1
Misogynists with an Urge to Merge? Ambivalent Attitudes towards Women in Rapists
Eveline E. Schippers
University of Amsterdam, De Forensische Zorgspecialisten
Eveline E. Schippers, Department of Psychology, University of Amsterdam; Research Department, De Forensische Zorgspecialisten.
This research was part of a larger research project supported by grant 2014.16 from the Kwaliteit Forensische Zorg at name of Smid – Bultens & Von Borries – Verschuere.
Correspondence concerning this article should be addressed to Eveline Schippers, Haarlemmerstraat 166B, 2312 GG Leiden, the Netherlands.
2 Abstract
It is important to understand the mechanisms underlying sexual coercion, in order to optimize treatment of convicted rapists and reduce risk of sexual recidivism. Rapists have been theorized to have hostile attitudes towards women, but some empirical evidence indicates that they also have relatively positive attitudes towards women. Current research investigated ambivalent attitudes towards women in rapists, a nonsexual offender control group, and a community control group (N=114) using explicit questionnaires (Hostility Towards Women Scale and Ambivalent Sexism Inventory) and implicit tests (IAT and AAT). Rapists unexpectedly showed less hostile sexism, nonsexual offenders showed more benevolent sexism, and rapists and nonsexual offenders both showed a weaker association between women and deceitful on the IAT, compared to community controls. The current results indicate that rapists have less hostile (more relatively positive) attitudes towards women than community controls. Relatively positive attitudes might urge rapists to be with women, rather than ignore them. This, together with a deviance of any kind (sexual or cognitive), might eventually lead to sexual offense. Findings have possible consequences for optimization of rapist treatment.
3 Misogynists with an Urge to Merge? Ambivalent Attitudes towards Women in Rapists
Fifteen percent of convicted sexual offenders reoffend with a sexual offense (Hanson & Morton-Bourgon, 2009). This rate is relatively low compared to other types of recidivism (Smid, 2014). Still, it is important to reduce sexual recidivism, in order to reduce the number of victims. Convicted rapists in TBS-clinics1 receive treatment to reduce the risk of recidivism. To optimize treatment, it is important to understand the mechanisms underlying sexual coercion. In this thesis, “sexual coercion” is used as an umbrella term for sexual aggression, sexual violence, rape, and related behaviors. We investigated rapists’ attitudes towards women. We considered only rapists from TBS-clinics with female adult victims, rather than child or male victims, for two reasons. First, the most common offense is with a male perpetrator and female victim (Lonsway & Fitzgerald, 1994). Second, child abusers may have different offense related attitudes than sexual offenders with adult victims (Helmus, Hanson, Babchishin, & Mann, 2013).
Negative Attitudes Towards Women
Negative attitudes2 towards women are associated with sexual coercion. Sexual coercion is related to, for example, rape myth acceptance beliefs (e.g., “women in short skirts ask for sex”), adversarial sexual beliefs (e.g., “women are out to take advantage of men”), sex role stereotyping (e.g., “a wife should never contradict her husband”), sexist attitudes (e.g., “men are superior to women”) and hostility towards women (e.g., “women irritate me”) (Burt, 1980; Lonsway & Fitzgerald, 1994). Hostility towards women is defined by Check (1984):
Hostility toward women is typically implicit, consisting of the mulling over of past attacks from women, rejections from women, and deprivations from women,
1 TBS (Ter Beschikking Stelling): court-ordered inpatient treatment for convicted offenders in the Netherlands. 2 In this paper, the term “attitude” is used to define cognitions as well as affect, as “the concept of attitude provides a highly useful integration of the influence of affect and cognition on behavior” (White, Donat, & Humphrey, 1996, General Discussion, par.1).
4
and may be inferred when aggression against a woman is motivated by a desire to hurt rather than by a desire to attain some extrinsic reinforcer. (p.35)
It is also closely related to certain beliefs about female sexuality, for instance, the idea that men and women are sexual adversaries (Check, Malamuth, Elias, & Barton, 1985). Hostility towards women has often been related to sexual coercion, mainly in college and community samples (e.g. Malamuth, 1986; Tharp et al., 2013), but also in forensic samples of nonsexual offenders (Robertson & Murachver, 2007), a mixed sample of child abusers and rapists (Widman, 2010), and samples of (juvenile) rapists (Johnson & Knight, 2000; Marshall & Hambley, 1996). In addition, it is significantly related to sexual recidivism (Hanson, Harris, Scott, & Helmus, 2007; Mann, Hanson, & Thornton, 2010). Hostility towards women was more strongly associated with sexual coercion than sexual compulsivity and hypermasculinity in studies including hostility towards women and other attitudes (Johnson & Knight, 2000; Malamuth, Sockloskie, Koss, & Tanaka, 1991). Furthermore, hostility towards women mainly accounted for the variance in sexist perceptions and rape supportive beliefs in a community sample (Forbes, Adams-Curtis, & White, 2004). Lonsway and Fitzgerald (1995) suggest that hostility towards women is the core of a “multidimensional misogyny”, including other adversarial or sexist cognitions (p.709). In sum, hostility towards women appears to be a core variable in relation to sexual coercion and will therefore be assessed in the current research.
Ambivalent Attitudes Towards Women
Research focuses on men’s negative attitudes towards women in relation to sexual coercion. However, sexual coercive men might have relatively positive attitudes towards women as well. A few examples illustrate this. First, sexism comprises two components that are distinct, but not necessarily conflicting: hostile sexism and benevolent sexism (Abrams, Viki, Masser, & Bohner, 2003; Glick & Fiske, 1996). Hostile sexism is a dominative and negative sexist view (e.g., “women fail to appreciate all that men do for them”). Benevolent
5
sexism is relatively and subjectively positive (e.g., “women must be rescued first from a sinking ship”), however, albeit the protective undertone, it still draws upon women’s inferiority. It is not clear how both elements relate to sexual coercion. Only hostile (not benevolent) sexism predicted self-reported proclivity for sexual coercion in community samples (Abrams et al., 2003; Forbes et al., 2004). Ambivalent sexism has not been studied in samples of actual sexual offenders yet, to the knowledge of the author. Second, Berliant, Nunes, and Sperling (2011) explicitly measured hostility towards women and implicitly measured positive versus negative evaluation of women in a community sample. Subjects who reported some likelihood to rape and high hostility towards women showed less negative evaluation of women, when compared to no likelihood to rape and low hostility. In other words: hostility towards women equaled a
less negative evaluation of women, when related to sexual coercion. A final example of
relatively positive attitudes towards women comes from clinical experience. Treatment providers experience that many rapists show a certain urge to approach women and a strong desire to have a (sexual) relationship with a woman (W.J. Smid, personal communication, June 19, 2015). Together, these examples justify that I investigate ambivalent attitudes towards women in relation to sexual coercion. Men who hate women may at the same time desire women and feel the “urge to merge” with women. Misogynists’ urge to merge might make them vulnerable to commit sexual offense. It must be noted that negative and positive attitudes can co-exist within a person and do not need to exclude each other (Larsen, McGraw, & Cacioppo, 2001). Ambivalence indicates the presence of both positive and negative feelings and is different from indifference, which indicates the absence of positive or negative feelings (Kaplan, 1972).
