual contact to rationalize the sex offending
behavior, or assign it less punishment. This result is consistent with sex offenders
clinically observed cognitive distortion that sexual contact not involving intercouse is
relatively harmless (Salter, 1988). Child molesters also prescribed significantly less
treatment for the offender than did controls if the sexual contact involved touching only.
Although differences were found between child molesters and controls on the child
vignettes, most of the differences appeared to be attributable to the nonfamilial admitters,
a subtype within the child molester group Most of the significant differences were found
between the nonfamilial admitters and the five other groups combined. In contrast to the
other four sex offender groups, the nonfamilial admitters showed the response pattern that
was expected based on the Stermac & Segals (1989) results. Across several variables,
the nonfamilial admitters showed more cognitive distortions than the other groups when
the childs response was positive, but scored more similarly to the other groups when the


112
childs response was clearly negative. The nonfamilial admitters also tended to show more
cognitive distortions than other groups when the childs response was passive, however,
these differences only approached significance when the strict criterion of the Scheffe test
was used. Nonfamilial admitters showed this pattern on the variables of child benefit,
child enjoyment, child desire, and child responsibility. That is, when the child smiled, the
nonfamilial admitters attributed significantly more benefit, desire, enjoyment, and
responsibility to the child than did the other five groups. Nonfamilial admitters also rated
other men as more likely to molest children than did the other five groups. These results
suggest a severe lack of insight among nonfamilial admitters regarding the childs
perspective.
These results also suggest that nonfamilial admitters tend to read more than
others into the childs response when it is positive, perhaps in order to justify the behavior.
This is often seen among sex offenders in treatment settings who admit to rationalizing
their behavior by claiming that the child really enjoyed the contact since she didnt cry
(Salter, 1988). It appears that nonfamilial admitters may also use the excuse that other
men molest fairly often, thus making the behavior more acceptable. This particular
cognitive distortion is also seen frequently among sex offenders in treatment.
Interestingly, nonfamilial admitters did not show cognitive distortions regarding
the mans responsibility for the sexual interaction or his need for treatment. This suggests
that, although the nonfamilial admitters attribute more responsiblility, benefit, desire, and
enjoyment to the child, they do not show cognitive distortions regarding the mans role in
the sexual interaction. In fact, they were the only sex offender group which did not


113
prescribe significantly less punishment than inmate controls when the child cried. This
suggests that nonfamilial admitters may recognize that the child molester should be
punished, although they continue to maintain cognitive distortions about the childs
involvement. This also indicates that the differences between child molesters and controls
on punishment and treatment ratings are not attributable to the nonfamilial admitters.
Although the results suggest that these cognitive distortions are found only among
nonfamilial admitters, it may be that this is the only group which will admit to these
cognitive distortions on a face valid questionnaire. The data collected from the individual
interviews support this hypothesis, as all of the four child molester subgroups showed
cognitive distortions when responding to verbal vignettes depicting sexual contact
between a child and an adult. The less structured format of the interview allowed more
opportunity for subjects to express their views in their own words, and thus, potentially
reveal more cognitive distortions. As expected, each of the four child molester groups
revealed more cognitive distortions than the controls. Statements which revealed a
permissive attitude toward adult sexual contact with children were considered to be
cognitive distortions. For example, the first question asked how the 35-year old man feels
about the 9-year old girl with whom he is sexually involved. A common cognitive
distortion found among the child molesters involved the belief that the man loved the girl
and was appropriately expressing his love for her in a sexual way. In contrast, the controls
tended to express the idea that the man was exploiting this child for his own selfish sexual
gratification.


19
offenders to present themselves in a desirable light in order to facilitate early release from
jail. The authors conclude, however, that nonspecific empathy does not appear to be child
molesters' main problem. Rather, they recommend focusing on child molesters' empathy
toward specific targets such as their victims. Finkelhor & Lewis (1987) also suggest
exploration of this more specific type of empathy since otherwise-empathic men
sometimes sexually exploit children, (p.76)
Given the inconsistent results in the literature, the present study examined
generalized empathy among child molesters. Specific empathy for children was not
assessed since an assessment instrument of this type could not be found. Child molesters
hypothesized repression was thought to be related to a lack of empathy since repression
might allow the offender to block out the victim's response, thus resulting in impaired
empathy Lack of empathy might then facilitate the commission of the offense since
empathy is thought to play an important function in the inhibition of aggressive or
antisocial acts toward others, with vicarious experience of the victim's pain serving as a
deterrent (Miller & Eisenberg, 1988). Miller and Eisenberg (1988) conducted a meta
analysis of studies addressing this issue and found empathy to be negatively related to
aggression and antisocial, externalizing behaviors.
Empathy has been described in the clinical literature as both a cognitive and affective
response (Miller & Eisenberg, 1988). Empathy has been defined as the ability to take the
viewpoint of the other person and to respond affectively to the other's perceived emotional
experience (Lorr, Youniss, & Stefic, 1991). Since empathy involves achieving a cognitive
understanding of the other person's feelings, a lack of empathy might manifest itself in cognitive


Other Men Ratings
82
C CC IA ID NFA NFD
C= Inmate Controls
CC Community Controls
IA= Incestuous Admitters
ID= Incestuous Deniers
NFA= Nonfamilial Admitters
NFD= Nonfamilial Deniers
Group
Figure 13. Mean group ratings on Other Men.


30
complicity. Test-retest reliability data was not available for this measure. Thus, the present
study included this analysis (see the Results section).
In addition to the twelve vignettes with a child, twelve vignettes with an adult were
designed for the current study. This set of vignettes was designed to assess attitudes toward
aggressive sexual contact with an adult female. The same vignette scenes and questions were
used as in the child vignettes, although the female portrayed was a 30-year old female. In
addition to the above questions asked regarding the woman (instead of the child), the following
questions were added (the variable names are in parentheses): 1) Do you think the woman was
sexually attracted to the man? (Womans Sexual Attraction) 2) Do you think the man felt
anxious? (Mans Anxiety) 3) Do you think the man felt confident that the woman was attracted
to him? (Mans Confidence) These questions were added in order to assess the subject's
perception of the man's anxiety and confidence, as well as the subject's perception of the
woman's degree of sexual attraction to the man. Since child molesters are thought to have
difficulty with relationships with adult women, they were expected to score differently from
controls on these variables.
Cognition questionnaire
The Cognition Questionnaire (Abel, Becker, Cunningham-Rathner, Rouleau, et al.
1984) was also administered as a second measure of cognitive distortions regarding sexual
contact with children. This instrument contains 29 items and was developed for use with adult
child molesters. The subject reads statements which reflect values about adult sexual contact
with children and selects a response ranging from 1-5 (strongly agree strongly disagree).
Sample items on the Cognition Questionnaire include: 1) A child 13 or younger can make


65
F(5,1025)= 3.171, p=. 008. Figure 18 illustrates how both types of sex offenders showed
the same pattern in their ratings of the womans enjoyment when she smiled or was
passive, but differed when the woman cried. Both groups rated the womans degree of
enjoyment as dependent on the level of sexual contact when she smiled or was passive.
When she cried, however, nonfamilial offenders continued to rate her enjoyment as
dependent on the level of sexual contact, while the incestuous offenders tended to rate her
enjoyment as roughly the same across levels of sexual contact. On the ejaculation
vignettes, the inmate controls attributed significantly more enjoyment to the woman than
the other five groups, F(l,205)=l 1.337, p=.001 (see Table 14).
Groups differed significantly in their ratings of the womans desire for the sexual
interaction, F(5,205)=2.283, p=. 048 (see Table 15 & Figure 19). The planned comparison
between child molesters and controls on the Group X Woman X Sex interaction
approached significance, F(5,1025)= 2.202, p=.052. Figure 20 shows that the same
pattern was found between child molesters and controls as was found between nonfamilial
and incestuous offenders on the variable of Womans Enjoyment. Figure 20 illustrates
that the child molesters and controls showed the same pattern in their ratings of the
womans desire when the woman smiled or was passive, but differed when the woman
cried. When she cried, the child molesters ratings of the womans desire were more
dependent on the level of sexual contact while the controls ratings were roughly the same
across levels of sexual contact.
There was also a significant difference among the groups in their ratings of the
womans sexual attraction to the man, F(5,205)=3.723, p=.003 (see Figure 21).


76
NFA Nonfamilial Admitters
NFD Nonfamilial Deniers
^ IA Incestuous Admitters
* ID Incestuous Deniers
^ CC Community Controls
O C Inmate Controls
Child Response
Figure 7. Mean Child Enjoyment ratings for levels of child response.


62
Analysis of variance was conducted with Group as the between subjects factor,
Womans Response (Woman) and Level of Sexual Contact (Sex) as the within subjects
factors, and Factor 1 as the dependent measure. There was a significant main effect of
Group, F(5,202)=3.124, p=010. Planned comparisons, however, showed no significant
differences between child molesters and controls, incestuous and nonfamilial offenders, or
admitters and deniers. The significant group effect appears to be related to the inmate
controls who scored higher on this factor than the other five groups, although this
difference was not significant using the Scheffe test, F(l,205)= 10.096, p=.02.
Analysis of variance using Factor 2 as the dependent measure also showed a
significant main effect of Group, F(5,205)=3.436, p=.005. Deniers scored significantly
different from admitters on Factor 2, F(l,202)=10.560, p=.001. Much of this difference
appears to be attributable to nonfamilial deniers who scored higher than the other groups
on this factor, although not significantly higher using the post-hoc Scheffe test,
F(l,205)=8.862, p=.003.
Multivariate analysis of variance. Based on the results of the principal components
analysis, the nine variables on the adult vignettes were divided into two variable sets, one
relating to the Womans Complicity and the other relating to the Mans Role A doubly
multivariate analysis of variance was conducted with each variable set. Social desirability
was initially used as a covariate in the analysis, however, it was found to be nonsignificant
with Womans Complicity, F(6,197)=1.678, p=. 128, and with Mans Role,
F(3,200)=2.351, p=,074, so it was dropped from further analysis.


CHAPTER 2
METHODS
Subjects
There was a total of 214 subjects that completed the study. Subjects were divided into
six groups: nonfamilial heterosexual male child molesters [admitters (n=31) and deniers
(n=32)], incestuous heterosexual male child molesters (admitters (n=26) and deniers (n=34)],
nonsex offending male inmates (n=60), and male community controls (n=31). All subjects
were at least 18 years of age. Subjects were recruited from the following correctional facilities
in Florida: Union Correctional Institution, New River Correctional Institution, and Marion
Correctional Institution.
In order to qualify for the study, subjects in the child molester groups must have been
convicted of a sexual offense against a female who was under the age of 12. They could also
not report current or prior participation in an inpatient sex offender treatment program since
differential treatment histories could serve as a confounding variable. The non-sex offending
inmate group was selected on the basis of having no prior record of sex offenses according to
self-report and their prison files. The community control group was recruited through local
advertisements posted in areas designed to attract a low socioeconomic group. In addition,
subjects were solicited at local construction companies and various volunteer organizations.
Child molesters were divided into admitters and deniers based on their written self-
report to the question Did you commit the crime(s) that you are charged with? If not, please
26


128
Bryer, J.B., Nelson, B.A., Miller, J.B., & Krol, P.A. (1987). Childhood physical and sexual
abuse as factors in adult psychiatric illness. American Journal of Psychiatry. 144. 1426-
1430.
Byrne, D. (1961). The Repression-Sensitization Scale: Rationale, reliability, and validity
Journal of Personality. 29,334-349.
Byrne, D., Bany, J., & Nelson, D (1963). Relation of the Revised Repression-Sensitization
Scale to Measures of Self-Description. Psychological Reports. 13,323-334.
Camras, L.A., Grow, J.G., & Ribordy, S.C. (1983). Recognition of emotional expression by
abused children. Journal of Clinical Child Psychology. 12, 325-328.
Costanzo, M. & Archer, D. (1989). Interpreting the expressive behavior of others: The
Interpersonal Perception Task. Journal of Nonverbal Behavior. 13 (41. 225-243.
Crowne, D., & Marlowe, D. (1960). A new scale of social desirability independent of
psychopathology. Journal of Consulting Psychology. 24, 349-354.
Dahlstrom, W.G., Welsh, G.S., & Dahlstrom, L.E. (1975). An MMPI handbook: VolU
Research Applications. Minneapolis, MN: University of Minnesota Press.
Davis, M.H. (1979). Individual differences in empathy: A multidisciplinary approach
(Doctoral dissertation, University of Texas, Austin, 1979). Dissertation Abstracts
International. 40, 3480B.
Davis, M.H. (1980). A multidimensional approach to individual differences in empathy JSAS
Catalog of Selected Documents in Psychology. JO, 85.
Davis, M.H. (1983). Measuring individual differences in empathy: Evidence for a multi
dimensional approach. Journal of Personality and Social Psychology. 44 ill. 113-126.
Dill, D., Chu, J., Grob, M., & Eisen, S. (1991). The reliability of abuse history reports: A
comparison of two inquiry formats. Comprehensive Psychiatry. 32(2), 166-169.
Dwyer, M., Rosser, S., & Sawyer, S. (1992). Dissociative experiences of sexual offenders: A
comparison between two outpatient groups and those found to be falsely accused. Special
Issue: Sex offender treatment: Psychological and medical approaches Journal of Offender
Rehabilitation. 18(3-41. 49-58.
Erickson, W.D., Luxenberg, M.G., Walbek, N.H., Seely, R.K. (1987). Frequency of MMPI
two-point code types among sex offenders. Journal of Consulting and Clinical Psychology.
55(41, 566-570.


72
Admitters Deniers
Figure 3. Mean social desirability ratings across groups.


32
The PONS was standardized on 492 subjects. Test-retest reliabilities average .81 with
an interval of 6 to 8 weeks between testing, although they vary over the three dimensions of
voice tone (lowest reliability), facial expressions, and body movements (Rosenthal et al., 1979).
The internal consistency coefficients were reported as 86 for total score and as follows for the
affective quadrants: positive-submissive (.78), positive-dominant (.84), negative-submissive
(.87), and negative-dominant (.87).
Principal components analysis on the eleven channel accuracy scores yielded four
factors measuring: 1 )decoding of facial cues, 2) decoding of randomized spliced cues, 3)
decoding of content-filtered cues, and 4) decoding of cues from the body without the face.
Principal components analysis was also conducted for the negative and positive affect scenes,
dominant and submissive channels, and the twenty scenes. The factor analysis results of the
twenty scenes suggests that decoding skill on the PONS can be subdivided into six skills
involving decoding cues of interpersonal control, cues of contrition, cues of strong but
controlled affects, cues of separation, cues of dependence, and cues associated with recitation.
The results of these factor analyses provide support for the various channels and affective
quadrants assessed by the PONS.
Interpersonal perception task
General social perception was measured using the Interpersonal Perception Task (IPT)
(Costanzo & Archer, 1989) which "is designed to assess accuracy in interpreting the expressive
behavior of others" (p.231). The IPT is a 38-minute videotape depicting naturalistic full-
channel sequences of behavior. There are 30 scenes which show one to four people. A
question appears on the screen before each scene, requiring the viewer to answer the question


104
Table 12
Rotated loadings for two factors derived from principal components analysis on the adult
vignettes
Components
Variables
1
2
Woman Desire
.915
.092
Woman Enjoyment
.905
.054
Woman Attraction
819
.228
Woman Respon
.814
.123
Woman Benefit
.626
-.058
Woman Harm
-.505
.130
Man Anxiety
.048
.766
Man Confidence
.321
.725
Man Respon.
-.400
.594
Table 13
Group means on Benefit to Woman across vignettes.
Incest Admitters Inmate Controls
M
2.40b
M
2.90a
SD
.14
SD
.09
N
25
N
60
Incest Deniers
M
2.62
Community Controls
M 2.59
SD
.12
SD
.13
N
33
N
30
Nonfamilial Admitters
M
2.92
Nonfamilial Deniers
M 2.80
SD
.13
SD
.12
N
31
N
32
Note. Means with different subscripts differ significantly at p<05, using Bonferroni
correction.


59
expected. Nonfamilial offenders rated other men as more likely to molest children than
did incestuous offenders, although this difference only approached significance,
F(l,202)=3.423, p=.066.
As predicted, admitters were more likely than deniers to rate other men as willing
to engage in sexual contact with a child, F(l,202)=3.853, p=.051, although this difference
only approached significance. A post-hoc comparison showed that, compared to the other
five groups, nonfamilial admitters rated other men as significantly more likely to molest
children, F(l,202)=12.625, p=.000 (see Table 11). There were no significant differences
between child molesters and controls, nonfamilial and incestuous offenders, and admitters
and deniers on the Group X Child interaction.
Analysis of Seriousness of Offense variables. As predicted, the groups differed
significantly in their ratings of the amount of punishment the molester should receive,
F(5,202)=6.623, p=.000. The Group X Sex Contact interaction was also significant,
F(15,606)=2.954, p=.003. Planned comparisons showed that the child molesters
prescribed significantly less punishment than the controls, (1,202)=28.90, p=.000. In
contrast to results on the other eight variables, this significant result did not appear to be
largely affected by the pattern of the nonfamilial admitters. Child molesters and controls
also differed significantly on the Group X Sex interaction, F(3,606)=9.905, p=.000.
Figure 14 shows that the child molesters prescribed significantly less punishment than the
controls across all four levels of sexual contact: touching F(l,202)=33 .470, p=.000,
rubbing F(l,202)=28.924, p=.000, fondling F( 1,202)=18.428, p=.000, and ejaculation
F( 1,202)=11.647, p=.001. The difference between the child molesters and controls was


The child molesters also tended to express the belief that the girl must love the
man or that she would not be sexually involved with him. They tended to view her as
consenting to the sexual interaction and as participating as an equal partner in the
relationship. In contrast, the controls tended to describe the child as feeling afraid and
unsure about the sexual interaction. In addition, child molesters most often indicated that
they would not interfere with this relationship if they were friends with the offender In
contrast, the controls usually indicated that they would try to intervene in some way in
order to stop the sexual abuse from continuing. Most of the control subjects also
indicated that they would either take physical action against the perpetrator or would
notify the police of the sexual offense In sum, the data from the interviews offer support
for the notion that child molesters show cognitive distortions, although their endorsement
of these distortions may vary according to the structure and format of the assessment
method.
Although child molesters as a group show more cognitive distortions than
controls, differences were found among the subtypes of child molesters. The results from
the child vignettes offered strong support for the hypothesis that nonfamilial offenders
show more cognitive distortions than incestuous offenders. In contrast to the incestuous
offenders, the nonfamilial offenders attributed more benefit and desire to the child, and
attributed more enjoyment and responsibility to the child when the child smiled or was
passive. Nonfamilial offenders ratings did not differ from those of incestuous offenders
when the child cried. Thus, nonfamilial offenders, particularly nonfamilial admitters,
showed significant differences compared to incestuous offenders in their perception of the


67
the most part, the deniers showed the same pattern of ratings as the deniers across the
levels of sexual contact and the womans response.
A post-hoc comparison showed that the nonfamilial deniers ascribed significantly
less responsibility to the man than the other five groups combined, F(l,205)=l 1.789,
p=.001 (see Table 17). Thus, much of the difference between the admitters and deniers
may have been a result of the extreme pattern of the nonfamilial deniers.
Comparison of child versus adult vignettes
Univariate analyses of variance were conducted for each variable with type of
vignette (child or adult) as the repeated measure. This analysis was conducted in order to
examine if the subjects gave different ratings on the same variables in the child versus the
adult vignettes. As expected, subjects differed significantly in their ratings on each
variable on the child versus adult vignettes Subjects attributed significantly more benefit
to the woman than to the child, F(l,199)=433.662, p=.000, and attributed more harm to
the child, F(l,199)=322.172, p=.000. Subjects also attributed more enjoyment and desire
to the woman, F(l,199)=377.155, p=.000 and F(l,199)=671.605=.000, respectively The
woman was assigned significantly more responsibility for the sexual contact than the child,
F(l,199)=704.403, p=.000, and more responsibility was attributed to the man for sexual
contact with the child, F(l,199)=301.838, p=.000. Groups differed significantly in their
ratings across the vignettes on the variable of females enjoyment, F(5,199)=3.713, p= 003
(see Figure 24). Compared to the other groups, nonfamilial admitters rated the child as
deriving more enjoyment, yet scored similarly to others in their ratings of the womans
level of enjoyment.


48
comparisons using the Bonferroni correction showed that the nonfamilial deniers endorsed
significantly more cognitive distortions than the community control group.
Clinical Vignettes
Child vignettes
Child molesters were expected to show more cognitive distortions on the child
vignettes than the controls, and admitters were expected to show more cognitive
distortions than the deniers Based on Stermac & Segals (1989) results, child molesters
were expected to ascribe more benefit to the child than control groups when the childs
response was positive or ambiguous, but to show no difference when the childs response
was clearly negative The variable of benefit was a composite of the variables benefit to
child, harm to child, and childs enjoyment. Compared to controls, child molesters were
also expected to attribute more responsibility and desire to the child and less responsibility
to the adult.
Test-retest reliability. The child vignettes were readministered to subjects after a 3
to 4 week interval. Scores on each of the nine variables were summed across the twelve
vignettes for each administration and Pearson correlations were calculated to measure
test-retest reliability from Trial 1 to Trial 2. The test-retest reliability for the variables is as
follows: Child Benefit (r=.83), Child Harm (r= 82), Child Enjoyment (r=.83), Child
Desire (r=.66), Mans Responsibility (r=.71), Treatment (r=,80), Other Men (r=.81), and
Punishment (r=.72). Analysis of the specific variables on each vignette also showed
strong correlations between the first and second administrations, demonstrating that
subjects tended to respond in a consistent manner on this measure over time.


20
distortions about the other person's feelings. Thus, the cognitive distortions seen in child
molesters could be related to a lack of empathy as well as repression, both of which could be
related to poor decoding skills.
Child molesters' perceived lack of empathy could also be related to their reported
histories of abuse. Miller and Eisenberg (1988) concluded from their meta-analysis that abused
subjects tend to have lower levels of empathy than nonabused subjects. Abused children are
thought to have little opportunity to learn to identify others' affective states or experience
empathic responding themselves since their needs and feelings are not recognized or
appropriately responded to in interactions with their abusive parents. Thus, abused children
tend to lack models of appropriate behavioral and emotional responding to others in need, and
as a result, show deficiencies in empathy which remain throughout adulthood.
In sum, child molesters' perceived lack of empathy was thought to be related to
their history of childhood abuse, repression, and poor decoding ability. In addition, lack
of empathy may play a role in their cognitive distortions regarding sexual contact with
children.
Rationale for the Proposed Study
This study focused on two aspects of child molesters' social information processing:
decoding skill and cognitive distortions. The literature has indicated that child molesters are
socially deficient, however, the nature of this deficiency is unclear. That is, this deficiency may
result from any number of problems in their cognitive operations, including deficiencies in their
decoding skill. Since the first step in exploring deficiencies in social information processing


108
Table 17
Group means on Mans Responsibility across vignettes.
Incest Admitters Inmate Controls
M
4.00**
M
3.70
SD
.11
SD
.07
N
25
N
60
Incest Deniers
M
3.58*
Community Controls
M 3.79
SD
.10
SD
.10
N
33
N
30
Nonfamilial Admitters
M
3.88b**
Nonfamilial Deniers
M 3.42a*
SD
.10
SD
.10
N
31
N
32
Note. Means with different subscripts differ significantly at g<05, using Bonferroni
correction. Mean with subscript a" differs significantly from other five groups combined,
p=.001.
Note. The two means denoted significantly differ from the means denoted **, p=.000.


Mean Treatment Ratings
84
C Controls
CM Child Molesters
Level of Sexual Contact
Figure 15. Mean treatment ratings for levels of sexual contact.


33
based on the scene that follows. During the six second interval following the scene, the viewer
chooses the correct answer from a multiple choice format. For example, the first scene shows
an adult male and female having a conversation with two seven year-old children The viewer is
asked "Who is the child of the two adults?" Since the questions are based on actual,
naturalistic scenes there is an objective criterion of accurate judgment for each item.
The IPT was standardized on 438 college students. The accuracy rates for the scenes
are reported as comparable to those for other similar measures. Accuracy rates are listed for
each scene. Test-retest reliability using 46 subjects after a five week break was .70. Internal
consistency was .52 using the KR-20, which the author suggests may reflect the diversity and
relatively small number of scenes. Subjects who scored in the high social sensitivity group
based on 17 peer ratings scored significantly higher in the IPT than subjects in the low social
sensitivity group, demonstrating construct validity. In addition, the correlation between IPT
scores and social sensitivity scores based on peer ratings was r=.48. The authors also report
significant positive correlations between the IPT and two other standardized measures of ability
to interpret the expressive behavior of others.
Measure of Empathy
Empathy was measured using the Interpersonal Reactivity Index (IRI, Davis, 1979)
which Marshall et al. (1993) cited as the best available measure for assessing the complex
nature of empathy. The IRI is a 28-item self-report questionnaire with four 7-item subscales
measuring different aspects of empathy. The four subscales are 1) perspective-taking, 2)
fantasy 3) empathic concern, and 4) personal distress. The Perspective-Taking (PT) scale
assesses the tendency to adopt others' psychological points of view. The Fantasy scale (FS)


132
Salter, A. (1988). Treating Child Sex Offenders and Victims. Sage: Beverly Hills, CA.
Segal, Z.V. & Marshall, W. (1985). Heterosexual social skills in a population of rapists and
child molesters. Journal of Consulting and Clinical Psychology. 53, 55-63.
Segal, Z.V., & Marshall, W. (1985b). Heterosexual social skills in a population of rapists and
child molesters. Journal of Consulting and Clinical Psychology. 53.
55-63.
Segal, Z.V., & Marshall, W. (1986). Discrepancies between self-efficacy predictions and
actual performance in a population of rapists and child molesters. Cognitive Therapy and
Research- 1CX33. 363-376.
Segal, Z.V., & Stermac, L.E. (1990). The role of cognition in sexual assault In Marshall,
W.L., Laws, D.R., & Barbaree, H E. (Eds ). The Handbook of Sexual Assault: Issues.
Theories, and Treatment of the Offender (pp. 101-120). New York: Plenum Press.
Stermac, L.E., & Segal, Z. (1989). Adult sexual contact with children: An examination of
cognitive factors Behavior Therapy. 20, 573-584.
Wechsler, D. (1981). Wechsler Adult Intelligence Scale-Revised Manual. The Psychological
Corporation: Cleveland, OH.


Mean Child Responsibility Ratings
81
Crying Passive Smiling
A NFA
NFD
a IA
* ID
* CC
C
Child Response
Nonfamilial Admitters
Nonfamilial Deniers
Incestuous Admitters
Incestuous Deniers
Community Controls
Inmate Controls
Figure 12. Mean Child Responsibility ratings for levels of child response.


11
Christophe, 1986). Abel et al. (1981) found, however, that incestuous offenders showed
sexual arousal to stimuli depicting sexual contact with children.
Pawlak, Boulet, and Bradford (1991) found differences between incestuous and
nonfamilial offenders on three of the subscales on the Derogatis Sexual Functioning Inventory
Incestuous offenders scored higher than nonfamilial offenders on sexual experience and sexual
satisfaction. The authors note, however, that this difference is difficult to interpret since these
subscale scores are based on factors which may be largely affected by involvement with a
regular sexual partner, and unfortunately, marital status differed greatly across the two groups.
As expected, incestuous offenders tended to be married while the majority of nonfamilial
offenders were single. In addition, nonfamilial offenders scored higher than incestuous
offenders on sexual fantasy, meaning that the nonfamilial offenders endorsed more items
depicting common sexual themes. This higher score on sexual fantasy may reflect a greater
need among nonfamilial offenders to resort to fantasy since their sexual experience and sexual
satisfaction are lacking. Although the two groups differed on these three subscales, the subtest
scores overall did not effectively discriminate between the two groups.
Marshall, Barbaree, and Christophe (1986) found differences between incestuous
and nonfamilial offenders in their erectile responses to sexual stimuli. Compared to
incestuous offenders and nonoffenders, the nonfamilial offenders showed greater arousal
to the stimuli involving children. Incestuous offenders, on the other hand, showed less
arousal to adult female stimuli than the other two groups. In a later study, Barbaree and
Marshall (1989) found that incestuous offenders showed either a non-discriminating sexual
arousal profile or an adult preference pattern while the nonfamilial group was more


69
hypotheses were not confirmed. Child molesters did not differ significantly from controls,
F(l,201)=2.434, p=. 120, and admitters did not differ from deniers, F(l,201)=362, p=,548
on the Repression-Sensitization Scale. An ANCOVA showed a significant effect for
social desirability, F(l, 201)= 10.938, p=.001, but a nonsignificant effect for Group,
F(5,201)=.804, p=.548 (see Table 3). The correlation of the repression scores and
empathy total score was r=.188, X2=7.391, p=.007, indicating that higher scores on
repression (which represent a tendency toward sensitization rather than repression) were
positively correlated with higher scores on empathy.


71
A NFA
NFD
IA
* m
Nonfamilial Admitters
Nonfamilial Demers
Incestuous Admitters
Incestuous Deniers
Community Controls
Inmate Controls
Figure 2. Mean social desirability ratings across groups.


The principal components analysis was conducted to reduce the number of
variables in an effort to minimize the redundancy associated with results involving multiple
highly correlated dependent measures. Thus, the two factors were used as dependent
measures in analyses of variance. An ANOVA conducted with Group as the between
subjects variable and Factor 1, Child Complicity, as the dependent measure yielded a
significant main effect of group, F(5,202)= 3 .91, p=.002. Planned comparisons showed
that child molesters did not attribute more complicity to the child than controls, F( 1,202)=
.369, p=.544. A post-hoc comparison, however, showed that nonfamilial admitters, one
of the child molester groups, attributed significantly more complicity than all of the other
groups combined, F(l,202)= 16.835, which was significant using the Scheffe test. As
expected, nonfamilial sex offenders as a group attributed more complicity than incestuous
offenders, F(l,202)= 6.050, p=.015, however, much of this difference may be attributed to
the nonfamilial admitters. Similarly, the admitters scored higher on this factor than the
deniers, F(l,202)=3.638, p=,058, although much of this difference is also probably
attributable to the nonfamilial admitters.
An ANOVA with Factor 2 as the dependent measure also yielded a significant
main effect of group, F(5,202)= 2.480, p=.033, indicating a significant difference among
groups in their attitudes regarding the mans responsibility and the seriousness of the
sexual behavior Planned comparisons showed that child molesters rated the sexual
behavior as less serious and less deserving of severe punishment than controls,
F(l,202)=8.040, p=,005. There were no significant differences between admitters and
deniers or incestuous and nonfamilial offenders.


14
were not less responsive to child cues. Thus, deniers tend to show a different pattern of
sexual responsivity, suggesting that they may constitute a clinically distinct subgroup.
Given these differences, deniers and admitters may also show different response patterns
on other variables such as on cognitions and social perception. It is interesting to note that
the adolescent sex offenders in the Becker et al. (1992) study showed the greatest sexual
response to age-appropriate female cues, supporting the hypothesis that child molesters would
prefer to have a sexual relationship with a peer, but may turn to younger females because of
difficulty with initiating or maintaining a sexual relationship with an adult female. This finding
is consistent with social skills research and anecdotal information on child molesters (Marshall,
Laws, & Barbaree, 1990) which suggests that sexual assaulters have difficulty forming and
maintaining healthy intimate relationships with women because of their lack of social
competence.
Haywood, Grossman, and Hardy (1993) also found significant differences between
admitters and deniers. Deniers showed more minimization than admitters on the Fake-
Good Scale on the Sixteen Personality Factors Inventory while admitters showed more
exaggeration of problems on the Fake-Bad scale. Both scales were able to correctly
discriminate between admitters and deniers. Deniers were more likely to be defensive
about undesirable personality traits than were admitters. Interestingly, the personality
factor which was most highly correlated with the Fake-Good scale was emotional
sensitivity. This scale is associated with "affective maturity, lack of anxiety, and ability to


controls). Within the six levels, comparisons were made between child molesters and
controls, nonfamilial and incestuous offenders, and admitters and deniers.
28
Group served as the sole independent variable in the analysis of all of the
dependent variables except for the cognitive distortions. This dependent variable was
assessed using clinical vignettes which contained two independent variables which varied
across the vignettes Level of Sexual Contact (4 levels) and Females Response (3 levels).
Thus, a 6x4x3 design was used in the analysis of the vignette data, with Group as a
between subjects variable and Level of Sexual Contact and Females Response as within
subjects variables. The following section lists each of the dependent variables and the
specific measures used.
Dependent variables
Measures of Cognition
Clinical vignettes
Cognitions regarding sexual contact were measured using the clinical vignettes from
Stermac and Segal's (1989) study. There were twelve vignettes describing sexual contact
between a male adult and seven year old child. The vignettes were designed by Stermac and
Segal (1989) based on documented reports of actual incidents between children and adults.
There are two independent factors in each vignette the degree of sexual contact and the
child's response to the contact. The levels of sexual contact include: 1) touching only, 2)
rubbing genital area over clothing (fondling), 3) fondling genitals under clothing (no clothes),
and 4) genital contact with ejaculation. The levels of the child's response include: 1) smiling,
2) passive/no response, and 3) crying with resistance.


3
general assertion subscale on the Callner-Ross Assertiveness Scale. In a follow-up study, Segal
and Marshall (1986) investigated social perception in child molesters in heterosexual situations
by having them interact with a female confederate. Compared with other sex offenders and
community controls, child molesters were found to be significantly worse at prediction and
evaluation of their own performance. Taken together, these findings suggest that child
molesters perceive themselves as less socially skilled and assertive than others, and are more
anxious than others about their social interaction. Given the inconsistent results, it is unclear
whether or not child molesters are actually less socially skilled than rapists and non-sex-
offending inmate controls.
Although the literature generally indicates lower social competence among child
molesters, there is currently no specific theory or clear pattern of deficits regarding child
molesters' social information processing Their interpersonal functioning may be affected by a
number of factors including accuracy of processing social cues and information. In addition to
appropriate response skills, "ability to accurately read the social environment," or decoding
skill, is necessary for social performance (Morrison & Bellack, 1981, p.70). Decoding skills
refer to "the afferent processes involved in the accurate reception, perception, and
interpretation of incoming sensory information (McFall, 1990, p. 313)." Ifa man does not
receive a woman's cues or misperceives or misinterprets these cues, then his social behavior
toward the woman is more likely to be inappropriate. Based on his social information
processing model, McFall (1990) suggests that the logical place to begin exploring social
information processing is at the beginning of the input process with decoding skills. This is the
point where the information input is more or less equivalent for everyone.


43
than deniers, Z=2.158, p=.015. The differences between incestuous and nonfamilial sex
offenders was not significant, Z=1.276, p=.202, two-tailed.
In order to examine differences between abused and nonabused subjects, several
one-way ANOVAs were conducted with abuse history as the between subjects factor and
total scores from the IPT, PONS, empathy measure, Cognition Questionnaire, Social
Desirability Scale, and Repression Scale as the dependent measures. Several ANOVAs
were used because group differences on each variables were of interest. There was a
significant difference between abused and nonabused subjects on repression,
F(l,208)=4.952, p=.027, indicating that those who denied a history of abuse scored higher
on repression than those reporting a history of abuse (see Table 1). When Group was
added as a second between subjects factor, however, the main effect of abuse history
became nonsignificant, F(l,198)=2.972, p=.086. There was also a significant main effect
of abuse history on social desirability, F(l,210)=8 491, p= 004, with those reporting a
history of abuse scoring significantly lower than nonabused subjects on social desirability
(see Table 1). When Group was added as the second between subjects factor, there was a
significant main effect of Group, F(5,200)=4.238, p=.001, and a significant Abuse X
Group interaction, F(5,200)=3.032, p=.012. Figure 2 shows that there was no consistent
pattern between abused and nonabused subjects across the six groups.
To examine the relationship between abuse history and admission/denial, an
ANOVA was conducted with Abuse and Admit as the two between subjects variables.
The community control subjects were not included in this analysis since the variable of
admission/denial was not relevant to them. There was a significant main effect of Abuse,


97
Table 4
Mean scores on the Cognition Questionnaire.
Incest Admitters
Inmate Controls
M
128.09
M
130.24
SD
3.02
SD
1.86
N
22
N
55
Incest Deniers
Community Controls
M
130 88
M
134.64*
SD
2.58
SD
2.56
N
29
N
29
Nonfamilial Admitters
M
124.27
SD
2.61
N
28
Nonfamilial Deniers
M
124.00*
SD
2.47
N
32
TOTALS:
Child molesters
Controls
M
126.81
M
132.44
SD
2.67
SD
2.21
N
111
N
84


101
Table 9
Group means on Child Enjoyment when the childs response was positive.
Incest Admitters Inmate Controls
M
2.28b
M
2.34b
SD
.20
SD
.12
N
23
N
61
Incest Deniers
Community Controls
M
2.41
M
2.07b
SD
.17
SD
.18
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
3.09a
M
2.39
SD
.17
SD
.17
N
31
N
31
Note. Means with different subscripts differ significantly at p<05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.


