Sexual offenders' perceptions of their early interpersonal relationships: an attachment perspective.

In recent years, researchers and clinicians have focused increasingly on the nature of sex offenders' intimate relationships and their early attachment experiences (Marshall, 1989). Basically, three overlapping but distinct lines of research can be identified. The first area of research is concerned with investigating sexual offenders' adult romantic relationships and the interactions between attachment style, intimacy deficits, and offending behavior (Ward, Hudson, & Marshall, 1996; Ward, McCormack, & Hudson, 1997). The second line of research is concerned with offenders' early developmental experiences, documenting the variables that predict later sexual aggression (Prentky et al., 1989). The third type of research has examined adolescent sexual offenders' interactions with their caregivers and peers (e.g., Blaske, Borduin, Henggeler, & Mann, 1989). Despite the value of the above research findings, the link between the quality of early interpersonal relationships and interpersonal dysfunction in adult offenders is not well understood. This study represents an attempt to integrate the recent findings from these different areas of research using attachment theory as a basic explanatory framework.

According to attachment theory, individuals' interpersonal schemas and strategies are shaped by their cumulative experiences with other people. Furthermore, it is argued that the interpersonal relationships that have the most influence on a person's working model of relationships are his or her parents or primary caregivers. Therefore, in view of the apparent difficulty sex offenders have in establishing intimate relationships with other adults, it has been maintained that they had destructive and negative relationships with their parents. In fact, the literature suggests that sexual offenders typically perceive their mothers more positively than their fathers, although this difference appears only to be one of degree (Hazelwood & Warren, 1989; Tingle, Barnard, Robbin, Newman, & Hutchinson, 1986). For example, in one study 36% of sex offenders described their relationship with their mothers as warm and close, and a further 31% described their mothers as cold, distant, uncaring, indifferent, hostile, and aggressive (Hazelwood & Warren, 1989). Sexual offenders' perceptions of their mothers appear to range from positive to negative, a finding that provides us with little predictive utility. Further information on this issue is provided by Tingle et al. (1986), who found that although the majority of child molesters (83%) reported that their relationships with their mothers were close, only a quarter of these described their mother as someone to whom they could turn to with a problem (Tingle et al., 1986). These authors suggested that the relationship between child molesters and their mothers is best characterized as dependent rather than reciprocal in nature.

A number of more specific difficulties have been noted in the relationships between sexual offenders and their mothers. Blaske et al. (1989) compared adolescents who committed a sexual offence with other nondelinquent adolescents and found lower rates of positive mother-son communication in the sex-offender group. With respect to differences across sexual offender types, rapists were found to have significantly more arguments with their mothers than child molesters (Tingle et. al., 1986). There is also evidence that sexual offenders identify less with their mothers than do members of other offender groups (Levant & Bass, 1991).

Traditionally, the role of the father in the etiology of an individual's sexual offending is seen as insignificant (Tingle et al., 1986). This perspective may have originated from the absence and lack of involvement of fathers in the early upbringing of many sexual offenders. However, the picture appears to be more complex than these results suggest. Of those sexual offenders who reported a father present during their childhood, the relationship between the father and the individual concerned was typically described as more problematic and negative than that between mother and son (Lisak & Roth, 1990). Specifically, a large percentage of sexual offenders (57%) described their fathers as cold, distant, hostile, and aggressive, with fewer (18%) crediting their fathers with positive qualities such as warmth (Lisak, 1994). This negative perception may be related to the high rates of physical abuse inflicted by both biological fathers and stepfathers on sexual offenders (Kahn & Chambers, 1991). Moreover, sexual offenders appear to identify less with their fathers than do other offender groups (Levant & Bass, 1991). In turn, rapists' relationships with their fathers have been reported to be more distant than child molesters (Tingle et al., 1986). A negative view of the relationship between rapists and their fathers was associated with a need for power and control, as well as with anger and hostility toward women (Hazelwood & Warren, 1989; Lisak & Roth, 1990). In contrast, child molesters reported equivalent rates of maternal and paternal rejection to nonoffender groups (Marshall & Mazzucco, 1995). In combination, these findings suggest that fathers of sexual offenders do play a significant role in the development of sexually aggressive tendencies. This may be a function of either their lack of involvement in the upbringing of their sons or of the violence they inflicted upon them.

Another important source of disruption to early interpersonal relationships is the loss of caregivers. In a study by Ryan and Lane (1991), over half of their juvenile sexual offenders were found to have experienced some form of parental loss through death, divorce, or separation. Sexual offenders may be less likely than nonsexual offenders to have an intact family of origin, and this fact may be partly responsible for their subsequent interpersonal problems. However, the research on this issue is rather patchy. There is some confusion over whether rapists or child molesters are more likely to have parents with an intact marriage (Seghorn, Prentky, & Boucher, 1987; Tingle et al., 1986), although the available data suggests that the parents of rapists are less likely to be legally married in the initial instance than those of child molesters (Saunders, Awad, & White, 1986).