Explicit Versus Implicit
Sexual coercion research is mainly based on explicit questionnaires. This is easy and time effective, but also has some disadvantages. Questionnaires can be contaminated by social
6
desirable answering and may suffer from lack of introspection. It is useful to measure attitudes both explicitly and implicitly, since explicit attitudes are not necessarily the same as implicit attitudes (Widman, 2010; Wilson, Lindsey, & Schooler, 2000). Explicit questionnaires measure conscious and deliberate attitudes, while implicit tests measure more spontaneous and automatic attitudes that may be beyond one’s conscious control (De Houwer, 2006). Assessment of a specific attitude is thus more complete with explicit and implicit measurement. It seems that the combination of implicit and explicit measurement provides partly distinct, but complementary information (Nunes, Hermann, & Ratcliffe, 2013; Snowden, Craig, & Gray, 2011). In the current research, I used explicit questionnaires and implicit tests to assess ambivalence in rapists.
As implicit measures, I used the Implicit Association Test (IAT) and the Approach-Avoidance Task (AAT). The IAT works by assigning stimuli to pairs of categories, where a relative fast categorization indicates a close association between categories (Greenwald, McGhee, & Schwartz, 1998). The IAT is frequently used, especially to assess child abusers’ attitudes (Snowden et al., 2011), but also to predict rape attitudes (Nunes et al., 2013). The IAT knows some drawbacks. It showed similar to lower discrimination of groups, compared with other implicit measures such as viewing time, choice reaction time, or rapid serial visual representation (Babchishin, Nunes, & Hermann, 2013). The IAT also works on a category level (e.g., “good” versus “bad”). This may push attitudes more towards extremes (Widman, 2010), for example when one is forced to choose “good” over “slightly better than neutral”. It may also cause miscategorization (Babchishin et al., 2013), when “rape” is, for instance, assigned to “good” because it is a “four-letter word” instead of “positive valence”. Yet, the IAT is relatively straightforward, cost effective, and noninvasive. It is frequently used and reviewed, and can easily be adapted to new categories or stimuli (Snowden et al., 2011). The AAT assesses relative approach and avoidance tendency towards stimuli (Rinck & Becker, 2007). In
7
general, positive attitudes towards objects are associated with an automatic approach tendency, whereas negative attitudes are associated with avoidance (Fazio, 2001). To the knowledge of the author, the AAT has not been used in sex offender samples before. However, there are some examples of the AAT in sexual research. One study assessed approach and avoidance tendencies towards pornographic stimuli in a community sample (Snagowski & Brand, 2015). Subjects that reported a high tendency towards cybersex showed either avoidance or approach tendencies towards pornographic stimuli, but not to neutral stimuli. Individuals with a low tendency towards cybersex showed neither approach nor avoidance tendencies towards pornographic stimuli. Furthermore, approach/avoidance tendency towards pornographic stimuli was associated with self-reported problematic sexual behavior (e.g., loss of control during sex, high sexual excitement). Thus, approach and avoidance tendencies towards pornographic stimuli were related to cybersex tendency, as well as to sexual problematic behavior. Another study validated the use of the AAT for erotic stimuli in a male community sample (Hofmann, Friese, & Gschwendner, 2009). Subjects showed an approach tendency towards erotic stimuli, but not towards art stimuli. Also, approach/avoidance tendency discriminated between subjects who were single and who were in a relationship. Men in a relationship showed shorter avoidance latencies for sexual stimuli than single men. Approach latencies for sexual stimuli were the same for both groups. According to the researchers, men in a relationship might have facilitated avoidance reactions, in order not to not fall for alternatives. These examples show that the AAT can be used to assess approach and avoidance tendencies in sexual research. Rapists might exhibit a stronger approach tendency (positive valence) towards women than nonoffenders.
Current Research
In sum, there is evidence that both hostile and positive attitudes towards women relate to sexual coercion. Hostile means that misogynistic men judge female characteristics in a
8
hostile, negative way, for example, “women are manipulative”. Positive means that misogynistic men judge female characteristics as relatively positive, which is not necessarily positive to the woman herself, for example, “women fulfill men’s needs”. The current research investigated these ambivalent attitudes towards women in rapists with implicit tests and explicit questionnaires. A field study, inquiring expert opinion, was conducted to clarify the nature of positive and hostile attitudes towards women in rapists. The field study enabled me to select the most appropriate measures and stimuli for the experimental study, resulting in the concepts “women are deceitful” and “women are prestige objects” for the implicit measures (see section Field Study below). The experimental study assessed ambivalent attitudes in rapists, compared to a nonsexual offender control group and a nonoffender community control group. It was hypothesized that rapists would show more hostile as well as more positive attitudes towards women compared to nonsexual offenders and community controls. More specifically, it was expected that rapists would show a stronger implicit association between woman and deceitful (versus woman and honest); that rapists would show a stronger implicit association between
wanting and women as prestige object (versus not wanting and women as prestige object); and
that rapists would show more approach tendency towards women. It was also expected, that rapists would score higher on hostility towards women, hostile sexism, and benevolent sexism compared to nonsexual offenders and community controls.
Field Study
A field study was conducted for two reasons. First, I could acquire more evidence for the idea that rapists may show positive attitudes towards women, in addition to hostility towards women. Second, I could clarify the nature of positive and hostile attitudes, by identifying the most appropriate stimuli for our implicit measures. A review of sexual coercion literature yielded a number of possible positive and hostile attitudes towards women, e.g. “women are beautiful creatures” and “women cannot be trusted”. These attitudes were combined in a list,
9
which was distributed among sexual offender therapists (N=17) from three TBS-clinics. Respondents rated to what extent proposed attitudes were perceived in sexual offenders in case load on a 4-point scale, ranging from 1 (nearly never encountered) to 4 (almost always
encountered)3. The two attitudes with the highest means were the hostile attitude “women are deceitful” (M=2.84, SD=0.93) and the positive attitude “women are prestige objects” (M=2.66,
SD=0.91). The results indicated that positive attitudes towards women were present among
rapists and that “woman are deceitful” and “women are prestige objects” were the most commonly encountered attitudes towards women. These attitudes were subsequently used to develop a Deceitful-IAT and a Prestige-IAT (see Method section).
Method Participants
The current research included patients from three Dutch TBS-clinics. The rapist group (n=32) included convicted rapists for whom the majority of their sexual victims was older than 16 years. A nonsexual offender group (n=52) consisted of patients convicted of violent but not sexual offense. This group was included as a patient control group that shared its environment with the rapist group. Patients could not participate when they had acute psychotic symptoms or high doses of psychotropic medication that would significantly interfere with their reaction time. Because TBS-patients might respond differently than healthy men, a community control group was included. The community control group was acquired by advertisement (n=3), snowball sampling (n=23), and street recruitment (n=4; total community control group n=30). The community control group was screened on the following criteria: no self-reported acute treatment of any psychiatric or psychological disorder; no self-reported history of offense; nonacademic education level (Dutch: Hoger Beroepsonderwijs). Groups were matched as closely as possible on age, country of birth, and intelligence. All participants (N=114) were
10
male, reported a sufficient understanding of Dutch, and reported sexual interest in women (not necessarily exclusive). An overview of demographic information of the subgroups is provided in Table 1.