125
future study might focus on differences between ones self-report of empathy versus other
measures of empathy such as more objective behavioral measures.
Summary of Findings
In sum, the major findings of this study were: 1) Child molesters did not show a
deficit in decoding skill or social perception; 2) Child molesters showed more cognitive
distortions than controls regarding sexual contact with children, however, most of these
distortions were found among the nonfamilial admitters; 3) Nonfamilial admitters showed
cognitive distortions on both the child and adult vignettes; 4) Incestuous offenders showed
significantly fewer cognitive distortions than nonfamilial offenders, however, incestuous
offenders did show some cognitive distortions; 4) Deniers scored higher on social
desirability than admitters; 5) Child molesters did not differ significantly from controls on
measures of empathy, although nonfamilial offenders scored higher on empathy than
incestuous offenders; and 6) There were no differences among groups on a measure of
repression.
Conclusions
In conclusion, these findings highlight the importance of subdividing child
molesters into groups based on their relationship to the victim. Subdividing child
molesters based on admission or denial appears to be of less value, given that the primary
difference between admitters and deniers involves degree of social desirability. Admission
or denial of the crime is more likely to change as the offender enters treatment and is thus
less stable as a defining characteristic of the child molester. The offenders relationship to
the victim, however, appears to be a much more valuable distinction, given that


25
unconscious defense mechanisms In this case, deniers would be expected to score significantly
higher on social desirability than the other groups, but not higher on repression. Another
possible alternative is that the child molesters may not be deficient on general nonverbal
decoding skill, but may show a deficiency on the more complex social perception task. Thus,
their deficiency may be more related to difficulties in decision skills than in decoding skills.


66
Unexpectedly, the inmate control group rated the woman as more sexually attracted to the
man than did the other five groups F(l,205)=16.847, p=.000 (see Table 16 & Figure 21).
There were no significant differences between child molesters and controls,
F(l,205)=1.360, p= 245.
Groups did not significantly differ in their ratings of the harm done to the woman
F(5,205)=1.933, p=.090. Planned comparisons showed that the difference between child
molesters and controls on the Group X Sex interaction, F(3,615)=2.609, p= 051
approached significance. Figure 22 shows that child molesters ascribed more harm to the
woman than did controls across all levels of sexual contact, but particularly when the
sexual contact involved touching or fondling. Both groups scored similarly when the
sexual contact involved genital contact or ejaculation.
There were no significant differences among groups on the ratings of the womans
responsibility for the sexual interaction F(5,205)=l .492, p=. 194. There were also no
significant interactions with Group.
Analysis of Mans Role" variables. Although there were no significant differences
among groups on the doubly MANOVA using these variables, further analysis was
conducted in order to examine the difference found between admitters and deniers on
Factor 2. ANOVAs were conducted for the variables of Mans Responsibility, Mans
Anxiety, and Mans Confidence. Deniers attributed significantly less responsibility to the
man than did admitters, F(l,205)=18,652, p=.000. Deniers also differed from admitters
on the Group X Woman X Sex interaction, F(5,1025)=2.934, p= 012, Figure 23 shows
that the deniers consistently attributed less responsibility to the man across vignettes For


131
Mehrabian, A., & Epstein, N.A. (1972). A measure of emotional empathy. Journal of
Personality. 40, 523-543.
Merbaum, M. (1972). The simulation of normal and MMPI profiles by repressors and
sensitizers. Journal of Consulting and Clinical Psychology. 39. 171.
Miller, P. & Eisenberg, N. (1988). The relation of empathy to aggressive and
extemalizing/antisocial behavior. Psychological Bulletin. 103(3). 324-344.
Morrison, R.L., & Bellack, A S. (1981). The role of social perception in social skills. Behavior
Therapy. 12,69-79.
ODonohue, W., & Letoumeau, E. (1993). A brief group treatment for the modification of
denial in child sexual abusers: Outcome and follow-up. Child Abuse & Neglect. 17. 299-
304.
Overholser, J., & Beck, S. (1986). Multimethod assessment of rapists, child molesters, and
three control groups on behavioral and psychological measures. Journal of Consulting and
Clinical Psychology. 54153. 682-687
Pawlak, A.E., Boulet, J R, Bradford, J.M.W. (1991). Discriminant analysis of a sexual-
functioning inventory with intrafamilial and extrafamilial child molesters. Archives of
Sexual Behavior. 20(1), 27-34.
Putnam, F.W. (1985). Dissociation as a response to extreme trauma. In K.P Kluft (Ed ), The
childhood antecedents of multiple personality (pp. 66-97). Washington, DC: American
Psychiatric Press.
Quinsey, V.L. (1986). Men who have sex with children. In D.N. Weistub (Ed ), Law and
mental health: International perspectives (vol. 2; pp 140-172). New York: Pergamon
Press.
Riggio, R, Tucker, J., & CofFaro, D (1989). Social skills and empathy. Personality and
Individual Differences. 10(1), 93-99
Rogers, R, & Dickey, R (1991) Denial and minimization among sex offenders: A review of
competing models of deception. Annals of Sex Research. 4. 49-63.
Rook, K.S., & Hammer, C.L. (1977). A cognitive perspective on the experience of sexual
arousal. Journal of Social Issues. 33. 7-29.
Rosenthal, R, Hall, J.A., DiMatteo, M R, Rogers, P.L., & Archer, D. (1979). Sensitivity to
nonverbal communication: The PONS Test. Baltimore. Johns Hopkins University Press.


Mean Woman Benefit Ratings
Level of Sexual Contact
Figure 17. Mean Woman Benefit ratings for levels of sexual contact.


15
deal with frustration (p. 187)." Based on this finding, one might expect deniers to score
higher on empathy in the current study, in an effort to present themselves as emotionally
sensitive.
In a more recent study, Haywood and Grossman (1994) found that deniers
reported significantly less sexual interest in children than did admitters or normal control
subjects. Deniers also reported more sexual interest in women than did admitters.
However, both admitters and deniers reported significantly less sexual interest in adult
women than the controls. Among the child molesters, low self-reported sexual interest in
children was significantly correlated with minimization of psychopathology on the F-K
index and the positive malingering scale. The authors conclude that, for the sex offender
population, subjective measures of sexual interest are "highly susceptible to distortion or
minimization of deviant sexual arousal (p.338).
Abused Versus Nonabused Subjects
In addition to differences between admitters and deniers, Becker et al. (1992) also
found differences between abused and nonabused groups. Subjects who reported a history
of physical or sexual abuse were more sexually responsive in general than subjects with no
reported history Abused subjects also showed a greater proportion of nondiscriminator
profiles. These findings suggest the need to differentiate child molesters based on abuse
history as well as by denial.
Overall, there has been little research in the area of cognitive factors in child
molestation, with only one empirically sound study to date (Stermac & Segal, 1989). Given


98
Table 5
Pearson correlation matrix for the variables on the child vignettes.
1
2
3
4
5
6
7
1) child harm
1.000
2) child desire
-0.347
1.000
3) child respons.
-0.349
0.836
1.000
4) man respons.
0.603
-0.522
-0.530
1.000
5) treatment
0.655
-0.442
-0.442
0.770
1.000
6) other men
-0.174
0.436
0.435
-0.267
-0.269
1.000
7) child enjoy.
-0.330
0.897
0.754
-0.469
-0.400
0.401
1.000
8) punishment
0.419
-0.331
-0.316
0.464
0.688
-0.207
-0.389
9) child benefit
-0.288
0.619
0.613
-0.395
-0.352
0.439
0.603
8
1.000
-0.293
number of observations: 208
Table 6
Rotated loadings for two factors derived from principal components analysis on the child
vignettes
Components
Variables
1
2
Child Desire
.891
.270
Child Respon.
.859
.260
Child Enjoyment
.847
.281
Child Benefit
.760
.197
Other Men
.618
.089
Treatment
-.227
-.907
Child Harm
-.146
-.793
Mans Respon.
-.352
-.779
Punishment
-.171
-.743


94
C= Inmate Controls
CC= Community Controls
IA= Incestuous Admitters
ED= Incestuous Demers
NFA= Nonfamilial Admitters
NFD= Nonfamilial Deniers
Group
Figure 25. Mean group ratings on empathy total score.


105
Table 14
Group means on Woman Enjoyment on ejaculation vignettes.
Incest Admitters Inmate Controls
M
2.80
M
3.07a
SD
.11
SD
.07
N
25
N
60
Incest Deniers
Community Controls
M
2.70b
M
2.69b
SD
.10
SD
.10
N
33
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.76
M
2.95
SD
.10
SD
.10
N
31
N
32
Note. Means with different subscripts differ significantly at g<05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.001


nonfamilial offenders. Thus, one might expect incestuous offenders to show fewer
cognitive distortions than their nonfamilial counterparts regarding sexual contact with
children.
Admitters Versus Deniers
Another distinction among child molesters which is often overlooked in the
research is whether or not the alleged offender admits to having committed the crime.
Many studies have focused on those who admit to their crime, however, the majority of
sex offenders deny culpability (Salter, 1988; French, 1988; Rogers & Dickey, 1991). Thus,
the sample of offenders being studied is not representative of the majority of offenders. In
addition, studies which combine admitters and deniers may yield confusing results since
several recent studies have shown that these two groups differ in significant ways (Becker,
Kaplan, & Tenke, 1992; Haywood, Grossman, & Hardy, 1993; Haywood & Grossman,
1994).
Becker, Kaplan, and Tenke (1992) found significant differences in the erectile
response profiles of admitters and deniers. They found that a disproportionate number of
deniers failed to show a significant sexual response to cues depicting any age category
(nonresponders). While only 18% of the admitters were classified as nonresponders, 58%
of the deniers fell in this category Although the deniers tended to be less responsive in
general, they were maximally responsive to child cues. Admitters, in contrast, were more
responsive to peer cues, but also responded to child and adult cues. Forty-two percent of
the admitters showed peer or adult profiles while only 11% of the deniers were in these
categories. Deniers were significantly less responsive to adult cues than admitters, but


95
Table 1
Mean scores on repression and social desirability for self-reported abused and nonabused
subjects.
Repression-Sensitization
Social Desirability
Nonabused
M
39.88*
18,47**
SD
2.01
.52
N
127
126
Abused
M
47.00*
16.08**
SD
2.49
.63
N
83
86
*£=.027. **£=004.
Table 2
Mean scores on social desirability as a function of abuse history and admission or denial.
Admitters
Deniers
Nonabused
M
17.59
19.52*
SD
.82
.83
N
54
52
Abused
M
14.93*
18.06
SD
.92
1.01
N
42
35
Total
M
15.23**
19.70**
SD
82
.83
N
54
52
*£=.002. **£=.000.


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
SOCIAL INFORMATION PROCESSING IN CHILD MOLESTERS: AN EXAMINATION
OF DECODING SKILL AND COGNITIVE FACTORS
By
Julie C. Medlin
December 1995
Chairperson: Jacquelin Goldman
Major Department: Clinical and Health Psychology
This study examined social information processing in heterosexual child molesters.
More specifically, this study examined child molesters decoding skill and cognitions regarding
sexual contact with children. The literature has indicated that child molesters are socially
deficient; however, the nature of this deficiency is unclear. That is, this deficiency may result
from any number of problems in their cognitive operations, including deficiencies in decoding
skill. Thus, the present study examined child molesters' decoding skill, or ability to "read"
others' cues. Contrary to expectations, child molesters did not show a deficit in decoding skill
or social perception when compared with control groups. Consistent with expectations, child
molesters showed cognitive distortions regarding sexual contact with children. All four sex
offender groups (nonfamilial admitter, nonfamilial denier, incestuous admitter, incestuous
vi


115
childs role in the sexual contact. It is important to note, however, that many of these
differences might have resulted from the extreme pattern of the nonfamilial admitters, and
may therefore not hold for nonfamilial deniers
Compared to incestuous offenders, nonfamilial offenders also rated other men as
more likely to molest children, although this difference only approached significance. The
difference between the nonfamilial and incestuous offenders also approached significance
on the Abel & Becker Cognition Questionnaire, indicating that nonfamilial offenders
endorsed more cognitive distortions than incestuous offenders regarding sexual contact
with children. In addition, the interview data also indicate that the nonfamilial offenders
showed more cognitive distortions than the incestuous offenders. The incestuous
offenders, however, showed more cognitive distortions than the control groups, which
would be expected given clinical observations noted in the literature (Salter, 1988). Thus,
incestuous offenders seem to show some cognitive distortions regarding sexual contact
with children, although show fewer cognitive distortions than nonfamilial offenders This
finding offers support for the literature suggesting that nonfamilial and incestuous
offenders molest for different reasons If nonfamilial offenders tend to molest because
they prefer children, then one might expect them to show more severe cognitive
distortions rationalizing their lifestyle preference. On the other hand, if incestuous
offenders tend to be situational offenders, then one might expect them to show fewer
cognitive distortions since they would be more likely to show a sexual preference for
adults rather than children Thus, incestuous offenders may have less of a strong need to
rationalize their sexual deviancy since it is not based on their primary sexual interest,


50
composite variable as did Stermac & Segal (1989), these two variables were only two of
the nine variables that were all highly correlated with each other. Thus, specifying these
two particular variables for combination did not seem to be statistically warranted.
Instead, a principal components analysis was conducted in order to reduce the number of
variables in a more statistically appropriate manner.
Principal components analysis of the child vignettes. In addition to reduction of
the number of variables, principal component analyses with varimax rotation were
conducted in order to examine the relationships among the variables. A principal
components analysis was conducted on the nine variables collapsed across levels of sexual
contact and childs response. The analysis yielded two factors, accounting for 38.16%
and 31.96% of the total variance. The variables with high loadings on Factor 1 included
child desire, child responsibility, child enjoyment, child benefit, and other men (see Table
6). Thus, this factor appears to measure the childs reaction and complicity in the sexual
contact. In contrast. Factor 2 measures the mans responsibility for the sexual contact as
well as the seriousness of the behavior. The variables with high loadings on this factor
included the mans responsibility, the harm to child, the mans need for treatment, and the
amount of punishment prescribed for the man.
Although the analysis of the nine variables summed across vignettes yielded two
meaningful factors, it was important to conduct principal components analyses for each of
the twelve vignettes to see if these two factors were stable within each of the vignettes. As
expected, the same factor structure was found in each of the twelve vignettes, indicating
that these same variables tend to cluster together within each vignette.


40
Scale, the empathy total score on the Interpersonal Reactivity Index, and the four subscale
scores measuring perspective-taking, fantasy, empathic concern, and personal distress.
Procedures
Prison files were reviewed in order to select subjects who met the criteria for the study.
All records of inmates with a child sex offense were reviewed. In order to generate a random
list for the control group, every fifth name was selected from the prison roster. These files
were reviewed and those inmates with no history of a sex offense were assigned to the nonsex
offending control group. Subjects selected for both groups were sent a letter which described
the nature of the study and listed the date and time of an informational meeting they could
attend to learn more about the study. Approximately ten subjects were scheduled for each
initial group meeting.
At the informational meeting, subjects were provided with a complete description of
the project. Those who volunteered were asked to complete the Introductory Questionnaire,
the Social Desirability Scale, and the Interpersonal Reactivity Index. Subjects were then
administered the IPT, and asked to complete one of the two sets of vignettes based on random
assignment. At the end of the first session, subjects were scheduled to return for an individual
session and the second group session within the next seven days.
In the individual session, subjects were administered a semistructured interview and
the Vocabulary Scale from the WAIS-R. In the second group session, subjects were
administered the R-S scale and the LEQ, followed by the PONS test. Subjects were then
asked to complete the second set of vignettes.


107
Table 16
Group means on Womans Sexual Attraction across vignettes.
Incest Admitters Inmate Controls
M
2.89
M
3.23
SD
.10
SD
.07
N
25
N
60
Incest Deniers
M
2.91
Community Controls
M 2.80
SD
.09
SD
.09
N
33
N
30
Nonfamilial Admitters
M
2.97
Nonfamilial Deniers
M 2.94
SD
.09
SD
.09
N
31
N
32
* differs significantly from other five groups combined, p= 000.


42
socioeconomic status, F(5,208)= 576, p=.718 or vocabulary score F(5,208)=1.942,
p=.089. There was a significant difference, however, on age, F(5,208)=8.518, p= 000.
The mean ages of the groups were 29.86 for the inmate control group, 34.74 for the
community control group, 38.89 for the incestuous admitters, 38.32 for the incestuous
deniers, 40.45 for the nonfamilial admitters, and 39.72 for the nonfamilial deniers. A
Scheffe post-hoc analysis showed that the inmate control group was significantly younger
than each of the four child molester groups. There were no significant differences among
the other groups. A contrast between child molesters and controls revealed that the child
molesters were significantly older than controls, F(l,208)=27.421, p=.000. Other studies
have also found that child molesters were significantly older than controls (Haywood &
Grossman, 1994). This age difference is not surprising since the general inmate
population tends to be younger than the sex offender population, by virtue of their shorter
prison sentences in Florida and sex offenders older age at arrest.
It was expected that a greater proportion of child molesters than controls would
report a history of physical and/or sexual abuse. In order to examine reported abuse
history across groups, a chi-square analysis was performed. There was a significant
difference among the six groups, X2(6)=13.626, p=.018. The following percentage of
each group reported a history of abuse: Incestuous Admitters (61.54%), Nonfamilial
Admitters (54.84%), Incestuous Deniers (47.06%), Nonfamilial Deniers (31.25%),
Controls (30%), and Community Controls (29%) (see Figure 1). One-tailed planned
comparisons showed that the child molesters reported significantly more abuse than
controls as expected, Z=2.862, p=.002, and admitters reported significantly more abuse


46
score significantly higher than all of the other groups, which was frequently the case with
the nonfamilial admitters. The Scheffe test was used to conduct these post hoc
comparisons.
Analysis of Decoding Skill
Child molesters were predicted to score lower on decoding measures than control
subjects. Contrary to expectations, there were no significant differences between child
molesters and controls on the total score of the Profile of Nonverbal Sensitivity Test
(PONS), F(2,203)=.018, p=892. There were also no significant differences between
admitters and deniers, or between incestuous and nonfamilial sex offenders. An ANOVA
conducted with the six groups also showed no significant difference among the groups
F(5,203)=.737, p=.597 on PONS total score (see Table 3). An ANOVA with the four
PONS subscale scores as the repeated measures revealed no significant differences among
groups, F(5,203)=789, p=.559. As expected, subjects showed significant differences
across the four subscales, F(3,609)=122.449, p=.000. It was predicted that the decoding
scores would be significantly correlated with the empathy scores. Contrary to
expectations, the correlation of the PONS total score with the IRI total score was not
significant, r= 039.
Consistent with the results on the PONS, there were no significant differences
between child molesters and controls, admitters and deniers, or between incestuous and
nonfamilial sex offenders on the Interpersonal Perception Task (IPT). An ANOVA also
showed no significant differences among the six groups on the IPT score, F(5,207)=.203,
p= 961 (see Table 3) or on the subjects estimation of their IPT score F(5,184)=.549,


CHAPTER 1
INTRODUCTION AND REVIEW OF LITERATURE
Child molestation is a major social problem which has only recently begun to receive
adequate research attention over the past decade (Knight, Rosenberg, & Schneider, 1985;
Abel, Rouleau, & Cunningham-Rathner, 1985; Marshall, Laws, & Barbaree, 1990). Despite
these research efforts, few theories have emerged to explain this phenomenon. Two areas of
potential relevance to theory construction include child molesters' social perception and their
cognitions about sexual interaction with children Research in these areas suggests a deficiency
in child molester's social information processing, and highlights the need for a more focused
investigation of these variables and their potential relationship.
Social Skills of Child Molesters
There have been numerous studies focusing on the social competence of child
molesters (Segal & Stermac, 1989). The social deficit model of child molestation suggests that
child molesters lack social skills and, as a result, have unsatisfactory heterosocial relationships
(Groth & Bimbaum,1979). The clinical literature has frequently characterized child molesters
as having impaired social relationships and discomfort with adult sexuality (Hobson, Boland,
Jamieson, 1985). More specifically, Finkelhor (1984, 1986) proposed a four-factor model of
child molestation postulating that child molesters turn to children partly because alternative
sources of sexual gratification have been blocked or inhibited.
1


Child Benefit Ratings
74
C CC IA ID NFA NFD
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Deniers
Nonfamilial Admitters
Nonfamilial Deniers
Group
Figure 5. Mean group ratings on Benefit to Child.


Sexual Attraction Ratings
90
Group
C=
cc=
IA=
ID=
NFA=
NFD=
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Deniers
Nonfamilial Admitters
Nonfamilial Deniers
Figure 21. Mean group ratings on Sexual Attraction.


Mean Punishment Ratings
83
12 3 4
C Controls
CM Child Molesters
Level of Sexual Contact
Figure 14. Mean punishment ratings for levels of sexual contact.


38
body?" A supplement to the questionnaire asked subjects for specifics concerning the
frequency and period of time during which the abuse occurred. In order to facilitate
comparisons with the Becker (1992) study, the variable of abuse encompassed both physical
and sexual abuse. Thus, within this study, the term a history of abuse refers to a self-
reported history of physical and/or sexual abuse.
Measure of Social Status
The Hollingshead Four Factor Index of Social Status (Hollingshead, 1975) was used
to estimate subjects' socioeconomic status. This index assesses social status based on
education, occupation, sex, and marital status. The occupation factor is based on the
approximately 450 occupational titles and codes of the 1970 United States census, and is
graded on a 9-point scale. The education factor is based on the number of years of schooling
and is scored on a 7-point scale. Scores from the occupation, marital status, and education
factors are combined to yield a social status score ranging from 8 to 66.
The Four Factor Index was found to be reliable and valid based on data derived from
the 1970 Census and the National Opinion Research Council. Social strata categories were
validated using factor analysis and criterion validity was established by comparing occupation
index scores with prestige scores developed by the National Opinion Research Center (r=.93).
Significant differences were found among these categories when mass communication data
were used as criteria of social behavior (Hollingshead, 1975). The Four Factor Index also
correlates highly with other measures of socioeconomic status: the Revised Duncan
Socioeconomic Index (r=.79) and the Siegel 1965 NORC Prestige Scale (r=.73).
Measure of Intelligence


130
Hobson, W., Boland, C., & Jamieson, D. (1985). Dangerous sexual offenders. Medical
Aspects of Human Sexuality. 19(21. 104-119.
Hollin, C. & Howells, K. (1991). Clinical Approaches to Sex Offenders and Their Victims.
John Wiley & Sons: Chichester, England.
Hollingshead, A.B. (1975). Four factor index of social status. Unpublished manuscript, Yale
University, New Haven.
Horley, J. (1988). Cognitions of child sexual abusers. The Journal of Sex Research. 25 (41,
542-545.
Howells, K. (1978). Some meanings of children for pedophiles. In M. Cook & G. Wilson
(Eds ), Love and Attraction (pp.57-82). London: Pergamon Press.
Howells, K. (1981). Adult sexual interest in children: Considerations relevant to theories of
aetiology. In MCook & K. Howells (Eds), Adult Sexual Interest in Children (pp.55-94).
New York: Academic Press.
Knight, R_, Rosenberg, R, & Schneider, B. (1985). Classification of sex offenders:
Perspectives, methods, and validation. In A.W. Burgess (Ed ), Rape and sexual assault: a
research handbook (pp. 223-293). New York: Garland.
Lipton, D.N., McDonel, E.C., & McFall, RM. (1987). Heterosocial perception in rapists.
Journal of Consulting and Clinical Psychology. 55 (11.17-21.
Lorr, M., Youniss, R, & Stefic, E. (1991). An inventory of social skills. Journal of
Personality Assessment 57(31, 506-520.
Marshall, W.L., Barbaree, H.E., Christophe, D. (1986). Sexual offenders against female
children: Sexual preferences for age of victims and type of behaviour. Canadian Journal of
Behavioral Science. 18(41.424-437.
Marshall, W.L., Jones, R, Hudson, S.M., & McDonald, E. (1993). Generalized empathy in
child molesters. Journal of Child Sexual Abuse. 2(41. 61-68.
Marshall, W.L., Laws, D.R, & Barbaree, HE. (Eds.) (1990). The Handbook of Sexual
Assault: Issues. Theories, and Treatment of the Offender. New York: Plenum Press.
Mayer, A. (1988). Sex offenders: Approaches to understanding and management. Holmes
Beach, FL: Learning Publications, Inc.
McFall, RM. (1990). The enhancement of social skills: An information processing analysis. In
Marshall, W.L., Laws, D.R, & Barbaree, HE. (Eds.). The Handbook of Sexual Assault:
Issues. Theories, and Treatment of the Offender (pp. 1-30). New York: Plenum Press.


96
Table 3
Group means and standard deviations for dependent measures.
Group Empathy Repression Cog. Que. IPT PONS Vocab
Nonfamilial
Admitters
M
SD
62.68
2.25
45.14
4.23
120.86*
3.32
13.71
58
153.73
3.74
34.75
2.90
Nonfamilial
Deniers
M
SD
64.60*
2.17
44.37
4.08
122.23
3.20
13.73
.56
152.83
3.61
28.57
2.80
Incest
Admitters
M
SD
62.37
2.54
47.77
4.77
129.96
3.74
14.55
66
159.50
4.21
38.05
3.28
Incest
Deniers
M
SD
55.00
2.17
38.33
4.08
126.17
3.20
14.20
.56
149.37
3.61
33.20
2.80
Inmate
Controls
M
SD
58.95
1.59
41.59
2.99
127.77
2.34
14.30
.41
152.33
2.64
29.39
2.05
Community
Controls
M
SD
58.53
2.17
37.47
4.08
134.90*
3.20
14.60
.56
153.38
3.61
36.50
2.80
'p-026
p=.015


4
Sexual assaulters in general have been hypothesized to have difficulty processing social
information from women as well as show a tendency to misinterpret social messages. Lipton,
McDonel, and McFall (1987) investigated this hypothesis, comparing rapists, violent
nonrapists, and nonviolent nonrapists on assessments of heterosocial cue-reading accuracy in
intimate and first-date situations. Rapists were found to miscontrue women's cues, particularly
cues involving negative moods in first-date situations, suggesting a deficit in their social
information processing. There appears to have been no such study conducted with child
molesters examining their ability to read others' cues. In fact, there have been few studies of
cognitive operations in sexual offenders in general. Cognitive operations refers to the various
ways in which the information processing system operates, including attention allocation,
encoding, decoding, or control processes.
Such studies are needed since child molesters' social skill deficiencies and cognitions
may be related to a difficulty in reading others' cues. It has been noted clinically that child
molesters misidentify emotions (Hobson, Boland, & Jamieson, 1985). Another reason to
expect that child molesters might be generally deficient in this skill is that low social
competence (Feldman & Rime, 1991) and an abuse history (Camras, Grow, & Ribordy, 1983)
are often associated with deficiencies in decoding. The latter variable is important to account
for since many child molesters (estimates range from 25% to 75%) report a history of abuse
(Mayer, 1988; Hamilton & Ondrovik, 1993) and those who report abuse may comprise a
distinct clinical subgroup (Becker, Kaplan, & Tenke, 1992).
In addition, child molesters are often characterized as lacking empathy (Hobson,
Boland, & Jamieson, 1985), and ability to read others' cues is believed to be a key component


BIOGRAPHICAL SKETCH
Julie Christine Medlin received her Bachelors of Arts degree from Harvard College in
1989. She worked for two years as a research assistant at the Center for Drug Treatment and
Research in Washington, D C She then enrolled in the doctoral program in the Department of
Clinical and Health Psychology at the University of Florida. She received her Master of
Science degree in 1993 and completed her clinical psychology internship in the Department of
Psychiatry, University of North Carolina at Chapel Hill, in 1995. Her PhD in clinical
psychology is expected to be granted in August, 1995. Ms. Medlin plans to pursue a career in
forensic psychology involving the treatment and evaluation of a sex offender population. She
also hopes to pursue a teaching and research career
133


Mean Child Desire Ratings
78
Child Response
Nonfamilial Offenders
Incestuous Offenders
Figure 9. Mean Child Desire ratings for levels of child response.


85
3.0
C CC IA ID NFA NFD
Group
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers
Figure 16. Mean group ratings on Benefit to Woman


7
community-based child molesters on their actual and ideal views of themselves. The measure
consisted of pairs of bipolar variables which reflect aspects of an individual's personality traits,
physical characteristics, and value orientations. Subjects located themselves on these
dimensions for their actual versus ideal selves. The largest discrepancies between the two
ratings were found on adjectives related to interactions with others. Although the results
suggest that child molesters feel less socially competent than they desire, comparisons with
nonoffender controls cannot be made since this group was not included in the study.
A later study on cognitive factors (Howells, 1978) investigated heterosexual child
molesters' perceptions of adult females, adult males, and children using the repertory grid
technique. The results indicated that child molesters employed more constructs related to
dominance in their descriptions of both children and adults. Children were perceived as being
at the nondominant pole, as "undemanding" and "passive" while adults were viewed as more
"domineering" and "demanding." The results also suggested that child molesters may be
attracted to children for more than sexual reasons, such as for their innocence and capacity for
unconditional love. Unfortunately, these results are only suggestive given the small sample size
of ten child molesters, problems in the statistical analysis, and the failure of a follow-up study
(Horley, 1988) to replicate these findings.
A more recent study (Abel et al., 1989) has focused on child molesters' cognitions
concerning child molestation. This study used a 29-item Likert scale and found 6 reliable
factors covering the general area of harm caused to the child by the molestation. The factors
successfully separated the child molesters from the control group, suggesting that child
molesters show cognitive distortions minimizing the harm of the molestation.


100
Table 8
Group means on Child Responsibility when the childs response was positive.
Incest Admitters
Inmate Controls
M
1 69b
M
1.91
SB
.19
SD
.11
N
23
N
61
Incest Deniers
Community Controls
M
1.84
M
1.54b
SD
.16
SD
.16
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.44a
M
1.94
SD
.16
SD
.16
N
31
N
32
Note. Means with different subscripts differ significantly at [><05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=,000.


34
assesses the subject's tendency to "transpose themselves imaginatively into the feelings and
actions of fictitious characters in books, movies, and plays (Davis, p. 114)." The Empathic
Concern (EC) scale measures feelings of sympathy and concern for others, and the Personal
Distress (PD) scale measures the subject's feelings of personal anxiety in tense interpersonal
situations.
The IRI was developed from a pool of 50 items some of which were adapted from
other measures of empathy such as the Mehrabian & Epstein emotional empathy scale (1972).
New items were added to measure cognitive aspects of empathy such as the ability to adopt
different perspectives Factor analysis revealed four factors which now form the four
subscales. This factor structure has been shown to be stable when repeatedly administered to
different samples (Davis, 1983). The final 28-item version of the IRI was standardized on 579
male and 582 female college students. Internal reliabilities range from .71 to .77 and test-retest
reliabilities range from .62 to .71 (Davis, 1980).
Davis (1983) cites data supporting the validity of the four IRI subscales and his
multidimensional view of empathy The four subscales which represent different aspects of
empathy showed distinctive patterns of relationships with measures of social functioning, self
esteem, emotionality, sensitivity to others, and unidimensional measures of empathy The
Hogan Empathy Scale which is more cognitive in its approach showed highest correlations
(r=.40) with the cognitive PT scale, as expected The PT scale also showed the least
association (r=.20) of the four IRI scales with the Mehrabian and Epstein Emotional Empathy
Scale which taps emotional rather than cognitive empathy. In addition, the FS and EC scales
showed greater associations with the Emotional and Empathy Scale (r=.52 and .60,


52
Multivariate analysis of variance. The principal components analysis reduced the
variables into two sets one relating to the Childs Complicity and another relating to the
Seriousness of the Offense. A doubly multivariate analysis of variance was conducted on
each of the two sets of variables. The Child Complicity variable set consisted of Child
Benefit, Child Enjoyment, Child Desire, Child Responsibility, and Other Men. The
variable set relating to the Seriousness of the Offense consisted of Child Harm, Mans
Responsibility, Mans Need for Treatment, and Punishment. Group served as the between
subjects factor with Childs Response (Child) and Level of Sexual Contact (Sex) as the
within subjects factors. The dependent measures were the variables in each variable set.
Social desirability was used as a covariate in the initial analysis, however, it was not
significant in the analysis of Child Complicity, F(5,194)=.422, p= 833, or Seriousness of
the Offense, F(4,195)=.539, p= 707, using the Wilks criterion, so it was dropped from
further analysis.
The doubly multivariate analysis for Child Complicity yielded a significant main
effect for Group, F(25, 733,32)=1.54, p=.046, using Wilks criterion. There was also a
significant Group X Child interaction, F(50, 1818.52)= 1.47, p=.018, using the averaged
multivariate tests of significance. Planned comparisons were conducted between child
molesters and controls, incestuous and nonfamilial offenders, and admitters and deniers.
Incestuous offenders scored significantly different from nonfamilial offenders on Child
Complicity, F(5,197)=2.921, p=.014, using the Wilks criterion. There was no significant
difference between child molesters and controls, F(5,197)=1.410, p=.222, or between
admitters and deniers, F(5,197)=.8372, p=.528.


s
In a similar study, Stermac and Segal (1989) tested the hypothesis that child molesters'
cognitions are more permissive and accepting of sexual contact with children. They used
written vignettes describing sexual contact between an adult and child. The vignettes contained
varying degrees of sexual contact (touching fondling no clothes, ejaculation) and varying
responses from the child (smiling passive, crying). After reading each vignette, child molesters
and controls responded to questions assessing their view of the adult's and child's behavior.
The results indicate that child molesters ascribed more benefits of the molestation to
the child than did the controls. Unlike the controls, the child molesters did not significantly
decrease their amount of ascribed benefit to the child as the amount of sexual contact
increased. The child molesters also differed significantly from the controls in their perceptions
of benefit based on the child's response, although their response was similar to that of the
controls when the child's response was negative. Compared with the controls, the child
molesters attributed more responsibility to the child and less overall responsibility to the adult.
In general, the child molesters' ratings were less affected by the amount of sexual contact or the
child's response.
The child molester group also showed significant differences from controls on the
Cognition Questionnaire (Abel, Becker, & Cunningham-Rathner, 1984), endorsing more items
indicative of permissive beliefs and attitudes towards adult sexual contact with children. These
results are consistent with clinical reports of child molesters' distorted cognitions regarding
sexual contact with children (Abel, Becker, & Cunningham-Rathner, 1984; Abel, Mittleman,
& Becker, 1985; Finkelhor, 1984).


106
Table 15
Group means on Womans Desire across vignettes.
Incest Admitters Inmate Controls
M
2.72
M
2.92
SD
.09
SD
.06
N
25
N
60
Incest Deniers
M
2.72
Community Controls
M 2.64
SD
.08
SD
.08
N
33
N
30
Nonfamilial Admitters
M
2.88
Nonfamilial Deniers
M 2.82
SD
.08
SD
.08
N
31
N
32
Note. Means with different subscripts differ significantly at ¡><.05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.001


73
140
C CC IA ID NFA NFD
Group
Figure 4. Mean scores across groups on the Cognition Questionnaire.
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers


denier) prescribed significantly less punishment and treatment for a child molester than did the
control groups. Most of the other cognitive distortions, however, were found among
nonfamilial admitters rather than among the other sex offender groups. When reading vignettes
portraying sexual interaction with a female child, nonfamilial admitters attributed significantly
more benefit, responsibility, desire, and enjoyment to the child than the other sex offender
groups and control groups. As expected, incestuous offenders differed significantly from
nonfamilial offenders in that they scored more similarly to controls. In addition, sex
offenders who denied having committed their sexual crimes generally endorsed fewer
cognitive distortions than admitters and showed more social desirability than admitters.
Contrary to expectations, child molesters did not differ from controls in their ratings of anxiety
in a sexual situation with an adult female.
Other variables examined included empathy, repression, and abuse history. Contrary
to predictions, child molesters did not score significantly lower on empathy than controls.
There was also the unexpected finding that nonfamilial offenders scored higher than
incestuous offenders on empathy. There were no differences among groups on repression.
Consistent with expectations, more child molesters than controls reported having been
physically or sexually abused as a child, as well as more admitters than deniers. In sum,
significant differences were found between incestuous and nonfamilial offenders, as well as
vii
between admitters and deniers.