Regardless of whether one or both parents are present in the family home, the environments of sexual offenders are characterized by many features which have the potential to damage the quality of early interpersonal relationships. One such feature is the presence of physical abuse, which has been reported at high rates in the histories of sexual offenders (e.g., Ryan & Lane, 1991). This abuse is most often carried out by biological fathers (44%) and stepfathers (20%) (Kahn & Chambers, 1991). The presence of physical abuse is unlikely to be specific to sexual offenders, as rates also tend to be high for nonsexual offenders (Lewis, Shanock, & Pincus, 1981). Interestingly, physical abuse tends to be more predictive of nonsexual aggression than sexual offenses (Prentky et al., 1989). It is reasonable to conclude that the experience of physical violence is likely to result in the development of insecure attachment and the associated beliefs that relationships are inherently dangerous and other people unreliable.

The occurrence of sexual abuse as a child may occur in combination with physical violence or exist as a separate problem. Estimates of the prevalence of sexual abuse in sexual offenders range from 9% to 47% (Fagan & Wexler, 1988). Milner and Robertson (1990) noted that family sexual abuse is more common in the family backgrounds of sexual offenders than other types of offenders. Most researchers agree that sexual abuse is more than twice as likely to be present in a sexual offender as in a nonsexual offender. Across subtypes of sexual offenders, child molesters are approximately twice as likely to be sexually abused than rapists (Seghorn et al., 1987). However, rapists are more likely to have been abused by a family member while child molesters are more likely to have been abused by nonfamily members (Seghorn et al., 1987). Thus, a history of sexual victimization and sexual deviation within the home has been found to be highly predictive of sexual aggression (Prentky et al., 1989). Furthermore, the nature of sexual abuse means that it is capable of disrupting attachment in multiple ways: for example, reducing individuals' ability to regulate negative affective states or leading to the establishment of distrustful attitudes toward others.

The consistency of caregiver availability and response has important implications for attachment style. The only study to specifically examine this variable used the term caregiver inconstancy to describe the stability of the primary caregiver relationship, as measured by the length of time spent with a single caregiver (Prentky et al., 1989). Prentky and his colleagues concluded that caregiver inconstancy was highly predictive of the severity of future sexual aggression. Other variables which have been found to be common in the parents of sexual offenders--adversely affecting their ability to provide consistent, predictable care-giving--include problems with substance abuse (e.g., Ryan & Lane, 1991) and involvement in the criminal justice system (e.g., Fagen & Wexler, 1988).

Once established, attachment styles can prove refractory to change because of the tendency to result in self-confirming consequences. For example, an avoidantly-attached individual, who tends to isolate himself from other people, may fail to develop the necessary skills to establish intimate relationships. Therefore, any overtures to others are likely to be inappropriate and to result in rejection or adverse consequences. This will be interpreted by the person concerned as evidence that others are unreliable, and that it is safer for him to avoid future involvements. Therefore, those offenders who are insecurely attached to their caregivers are likely to have dysfunctional relationships with other individuals, such as peers. Blaske et al. (1989) concluded that sexual offenders typically have lower levels of emotional bonding with peers. Similarly, Tingle et al. (1986) reported that 86% of adult rapists and 74% of child molesters in their sample had few or no friends when young. The importance of peer groups for the development of adult attachment patterns, particularly in adolescence, is currently the focus of theoretical and empirical research in the attachment area (Hazan & Shaver, 1994).

In summary, there is ample evidence that sexual offenders have experienced the types of adverse early events frequently associated with the development of insecure attachment styles. Sexual offenders typically have negative relationships with both of their parents, identify with their parents less than do other offender groups, experience high rates of physical and sexual abuse, are more likely to experience the loss of caregivers, are less likely to have stable and constant relationships with caregivers, and typically communicate less with their parents. The adult attachment styles of these men are likely to reflect these cumulative experiences with relationships and to lead to problematic interpersonal expectancies, goals, and strategies.

RATIONALE AND RESEARCH QUESTIONS

The overall goal of this study was to examine offenders' perceptions of their early interpersonal experiences using grounded theory (a qualitative method). We know comparatively little about the way offenders represent their past experiences, and even less about the differences, if any, between sexual offenders' views of their early interpersonal relationships compared with the views of other types of offenders. Therefore, it seemed appropriate to use a method that would allow for maximum exploration of the derived data rather than impose an a priori set of categories on offenders' responses.

We had three specific aims in conducting this study. First, we sought to develop a fine-grained description of sexual offenders early experiences, particularly those aspects associated with their offense-related problems (i.e., emotional regulation difficulties, impaired autonomy, level of interpersonal trust, experiences of loss and caregiver inconsistency, and the degree to which offenders evaluated themselves in a negative manner). Second, we were curious whether or not sexual offenders, compared to other types of offenders, exhibited differences in the way they construed their early interpersonal experiences. The aim in investigating this issue was to establish whether there were differences between the different types of offenders in their early experiences, or rather, in their perceptions of past experiences. Third, we wanted to ascertain whether offenders had different perceptions of their early relationships with their mothers and fathers.

METHOD

A semi-structured interview format was used, as there is persuasive evidence that this type of assessment can provide a more accurate picture of individual's perceptions of their close relationships than can questionnaire data (e.g., Ward et al., 1997). Accuracy may be enhanced because individuals' interpersonal schemas are often tacit and therefore not easily accessible by direct questioning. They may not be aware of their underlying relationship models, and when asked directly about these may unconsciously distort their responses in some way. By allowing offenders to respond more freely to open-ended questions it is possible to develop a richer understanding of their underlying beliefs and models.