Instruments
Demographic information form. Participants completed a short questionnaire about
their age, country of birth, educational level, and current profession. Information regarding current and past relationships was recorded by questioning marital status, length of longest romantic relationship with a woman, and having ever lived with a woman in a romantic relationship longer than 2 years.
Table 1
Demographic Information of Subgroups
Rapists Nonsexual offenders Community Controls Age (years) 43.50 (12.18) 38.66 (9.81) 36.13 (13.27) Country of Birth Netherlands Other 78% 22% 71% 29% 90% 10% DART IQ 89.17 (10.09) 86.98 (12.41) 93.43 (13.15) Marital Status Single Relationship 81% 19% 75% 25% 40% 60%
Ever Lived Together >2 years 72% 52% 47%
Note. DART IQ: IQ score on the Dutch Adult Reading Test. Ever Lived Together >2 years:
percentage that ever lived together with a woman in a romantic relationship for more than 2 years. Percentages are provided or mean scores with standard deviation between brackets.
11 Deceitful-IAT. The Deceitful-IAT was designed for the current study to assess the
association between women and deceitful (original by Greenwald et al., 1998). The attribute category was “deceitful” versus “honest”, the target category was “man” versus “woman” (Dutch: respectively vals versus betrouwbaar, and man versus vrouw). All category labels had five matched stimulus words, for example “she” or “lady” for woman (see Appendix, Table A1). Stimulus words for the attribute category were chosen to imply deliberateness behind the characteristics, e.g., rather “reliable” and “mean” than “nice” and “dumb”. Category words were presented on the left and on the right side of the screen. Stimulus words were presented in the middle and had to be assigned to either the left category by pressing the E-key, or to the right category by pressing the I-key. The E- and I-key were marked with a white sticker. In block 1, the attribute category (deceitful/honest) was practiced in 20 trials (see Table 2). In block 2, the target category (man/woman) was practiced in 20 trials. In block 3 (20 trials) and 4 (40 trials), both categories were presented in combination. Stimulus words from attribute and target category were randomly presented and had to be assigned to the left or the right side, whilst paying no attention to the irrelevant category. In block 5, the target category (man/woman) switched sides, and the new position was practiced in 20 trials. In block 6 (20 trials) and 7 (40 trials), the attribute and target category were paired again, but now with the reverse positions. Each subject completed a total of 60 practice trials and 120 test trials. Participants were instructed to respond as fast and accurate as possible. After an incorrect response, a correct response was required to continue. Blocks were counterbalanced such that half of the participants started with the women-deceitful and men-honest combination, and the other half started with the women-honest and men-deceitful combination.
12
Table 2
Deceitful-IAT Design
Block N Trials Left Key Right Key
1 (practice) 20 honest deceitful
2 (practice) 20 man woman
3 (test) 20 man + honest woman + deceitful 4 (test) 40 man + honest woman + deceitful
5 (practice) 20 woman man
6 (test) 20 woman + honest man + deceitful 7 (test) 40 woman + honest man + deceitful
IAT data were analyzed conform the procedure from Greenwald, Nosek, and Banaji (2003). Only data from blocks 3, 4, 6, and 7 were analyzed4. Trials with latencies larger than 10.000 ms were eliminated. The specific combinations of woman-deceitful and man-honest, and woman-honest and man-deceitful were coded. All blocks with the same combination were collapsed, regardless whether they were counterbalanced as first or second in the test. For each subject, means were calculated for each of the blocks 3, 4, 6, and 7. For each subject, the standard deviation over all blocks was calculated. An individual difference score was calculated by subtracting the two means of the women-deceitful and man-honest combination from the two means of the women-honest and man-deceitful combination (Mwomenhonest – Mwomendeceitful). The difference score was divided by the standard deviation and the two resulting numbers were averaged. This resulted in the final D-score for the Deceitful-IAT per subject. A positive Deceitful-IAT D-score indicated that participants responded faster on the women-deceitful (and
4 The original Greenwald et al. (1998) IAT included seven blocks, of which only blocks 4 and 7 were used in statistical analyses. The IAT still knows seven blocks, but the Greenwald et al. (2003) procedure uses data from blocks 3 and 6 as well.
13
man-honest) combination, indicating that these were closer associated than women-honest (and man-deceitful), i.e. Mwomenhonest > Mwomendeceitful. A negative Deceitful-IAT D-score indicated a closer association between women-honest and between man-deceitful. In the current experiment, the Deceitful-IAT had a good split-half reliability (odd-even method), with Spearman-Brown rs = .97.
Prestige-IAT. The Prestige-IAT was designed for the current study to assess the
association between wanting and women as prestige object. The attribute category was “women as prestige object” versus “women as equal partner” (named prestige and equal hereafter), the target category was “wanting” versus “not wanting” (Dutch: ik wil wel versus ik wil niet). The target category had five matched stimulus words expressing wanting, and five expressing not-wanting (see Appendix, Table A2). The attribute category used five stimulus pictures for prestige and five for equal. Stimulus pictures were all black and white pictures expressing either prestige scenes (e.g., women as “trophy wives”) or equal scenes (e.g., equal relationships). Pictures were matched on brightness, orientation, and type of scene (e.g., a prestige picture of a sport jock with a cheerleader was matched with an equal picture of a man and women practicing sports together). Pictures were also matched on position, posture, age and attractiveness of depicted people. The series of pictures was validated in a stimulus validation study (see Prestige-IAT stimulus validation section). For the Prestige-IAT, the same procedure was applied as for the Deceitful-IAT. Here, a positive Prestige-IAT D-score indicated that wanting-prestige (and not-wanting-equal) were closer associated than not-wanting-prestige (and wanting-equal). A negative Prestige-IAT D-score indicated a closer association between not-wanting-prestige and between wanting-equal. In the current experiment, the Prestige-IAT had a good split-half reliability (odd-even method), with Spearman-Brown rs = .98.
Prestige-IAT stimulus validation. An online questionnaire was distributed among 91 Dutch respondents (46 male, age M=28.29, SD=12.95), recruited by snowball sampling. They
14
rated to what extent the depicted women were a prestige object to the depicted men on a scale from 1 (not prestigious at all) to 10 (very prestigious) for all 10 Prestige-IAT stimulus pictures and two attribute label pictures. Bivariate correlations were calculated (scores not provided). This could show if the prestige pictures were related to each other and thus supported a similar prestige concept and if the equal pictures related to each other in support of a similar equal concept. All prestige pictures were correlated, with exception of one picture that did not correlate to any other picture. The inadequate prestige picture was replaced. To test the new stimulus set, a new online questionnaire was distributed among 51 different Dutch respondents, recruited by snowball sampling (22 male, age M=25.67, SD=7.03). Bivariate correlations were calculated again. All prestige stimuli and all equal stimuli were significantly related to each other, with exception of two prestige pictures that did not have a correlation with each other, but did correlate with all other pictures (scores not provided). This means that the respondents identified all prestige pictures to be similarly prestigious, and all equal pictures to be similarly equal. An independent samples t-test showed that rating differed significantly for equal (M=4.30, SD=.18) and prestige stimuli (M=7.41, SD=.82), t(12)=–9.83, p<.001, d=5.24 (i.e., a very large effect size, according to Cohen [1992]). This means that all prestige pictures had a higher prestige score and the respondents differentiated between prestige and equal pictures. The final stimulus set was used in the Prestige-IAT.