SOCIAL INFORMATION PROCESSING IN CHILD MOLESTERS: AN
EXAMINATION OF DECODING SKILL AND COGNITIVE FACTORS
By
JULIE C. MEDLIN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1995


Woman Desire Ratings
88
C CC IA ID NFA NFD
Group
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers
Figure 19. Mean group ratings on Woman Desire.


21
begins with an examination of decoding skill (McFall, 1990), the present study examined this
variable among child molesters.
Given child molesters' low social competence and the recent finding of a decoding
deficiency found among other sex offenders (Lipton, McDonel, & McFall, 1987), it was
expected that child molesters would also show a deficiency in decoding skill when compared to
non-sex-offending control groups. In addition to differences between child molesters and
controls, differences between admitters and deniers were examined since a preliminary finding
suggested that deniers may represent a distinct clinical subgroup (Becker et al., 1992). In
addition to examining nonverbal decoding, this study examined child molesters' performance on
a more complex social perception task which involves the next step in McFall's (1990) social
information processing model decision skills. Child molesters were expected to perform as
poorly on this measure as on the decoding task since the social perception task involves
decoding as well as decision skills If the child molesters did not show a deficiency in decoding
skills, then this social perception task would allow for an examination of child molesters at the
next level of social information processing decision skills.
Since child molesters were expected to show deficits in decoding skill, they were also
expected to show deficits in empathy since empathy requires good decoding skills. This notion
of a deficit in empathy is consistent with clinical observation of child molesters, and has
received some empirical support, although the research in this area has yielded Inconsistent
findings. Thus, another reason for examining empathy among this group stems from the
inconsistent findings in the literature. Unlike previous studies, the current investigation used a
more appropriate measure of empathy which assesses a multifaceted notion of empathy, and in


44
F(l,179)=5.266, p=,023, and a significant main effect of Admit, F(l,179)=7.905, p=.005,
with a nonsignificant Abuse X Admit interaction, F(l,179)=.447, p=.505 (see Table 2).
Figure 3 shows that subjects reporting no history of abuse scored significantly higher on
social desirability than those reporting abuse In addition, inmates who denied having
committed their crimes scored significantly higher on social desirability than those who
admitted to committing their crimes. Thus, deniers who did not report abuse showed the
highest social desirability while admitters who reported a history of abuse showed the
lowest social desirability.
No significant differences were expected between child molesters and controls on
social desirability. One outlier was identified by the studentized residual value and was
dropped from the analysis. As expected, a planned comparison revealed no significant
differences in social desirability between child molesters and controls, F(l,205)=.337,
p=.562. An ANOVA with the six groups, however, yielded a significant main effect for
Group F(5,205)=4.258, p=.001. As expected, deniers scored significantly higher than
admitters on social desirability, F(l,205)=16.769, p=.000 (see Table 2). There were no
significant differences between incestuous and nonfamilial sex offenders, F( 1,205)= 2.899,
p=.090. All possible pairwise comparisons using the Bonferroni correction showed that
incestuous deniers and nonfamilial deniers scored significantly higher than incestuous
admitters on social desirability, p=.003, p=.002, respectively.
Method of Analysis
The dependent variables were analyzed using ANCOVAs or MANCOVAs.
Decoding skill, empathy, repression, and cognitive distortions on the Cognition


36
function of a socially desirable response style (Byrne, 1961). There is some evidence to
suggest some independence between repression-sensitization and social desirability (Byrne,
1961). In a study by Merbaum (1972), extreme sensitizers and repressers were asked to retake
the test three months after having initially completed the R-S scale. In this second
administration they were asked to respond in a way that would make them look as good as
possible. The scores of both groups shifted in the repressor direction, however, the sensitizers
still scored higher in the direction of sensitization than did repressers Thus, differences
between on repression-sensitization remained despite the instructions to adopt a socially
desirable response style
Measure of Social Desirability
The Marlowe-Crowne Social Desirability Scale (MCSDS) (Crowne & Marlowe, 1960)
was used to control for socially desirable response sets. The MCSDS is a 33-item scale which
was originally developed to discriminate the effects of item content from the subject's need to
present him or herself in a socially desirable light. As a result, the MCSDS focuses on
culturally sanctioned and approved behaviors that do not occur frequently. Sample items from
this questionnaire include: 1) I never hesitate to go out of my way to help someone in trouble,
and 2) I am always careful about my manner of dress. The internal consistency coefficient is
reported at .88, and the test-retest reliability is .89 for a sample of normal 39 college students
Normative data is presented for 28 males and 27 females (Crowne & Marlowe, 1960). The
MCSDS is highly correlated with the Edwards Social Desirability Scale ( 35) and the following
MMPI subscales: L (.54), K (.40), and F (.36). The MCSDS has been used in numerous
studies, including the Stermac and Segal (1989) study.


63
Groups differed significantly on the Womans Complicity, F(30,802)= 1.634,
p=.018, using Wilks criterion. There were no significant interactions with Group.
Planned comparisons revealed significant differences between incestuous and nonfamilial
offenders, F(6,200)=2.186, p=046. There were no significant differences between child
molesters and controls, or between admitters and deniers. There were no significant
differences on the planned comparisons for the Group X Woman interaction. The
difference between nonfamilial and incestuous offenders approached significance on the
Group X Sex interaction, F(18,1725.83)=1.555, p=.064, using the averaged multivariate
tests. Child molesters and controls differed significantly on the Group X Woman X Sex
interaction, F(36,170)=1.919, p=.003.
There was no significant main effect of Group on the Mans Role,
F(15,560.80)=1.363, p=. 160. There were also no significant interactions with Group.
Planned comparisons also showed no significant differences on the main effect of Group
or interactions with Group. These results do not support the hypothesis that child
molesters would rate more anxiety and less confidence to a man in a sexual interaction
with a woman.
Analysis of Womans Complicity variables. In order to further examine the
significant results from the doubly MANOVA, univariate analyses of variance were
conducted for each variable comprising Womans Complicity. Group served as the
between subjects factor with Womans Response and Level of Sexual Contact as the
within subjects factors. For each of these variables there were significant main effects on
Womans Response and Level of Sexual Contact, indicating that subjects ratings differed


27
explain why you disagree with the charge(s). Their admission or denial was also assessed
during a semi-structured interview in which they were asked to describe their charge and
whether or not they committed the crime.
Child molesters were also subdivided into incestuous and nonfamilial offender groups
based on information in their prison file describing their relationship to the victim. Since this
information was not always available in the prison records, this information was also gathered
during an individual interview in which the inmate was asked to describe the nature of his
relationship to the victim. Sex offenders who molested their daughters or stepdaughters were
considered to be incestuous offenders, as well as those molesting any female child living in their
home under their custody or authority. Thus, if the molester was involved in some kind of
parental relationship with the victim, as in the case of a live-in boyfriend who molests his
girlfriends children, then he was considered to be an incestuous or familial offender. This
broad definition of incest was used since it is the dynamics of the relationship that are
considered to be most important, and since most studies on incest have used this broader
definition (Marshall et al., 1986).
Research Design
In order to test the hypotheses, the six subject groups were administered a battery
of tests measuring the following variables: cognitive distortions regarding sexual contact
with children, decoding ability, empathy, repression, and social desirability. Background
variables included age, socioeconomic status, intelligence, and history of abuse. Group
served as the main independent variable with six levels (nonfamilial admitters, nonfamilial
deniers, incestuous admitters, incestuous deniers, community controls, and inmate


124
history of abuse scored lower on social desirability than those denying a history of abuse.
Again, this is not surprising since admitting to a history of abuse on a questionnaire
involves some suspension of social desirability concerns. Admission or denial of the crime
is also related to social desirability in that those who admit to their crime and report
histories of abuse score significantly lower on social desirability than those who deny their
crime and deny histories of abuse Thus, inmates who deny committing their crime and
deny having a history of abuse will tend to deny faults in general. This highlights, once
again, the importance of denial in subjects self-reporting.
Contrary to expectations, abused subjects did not score lower than nonabused
subjects on empathy, suggesting that a history of abuse may not impair ones capacity for
empathy as Miller and Eisenberg (1988) found in their meta-analysis. A more likely
interpretation, however, is that this different result may be attributed to the use of a
predominantly incarcerated sample in contrast to the nonincarcerated samples typically
used in the empathy literature. Thus, it may be that incarcerated samples tend to rate
themselves as higher on empathy than nonincarcerated samples in general in an effort to
present themselves in a positive light and maintain that they are good people although
they are in prison. This seemed to be the case with the child molesters who scored higher
on empathic concern. It may also be that many subjects underreported their histories of
abuse as a result of their tendency to deny faults, thus resulting in some abused subjects
being misclassified as nonabused subjects. In any case, this finding merits further
investigation in future research since the reasons for the different results remain unclear. A


126
nonfamilial and incestuous offenders differed significantly in their degree of cognitive
distortions. These results also offer support for the notion that incestuous molesters tend
to be situational offenders while nonfamilial molesters tend to be sexual-preference
offenders.
Future research might focus on how to incorporate these differences into treatment
programs for incestuous versus nonfamilial offenders. For example, treatment for
nonfamilial offenders might focus on their self-centered distortions and lack of sensitivity
to their victims while treatment for incestuous offenders might focus on the more subtle
distortions they used to allow themselves to offend Future research on empathy might
also focus on child molesters specific empathy for their victims, since it is possible that
they might possess only a specific deficiency in empathy. Clearly, more
research is needed in order to fully understand the role of empathy and cognitive
distortions in sex offending behavior so this knowledge can be used to enhance treatment
with a highly deviant and dangerous clinical population.


37
Self-Report of Abuse History
The Life Experiences Questionnaire (LEQ) developed by Bryer et a). (1987) was used
to assess abuse history. Bryer et al. (1987) administered the LEQ to 66 females upon
admission to a private psychiatric hospital. They found that the more severely disturbed
patients were more likely to report having been abused in childhood. A recent study showed
that female inpatients were more likely to report a history of abuse on the LEQ than during an
intake interview (Dill, D., Chu, J., Grob, M., & Eisen, S., 1991). There was low agreement
between the LEQ and intake format in regard to reported histories of physical abuse, k=.32.
The authors note that the low agreement was largely due to subjects who reported abuse on
the LEQ but not during the intake interview. 59% of subjects reported physical abuse on the
LEQ while only 26% reported this type of abuse in the interview Ninety-two percent of
subjects reporting abuse during the interview also reported abuse on the LEQ, however, 47%
of those who did not report abuse during the interview later reported abuse on the LEQ.
Agreement between interview and LEQ reports of sexual abuse were also fairly low, k=.44
52% of subjects reported a history of sexual abuse on the LEQ while only 35% reported this
history in the intake interview These results strongly support the need to assess abuse history
in a written format since many subjects may be reluctant to report abuse history during an
interview.
In the present study, physical abuse was defined as an affirmative response to the LEQ
question "were you ever hit really hard, kicked, punched, stabbed, or thrown down?
Similarly, sexual abuse was defined as an affirmative response to the question "were you ever
pressured against your will into forced contact with the sexual parts of your body or [his/her]


I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
acqueiin Goldman, Chair
Professor of Clinical and Health Psychology
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
'h'i fo
Michael Perri
Professor of Clinical and Health Psychology
1 certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
4^$ \aS\J2~-
James Johnson / \
Professor of Clinical-and Health Psychology
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
jbA, .
James Algina
Piofessor of Foundatii
;of Education


54
Child Complicity variables and the Group X Sex interaction examined for the Seriousness
of the Offense variables
Analysis of Child Complicity variables. As expected, there was a significant
difference between groups on the variable Benefit to Child, F(5,202)=3.380, p=.006 (see
Figure 5). Contrary to expectations, there were no significant differences between child
molesters and controls in their overall benefit ratings, however, there was a significant
finding regarding the nonfamilial admitters. A post-hoc comparison showed that the
nonfamilial admitters ascribed significantly more benefit to the child than the other five
groups, F(l,202)=14.974, using the Scheffe test. Nonfamilial sex offenders attributed
more benefit to the child than did incestuous offenders, F(l,202)=9.214, p=.003, as
predicted. Contrary to expectations, there were no significant differences between
admitters and deniers, F(l,202)=1.292, p=257.
There was a significant difference between child molesters and controls on the
Group X Childs Response interaction, F(2,404)=4.070, p=.018. There were no
significant differences between admitters and deniers or nonfamilial and incestuous
offenders on this interaction.
Groups differed significantly on vignettes in which the child smiled,
F(5,202)=4.106, p=.001, and gave a passive response, F(5,202)=4.403, p=.001. There
were no significant differences among groups, however, on the vignettes in which the child
cried, F(5,202)=1.422, p=.218. This finding is consistent with Stermac and Segals (1989)
results. Planned comparisons showed that the child molesters did not significantly differ
from controls at any level of the childs response. A post-hoc comparison, however,


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ESFOB0TT2_X824OX INGEST_TIME 2015-03-27T19:12:54Z PACKAGE AA00029758_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


I
C CC IA ID NFA NFD
Vignettes:
* ADULT
D CHILD
Group
Figure 24. Mean group ratings on Child Enjoyment and Woman Enjoyment.


ACKNOWLEDGEMENTS
I would like to thank Jacquelin Goldman, Ph D., ABPP, my dissertation chair, for
supporting me in pursuing my interest in research with a particularly difficult forensic
population. She encouraged me to follow my true research interests rather than design a dull,
but easy, research study. I also thank her for helping me make some difficult design decisions
during the data collection process, and then later in the statistical analysis process. Overall, she
allowed me to develop into an independent researcher, as she gave me the freedom to be
creative and encouraged me to take risks. As a result, I have truly enjoyed and have learned
from my research pursuits in graduate school.
I would also like to thank my dissertation committee: Michael Peni, Ph D., James
Johnson, Ph D., James Algina, Ph D., and Otto von Mering. I offer my thanks for their
patience during my proposal meeting when I explained entirely too many hypotheses and a
complex design. I appreciate their willingness to allow me to venture into such a difficult
project potentially fraught with many practical problems, given the subject population and
prison environment
The Florida Department of Corrections also deserves enormous gratitude for allowing
me to conduct research in their prison system. In particular, I would like to thank William
Bales, PhD., Chief of the Bureau of Planning, Research, and Statistics for the Florida
Department of Corrections. He granted me permission to conduct my research at various
ii


CHAPTER 4
DISCUSSION
The results confirm some of the hypotheses and disconfirm some of the other
hypotheses. The first hypothesis regarding decoding skill was not supported by the data.
Contrary to expectations, there were no significant differences among groups on decoding
skill or social perception. Thus, these results suggest that child molesters sexual
offending does not appear to be related to a deficiency in perceiving others cues
accurately. Although several of the IPT scenes involved childrens cues, it was not
possible to score decoding ability based on childrens cues alone. Thus, it is possible that
child molesters may show a deficit in reading these particular cues. It may be also be that
these measures do not truly tap decoding skill or social perception, and that a more
sophisticated measure might be more likely to reveal a deficiency.
The second hypothesis proposed that child molesters would show more distorted
cognitions than controls regarding sexual contact with a female child. This hypothesis was
confirmed, as child molesters showed more cognitive distortions than controls on the Abel
and Becker Cognition Questionnaire and on the child vignettes. Thus, child molesters
endorsed more deviant beliefs regarding sexual contact with children. On the child
vignettes, child molesters ascribed more benefit to the child than did controls, and also
ascribed more enjoyment to the child when she was smiling. Child molesters also tended
to attribute more responsibility than controls to the child when she was smiling, although
110


117
the mans responsibility for the sexual behavior, although nonfamilial offenders also
attribute responsibility to the child.
Another finding which contrasts with the results of Stermac & Segal (1989) is the
lack of a difference between child molesters and controls in their ratings of the amount of
harm inflicted on the child Stermac & Segal (1989) combined the variable child harm
with child enjoyment and child benefit to form a composite variable. They found that child
molesters scored higher on this composite variable than did controls, indicating that child
molesters attributed more benefit and less harm to the child than did controls. The results
of the present study on the variable of child harm are not directly comparable to Stermac
& Segals (1989) composite variable since results on child harm cannot be extracted from
this composite variable Differences were found between the nonfamilial admitters and
other groups on the variable of benefit to the child, suggesting that the difference on the
composite variable might have been largely attributed to the individual variable of benefit
rather than to child harm. In any case, the results of the present study suggest that child
molesters have some awareness of the harm done to the child or, at least, possess an
awareness of the socially appropriate response.
The third hypothesis suggested that deniers would show fewer cognitive
distortions than admitters. The results provide minimal support for this hypothesis, as
most of the findings only approached significance and appeared to have been largely
affected by the extreme pattern of the nonfamilial admitters. Like the nonfamilial
admitters, admitters as a group tended to rate the child as more desirous of the sexual
contact than did deniers when the child smiled or was passive, but rated them similarly to


64
significantly as the level of sexual contact and the womans response to the sexual contact
changed. There was a significant interaction between these two variables on every
variable except for Womans Harm.
On the adult vignettes, groups were expected to differ only on the variables of
Mans Anxiety and Mans Confidence. Contrary to expectations, groups differed
significantly in their ratings of the amount of benefit the woman received from the sexual
contact, F(5,205)=2.818, p=.017 (see Figure 16 & Table 13). There was the unexpected
finding that nonfamilial sex offenders rated significantly more benefit to the woman than
did incestuous offenders, F(l,205)=7.569, p=.006, and showed a different pattern in their
ratings across the levels of sexual contact, F(3,615)=3.259, p=.021. Figure 17 shows that
the incestuous offenders decreased their ratings of the womans benefit as the degree of
sexual contact increased, while the nonfamilial offenders did not show this decrease in
their ratings. In contrast, the nonfamilial offenders rated the womans benefit as increasing
as the sexual contact progressed from touching only to fondling and ejaculation.
Groups also differed significantly in their ratings of the degree to which the woman
enjoyed the sexual contact, F(5,205)=2.509, p=.031. There was also a significant Group
X Sex interaction, F(15,615)=l .831, p=.027. There were no significant differences among
the groups on the touching or fondling vignettes, however, there were significant
group differences on the no clothes F(5,205)=2.595, p=.027, and ejaculation
vignettes F(5,205)=3.140, p=.009.
Nonfamilial offenders differed from incestuous offenders on the Group X Sex
interaction, F(3,615)=3.976, p=.008, and on the Group X Woman X Sex interaction,


103
Table 11
Group means on Other Men across vignettes.
Incest Admitters Inmate Controls
M
1.86
M
1.98
SD
.18
SD
.11
N
23
N
61
Incest Deniers
Community Controls
M
1.79
M
1.63b
SD
.16
SD
.16
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.42a
M
1 84
SD
.16
SD
.16
N
31
N
32
Note. Means with different subscripts differ significantly at p< 05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.


23
show no differences on these same variables relating to sexual contact with an adult. That is,
child molesters cognitive distortions relating to sexual contact were hypothesized to be largely
limited to interactions with children, however, child molesters were expected to rate more
anxiety than controls in a sexual situation with an adult female. This rating of anxiety was
expected to reflect child molesters lack of confidence and diflBculty with adult heterosexual
relationships. No differences were expected between admitters and deniers on the adult
vignettes.
Social desirability was also examined since denial among sex offenders is thought to be
at least partially related to a desire to present oneself in a positive light. Deniers were expected
to show more social desirability than admitters, while child molesters as a group were not
expected to differ significantly from controls. Repression was also examined since denial was
also thought to be partly related to unconscious defense mechanisms Thus, deniers were
expected to show greater repression than admitters Repression was thought to contribute to
permissive cognitions and was expected to negatively correlate with level of empathy since
repression might help molesters distance themselves from their sexual offending behavior, thus
reducing their ability to empathize with their victims' emotional reaction.
Hypotheses
This study examined decoding skill and cognitive factors among incarcerated heterosexual
child molesters (admitters and deniers), non-sex offending inmates, and community controls. It
was expected that:
(1) Child molesters would score lower than the two control groups on the decoding and social
perception tasks, with no differences expected between admitters and deniers.


123
between the perpetrator and victim, and unrelated to whether the perpetrator admits or
denies having committed the crime.
Hypothesis #6 proposed that child molesters would show more repression than
controls. Contrary to expectations, there were no significant differences among groups on
repression. This finding suggests that deniers are not characterized by a deeper level of
repression than admitters. Deniers do differ from admitters, however, in their level of
social desirability. As expected, deniers showed significantly more social desirability than
admitters. Thus, denial of child molesting behavior appears to be related to concerns of
social desirability rather than to differences in level of repression. This finding suggests
that deniers may have some motivation for treatment, given that they seem to be
motivated to please others by presenting themselves in a socially desirable light. This
need for social desirability might help them in treatment if they were to enter a treatment
group comprised of admitters in which telling the truth was considered to be the most
socially desirable action. Thus, treatment programs might target deniers by creating a
social climate within the program which places a high value on admission of ones crime
and offers social rewards for this behavior.
In addition to highlighting the need to subdivide child molesters based on denial
and the relationship to the victim, the results of the present study also highlight the
importance of offenders self-reported histories of abuse Not surprisingly, subjects
reporting histories of abuse scored lower on repression than those denying histories of
abuse. This is understandable since one would need to face ones problems, and thus not
repress, in order to admit to a history of abuse. In addition, subjects who reported a


22
addition, examined this variable among subgroups of child molesters familial versus
nonfamilial and admitters versus deniers.
The second aspect of the social information processing examined was child molesters'
cognitions regarding sexual contact. Part of this investigation involved an attempt to replicate
Stermac and Segal's (1989) results, with the modification that child molesters were subdivided
into appropriate subgroups Based on Stermac & Segals (1989) results, it was expected that
child molesters as a group would ascribe significantly more benefit to the child than control
groups when the childs response was positive or ambiguous, and would show no difference
from controls when the childs response was clearly negative. The child molesters were also
expected to attribute more responsibility and desire to the child and less overall responsibility to
the adult.
In addition, child molesters were expected to endorse more items on the Cognition
Questionnaire which are indicative of permissive beliefs and attitudes toward sexual contact
with children. Within the child molester group, deniers were expected to endorse fewer
permissive cognitions than admitters on both the vignettes and the Cognition Questionnaire.
This was expected since deniers show the most extreme form of denial and thus are not likely
to admit to cognitive distortions involving some acknowledgement of a sexual interest in
children. In addition, these measures of cognitive distortions tend to be face valid and thus
responses to these measures were expected to be highly influenced by social desirability.
In addition to replicating Stermac and Segal's (1989) results, this study investigated
child molesters' cognitions regarding sexual contact with adults. In contrast to their expected
distorted cognitions regarding sexual contact with children, child molesters were expected to


12
heterogeneous, displaying a variety of profiles ranging from nondiscriminating, to child or
adult preferences. Consistent with these results, a recent study found that incestuous
offenders reported less sexual interest in children than did nonfamilial offenders on the Pictorial
Sexual Interest Card Sort (Haywood and Grossman, 1994).
Differences have also been found between the MMPI profiles of incestuous versus
nonfamilial offenders (Erickson et al., 1987). In general, the incestuous offenders showed a
lesser degree of disturbance than the nonfamilial offenders. Among offenders who were the
biological fathers, 4-3 was the most common profile; however, this code type was equally
prevalent among the control groups. The authors suggest that the 4-3/3-4 profile describes
"chronic anger, overcontrolled hostility, marital discord, and passive-aggressive personalities,"
with the 4-3 characterized by more overt acting-out while the 3-4 is characterized by more
passivity (p.568).
Among incestuous stepfathers, 4-7/7-4 was the most common profile. The authors
associate this profile with "insensitivity to others, brooding interposed with episodic acting-out,
and alcoholism" (p.568). For the nonfamilial offenders, the 4-2/2-4 and 4-8/8-4 profile was
more common than among the incestous offenders. The 2-point peak code for the mean
profile of the child molesters was a 4-2. This same code has been found in several other studies
and is considered to be consistent with the clinical profile portraying child molesters as passive,
dependent, socially uncomfortable, and impulsive (Erickson et al., 1987).
Although these studies have yielded somewhat inconsistent results, the majority of
the studies have found that incestuous offenders appear more similar to controls than do


119
provocative questions in their own words, thus making it easier for them to reveal
cognitive distortions in their thinking. In addition, the interview was conducted
individually while the vignettes were completed in a group testing situation in which
inmates were sitting next to each other. There may have been strong demand
characteristics for circling particular answers in this group situation since it is generally
considered to be dangerous to admit to permissive attitudes toward sexual contact with
children while in a prison environment. Thus, there seem to be several plausible
explanations for the differences between the vignette and interview data, all of which
relate to concerns regarding social desirability, which was found to be particularly high
among deniers.
The fourth hypothesis suggested that child molesters would not show the same
cognitive distortions regarding sexual contact with an adult as they did regarding sexual
contact with a child. The results generally supported this hypothesis, however, the child
molesters differed from controls in their ratings of the womans desire. When the
womans response was clearly negative, the child molesters rated the womans desire for
the sexual contact as more dependent on the degree of sexual contact than did the
controls. This difference was slight, however, and was not found on other variables. In
fact, the child molesters rated the sexual touching and fondling as more harmful for the
woman than did the controls. This suggests that child molesters may be able to empathize
more with an adult victim than with a child in the age range of their own victim. Thus,
their clinically noted lack of empathy may be specific to their victims, or to children, rather
than generalizare to adults.


6
results may point to a deficit in the cognitive processing that precedes social behavior Clearly,
there needs to be more research focusing on the nature of this deficiency in child molesters
social information processing.
Cognitions of Child Molesters
Possible deficiencies in social information processing may be related to distorted
cognitions reported among child molesters (Stermac & Segal, 1989). In contrast to the
plethora of research conducted on child molesters' social skills, very little research has been
conducted on child molesters' cognitions, although cognition is believed to play an important
role in any form of human sexual arousal (Rook & Hammer, 1977). Most of the studies
focusing on child molesters' cognitions have focused on cognitive products that are thoughts or
images that form as a result of input of information and the interaction of cognitive structures,
propositions, and operations. Examples of cognitive products include self-statements,
attributions, and inferences.
Investigation of child molesters' cognitions is important as cognitions may play a
number of roles in the commission of a sexual offense. For example, specific cognitions may
contribute to the development of adults' sexual interest in children or they may be post hoc
rationalizations that faciliate continued offending. It is also possible that these cognitions are an
epiphenomena of deviant sexual arousal which would disappear once arousal was eliminated
(Stermac & Segal, 1989).
One of the first studies to examine cognitive factors among child molesters was
conducted by Frisbie, Vanasek, and Dingman (1967). Using the semantic differential, these
authors investigated child molesters' self view using ratings from institutionalized and


118
deniers when the child cried. The admitters showed a significantly different pattern from
deniers across the childs response, but did not significantly differ from deniers at the
individual levels of the childs response. The deniers also rated the man in the adult
vignettes as having less responsibility for the sexual interaction than did the admitters,
however, this result was largely attributable to the nonfamilial deniers who gave higher
ratings than the other five groups. This finding suggests that nonfamilial deniers, like
nonfamilial admitters, do show cognitive distortions regarding sexual interaction, although
they did not show many cognitive distortions on the child vignettes. This may reflect their
desire to present themselves in a good light since the child vignettes are much more face
valid than the adult vignettes. In sum, there were few differences between admitters and
deniers, and the few differences that emerged appeared to be more related to the extreme
pattern of the nonfamilial offenders Thus, the offenders relationship to the victim
appeared to be a more important variable than the offenders admission or denial of the
crime.
In contrast to the results from the vignettes, the interview data suggest that deniers
do show cognitive distortions regarding sexual contact with children, although they tend
to show fewer distortions than admitters. Deniers also tended to show more subtle types
of distortions regarding the genuineness of the mans feelings for the girl or vice versa.
Although the results suggest that social desirability was not significantly correlated with
the responses on the vignettes, it seems likely that the Likert scale format presented more
demands for social desirability given the structure and face validity of the answers. In
contrast, the open-ended question format of the interview forced the subjects to answer


55
showed that nonfamilial admitters ascribed significantly more benefit to the child than the
other groups when the child smiled, F(l,202)=19.907, p=000. This post-hoc contrast
was not significant when the child was passive or cried. Bonferroni adjusted comparisons,
however, showed that the nonfamilial admitters gave significantly higher ratings on the
passive vignettes than every group except for the nonfamilial deniers.
As expected, groups differed in their ratings of the childs enjoyment, however, the
difference only approached significance, F(5,202)=2.192, p=.057. Planned comparisons
showed no significant differences between child molesters and controls, F(l,202)=1.603,
p=.207, or between admitters and deniers, F(l,202)=1.484, p=.225, however, nonfamilial
offenders gave significantly higher ratings of child enjoyment than incestuous offenders,
F(l,202)=3.984, p=.047, as predicted.
There was also a significant Group X Childs Response interaction
F(10,404)=4.00, p=.000, indicating that the groups ratings of child enjoyment differed
across the levels of the childs response. More specifically, the groups differed
significantly in their ratings of child enjoyment on the vignettes in which the child smiled,
F(5,202)=3.997, p=.002, but did not show significant differences when the child cried or
was passive.
There was a significant Group X Child interaction for the planned comparison
between child molesters and controls, F(2,404)=8.443, p=.000, and for the comparison
between nonfamilial and incestuous offenders, F(2,404)=.019 (see Figure 6). When the
child was smiling, child molesters ascribed more enjoyment to the child than did controls,
F(l,202)=5.919, p=.016. Child molesters did not significantly differ from controls when


58
F(10,404)=3.739, p= 000 (see Figure 12). Planned comparisons showed significant
differences between child molesters and controls, F(2,404)=3.751, p=.024 and between
incestuous and nonfamilial offenders, F(2,404)=6.822, p=. 001 on the Group X Child
interaction. Admitters and deniers did not significantly differ on this interaction,
F(2,404)=l 479, p=.229.
On the smile vignettes, the difference between child molesters and controls
approached significance, F(l,202)=3.726, p=.055. There was no significant difference
between child molesters and controls when the child was passive, F(l,202)=2.764,
p=.098, or when the child cried, F(l,202)=.012, p=.911. Nonfamilial offenders ascribed
more responsibility to the child than did incestuous offenders when the child smiled,
F(l,202)=6.616, p=.011 and was passive, F(l,202)=5.585, p=.019, but not when the child
cried, F(l,202)=.022, p= 882.
The post-hoc comparison between nonfamilial admitters and the other five groups
approached significance using the Scheffe test, F(2,404)=16.365. Nonfamilial admitters
showed a similar pattern on this variable as on Child Benefit, Enjoyment and Desire
Using the Scheffe test, the nonfamilial admitters ascribed significantly more responsibility
to the child than the other groups when the child smiled, F(l,202)=14.129, p=.000, but
not when the child was passive or cried.
As expected, groups differed significantly in their ratings of the likelihood of other
men engaging in sexual contact with children, F(5,202)=2.837, p=.017 (see Figure 13).
Child molesters did not differ significantly from controls as expected, F(l,202)=3.0311,
p= 083. Nonfamilial sex offenders, however, did differ from incestuous offenders as


Mean Woman Desire Ratings
89
Woman Smiling Woman Passive Woman Crying
Level of Sexual Contact
l=Touching 2=Fondling 3=No Clothes 4=Ejaculation
Figure 20. Mean Woman Desire ratings for levels of sexual contact and woman's response.


institutions within the DOC system. I would also like to thank Nancy Haynes, Ph D, Senior
Psychologist; William Cook, Human Services Program Director; Joe Thornton, M.D.,
Medical Executive Director; and Dennis ONeill, Superintendent at Union Correctional
Institution. They helped make my research a reality by offering to let me conduct research at
their institution. Without their help, I simply would not have been able to conduct this
research!! In particular, Nancy Haynes, Ph D., and William Cook were incredibly generous
and giving in welcoming me to UCI and helping me get my data collection started. I cannot
express just how thankful I am for their help! I would also like to thank Don Kimbrell, Ph D.,
at the New River Correctional Institution for facilitating my entrance into the New River East
and West institutions. In addition, I send my thanks to the staff at the Marion Correctional
Institution.
I also extend my thanks to my dedicated research assistants: Jennie Sherrick, Nicole
Mahoney, Barry Kays, Laura McCoy, and Mike Britt. They were priceless assistants as they
were flexible in their schedules and their attitudes as I bounced them from prison to prison,
assigning difficult and varying tasks. I was especially grateful for their patience as they had to
keep up with my cra2y, breakneck schedule. I would not have been able to finish on time
without their hours of tireless work!
Last, but certainly not least, I would like to thank my loving husband who has been put
through the wringer while I have been driven to finish my dissertation. He has been patient and
understanding while I have maintained an unforgiving schedule and have been completely-
obsessed with my research. I would not have been able to get through this difficult time were it
not for his reassurance and unwavering support!
¡ii


Conclusions 125
REFERENCES 127
BIOGRAPHICAL SKETCH 133


18
Two studies reported a generalized deficit in empathy among child molesters. One
study reported that 35.8% of child molesters lacked empathy, however, the method for
assessing empathy was unclear (Abel, Mittleman, & Becker, 1985). Another study
reported significant differences between sex offenders and community subjects using
Hogans Empathy Scale (Rice, Chaplin, Harris, & Coutts, 1990; as referenced in Marshall
et al., 1993). Marshall et al. (1993) suggest that the samples used in these studies may not
be representative of the general sex offender population, as they consisted of a more
deviant and chronic group of sex offenders than is typically found in a general prison
population or outpatient clinic. Marshall et al. (1993) also point out the importance of the
method used to assess empathy. A variety of assessment instruments measure different and
limited aspects of empathy They cite a study conducted by Seto (1992) which reported
that incarcerated rapists showed no empathy deficiency on the Mehrabian and Epstein
(1972) measure of empathy, but did show a deficit on Hogans scale.
In their 1993 study, Marshall et al. used the Interpersonal Reactivity Index (Davis,
1983), citing it as the best available measure for assessing the complex nature of empathy.
They found no deficits in empathy among incarcerated child molesters when compared
with controls. In contrast, outpatient child molesters did show a deficiency in empathy.
More specifically, child molesters scored significantly lower on the total IR1 score and on
the Fantasy subscale. Unfortunately, the study did not differentiate between familial and
nonfamilial offenders or between homosexual and heterosexual offenders, thus introducing
a potentially confounding factor. Nonetheless, the authors suggest that the differences
among the child molesters may have resulted from motivation among the incarcerated


39
The Vocabulary Scale of the Wechsler Adult Intelligence Scale-Revised was used as a
rough estimate of level of intellectual functioning. The Vocabulary Scale contains 35 items or
vocabulary words which the subject is asked to define. Responses are scored on a 0-2 point
scale. Split-half reliability for the Vocabulary Scale ranges from .94 to .96 across the 16-74
age range and the test-retest reliability is 93 for ages 25-34, and .91 for ages 45-54 (Wechsler,
1981). The Vocabulary score is highly correlated with the WAIS-R Full Scale score (r=.85)
and the Verbal Scale score (r=.90). The WAIS-R was standardized on a sample of 1,880
subjects stratified according to age, sex, race, geographic region, education, and occupation.
Summary of the Variables.
The background variables included age, socioeconomic status score on the
Hollingshead Four Factor Index of Social Status, vocabulary score on the WAIS-R, and self-
report of physical or sexual abuse as assessed by the Life Experiences Questionnaire The
dependent variables used to assess cognitive distortions included the total score on the
Cognition Questionnaire, and the set variables on the child and adult versions of the written
vignettes: 1) Females Enjoyment of the sexual interaction, 2) Females Desire, 3) Females
Benefit, 4) Females Responsibility, 5) Females Harm, and 6) the Mans Responsibility There
were two additional variables on the adult female vignettes including the Womans Sexual
Attraction to the man and the Mans Anxiety regarding the sexual interaction.
Decoding ability was assessed using the total scores from the Profile of Nonverbal
Sensitivity Test and the Interpersonal Perception Task. Other dependent variables included the
total scores on the Repression-Sensitization Scale and the Marlowe-Crowne Social Desirability


2
Segal and Marshall (1985) conducted one of the first studies documenting a social skill
deficit in child molesters. The authors compared rapists, child molesters, non-sexual offenders,
and community control subjects on social skills measures of videotaped conversations with
females. Although all incarcerated groups were less socially skilled than the community control
group, the child molesters also demonstrated a skill deficit relative to the rapists and inmate
control group.
In a similar study using behavioral assessment of controlled social interaction, both
rapists and child molesters were significantly less skilled than community control subjects on
measures of form of conversation and general level of social skill (Overholser & Beck, 1986).
Although child molesters did not show a social skills deficit relative to the rapists in this study,
they did show a trend toward lower assertiveness than rapists and controls on a questionnaire
assessment of assertiveness. In addition, the child molesters displayed significantly more fear of
being negatively evaluated than did rapists and controls.
Segal and Marshall (1985b) also found that child molesters showed more heterosocial
inadequacy than rapists. Compared to the rapist group, child molesters rated themselves as less
skilled and as more anxious in a heterosexual interaction with a confederate. They also scored
significantly higher than rapists on the Social Avoidance and Distress Scale (SADS). The
authors note that this social inadequacy appears to be largely cognitive, given that the child
molesters were rated as equally competent in the social situation as the other groups by both
the confederate and independent judges. Consistent with the results of the Overholser and
Beck (1986) study, child molesters rated themselves as less assertive in accepting positive and
negative feedback from others and scored significantly lower than rapists and controls on the


24
(2) Child molesters would show distorted cognitions regarding sexual contact with a female
child as measured by the vignettes and Cognition Questionnaire. More specifically, child
molesters would ascribe more benefit, desire, and responsibility to the child, and less
responsibility and punishment for the man Nonfamilial sex offenders would also show more
cognitive distortions than incestuous offenders.
(3) Deniers would show fewer cognitive distortions than admitters on both measures regarding
sexual contact with children.
(4) On the adult vignettes, child molesters would not show cognitive distortions on the same
variables as they would on the child vignettes. However, child molesters would rate more
anxiety than controls regarding adult sexual interaction with an adult female.
(5) Child molesters would demonstrate less empathy than controls; deniers would show less
empathy than admitters. Empathy scores would correlate positively with decoding score and
negatively with repression.
(6) Child molesters would endorse more symptoms of repression than controls. Deniers
would show more repression than admitters.
Alternative Hypotheses
Child molesters' cognitive distortions could be related to sexual contact in general,
rather than specific to sexual contact with children. In that case, child molesters would be
expected to show the same cognitive distortions on the vignettes presenting sexual contact
with an adult female as in the vignettes presenting sexual contact with a female child.
Deniers may not show greater repression, but may show more social desirability. Thus,
denial might be more related to perceived demands for social desirability and less related to


17
seriousness of his offense, and may facilitate the development and maintenance of cognitive
distortions to rationalize or justify the behavior. Assuming that deniers would show more
repression than admitters, one might expect them to show more cognitive distortions regarding
sexual contact with children, however, as several studies suggest (Becker et al., 1992;
Haywood et al., 1993; Haywood & Grossman, 1994), deniers appear to be effective at
suppressing both deviant sexual responses as well as deviant thoughts. Thus, deniers would be
expected to have more cognitive distortions, but would be less likely to show these distortions,
especially on a face valid measure.
In sum, since some form of denial is common among most child molesters, it might be
expected that child molesters as a group would show more repression than non-child
molesters. The subgroup of deniers might also be expected to show more repression than child
molesters who admit to the offense.
Empathy
Like repression, empathy is another variable which might be related to decoding
skill and cognitive distortions among child molesters. Clinicians have long noted that sex
offenders show little empathy for their victims. Many authors have also suggested,
however, that this deficit in empathy may be generalized in the sense that it applies toward
all people, rather than applying solely to the victim (Marshall, Jones, Hudson, &
McDonald, 1993). The few studies that have measured empathy among sex offenders
have focused on this concept of generalized empathy, and have yielded inconsistent
results.