Participants and Setting

The child molester participants in this research were involved in the Kia Marama Sex Offender Treatment Program (Hudson, Marshall, Ward, Johnston, & Jones, 1995), operating within a medium security prison in New Zealand. The majority of these men participated before treatment. The other groups, sexual offenders with adult victims (referred to as rapists for convenience), violent nonsexual offenders (the violent group), and nonsexual, nonviolent offenders (the nonviolent group) were recruited from other areas of the same medium security prison as the child molesters and another medium security facility within the same geographic area. None of the men from the comparison groups was in treatment.

The study participants consisted of 55 men who had offended sexually against children, 30 men who had offended sexually against adult women, 32 men with violent offenses, and 30 men incarcerated for neither sexual nor violent offenses. This classification was exclusive. The offense records of all participants were reviewed to ensure that men in the child molester group had no offenses against adult victims, rapists had no offenses against children, men in the violent group had no sexual offenses, and, finally, men in the nonviolent group had neither sexual nor violent offenses in their criminal histories. The offenses committed by the child molesters ranged from masturbating in front of a child to completed intercourse or sodomy with force. Offenses by the rapists ranged from sexual violations (indecent assaults) to a predatory, sadistic rape of an adult woman, with high levels of force and violence. The violent offenders had been convicted of offenses ranging from a fight in a bar to a premeditated and excessively violent murder. The nonviolent offenders reflected convictions for driving offenses, drug-related crimes, burglary, and fraud.

There were significant differences between the groups in age, with child molesters being significantly older than all the other groups and rapists being significantly older than both the violent and nonviolent groups. Similarly, there were significant differences across the groups in length of offending history, with rapists and child molesters showing the most extensive history, then violent offenders, and finally nonviolent offenders showing the shortest history. Finally, there were also significant differences between the groups with respect to length of current sentence, with rapists and violent offenders not being discriminable from each other but with the longest sentences, followed by child molesters, and finally the nonviolent offenders with significantly shorter sentence lengths. These differences between the groups were as expected and were not germane to the research hypotheses.

Procedure

In this study, we used a qualitative method, grounded theory, to transform the descriptive data generated by the interview into categories. Grounded theory consists of a set of systematic procedures that seek to inductively derive a theory, or a set of categories, from qualitative data (Strauss & Corbin, 1990). Typically, concepts are inductively derived from an initial set of qualitative descriptions or scripts, which, once coded into rudimentary conceptual categories, lead to the collection either of more descriptions or of quantitative data. The next step is the deduction of predictions or hypotheses concerning the ability of the provisional categories to account for new protocols. If they fail to accommodate the new data then new categories are formulated and the process continues. Therefore, there is a progressive development of categories as the research project unfolds. Researchers guided by the grounded theory approach are able to add questions, choose different samples, and explore significant areas of interest depending upon the results of preliminary data analysis. The whole process of category building is dynamic and extremely sensitive to patterns detected in the data.

The current understanding of the nature and quality of early interpersonal experiences in sexual offenders, and their relationship with sexual crimes, is still quite limited. In view of this, it was believed to be a useful strategy to let the offenders' relatively unstructured descriptions of their early interpersonal experiences provide the basis of category development. A unique feature of the approach in this study is the use of both qualitative and quantitative methods in the generation and application of the categories. This approach has been successfully used before to generate a description of intimacy deficits in sexual offenders (Ward et al., 1997).

Each of the participants was interviewed about his early interpersonal experiences. The format for the interviews was in part a semi-structured one, with 35 questions being selected from the attachment research literature (e.g., Hazan & Shaver, 1994). These covered parental responsiveness to negative emotion, attention and support, consistency of parental behavior, perception of safety and security within the home environment, degree of supportiveness or respect for them as individuals, and an overall evaluation of the relationship--for example, "How would your mother (and father if present) respond if you got upset?" (responsiveness), "Do you feel they treated you as an individual? (respect for individuality), and "Did they tend to deal with you in a consistent way?" (parental consistency). The interviewer was a graduate clinical psychology student who was experienced in interviewing offender groups. Each interview took approximately 1 hour.

In step one the interview summaries, which contained no identifying details with respect to offender type, were subjected to a grounded theory analysis in order to develop a set of descriptive categories concerning offenders' perceptions of their early interpersonal relationships (Strauss & Corbin, 1990). Each of the categories was then dimensionalized on a 7-point Likert scale. A member of the research team (SH) who did not play a part in their original development took a 20% sample of the protocols as a check on the degree of category "saturation" (content validity). He read each interview transcript and noted whether the data could be fitted into the categories.

In step two, the categories were used to rate each participant's responses on the associated dimensions. As a check on the reliability of rating, a second researcher (KM) rated a random sample of 20% of the protocols on the various dimensions. There were no identifying details about offender type on the protocols, and therefore the researchers performing the two reliability checks were blind to which groups the subjects belonged.

RESULTS

Step One: Category Development

The grounded theory approach (Strauss & Corbin 1990) was applied to a comprehensive list of the responses provided by the participants to questions about their early interpersonal experiences. Initial coding of the raw information into basic categories occurred by grouping together the responses of similar meanings. The resulting categories were collapsed into 14 final conceptual categories, which were each dimensionalized using a 7-point Likert scale. This allowed us to rate each participant's degree of fit within each category. The categories were as follows.