Approach-Avoidance Task (AAT). The AAT assessed approach/avoidance tendency
towards women and men (original by Rinck & Becker, 2007). In the AAT, subjects are instructed to pull or push a joystick based on the presented stimulus, i.e., picture of a man or a woman. When pulled, the picture enlarges such that the participant virtually approaches the picture. When pushed, the picture size decreases such that the participant virtually avoids the picture. A relative faster pull reaction indicates an approach bias to the stimuli, a relative faster push reaction indicates an avoidance bias. The instructions of Rink and Becker (2007) were
15
followed for the experimental setup (see p.109 for a detailed description). Stimuli were eight whole-body pictures of women and eight whole-body pictures of men standing in front of a white background. Pictures were matched on ethnicity, clothing, color, posture, age, and attractiveness of the depicted people. Push/pull instructions changed halfway through the assessment (women-pull became women-push, and vice versa). Every test block was preceeded by a practice block (see Table 3). Each participant completed a total of 32 practice trials and 128 test trials. Participants were instructed to respond as accurate and fast as possible. Blocks were counterbalanced such that half of the participants started with the woman-pull instruction and the other half with woman-push.
The AAT data were analyzed with the same procedure as the IAT data, using standardized difference scores instead of means or medians (cf. Greenwald et al. 2003). Standardized D-scores are less vulnerable to individual differences in response latencies, and provided stronger results for the AAT than the original scoring algorithm (Wiers, Eberl, Rinck, Becker, & Lindenmeyer, 2011). Here, only data from testing blocks 2 and 4 were analyzed, and only correct trials were used. Latency was the time to finish a reach with the joystick. A positive AAT D-score means that the mean latency for woman-pull blocks was shorter than the mean latency for woman-push blocks (Mwomanpush > Mwomanpull). In this case subjects showed an approach towards women (and avoidance of men) rather than avoidance of women (and approach towards men). A negative AAT D-score indicated an avoidance of women (and approach towards men) rather than approach towards women (and avoidance of men). In the current experiment, the AAT had a good split-half reliability (odd-even method), with Spearman-Brown rs = .97.
16
Table 3
AAT Design
Block N trials Instructions 1 (practice) 16 man-pull, woman-push 2 (test) 64 man-pull, woman-push 3 (practice) 16 man-push, woman-pull 4 (test) 64 man-push, woman-pull
Dutch Adult Reading Test. The Dutch Adult Reading Test is a quick and reliable
estimator of verbal IQ (Schmand, Lindeboom, & Van Harskamp, 1992). The test requires the reading of 50 words that have irregular or unusual pronunciation. Pronunciation of words is scored as right (two points), doubt (one point), or wrong (zero points), and a sum score is calculated, which is converted into an IQ score. The Dutch Adult Reading Test underestimates IQ, especially for lower IQ scores (De Lugt, Ponds, Rozendaal, Houx, & Jolles, 1995). However, because of its easy and quick administration it was used as a rough IQ estimate to match the groups.
Balanced Inventory of Desirable Responding. The short Dutch version of the
Balanced Inventory of Desirable Responding estimates social desirable responding with two scales of each 10 items (Paulhus & John, 1998). The Self Deceptive Enhancement scale measures honest, but unconsciously inflated self-descriptions, for example, “Once I’ve made up my mind, other people can seldom change my opinion”. The Impression Management scale measures consciously inflated self-descriptions, for example, “I never take things that don’t belong to me”. Participants answer on a 5-point Likert scale, ranging from 1 (disagree strongly) to 5 (agree strongly). After reversing negatively keyed items, a sum score is calculated for each scale, with higher scores indicating higher self-deceptive enhancement and higher impression
17
management (Stöber, Dette, & Musch, 2010). In the current experiment, both scales had an unacceptable reliability, with Self Deceptive Enhancement Cronbach’s α=.46 and Impression Management α=.59. This is remarkable, since the scales usually show better reliability with α on average respectively mid-.60 and low .70 (Steenkamp, De Jong, & Baumgartner, 2010). I inspected Cronbach’s α at the group levels. For the Self Deceptive Enhancement scale, rapists
α=.18, nonsexual offenders α=.54, and community controls α=.51. For the Impression
Management scale, rapists α=.70, nonsexual offenders α=.53, community controls α=.59. This means that the items of the Self Deceptive Enhancement scale related less to each other for rapists than for nonsexual offenders and community controls; and that the items of the Impression Management scale related more to each other for rapists than for the other groups. No relevant explanations could be found for this pattern. It was decided to still use and report upon these scales, but to interpret their results with caution.
Ambivalent Sexism Inventory. Hostile and benevolent sexism were measured
explicitly with concordant scales of the Ambivalent Sexism Inventory (Glick & Fiske, 1996), which has 22 statements concerning men and women and their relationships in contemporary society, such as “women should be cherished and protected by men” (benevolent) or “women seek to gain power by getting control over men” (hostile). Participants answer on a 6-point Likert scale, ranging from 0 (disagree strongly) to 5 (agree strongly). A hostile sexism score and a benevolent sexism score were calculated separately by averaging concordant items. For the current experiment, a translation of the scale was made. It had a sufficient reliability, with Cronbach’s α = .78 for the Hostile Sexism scale and α = .69 for the Benevolent Sexism scale.
Revised Hostility Towards Women Scale. Hostility towards women was measured
explicitly with the revised Hostility Towards Women Scale (original by Check, 1984; revision by Lonsway & Fitzgerald, 1995). This scale contains 10 items that have to be answered with true or false, such as “When it really comes down to it, a lot of women are deceitful” or “I am
18
easily angered by women”. A total sum score is calculated after recoding reversed items. A translation of the revised scale was made for the current experiment. The translated scale appeared to be of sufficient reliability, with Cronbach’s α=.69.
Procedure
Participants were informed that they joined a research about attitudes towards women. It was emphasized to the TBS-patients that treatment providers and staff had no insight in the results. After obtaining informed consent, participants were invited to either a laboratory room (the community controls), or a quiet staff meeting room in the TBS-clinics (the offender groups). Participants completed a battery of tests and questionnaires on a laptop in approximately one hour, after which they received a €10-compensation. Current research used instruments from testing sessions of a larger research project. Not all administered instruments were used for this thesis. Instruments were administered in the following order. First, all participants completed the demographic information form. Second and third, the Deceitful-IAT and Prestige-IAT were completed. Because prior IAT-experience can cause learning effects (Greenwald et al., 2003), the order of Deceitful-IAT and Prestige-IAT was counterbalanced such that half of the participants started with the Deceitful-IAT and other half with the Prestige-IAT. Fourth, the AAT was completed. Then, an explicit rating of the stimuli of both IATs and AAT was administered. After this, a number of questionnaires was assessed, consisting of (in this order): the translated Ambivalent Sexism Inventory, the translated revised Hostility Towards Women Scale, the Dutch version of the Balanced Inventory of Desirable Responding, the Dutch version of the Social Potency Scale of the Multidimensional Personality Questionnaire to measure dominance (Eigenhuis, Kamphuis, & Noordhof, 2012), and finally a translated version of the Sexual Narcissism Scale to measure sexual narcissism (Widman & McNulty, 2010). As a final measure, the Dutch Adult Reading Test was administered. Afterwards participants were debriefed about the specific topic of the research. All implicit tests
19
and questionnaires were programmed and administered using Inquisit 4 (Millisecond Software). The AAT was programmed in JavaScript Project Builder and administered using Joystick Program. The Dutch Adult Reading Test was administered with pen and paper. This thesis is part of a larger research project about hostility towards women, which is a collaboration between three TBS-clinics (Van der Hoeven-Kliniek Utrecht, Pompekliniek Nijmegen, Oostvaarderskliniek Almere) and the University of Amsterdam.