Mean Child Enjoyment Ratings
75
Child Response
Figure 6. Mean Child Enjoyment ratings for levels of child response.


102
Table 10
Group means on Child Desire when the childs response was positive.
Incest Admitters
M
2.12
Inmate Controls
M 2.20b
SD
.20
SD
.12
N
23
N
61
Incest Deniers
M
2.21
Community Controls
M 1.91b
SD
.17
SD
.17
N
31
N
30
Nonfamilial Admitters
M
2.82a
Nonfamilial Deniers
M 2.08b
SD
.17
SD
.17
N
31
N
32
Note. Means with different subscripts differ significantly at p< 05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.


49
Reduction of variables. In order to facilitate comparison with the results of
Stermac & Segal (1989), Pearson correlation matrices were conducted for each of the
twelve child vignettes to see which variables were highly correlated across levels of sexual
contact and childs response (see Table 5). This matrix shows that the nine variables are
highly correlated with each other. Stermac & Segal (1989) addressed this issue of
multicollinearity in their study by combining highly correlated variables into two
composite variables. They found that Benefit to Child, Harm to Child, and the Childs
Enjoyment were highly correlated. Child Benefit and Enjoyment were negatively
correlated with Child Harm, with correlation coefficients ranging from .32 to .65. They
combined these variables into a composite variable called Benefit to the Child.
In the present study, the negative correlation coefficients between Child Benefit
and Enjoyment and Child Harm ranged from .09 to .34. Child Benefit and Child
Enjoyment showed correlation coefficients ranging from 18 to .59. Since these
correlations were low and spanned a large range, these three variables were not combined
into a composite variable. When the variables were summed across levels of childs
response and sexual contact, Child Benefit correlated with Enjoyment at 60 and with
Harm at -.30. The latter two variables correlated -.35.
Stermac & Segal (1989) also formed a second composite variable called Child
Complicity from the variables of Child Responsibility and Child Desire. Across the levels
of sexual contact and childs response, these two variables showed correlations ranging
from .37 to .71. In the present study, these correlations ranged from .40 to .75, closely
replicating Stermac & Segals (1989) findings. Although we could have formed a


47
p=.739. It was expected that empathy scores would be significantly positively correlated
with decoding scores. Consistent with the results from the PONS, the correlation of the
IPT score with the IRI total score was not significant, r=. 110.
Analysis of Cognitive Distortions
Cognition Questionnaire
It was predicted that child molesters would show more cognitive distortions than
controls on the Cognition Questionnaire which assesses ones attitude toward sexual
contact with children. Seven outliers were identified based on studentized residual values
and these cases were dropped from the analysis. A planned comparison revealed that child
molesters did endorse significantly more cognitive distortions about sexual contact with
children than did the control groups, F(l,188)=7.478, p=.007 (see Table 4). Lower scores
on this measure indicate more cognitive distortion. An ANOVA with the six groups
revealed a significant main effect for Group F(5,188)=2.862, p=.023 and social
desirability, F(l,188)=6.929, p=.009 (see Figure 4). It was also predicted that deniers
would show fewer cognitive distortions than admitters, however, planned comparisons
showed no significant differences between these two groups F(l,188)=.210, p=.648
Nonfamilial offenders were expected to show more cognitive distortions than incestuous
offenders. Consistent with this prediction, the difference between incestuous and
nonfamilial sex offenders was significant, F(l,188)=4.041, p=.046, with nonfamilial sex
offenders endorsing more cognitive distortions than incestuous offenders Pairwise


Mean Child Enjoyment Ratings
79
Child Response
Nonfamilial Offenders
Incestuous Offenders
Figure 10. Mean Child Desire ratings for levels of child response.




116
although, they may still have a need to rationalize their incestuous behavior to some
degree.
Contrary to expectations, child molesters did not differ from controls in their
ratings of the mans degree of responsibility for the sexual interaction. These results
contrast with those of Stermac & Segal (1989) who found that child molesters attributed
less responsibility overall to the man, and that their ratings depended more on the level of
sexual contact and childs response than did the controls ratings. The child molesters
ratings differed from controls on all levels of sexual contact, but were similar to the
controls ratings when the childs response was negative. Based on these results, it was
expected that child molesters in the present study would show the same pattern of
responding. The reason for these different results is unclear. It may be that the difference
in sample composition might account for this difference. Stermac & Segal (1989)
included homosexual and heterosexual child molesters in their sample, while the present
study included only heterosexual child molesters It may be that homosexual molesters
attribute less responsibility to to the man than heterosexual molesters, thus resulting in
different results. Another possible explanation might be related to differences in the
forensic settings where the data was collected and differential demands for social
desirability in these settings. It may be that child molesters volunteering for treatment at a
psychiatric facility may have fewer social desirability concerns than incarcerated child
molesters not participating in a treatment program, and thus incarcerated molesters may be
more reluctant to admit to obvious cognition distortions regarding the mans
responsibility. In any case, the present finding suggests that child molesters are aware of


Mean Man Responsibility Ratings
92
Figure 23. Mean Man Responsibility ratings for levels of sexual contact and child's response.


CHAPTER 3
RESULTS
Based on the literature, it was predicted that: 1) child molesters would score
lower than controls on decoding tasks; 2) child molesters would show more cognitive
distortions than controls regarding sexual contact with children, and nonfamilial offenders
would show more cognitive distortions than incestuous offenders; 3) deniers would show
fewer cognitive distortions than admitters, 4) child molesters would attribute more anxiety
to the man in the adult vignettes than would controls; 5) child molesters would show less
empathy than controls, and deniers would show less empathy than admitters; and 6) child
molesters would show more repression than controls, and deniers would show more
repression than admitters.
Background Variables
Background variables included age, socioeconomic status, vocabulary scores,
social desirability, and self-reported abuse history. In order to examine among-group
differences on these variables, several one-way ANOVAs were conducted with Group
(incestuous admitters, incestuous deniers, nonfamilial admitters, nonfamilial deniers, non
sex offending inmate controls, community controls) as the between subjects factor and
age, socioeconomic status, and vocabulary scores as the dependent measures. Several
ANOVAs were conducted rather than a MANOVA since the primary interest was to
examine group differences on each variable. There were no significant differences on
41


35
respectively). Davis (personal communication, 1994) recommends examining each of the
subscale scores separately rather than using the total IRI score since each of the subscales
appears to measure a different aspect of empathy. Other researchers, however, argue that the
IRI total score may serve as an overall measure of empathy.
Measure of Repression
The Repression-Sensitization Scale (Bryne, 1961) was used to measure repression.
The R-S scale is composed of 127 items from six Minnesota Multiphasic Personality Inventory
(MMPI) scales, three measuring sensitization and three measuring repression. The range of
possible scores is 0 to 127; high scores indicate sensitization while low scores indicate
repression. The behavioral dimension of repression-sensitization is viewed as a continnum,
with the repressive extreme involving avoidance defenses such as denial and the sensitizing
extreme involving approach defenses such as intellectualization.
The R-S scale was standardized on 733 male and 571 female college students (Byrne,
Barry, & Nelson, 1963). Test-retest reliability over three months was .82 and split-half
reliability was .94. In terms of face validity, seven of 10 judges showed agreement on 90% of
the items when asked to respond as they thought a repressor would respond. R-S scores have
been found to be consistently related to self-report measures More specifically, scores on the
R-S scale were related to self-ideal discrepancy (r= 63) and negative self-description (r=.68),
supporting the notion that individuals who use repressive defense mechanisms are less likely to
verbalize negative self-descriptions than individuals with sensitizing defense mechanisms.
Correlations between the R-S scale and Marlowe-Crowne Social Desirability Scale
range from -.37 to -.49, suggesting that repression-sensitization is partially, but not entirely a


10
Subtypes of Child Molesters
Nonfamilial Versus Incestuous Child Molesters
Unfortunately, many of the studies cited above did not subdivide their samples of
child molesters based on gender of the victim or relationship to the victim. This tendency
to lump many types of sex offenders into one general category has been cited as one of the
potential reasons for inconsistent results (Howells, 1981). A variety of studies over the
years have demonstrated the need to differentiate between heterosexual and homosexual
offenders, and between nonfamilial and familial offenders. Howells (1981) offers the
conceptualization that child molesters fall on a continuum, ranging from pedophiles with a
primary sexual preference for children, to persons with a primary sexual preference for
adults. Most reseachers in the area agree that not all child molesters are pedophiles.
Thus, some theorists propose a distinction between molestation based on situational
factors versus molestation based on sexual preference. For example, Pawlak, Boulet, and
Bradford (1991) suggest that some child molestation may be related to situational factors
interacting with various types of psychopathology.
Many theorists view incestuous offenders as situational offenders and nonfamilial
offenders as more sexual preference offenders (Pawlak, Boulet, & Bradford, 1991). Thus,
incestuous offenders deviant sexual behavior is thought to be less related to a primary
sexual interest in children and more related to taking advantage of easy sexual access
Consistent with this view, incestuous offenders showed more age appropriate preferences
and recidivated at lower rates than nonfamilial child molesters (Marshall, Barbaree, &


9
Although the results suggest a difference in cognition, it is unclear if these cognitions or
schemata develop as post hoc rationalizations for their behavior or if they were formed before
the behavior. Abel et al. (1984) suggest that some offenders who find themselves sexually
attracted to children respond to this dilemma by modifying their cognitions in order to justify
their deviant behavior. These cognitive distortions may also be viewed in terms of Finkelhofs
(1984, 1986) theoretical model of child molestation which describes various disinhibitors such
as alcohol use or psychosis which enable child molesters to overcome their internal inhibitions.
Thus, cognitive distortions may be viewed as another type of internal disinhibitor (Hollin &
Howells, 1991).
The results from the Stermac and Segal (1989) study suggest that the child molesters
were less sensitive or responsive to the child's cues than were the controls, suggesting a lack of
empathy. As the authors indicated, child molesters perceptions may be modified only in
response to a clearly negative and unambiguous response such as the child's crying. It is
important to note that the vignettes in this study were written and did not require the subjects
to "read" or interpret others' cues. It is likely that in real life situations, the child's cues would
probably not be as clearly portrayed as they are in a written vignette. Since child molesters
tend to interpret ambiguous responses on the child's part as positive within the structure of a
simple written vignette, they may also have difficulty reading others' ambiguous nonverbal cues
in real life situations. It may be that child molesters have difficulty accurately reading others'
cues and that this deficiency in information processing contributes to permissive cognitions.


5
of empathy (Hall, 1979; Riggio, Tucker, & Coffaro, 1989; N.D. Feshbach, 1978). That is, one
can experience true empathic concern only if he or she is nonverbally sensitive in order to
vicariously experience the other's emotional state (Mehrabian & Epstein, 1972; Riggio et al.,
1989). Thus, child molesters' clinically noted lack of empathy may result from a deficiency in
decoding and may also contribute to the commission of the offense by serving as a disinhibitor.
In addition, a decoding deficiency might also be related to cognitive distortions found among
child molesters (Stermac & Segal, 1989).
According to McFalTs (1990) social information processing model, the next two steps
following the decoding process are the decision skill process and the enactment skill process,
respectively. Decision skills involve "generating response options, matching these to task
demands, selecting the best option, searching for that option in the behavioral repertoire, and
evaluating the subjective utility of that option's likely outcomes relative to the likely outcomes
of other options (McFall, 1990, p. 313)." Enactment skills then involve carrying out the
behavioral option selected. Inappropriate social behavior may result from a problem at any of
these three levels.
A study conducted by Barbaree, Marshall, and Connor (1988; as referenced in
Marshall, Laws, & Barbaree, 1990) suggests that child molesters may not have difficulty with
decision skills, but may have problems with enactment skills. This study assessed child
molesters' social problem-solving skills in social situations involving children and adults Child
molesters were found to be as skilled as nonoffenders in recognizing problems and generating
alternative solutions, but tended to choose socially unacceptable solutions. They also failed to
recognize the likely negative outcomes of these solutions. The authors suggest that these


99
Table 7
Group means on Benefit to Child across vignettes
Incest Admitters Inmate Controls
M
1.17b
M
1.52
SD
.17
SD
.11
N
23
N
61
Incest Deniers
Community Controls
M
1.37b
M
1.38b
SD
.15
SD
.15
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.01a
M
1.37
SD
.15
SD
.15
N
31
N
31
Note. Means with different subscripts differ significantly at p< 05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.


SOCIAL INFORMATION PROCESSING IN CHILD MOLESTERS: AN
EXAMINATION OF DECODING SKILL AND COGNITIVE FACTORS
By
JULIE C. MEDLIN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1995

ACKNOWLEDGEMENTS
I would like to thank Jacquelin Goldman, Ph D., ABPP, my dissertation chair, for
supporting me in pursuing my interest in research with a particularly difficult forensic
population. She encouraged me to follow my true research interests rather than design a dull,
but easy, research study. I also thank her for helping me make some difficult design decisions
during the data collection process, and then later in the statistical analysis process. Overall, she
allowed me to develop into an independent researcher, as she gave me the freedom to be
creative and encouraged me to take risks. As a result, I have truly enjoyed and have learned
from my research pursuits in graduate school.
I would also like to thank my dissertation committee: Michael Peni, Ph D., James
Johnson, Ph D., James Algina, Ph D., and Otto von Mering. I offer my thanks for their
patience during my proposal meeting when I explained entirely too many hypotheses and a
complex design. I appreciate their willingness to allow me to venture into such a difficult
project potentially fraught with many practical problems, given the subject population and
prison environment
The Florida Department of Corrections also deserves enormous gratitude for allowing
me to conduct research in their prison system. In particular, I would like to thank William
Bales, PhD., Chief of the Bureau of Planning, Research, and Statistics for the Florida
Department of Corrections. He granted me permission to conduct my research at various
ii

institutions within the DOC system. I would also like to thank Nancy Haynes, Ph D, Senior
Psychologist; William Cook, Human Services Program Director; Joe Thornton, M.D.,
Medical Executive Director; and Dennis ONeill, Superintendent at Union Correctional
Institution. They helped make my research a reality by offering to let me conduct research at
their institution. Without their help, I simply would not have been able to conduct this
research!! In particular, Nancy Haynes, Ph D., and William Cook were incredibly generous
and giving in welcoming me to UCI and helping me get my data collection started. I cannot
express just how thankful I am for their help! I would also like to thank Don Kimbrell, Ph D.,
at the New River Correctional Institution for facilitating my entrance into the New River East
and West institutions. In addition, I send my thanks to the staff at the Marion Correctional
Institution.
I also extend my thanks to my dedicated research assistants: Jennie Sherrick, Nicole
Mahoney, Barry Kays, Laura McCoy, and Mike Britt. They were priceless assistants as they
were flexible in their schedules and their attitudes as I bounced them from prison to prison,
assigning difficult and varying tasks. I was especially grateful for their patience as they had to
keep up with my cra2y, breakneck schedule. I would not have been able to finish on time
without their hours of tireless work!
Last, but certainly not least, I would like to thank my loving husband who has been put
through the wringer while I have been driven to finish my dissertation. He has been patient and
understanding while I have maintained an unforgiving schedule and have been completely-
obsessed with my research. I would not have been able to get through this difficult time were it
not for his reassurance and unwavering support!
¡ii

TABLE OF CONTENTS
Pace
ACKNOWLEDGEMENTS ii
ABSTRACT vi
CHAPTERS
1 INTRODUCTION AND REVIEW OF LITERATURE 1
Social Skills of Child Molesters 1
Cognitions of Child Molesters 6
Subtypes ofChildMolesters 10
Other Variables Related to Decoding Skill and Cognitions 16
Rationale for the Proposed Study 20
Hypotheses 23
Alternative Hypotheses 24
2 METHODS 26
Subjects 26
Research Design 27
Dependent Variables 28
Procedures 39
3 RESULTS 41
Background Variables 41
Method of Analysis 44
Analysis of Decoding Skill 46
Analysis of Cognitive Distortions 47
Comparison of Child Versus Adult Vignettes 66
Analysis of Empathy 67
Analysis of Repression 68
4 DISCUSSION 110
Summary of Findings 125
iv

Conclusions 125
REFERENCES 127
BIOGRAPHICAL SKETCH 133

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
SOCIAL INFORMATION PROCESSING IN CHILD MOLESTERS: AN EXAMINATION
OF DECODING SKILL AND COGNITIVE FACTORS
By
Julie C. Medlin
December 1995
Chairperson: Jacquelin Goldman
Major Department: Clinical and Health Psychology
This study examined social information processing in heterosexual child molesters.
More specifically, this study examined child molesters decoding skill and cognitions regarding
sexual contact with children. The literature has indicated that child molesters are socially
deficient; however, the nature of this deficiency is unclear. That is, this deficiency may result
from any number of problems in their cognitive operations, including deficiencies in decoding
skill. Thus, the present study examined child molesters' decoding skill, or ability to "read"
others' cues. Contrary to expectations, child molesters did not show a deficit in decoding skill
or social perception when compared with control groups. Consistent with expectations, child
molesters showed cognitive distortions regarding sexual contact with children. All four sex
offender groups (nonfamilial admitter, nonfamilial denier, incestuous admitter, incestuous
vi

denier) prescribed significantly less punishment and treatment for a child molester than did the
control groups. Most of the other cognitive distortions, however, were found among
nonfamilial admitters rather than among the other sex offender groups. When reading vignettes
portraying sexual interaction with a female child, nonfamilial admitters attributed significantly
more benefit, responsibility, desire, and enjoyment to the child than the other sex offender
groups and control groups. As expected, incestuous offenders differed significantly from
nonfamilial offenders in that they scored more similarly to controls. In addition, sex
offenders who denied having committed their sexual crimes generally endorsed fewer
cognitive distortions than admitters and showed more social desirability than admitters.
Contrary to expectations, child molesters did not differ from controls in their ratings of anxiety
in a sexual situation with an adult female.
Other variables examined included empathy, repression, and abuse history. Contrary
to predictions, child molesters did not score significantly lower on empathy than controls.
There was also the unexpected finding that nonfamilial offenders scored higher than
incestuous offenders on empathy. There were no differences among groups on repression.
Consistent with expectations, more child molesters than controls reported having been
physically or sexually abused as a child, as well as more admitters than deniers. In sum,
significant differences were found between incestuous and nonfamilial offenders, as well as
vii
between admitters and deniers.

CHAPTER 1
INTRODUCTION AND REVIEW OF LITERATURE
Child molestation is a major social problem which has only recently begun to receive
adequate research attention over the past decade (Knight, Rosenberg, & Schneider, 1985;
Abel, Rouleau, & Cunningham-Rathner, 1985; Marshall, Laws, & Barbaree, 1990). Despite
these research efforts, few theories have emerged to explain this phenomenon. Two areas of
potential relevance to theory construction include child molesters' social perception and their
cognitions about sexual interaction with children Research in these areas suggests a deficiency
in child molester's social information processing, and highlights the need for a more focused
investigation of these variables and their potential relationship.
Social Skills of Child Molesters
There have been numerous studies focusing on the social competence of child
molesters (Segal & Stermac, 1989). The social deficit model of child molestation suggests that
child molesters lack social skills and, as a result, have unsatisfactory heterosocial relationships
(Groth & Bimbaum,1979). The clinical literature has frequently characterized child molesters
as having impaired social relationships and discomfort with adult sexuality (Hobson, Boland,
Jamieson, 1985). More specifically, Finkelhor (1984, 1986) proposed a four-factor model of
child molestation postulating that child molesters turn to children partly because alternative
sources of sexual gratification have been blocked or inhibited.
1

2
Segal and Marshall (1985) conducted one of the first studies documenting a social skill
deficit in child molesters. The authors compared rapists, child molesters, non-sexual offenders,
and community control subjects on social skills measures of videotaped conversations with
females. Although all incarcerated groups were less socially skilled than the community control
group, the child molesters also demonstrated a skill deficit relative to the rapists and inmate
control group.
In a similar study using behavioral assessment of controlled social interaction, both
rapists and child molesters were significantly less skilled than community control subjects on
measures of form of conversation and general level of social skill (Overholser & Beck, 1986).
Although child molesters did not show a social skills deficit relative to the rapists in this study,
they did show a trend toward lower assertiveness than rapists and controls on a questionnaire
assessment of assertiveness. In addition, the child molesters displayed significantly more fear of
being negatively evaluated than did rapists and controls.
Segal and Marshall (1985b) also found that child molesters showed more heterosocial
inadequacy than rapists. Compared to the rapist group, child molesters rated themselves as less
skilled and as more anxious in a heterosexual interaction with a confederate. They also scored
significantly higher than rapists on the Social Avoidance and Distress Scale (SADS). The
authors note that this social inadequacy appears to be largely cognitive, given that the child
molesters were rated as equally competent in the social situation as the other groups by both
the confederate and independent judges. Consistent with the results of the Overholser and
Beck (1986) study, child molesters rated themselves as less assertive in accepting positive and
negative feedback from others and scored significantly lower than rapists and controls on the

3
general assertion subscale on the Callner-Ross Assertiveness Scale. In a follow-up study, Segal
and Marshall (1986) investigated social perception in child molesters in heterosexual situations
by having them interact with a female confederate. Compared with other sex offenders and
community controls, child molesters were found to be significantly worse at prediction and
evaluation of their own performance. Taken together, these findings suggest that child
molesters perceive themselves as less socially skilled and assertive than others, and are more
anxious than others about their social interaction. Given the inconsistent results, it is unclear
whether or not child molesters are actually less socially skilled than rapists and non-sex-
offending inmate controls.
Although the literature generally indicates lower social competence among child
molesters, there is currently no specific theory or clear pattern of deficits regarding child
molesters' social information processing Their interpersonal functioning may be affected by a
number of factors including accuracy of processing social cues and information. In addition to
appropriate response skills, "ability to accurately read the social environment," or decoding
skill, is necessary for social performance (Morrison & Bellack, 1981, p.70). Decoding skills
refer to "the afferent processes involved in the accurate reception, perception, and
interpretation of incoming sensory information (McFall, 1990, p. 313)." Ifa man does not
receive a woman's cues or misperceives or misinterprets these cues, then his social behavior
toward the woman is more likely to be inappropriate. Based on his social information
processing model, McFall (1990) suggests that the logical place to begin exploring social
information processing is at the beginning of the input process with decoding skills. This is the
point where the information input is more or less equivalent for everyone.

4
Sexual assaulters in general have been hypothesized to have difficulty processing social
information from women as well as show a tendency to misinterpret social messages. Lipton,
McDonel, and McFall (1987) investigated this hypothesis, comparing rapists, violent
nonrapists, and nonviolent nonrapists on assessments of heterosocial cue-reading accuracy in
intimate and first-date situations. Rapists were found to miscontrue women's cues, particularly
cues involving negative moods in first-date situations, suggesting a deficit in their social
information processing. There appears to have been no such study conducted with child
molesters examining their ability to read others' cues. In fact, there have been few studies of
cognitive operations in sexual offenders in general. Cognitive operations refers to the various
ways in which the information processing system operates, including attention allocation,
encoding, decoding, or control processes.
Such studies are needed since child molesters' social skill deficiencies and cognitions
may be related to a difficulty in reading others' cues. It has been noted clinically that child
molesters misidentify emotions (Hobson, Boland, & Jamieson, 1985). Another reason to
expect that child molesters might be generally deficient in this skill is that low social
competence (Feldman & Rime, 1991) and an abuse history (Camras, Grow, & Ribordy, 1983)
are often associated with deficiencies in decoding. The latter variable is important to account
for since many child molesters (estimates range from 25% to 75%) report a history of abuse
(Mayer, 1988; Hamilton & Ondrovik, 1993) and those who report abuse may comprise a
distinct clinical subgroup (Becker, Kaplan, & Tenke, 1992).
In addition, child molesters are often characterized as lacking empathy (Hobson,
Boland, & Jamieson, 1985), and ability to read others' cues is believed to be a key component

5
of empathy (Hall, 1979; Riggio, Tucker, & Coffaro, 1989; N.D. Feshbach, 1978). That is, one
can experience true empathic concern only if he or she is nonverbally sensitive in order to
vicariously experience the other's emotional state (Mehrabian & Epstein, 1972; Riggio et al.,
1989). Thus, child molesters' clinically noted lack of empathy may result from a deficiency in
decoding and may also contribute to the commission of the offense by serving as a disinhibitor.
In addition, a decoding deficiency might also be related to cognitive distortions found among
child molesters (Stermac & Segal, 1989).
According to McFalTs (1990) social information processing model, the next two steps
following the decoding process are the decision skill process and the enactment skill process,
respectively. Decision skills involve "generating response options, matching these to task
demands, selecting the best option, searching for that option in the behavioral repertoire, and
evaluating the subjective utility of that option's likely outcomes relative to the likely outcomes
of other options (McFall, 1990, p. 313)." Enactment skills then involve carrying out the
behavioral option selected. Inappropriate social behavior may result from a problem at any of
these three levels.
A study conducted by Barbaree, Marshall, and Connor (1988; as referenced in
Marshall, Laws, & Barbaree, 1990) suggests that child molesters may not have difficulty with
decision skills, but may have problems with enactment skills. This study assessed child
molesters' social problem-solving skills in social situations involving children and adults Child
molesters were found to be as skilled as nonoffenders in recognizing problems and generating
alternative solutions, but tended to choose socially unacceptable solutions. They also failed to
recognize the likely negative outcomes of these solutions. The authors suggest that these

6
results may point to a deficit in the cognitive processing that precedes social behavior Clearly,
there needs to be more research focusing on the nature of this deficiency in child molesters
social information processing.
Cognitions of Child Molesters
Possible deficiencies in social information processing may be related to distorted
cognitions reported among child molesters (Stermac & Segal, 1989). In contrast to the
plethora of research conducted on child molesters' social skills, very little research has been
conducted on child molesters' cognitions, although cognition is believed to play an important
role in any form of human sexual arousal (Rook & Hammer, 1977). Most of the studies
focusing on child molesters' cognitions have focused on cognitive products that are thoughts or
images that form as a result of input of information and the interaction of cognitive structures,
propositions, and operations. Examples of cognitive products include self-statements,
attributions, and inferences.
Investigation of child molesters' cognitions is important as cognitions may play a
number of roles in the commission of a sexual offense. For example, specific cognitions may
contribute to the development of adults' sexual interest in children or they may be post hoc
rationalizations that faciliate continued offending. It is also possible that these cognitions are an
epiphenomena of deviant sexual arousal which would disappear once arousal was eliminated
(Stermac & Segal, 1989).
One of the first studies to examine cognitive factors among child molesters was
conducted by Frisbie, Vanasek, and Dingman (1967). Using the semantic differential, these
authors investigated child molesters' self view using ratings from institutionalized and

7
community-based child molesters on their actual and ideal views of themselves. The measure
consisted of pairs of bipolar variables which reflect aspects of an individual's personality traits,
physical characteristics, and value orientations. Subjects located themselves on these
dimensions for their actual versus ideal selves. The largest discrepancies between the two
ratings were found on adjectives related to interactions with others. Although the results
suggest that child molesters feel less socially competent than they desire, comparisons with
nonoffender controls cannot be made since this group was not included in the study.
A later study on cognitive factors (Howells, 1978) investigated heterosexual child
molesters' perceptions of adult females, adult males, and children using the repertory grid
technique. The results indicated that child molesters employed more constructs related to
dominance in their descriptions of both children and adults. Children were perceived as being
at the nondominant pole, as "undemanding" and "passive" while adults were viewed as more
"domineering" and "demanding." The results also suggested that child molesters may be
attracted to children for more than sexual reasons, such as for their innocence and capacity for
unconditional love. Unfortunately, these results are only suggestive given the small sample size
of ten child molesters, problems in the statistical analysis, and the failure of a follow-up study
(Horley, 1988) to replicate these findings.
A more recent study (Abel et al., 1989) has focused on child molesters' cognitions
concerning child molestation. This study used a 29-item Likert scale and found 6 reliable
factors covering the general area of harm caused to the child by the molestation. The factors
successfully separated the child molesters from the control group, suggesting that child
molesters show cognitive distortions minimizing the harm of the molestation.

s
In a similar study, Stermac and Segal (1989) tested the hypothesis that child molesters'
cognitions are more permissive and accepting of sexual contact with children. They used
written vignettes describing sexual contact between an adult and child. The vignettes contained
varying degrees of sexual contact (touching fondling no clothes, ejaculation) and varying
responses from the child (smiling passive, crying). After reading each vignette, child molesters
and controls responded to questions assessing their view of the adult's and child's behavior.
The results indicate that child molesters ascribed more benefits of the molestation to
the child than did the controls. Unlike the controls, the child molesters did not significantly
decrease their amount of ascribed benefit to the child as the amount of sexual contact
increased. The child molesters also differed significantly from the controls in their perceptions
of benefit based on the child's response, although their response was similar to that of the
controls when the child's response was negative. Compared with the controls, the child
molesters attributed more responsibility to the child and less overall responsibility to the adult.
In general, the child molesters' ratings were less affected by the amount of sexual contact or the
child's response.
The child molester group also showed significant differences from controls on the
Cognition Questionnaire (Abel, Becker, & Cunningham-Rathner, 1984), endorsing more items
indicative of permissive beliefs and attitudes towards adult sexual contact with children. These
results are consistent with clinical reports of child molesters' distorted cognitions regarding
sexual contact with children (Abel, Becker, & Cunningham-Rathner, 1984; Abel, Mittleman,
& Becker, 1985; Finkelhor, 1984).

9
Although the results suggest a difference in cognition, it is unclear if these cognitions or
schemata develop as post hoc rationalizations for their behavior or if they were formed before
the behavior. Abel et al. (1984) suggest that some offenders who find themselves sexually
attracted to children respond to this dilemma by modifying their cognitions in order to justify
their deviant behavior. These cognitive distortions may also be viewed in terms of Finkelhofs
(1984, 1986) theoretical model of child molestation which describes various disinhibitors such
as alcohol use or psychosis which enable child molesters to overcome their internal inhibitions.
Thus, cognitive distortions may be viewed as another type of internal disinhibitor (Hollin &
Howells, 1991).
The results from the Stermac and Segal (1989) study suggest that the child molesters
were less sensitive or responsive to the child's cues than were the controls, suggesting a lack of
empathy. As the authors indicated, child molesters perceptions may be modified only in
response to a clearly negative and unambiguous response such as the child's crying. It is
important to note that the vignettes in this study were written and did not require the subjects
to "read" or interpret others' cues. It is likely that in real life situations, the child's cues would
probably not be as clearly portrayed as they are in a written vignette. Since child molesters
tend to interpret ambiguous responses on the child's part as positive within the structure of a
simple written vignette, they may also have difficulty reading others' ambiguous nonverbal cues
in real life situations. It may be that child molesters have difficulty accurately reading others'
cues and that this deficiency in information processing contributes to permissive cognitions.