Caregiver relationship variables. Four initial categories emerged which related to aspects of the parent's behavior and style. These categories were rated separately for the mother and father. The first of these was Responsiveness, which encompassed perceptions surrounding the caregiver's responsiveness, availability, and support, with scale anchors of responsiveness and neglect at either end (1= neglect; 7= responsive). The anchoring of each Likert scale in this manner was done for each category. An example of responsiveness was "She would comfort me and listen to me"; an example of neglect was "She wouldn't notice" and "I can't remember any help or support." A category of Consistency emerged as another construct not reducible by more abstract analysis, and thus became the second category. Included under this heading were features of the caregiver's' behavior relating to their consistency, reliability, and predictability. Perceptions such as "Life was unpredictable day by day, didn't know what was happening" or "Mum was up and down like a yoyo" were rated as very inconsistent on this dimension. Alternatively, a very consistent caregiver would be described with perceptions like "He was consistently abusive" or "I could always rely on her word." A category called Acceptance emerged to describe the extent to which participants felt accepted, loved, and approved of by their parents as well as their general experience of feeling part of the family. A participant was rated as experiencing rejection on this dimension if he said something like "Father picked on me" or "I was the black sheep of the family" while those who received a rating of acceptance may have responded with "I felt welcomed and loved." The anchors were rejection and acceptance respectively. Boundaries was the name given to the category that described the approach to parenting by caregivers with particular respect to supervision and discipline. One anchor was Lax/under controlled parenting ("I didn't know where I stood") with the other being firm ("My parents had fair ground rules").

Self variables. The next three categories to emerge reflected aspects of the individual's functioning within the attachment relationship. The first of these was a category described as Emotional Regulation, which included the participants' perceptions about their identification and modulation of emotion, especially the degree to which emotional states were shared with others and support gained. One pole was labeled defended ("If I was in a mood I wanted space") and the other pole labeled expressive ("I would go to mum when upset"). The individual's sense of autonomy, advocacy for their rights and personal opinions, and experiences of separateness and mastery were grouped together under the category of Autonomy. The anchors here were autonomous ("I would go exploring by myself") and enmeshment ("I was very protective of mother"). Self-evaluation was a category that captured the participant's general evaluation and perceptions of the self (as a child) and the anchors were negative ("I was very under-confident") and positive ("I did well at school" or "I was a skating champ, good at archery too").

Contextual variables. Reflecting more general aspects of the family environment and developmental context were a number of categories described as contextual variables. Initially, sexual abuse, physical abuse, and parental loss were all coded under one overarching category called Developmental Trauma; however, it became clear as analysis progressed that these were conceptually different factors and as such warranted separate categories. Sexual Abuse and Deviation was a category that incorporated perceptions of sexual deviation and abuse. The anchors were absent, where there were no indications of exposure to sexual abuse or sexually deviant behavior, and extreme ("I blocked out a lot of early memories due to sexual abuse"). Similarly, a category called Physical Abuse emerged to account for physical violence. Again, the anchors were absent (no mention) and extreme ("He beat the shit out of us for nothing all the time"). The category labeled Loss included events such as loss of parents, grandparents, pets and friends, with anchors of absent and severe ("felt really cut up when mother died"). The categories of Level of Conflict, Degree of Disunity, Violence Between Parents, and Hostile Atmosphere were initially coded under separate categories. Subsequently these were incorporated within an overarching category labeled Conflict. The anchors were absent ("home life was okay, no problems") and substantial where there were few skills to resolve disputes positively and a hostile atmosphere evident within the home ("My mother had a turbulent relationship with my stepfather" or "Everyone was mad--violence all the time").

Perceptions relating to safety and security within the home environment, and the degree to which the participant's need for security was met, were simply labeled Safety. This category varied from safe ("felt safe and secure, had no fears, nothing unpredictable happened") to dangerous ("I constantly feared Dad would find out I'd done something wrong" or "I never felt safe so didn't spend much time at home"). A final category, Positive Mediating Interactions, encompassed constructive interpersonal experiences that occurred outside of the primary caregiver relationships and which provided the individual with an experience of positive attachment. The anchors were absent and substantial (e.g., positive interactions with grandparents).

The content validity checks (category saturation) were satisfactory with no piece of interview data being unable to be coded within the 14 categories. As a further reliability check, all of the data were independently coded on the 7-point Likert scales derived from the categories. The results were acceptable, with an overall reliability coefficient of r = .79. Coefficients between the two raters varied from .60 to .90 for caregiver relationship variables, from .57 to .82 for self variables, and from .77 to .95 for contextual variables.