Statistical Analyses
All statistical analyses were performed using SPSS Statistics 17.0. First, relevant demographic variables were compared between the rapist group, nonsexual offender group, and community control group. A one-way ANOVA tested whether groups were different on the Dutch Adult Reading Test, age, Self Deceptive Enhancement scale, and Impression Management scale. Post hoc Hochberg GT2 tests were used to localize the difference between groups. A Pearson Chi-square test assessed the effect of group on country of birth, marital status, and ever living together with a women for more than two years. Any variable that differed significantly between the groups was incorporated in the main analyses as covariate. The assumptions for ANOVA were assessed for the experimental variables hostile sexism score, benevolent sexism score, hostility towards women score, Deceitful-IAT D-score, Prestige-IAT D-score, and AAT D-score. The Shapiro-Wilks test explored normal distribution of the variables, and Levene’s test assessed homogeneity of variance, both at an α-level of .05. Finally, separate ANCOVAs assessed group differences on the Deceitful-IAT D-score, hostility towards women score, hostile sexism score, Prestige-IAT D-score, AAT D-score, and benevolent sexism score5. In case of significant results at a α-level of .05, follow-up ANCOVAs were conducted to see which groups differed. Effect sizes are reported conform Cohen’s
5 Five ANCOVAs were conducted simultaneously, which may inflate chances of Type I error. I did not make a correction for multiple comparisons, because I did not want to increase possibility of a Type II error, i.e. lose power to detect real differences. Results should be interpreted with this taken into consideration.
20
standards, where d=.20 is small, d=.50 is medium, and d=.80 is large (Cohen, 1992). Experimental power was good (.81) for separate ANCOVAs with two covariates, a sample of
N=114, an α-level of .05, and a medium effect size of d=.58.
Results Preliminary Analyses
A one-way ANOVA showed that the groups differed significantly on age,
F(2,109)=3.347, p=.039, d=.41. A post hoc Hochberg GT2 test indicated that the mean score
of the rapist group was significantly higher (7.37 years) than the community control group. Groups did not differ significantly on the Dutch Adult Reading Test, F(2,108)=2.714, p=.071,
d=.46, nor on the Self Deceptive Enhancement scale, F(2,110)=2.017, p=.138, d=.38, nor on
the Impression Management scale, F(2,110)=.119, p=.888, d=.09. All groups were thus equal in IQ-estimation and social desirable responding. A Pearson Chi-square test was conducted to test whether groups differed significantly on country of birth, marital status, and ever living together with a women for more than two years. Group differed significantly on marital status,
χ2(2, N=114)=14.443, p=.001, d=.77 (see Table 1 for frequencies). There was no significant effect of group on country of birth, χ2(2, N=114)=3.947, p=.139, d=.39, nor on ever living together with a women for more than two years, χ2(2, N=114)=4.687, p=.096, d=.41. Age and marital status differed significantly between groups. Separate ANOVAs were conducted to test the interaction of age and group, and the interaction of marital status and group on the dependent variables. There were no interactions significantly affecting any dependent variable at p=.05, therefore, age and marital status were included as covariates in subsequent analyses. For the sake of completeness, footnotes 6,7 report statistical results of analyses without the covariates.
The Shapiro-Wilks test for normality indicated that the data were statistically normal (p>.05) for the hostile sexism score, benevolent sexism score, Deceitful-IAT D-score, Prestige-IAT D-score, and AAT D-score. The hostility towards women score, however, was not
21
normally distributed, Shapiro-Wilk’s W(113)=.841, p<.001 (see Exploratory Analyses section below). Instead of an ANCOVA, hostility towards women was thus subsequently assessed with a Kruskal-Wallis H test (without the covariates age and marital status). Additionally, the homogeneity of variance assumption was met for all experimental variables, with Levene’s F having p-values above .05 for all variables. Correlation analyses assessed the relationships among the experimental variables (see Table 4). Hostile sexism and hostility towards women were moderately correlated, r=.38, p<.01, while the rest of the variables showed no relation to each other. This means that there is sufficient independence among the variables to conduct ANOVAs. Inspection of error percentages of the implicit tasks showed that the Prestige-IAT was a somewhat difficult test, with 22% of participants showing mean error percentages above 20. In comparison, only 1% on the Deceitful-IAT and 5% on the AAT showed mean error percentages above 20. One subject was excluded for all analyses except the AAT, because of insufficient understanding of Dutch and incomplete tests. After inspection of boxplots for outliers, one subject was excluded for analysis of the Hostility Towards Women Scale and one subject was excluded for analysis of the Deceitful-IAT because of extreme scores (i.e. respectively 4.88 SD and 2.98 SD from the means). This resulted in 112 subjects for analysis of the Deceitful-IAT, 113 for the Hostile Sexism scale, the Benevolent Sexism scale, the Prestige-IAT and the Hostility Towards Women Scale, and 114 for the AAT.
22
Table 4
Bivariate Pearson Correlations Between Experimental Variables
Hostile Sexism Benevolent Sexism Hostility Tw.Women Deceitful-IAT Prestige-IAT AAT 1. Hostile Sexism - 2. Benevolent Sexism .13 -
3. Hostility Towards Women .38** -.06 -
4. Deceitful-IAT .17 .01 .00 -
5. Prestige-IAT -.01 -.01 .11 .04 -
6. AAT .05 .17 .05 -.15 .10 -
** p<.01.
Attitudes Towards Women
Groups differed significantly on hostile sexism, F(2,108)=3.203, p=.045, d=.496. Follow-up analyses showed that rapists scored significantly lower than community controls,
F(1,57)=8.167, p=.006, d=.76 (see Table 5 for estimated marginal means). The difference
between rapists and nonsexual offenders was not significant, F(1,79)=2.695, p=.105, d=.37, nor was the difference between nonsexual offenders and community controls, F(1,78)=1.879,
p=.174, d=.31. This means that rapists showed significantly less hostile sexism than community
controls. Also, groups differed significantly on benevolent sexism, F(2,108)=3.528, p=.033,
d=.51. Follow-up analysis showed that rapists did not differ significantly from nonsexual
offenders, F(1,79)=.770, p=.383, d=.20, nor from community controls, F(1,57)=2.580, p=.114,
6 ANOVAs assessed the effect of group on all dependent variables (except hostility towards women) without covariates. Group differed significantly on the Deceitful-IAT, F(2,110)=4.150, p=.018, d=.55. Post hoc analyses showed that rapists (M=.03, SD=.38) and nonsexual controls (M=.04, SD=.39) scored significantly lower than community controls (M=.27, SD=.39). Group did not differ significantly on hostile sexism, F(2,110)=2.468, p=.089, d=.42, nor on benevolent sexism, F(2,110)=2.631, p=.077, d=.44, nor on Prestige-IAT score, F(2,110)=1.153, p=.320, d=.29, nor on AAT score, F(2,111)=1.472, p=.234, d=.33.