10
Subtypes of Child Molesters
Nonfamilial Versus Incestuous Child Molesters
Unfortunately, many of the studies cited above did not subdivide their samples of
child molesters based on gender of the victim or relationship to the victim. This tendency
to lump many types of sex offenders into one general category has been cited as one of the
potential reasons for inconsistent results (Howells, 1981). A variety of studies over the
years have demonstrated the need to differentiate between heterosexual and homosexual
offenders, and between nonfamilial and familial offenders. Howells (1981) offers the
conceptualization that child molesters fall on a continuum, ranging from pedophiles with a
primary sexual preference for children, to persons with a primary sexual preference for
adults. Most reseachers in the area agree that not all child molesters are pedophiles.
Thus, some theorists propose a distinction between molestation based on situational
factors versus molestation based on sexual preference. For example, Pawlak, Boulet, and
Bradford (1991) suggest that some child molestation may be related to situational factors
interacting with various types of psychopathology.
Many theorists view incestuous offenders as situational offenders and nonfamilial
offenders as more sexual preference offenders (Pawlak, Boulet, & Bradford, 1991). Thus,
incestuous offenders deviant sexual behavior is thought to be less related to a primary
sexual interest in children and more related to taking advantage of easy sexual access
Consistent with this view, incestuous offenders showed more age appropriate preferences
and recidivated at lower rates than nonfamilial child molesters (Marshall, Barbaree, &

11
Christophe, 1986). Abel et al. (1981) found, however, that incestuous offenders showed
sexual arousal to stimuli depicting sexual contact with children.
Pawlak, Boulet, and Bradford (1991) found differences between incestuous and
nonfamilial offenders on three of the subscales on the Derogatis Sexual Functioning Inventory
Incestuous offenders scored higher than nonfamilial offenders on sexual experience and sexual
satisfaction. The authors note, however, that this difference is difficult to interpret since these
subscale scores are based on factors which may be largely affected by involvement with a
regular sexual partner, and unfortunately, marital status differed greatly across the two groups.
As expected, incestuous offenders tended to be married while the majority of nonfamilial
offenders were single. In addition, nonfamilial offenders scored higher than incestuous
offenders on sexual fantasy, meaning that the nonfamilial offenders endorsed more items
depicting common sexual themes. This higher score on sexual fantasy may reflect a greater
need among nonfamilial offenders to resort to fantasy since their sexual experience and sexual
satisfaction are lacking. Although the two groups differed on these three subscales, the subtest
scores overall did not effectively discriminate between the two groups.
Marshall, Barbaree, and Christophe (1986) found differences between incestuous
and nonfamilial offenders in their erectile responses to sexual stimuli. Compared to
incestuous offenders and nonoffenders, the nonfamilial offenders showed greater arousal
to the stimuli involving children. Incestuous offenders, on the other hand, showed less
arousal to adult female stimuli than the other two groups. In a later study, Barbaree and
Marshall (1989) found that incestuous offenders showed either a non-discriminating sexual
arousal profile or an adult preference pattern while the nonfamilial group was more

12
heterogeneous, displaying a variety of profiles ranging from nondiscriminating, to child or
adult preferences. Consistent with these results, a recent study found that incestuous
offenders reported less sexual interest in children than did nonfamilial offenders on the Pictorial
Sexual Interest Card Sort (Haywood and Grossman, 1994).
Differences have also been found between the MMPI profiles of incestuous versus
nonfamilial offenders (Erickson et al., 1987). In general, the incestuous offenders showed a
lesser degree of disturbance than the nonfamilial offenders. Among offenders who were the
biological fathers, 4-3 was the most common profile; however, this code type was equally
prevalent among the control groups. The authors suggest that the 4-3/3-4 profile describes
"chronic anger, overcontrolled hostility, marital discord, and passive-aggressive personalities,"
with the 4-3 characterized by more overt acting-out while the 3-4 is characterized by more
passivity (p.568).
Among incestuous stepfathers, 4-7/7-4 was the most common profile. The authors
associate this profile with "insensitivity to others, brooding interposed with episodic acting-out,
and alcoholism" (p.568). For the nonfamilial offenders, the 4-2/2-4 and 4-8/8-4 profile was
more common than among the incestous offenders. The 2-point peak code for the mean
profile of the child molesters was a 4-2. This same code has been found in several other studies
and is considered to be consistent with the clinical profile portraying child molesters as passive,
dependent, socially uncomfortable, and impulsive (Erickson et al., 1987).
Although these studies have yielded somewhat inconsistent results, the majority of
the studies have found that incestuous offenders appear more similar to controls than do

nonfamilial offenders. Thus, one might expect incestuous offenders to show fewer
cognitive distortions than their nonfamilial counterparts regarding sexual contact with
children.
Admitters Versus Deniers
Another distinction among child molesters which is often overlooked in the
research is whether or not the alleged offender admits to having committed the crime.
Many studies have focused on those who admit to their crime, however, the majority of
sex offenders deny culpability (Salter, 1988; French, 1988; Rogers & Dickey, 1991). Thus,
the sample of offenders being studied is not representative of the majority of offenders. In
addition, studies which combine admitters and deniers may yield confusing results since
several recent studies have shown that these two groups differ in significant ways (Becker,
Kaplan, & Tenke, 1992; Haywood, Grossman, & Hardy, 1993; Haywood & Grossman,
1994).
Becker, Kaplan, and Tenke (1992) found significant differences in the erectile
response profiles of admitters and deniers. They found that a disproportionate number of
deniers failed to show a significant sexual response to cues depicting any age category
(nonresponders). While only 18% of the admitters were classified as nonresponders, 58%
of the deniers fell in this category Although the deniers tended to be less responsive in
general, they were maximally responsive to child cues. Admitters, in contrast, were more
responsive to peer cues, but also responded to child and adult cues. Forty-two percent of
the admitters showed peer or adult profiles while only 11% of the deniers were in these
categories. Deniers were significantly less responsive to adult cues than admitters, but

14
were not less responsive to child cues. Thus, deniers tend to show a different pattern of
sexual responsivity, suggesting that they may constitute a clinically distinct subgroup.
Given these differences, deniers and admitters may also show different response patterns
on other variables such as on cognitions and social perception. It is interesting to note that
the adolescent sex offenders in the Becker et al. (1992) study showed the greatest sexual
response to age-appropriate female cues, supporting the hypothesis that child molesters would
prefer to have a sexual relationship with a peer, but may turn to younger females because of
difficulty with initiating or maintaining a sexual relationship with an adult female. This finding
is consistent with social skills research and anecdotal information on child molesters (Marshall,
Laws, & Barbaree, 1990) which suggests that sexual assaulters have difficulty forming and
maintaining healthy intimate relationships with women because of their lack of social
competence.
Haywood, Grossman, and Hardy (1993) also found significant differences between
admitters and deniers. Deniers showed more minimization than admitters on the Fake-
Good Scale on the Sixteen Personality Factors Inventory while admitters showed more
exaggeration of problems on the Fake-Bad scale. Both scales were able to correctly
discriminate between admitters and deniers. Deniers were more likely to be defensive
about undesirable personality traits than were admitters. Interestingly, the personality
factor which was most highly correlated with the Fake-Good scale was emotional
sensitivity. This scale is associated with "affective maturity, lack of anxiety, and ability to

15
deal with frustration (p. 187)." Based on this finding, one might expect deniers to score
higher on empathy in the current study, in an effort to present themselves as emotionally
sensitive.
In a more recent study, Haywood and Grossman (1994) found that deniers
reported significantly less sexual interest in children than did admitters or normal control
subjects. Deniers also reported more sexual interest in women than did admitters.
However, both admitters and deniers reported significantly less sexual interest in adult
women than the controls. Among the child molesters, low self-reported sexual interest in
children was significantly correlated with minimization of psychopathology on the F-K
index and the positive malingering scale. The authors conclude that, for the sex offender
population, subjective measures of sexual interest are "highly susceptible to distortion or
minimization of deviant sexual arousal (p.338).
Abused Versus Nonabused Subjects
In addition to differences between admitters and deniers, Becker et al. (1992) also
found differences between abused and nonabused groups. Subjects who reported a history
of physical or sexual abuse were more sexually responsive in general than subjects with no
reported history Abused subjects also showed a greater proportion of nondiscriminator
profiles. These findings suggest the need to differentiate child molesters based on abuse
history as well as by denial.
Overall, there has been little research in the area of cognitive factors in child
molestation, with only one empirically sound study to date (Stermac & Segal, 1989). Given

16
this lack of research, the role of these cognitions remains unclear, particularly in regards to how
they interact with other variables to facilitate the commission of the offense (Quinsey, 1986).
Other Variables Related to Decoding Skill and Cognitions
Repression
Prior to the past decade, child molesters denial and defensiveness was viewed as
related to massive repression and unconscious denial (Rogers & Dickey, 1991). Rogers &
Dickey (1991) suggest that the pathogenic model was largely discarded in the past decade,
however, as a result of growing disenchanment with psychoanalytic theory and its
problems with clinical efficacy rather than on the basis of research findings. The results of
the study by Becker, Kaplan, and Tenke (1992), however, raise the possibility that
repression may play a role in child molesters denial since the deniers were sexually
nonresponsive to all sexual cues presented Although it is possible that the deniers were
able to consciously control their sexual arousal, repression might better account for these
deniers' ability to suppress their sexual impulses to prevent them from physiologically
responding to sexual cues. That is, the denial may involve more than a conscious attempt
to deny the crime in order to avoid perceived negative legal consequences; instead, or in
addition, it may involve an unconscious defense mechanism employed as a means of
controlling anxiety and guilt.
People who use repression as a primary defense mechanism often "distort their own
self-evaluative responses as a way to make themselves more comfortable, to block perceptions
of behavioral outcomes, and to avoid disturbing or uncomfortable implications in a variety of
situations (Dahlstrom, p. 104)." Thus, repression might help the child molester deny the

17
seriousness of his offense, and may facilitate the development and maintenance of cognitive
distortions to rationalize or justify the behavior. Assuming that deniers would show more
repression than admitters, one might expect them to show more cognitive distortions regarding
sexual contact with children, however, as several studies suggest (Becker et al., 1992;
Haywood et al., 1993; Haywood & Grossman, 1994), deniers appear to be effective at
suppressing both deviant sexual responses as well as deviant thoughts. Thus, deniers would be
expected to have more cognitive distortions, but would be less likely to show these distortions,
especially on a face valid measure.
In sum, since some form of denial is common among most child molesters, it might be
expected that child molesters as a group would show more repression than non-child
molesters. The subgroup of deniers might also be expected to show more repression than child
molesters who admit to the offense.
Empathy
Like repression, empathy is another variable which might be related to decoding
skill and cognitive distortions among child molesters. Clinicians have long noted that sex
offenders show little empathy for their victims. Many authors have also suggested,
however, that this deficit in empathy may be generalized in the sense that it applies toward
all people, rather than applying solely to the victim (Marshall, Jones, Hudson, &
McDonald, 1993). The few studies that have measured empathy among sex offenders
have focused on this concept of generalized empathy, and have yielded inconsistent
results.

18
Two studies reported a generalized deficit in empathy among child molesters. One
study reported that 35.8% of child molesters lacked empathy, however, the method for
assessing empathy was unclear (Abel, Mittleman, & Becker, 1985). Another study
reported significant differences between sex offenders and community subjects using
Hogans Empathy Scale (Rice, Chaplin, Harris, & Coutts, 1990; as referenced in Marshall
et al., 1993). Marshall et al. (1993) suggest that the samples used in these studies may not
be representative of the general sex offender population, as they consisted of a more
deviant and chronic group of sex offenders than is typically found in a general prison
population or outpatient clinic. Marshall et al. (1993) also point out the importance of the
method used to assess empathy. A variety of assessment instruments measure different and
limited aspects of empathy They cite a study conducted by Seto (1992) which reported
that incarcerated rapists showed no empathy deficiency on the Mehrabian and Epstein
(1972) measure of empathy, but did show a deficit on Hogans scale.
In their 1993 study, Marshall et al. used the Interpersonal Reactivity Index (Davis,
1983), citing it as the best available measure for assessing the complex nature of empathy.
They found no deficits in empathy among incarcerated child molesters when compared
with controls. In contrast, outpatient child molesters did show a deficiency in empathy.
More specifically, child molesters scored significantly lower on the total IR1 score and on
the Fantasy subscale. Unfortunately, the study did not differentiate between familial and
nonfamilial offenders or between homosexual and heterosexual offenders, thus introducing
a potentially confounding factor. Nonetheless, the authors suggest that the differences
among the child molesters may have resulted from motivation among the incarcerated

19
offenders to present themselves in a desirable light in order to facilitate early release from
jail. The authors conclude, however, that nonspecific empathy does not appear to be child
molesters' main problem. Rather, they recommend focusing on child molesters' empathy
toward specific targets such as their victims. Finkelhor & Lewis (1987) also suggest
exploration of this more specific type of empathy since otherwise-empathic men
sometimes sexually exploit children, (p.76)
Given the inconsistent results in the literature, the present study examined
generalized empathy among child molesters. Specific empathy for children was not
assessed since an assessment instrument of this type could not be found. Child molesters
hypothesized repression was thought to be related to a lack of empathy since repression
might allow the offender to block out the victim's response, thus resulting in impaired
empathy Lack of empathy might then facilitate the commission of the offense since
empathy is thought to play an important function in the inhibition of aggressive or
antisocial acts toward others, with vicarious experience of the victim's pain serving as a
deterrent (Miller & Eisenberg, 1988). Miller and Eisenberg (1988) conducted a meta
analysis of studies addressing this issue and found empathy to be negatively related to
aggression and antisocial, externalizing behaviors.
Empathy has been described in the clinical literature as both a cognitive and affective
response (Miller & Eisenberg, 1988). Empathy has been defined as the ability to take the
viewpoint of the other person and to respond affectively to the other's perceived emotional
experience (Lorr, Youniss, & Stefic, 1991). Since empathy involves achieving a cognitive
understanding of the other person's feelings, a lack of empathy might manifest itself in cognitive

20
distortions about the other person's feelings. Thus, the cognitive distortions seen in child
molesters could be related to a lack of empathy as well as repression, both of which could be
related to poor decoding skills.
Child molesters' perceived lack of empathy could also be related to their reported
histories of abuse. Miller and Eisenberg (1988) concluded from their meta-analysis that abused
subjects tend to have lower levels of empathy than nonabused subjects. Abused children are
thought to have little opportunity to learn to identify others' affective states or experience
empathic responding themselves since their needs and feelings are not recognized or
appropriately responded to in interactions with their abusive parents. Thus, abused children
tend to lack models of appropriate behavioral and emotional responding to others in need, and
as a result, show deficiencies in empathy which remain throughout adulthood.
In sum, child molesters' perceived lack of empathy was thought to be related to
their history of childhood abuse, repression, and poor decoding ability. In addition, lack
of empathy may play a role in their cognitive distortions regarding sexual contact with
children.
Rationale for the Proposed Study
This study focused on two aspects of child molesters' social information processing:
decoding skill and cognitive distortions. The literature has indicated that child molesters are
socially deficient, however, the nature of this deficiency is unclear. That is, this deficiency may
result from any number of problems in their cognitive operations, including deficiencies in their
decoding skill. Since the first step in exploring deficiencies in social information processing

21
begins with an examination of decoding skill (McFall, 1990), the present study examined this
variable among child molesters.
Given child molesters' low social competence and the recent finding of a decoding
deficiency found among other sex offenders (Lipton, McDonel, & McFall, 1987), it was
expected that child molesters would also show a deficiency in decoding skill when compared to
non-sex-offending control groups. In addition to differences between child molesters and
controls, differences between admitters and deniers were examined since a preliminary finding
suggested that deniers may represent a distinct clinical subgroup (Becker et al., 1992). In
addition to examining nonverbal decoding, this study examined child molesters' performance on
a more complex social perception task which involves the next step in McFall's (1990) social
information processing model decision skills. Child molesters were expected to perform as
poorly on this measure as on the decoding task since the social perception task involves
decoding as well as decision skills If the child molesters did not show a deficiency in decoding
skills, then this social perception task would allow for an examination of child molesters at the
next level of social information processing decision skills.
Since child molesters were expected to show deficits in decoding skill, they were also
expected to show deficits in empathy since empathy requires good decoding skills. This notion
of a deficit in empathy is consistent with clinical observation of child molesters, and has
received some empirical support, although the research in this area has yielded Inconsistent
findings. Thus, another reason for examining empathy among this group stems from the
inconsistent findings in the literature. Unlike previous studies, the current investigation used a
more appropriate measure of empathy which assesses a multifaceted notion of empathy, and in

22
addition, examined this variable among subgroups of child molesters familial versus
nonfamilial and admitters versus deniers.
The second aspect of the social information processing examined was child molesters'
cognitions regarding sexual contact. Part of this investigation involved an attempt to replicate
Stermac and Segal's (1989) results, with the modification that child molesters were subdivided
into appropriate subgroups Based on Stermac & Segals (1989) results, it was expected that
child molesters as a group would ascribe significantly more benefit to the child than control
groups when the childs response was positive or ambiguous, and would show no difference
from controls when the childs response was clearly negative. The child molesters were also
expected to attribute more responsibility and desire to the child and less overall responsibility to
the adult.
In addition, child molesters were expected to endorse more items on the Cognition
Questionnaire which are indicative of permissive beliefs and attitudes toward sexual contact
with children. Within the child molester group, deniers were expected to endorse fewer
permissive cognitions than admitters on both the vignettes and the Cognition Questionnaire.
This was expected since deniers show the most extreme form of denial and thus are not likely
to admit to cognitive distortions involving some acknowledgement of a sexual interest in
children. In addition, these measures of cognitive distortions tend to be face valid and thus
responses to these measures were expected to be highly influenced by social desirability.
In addition to replicating Stermac and Segal's (1989) results, this study investigated
child molesters' cognitions regarding sexual contact with adults. In contrast to their expected
distorted cognitions regarding sexual contact with children, child molesters were expected to

23
show no differences on these same variables relating to sexual contact with an adult. That is,
child molesters cognitive distortions relating to sexual contact were hypothesized to be largely
limited to interactions with children, however, child molesters were expected to rate more
anxiety than controls in a sexual situation with an adult female. This rating of anxiety was
expected to reflect child molesters lack of confidence and diflBculty with adult heterosexual
relationships. No differences were expected between admitters and deniers on the adult
vignettes.
Social desirability was also examined since denial among sex offenders is thought to be
at least partially related to a desire to present oneself in a positive light. Deniers were expected
to show more social desirability than admitters, while child molesters as a group were not
expected to differ significantly from controls. Repression was also examined since denial was
also thought to be partly related to unconscious defense mechanisms Thus, deniers were
expected to show greater repression than admitters Repression was thought to contribute to
permissive cognitions and was expected to negatively correlate with level of empathy since
repression might help molesters distance themselves from their sexual offending behavior, thus
reducing their ability to empathize with their victims' emotional reaction.
Hypotheses
This study examined decoding skill and cognitive factors among incarcerated heterosexual
child molesters (admitters and deniers), non-sex offending inmates, and community controls. It
was expected that:
(1) Child molesters would score lower than the two control groups on the decoding and social
perception tasks, with no differences expected between admitters and deniers.

24
(2) Child molesters would show distorted cognitions regarding sexual contact with a female
child as measured by the vignettes and Cognition Questionnaire. More specifically, child
molesters would ascribe more benefit, desire, and responsibility to the child, and less
responsibility and punishment for the man Nonfamilial sex offenders would also show more
cognitive distortions than incestuous offenders.
(3) Deniers would show fewer cognitive distortions than admitters on both measures regarding
sexual contact with children.
(4) On the adult vignettes, child molesters would not show cognitive distortions on the same
variables as they would on the child vignettes. However, child molesters would rate more
anxiety than controls regarding adult sexual interaction with an adult female.
(5) Child molesters would demonstrate less empathy than controls; deniers would show less
empathy than admitters. Empathy scores would correlate positively with decoding score and
negatively with repression.
(6) Child molesters would endorse more symptoms of repression than controls. Deniers
would show more repression than admitters.
Alternative Hypotheses
Child molesters' cognitive distortions could be related to sexual contact in general,
rather than specific to sexual contact with children. In that case, child molesters would be
expected to show the same cognitive distortions on the vignettes presenting sexual contact
with an adult female as in the vignettes presenting sexual contact with a female child.
Deniers may not show greater repression, but may show more social desirability. Thus,
denial might be more related to perceived demands for social desirability and less related to

25
unconscious defense mechanisms In this case, deniers would be expected to score significantly
higher on social desirability than the other groups, but not higher on repression. Another
possible alternative is that the child molesters may not be deficient on general nonverbal
decoding skill, but may show a deficiency on the more complex social perception task. Thus,
their deficiency may be more related to difficulties in decision skills than in decoding skills.

CHAPTER 2
METHODS
Subjects
There was a total of 214 subjects that completed the study. Subjects were divided into
six groups: nonfamilial heterosexual male child molesters [admitters (n=31) and deniers
(n=32)], incestuous heterosexual male child molesters (admitters (n=26) and deniers (n=34)],
nonsex offending male inmates (n=60), and male community controls (n=31). All subjects
were at least 18 years of age. Subjects were recruited from the following correctional facilities
in Florida: Union Correctional Institution, New River Correctional Institution, and Marion
Correctional Institution.
In order to qualify for the study, subjects in the child molester groups must have been
convicted of a sexual offense against a female who was under the age of 12. They could also
not report current or prior participation in an inpatient sex offender treatment program since
differential treatment histories could serve as a confounding variable. The non-sex offending
inmate group was selected on the basis of having no prior record of sex offenses according to
self-report and their prison files. The community control group was recruited through local
advertisements posted in areas designed to attract a low socioeconomic group. In addition,
subjects were solicited at local construction companies and various volunteer organizations.
Child molesters were divided into admitters and deniers based on their written self-
report to the question Did you commit the crime(s) that you are charged with? If not, please
26

27
explain why you disagree with the charge(s). Their admission or denial was also assessed
during a semi-structured interview in which they were asked to describe their charge and
whether or not they committed the crime.
Child molesters were also subdivided into incestuous and nonfamilial offender groups
based on information in their prison file describing their relationship to the victim. Since this
information was not always available in the prison records, this information was also gathered
during an individual interview in which the inmate was asked to describe the nature of his
relationship to the victim. Sex offenders who molested their daughters or stepdaughters were
considered to be incestuous offenders, as well as those molesting any female child living in their
home under their custody or authority. Thus, if the molester was involved in some kind of
parental relationship with the victim, as in the case of a live-in boyfriend who molests his
girlfriends children, then he was considered to be an incestuous or familial offender. This
broad definition of incest was used since it is the dynamics of the relationship that are
considered to be most important, and since most studies on incest have used this broader
definition (Marshall et al., 1986).
Research Design
In order to test the hypotheses, the six subject groups were administered a battery
of tests measuring the following variables: cognitive distortions regarding sexual contact
with children, decoding ability, empathy, repression, and social desirability. Background
variables included age, socioeconomic status, intelligence, and history of abuse. Group
served as the main independent variable with six levels (nonfamilial admitters, nonfamilial
deniers, incestuous admitters, incestuous deniers, community controls, and inmate

controls). Within the six levels, comparisons were made between child molesters and
controls, nonfamilial and incestuous offenders, and admitters and deniers.
28
Group served as the sole independent variable in the analysis of all of the
dependent variables except for the cognitive distortions. This dependent variable was
assessed using clinical vignettes which contained two independent variables which varied
across the vignettes Level of Sexual Contact (4 levels) and Females Response (3 levels).
Thus, a 6x4x3 design was used in the analysis of the vignette data, with Group as a
between subjects variable and Level of Sexual Contact and Females Response as within
subjects variables. The following section lists each of the dependent variables and the
specific measures used.
Dependent variables
Measures of Cognition
Clinical vignettes
Cognitions regarding sexual contact were measured using the clinical vignettes from
Stermac and Segal's (1989) study. There were twelve vignettes describing sexual contact
between a male adult and seven year old child. The vignettes were designed by Stermac and
Segal (1989) based on documented reports of actual incidents between children and adults.
There are two independent factors in each vignette the degree of sexual contact and the
child's response to the contact. The levels of sexual contact include: 1) touching only, 2)
rubbing genital area over clothing (fondling), 3) fondling genitals under clothing (no clothes),
and 4) genital contact with ejaculation. The levels of the child's response include: 1) smiling,
2) passive/no response, and 3) crying with resistance.

29
The subjects were asked to read each vignette and respond to six questions using a five
point Likert-type rating scale ranging from 1 (not at all) to 5 (definitely). The questions were
designed to reflect child molesters' central cognitive beliefs about adult sexual contact with
children, as described in previous clinical investigations. Subjects were asked the following
questions (the variable name for each question is in parentheses following the question): 1) Do
you think the child enjoyed what happened? (Child Enjoyment) 2) Do you think the child
wanted this to happen? (Child Desire) 3) Do you think the child could benefit from this
experience? (Child Benefit) 4) Do you think the child was responsible for what happened?
(Child Responsibility) 5) Do you think the child could be harmed by this experience? (Child
Harm) 6) Do you think the man was responsible for what happened? (Man Responsibility)
In order to assure face validity, Stermac and Segal (1989) asked subjects in a pilot
study to rank order the vignettes in terms of degree of sexual contact between the child and
adult. They were also given a multiple choice format and were asked to determine the child's
response. The authors reported complete agreement among subjects on these tasks.
Stermac and Segal (1989) conducted Pearson correlation matrices for each of the
twelve vignettes in order to examine correlations among the variables. They found that child
benefit, harm, and enjoyment were highly correlated across the levels of sexual contact and
childs response. Child benefit and enjoyment were negatively correlated with child harm, with
correlation coefficients ranging from .32 to .65. Child responsibility and childs desire were
also found to be highly correlated, with correlation coefficients ranging from .37 to .71. For
analysis purposes, the authors combined child benefit, enjoyment, and harm into a composite
variable called benefit to child, and combined child responsibility and desire into child

30
complicity. Test-retest reliability data was not available for this measure. Thus, the present
study included this analysis (see the Results section).
In addition to the twelve vignettes with a child, twelve vignettes with an adult were
designed for the current study. This set of vignettes was designed to assess attitudes toward
aggressive sexual contact with an adult female. The same vignette scenes and questions were
used as in the child vignettes, although the female portrayed was a 30-year old female. In
addition to the above questions asked regarding the woman (instead of the child), the following
questions were added (the variable names are in parentheses): 1) Do you think the woman was
sexually attracted to the man? (Womans Sexual Attraction) 2) Do you think the man felt
anxious? (Mans Anxiety) 3) Do you think the man felt confident that the woman was attracted
to him? (Mans Confidence) These questions were added in order to assess the subject's
perception of the man's anxiety and confidence, as well as the subject's perception of the
woman's degree of sexual attraction to the man. Since child molesters are thought to have
difficulty with relationships with adult women, they were expected to score differently from
controls on these variables.
Cognition questionnaire
The Cognition Questionnaire (Abel, Becker, Cunningham-Rathner, Rouleau, et al.
1984) was also administered as a second measure of cognitive distortions regarding sexual
contact with children. This instrument contains 29 items and was developed for use with adult
child molesters. The subject reads statements which reflect values about adult sexual contact
with children and selects a response ranging from 1-5 (strongly agree strongly disagree).
Sample items on the Cognition Questionnaire include: 1) A child 13 or younger can make

31
her(his) own decision as to whether she (he) wants to have sex with an adult or not, and 2) Sex
between a 13-year old (or younger) child and an adult causes the child no emotional problems.
Test-retest reliability for this questionnaire has been reported as .76 over a three week interval
(Abel, Becker, Cunningham-Rathner, Rouleau, et al. 1984). Six subscales were identified with
coefficient alphas ranging from .59 to .71. Scores on the questionnaire distinguished child
molesters from normals, demonstrating discriminant validity.
Measures of Decoding/Social Perception
Profile of nonverbal sensitivity
The Profile of Nonverbal Sensitivity (PONS) (Rosenthal et al., 1979) was used as a
general measure of nonverbal decoding ability. This widely used measure has the benefit of
distinguishing between several channels of nonverbal communication and various types of cues.
The PONS is a 47-minute videotape which is comprised of220 two second film clips in which
eleven nonverbal channels are isolated: three "pure" visual (face, body, figure) and two ''pure"
auditory channels (random spliced, content free), with six additional paired combinations In
each film segment, one of twenty affectively charged interpersonal situations is portrayed by a
woman who is not an actress. The twenty situations can be categorized into four affective
quadrants: positive-dominant, positive-submissive, negative-dominant, and negative-
submissive. Subjects are asked to choose which of two statements best describes the behavior
in each film clip. For example, on the first item the subject must decide whether the woman in
the film clip is expressing jealous anger or talking to a lost child. In another item the
respondent must choose between leaving on a trip or saying a prayer. The test was
designed to measure one's ability to read" and interpret nonverbal cues.

32
The PONS was standardized on 492 subjects. Test-retest reliabilities average .81 with
an interval of 6 to 8 weeks between testing, although they vary over the three dimensions of
voice tone (lowest reliability), facial expressions, and body movements (Rosenthal et al., 1979).
The internal consistency coefficients were reported as 86 for total score and as follows for the
affective quadrants: positive-submissive (.78), positive-dominant (.84), negative-submissive
(.87), and negative-dominant (.87).
Principal components analysis on the eleven channel accuracy scores yielded four
factors measuring: 1 )decoding of facial cues, 2) decoding of randomized spliced cues, 3)
decoding of content-filtered cues, and 4) decoding of cues from the body without the face.
Principal components analysis was also conducted for the negative and positive affect scenes,
dominant and submissive channels, and the twenty scenes. The factor analysis results of the
twenty scenes suggests that decoding skill on the PONS can be subdivided into six skills
involving decoding cues of interpersonal control, cues of contrition, cues of strong but
controlled affects, cues of separation, cues of dependence, and cues associated with recitation.
The results of these factor analyses provide support for the various channels and affective
quadrants assessed by the PONS.
Interpersonal perception task
General social perception was measured using the Interpersonal Perception Task (IPT)
(Costanzo & Archer, 1989) which "is designed to assess accuracy in interpreting the expressive
behavior of others" (p.231). The IPT is a 38-minute videotape depicting naturalistic full-
channel sequences of behavior. There are 30 scenes which show one to four people. A
question appears on the screen before each scene, requiring the viewer to answer the question

33
based on the scene that follows. During the six second interval following the scene, the viewer
chooses the correct answer from a multiple choice format. For example, the first scene shows
an adult male and female having a conversation with two seven year-old children The viewer is
asked "Who is the child of the two adults?" Since the questions are based on actual,
naturalistic scenes there is an objective criterion of accurate judgment for each item.
The IPT was standardized on 438 college students. The accuracy rates for the scenes
are reported as comparable to those for other similar measures. Accuracy rates are listed for
each scene. Test-retest reliability using 46 subjects after a five week break was .70. Internal
consistency was .52 using the KR-20, which the author suggests may reflect the diversity and
relatively small number of scenes. Subjects who scored in the high social sensitivity group
based on 17 peer ratings scored significantly higher in the IPT than subjects in the low social
sensitivity group, demonstrating construct validity. In addition, the correlation between IPT
scores and social sensitivity scores based on peer ratings was r=.48. The authors also report
significant positive correlations between the IPT and two other standardized measures of ability
to interpret the expressive behavior of others.
Measure of Empathy
Empathy was measured using the Interpersonal Reactivity Index (IRI, Davis, 1979)
which Marshall et al. (1993) cited as the best available measure for assessing the complex
nature of empathy. The IRI is a 28-item self-report questionnaire with four 7-item subscales
measuring different aspects of empathy. The four subscales are 1) perspective-taking, 2)
fantasy 3) empathic concern, and 4) personal distress. The Perspective-Taking (PT) scale
assesses the tendency to adopt others' psychological points of view. The Fantasy scale (FS)

34
assesses the subject's tendency to "transpose themselves imaginatively into the feelings and
actions of fictitious characters in books, movies, and plays (Davis, p. 114)." The Empathic
Concern (EC) scale measures feelings of sympathy and concern for others, and the Personal
Distress (PD) scale measures the subject's feelings of personal anxiety in tense interpersonal
situations.
The IRI was developed from a pool of 50 items some of which were adapted from
other measures of empathy such as the Mehrabian & Epstein emotional empathy scale (1972).
New items were added to measure cognitive aspects of empathy such as the ability to adopt
different perspectives Factor analysis revealed four factors which now form the four
subscales. This factor structure has been shown to be stable when repeatedly administered to
different samples (Davis, 1983). The final 28-item version of the IRI was standardized on 579
male and 582 female college students. Internal reliabilities range from .71 to .77 and test-retest
reliabilities range from .62 to .71 (Davis, 1980).
Davis (1983) cites data supporting the validity of the four IRI subscales and his
multidimensional view of empathy The four subscales which represent different aspects of
empathy showed distinctive patterns of relationships with measures of social functioning, self
esteem, emotionality, sensitivity to others, and unidimensional measures of empathy The
Hogan Empathy Scale which is more cognitive in its approach showed highest correlations
(r=.40) with the cognitive PT scale, as expected The PT scale also showed the least
association (r=.20) of the four IRI scales with the Mehrabian and Epstein Emotional Empathy
Scale which taps emotional rather than cognitive empathy. In addition, the FS and EC scales
showed greater associations with the Emotional and Empathy Scale (r=.52 and .60,

35
respectively). Davis (personal communication, 1994) recommends examining each of the
subscale scores separately rather than using the total IRI score since each of the subscales
appears to measure a different aspect of empathy. Other researchers, however, argue that the
IRI total score may serve as an overall measure of empathy.
Measure of Repression
The Repression-Sensitization Scale (Bryne, 1961) was used to measure repression.
The R-S scale is composed of 127 items from six Minnesota Multiphasic Personality Inventory
(MMPI) scales, three measuring sensitization and three measuring repression. The range of
possible scores is 0 to 127; high scores indicate sensitization while low scores indicate
repression. The behavioral dimension of repression-sensitization is viewed as a continnum,
with the repressive extreme involving avoidance defenses such as denial and the sensitizing
extreme involving approach defenses such as intellectualization.
The R-S scale was standardized on 733 male and 571 female college students (Byrne,
Barry, & Nelson, 1963). Test-retest reliability over three months was .82 and split-half
reliability was .94. In terms of face validity, seven of 10 judges showed agreement on 90% of
the items when asked to respond as they thought a repressor would respond. R-S scores have
been found to be consistently related to self-report measures More specifically, scores on the
R-S scale were related to self-ideal discrepancy (r= 63) and negative self-description (r=.68),
supporting the notion that individuals who use repressive defense mechanisms are less likely to
verbalize negative self-descriptions than individuals with sensitizing defense mechanisms.
Correlations between the R-S scale and Marlowe-Crowne Social Desirability Scale
range from -.37 to -.49, suggesting that repression-sensitization is partially, but not entirely a

36
function of a socially desirable response style (Byrne, 1961). There is some evidence to
suggest some independence between repression-sensitization and social desirability (Byrne,
1961). In a study by Merbaum (1972), extreme sensitizers and repressers were asked to retake
the test three months after having initially completed the R-S scale. In this second
administration they were asked to respond in a way that would make them look as good as
possible. The scores of both groups shifted in the repressor direction, however, the sensitizers
still scored higher in the direction of sensitization than did repressers Thus, differences
between on repression-sensitization remained despite the instructions to adopt a socially
desirable response style
Measure of Social Desirability
The Marlowe-Crowne Social Desirability Scale (MCSDS) (Crowne & Marlowe, 1960)
was used to control for socially desirable response sets. The MCSDS is a 33-item scale which
was originally developed to discriminate the effects of item content from the subject's need to
present him or herself in a socially desirable light. As a result, the MCSDS focuses on
culturally sanctioned and approved behaviors that do not occur frequently. Sample items from
this questionnaire include: 1) I never hesitate to go out of my way to help someone in trouble,
and 2) I am always careful about my manner of dress. The internal consistency coefficient is
reported at .88, and the test-retest reliability is .89 for a sample of normal 39 college students
Normative data is presented for 28 males and 27 females (Crowne & Marlowe, 1960). The
MCSDS is highly correlated with the Edwards Social Desirability Scale ( 35) and the following
MMPI subscales: L (.54), K (.40), and F (.36). The MCSDS has been used in numerous
studies, including the Stermac and Segal (1989) study.