Step Two: Offender Type Comparisons

Caregiver relationship variables. There were no significant differences among the four offender groups on the Responsiveness--Mother dimension, F(3,140) = .81, p = .49, ns, (M = 4.29, 4.33, 4.69, 4.17, and SD = 1.11, 1.56, 1.63, 1.32 for child molesters, rapists, violent offenders and nonviolent offenders respectively). However the groups did differ significantly on the Responsiveness--Father dimension F(3,131) = 4.55, p = .005, (M = 4.88, 5.72, 5.46, 4.77, and SD = 1.24, 1.19, 1.20, 1.14 for child molesters, rapists, violent offenders, and nonviolent offenders respectively). Post hoc testing (Fisher's PLSD) suggested that child molesters and nonviolent offenders were indiscriminable with respect to these scores, but that both of these groups described their fathers as significantly more responsive than did rapists and violent offenders. No differences were found for the Consistency--Mother dimension, F(3,140) = .51, p = .68, ns, (M = 3.37, 3.73, 3.69, 3.77, and SD = 1.47, 1.76, 1.78, 1.98 for child molesters, rapists, violent offenders, and non-violent offenders respectively) nor for the Consistency--Father dimension, F(3,131) = .80, p = .50, ns, (M = 3.58, 4.03, 4.14, 3.92, and SD = 1.58, 1.84, 1.82, 1.88 respectively). There were no significant differences across the groups on their ratings on the Acceptance--Mother dimension, F(3,140) = 1.25, p = .29, ns, (M = 4.31, 4.1, 4.76, 4.37, and SD = 0.98, 1.73, 1.48, 1.33 respectively) nor on the Acceptance--Father dimension, F(3,131) = 2.197, p = .09, ns, (M = 4.86, 5.21, 5.04, 4.35, and SD = 1.14, 1.52, 1.35, 1.29 respectively).

Finally, the four groups differed significantly on the Boundaries--Mother dimension, F(3,140) = 6.49, p = .0004, (M = 4.44, 3.3, 3.83, 4.37, and SD = 1.07, 1.34, 1.37, 1.22 respectively), with child molesters and nonviolent offenders being indiscriminable, but with firmer reported boundaries than the mothers of rapists. Child molesters were also rating more highly on this dimension than violent offenders. However, rapists were indiscriminable from violent offenders, who in turn were indiscriminable from nonviolent offenders. There was also a significant difference between the groups on the Boundaries--Father dimension, F(3,131) = 5.80, p = .0009, (M = 4.76, 3.62, 3.89, 4.46, and SD = 1.09, 1.35, 1.34, 1.48 respectively). Again, child molesters and nonviolent offenders were indiscriminable and their fathers were both rated as having significantly firmer boundaries than rapists, with the fathers of child molesters also being rated more highly on this dimension that violent offenders. Rapists were indiscriminable from violent offenders, who in turn were indiscriminable from nonviolent offenders.

Self variables. The four groups did not differ significantly on the Emotional Regulation dimension, F(3,139) = 1.44, p = .233, ns, (M = 2.63, 2.63, 2.13, 2.57, and SD = 1.02, 1.40, 0.9, 1.19 for child molesters, rapists, violent offenders and nonviolent offenders respectively). However, there was a significant difference between the groups with respect to the Autonomy dimension, F(3,142) = 6.11, p = .0006, (M = 3.98, 3.20, 3.40, 2.97, and SD = 1.37, 0.88, 0.77, 1.22 for child molesters, rapists, violent offenders and nonviolent offenders respectively). Child molesters were rated as significantly more enmeshed than nonviolent offenders and 'rapists, with violent offenders falling in the middle ground. The four groups also differed significantly on the Self-evaluation dimension, F(3,142) = 2.84, p = .04, (M = 3.43, 3.67, 3.33, 3.9, and SD = 0.69, 0.92, 0.76, 1.09 respectively). Child molesters and violent offenders, who were indiscriminable, were judged significantly more negative in their self-evaluation than nonviolent offenders were, with rapists occupying the middle ground.

Contextual variables. The groups differed significantly on their ratings on the Sexual Abuse dimension, F(3,142) = 2.81, p = .04, (M = 1.91, 1.23, 1.5, 1.23, and SD = 1.52, 0.90, 1.17, 0.97 for child molesters, rapists, violent offenders and nonviolent offenders respectively). Child molesters were rated as reporting significantly more sexual abuse than rapists and nonviolent offenders, with violent offenders occupying the middle ground. On the physical abuse dimension there were also significant differences across the groups, F(3,136) = 2.64 p = .05, (M = 3.54, 4.23, 4.29, 3.28, and SD = 1.64, 1.65, 2.00, 1.65 for respectively). Violent offenders and rapists, who were indiscriminable on this dimension, were rated as perceiving significantly more physical abuse than nonviolent offenders perceived, with child molesters in the middle ground.

The four groups did not differ significantly on the Loss dimension, F(3,142) = 1.07, p = .365, ns, (M = 2.17, 1.87, 1.8, 1.77, and SD = 1.30, 1.22, 0.96, 1.10 for child molesters, rapists, violent offenders, and nonviolent offenders respectively), on the Conflict dimension, F(3,142) = 0.76, p = .52, ns, (M = 2.36, 2.83, 2.83, 2.73, and SD = 1.69, 1.68, 1.72, 1.78 respectively), nor on the Positive Mediating Interactions dimension, F(3,142) = 3.74, p = .77, ns, (M = 1.26, 1.43, 1.33, 1.40, and SD = .59, .94, 0.76, .89 respectively). However, they did differ significantly on the Safety dimension, F(3,142) = 2.89, p = .03, (M = 2.87, 3.83, 3.8, 3.1, and SD = 1.71, 1.80, 1.73, 1.86 respectively). Child molesters were rated as perceiving significantly more safety than rapists and violent offenders, with nonviolent offenders falling in the middle ground.