23 d=.42. Nonsexual offenders scored significantly higher than community controls, F(1,78)=8.251, p=.005, d=.65. This means that nonsexual offenders showed more benevolent
sexism than community controls. Also, groups differed significantly on the Deceitful-IAT,
F(2,107)=3.985, p=.021, d=.54. All estimated marginal means were positive, indicating that all
groups had a closer association between women-deceitful (and men-honest) than between women-honest (and men-deceitful). Follow-up analyses showed that rapists did not differ significantly from nonsexual offenders, F(1,78)=.037, p=.849, d=.00, and differed marginally significantly from community controls, F(1,57)=3.765, p=.057, d=.51. Nonsexual offenders scored significantly lower than community controls, F(1,77)=9.699, p=.003, d=.71. This means that a trend is visible where rapists have a weaker association between women and deceitful than community controls. Nonsexual offenders had a significant weaker association between women and deceitful than community controls. Groups did not differ significantly on the Prestige-IAT, F(2,108)=1.740, p=.180, d=.36. For the sake of completion, I mention that all estimated marginal means were negative, indicating that all groups had a closer association between wanting-prestige (and wanting-equal) than between wanting-prestige (and not-wanting-equal). Also, groups did not differ significantly on the AAT, F(2,109)=2.662, p=.074,
d=.44, which means that all groups were similar in their approach/avoidance tendency towards
women. However, the p-value was close to significant and the effect size small to moderate. Both rapists and nonsexual offenders had positive scores, whereas community controls had a negative score, as indicated by the estimated marginal means (i.e. group means when controlling for covariates at their mean scores). Rapists and nonsexual offenders thus showed an approach towards women, and community controls showed avoidance of women, however, the difference between the groups was only close to significant. Groups did not differ significantly on hostility towards women, χ2(2, N=112)=4.540, p=.103, d=.41. To have an overview of the differences between groups, means are displayed in Figure 1.
24 Table 5
Group Means of Dependent Variables for Total and Split Sample
Instrument Rapists Nonsexual
Offenders Community Controls Number (n) Single Relationship 26 6 39 13 12 18 Hostile Sexism Total
Single Relationship 1.61a 1.60a 1.54 1.86 1.87b 1.91 2.11a 2.45ab 1.85 Benevolent Sexism Total
Single Relationship 2.02 1.98 1.89 2.13a 2.07a 2.29 1.69a 1.55a 1.90 Deceitful-IAT Total Single Relationship .03† -.01a .25 .06a .10b -.06 .29†a .37ab .19 Prestige-IAT Total Single Relationship -.52 -.57 -.47 -.38 -.49 -.11 -.57 -.51 -.51 AAT Total Single Relationship .16 .11 .28 .16 .11 .26 -.04 -.09 .07 Hostility Towards Women Total
Single Relationship 46.50 31.32 16.00 61.51 39.71 22.69 58.32 46.50 17.33
Note. Estimated Marginal Means are displayed, with covariates age (total sample and split samples)
and marital status (total sample) evaluated at their mean scores. For Hostility Towards Women, mean ranks are provided. Means that share subscript a or b are significantly different at p<.05. † indicates a marginal significant difference at p=.057.
25
26
The Hostility Towards Women Scale was characterized by very low scores. The scale was reconsidered in exploratory analyses. Scores were categorized into “sum score of zero” (36% of participants) and “sum score higher than zero”. A comparison of the categories “zero” and “higher than zero” between groups, revealed no significant difference between groups, χ2(2,
N=112)=4.508, p=.105, d=.41.
The error percentages of the implicit tasks were analyzed. This could show if there was an effect of group on error percentage (instead of reaction time) in certain conditions. For the AAT, mean error percentages were calculated for the conditions men-pull, men-push, women-pull, and women-push. An ANOVA showed that group did not significantly differentiate the error conditions, man-push F(2,111)=.215, p=.807, d=.13; man-pull F(2,111)=.300, p=.741,
d=.14; woman-push F(2,111)=.387, p=.680, d=.17; woman-pull F(2,111)=.236, p=.790, d=.13.
For the Deceitful-IAT, mean error percentages were calculated for the men-honest/women-deceitful condition and the men-men-honest/women-deceitful/women-honest condition. An ANOVA showed that group did not significantly differentiate the error conditions, women-deceitful F(2,109)=1.721,
p=.184, d=.36; women-honest F(2,109)=.047, p=.954, d=.06. For the Prestige-IAT, mean error
percentages were calculated for the wanting-prestige/wanting-equal condition and the not-wanting-prestige/wanting-equal condition. An ANOVA showed that group did not significantly differentiate the error conditions, wanting-prestige F(2,110)=1.118, p=.331, d=.29; women-honest F(2,110)=1.165, p=.316, d=.29. This means that all groups had equal error percentages for all conditions on the AAT, the Deceitful-IAT, and the Prestige-IAT.
Since marital status differed significantly between groups, it was included in the ANCOVAs. However, marital status was scored as a dichotomous variable (single versus in a relationship), which makes its use in an ANCOVA questionable. Therefore, analyses were re-done with the sample divided into participants who were single and who were in a relationship.
27
Subsample n was very small in some cases (rapists in a relationship n=6), which results in low experimental power. Therefore, the following results must be used in addition to the main results and must be interpreted with caution. Groups differed significantly on hostile sexism for singles, F(2,72)=4.774, p=.011, d=.73, but not for men in a relationship, F(2,33)=.650, p=.529,
d=.407 (see Table 5). Single rapists scored significantly lower than single community controls,
F(1,34)=8.731, p=.006, d=1.01, and single nonsexual offenders scored significantly lower than
single community controls, F(1,48)=4.806, p=.033, d=.63. Groups did differ marginally significantly on benevolent sexism for singles, F(2,72)=2.945, p=.059, d=.57, and did not differ significantly for men in a relationship, F(2,33)=.994, p=.381, d=.49. Single nonsexual offenders scored significantly higher than single community controls on benevolent sexism,
F(1,48)=7.141, p=.010, d=.77. Groups differed significantly on the Deceitful-IAT for singles, F(2,71)=3.653, p=.031, d=.64, but not for men in a relationship, F(2,33)=2.335, p=.113, d=.75.