37
Self-Report of Abuse History
The Life Experiences Questionnaire (LEQ) developed by Bryer et a). (1987) was used
to assess abuse history. Bryer et al. (1987) administered the LEQ to 66 females upon
admission to a private psychiatric hospital. They found that the more severely disturbed
patients were more likely to report having been abused in childhood. A recent study showed
that female inpatients were more likely to report a history of abuse on the LEQ than during an
intake interview (Dill, D., Chu, J., Grob, M., & Eisen, S., 1991). There was low agreement
between the LEQ and intake format in regard to reported histories of physical abuse, k=.32.
The authors note that the low agreement was largely due to subjects who reported abuse on
the LEQ but not during the intake interview. 59% of subjects reported physical abuse on the
LEQ while only 26% reported this type of abuse in the interview Ninety-two percent of
subjects reporting abuse during the interview also reported abuse on the LEQ, however, 47%
of those who did not report abuse during the interview later reported abuse on the LEQ.
Agreement between interview and LEQ reports of sexual abuse were also fairly low, k=.44
52% of subjects reported a history of sexual abuse on the LEQ while only 35% reported this
history in the intake interview These results strongly support the need to assess abuse history
in a written format since many subjects may be reluctant to report abuse history during an
interview.
In the present study, physical abuse was defined as an affirmative response to the LEQ
question "were you ever hit really hard, kicked, punched, stabbed, or thrown down?
Similarly, sexual abuse was defined as an affirmative response to the question "were you ever
pressured against your will into forced contact with the sexual parts of your body or [his/her]

38
body?" A supplement to the questionnaire asked subjects for specifics concerning the
frequency and period of time during which the abuse occurred. In order to facilitate
comparisons with the Becker (1992) study, the variable of abuse encompassed both physical
and sexual abuse. Thus, within this study, the term a history of abuse refers to a self-
reported history of physical and/or sexual abuse.
Measure of Social Status
The Hollingshead Four Factor Index of Social Status (Hollingshead, 1975) was used
to estimate subjects' socioeconomic status. This index assesses social status based on
education, occupation, sex, and marital status. The occupation factor is based on the
approximately 450 occupational titles and codes of the 1970 United States census, and is
graded on a 9-point scale. The education factor is based on the number of years of schooling
and is scored on a 7-point scale. Scores from the occupation, marital status, and education
factors are combined to yield a social status score ranging from 8 to 66.
The Four Factor Index was found to be reliable and valid based on data derived from
the 1970 Census and the National Opinion Research Council. Social strata categories were
validated using factor analysis and criterion validity was established by comparing occupation
index scores with prestige scores developed by the National Opinion Research Center (r=.93).
Significant differences were found among these categories when mass communication data
were used as criteria of social behavior (Hollingshead, 1975). The Four Factor Index also
correlates highly with other measures of socioeconomic status: the Revised Duncan
Socioeconomic Index (r=.79) and the Siegel 1965 NORC Prestige Scale (r=.73).
Measure of Intelligence

39
The Vocabulary Scale of the Wechsler Adult Intelligence Scale-Revised was used as a
rough estimate of level of intellectual functioning. The Vocabulary Scale contains 35 items or
vocabulary words which the subject is asked to define. Responses are scored on a 0-2 point
scale. Split-half reliability for the Vocabulary Scale ranges from .94 to .96 across the 16-74
age range and the test-retest reliability is 93 for ages 25-34, and .91 for ages 45-54 (Wechsler,
1981). The Vocabulary score is highly correlated with the WAIS-R Full Scale score (r=.85)
and the Verbal Scale score (r=.90). The WAIS-R was standardized on a sample of 1,880
subjects stratified according to age, sex, race, geographic region, education, and occupation.
Summary of the Variables.
The background variables included age, socioeconomic status score on the
Hollingshead Four Factor Index of Social Status, vocabulary score on the WAIS-R, and self-
report of physical or sexual abuse as assessed by the Life Experiences Questionnaire The
dependent variables used to assess cognitive distortions included the total score on the
Cognition Questionnaire, and the set variables on the child and adult versions of the written
vignettes: 1) Females Enjoyment of the sexual interaction, 2) Females Desire, 3) Females
Benefit, 4) Females Responsibility, 5) Females Harm, and 6) the Mans Responsibility There
were two additional variables on the adult female vignettes including the Womans Sexual
Attraction to the man and the Mans Anxiety regarding the sexual interaction.
Decoding ability was assessed using the total scores from the Profile of Nonverbal
Sensitivity Test and the Interpersonal Perception Task. Other dependent variables included the
total scores on the Repression-Sensitization Scale and the Marlowe-Crowne Social Desirability

40
Scale, the empathy total score on the Interpersonal Reactivity Index, and the four subscale
scores measuring perspective-taking, fantasy, empathic concern, and personal distress.
Procedures
Prison files were reviewed in order to select subjects who met the criteria for the study.
All records of inmates with a child sex offense were reviewed. In order to generate a random
list for the control group, every fifth name was selected from the prison roster. These files
were reviewed and those inmates with no history of a sex offense were assigned to the nonsex
offending control group. Subjects selected for both groups were sent a letter which described
the nature of the study and listed the date and time of an informational meeting they could
attend to learn more about the study. Approximately ten subjects were scheduled for each
initial group meeting.
At the informational meeting, subjects were provided with a complete description of
the project. Those who volunteered were asked to complete the Introductory Questionnaire,
the Social Desirability Scale, and the Interpersonal Reactivity Index. Subjects were then
administered the IPT, and asked to complete one of the two sets of vignettes based on random
assignment. At the end of the first session, subjects were scheduled to return for an individual
session and the second group session within the next seven days.
In the individual session, subjects were administered a semistructured interview and
the Vocabulary Scale from the WAIS-R. In the second group session, subjects were
administered the R-S scale and the LEQ, followed by the PONS test. Subjects were then
asked to complete the second set of vignettes.

CHAPTER 3
RESULTS
Based on the literature, it was predicted that: 1) child molesters would score
lower than controls on decoding tasks; 2) child molesters would show more cognitive
distortions than controls regarding sexual contact with children, and nonfamilial offenders
would show more cognitive distortions than incestuous offenders; 3) deniers would show
fewer cognitive distortions than admitters, 4) child molesters would attribute more anxiety
to the man in the adult vignettes than would controls; 5) child molesters would show less
empathy than controls, and deniers would show less empathy than admitters; and 6) child
molesters would show more repression than controls, and deniers would show more
repression than admitters.
Background Variables
Background variables included age, socioeconomic status, vocabulary scores,
social desirability, and self-reported abuse history. In order to examine among-group
differences on these variables, several one-way ANOVAs were conducted with Group
(incestuous admitters, incestuous deniers, nonfamilial admitters, nonfamilial deniers, non
sex offending inmate controls, community controls) as the between subjects factor and
age, socioeconomic status, and vocabulary scores as the dependent measures. Several
ANOVAs were conducted rather than a MANOVA since the primary interest was to
examine group differences on each variable. There were no significant differences on
41

42
socioeconomic status, F(5,208)= 576, p=.718 or vocabulary score F(5,208)=1.942,
p=.089. There was a significant difference, however, on age, F(5,208)=8.518, p= 000.
The mean ages of the groups were 29.86 for the inmate control group, 34.74 for the
community control group, 38.89 for the incestuous admitters, 38.32 for the incestuous
deniers, 40.45 for the nonfamilial admitters, and 39.72 for the nonfamilial deniers. A
Scheffe post-hoc analysis showed that the inmate control group was significantly younger
than each of the four child molester groups. There were no significant differences among
the other groups. A contrast between child molesters and controls revealed that the child
molesters were significantly older than controls, F(l,208)=27.421, p=.000. Other studies
have also found that child molesters were significantly older than controls (Haywood &
Grossman, 1994). This age difference is not surprising since the general inmate
population tends to be younger than the sex offender population, by virtue of their shorter
prison sentences in Florida and sex offenders older age at arrest.
It was expected that a greater proportion of child molesters than controls would
report a history of physical and/or sexual abuse. In order to examine reported abuse
history across groups, a chi-square analysis was performed. There was a significant
difference among the six groups, X2(6)=13.626, p=.018. The following percentage of
each group reported a history of abuse: Incestuous Admitters (61.54%), Nonfamilial
Admitters (54.84%), Incestuous Deniers (47.06%), Nonfamilial Deniers (31.25%),
Controls (30%), and Community Controls (29%) (see Figure 1). One-tailed planned
comparisons showed that the child molesters reported significantly more abuse than
controls as expected, Z=2.862, p=.002, and admitters reported significantly more abuse

43
than deniers, Z=2.158, p=.015. The differences between incestuous and nonfamilial sex
offenders was not significant, Z=1.276, p=.202, two-tailed.
In order to examine differences between abused and nonabused subjects, several
one-way ANOVAs were conducted with abuse history as the between subjects factor and
total scores from the IPT, PONS, empathy measure, Cognition Questionnaire, Social
Desirability Scale, and Repression Scale as the dependent measures. Several ANOVAs
were used because group differences on each variables were of interest. There was a
significant difference between abused and nonabused subjects on repression,
F(l,208)=4.952, p=.027, indicating that those who denied a history of abuse scored higher
on repression than those reporting a history of abuse (see Table 1). When Group was
added as a second between subjects factor, however, the main effect of abuse history
became nonsignificant, F(l,198)=2.972, p=.086. There was also a significant main effect
of abuse history on social desirability, F(l,210)=8 491, p= 004, with those reporting a
history of abuse scoring significantly lower than nonabused subjects on social desirability
(see Table 1). When Group was added as the second between subjects factor, there was a
significant main effect of Group, F(5,200)=4.238, p=.001, and a significant Abuse X
Group interaction, F(5,200)=3.032, p=.012. Figure 2 shows that there was no consistent
pattern between abused and nonabused subjects across the six groups.
To examine the relationship between abuse history and admission/denial, an
ANOVA was conducted with Abuse and Admit as the two between subjects variables.
The community control subjects were not included in this analysis since the variable of
admission/denial was not relevant to them. There was a significant main effect of Abuse,

44
F(l,179)=5.266, p=,023, and a significant main effect of Admit, F(l,179)=7.905, p=.005,
with a nonsignificant Abuse X Admit interaction, F(l,179)=.447, p=.505 (see Table 2).
Figure 3 shows that subjects reporting no history of abuse scored significantly higher on
social desirability than those reporting abuse In addition, inmates who denied having
committed their crimes scored significantly higher on social desirability than those who
admitted to committing their crimes. Thus, deniers who did not report abuse showed the
highest social desirability while admitters who reported a history of abuse showed the
lowest social desirability.
No significant differences were expected between child molesters and controls on
social desirability. One outlier was identified by the studentized residual value and was
dropped from the analysis. As expected, a planned comparison revealed no significant
differences in social desirability between child molesters and controls, F(l,205)=.337,
p=.562. An ANOVA with the six groups, however, yielded a significant main effect for
Group F(5,205)=4.258, p=.001. As expected, deniers scored significantly higher than
admitters on social desirability, F(l,205)=16.769, p=.000 (see Table 2). There were no
significant differences between incestuous and nonfamilial sex offenders, F( 1,205)= 2.899,
p=.090. All possible pairwise comparisons using the Bonferroni correction showed that
incestuous deniers and nonfamilial deniers scored significantly higher than incestuous
admitters on social desirability, p=.003, p=.002, respectively.
Method of Analysis
The dependent variables were analyzed using ANCOVAs or MANCOVAs.
Decoding skill, empathy, repression, and cognitive distortions on the Cognition

45
Questionnaire were analyzed using separate analyses of variance. Group served as the
between subjects variable across analyses, and social desirability served as the covariate in
the analysis of the Cognition Questionnaire, empathy, and repression scores. The
assumptions for ANOVA were met for each of the dependent variables.
The vignette data was analyzed using principal components analysis followed by a
multivariate analysis of covariance For the univariate and multivariate analyses, planned
comparisons were conducted between child molesters and controls, admitters and deniers,
and between incestuous and nonfamilial sex offenders. Each of these comparisons was
considered to be its own separate family since each was thought to involve a conceptually
distinct comparison. Thus, the Bonferroni correction was not applied to these three
comparisons. The Bonferroni correction was used, however, when all possible pairwise
comparisons were conducted. All possible pairwise comparisons were examined since it
was expected that there might be differences among the six groups other than those
among the three families. More specific hypotheses than those regarding the three families
were not proposed since there has been little research on the differences among these
particular subgroups of child molesters. Thus, for example, differences between
incestuous deniers and nonfamilial admitters might be of interest, however, there was no
specific hypothesis regarding these two groups since there has been no research to date in
this area.
In addition to the planned comparisons and Bonferroni pairwise comparisons,
some specific post hoc comparisons were conducted between a particular group and the
five other groups combined. This contrast was conducted when one group seemed to

46
score significantly higher than all of the other groups, which was frequently the case with
the nonfamilial admitters. The Scheffe test was used to conduct these post hoc
comparisons.
Analysis of Decoding Skill
Child molesters were predicted to score lower on decoding measures than control
subjects. Contrary to expectations, there were no significant differences between child
molesters and controls on the total score of the Profile of Nonverbal Sensitivity Test
(PONS), F(2,203)=.018, p=892. There were also no significant differences between
admitters and deniers, or between incestuous and nonfamilial sex offenders. An ANOVA
conducted with the six groups also showed no significant difference among the groups
F(5,203)=.737, p=.597 on PONS total score (see Table 3). An ANOVA with the four
PONS subscale scores as the repeated measures revealed no significant differences among
groups, F(5,203)=789, p=.559. As expected, subjects showed significant differences
across the four subscales, F(3,609)=122.449, p=.000. It was predicted that the decoding
scores would be significantly correlated with the empathy scores. Contrary to
expectations, the correlation of the PONS total score with the IRI total score was not
significant, r= 039.
Consistent with the results on the PONS, there were no significant differences
between child molesters and controls, admitters and deniers, or between incestuous and
nonfamilial sex offenders on the Interpersonal Perception Task (IPT). An ANOVA also
showed no significant differences among the six groups on the IPT score, F(5,207)=.203,
p= 961 (see Table 3) or on the subjects estimation of their IPT score F(5,184)=.549,

47
p=.739. It was expected that empathy scores would be significantly positively correlated
with decoding scores. Consistent with the results from the PONS, the correlation of the
IPT score with the IRI total score was not significant, r=. 110.
Analysis of Cognitive Distortions
Cognition Questionnaire
It was predicted that child molesters would show more cognitive distortions than
controls on the Cognition Questionnaire which assesses ones attitude toward sexual
contact with children. Seven outliers were identified based on studentized residual values
and these cases were dropped from the analysis. A planned comparison revealed that child
molesters did endorse significantly more cognitive distortions about sexual contact with
children than did the control groups, F(l,188)=7.478, p=.007 (see Table 4). Lower scores
on this measure indicate more cognitive distortion. An ANOVA with the six groups
revealed a significant main effect for Group F(5,188)=2.862, p=.023 and social
desirability, F(l,188)=6.929, p=.009 (see Figure 4). It was also predicted that deniers
would show fewer cognitive distortions than admitters, however, planned comparisons
showed no significant differences between these two groups F(l,188)=.210, p=.648
Nonfamilial offenders were expected to show more cognitive distortions than incestuous
offenders. Consistent with this prediction, the difference between incestuous and
nonfamilial sex offenders was significant, F(l,188)=4.041, p=.046, with nonfamilial sex
offenders endorsing more cognitive distortions than incestuous offenders Pairwise

48
comparisons using the Bonferroni correction showed that the nonfamilial deniers endorsed
significantly more cognitive distortions than the community control group.
Clinical Vignettes
Child vignettes
Child molesters were expected to show more cognitive distortions on the child
vignettes than the controls, and admitters were expected to show more cognitive
distortions than the deniers Based on Stermac & Segals (1989) results, child molesters
were expected to ascribe more benefit to the child than control groups when the childs
response was positive or ambiguous, but to show no difference when the childs response
was clearly negative The variable of benefit was a composite of the variables benefit to
child, harm to child, and childs enjoyment. Compared to controls, child molesters were
also expected to attribute more responsibility and desire to the child and less responsibility
to the adult.
Test-retest reliability. The child vignettes were readministered to subjects after a 3
to 4 week interval. Scores on each of the nine variables were summed across the twelve
vignettes for each administration and Pearson correlations were calculated to measure
test-retest reliability from Trial 1 to Trial 2. The test-retest reliability for the variables is as
follows: Child Benefit (r=.83), Child Harm (r= 82), Child Enjoyment (r=.83), Child
Desire (r=.66), Mans Responsibility (r=.71), Treatment (r=,80), Other Men (r=.81), and
Punishment (r=.72). Analysis of the specific variables on each vignette also showed
strong correlations between the first and second administrations, demonstrating that
subjects tended to respond in a consistent manner on this measure over time.

49
Reduction of variables. In order to facilitate comparison with the results of
Stermac & Segal (1989), Pearson correlation matrices were conducted for each of the
twelve child vignettes to see which variables were highly correlated across levels of sexual
contact and childs response (see Table 5). This matrix shows that the nine variables are
highly correlated with each other. Stermac & Segal (1989) addressed this issue of
multicollinearity in their study by combining highly correlated variables into two
composite variables. They found that Benefit to Child, Harm to Child, and the Childs
Enjoyment were highly correlated. Child Benefit and Enjoyment were negatively
correlated with Child Harm, with correlation coefficients ranging from .32 to .65. They
combined these variables into a composite variable called Benefit to the Child.
In the present study, the negative correlation coefficients between Child Benefit
and Enjoyment and Child Harm ranged from .09 to .34. Child Benefit and Child
Enjoyment showed correlation coefficients ranging from 18 to .59. Since these
correlations were low and spanned a large range, these three variables were not combined
into a composite variable. When the variables were summed across levels of childs
response and sexual contact, Child Benefit correlated with Enjoyment at 60 and with
Harm at -.30. The latter two variables correlated -.35.
Stermac & Segal (1989) also formed a second composite variable called Child
Complicity from the variables of Child Responsibility and Child Desire. Across the levels
of sexual contact and childs response, these two variables showed correlations ranging
from .37 to .71. In the present study, these correlations ranged from .40 to .75, closely
replicating Stermac & Segals (1989) findings. Although we could have formed a

50
composite variable as did Stermac & Segal (1989), these two variables were only two of
the nine variables that were all highly correlated with each other. Thus, specifying these
two particular variables for combination did not seem to be statistically warranted.
Instead, a principal components analysis was conducted in order to reduce the number of
variables in a more statistically appropriate manner.
Principal components analysis of the child vignettes. In addition to reduction of
the number of variables, principal component analyses with varimax rotation were
conducted in order to examine the relationships among the variables. A principal
components analysis was conducted on the nine variables collapsed across levels of sexual
contact and childs response. The analysis yielded two factors, accounting for 38.16%
and 31.96% of the total variance. The variables with high loadings on Factor 1 included
child desire, child responsibility, child enjoyment, child benefit, and other men (see Table
6). Thus, this factor appears to measure the childs reaction and complicity in the sexual
contact. In contrast. Factor 2 measures the mans responsibility for the sexual contact as
well as the seriousness of the behavior. The variables with high loadings on this factor
included the mans responsibility, the harm to child, the mans need for treatment, and the
amount of punishment prescribed for the man.
Although the analysis of the nine variables summed across vignettes yielded two
meaningful factors, it was important to conduct principal components analyses for each of
the twelve vignettes to see if these two factors were stable within each of the vignettes. As
expected, the same factor structure was found in each of the twelve vignettes, indicating
that these same variables tend to cluster together within each vignette.

The principal components analysis was conducted to reduce the number of
variables in an effort to minimize the redundancy associated with results involving multiple
highly correlated dependent measures. Thus, the two factors were used as dependent
measures in analyses of variance. An ANOVA conducted with Group as the between
subjects variable and Factor 1, Child Complicity, as the dependent measure yielded a
significant main effect of group, F(5,202)= 3 .91, p=.002. Planned comparisons showed
that child molesters did not attribute more complicity to the child than controls, F( 1,202)=
.369, p=.544. A post-hoc comparison, however, showed that nonfamilial admitters, one
of the child molester groups, attributed significantly more complicity than all of the other
groups combined, F(l,202)= 16.835, which was significant using the Scheffe test. As
expected, nonfamilial sex offenders as a group attributed more complicity than incestuous
offenders, F(l,202)= 6.050, p=.015, however, much of this difference may be attributed to
the nonfamilial admitters. Similarly, the admitters scored higher on this factor than the
deniers, F(l,202)=3.638, p=,058, although much of this difference is also probably
attributable to the nonfamilial admitters.
An ANOVA with Factor 2 as the dependent measure also yielded a significant
main effect of group, F(5,202)= 2.480, p=.033, indicating a significant difference among
groups in their attitudes regarding the mans responsibility and the seriousness of the
sexual behavior Planned comparisons showed that child molesters rated the sexual
behavior as less serious and less deserving of severe punishment than controls,
F(l,202)=8.040, p=,005. There were no significant differences between admitters and
deniers or incestuous and nonfamilial offenders.

52
Multivariate analysis of variance. The principal components analysis reduced the
variables into two sets one relating to the Childs Complicity and another relating to the
Seriousness of the Offense. A doubly multivariate analysis of variance was conducted on
each of the two sets of variables. The Child Complicity variable set consisted of Child
Benefit, Child Enjoyment, Child Desire, Child Responsibility, and Other Men. The
variable set relating to the Seriousness of the Offense consisted of Child Harm, Mans
Responsibility, Mans Need for Treatment, and Punishment. Group served as the between
subjects factor with Childs Response (Child) and Level of Sexual Contact (Sex) as the
within subjects factors. The dependent measures were the variables in each variable set.
Social desirability was used as a covariate in the initial analysis, however, it was not
significant in the analysis of Child Complicity, F(5,194)=.422, p= 833, or Seriousness of
the Offense, F(4,195)=.539, p= 707, using the Wilks criterion, so it was dropped from
further analysis.
The doubly multivariate analysis for Child Complicity yielded a significant main
effect for Group, F(25, 733,32)=1.54, p=.046, using Wilks criterion. There was also a
significant Group X Child interaction, F(50, 1818.52)= 1.47, p=.018, using the averaged
multivariate tests of significance. Planned comparisons were conducted between child
molesters and controls, incestuous and nonfamilial offenders, and admitters and deniers.
Incestuous offenders scored significantly different from nonfamilial offenders on Child
Complicity, F(5,197)=2.921, p=.014, using the Wilks criterion. There was no significant
difference between child molesters and controls, F(5,197)=1.410, p=.222, or between
admitters and deniers, F(5,197)=.8372, p=.528.

53
Child molesters and controls differed significantly on the Group X Child
interaction, F(10, 796)=2.134, p=.020, using the averaged multivariate test of significance.
The difference between incestuous and nonfamilial offenders on this interaction
approached significance, F(10, 796)=1.817, p=.054, on the averaged tests. There were no
significant differences among groups on the Group X Sex interaction. The difference
between child molesters and controls approached significance on the Group X Child X
Sex interaction, F(30, 4810)=1.428, p=.061, using the averaged multivariate tests.
The doubly MANOVA on the Seriousness of the Offense yielded a significant main
effect of Group, F(20, 657.64)=2.34, p=.001, and a significant Group X Sex interaction,
F(60, 893.47)=1.522, p=.008. Planned comparisons revealed that the child molesters
differed significantly from controls on the Seriousness of the Offense, F(4,198)=6.015,
p=.000. There were no significant differences among the other groups.
Child molesters differed from controls on the Group X Sex interaction,
F(12,190)=2.505, p=.004. The difference between admitters and deniers on the Group X
Sex interaction approached significance, F(12,1587.74)=1.744, p= .052, using the
averaged multivariate test of significance. The Group X Child X Sex interaction
approached significance for the planned comparison between admitters and deniers, F(24,
178)=1.516, p=067.
Analysis of variance. The significant MANOVA results were followed by
univariate analyses of variance on each of the nine variables. Group effects were
examined on each of the variables, with the Group X Child interaction examined for the

54
Child Complicity variables and the Group X Sex interaction examined for the Seriousness
of the Offense variables
Analysis of Child Complicity variables. As expected, there was a significant
difference between groups on the variable Benefit to Child, F(5,202)=3.380, p=.006 (see
Figure 5). Contrary to expectations, there were no significant differences between child
molesters and controls in their overall benefit ratings, however, there was a significant
finding regarding the nonfamilial admitters. A post-hoc comparison showed that the
nonfamilial admitters ascribed significantly more benefit to the child than the other five
groups, F(l,202)=14.974, using the Scheffe test. Nonfamilial sex offenders attributed
more benefit to the child than did incestuous offenders, F(l,202)=9.214, p=.003, as
predicted. Contrary to expectations, there were no significant differences between
admitters and deniers, F(l,202)=1.292, p=257.
There was a significant difference between child molesters and controls on the
Group X Childs Response interaction, F(2,404)=4.070, p=.018. There were no
significant differences between admitters and deniers or nonfamilial and incestuous
offenders on this interaction.
Groups differed significantly on vignettes in which the child smiled,
F(5,202)=4.106, p=.001, and gave a passive response, F(5,202)=4.403, p=.001. There
were no significant differences among groups, however, on the vignettes in which the child
cried, F(5,202)=1.422, p=.218. This finding is consistent with Stermac and Segals (1989)
results. Planned comparisons showed that the child molesters did not significantly differ
from controls at any level of the childs response. A post-hoc comparison, however,

55
showed that nonfamilial admitters ascribed significantly more benefit to the child than the
other groups when the child smiled, F(l,202)=19.907, p=000. This post-hoc contrast
was not significant when the child was passive or cried. Bonferroni adjusted comparisons,
however, showed that the nonfamilial admitters gave significantly higher ratings on the
passive vignettes than every group except for the nonfamilial deniers.
As expected, groups differed in their ratings of the childs enjoyment, however, the
difference only approached significance, F(5,202)=2.192, p=.057. Planned comparisons
showed no significant differences between child molesters and controls, F(l,202)=1.603,
p=.207, or between admitters and deniers, F(l,202)=1.484, p=.225, however, nonfamilial
offenders gave significantly higher ratings of child enjoyment than incestuous offenders,
F(l,202)=3.984, p=.047, as predicted.
There was also a significant Group X Childs Response interaction
F(10,404)=4.00, p=.000, indicating that the groups ratings of child enjoyment differed
across the levels of the childs response. More specifically, the groups differed
significantly in their ratings of child enjoyment on the vignettes in which the child smiled,
F(5,202)=3.997, p=.002, but did not show significant differences when the child cried or
was passive.
There was a significant Group X Child interaction for the planned comparison
between child molesters and controls, F(2,404)=8.443, p=.000, and for the comparison
between nonfamilial and incestuous offenders, F(2,404)=.019 (see Figure 6). When the
child was smiling, child molesters ascribed more enjoyment to the child than did controls,
F(l,202)=5.919, p=.016. Child molesters did not significantly differ from controls when

56
the child was passive, F(l,202)=1.507, p= 221, but did differ when the child cried,
F(l,202)=4.873, p=.028. Child molesters gave higher child enjoyment ratings than
controls when the child smiled, but gave lower ratings than controls when the child cried.
Nonfamilial offenders ascribed more enjoyment to the child than did incestuous offenders
when the child smiled, F(l,202)=4.798, p=.030, and was passive, F(l,202)=4.153, p=.043,
but did not differ from incestuous offenders when the child cried, F(l,202)=.065, p=.799.
The post-hoc comparison between nonfamilial admitters and the other five groups
approached significance on the Group X Child interaction, F(2,404)=15.706, using the
Scheffe test. Nonfamilial admitters attributed significantly more enjoyment to the child
than the other five groups when the child smiled, F(l,202)=17.561, p=.000, but not when
the child was passive, F(l,202)=7.839, p=.006, or when the child cried, F(l,202)=1.444,
p=.231 (see Table 7, Figure 7).
Planned comparisons showed no significant differences between child molesters
and controls on their ratings of the childs desire for the sexual interaction, F(l,202)=.713,
p=.399 There were also no significant differences between admitters and deniers,
F(l,202)=1.438, p=.232, or between incestuous and nonfamilial offenders,
F( 1,202)= 1.780, p=.184.
Groups differed significantly, however, in their ratings of the childs desire
depending on the childs response. This was indicated by a significant Group X Child
interaction on Child Desire, F(10,404)=4.101, p=.000. Groups differed significantly when
the child smiled, F(5,202)=3.425, p=.005, but not when the child was passive or cried (see
Figure 8). Planned comparisons on the Group X Child interaction showed that child

57
molesters scored significantly different from controls in their ratings across the levels of
the childs response, F(2,404)=7.186, p=.001. The difference between between admitters
and deniers was also significant, F(2,404)=4.327, p=.014 (see Figure 9), as was the
difference between incestuous and nonfamilial offenders, F(2,404)=4.750, p= 009 (see
Figure 10).
Child molesters did not differ from controls in their desire ratings when the child
smiled, F(l,202)=3.440, p=.065, or when the child was passive, F(l,202)=1.378, p=.242.
The child molesters ratings were significantly lower than those of controls, however,
when the child cried, F(l,202)=4.449, p=,036 (see Figure 11). Nonfamilial offenders did
not score significantly different from incestuous offenders when the child smiled,
F(l,202)=2.665, p=. 104, was passive, F(l,202)=3.679, p=.057, or cried; F(l,202)=1.081,
p=.300.
The post-hoc comparison between the nonfamilial admitters and the five groups
approached significance using the Scheffe test, F(2,404)=16.234. Nonfamilial admitters
ascribed more desire to the child than the other groups when the child smiled,
F(l,202)=15 262, p=,000, but not when the child was passive, F(l,202)= 7.883, p=.005 or
cried (Table 10). Figure 11 illustrates that nonfamilial admitters showed the same pattern
on this variable as on Child Benefit and Enjoyment. Nonfamilial admitters attributed
significantly more desire to the child than did the other groups when the child smiled, but
scored the same as others when the childs response was ambiguous or clearly negative
Groups differed significantly in their ratings of the childs responsibility,
F(5,205)=2.703, p=.022. There was also a significant Group X Child interaction,

58
F(10,404)=3.739, p= 000 (see Figure 12). Planned comparisons showed significant
differences between child molesters and controls, F(2,404)=3.751, p=.024 and between
incestuous and nonfamilial offenders, F(2,404)=6.822, p=. 001 on the Group X Child
interaction. Admitters and deniers did not significantly differ on this interaction,
F(2,404)=l 479, p=.229.
On the smile vignettes, the difference between child molesters and controls
approached significance, F(l,202)=3.726, p=.055. There was no significant difference
between child molesters and controls when the child was passive, F(l,202)=2.764,
p=.098, or when the child cried, F(l,202)=.012, p=.911. Nonfamilial offenders ascribed
more responsibility to the child than did incestuous offenders when the child smiled,
F(l,202)=6.616, p=.011 and was passive, F(l,202)=5.585, p=.019, but not when the child
cried, F(l,202)=.022, p= 882.
The post-hoc comparison between nonfamilial admitters and the other five groups
approached significance using the Scheffe test, F(2,404)=16.365. Nonfamilial admitters
showed a similar pattern on this variable as on Child Benefit, Enjoyment and Desire
Using the Scheffe test, the nonfamilial admitters ascribed significantly more responsibility
to the child than the other groups when the child smiled, F(l,202)=14.129, p=.000, but
not when the child was passive or cried.
As expected, groups differed significantly in their ratings of the likelihood of other
men engaging in sexual contact with children, F(5,202)=2.837, p=.017 (see Figure 13).
Child molesters did not differ significantly from controls as expected, F(l,202)=3.0311,
p= 083. Nonfamilial sex offenders, however, did differ from incestuous offenders as

59
expected. Nonfamilial offenders rated other men as more likely to molest children than
did incestuous offenders, although this difference only approached significance,
F(l,202)=3.423, p=.066.
As predicted, admitters were more likely than deniers to rate other men as willing
to engage in sexual contact with a child, F(l,202)=3.853, p=.051, although this difference
only approached significance. A post-hoc comparison showed that, compared to the other
five groups, nonfamilial admitters rated other men as significantly more likely to molest
children, F(l,202)=12.625, p=.000 (see Table 11). There were no significant differences
between child molesters and controls, nonfamilial and incestuous offenders, and admitters
and deniers on the Group X Child interaction.
Analysis of Seriousness of Offense variables. As predicted, the groups differed
significantly in their ratings of the amount of punishment the molester should receive,
F(5,202)=6.623, p=.000. The Group X Sex Contact interaction was also significant,
F(15,606)=2.954, p=.003. Planned comparisons showed that the child molesters
prescribed significantly less punishment than the controls, (1,202)=28.90, p=.000. In
contrast to results on the other eight variables, this significant result did not appear to be
largely affected by the pattern of the nonfamilial admitters. Child molesters and controls
also differed significantly on the Group X Sex interaction, F(3,606)=9.905, p=.000.
Figure 14 shows that the child molesters prescribed significantly less punishment than the
controls across all four levels of sexual contact: touching F(l,202)=33 .470, p=.000,
rubbing F(l,202)=28.924, p=.000, fondling F( 1,202)=18.428, p=.000, and ejaculation
F( 1,202)=11.647, p=.001. The difference between the child molesters and controls was

60
largest when the the sexual contact involved touching only, and this difference decreased
as the level of sexual contact increased. There were no significant differences between
nonfamilial and incestuous offenders, F(3,606)=1.968, p=. 118, or between admitters and
deniers, F(3,606)=1.736, p=. 158 on the Group X Sex interaction.
There were no significant differences among groups on Child Harm,
F(5,202)=.662, p=.653 and there were no significant interactions with Group. There were
also no significant differences among groups on Man Responsibility F(5,202)=1.227,
p=.298 and no significant interactions with Group.
There were no significant differences among the groups on Treatment,
F(5,202)=l 652, p=.148, however, child molesters differed significantly from controls on
the Group X Sex interaction, F(3,606)=4.451, p=.004. Figure 15 shows that the child
molesters consistently prescribed less treatment than controls across all levels of sexual
contact. The difference between groups was significant when the sexual contact involved
touching only, F(l,202)=9.068, p=.003, but not when the sexual contact increased to
rubbing F(l,202)=2.593, p=. 109, fondling F(l,202)=2.966, p=.087, or ejaculation
F(l,202)=1.784, p=. 183, This pattern was similar to the child molesters pattern of
ratings on the variable of punishment. These results suggest that the finding of differences
between child molesters and controls on Factor 2, or Seriousness of Offense, can be
attributed to child molesters different views on punishment and treatment, rather than to
differences in ratings of the mans responsibility or harm to the child.
As expected, there were no significant correlations between each of the nine
variables and scores on the Social Desirability Scale. This finding is consistent with the

61
results of the Stermac and Segal (1989) study. Correlation coefficients ranged from .008
to .052.
Adult vignettes
Principal components analysis. The data from the adult vignettes was analyzed in
the same manner as the child vignettes. A principal components analysis was conducted
on the nine variables collapsed across the levels of sexual contact and the womans
response. The analysis yielded two factors which exceeded the eigenvalue criterion of 1.
Factor 1 accounted for 43.37% of the total variance while Factor 2 accounted for 17.38%
of the variance. The component loadings listed in Table 12 indicate that these two factors
are similar to the two factors derived on the child vignettes. The females Desire,
Enjoyment, Responsibility, and Benefit loaded highly on Factor 1 in both the child and
adult vignettes. In addition, Factor 2 seems to tap the mans role in both vignettes. On the
adult vignettes, Factor 2 taps the Mans Anxiety, Confidence, and Responsibility for the
sexual interaction. Although some of these specific variables were not included on the
child vignettes, these variables focus on the mans role as do the variables which loaded
highly on Factor 2 on the child vignettes.
Principal components analyses were conducted for each of the 12 vignettes. The
analyses yielded 2 to 3 factors for each vignette, similar to the factor structure found in the
analysis of the nine means. Within each vignette, the first factor accounted for 33.01%-
42.03% of the variance and the second factor accounted for 13 05%-19.89% of the
variance. These results are consistent with the results of the analysis using the nine means.

62
Analysis of variance was conducted with Group as the between subjects factor,
Womans Response (Woman) and Level of Sexual Contact (Sex) as the within subjects
factors, and Factor 1 as the dependent measure. There was a significant main effect of
Group, F(5,202)=3.124, p=010. Planned comparisons, however, showed no significant
differences between child molesters and controls, incestuous and nonfamilial offenders, or
admitters and deniers. The significant group effect appears to be related to the inmate
controls who scored higher on this factor than the other five groups, although this
difference was not significant using the Scheffe test, F(l,205)= 10.096, p=.02.
Analysis of variance using Factor 2 as the dependent measure also showed a
significant main effect of Group, F(5,205)=3.436, p=.005. Deniers scored significantly
different from admitters on Factor 2, F(l,202)=10.560, p=.001. Much of this difference
appears to be attributable to nonfamilial deniers who scored higher than the other groups
on this factor, although not significantly higher using the post-hoc Scheffe test,
F(l,205)=8.862, p=.003.
Multivariate analysis of variance. Based on the results of the principal components
analysis, the nine variables on the adult vignettes were divided into two variable sets, one
relating to the Womans Complicity and the other relating to the Mans Role A doubly
multivariate analysis of variance was conducted with each variable set. Social desirability
was initially used as a covariate in the analysis, however, it was found to be nonsignificant
with Womans Complicity, F(6,197)=1.678, p=. 128, and with Mans Role,
F(3,200)=2.351, p=,074, so it was dropped from further analysis.