Comparisons of Mothers and Fathers

Mothers and fathers were compared on the four relevant dimensions using only complete data sets. There was a significant difference on the Responsivity dimension, F(1,126) = 38.50, p = .0001, (M = 4.29, 4.31, 4.70, 4.27, 4.38 for mothers and 4.88, 5.72, 5.41, 4.77, 5.15 for fathers, for child molesters, rapists, violent offenders, nonviolent offenders, and overall respectively). Fathers were perceived as less responsive overall than mothers. There was also an interaction effect between offender type and gender of parent, F(3.126) = 2.57, p = .05; specifically, rapists perceived their fathers as significantly less responsive than their mothers. There was also a significant difference on the Consistency dimension, F(1,126) = 5.45, p = .02, (M = 3.31, 3.66, 3.74, 3.77, 3.57 for mothers and 3.60, 4.03, 4.11, 3.92, 3.87 for fathers, for child molesters, rapists, violent offenders, nonviolent offenders, and overall respectively); that is, fathers were perceived as more inconsistent than were mothers.

There was a significant difference between mothers and fathers on the Acceptance dimension, F(1,126) = 15.47, p = .0001, (M = 4.29, 4.03, 4.78, 4.46, 4.37 for mothers and 4.83, 5.21, 4.96, 4.35, 4.85 for fathers, for child molesters, rapists, violent offenders, nonviolent offenders and overall respectively), with fathers perceived as more rejecting overall than mothers. The interaction between offender type and parent gender was significant, F(3,126) = 4.47, p = .005; specifically, rapists perceived their fathers as least accepting, followed by child molesters, with no difference between violent offenders and nonviolent offenders. There was also a difference between mothers and fathers on the Boundaries dimension, F(1,126) = 6.70, p = .01, (M = 4.42, 3.24, 3.93, 4.27, 4.02 for mothers and 4.73, 3.62, 3.93, 4.46, 4.27 for fathers, for child molesters, rapists, violent offenders, non-violent offenders, and overall respectively), with fathers being perceived as having firmer boundaries than mothers.

DISCUSSION

The grounded theory analysis of the interview data revealed that 14 categories were able to account for the participants' perceptions of their early interpersonal experiences. That these categories were not further reducible points to the diversity and complexity of offenders' perceptions of their early interpersonal experiences. The categories of responsiveness, consistency, acceptance, and boundaries all emerged as significant aspects of the offenders' perceptions of their relationships with their caregivers. Emotional regulation, autonomy, and self-evaluation appeared as important facets of the offenders' experience of the self in interaction with their attachment figures. Finally, experiences of sexual deviation and abuse, physical abuse, loss, conflict, safety, and positive mediating interactions arose as meaningful contextual factors.

The derivation of the categories suggests that most of the offenders described their interactions with caregivers as involving high levels of neglect and rejection and low levels of supervision, discipline, and consistency. Generally, they tended to evaluate themselves negatively and appeared to manage emotions in a defensive or avoidant fashion rather than to directly seek support from other people. Perceptions of physical abuse were strikingly high, with a number of offenders also describing past sexual abuse, loss of caregivers, and conflict within the family home. Sadly, the early interpersonal experience described least often by all offender groups was the presence of a positive mediating interaction; there appear to have been very few positive relationships in their lives. With such negative and disruptive early interpersonal experiences, it is of no surprise that the majority of the offenders (over 75%) in this study reported insecure attachment styles as adults. This is in contrast to normative samples where the majority (55-65%) of individuals rate themselves as securely attached.

The second aim of the study was to use the above categories to investigate differences between the various offender groups. The most salient finding from this analysis was that rapists and violent offenders shared a number of features of their early interpersonal relationships in common. Specifically, the fathers of rapists and violent offenders were less responsive than were the fathers of other groups, both mothers and fathers provided looser boundaries for rapists and violent offenders than for other offender types, and both rapists and violent offenders experienced more physical abuse and felt less safe than the other groups of offenders. These results are consistent with findings documented in the literature. As noted earlier, this research suggests that higher rates of physical abuse in childhood are related to greater perceptions of danger and higher levels of coercive and aggressive behavior (i.e., features more characteristic of rape and violent offending than child molestation or nonviolent offenses). Furthermore, while no specific hypotheses were made with respect to parental style and boundaries, aggressive, coercive behaviors may function during childhood to secure the attention of unresponsive caregivers. The presence of aggressive and hostile behavior may also reflect the existence of a dismissive attachment style, and indicate that rapists and violent offenders may be more alike in this respect than other types of offenders (Hudson et al., 1997; Ward et al., 1997).

The significant finding that fathers of rapists and violent offenders were less responsive to their needs, in combination with the nonsignificant trend for the fathers of rapists to be more rejecting than the fathers of nonviolent offenders, points to the importance of this relationship for both of these groups. This is consistent with previous research, which has found that sexual offenders typically have more negative, problematic relationships with their fathers than they do with their mothers, and that this is more evident in rapists than in child molesters (Lisak, 1994). Although this result is consistent with the empirical literature, it conflicts with the popular myths surrounding the crucial role of the mother in the causation of rapists' hostility towards women. It seems that the role of the father in the development of both violent men and those who rape adult women is in need of greater recognition and understanding.