Single rapists scored significantly lower than single community controls, F(1,34)=4.284,
p=.046, d=.71, and single nonsexual offenders scored significantly lower than single
community controls, F(1,47)=5.905, p=.019, d=.71. The estimated marginal means for single community controls and nonsexual offenders were positive, indicating a closer association between women and deceitful than women and honest. However, the mean for single rapists was negative, indicating a slightly closer association between women and honest than women and deceitful. Groups did not differ significantly on the Prestige-IAT for singles, F(2,72)=.222,
p=.802, d=.16, nor for men in a relationship, F(2,33)=2.946, p=.066, d=.84. Groups did not
7 ANOVAs assessed the effect of group on all dependent variables (except hostility towards women) without the covariate age. First, the results for the single subsample are listed. Groups differed significantly on hostile sexism, F(2,73)=3.531, p=.034 (rapists scored lower than community controls), on benevolent sexism, F(2,73)=3.210, p=.046 (nonsexual offenders scored higher than community controls), and on the Deceitful-IAT, F(2,73)=4.675, p=.012 (rapists and nonsexual offenders scored lower than community controls). Groups did not differ significantly on hostility towards women, F(2,73)=1.774, p=.177, nor on the Prestige-IAT, F(2,73)=.158, p=.854, nor on the AAT, F(2,74)=1.297, p=.280.
Second, the results of ANOVAs for the relationship subsample are listed. Groups did not differ significantly on hostile sexism, F(2,34)=.621, p=.543, nor on benevolent sexism, F(2,34)=.915, p=.410, nor on hostility towards women, F(2,34)=1.692, p=.199, nor on the Deceitful-IAT, F(2,34)=2.311, p=.115, nor on the Prestige-IAT, F(2,34)=.3.031, p=.061, nor on the AAT, F(2,34)=1.356, p=.271.
28
differ significantly on the AAT for singles, F(2,73)=1.249, p=.293, d=.37, nor for men in a relationship, F(2,33)=1.314, p=.282, d=.57. Groups did not differ significantly on hostility towards women for singles, χ2(2, N=75)=4.715, p=.095, d=.52, nor for men in a relationship,
χ2(2, N=37)=2.571, p=.277, d=.54. Estimated marginal means and means ranks (for hostility towards women) for total and split sample are displayed in Figure 1. All variables show clear trends that are similar for total and split sample, with exception from the Deceitful-IAT. Here, a possible interaction between group and marital status seems present, because rapists in a relation deviate from the trend. Also, community controls in a relationship showed approach towards women, while single community controls show avoidance. Also, rapists in a relationship show the least benevolent sexism, compared to the control groups, whereas single rapists showed the most benevolent sexism.
Discussion
It was hypothesized that rapists would show more hostile and more positive attitudes towards women than nonsexual offenders and community controls. More specifically, it was expected that rapists would show a stronger implicit association between women and deceitful (versus women and honest), that rapists would show more approach tendency towards women, and that rapists would score higher on hostility towards women, hostile sexism, and benevolent sexism compared to nonsexual offenders and community controls. In the current research, rapists showed less hostile sexism than community controls. Nonsexual offenders showed more benevolent sexism than community controls. Rapists and nonsexual offenders showed a weaker association between women and deceitful than community controls. Groups did not differ significantly in hostility towards women, approach/avoidance tendency towards women, and association between not-wanting and prestige. Current data indicate that differences are more pronounced in single men, but this must be interpreted with caution since there are so few TBS-patients with a relationship. Men in a relationship show some remarkable deviations from the
29
trends, even though analyses showed no significant interactions between group and marital status on the dependent variables. Marital status seems to be an important factor and could be addressed in future research. Current results are partly unexpected, but relatively consistent. I used a battery of six tests, which to some extent all point to the same result: Rapists show the least hostile attitudes towards women, and thus more relatively positive attitudes towards women than community controls. The difference between rapists and nonsexual offenders was never significant, but visual inspection of trends indicates that rapists might have less hostile attitudes towards women than nonsexual offenders (true for all variables, except benevolent sexism and AAT score).
A few arguments might explain why rapists showed less hostile attitudes towards women than community controls. First, TBS-patients receive frequent treatment to address deviant cognitions. TBS-treatment might enhance offenders’ attitudes so much that they value women as more positive than community controls. Future research should incorporate time spent in treatment in analyses to investigate this possibility. Second, abstinence from sex and women in the clinic might have affected TBS-patients. Previous research showed that rapists experience loneliness (Marshall & Hambley, 1996; Seidman, Marshall, Hudson, & Robertson, 1994). Loneliness may push desires towards love and relationships. This may explain why offenders value women as more positive. Third, less hostile attitudes towards women might actually relate to sexual offense. Rapists with merely hostile attitudes towards women could choose not to interact with women. Positive attitudes and an automatic approach tendency might cause rapists to desire interaction with women. It might be that relatively positive attitudes urge them to be with women –to “merge” with women– rather than ignore them. Together with a deviance of any kind (sexual or cognitive), this may eventually lead to sexual offense.
30 Discrimination Rapists and Nonsexual Offenders
Statistical analyses could not discriminate between rapists and nonsexual offenders. There are a few explanations for this. First, both offender groups might have more socially desirable responses than community controls. We controlled for this option by assessing social desirability and found no differences between the groups. However, our instruments were less reliable than previously reported, which may have caused different results. Second, the shared environment in TBS-clinics (e.g., treatment) might have influenced rapists’ and nonsexual offenders’ attitudes in such a way that they are comparable to each other, but different from community controls. Third, the nonsexual offender group was a heterogeneous group. It consisted of patients convicted of many different offenses, including patients convicted of partner violence. It may be that this subgroup is more comparable to rapists in their attitudes towards women and obscures group differences. Fourth, a possible interaction between group and marital status may have manipulated the results. TBS-patients are more often single than the community controls, which might make the offender groups similar. I accounted for this by using marital status as a covariate, and by using a split sample of singles and men in a relationship in statistical analyses. However, as reported, these results must be interpreted with caution. Finally, it may be that current findings relate to general (violent) offense rather than sexual offense. For instance, a higher automatic approach tendency might be related to disinhibition or impulsivity, which may be similar across both offender groups.
Nonsignificant Results
Rapists did not show the most hostile attitudes towards women, as had been predicted by the literature. This might be explained by some arguments. First, the sample of Dutch TBS-patients might not be comparable to previous research consisting of prison samples and college samples. Although TBS-patients are convicted for offense, they are not the same as rapists in prison samples. Also, TBS-treatment might have influenced rapists’ attitudes in such a way that
31
they are less negative. As a final remark, albeit the group differences on the AAT were nonsignificant, they were informative to some extent. All groups showed an approach tendency towards women, whereas single community controls showed avoidance of women. This is in line with previous research, where approach/avoidance tendency could discriminate between single and attached men (Hofmann et al., 2009). In this research, men in a relationship showed facilitated avoidance, whereas in the current findings single men showed more avoidance.