63
Groups differed significantly on the Womans Complicity, F(30,802)= 1.634,
p=.018, using Wilks criterion. There were no significant interactions with Group.
Planned comparisons revealed significant differences between incestuous and nonfamilial
offenders, F(6,200)=2.186, p=046. There were no significant differences between child
molesters and controls, or between admitters and deniers. There were no significant
differences on the planned comparisons for the Group X Woman interaction. The
difference between nonfamilial and incestuous offenders approached significance on the
Group X Sex interaction, F(18,1725.83)=1.555, p=.064, using the averaged multivariate
tests. Child molesters and controls differed significantly on the Group X Woman X Sex
interaction, F(36,170)=1.919, p=.003.
There was no significant main effect of Group on the Mans Role,
F(15,560.80)=1.363, p=. 160. There were also no significant interactions with Group.
Planned comparisons also showed no significant differences on the main effect of Group
or interactions with Group. These results do not support the hypothesis that child
molesters would rate more anxiety and less confidence to a man in a sexual interaction
with a woman.
Analysis of Womans Complicity variables. In order to further examine the
significant results from the doubly MANOVA, univariate analyses of variance were
conducted for each variable comprising Womans Complicity. Group served as the
between subjects factor with Womans Response and Level of Sexual Contact as the
within subjects factors. For each of these variables there were significant main effects on
Womans Response and Level of Sexual Contact, indicating that subjects ratings differed

64
significantly as the level of sexual contact and the womans response to the sexual contact
changed. There was a significant interaction between these two variables on every
variable except for Womans Harm.
On the adult vignettes, groups were expected to differ only on the variables of
Mans Anxiety and Mans Confidence. Contrary to expectations, groups differed
significantly in their ratings of the amount of benefit the woman received from the sexual
contact, F(5,205)=2.818, p=.017 (see Figure 16 & Table 13). There was the unexpected
finding that nonfamilial sex offenders rated significantly more benefit to the woman than
did incestuous offenders, F(l,205)=7.569, p=.006, and showed a different pattern in their
ratings across the levels of sexual contact, F(3,615)=3.259, p=.021. Figure 17 shows that
the incestuous offenders decreased their ratings of the womans benefit as the degree of
sexual contact increased, while the nonfamilial offenders did not show this decrease in
their ratings. In contrast, the nonfamilial offenders rated the womans benefit as increasing
as the sexual contact progressed from touching only to fondling and ejaculation.
Groups also differed significantly in their ratings of the degree to which the woman
enjoyed the sexual contact, F(5,205)=2.509, p=.031. There was also a significant Group
X Sex interaction, F(15,615)=l .831, p=.027. There were no significant differences among
the groups on the touching or fondling vignettes, however, there were significant
group differences on the no clothes F(5,205)=2.595, p=.027, and ejaculation
vignettes F(5,205)=3.140, p=.009.
Nonfamilial offenders differed from incestuous offenders on the Group X Sex
interaction, F(3,615)=3.976, p=.008, and on the Group X Woman X Sex interaction,

65
F(5,1025)= 3.171, p=. 008. Figure 18 illustrates how both types of sex offenders showed
the same pattern in their ratings of the womans enjoyment when she smiled or was
passive, but differed when the woman cried. Both groups rated the womans degree of
enjoyment as dependent on the level of sexual contact when she smiled or was passive.
When she cried, however, nonfamilial offenders continued to rate her enjoyment as
dependent on the level of sexual contact, while the incestuous offenders tended to rate her
enjoyment as roughly the same across levels of sexual contact. On the ejaculation
vignettes, the inmate controls attributed significantly more enjoyment to the woman than
the other five groups, F(l,205)=l 1.337, p=.001 (see Table 14).
Groups differed significantly in their ratings of the womans desire for the sexual
interaction, F(5,205)=2.283, p=. 048 (see Table 15 & Figure 19). The planned comparison
between child molesters and controls on the Group X Woman X Sex interaction
approached significance, F(5,1025)= 2.202, p=.052. Figure 20 shows that the same
pattern was found between child molesters and controls as was found between nonfamilial
and incestuous offenders on the variable of Womans Enjoyment. Figure 20 illustrates
that the child molesters and controls showed the same pattern in their ratings of the
womans desire when the woman smiled or was passive, but differed when the woman
cried. When she cried, the child molesters ratings of the womans desire were more
dependent on the level of sexual contact while the controls ratings were roughly the same
across levels of sexual contact.
There was also a significant difference among the groups in their ratings of the
womans sexual attraction to the man, F(5,205)=3.723, p=.003 (see Figure 21).

66
Unexpectedly, the inmate control group rated the woman as more sexually attracted to the
man than did the other five groups F(l,205)=16.847, p=.000 (see Table 16 & Figure 21).
There were no significant differences between child molesters and controls,
F(l,205)=1.360, p= 245.
Groups did not significantly differ in their ratings of the harm done to the woman
F(5,205)=1.933, p=.090. Planned comparisons showed that the difference between child
molesters and controls on the Group X Sex interaction, F(3,615)=2.609, p= 051
approached significance. Figure 22 shows that child molesters ascribed more harm to the
woman than did controls across all levels of sexual contact, but particularly when the
sexual contact involved touching or fondling. Both groups scored similarly when the
sexual contact involved genital contact or ejaculation.
There were no significant differences among groups on the ratings of the womans
responsibility for the sexual interaction F(5,205)=l .492, p=. 194. There were also no
significant interactions with Group.
Analysis of Mans Role" variables. Although there were no significant differences
among groups on the doubly MANOVA using these variables, further analysis was
conducted in order to examine the difference found between admitters and deniers on
Factor 2. ANOVAs were conducted for the variables of Mans Responsibility, Mans
Anxiety, and Mans Confidence. Deniers attributed significantly less responsibility to the
man than did admitters, F(l,205)=18,652, p=.000. Deniers also differed from admitters
on the Group X Woman X Sex interaction, F(5,1025)=2.934, p= 012, Figure 23 shows
that the deniers consistently attributed less responsibility to the man across vignettes For

67
the most part, the deniers showed the same pattern of ratings as the deniers across the
levels of sexual contact and the womans response.
A post-hoc comparison showed that the nonfamilial deniers ascribed significantly
less responsibility to the man than the other five groups combined, F(l,205)=l 1.789,
p=.001 (see Table 17). Thus, much of the difference between the admitters and deniers
may have been a result of the extreme pattern of the nonfamilial deniers.
Comparison of child versus adult vignettes
Univariate analyses of variance were conducted for each variable with type of
vignette (child or adult) as the repeated measure. This analysis was conducted in order to
examine if the subjects gave different ratings on the same variables in the child versus the
adult vignettes. As expected, subjects differed significantly in their ratings on each
variable on the child versus adult vignettes Subjects attributed significantly more benefit
to the woman than to the child, F(l,199)=433.662, p=.000, and attributed more harm to
the child, F(l,199)=322.172, p=.000. Subjects also attributed more enjoyment and desire
to the woman, F(l,199)=377.155, p=.000 and F(l,199)=671.605=.000, respectively The
woman was assigned significantly more responsibility for the sexual contact than the child,
F(l,199)=704.403, p=.000, and more responsibility was attributed to the man for sexual
contact with the child, F(l,199)=301.838, p=.000. Groups differed significantly in their
ratings across the vignettes on the variable of females enjoyment, F(5,199)=3.713, p= 003
(see Figure 24). Compared to the other groups, nonfamilial admitters rated the child as
deriving more enjoyment, yet scored similarly to others in their ratings of the womans
level of enjoyment.

Analysis of Empathy
Child molesters were expected to score lower on empathy than controls, and
68
deniers were expected to score lower than admitters Three outliers were identified and
removed from the analysis. Contrary to expectations, a planned comparison revealed no
significant difference between child molesters and controls on the empathy total score
F(l,199)=2.917, p=.089. A planned comparison showed that the nonfamilial sex
offenders scored significantly higher on empathy total score than the incestuous sex
offenders, F(l,199)=5.889, p=.016 (see Table 18). Contrary to predictions, there was no
significant difference between admitters and deniers on total empathy score,
F(l,199)=.623, p= 431. An ANCOVA revealed a significant main effect of group
F(5,199)=3 007,p=.012 (see Figure 25). The covariate was not significant,
F( 1,199)=.724, p=.396. All possible pairwise comparisons using the Bonferroni correction
showed that the nonfamilial deniers scored significantly higher than the incestuous deniers
on the total empathy score (p=.026).
In order to examine differences among the six groups on the four IRI subscales, an
ANOVA was conducted with group as the between subjects variable and the four subscale
scores as the repeated measures. The Group X Subscale interaction was not significant,
F(5,205)=1.36, p=. 162, indicating that differences among the groups did not vary across
subscales.
Analysis of Repression
It was expected that child molesters would score higher on repression than
controls, and that deniers would score higher on repression than admitters. These

69
hypotheses were not confirmed. Child molesters did not differ significantly from controls,
F(l,201)=2.434, p=. 120, and admitters did not differ from deniers, F(l,201)=362, p=,548
on the Repression-Sensitization Scale. An ANCOVA showed a significant effect for
social desirability, F(l, 201)= 10.938, p=.001, but a nonsignificant effect for Group,
F(5,201)=.804, p=.548 (see Table 3). The correlation of the repression scores and
empathy total score was r=.188, X2=7.391, p=.007, indicating that higher scores on
repression (which represent a tendency toward sensitization rather than repression) were
positively correlated with higher scores on empathy.

70
70
C CC IA ID NFA NFD
Group
Figure 1. Percentage of each group reporting a history of abuse.
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers

71
A NFA
NFD
IA
* m
Nonfamilial Admitters
Nonfamilial Demers
Incestuous Admitters
Incestuous Deniers
Community Controls
Inmate Controls
Figure 2. Mean social desirability ratings across groups.

72
Admitters Deniers
Figure 3. Mean social desirability ratings across groups.

73
140
C CC IA ID NFA NFD
Group
Figure 4. Mean scores across groups on the Cognition Questionnaire.
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers

Child Benefit Ratings
74
C CC IA ID NFA NFD
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Deniers
Nonfamilial Admitters
Nonfamilial Deniers
Group
Figure 5. Mean group ratings on Benefit to Child.

Mean Child Enjoyment Ratings
75
Child Response
Figure 6. Mean Child Enjoyment ratings for levels of child response.

76
NFA Nonfamilial Admitters
NFD Nonfamilial Deniers
^ IA Incestuous Admitters
* ID Incestuous Deniers
^ CC Community Controls
O C Inmate Controls
Child Response
Figure 7. Mean Child Enjoyment ratings for levels of child response.

Mean Child Desire Ratings
77
Crying Passive Smiling
Child Response
NFD
G IA
* ID
* CC
c
Nonfamilial Admitters
Nonfamilial Demers
Incestuous Admitters
Incestuous Deniers
Community Controls
Inmate Controls
Figure 8. Mean Child Desire ratings for levels of child response.

Mean Child Desire Ratings
78
Child Response
Nonfamilial Offenders
Incestuous Offenders
Figure 9. Mean Child Desire ratings for levels of child response.

Mean Child Enjoyment Ratings
79
Child Response
Nonfamilial Offenders
Incestuous Offenders
Figure 10. Mean Child Desire ratings for levels of child response.

Mean Child Desire Ratings
80
C Controls
O CM Child Molesters
Figure 11. Mean Child Desire ratings for levels of child response.

Mean Child Responsibility Ratings
81
Crying Passive Smiling
A NFA
NFD
a IA
* ID
* CC
C
Child Response
Nonfamilial Admitters
Nonfamilial Deniers
Incestuous Admitters
Incestuous Deniers
Community Controls
Inmate Controls
Figure 12. Mean Child Responsibility ratings for levels of child response.

Other Men Ratings
82
C CC IA ID NFA NFD
C= Inmate Controls
CC Community Controls
IA= Incestuous Admitters
ID= Incestuous Deniers
NFA= Nonfamilial Admitters
NFD= Nonfamilial Deniers
Group
Figure 13. Mean group ratings on Other Men.

Mean Punishment Ratings
83
12 3 4
C Controls
CM Child Molesters
Level of Sexual Contact
Figure 14. Mean punishment ratings for levels of sexual contact.

Mean Treatment Ratings
84
C Controls
CM Child Molesters
Level of Sexual Contact
Figure 15. Mean treatment ratings for levels of sexual contact.

85
3.0
C CC IA ID NFA NFD
Group
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers
Figure 16. Mean group ratings on Benefit to Woman

Mean Woman Benefit Ratings
Level of Sexual Contact
Figure 17. Mean Woman Benefit ratings for levels of sexual contact.

Mean Woman Enjoyment Ratings
87
Woman Smiling Woman Passive Woman Crying
Level of Sexual Contact
l=Touching 2=Fondling 3=No Clothes 4=Ejaculation
Figure 18. Mean Woman Enjoyment ratings for levels of sexual contact and childs response.

Woman Desire Ratings
88
C CC IA ID NFA NFD
Group
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Demers
Nonfamilial Admitters
Nonfamilial Demers
Figure 19. Mean group ratings on Woman Desire.

Mean Woman Desire Ratings
89
Woman Smiling Woman Passive Woman Crying
Level of Sexual Contact
l=Touching 2=Fondling 3=No Clothes 4=Ejaculation
Figure 20. Mean Woman Desire ratings for levels of sexual contact and woman's response.

Sexual Attraction Ratings
90
Group
C=
cc=
IA=
ID=
NFA=
NFD=
Inmate Controls
Community Controls
Incestuous Admitters
Incestuous Deniers
Nonfamilial Admitters
Nonfamilial Deniers
Figure 21. Mean group ratings on Sexual Attraction.

Mean Woman Harm Ratings
91
Level of Sexual Contact
Child Molesters
Controls
Figure 22. Mean Woman Harm ratings for levels of sexual contact.

Mean Man Responsibility Ratings
92
Figure 23. Mean Man Responsibility ratings for levels of sexual contact and child's response.

I
C CC IA ID NFA NFD
Vignettes:
* ADULT
D CHILD
Group
Figure 24. Mean group ratings on Child Enjoyment and Woman Enjoyment.

94
C= Inmate Controls
CC= Community Controls
IA= Incestuous Admitters
ED= Incestuous Demers
NFA= Nonfamilial Admitters
NFD= Nonfamilial Deniers
Group
Figure 25. Mean group ratings on empathy total score.

95
Table 1
Mean scores on repression and social desirability for self-reported abused and nonabused
subjects.
Repression-Sensitization
Social Desirability
Nonabused
M
39.88*
18,47**
SD
2.01
.52
N
127
126
Abused
M
47.00*
16.08**
SD
2.49
.63
N
83
86
*£=.027. **£=004.
Table 2
Mean scores on social desirability as a function of abuse history and admission or denial.
Admitters
Deniers
Nonabused
M
17.59
19.52*
SD
.82
.83
N
54
52
Abused
M
14.93*
18.06
SD
.92
1.01
N
42
35
Total
M
15.23**
19.70**
SD
82
.83
N
54
52
*£=.002. **£=.000.

96
Table 3
Group means and standard deviations for dependent measures.
Group Empathy Repression Cog. Que. IPT PONS Vocab
Nonfamilial
Admitters
M
SD
62.68
2.25
45.14
4.23
120.86*
3.32
13.71
58
153.73
3.74
34.75
2.90
Nonfamilial
Deniers
M
SD
64.60*
2.17
44.37
4.08
122.23
3.20
13.73
.56
152.83
3.61
28.57
2.80
Incest
Admitters
M
SD
62.37
2.54
47.77
4.77
129.96
3.74
14.55
66
159.50
4.21
38.05
3.28
Incest
Deniers
M
SD
55.00
2.17
38.33
4.08
126.17
3.20
14.20
.56
149.37
3.61
33.20
2.80
Inmate
Controls
M
SD
58.95
1.59
41.59
2.99
127.77
2.34
14.30
.41
152.33
2.64
29.39
2.05
Community
Controls
M
SD
58.53
2.17
37.47
4.08
134.90*
3.20
14.60
.56
153.38
3.61
36.50
2.80
'p-026
p=.015

97
Table 4
Mean scores on the Cognition Questionnaire.
Incest Admitters
Inmate Controls
M
128.09
M
130.24
SD
3.02
SD
1.86
N
22
N
55
Incest Deniers
Community Controls
M
130 88
M
134.64*
SD
2.58
SD
2.56
N
29
N
29
Nonfamilial Admitters
M
124.27
SD
2.61
N
28
Nonfamilial Deniers
M
124.00*
SD
2.47
N
32
TOTALS:
Child molesters
Controls
M
126.81
M
132.44
SD
2.67
SD
2.21
N
111
N
84

98
Table 5
Pearson correlation matrix for the variables on the child vignettes.
1
2
3
4
5
6
7
1) child harm
1.000
2) child desire
-0.347
1.000
3) child respons.
-0.349
0.836
1.000
4) man respons.
0.603
-0.522
-0.530
1.000
5) treatment
0.655
-0.442
-0.442
0.770
1.000
6) other men
-0.174
0.436
0.435
-0.267
-0.269
1.000
7) child enjoy.
-0.330
0.897
0.754
-0.469
-0.400
0.401
1.000
8) punishment
0.419
-0.331
-0.316
0.464
0.688
-0.207
-0.389
9) child benefit
-0.288
0.619
0.613
-0.395
-0.352
0.439
0.603
8
1.000
-0.293
number of observations: 208
Table 6
Rotated loadings for two factors derived from principal components analysis on the child
vignettes
Components
Variables
1
2
Child Desire
.891
.270
Child Respon.
.859
.260
Child Enjoyment
.847
.281
Child Benefit
.760
.197
Other Men
.618
.089
Treatment
-.227
-.907
Child Harm
-.146
-.793
Mans Respon.
-.352
-.779
Punishment
-.171
-.743

99
Table 7
Group means on Benefit to Child across vignettes
Incest Admitters Inmate Controls
M
1.17b
M
1.52
SD
.17
SD
.11
N
23
N
61
Incest Deniers
Community Controls
M
1.37b
M
1.38b
SD
.15
SD
.15
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.01a
M
1.37
SD
.15
SD
.15
N
31
N
31
Note. Means with different subscripts differ significantly at p< 05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.

100
Table 8
Group means on Child Responsibility when the childs response was positive.
Incest Admitters
Inmate Controls
M
1 69b
M
1.91
SB
.19
SD
.11
N
23
N
61
Incest Deniers
Community Controls
M
1.84
M
1.54b
SD
.16
SD
.16
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.44a
M
1.94
SD
.16
SD
.16
N
31
N
32
Note. Means with different subscripts differ significantly at [><05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=,000.

101
Table 9
Group means on Child Enjoyment when the childs response was positive.
Incest Admitters Inmate Controls
M
2.28b
M
2.34b
SD
.20
SD
.12
N
23
N
61
Incest Deniers
Community Controls
M
2.41
M
2.07b
SD
.17
SD
.18
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
3.09a
M
2.39
SD
.17
SD
.17
N
31
N
31
Note. Means with different subscripts differ significantly at p<05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.

102
Table 10
Group means on Child Desire when the childs response was positive.
Incest Admitters
M
2.12
Inmate Controls
M 2.20b
SD
.20
SD
.12
N
23
N
61
Incest Deniers
M
2.21
Community Controls
M 1.91b
SD
.17
SD
.17
N
31
N
30
Nonfamilial Admitters
M
2.82a
Nonfamilial Deniers
M 2.08b
SD
.17
SD
.17
N
31
N
32
Note. Means with different subscripts differ significantly at p< 05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.

103
Table 11
Group means on Other Men across vignettes.
Incest Admitters Inmate Controls
M
1.86
M
1.98
SD
.18
SD
.11
N
23
N
61
Incest Deniers
Community Controls
M
1.79
M
1.63b
SD
.16
SD
.16
N
31
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.42a
M
1 84
SD
.16
SD
.16
N
31
N
32
Note. Means with different subscripts differ significantly at p< 05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.000.

104
Table 12
Rotated loadings for two factors derived from principal components analysis on the adult
vignettes
Components
Variables
1
2
Woman Desire
.915
.092
Woman Enjoyment
.905
.054
Woman Attraction
819
.228
Woman Respon
.814
.123
Woman Benefit
.626
-.058
Woman Harm
-.505
.130
Man Anxiety
.048
.766
Man Confidence
.321
.725
Man Respon.
-.400
.594
Table 13
Group means on Benefit to Woman across vignettes.
Incest Admitters Inmate Controls
M
2.40b
M
2.90a
SD
.14
SD
.09
N
25
N
60
Incest Deniers
M
2.62
Community Controls
M 2.59
SD
.12
SD
.13
N
33
N
30
Nonfamilial Admitters
M
2.92
Nonfamilial Deniers
M 2.80
SD
.13
SD
.12
N
31
N
32
Note. Means with different subscripts differ significantly at p<05, using Bonferroni
correction.

105
Table 14
Group means on Woman Enjoyment on ejaculation vignettes.
Incest Admitters Inmate Controls
M
2.80
M
3.07a
SD
.11
SD
.07
N
25
N
60
Incest Deniers
Community Controls
M
2.70b
M
2.69b
SD
.10
SD
.10
N
33
N
30
Nonfamilial Admitters
Nonfamilial Deniers
M
2.76
M
2.95
SD
.10
SD
.10
N
31
N
32
Note. Means with different subscripts differ significantly at g<05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.001

106
Table 15
Group means on Womans Desire across vignettes.
Incest Admitters Inmate Controls
M
2.72
M
2.92
SD
.09
SD
.06
N
25
N
60
Incest Deniers
M
2.72
Community Controls
M 2.64
SD
.08
SD
.08
N
33
N
30
Nonfamilial Admitters
M
2.88
Nonfamilial Deniers
M 2.82
SD
.08
SD
.08
N
31
N
32
Note. Means with different subscripts differ significantly at ¡><.05, using Bonferroni
correction. Mean with subscript a differs significantly from other five groups combined,
p=.001

107
Table 16
Group means on Womans Sexual Attraction across vignettes.
Incest Admitters Inmate Controls
M
2.89
M
3.23
SD
.10
SD
.07
N
25
N
60
Incest Deniers
M
2.91
Community Controls
M 2.80
SD
.09
SD
.09
N
33
N
30
Nonfamilial Admitters
M
2.97
Nonfamilial Deniers
M 2.94
SD
.09
SD
.09
N
31
N
32
* differs significantly from other five groups combined, p= 000.

108
Table 17
Group means on Mans Responsibility across vignettes.
Incest Admitters Inmate Controls
M
4.00**
M
3.70
SD
.11
SD
.07
N
25
N
60
Incest Deniers
M
3.58*
Community Controls
M 3.79
SD
.10
SD
.10
N
33
N
30
Nonfamilial Admitters
M
3.88b**
Nonfamilial Deniers
M 3.42a*
SD
.10
SD
.10
N
31
N
32
Note. Means with different subscripts differ significantly at g<05, using Bonferroni
correction. Mean with subscript a" differs significantly from other five groups combined,
p=.001.
Note. The two means denoted significantly differ from the means denoted **, p=.000.

109
Table 18
Group means on empathy total score.
Nonfamilial Sex Offenders
M
SD
N
64 03
2.00
61
Incestuous Sex Offenders
M
SD
N
59.09
2.10
57
p=.016.

CHAPTER 4
DISCUSSION
The results confirm some of the hypotheses and disconfirm some of the other
hypotheses. The first hypothesis regarding decoding skill was not supported by the data.
Contrary to expectations, there were no significant differences among groups on decoding
skill or social perception. Thus, these results suggest that child molesters sexual
offending does not appear to be related to a deficiency in perceiving others cues
accurately. Although several of the IPT scenes involved childrens cues, it was not
possible to score decoding ability based on childrens cues alone. Thus, it is possible that
child molesters may show a deficit in reading these particular cues. It may be also be that
these measures do not truly tap decoding skill or social perception, and that a more
sophisticated measure might be more likely to reveal a deficiency.
The second hypothesis proposed that child molesters would show more distorted
cognitions than controls regarding sexual contact with a female child. This hypothesis was
confirmed, as child molesters showed more cognitive distortions than controls on the Abel
and Becker Cognition Questionnaire and on the child vignettes. Thus, child molesters
endorsed more deviant beliefs regarding sexual contact with children. On the child
vignettes, child molesters ascribed more benefit to the child than did controls, and also
ascribed more enjoyment to the child when she was smiling. Child molesters also tended
to attribute more responsibility than controls to the child when she was smiling, although
110

Ill
this difference only approached significance. Interestingly, the child molesters gave
significantly lower ratings than controls on child enjoyment and desire when the child
cried. This may reflect child molesters tendency to overcompensate in their ratings when
the childs response is clearly negative and the socially desirable response is more obvious.
It might also suggest that child molesters are more aware of childrens lack of enjoyment
and desire when their responses are clearly negative.
Child molesters also prescribed less punishment for the offender regardless of the
degree of sexual contact, although their ratings became more similar to those of controls
as the degree of sexual contact increased. Thus, the child molesters pattern of responses
suggest that they used the degree of sexual contact to rationalize the sex offending
behavior, or assign it less punishment. This result is consistent with sex offenders
clinically observed cognitive distortion that sexual contact not involving intercouse is
relatively harmless (Salter, 1988). Child molesters also prescribed significantly less
treatment for the offender than did controls if the sexual contact involved touching only.
Although differences were found between child molesters and controls on the child
vignettes, most of the differences appeared to be attributable to the nonfamilial admitters,
a subtype within the child molester group Most of the significant differences were found
between the nonfamilial admitters and the five other groups combined. In contrast to the
other four sex offender groups, the nonfamilial admitters showed the response pattern that
was expected based on the Stermac & Segals (1989) results. Across several variables,
the nonfamilial admitters showed more cognitive distortions than the other groups when
the childs response was positive, but scored more similarly to the other groups when the

112
childs response was clearly negative. The nonfamilial admitters also tended to show more
cognitive distortions than other groups when the childs response was passive, however,
these differences only approached significance when the strict criterion of the Scheffe test
was used. Nonfamilial admitters showed this pattern on the variables of child benefit,
child enjoyment, child desire, and child responsibility. That is, when the child smiled, the
nonfamilial admitters attributed significantly more benefit, desire, enjoyment, and
responsibility to the child than did the other five groups. Nonfamilial admitters also rated
other men as more likely to molest children than did the other five groups. These results
suggest a severe lack of insight among nonfamilial admitters regarding the childs
perspective.
These results also suggest that nonfamilial admitters tend to read more than
others into the childs response when it is positive, perhaps in order to justify the behavior.
This is often seen among sex offenders in treatment settings who admit to rationalizing
their behavior by claiming that the child really enjoyed the contact since she didnt cry
(Salter, 1988). It appears that nonfamilial admitters may also use the excuse that other
men molest fairly often, thus making the behavior more acceptable. This particular
cognitive distortion is also seen frequently among sex offenders in treatment.
Interestingly, nonfamilial admitters did not show cognitive distortions regarding
the mans responsibility for the sexual interaction or his need for treatment. This suggests
that, although the nonfamilial admitters attribute more responsiblility, benefit, desire, and
enjoyment to the child, they do not show cognitive distortions regarding the mans role in
the sexual interaction. In fact, they were the only sex offender group which did not

113
prescribe significantly less punishment than inmate controls when the child cried. This
suggests that nonfamilial admitters may recognize that the child molester should be
punished, although they continue to maintain cognitive distortions about the childs
involvement. This also indicates that the differences between child molesters and controls
on punishment and treatment ratings are not attributable to the nonfamilial admitters.
Although the results suggest that these cognitive distortions are found only among
nonfamilial admitters, it may be that this is the only group which will admit to these
cognitive distortions on a face valid questionnaire. The data collected from the individual
interviews support this hypothesis, as all of the four child molester subgroups showed
cognitive distortions when responding to verbal vignettes depicting sexual contact
between a child and an adult. The less structured format of the interview allowed more
opportunity for subjects to express their views in their own words, and thus, potentially
reveal more cognitive distortions. As expected, each of the four child molester groups
revealed more cognitive distortions than the controls. Statements which revealed a
permissive attitude toward adult sexual contact with children were considered to be
cognitive distortions. For example, the first question asked how the 35-year old man feels
about the 9-year old girl with whom he is sexually involved. A common cognitive
distortion found among the child molesters involved the belief that the man loved the girl
and was appropriately expressing his love for her in a sexual way. In contrast, the controls
tended to express the idea that the man was exploiting this child for his own selfish sexual
gratification.

The child molesters also tended to express the belief that the girl must love the
man or that she would not be sexually involved with him. They tended to view her as
consenting to the sexual interaction and as participating as an equal partner in the
relationship. In contrast, the controls tended to describe the child as feeling afraid and
unsure about the sexual interaction. In addition, child molesters most often indicated that
they would not interfere with this relationship if they were friends with the offender In
contrast, the controls usually indicated that they would try to intervene in some way in
order to stop the sexual abuse from continuing. Most of the control subjects also
indicated that they would either take physical action against the perpetrator or would
notify the police of the sexual offense In sum, the data from the interviews offer support
for the notion that child molesters show cognitive distortions, although their endorsement
of these distortions may vary according to the structure and format of the assessment
method.
Although child molesters as a group show more cognitive distortions than
controls, differences were found among the subtypes of child molesters. The results from
the child vignettes offered strong support for the hypothesis that nonfamilial offenders
show more cognitive distortions than incestuous offenders. In contrast to the incestuous
offenders, the nonfamilial offenders attributed more benefit and desire to the child, and
attributed more enjoyment and responsibility to the child when the child smiled or was
passive. Nonfamilial offenders ratings did not differ from those of incestuous offenders
when the child cried. Thus, nonfamilial offenders, particularly nonfamilial admitters,
showed significant differences compared to incestuous offenders in their perception of the

115
childs role in the sexual contact. It is important to note, however, that many of these
differences might have resulted from the extreme pattern of the nonfamilial admitters, and
may therefore not hold for nonfamilial deniers
Compared to incestuous offenders, nonfamilial offenders also rated other men as
more likely to molest children, although this difference only approached significance. The
difference between the nonfamilial and incestuous offenders also approached significance
on the Abel & Becker Cognition Questionnaire, indicating that nonfamilial offenders
endorsed more cognitive distortions than incestuous offenders regarding sexual contact
with children. In addition, the interview data also indicate that the nonfamilial offenders
showed more cognitive distortions than the incestuous offenders. The incestuous
offenders, however, showed more cognitive distortions than the control groups, which
would be expected given clinical observations noted in the literature (Salter, 1988). Thus,
incestuous offenders seem to show some cognitive distortions regarding sexual contact
with children, although show fewer cognitive distortions than nonfamilial offenders This
finding offers support for the literature suggesting that nonfamilial and incestuous
offenders molest for different reasons If nonfamilial offenders tend to molest because
they prefer children, then one might expect them to show more severe cognitive
distortions rationalizing their lifestyle preference. On the other hand, if incestuous
offenders tend to be situational offenders, then one might expect them to show fewer
cognitive distortions since they would be more likely to show a sexual preference for
adults rather than children Thus, incestuous offenders may have less of a strong need to
rationalize their sexual deviancy since it is not based on their primary sexual interest,

116
although, they may still have a need to rationalize their incestuous behavior to some
degree.
Contrary to expectations, child molesters did not differ from controls in their
ratings of the mans degree of responsibility for the sexual interaction. These results
contrast with those of Stermac & Segal (1989) who found that child molesters attributed
less responsibility overall to the man, and that their ratings depended more on the level of
sexual contact and childs response than did the controls ratings. The child molesters
ratings differed from controls on all levels of sexual contact, but were similar to the
controls ratings when the childs response was negative. Based on these results, it was
expected that child molesters in the present study would show the same pattern of
responding. The reason for these different results is unclear. It may be that the difference
in sample composition might account for this difference. Stermac & Segal (1989)
included homosexual and heterosexual child molesters in their sample, while the present
study included only heterosexual child molesters It may be that homosexual molesters
attribute less responsibility to to the man than heterosexual molesters, thus resulting in
different results. Another possible explanation might be related to differences in the
forensic settings where the data was collected and differential demands for social
desirability in these settings. It may be that child molesters volunteering for treatment at a
psychiatric facility may have fewer social desirability concerns than incarcerated child
molesters not participating in a treatment program, and thus incarcerated molesters may be
more reluctant to admit to obvious cognition distortions regarding the mans
responsibility. In any case, the present finding suggests that child molesters are aware of

117
the mans responsibility for the sexual behavior, although nonfamilial offenders also
attribute responsibility to the child.
Another finding which contrasts with the results of Stermac & Segal (1989) is the
lack of a difference between child molesters and controls in their ratings of the amount of
harm inflicted on the child Stermac & Segal (1989) combined the variable child harm
with child enjoyment and child benefit to form a composite variable. They found that child
molesters scored higher on this composite variable than did controls, indicating that child
molesters attributed more benefit and less harm to the child than did controls. The results
of the present study on the variable of child harm are not directly comparable to Stermac
& Segals (1989) composite variable since results on child harm cannot be extracted from
this composite variable Differences were found between the nonfamilial admitters and
other groups on the variable of benefit to the child, suggesting that the difference on the
composite variable might have been largely attributed to the individual variable of benefit
rather than to child harm. In any case, the results of the present study suggest that child
molesters have some awareness of the harm done to the child or, at least, possess an
awareness of the socially appropriate response.
The third hypothesis suggested that deniers would show fewer cognitive
distortions than admitters. The results provide minimal support for this hypothesis, as
most of the findings only approached significance and appeared to have been largely
affected by the extreme pattern of the nonfamilial admitters. Like the nonfamilial
admitters, admitters as a group tended to rate the child as more desirous of the sexual
contact than did deniers when the child smiled or was passive, but rated them similarly to

118
deniers when the child cried. The admitters showed a significantly different pattern from
deniers across the childs response, but did not significantly differ from deniers at the
individual levels of the childs response. The deniers also rated the man in the adult
vignettes as having less responsibility for the sexual interaction than did the admitters,
however, this result was largely attributable to the nonfamilial deniers who gave higher
ratings than the other five groups. This finding suggests that nonfamilial deniers, like
nonfamilial admitters, do show cognitive distortions regarding sexual interaction, although
they did not show many cognitive distortions on the child vignettes. This may reflect their
desire to present themselves in a good light since the child vignettes are much more face
valid than the adult vignettes. In sum, there were few differences between admitters and
deniers, and the few differences that emerged appeared to be more related to the extreme
pattern of the nonfamilial offenders Thus, the offenders relationship to the victim
appeared to be a more important variable than the offenders admission or denial of the
crime.
In contrast to the results from the vignettes, the interview data suggest that deniers
do show cognitive distortions regarding sexual contact with children, although they tend
to show fewer distortions than admitters. Deniers also tended to show more subtle types
of distortions regarding the genuineness of the mans feelings for the girl or vice versa.
Although the results suggest that social desirability was not significantly correlated with
the responses on the vignettes, it seems likely that the Likert scale format presented more
demands for social desirability given the structure and face validity of the answers. In
contrast, the open-ended question format of the interview forced the subjects to answer

119
provocative questions in their own words, thus making it easier for them to reveal
cognitive distortions in their thinking. In addition, the interview was conducted
individually while the vignettes were completed in a group testing situation in which
inmates were sitting next to each other. There may have been strong demand
characteristics for circling particular answers in this group situation since it is generally
considered to be dangerous to admit to permissive attitudes toward sexual contact with
children while in a prison environment. Thus, there seem to be several plausible
explanations for the differences between the vignette and interview data, all of which
relate to concerns regarding social desirability, which was found to be particularly high
among deniers.
The fourth hypothesis suggested that child molesters would not show the same
cognitive distortions regarding sexual contact with an adult as they did regarding sexual
contact with a child. The results generally supported this hypothesis, however, the child
molesters differed from controls in their ratings of the womans desire. When the
womans response was clearly negative, the child molesters rated the womans desire for
the sexual contact as more dependent on the degree of sexual contact than did the
controls. This difference was slight, however, and was not found on other variables. In
fact, the child molesters rated the sexual touching and fondling as more harmful for the
woman than did the controls. This suggests that child molesters may be able to empathize
more with an adult victim than with a child in the age range of their own victim. Thus,
their clinically noted lack of empathy may be specific to their victims, or to children, rather
than generalizare to adults.