The finding that child molesters described significantly more sexual deviation and abuse than rapists and nonviolent offenders (with violent offenders in the middle) is also in line with the literature reviewed earlier: for example, the finding that child molesters are at least twice as likely to report sexual abuse than are rapists (e.g., Seghorn et al., 1987). The finding that violent offenders were also rated high on the dimension of sexual abuse was not predicted by the literature and represents a puzzle. One possible explanation could be that the combination of physical and sexual violence may reflect greater attachment disruption and subsequently an increased predisposition to commit violent acts. However, because we did not estimate the actual frequency of sexual abuse, it is possible that this difference represents more extreme sexual abuse for a few individuals as opposed to higher numbers of violent offenders experiencing sexual abuse.

Child molesters and violent offenders appear to evaluate themselves more negatively than the other offender groups. It has been previously argued that child molesters have a tendency toward fearful and preoccupied attachment styles, both characterized by negative views of the self (Ward et al., 1996). Our results are consistent with this earlier work and with additional research concluding that as a group child molesters tend to view themselves in a self-depreciating manner (Marshall & Mazzucco, 1995; Ward et al., 1997).

Negative evaluations of the self are related to a sense of personal unworthiness and to a reduced sense of autonomy. This relationship is reflected in the finding that child molesters experienced less autonomous attachment relationships than nonviolent offenders and rapists, with violent offenders again in the middle of the range. This finding fits with both previous research indicating that child molesters tend to experience dependent relationships with their mothers and the previously stated clinical observation that these individuals report a lack of clear boundaries in relationships with their parents (Tingle et al., 1986). The lesser degree of autonomy child molesters described may also be related to their greater tendency to exhibit fearful and preoccupied attachment styles: Individuals with these styles desire to become intimate with others but are fearful of rejection. The greater focus on others may be related to a higher degree of sensitivity to rejection and to a tendency to seek the approval of others and, ultimately, to less manifest violence in the context of offending and to more grooming and courtship behaviors. The greater reported autonomy of rapists could be related to their tendency to be dismissively attached as opposed to the more constructive or positive experience of autonomy evident in nonviolent offenders.

The categories where clear differences between the groups were not found were maternal responsiveness, maternal acceptance, consistency (both parents), and emotional regulation. In view of the importance that attachment literature places on maternal responsivity, it is unclear why we found no significant differences between the various types of offenders. Clearly the behavior of fathers had more impact on specific types of offenders than did that of mothers, at least within an incarcerated group of offenders. However, given that maternal responsiveness and acceptance were generally low (albeit less so than for fathers) for each of the offender groups in this study, it is possible that this may represent a general vulnerability factor to offending. Further investigation with community controls may help to clarify this point.

As previously stated, there was no difference across offender types on the category of emotional regulation. Again, this finding is somewhat puzzling and may simply reflect within-group heterogeneity. However, as noted earlier, the majority of individuals in this study appeared to be at least somewhat emotionally defended. This may be related to being a male residing in a prison environment. Whether this result reflects gender, a process of imprisonment, or a general vulnerability to offending is difficult to ascertain. Again, further investigation with a community control group would clarify this issue.

Finally, the fact that there was no significant difference between the various groups on the category of consistency is another puzzle. Previous research has linked inconsistency, in the form of chaotic environments and loss and instability of caregiver relationships, to the severity of aggression in sexual and violent offending (Prentky et al., 1989). In this study, we incorporated chaotic or inconsistent environments and the consistency of parenting style within the overall category, possibly obscuring any real difference. Furthermore, the large standard deviations for this category may indicate that significant within-group differences were being masked by the overall means. The severity of aggression involved in the sexual and violent crimes was not measured; therefore, this association was not able to be examined.

These results support the increasingly accepted notion that sexual offenders do not represent a homogenous group (Ward et al., 1996). Certainly, it seems that a simple distinction between sexual offenders and other criminal offenders is not warranted. In this study, there was not one factor on which both groups of sexual offenders differed from the other groups. In fact, it seems that, at least with respect to their early interpersonal relationships, rapists have more in common with violent offenders than they do with child molesters. This fact coheres with both the general literature and our previous findings on the relationship between intimacy variables and offender type (Ward et al., 1997). In many respects child molesters were more similar to nonviolent offenders, with only the presence of a history of sexual deviation or abuse and generally less autonomy in relationships distinguishing the two types of offenders.

The third aim of the study was to make use of the categories to examine the relative differences between the perceptions of mothers and fathers across the offender groups. In terms of responsiveness, rapists in particular perceived fathers as less responsive than mothers. In terms of the acceptance category, fathers were rated as less accepting (more rejecting) by rapists, and, to a lesser extent by child molesters. Finally, for all offender groups, fathers were rated as less consistent and more boundaried than were mothers. These results are consistent with previous research where negative relationships with fathers, in conjunction with a more benign relationship and looser boundaries with mothers, have been documented in the histories of both sexual and violent offenders. Once again, these findings highlight the importance of rejection and neglect by fathers in the histories of offenders.