120
Although child molesters as a group did not differ significantly from controls, there
were significant differences among subtypes of child molesters on the adult vignettes.
More specifically, nonfamilial and incestuous offenders differed in their ratings of the
womans enjoyment and benefit from the sexual contact. On the benefit ratings,
incestuous offenders attributed less benefit to the woman as the degree of sexual contact
increased, while the nonfamilial offenders tended to attribute more benefit. In addition,
the incestuous offenders ratings of the womans enjoyment became less dependent on the
degree of sexual contact when the womans response was clearly negative. In contrast,
the nonfamilial offenders tended to vary their ratings of the womans enjoyment based on
the degree of sexual contact, although she was portrayed as crying. These results suggest
that the nonfamilial offenders may focus more on the degree of sexual contact as the most
important variable in a sexual interaction rather than the womans response to the sexual
contact. This might reflect a self-centeredness among nonfamilial offenders which might
help explain their noted lack of success in relationships with adult women.
There were several unexpected findings regarding the inmate control group.
Compared to the other five groups, the inmate controls attributed significantly more
enjoyment to the woman on the ejaculation vignettes which portray nonmutual orgasm
with the man ejaculating after rubbing his penis between the womans legs. In addition,
the inmate controls gave significantly higher ratings than the other groups on the womans
sexual attraction to the man. These findings suggests that nonsex offending inmates tend
to view women as being sexually interested regardless of their outward response (smiling,
passive, or crying) and as enjoying sexual contact focused primarily on the mans

121
enjoyment. These results also suggest that cognitive distortions regarding sexual
interaction are not limited to sex offenders, particularly when they relate to adult women.
There were no significant differences among groups on the ratings of the mans
anxiety or the mans confidence, suggesting that child molesters did not view the man as
more anxious or less confident in the situation than did controls. While these questions
were designed to tap into the child molesters own anxiety and lack of confidence
regarding sexual contact with women, it may be that the subjects answered the questions
solely on the information in the vignette, without projecting any of their own feelings or
beliefs onto the man. Thus, the child molesters may have perceived the man in the
vignettes as confident without perceiving themselves as being confident in that same
situation.
Child molesters also did not differ from controls in their ratings of the womans
responsibility for the sexual contact. These results differ from those on the child vignettes
in that the nonfamilial admitters did not attribute significantly more responsibility to the
woman than the other groups as they did with the child. Thus, the nonfamilial admitters
cognitive distortion about the females responsibility was limited to sexual contact with
children and did not generalize to adult women.
Hypothesis #5 proposed that child molesters would show less empathy than
controls. Contrary to expectations, child molesters did not differ from controls in their
level of empathy. This finding is consistent with Marshall et als (1993) finding of no
differences between incarcerated child molesters and controls. It may be that child
molesters do not lack general empathy for others, but that they lack specific empathy for

122
their victims. It may also be that child molesters perceive themselves as empathic, and
therefore rate themselves as empathic on self-report measures of empathy. Thus, it might
be important to use behavioral measures rather than self-report measures of empathy for
this particular group. A future study might examine the differences between child
molesters self-perception of their empathy versus others perceptions of their capacity for
empathy.
Although there were no differences between child molesters and controls on
empathy, there was the unexpected finding that nonfamilial sex offenders scored
significantly higher on empathy than incestuous offenders. It is possible that nonfamililial
offenders may have greater empathy than incestuous offenders, however, it is more likely
that nonfamilial offenders just report themselves as being more empathic. Thus, this
finding might reflect a lack of insight regarding their own capacity for empathy This is
probably the case since the nonfamilial offenders responses on the child and adult
vignettes revealed a lack of empathy for the victims perspective. Unlike the incestuous
offenders, the nonfamilial admitters rated the child as benefitting more from the sexual
contact, as well as desiring and enjoying it more. Thus, it is not surprising that they might
also misperceive themselves as being empathic and attuned to the childs desires. If this is
the case, this finding might actually suggest that not only are nonfamilial offenders not
more empathic than incestuous offenders, they also show yet another cognitive distortion
compared to incestuous offenders.
Contrary to expectations, there were no differences between admitters and deniers
on empathy. Thus, level of empathy appears to be significantly related to the relationship

123
between the perpetrator and victim, and unrelated to whether the perpetrator admits or
denies having committed the crime.
Hypothesis #6 proposed that child molesters would show more repression than
controls. Contrary to expectations, there were no significant differences among groups on
repression. This finding suggests that deniers are not characterized by a deeper level of
repression than admitters. Deniers do differ from admitters, however, in their level of
social desirability. As expected, deniers showed significantly more social desirability than
admitters. Thus, denial of child molesting behavior appears to be related to concerns of
social desirability rather than to differences in level of repression. This finding suggests
that deniers may have some motivation for treatment, given that they seem to be
motivated to please others by presenting themselves in a socially desirable light. This
need for social desirability might help them in treatment if they were to enter a treatment
group comprised of admitters in which telling the truth was considered to be the most
socially desirable action. Thus, treatment programs might target deniers by creating a
social climate within the program which places a high value on admission of ones crime
and offers social rewards for this behavior.
In addition to highlighting the need to subdivide child molesters based on denial
and the relationship to the victim, the results of the present study also highlight the
importance of offenders self-reported histories of abuse Not surprisingly, subjects
reporting histories of abuse scored lower on repression than those denying histories of
abuse. This is understandable since one would need to face ones problems, and thus not
repress, in order to admit to a history of abuse. In addition, subjects who reported a

124
history of abuse scored lower on social desirability than those denying a history of abuse.
Again, this is not surprising since admitting to a history of abuse on a questionnaire
involves some suspension of social desirability concerns. Admission or denial of the crime
is also related to social desirability in that those who admit to their crime and report
histories of abuse score significantly lower on social desirability than those who deny their
crime and deny histories of abuse Thus, inmates who deny committing their crime and
deny having a history of abuse will tend to deny faults in general. This highlights, once
again, the importance of denial in subjects self-reporting.
Contrary to expectations, abused subjects did not score lower than nonabused
subjects on empathy, suggesting that a history of abuse may not impair ones capacity for
empathy as Miller and Eisenberg (1988) found in their meta-analysis. A more likely
interpretation, however, is that this different result may be attributed to the use of a
predominantly incarcerated sample in contrast to the nonincarcerated samples typically
used in the empathy literature. Thus, it may be that incarcerated samples tend to rate
themselves as higher on empathy than nonincarcerated samples in general in an effort to
present themselves in a positive light and maintain that they are good people although
they are in prison. This seemed to be the case with the child molesters who scored higher
on empathic concern. It may also be that many subjects underreported their histories of
abuse as a result of their tendency to deny faults, thus resulting in some abused subjects
being misclassified as nonabused subjects. In any case, this finding merits further
investigation in future research since the reasons for the different results remain unclear. A

125
future study might focus on differences between ones self-report of empathy versus other
measures of empathy such as more objective behavioral measures.
Summary of Findings
In sum, the major findings of this study were: 1) Child molesters did not show a
deficit in decoding skill or social perception; 2) Child molesters showed more cognitive
distortions than controls regarding sexual contact with children, however, most of these
distortions were found among the nonfamilial admitters; 3) Nonfamilial admitters showed
cognitive distortions on both the child and adult vignettes; 4) Incestuous offenders showed
significantly fewer cognitive distortions than nonfamilial offenders, however, incestuous
offenders did show some cognitive distortions; 4) Deniers scored higher on social
desirability than admitters; 5) Child molesters did not differ significantly from controls on
measures of empathy, although nonfamilial offenders scored higher on empathy than
incestuous offenders; and 6) There were no differences among groups on a measure of
repression.
Conclusions
In conclusion, these findings highlight the importance of subdividing child
molesters into groups based on their relationship to the victim. Subdividing child
molesters based on admission or denial appears to be of less value, given that the primary
difference between admitters and deniers involves degree of social desirability. Admission
or denial of the crime is more likely to change as the offender enters treatment and is thus
less stable as a defining characteristic of the child molester. The offenders relationship to
the victim, however, appears to be a much more valuable distinction, given that

126
nonfamilial and incestuous offenders differed significantly in their degree of cognitive
distortions. These results also offer support for the notion that incestuous molesters tend
to be situational offenders while nonfamilial molesters tend to be sexual-preference
offenders.
Future research might focus on how to incorporate these differences into treatment
programs for incestuous versus nonfamilial offenders. For example, treatment for
nonfamilial offenders might focus on their self-centered distortions and lack of sensitivity
to their victims while treatment for incestuous offenders might focus on the more subtle
distortions they used to allow themselves to offend Future research on empathy might
also focus on child molesters specific empathy for their victims, since it is possible that
they might possess only a specific deficiency in empathy. Clearly, more
research is needed in order to fully understand the role of empathy and cognitive
distortions in sex offending behavior so this knowledge can be used to enhance treatment
with a highly deviant and dangerous clinical population.

REFERENCES
Abel, G.G., Becker, J.V., & Cunningham-Rathner, J. (1984) Complications, consent, and
cognitions in sex between children and adults. International Journal of Law and Psychiatry.
7, 89-103.
Abel, G.G., Blanchard, E.B., & Barlow, D.H. (1981). Measurement of sexual arousal in
several paraphilias: The effects of stimulus modality, instructional set and stimulus content
on the objective. Behaviour Research and Therapy. 19(11. 25-33.
Abel, G.G., Gore, D.K., Holland, C.L., Camp, N., Becker, J.V., & Rathner, J. (1989). The
measurement of the cognitive distortions of child molesters. Annals of Sex Research. 2,
135-153.
Abel, G.G., Mittleman, M S., & Becker, J. (1985). Sexual offenders: Results of assessment
and recommendations for treatment. In M.H. Ben-Aron, S.J. Hucker & C D. Webster
(Eds), Clinical criminology: The assessment and treatment of criminal behaviour (dp. 191-
205). Toronto, University of Toronto and M & M Graphics, Ltd.
Abel, G., Rouleau, I, & Cunningham-Rathner, J. (1985). Sexually aggressive behavior In W.
Curran, A.L. McGarry, & S.A. Shah (Eds.), Forensic psychiatry and psychology:
Perspectives and standards for interdisciplinary practice (pp. 289-313). Philadelphia: F A.
Davis.
Barbaree, H E., & Marshall, W.L. (1989). Erectile responses among heterosexual child
molesters and matched non-offenders: Five distinct age preference profiles. Canadian
Journal ofBehavioral Science. 21(11. 70-81.
Becker, J., Kaplan, M., & Tenke, C. (1992). The relationship of abuse history, denial, and
erectile response profiles of adolescent sexual perpetrators. Behavior Therapy. 23, 87-97.
Bliss, E.L., and Larson, E.M. (1985) Sexual criminality and hypnotizability. Journal of
Nervous and Mental Disease. 173. 522-526.
Briere, J., & Runtz, M. (1989). Symptomatology associated with childhood sexual
victimization in a non-clinical adult sample. Child Abuse & Neglect. 12, 51-59.
127

128
Bryer, J.B., Nelson, B.A., Miller, J.B., & Krol, P.A. (1987). Childhood physical and sexual
abuse as factors in adult psychiatric illness. American Journal of Psychiatry. 144. 1426-
1430.
Byrne, D. (1961). The Repression-Sensitization Scale: Rationale, reliability, and validity
Journal of Personality. 29,334-349.
Byrne, D., Bany, J., & Nelson, D (1963). Relation of the Revised Repression-Sensitization
Scale to Measures of Self-Description. Psychological Reports. 13,323-334.
Camras, L.A., Grow, J.G., & Ribordy, S.C. (1983). Recognition of emotional expression by
abused children. Journal of Clinical Child Psychology. 12, 325-328.
Costanzo, M. & Archer, D. (1989). Interpreting the expressive behavior of others: The
Interpersonal Perception Task. Journal of Nonverbal Behavior. 13 (41. 225-243.
Crowne, D., & Marlowe, D. (1960). A new scale of social desirability independent of
psychopathology. Journal of Consulting Psychology. 24, 349-354.
Dahlstrom, W.G., Welsh, G.S., & Dahlstrom, L.E. (1975). An MMPI handbook: VolU
Research Applications. Minneapolis, MN: University of Minnesota Press.
Davis, M.H. (1979). Individual differences in empathy: A multidisciplinary approach
(Doctoral dissertation, University of Texas, Austin, 1979). Dissertation Abstracts
International. 40, 3480B.
Davis, M.H. (1980). A multidimensional approach to individual differences in empathy JSAS
Catalog of Selected Documents in Psychology. JO, 85.
Davis, M.H. (1983). Measuring individual differences in empathy: Evidence for a multi
dimensional approach. Journal of Personality and Social Psychology. 44 ill. 113-126.
Dill, D., Chu, J., Grob, M., & Eisen, S. (1991). The reliability of abuse history reports: A
comparison of two inquiry formats. Comprehensive Psychiatry. 32(2), 166-169.
Dwyer, M., Rosser, S., & Sawyer, S. (1992). Dissociative experiences of sexual offenders: A
comparison between two outpatient groups and those found to be falsely accused. Special
Issue: Sex offender treatment: Psychological and medical approaches Journal of Offender
Rehabilitation. 18(3-41. 49-58.
Erickson, W.D., Luxenberg, M.G., Walbek, N.H., Seely, R.K. (1987). Frequency of MMPI
two-point code types among sex offenders. Journal of Consulting and Clinical Psychology.
55(41, 566-570.

129
Feldman, R. (1992). Applications of Nonverbal Behavioral Theories and Research. Lawrence
Erlbaum Associates: Hillsdale, N.J.
Feldman, R. & Rime, B. (1991). Fundamentals of Nonverbal Behavior. Cambridge University
Press: Cambridge.
Feshbach, N. (1978). Studies of empathic behavior in children. In B.A. Maher (Ed.), Progress
in Experimental Personality Research (Vol. 8, pp. 1-47). New York: Academic Press.
Finkelhor, D. (1984). Child Sexual Abuse: New Theory and Research. New York: Free Press.
Finkelhor, D. (1986). A Sourcebook on Child Sexual Abuse. Beverly Hills, CA: Sage.
Finkelhor, D., & Lewis, I. (1988). An epidemiologic approach to the study of child
molestation. In Prentky, R. & Quinsey, V (Eds ), Annals of the New York Academy of
Sciences (Vol. 528, pp 64-78) New York: New York Academy of Sciences.
French, D. (1988). Distortion and lying as defense processes in the adolescent child molester.
Journal of Offender Counseling Services and Rehabilitation 13111. 27-37.
Frisbie, Vanasek, F., & Dingman, H. (1967). The self and the ideal self: Methodological study
of pedophiles. Psychological Reports. 20, 699-706.
Frischholz, E., Braun, B., Sachs, R., Schwartz, D. (1992). Construct validity of the
Dissociative Experiences Scale: H Its relationship to hypnotizability. American Journal of
Clinical Hypnosis. 35121. 145-152.
Groth, AN., & Bimbaum, H.J. (1979). Men who rape The psychology of the offender. New
York: Plenum.
Hall, J.A. (1979). Gender, gender roles, and nonverbal communication skills. In R. Rosenthal
(Ed.), Skill in nonverbal communication: Individual differences. Cambridge, MA:
Oelgeschlager, Gunn & Hain.
Hamilton, D., & Ondrovik, J. (1993, April 29-May 2). Dissociation in pedophile sex
offenders. Paper presented at the Ninth Annual Symposium in Forensic Psychology,
Santa Fe, NM.
Haywood, T.W., & Grossman, L.S. (1994). Denial of deviant sexual arousal and
psychopathology in child molesters. Behavior Therapy. 25, 327-340.
Haywood, T.W., Grossman, L.S., & Hardy, D W. (1993). Denial and social desirability in
clinical evaluations of alleged sex offenders. The Journal of Nervous and Mental Disease.
181(31. 183-188.

130
Hobson, W., Boland, C., & Jamieson, D. (1985). Dangerous sexual offenders. Medical
Aspects of Human Sexuality. 19(21. 104-119.
Hollin, C. & Howells, K. (1991). Clinical Approaches to Sex Offenders and Their Victims.
John Wiley & Sons: Chichester, England.
Hollingshead, A.B. (1975). Four factor index of social status. Unpublished manuscript, Yale
University, New Haven.
Horley, J. (1988). Cognitions of child sexual abusers. The Journal of Sex Research. 25 (41,
542-545.
Howells, K. (1978). Some meanings of children for pedophiles. In M. Cook & G. Wilson
(Eds ), Love and Attraction (pp.57-82). London: Pergamon Press.
Howells, K. (1981). Adult sexual interest in children: Considerations relevant to theories of
aetiology. In MCook & K. Howells (Eds), Adult Sexual Interest in Children (pp.55-94).
New York: Academic Press.
Knight, R_, Rosenberg, R, & Schneider, B. (1985). Classification of sex offenders:
Perspectives, methods, and validation. In A.W. Burgess (Ed ), Rape and sexual assault: a
research handbook (pp. 223-293). New York: Garland.
Lipton, D.N., McDonel, E.C., & McFall, RM. (1987). Heterosocial perception in rapists.
Journal of Consulting and Clinical Psychology. 55 (11.17-21.
Lorr, M., Youniss, R, & Stefic, E. (1991). An inventory of social skills. Journal of
Personality Assessment 57(31, 506-520.
Marshall, W.L., Barbaree, H.E., Christophe, D. (1986). Sexual offenders against female
children: Sexual preferences for age of victims and type of behaviour. Canadian Journal of
Behavioral Science. 18(41.424-437.
Marshall, W.L., Jones, R, Hudson, S.M., & McDonald, E. (1993). Generalized empathy in
child molesters. Journal of Child Sexual Abuse. 2(41. 61-68.
Marshall, W.L., Laws, D.R, & Barbaree, HE. (Eds.) (1990). The Handbook of Sexual
Assault: Issues. Theories, and Treatment of the Offender. New York: Plenum Press.
Mayer, A. (1988). Sex offenders: Approaches to understanding and management. Holmes
Beach, FL: Learning Publications, Inc.
McFall, RM. (1990). The enhancement of social skills: An information processing analysis. In
Marshall, W.L., Laws, D.R, & Barbaree, HE. (Eds.). The Handbook of Sexual Assault:
Issues. Theories, and Treatment of the Offender (pp. 1-30). New York: Plenum Press.

131
Mehrabian, A., & Epstein, N.A. (1972). A measure of emotional empathy. Journal of
Personality. 40, 523-543.
Merbaum, M. (1972). The simulation of normal and MMPI profiles by repressors and
sensitizers. Journal of Consulting and Clinical Psychology. 39. 171.
Miller, P. & Eisenberg, N. (1988). The relation of empathy to aggressive and
extemalizing/antisocial behavior. Psychological Bulletin. 103(3). 324-344.
Morrison, R.L., & Bellack, A S. (1981). The role of social perception in social skills. Behavior
Therapy. 12,69-79.
ODonohue, W., & Letoumeau, E. (1993). A brief group treatment for the modification of
denial in child sexual abusers: Outcome and follow-up. Child Abuse & Neglect. 17. 299-
304.
Overholser, J., & Beck, S. (1986). Multimethod assessment of rapists, child molesters, and
three control groups on behavioral and psychological measures. Journal of Consulting and
Clinical Psychology. 54153. 682-687
Pawlak, A.E., Boulet, J R, Bradford, J.M.W. (1991). Discriminant analysis of a sexual-
functioning inventory with intrafamilial and extrafamilial child molesters. Archives of
Sexual Behavior. 20(1), 27-34.
Putnam, F.W. (1985). Dissociation as a response to extreme trauma. In K.P Kluft (Ed ), The
childhood antecedents of multiple personality (pp. 66-97). Washington, DC: American
Psychiatric Press.
Quinsey, V.L. (1986). Men who have sex with children. In D.N. Weistub (Ed ), Law and
mental health: International perspectives (vol. 2; pp 140-172). New York: Pergamon
Press.
Riggio, R, Tucker, J., & CofFaro, D (1989). Social skills and empathy. Personality and
Individual Differences. 10(1), 93-99
Rogers, R, & Dickey, R (1991) Denial and minimization among sex offenders: A review of
competing models of deception. Annals of Sex Research. 4. 49-63.
Rook, K.S., & Hammer, C.L. (1977). A cognitive perspective on the experience of sexual
arousal. Journal of Social Issues. 33. 7-29.
Rosenthal, R, Hall, J.A., DiMatteo, M R, Rogers, P.L., & Archer, D. (1979). Sensitivity to
nonverbal communication: The PONS Test. Baltimore. Johns Hopkins University Press.

132
Salter, A. (1988). Treating Child Sex Offenders and Victims. Sage: Beverly Hills, CA.
Segal, Z.V. & Marshall, W. (1985). Heterosexual social skills in a population of rapists and
child molesters. Journal of Consulting and Clinical Psychology. 53, 55-63.
Segal, Z.V., & Marshall, W. (1985b). Heterosexual social skills in a population of rapists and
child molesters. Journal of Consulting and Clinical Psychology. 53.
55-63.
Segal, Z.V., & Marshall, W. (1986). Discrepancies between self-efficacy predictions and
actual performance in a population of rapists and child molesters. Cognitive Therapy and
Research- 1CX33. 363-376.
Segal, Z.V., & Stermac, L.E. (1990). The role of cognition in sexual assault In Marshall,
W.L., Laws, D.R., & Barbaree, H E. (Eds ). The Handbook of Sexual Assault: Issues.
Theories, and Treatment of the Offender (pp. 101-120). New York: Plenum Press.
Stermac, L.E., & Segal, Z. (1989). Adult sexual contact with children: An examination of
cognitive factors Behavior Therapy. 20, 573-584.
Wechsler, D. (1981). Wechsler Adult Intelligence Scale-Revised Manual. The Psychological
Corporation: Cleveland, OH.

BIOGRAPHICAL SKETCH
Julie Christine Medlin received her Bachelors of Arts degree from Harvard College in
1989. She worked for two years as a research assistant at the Center for Drug Treatment and
Research in Washington, D C She then enrolled in the doctoral program in the Department of
Clinical and Health Psychology at the University of Florida. She received her Master of
Science degree in 1993 and completed her clinical psychology internship in the Department of
Psychiatry, University of North Carolina at Chapel Hill, in 1995. Her PhD in clinical
psychology is expected to be granted in August, 1995. Ms. Medlin plans to pursue a career in
forensic psychology involving the treatment and evaluation of a sex offender population. She
also hopes to pursue a teaching and research career
133

I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
acqueiin Goldman, Chair
Professor of Clinical and Health Psychology
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
'h'i fo
Michael Perri
Professor of Clinical and Health Psychology
1 certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
4^$ \aS\J2~-
James Johnson / \
Professor of Clinical-and Health Psychology
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
jbA, .
James Algina
Piofessor of Foundatii
;of Education

I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
Professor of Psychology
This dissertation was submitted to the Graduate Faculty of the College of Health
Related Professions and to the Graduate School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
December, 1995
Dean, College of Health Related Professions
Dean, Graduate School




I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation
for the degree of Doctor of Philosophy.
Professor of Psychology
This dissertation was submitted to the Graduate Faculty of the College of Health
Related Professions and to the Graduate School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
December, 1995
Dean, College of Health Related Professions
Dean, Graduate School


16
this lack of research, the role of these cognitions remains unclear, particularly in regards to how
they interact with other variables to facilitate the commission of the offense (Quinsey, 1986).
Other Variables Related to Decoding Skill and Cognitions
Repression
Prior to the past decade, child molesters denial and defensiveness was viewed as
related to massive repression and unconscious denial (Rogers & Dickey, 1991). Rogers &
Dickey (1991) suggest that the pathogenic model was largely discarded in the past decade,
however, as a result of growing disenchanment with psychoanalytic theory and its
problems with clinical efficacy rather than on the basis of research findings. The results of
the study by Becker, Kaplan, and Tenke (1992), however, raise the possibility that
repression may play a role in child molesters denial since the deniers were sexually
nonresponsive to all sexual cues presented Although it is possible that the deniers were
able to consciously control their sexual arousal, repression might better account for these
deniers' ability to suppress their sexual impulses to prevent them from physiologically
responding to sexual cues. That is, the denial may involve more than a conscious attempt
to deny the crime in order to avoid perceived negative legal consequences; instead, or in
addition, it may involve an unconscious defense mechanism employed as a means of
controlling anxiety and guilt.
People who use repression as a primary defense mechanism often "distort their own
self-evaluative responses as a way to make themselves more comfortable, to block perceptions
of behavioral outcomes, and to avoid disturbing or uncomfortable implications in a variety of
situations (Dahlstrom, p. 104)." Thus, repression might help the child molester deny the


Analysis of Empathy
Child molesters were expected to score lower on empathy than controls, and
68
deniers were expected to score lower than admitters Three outliers were identified and
removed from the analysis. Contrary to expectations, a planned comparison revealed no
significant difference between child molesters and controls on the empathy total score
F(l,199)=2.917, p=.089. A planned comparison showed that the nonfamilial sex
offenders scored significantly higher on empathy total score than the incestuous sex
offenders, F(l,199)=5.889, p=.016 (see Table 18). Contrary to predictions, there was no
significant difference between admitters and deniers on total empathy score,
F(l,199)=.623, p= 431. An ANCOVA revealed a significant main effect of group
F(5,199)=3 007,p=.012 (see Figure 25). The covariate was not significant,
F( 1,199)=.724, p=.396. All possible pairwise comparisons using the Bonferroni correction
showed that the nonfamilial deniers scored significantly higher than the incestuous deniers
on the total empathy score (p=.026).
In order to examine differences among the six groups on the four IRI subscales, an
ANOVA was conducted with group as the between subjects variable and the four subscale
scores as the repeated measures. The Group X Subscale interaction was not significant,
F(5,205)=1.36, p=. 162, indicating that differences among the groups did not vary across
subscales.
Analysis of Repression
It was expected that child molesters would score higher on repression than
controls, and that deniers would score higher on repression than admitters. These


Mean Woman Harm Ratings
91
Level of Sexual Contact
Child Molesters
Controls
Figure 22. Mean Woman Harm ratings for levels of sexual contact.


57
molesters scored significantly different from controls in their ratings across the levels of
the childs response, F(2,404)=7.186, p=.001. The difference between between admitters
and deniers was also significant, F(2,404)=4.327, p=.014 (see Figure 9), as was the
difference between incestuous and nonfamilial offenders, F(2,404)=4.750, p= 009 (see
Figure 10).
Child molesters did not differ from controls in their desire ratings when the child
smiled, F(l,202)=3.440, p=.065, or when the child was passive, F(l,202)=1.378, p=.242.
The child molesters ratings were significantly lower than those of controls, however,
when the child cried, F(l,202)=4.449, p=,036 (see Figure 11). Nonfamilial offenders did
not score significantly different from incestuous offenders when the child smiled,
F(l,202)=2.665, p=. 104, was passive, F(l,202)=3.679, p=.057, or cried; F(l,202)=1.081,
p=.300.
The post-hoc comparison between the nonfamilial admitters and the five groups
approached significance using the Scheffe test, F(2,404)=16.234. Nonfamilial admitters
ascribed more desire to the child than the other groups when the child smiled,
F(l,202)=15 262, p=,000, but not when the child was passive, F(l,202)= 7.883, p=.005 or
cried (Table 10). Figure 11 illustrates that nonfamilial admitters showed the same pattern
on this variable as on Child Benefit and Enjoyment. Nonfamilial admitters attributed
significantly more desire to the child than did the other groups when the child smiled, but
scored the same as others when the childs response was ambiguous or clearly negative
Groups differed significantly in their ratings of the childs responsibility,
F(5,205)=2.703, p=.022. There was also a significant Group X Child interaction,


61
results of the Stermac and Segal (1989) study. Correlation coefficients ranged from .008
to .052.
Adult vignettes
Principal components analysis. The data from the adult vignettes was analyzed in
the same manner as the child vignettes. A principal components analysis was conducted
on the nine variables collapsed across the levels of sexual contact and the womans
response. The analysis yielded two factors which exceeded the eigenvalue criterion of 1.
Factor 1 accounted for 43.37% of the total variance while Factor 2 accounted for 17.38%
of the variance. The component loadings listed in Table 12 indicate that these two factors
are similar to the two factors derived on the child vignettes. The females Desire,
Enjoyment, Responsibility, and Benefit loaded highly on Factor 1 in both the child and
adult vignettes. In addition, Factor 2 seems to tap the mans role in both vignettes. On the
adult vignettes, Factor 2 taps the Mans Anxiety, Confidence, and Responsibility for the
sexual interaction. Although some of these specific variables were not included on the
child vignettes, these variables focus on the mans role as do the variables which loaded
highly on Factor 2 on the child vignettes.
Principal components analyses were conducted for each of the 12 vignettes. The
analyses yielded 2 to 3 factors for each vignette, similar to the factor structure found in the
analysis of the nine means. Within each vignette, the first factor accounted for 33.01%-
42.03% of the variance and the second factor accounted for 13 05%-19.89% of the
variance. These results are consistent with the results of the analysis using the nine means.


REFERENCES
Abel, G.G., Becker, J.V., & Cunningham-Rathner, J. (1984) Complications, consent, and
cognitions in sex between children and adults. International Journal of Law and Psychiatry.
7, 89-103.
Abel, G.G., Blanchard, E.B., & Barlow, D.H. (1981). Measurement of sexual arousal in
several paraphilias: The effects of stimulus modality, instructional set and stimulus content
on the objective. Behaviour Research and Therapy. 19(11. 25-33.
Abel, G.G., Gore, D.K., Holland, C.L., Camp, N., Becker, J.V., & Rathner, J. (1989). The
measurement of the cognitive distortions of child molesters. Annals of Sex Research. 2,
135-153.
Abel, G.G., Mittleman, M S., & Becker, J. (1985). Sexual offenders: Results of assessment
and recommendations for treatment. In M.H. Ben-Aron, S.J. Hucker & C D. Webster
(Eds), Clinical criminology: The assessment and treatment of criminal behaviour (dp. 191-
205). Toronto, University of Toronto and M & M Graphics, Ltd.
Abel, G., Rouleau, I, & Cunningham-Rathner, J. (1985). Sexually aggressive behavior In W.
Curran, A.L. McGarry, & S.A. Shah (Eds.), Forensic psychiatry and psychology:
Perspectives and standards for interdisciplinary practice (pp. 289-313). Philadelphia: F A.
Davis.
Barbaree, H E., & Marshall, W.L. (1989). Erectile responses among heterosexual child
molesters and matched non-offenders: Five distinct age preference profiles. Canadian
Journal ofBehavioral Science. 21(11. 70-81.
Becker, J., Kaplan, M., & Tenke, C. (1992). The relationship of abuse history, denial, and
erectile response profiles of adolescent sexual perpetrators. Behavior Therapy. 23, 87-97.
Bliss, E.L., and Larson, E.M. (1985) Sexual criminality and hypnotizability. Journal of
Nervous and Mental Disease. 173. 522-526.
Briere, J., & Runtz, M. (1989). Symptomatology associated with childhood sexual
victimization in a non-clinical adult sample. Child Abuse & Neglect. 12, 51-59.
127


56
the child was passive, F(l,202)=1.507, p= 221, but did differ when the child cried,
F(l,202)=4.873, p=.028. Child molesters gave higher child enjoyment ratings than
controls when the child smiled, but gave lower ratings than controls when the child cried.
Nonfamilial offenders ascribed more enjoyment to the child than did incestuous offenders
when the child smiled, F(l,202)=4.798, p=.030, and was passive, F(l,202)=4.153, p=.043,
but did not differ from incestuous offenders when the child cried, F(l,202)=.065, p=.799.
The post-hoc comparison between nonfamilial admitters and the other five groups
approached significance on the Group X Child interaction, F(2,404)=15.706, using the
Scheffe test. Nonfamilial admitters attributed significantly more enjoyment to the child
than the other five groups when the child smiled, F(l,202)=17.561, p=.000, but not when
the child was passive, F(l,202)=7.839, p=.006, or when the child cried, F(l,202)=1.444,
p=.231 (see Table 7, Figure 7).
Planned comparisons showed no significant differences between child molesters
and controls on their ratings of the childs desire for the sexual interaction, F(l,202)=.713,
p=.399 There were also no significant differences between admitters and deniers,
F(l,202)=1.438, p=.232, or between incestuous and nonfamilial offenders,
F( 1,202)= 1.780, p=.184.
Groups differed significantly, however, in their ratings of the childs desire
depending on the childs response. This was indicated by a significant Group X Child
interaction on Child Desire, F(10,404)=4.101, p=.000. Groups differed significantly when
the child smiled, F(5,202)=3.425, p=.005, but not when the child was passive or cried (see
Figure 8). Planned comparisons on the Group X Child interaction showed that child


129
Feldman, R. (1992). Applications of Nonverbal Behavioral Theories and Research. Lawrence
Erlbaum Associates: Hillsdale, N.J.
Feldman, R. & Rime, B. (1991). Fundamentals of Nonverbal Behavior. Cambridge University
Press: Cambridge.
Feshbach, N. (1978). Studies of empathic behavior in children. In B.A. Maher (Ed.), Progress
in Experimental Personality Research (Vol. 8, pp. 1-47). New York: Academic Press.
Finkelhor, D. (1984). Child Sexual Abuse: New Theory and Research. New York: Free Press.
Finkelhor, D. (1986). A Sourcebook on Child Sexual Abuse. Beverly Hills, CA: Sage.
Finkelhor, D., & Lewis, I. (1988). An epidemiologic approach to the study of child
molestation. In Prentky, R. & Quinsey, V (Eds ), Annals of the New York Academy of
Sciences (Vol. 528, pp 64-78) New York: New York Academy of Sciences.
French, D. (1988). Distortion and lying as defense processes in the adolescent child molester.
Journal of Offender Counseling Services and Rehabilitation 13111. 27-37.
Frisbie, Vanasek, F., & Dingman, H. (1967). The self and the ideal self: Methodological study
of pedophiles. Psychological Reports. 20, 699-706.
Frischholz, E., Braun, B., Sachs, R., Schwartz, D. (1992). Construct validity of the
Dissociative Experiences Scale: H Its relationship to hypnotizability. American Journal of
Clinical Hypnosis. 35121. 145-152.
Groth, AN., & Bimbaum, H.J. (1979). Men who rape The psychology of the offender. New
York: Plenum.
Hall, J.A. (1979). Gender, gender roles, and nonverbal communication skills. In R. Rosenthal
(Ed.), Skill in nonverbal communication: Individual differences. Cambridge, MA:
Oelgeschlager, Gunn & Hain.
Hamilton, D., & Ondrovik, J. (1993, April 29-May 2). Dissociation in pedophile sex
offenders. Paper presented at the Ninth Annual Symposium in Forensic Psychology,
Santa Fe, NM.
Haywood, T.W., & Grossman, L.S. (1994). Denial of deviant sexual arousal and
psychopathology in child molesters. Behavior Therapy. 25, 327-340.
Haywood, T.W., Grossman, L.S., & Hardy, D W. (1993). Denial and social desirability in
clinical evaluations of alleged sex offenders. The Journal of Nervous and Mental Disease.
181(31. 183-188.


Mean Woman Enjoyment Ratings
87
Woman Smiling Woman Passive Woman Crying
Level of Sexual Contact
l=Touching 2=Fondling 3=No Clothes 4=Ejaculation
Figure 18. Mean Woman Enjoyment ratings for levels of sexual contact and childs response.


60
largest when the the sexual contact involved touching only, and this difference decreased
as the level of sexual contact increased. There were no significant differences between
nonfamilial and incestuous offenders, F(3,606)=1.968, p=. 118, or between admitters and
deniers, F(3,606)=1.736, p=. 158 on the Group X Sex interaction.
There were no significant differences among groups on Child Harm,
F(5,202)=.662, p=.653 and there were no significant interactions with Group. There were
also no significant differences among groups on Man Responsibility F(5,202)=1.227,
p=.298 and no significant interactions with Group.
There were no significant differences among the groups on Treatment,
F(5,202)=l 652, p=.148, however, child molesters differed significantly from controls on
the Group X Sex interaction, F(3,606)=4.451, p=.004. Figure 15 shows that the child
molesters consistently prescribed less treatment than controls across all levels of sexual
contact. The difference between groups was significant when the sexual contact involved
touching only, F(l,202)=9.068, p=.003, but not when the sexual contact increased to
rubbing F(l,202)=2.593, p=. 109, fondling F(l,202)=2.966, p=.087, or ejaculation
F(l,202)=1.784, p=. 183, This pattern was similar to the child molesters pattern of
ratings on the variable of punishment. These results suggest that the finding of differences
between child molesters and controls on Factor 2, or Seriousness of Offense, can be
attributed to child molesters different views on punishment and treatment, rather than to
differences in ratings of the mans responsibility or harm to the child.
As expected, there were no significant correlations between each of the nine
variables and scores on the Social Desirability Scale. This finding is consistent with the


122
their victims. It may also be that child molesters perceive themselves as empathic, and
therefore rate themselves as empathic on self-report measures of empathy. Thus, it might
be important to use behavioral measures rather than self-report measures of empathy for
this particular group. A future study might examine the differences between child
molesters self-perception of their empathy versus others perceptions of their capacity for
empathy.
Although there were no differences between child molesters and controls on
empathy, there was the unexpected finding that nonfamilial sex offenders scored
significantly higher on empathy than incestuous offenders. It is possible that nonfamililial
offenders may have greater empathy than incestuous offenders, however, it is more likely
that nonfamilial offenders just report themselves as being more empathic. Thus, this
finding might reflect a lack of insight regarding their own capacity for empathy This is
probably the case since the nonfamilial offenders responses on the child and adult
vignettes revealed a lack of empathy for the victims perspective. Unlike the incestuous
offenders, the nonfamilial admitters rated the child as benefitting more from the sexual
contact, as well as desiring and enjoying it more. Thus, it is not surprising that they might
also misperceive themselves as being empathic and attuned to the childs desires. If this is
the case, this finding might actually suggest that not only are nonfamilial offenders not
more empathic than incestuous offenders, they also show yet another cognitive distortion
compared to incestuous offenders.
Contrary to expectations, there were no differences between admitters and deniers
on empathy. Thus, level of empathy appears to be significantly related to the relationship


TABLE OF CONTENTS
Pace
ACKNOWLEDGEMENTS ii
ABSTRACT vi
CHAPTERS
1 INTRODUCTION AND REVIEW OF LITERATURE 1
Social Skills of Child Molesters 1
Cognitions of Child Molesters 6
Subtypes ofChildMolesters 10
Other Variables Related to Decoding Skill and Cognitions 16
Rationale for the Proposed Study 20
Hypotheses 23
Alternative Hypotheses 24
2 METHODS 26
Subjects 26
Research Design 27
Dependent Variables 28
Procedures 39
3 RESULTS 41
Background Variables 41
Method of Analysis 44
Analysis of Decoding Skill 46
Analysis of Cognitive Distortions 47
Comparison of Child Versus Adult Vignettes 66
Analysis of Empathy 67
Analysis of Repression 68
4 DISCUSSION 110
Summary of Findings 125
iv