This study represents a preliminary, exploratory study, and it is clear that further research is needed to tease out more carefully the nature and severity of abuse in the early interpersonal experiences of sexual offenders. It is also clear that these issues are pertinent to nonsexual offenders, and research with this population will need to also consider these variables and their influences. An integrated measurement approach is favored where multiple methods are used (e.g., interview data, questionnaires, and experimental methods) to provide a clearer picture of why, and how, these experiences are related to sexual crimes.

In conclusion, the current study provides evidence that the early interpersonal experiences of offenders are overwhelmingly negative, and, for rapists and violent offenders, the most negative of these experiences were with their fathers. Another major conclusion is that negative early interpersonal experiences are characteristic of both violent and nonviolent offenders and are not specific to sexual offenders. It is possible that such experiences can be viewed as a generalized vulnerability factor resulting in a variety of offending patterns and life problems. The identification of variables further back in the etiological chain has major implications for the prevention of crime, particularly the early detection and treatment of individuals who are at risk of developing antisocial lifestyles and attitudes.

REFERENCES

Blaske, D. M., Borduin., C. M., Henggeler, S., & Mann, B. (1989). Individual, family and peer characteristics of adolescent sexual offenders and assaultive offenders. Developmental Psychology, 25, 846-855.

Fagan, J., & Wexler, S. (1988). Explanations of sexual assault among violent delinquents. Journal of Adolescent Research, 3, 363-385.

Hazan, C., & Shaver, P. (1994). Attachment as an organizational framework for research on close relationships. Psychological Inquiry, 5, 1-22.

Hazelwood, R. R., & Warren, J. (1989). The serial rapist: His characteristics and victims (conclusion). FBI Law Enforcement Bulletin, 58, 18-25.

Hudson, S. M., Marshall, W. L., Ward, T., Johnston, P. W., & Jones, R. L. (1995). Kia Marama: A cognitive behavioural programme for incarcerated child molesters. Behaviour Change, 12, 69-80.

Kahn, T. J., & Chambers, H. J. (1991). Assessing reoffence risk with juvenile sexual offenders. Child Welfare, 70, 333-345.

Levant, M. D., & Bass, B. A. (1991). Parental identification of rapists and pedophiles. Psychological Reports, 69, 463-466.

Lewis, D. O., Shanock, S. S., & Pincus, J. H. (1981). Juvenile male sexual assaulters. American Journal of Psychiatry; 136, 1194-1196.

Lisak, D. (1994). Subjective assessment of relationships with parents by sexually aggressive and nonaggressive men. Journal of Interpersonal Violence, 9, 399-411.

Lisak, D., & Roth, S. (1990). Motives and psychodynamics of self-reported, unincarcerated rapists. American Journal of Orthopsychiatry, 60, 268-280.

Marshall, W. L. (1989). Invited essay: Intimacy, loneliness and sexual offenders. Behaviour Research and Therapy, 27, 491-503.

Marshall, W. L., & Mazzucco, A. (1995). Self-esteem and parental attachments in child molesters. Sexual Abuse: A Journal of Research and Treatment, 7, 279-285.

Milner, T., & Robertson, M. (1990). Comparison of physical child abusers, intrafamilial sexual child abusers and child neglecters, Journal of Interpersonal Violence, 5, 37-48.

Prentky, R. A., Knight, R. A., Sims-Knight, J. E., Straus, H., Rokous, F., & Cerce, D. (1989). Developmental antecedents of sexual aggression. Development and Psychopathology, 1, 153-169.

Ryan, G., & Lane, S. (1991). Juvenile sexual offending: Causes, consequences and correction. Lexington, MA: Lexington Books.

Saunders, E., Awad, G. A., & White, G. (1986). Male adolescent sexual offenders: The offender and the offence. Canadian Journal of Psychiatry, 31, 542-549.

Seghorn, T. K. Prentky, R. A., & Boucher, R. J. (1987). Child abuse in the lives of sexually aggressive offenders. Journal of the American Academy of Child and Adolescent Psychiatry, 26, 262-267.

Strauss, A., & Corbin, J. (1990). Basics of qualitative research: Grounded theory procedures and techniques. Newbury Park, CA: Sage.

Tingle, D., Barnard, G. W., Robbin, L., Newman., G., & Hutchinson, D. (1986). Childhood and adolescent characteristics of pedophiles and rapists. International Journal of Law and Psychiatry, 9, 103-116.

Ward, T., Hudson, S. M., & Marshall, W. L. (1996). Attachment style in sex offenders: A preliminary study. The Journal of Sex Research, 33, 17-26.

Ward, T., McCormack, J., & Hudson, S. M. (1997). Sexual offenders perceptions' of their intimate relationships. Sexual Abuse: A Journal of Research and Treatment, 9, 57-73.

Julie McCormack and Stephen M. Hudson
University of Canterbury, New Zealand

Tony Ward
University of Melbourne, Australia

Sadly, Dr Steve Hudson died before this paper could appear. He was a creative and innovative researcher and a wonderful human being, and we will miss him.

Address correspondence to Dr. Tony Ward, Department of Criminology, University of Melbourne, 234 Queensberry Street, Melbourne 3010, Australia; e-mail: [e-mail].