Strengths and Limitations
The current conclusions have some limitations. Implicit measures cannot be used for individual diagnosis since they rely on group averages. This makes it harder to use current results in practice for individual purposes. Also, the IAT and AAT assess relative associations. For instance, the association women-deceitful is assessed simultaneously with the association men-honest in the Deceitful-IAT, whereas the instruction pull-women is assessed simultaneously with the instruction push-men in the AAT. It is impossible to disentangle whether any effect results from, for example, a women-deceitful or a men-honest association in this design. Furthermore, the Prestige-IAT appeared to be a difficult test, which might have influenced the results. More research is needed to develop a version of the Prestige-IAT that is easier to complete. Also, recruitment might have created a difference between the groups. The community control group was recruited with flyers and information letters that explicitly stated “what do you think of women?” to attract participants. TBS-patients were recruited without such explicit statements, in order to avoid suspiciousness and refusal of participation. The explicit statements may have resulted in a community control group with outspoken opinions, and therefore have created a community control group that is more different than the other groups. Finally, TBS-patients may not be directly comparable to prison samples, and therefore it is possible that results cannot be generalized to rapists in general. Current design also knows some strengths. I assessed hostile and positive attitudes towards women with both implicit and
32
explicit measures. This captured more variance than either measure alone and provided a broader basis to draw conclusions. Also, I used a TBS-patient sample instead of mere a community or college sample. This makes current findings more relevant for practical implications, for example in treatment. Finally, research to relative positive attitudes towards women among rapists is scarce. The current research contributes to the literature with new and remarkable findings.
Conclusion and Implications
Instead of finding that rapists are misogynists with an urge to merge, I found that rapists value women more positively than hostile, compared to community controls. Future research should assess whether or not positive attitudes emerge from treatment. It could be that relatively positive attitudes actually relate to offense. Positive attitudes might urge rapists to approach women, rather than ignore them. This, together with a deviance of any kind (sexual or cognitive), may eventually lead to sexual offense. Rapist treatment should be tailor-made and address the appropriate mechanisms to reduce the risk of sexual recidivism. Positive attitudes and their effect on behavior would be a new and import target for treatment and risk assessment. Instead of diminishing positive attitudes towards women, treatment might focus on the action tendencies or behavioral effects related to these attitudes. For example, the AAT has been used to retrain other attitudes (Wiers et al., 2011) and could easily be modified for re-training action tendencies towards women. Risk assessment of TBS-patients should incorporate positive attitudes towards women, to have a complete picture of offenders’ attitudes and risk of sexual recidivism, before allowing leaves into society. Also, it is important to incorporate positive attitudes towards women in future research to rapists’ attitudes.
33 References
Abrams, D., Viki, G. T., Masser, B., & Bohner, G. (2003). Perceptions of stranger and acquaintance rape: The role of benevolent and hostile sexism in victim blame and rape proclivity. Journal of Personality and Social Psychology, 84(1), 111–125. doi:10.1037/0022-3514.84.1.111
Babchishin, K. M., Nunes, K. L., & Hermann, C. A. (2013). The validity of implicit association test (IAT) measures of sexual attraction to children: A meta-analysis. Archives of Sexual
Behavior, 42(3), 487–499. doi:10.1007/s10508-012-0022-8
Berliant, J., Nunes, K. L., & Sperling, A. (2011). Is inconsistency between implicit and explicit
perceptions of women associated with likelihood to rape? Poster presented at the 30th
Annual Research and Treatment Conference of the Association for the Treatment of Sexual Abusers, Toronto, ON.
Burt, M. R. (1980). Cultural myths and supports for rape. Journal of Personality and Social
Psychology, 38(2), 217–230.
Check, J. V. P. (1984). The hostility towards women scale (Doctoral dissertation, University of Manitoba). Retrieved from http://hdl.handle.net/1993/23279
Check, J. V. P., Malamuth, N., Elias, B., & Barton, S. (1985). On hostile ground. Psychology
Today, 19(4), 56–61.
Cohen, J. (1992). A Power Primer. Psychological Bulletin, 112(1), 155–159.
De Houwer, J. (2006). What are implicit measures and why are we using them? Handbook of
Implicit Cognition and Addiction. doi:10.1037/0033-2909.127.6.853
De Lugt, M., Ponds, R., Rozendaal, N., Houx, P., & Jolles, J. (1995). The Dutch Adult Reading Test (DART): A measure of (premorbid) intelligence level? In Maastricht aging study:
34
Eigenhuis, A., Kamphuis, J. H., & Noordhof, A. (2012). Development and validation of the Dutch brief form of the Multidimensional Personality Questionnaire (MPQ-BF-NL).
Assessment, 20(5), 565–575. doi:10.1177/1073191112444920
Fazio, R. H. (2001). On the automatic activation of associated evaluations: An overview.
Cognition & Emotion, 15(2), 115–141. doi:10.1080/02699930125908
Forbes, G. B., Adams-Curtis, L. E., & White, K. B. (2004). First- and second-generation measures of sexism, rape myths and related beliefs, and hostility toward women: Their interrelationships and association with college students’ experiences with dating aggression and sexual coercion. Violence Against Women, 10(3), 236–261. doi:10.1177/1077801203256002
Glick, P., & Fiske, S. T. (1996). The ambivalent sexism inventory: Differentiating hostile and benevolent sexism. Journal of Personality and Social Psychology, 70(3), 491–512. Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual
differences in implicit cognition: The implicit association test. Journal of Personality and
Social Psychology, 74(6), 1464–1480. doi:10.1037//0022-3514.74.6.1464
Greenwald, A. G., Nosek, B. A, & Banaji, M. R. (2003). Understanding and using the implicit association test: I. An improved scoring algorithm. Journal of Personality and Social
Psychology, 85(2), 197–216. doi:10.1037/0022-3514.85.2.197
Hanson, R. K., Harris, A. J. R., Scott, T.-L., & Helmus, L. (2007). Assessing the risk of sexual
offenders on community supervision: The Dynamic Supervision Project (Report No.
2007-05). Ottawa: Public Safety Canada.
Hanson, R. K., & Morton-Bourgon, K. E. (2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 118 prediction studies. Psychological Assessment,
35
Helmus, L., Hanson, R. K., Babchishin, K. M., & Mann, R. E. (2013). Attitudes supportive of sexual offending predict recidivism: a meta-analysis. Trauma, Violence & Abuse, 14(1), 34–53. doi:10.1177/1524838012462244
Hofmann, W., Friese, M., & Gschwendner, T. (2009). Men on the “pull”: Automatic approach/avoidance tendencies and sexual interest behavior. Social Psychology, 40(2), 73–78. doi:10.1027/1864-9335.40.2.73
Johnson, G. M., & Knight, R. A. (2000). Developmental antecedents of sexual coercion in juvenile sexual offenders. Sexual Abuse: A Journal of Research and Treatment, 12(3), 165–178.
Kaplan, K. J. (1972). On the ambivalence-indifference problem in attitude theory and measurement: A suggested modification of the semantic differential technique.
Psychological Bulletin, 77(5), 361–372. doi:10.1037/h0032590
Larsen, J. T., McGraw, A. P., & Cacioppo, J. T. (2001). Can people feel happy and sad at the same time? Journal of Personality and Social Psychology, 81(4), 684–696. doi:10.1037/0022-3514.81.4.684
Lonsway, K. A., & Fitzgerald, L. F. (1994). Rape myths: In review. Psychologyof Women
Quarterly, 18, 133–164.
Lonsway, K. A., & Fitzgerald, L. F. (1995). Attitudinal antecedents of rape myth acceptance: A theoretical and empirical reexamination. Journal of Personality and Social Psychology,
68(4), 704–711. doi:10.1037//0022-3514.68.4.704
Malamuth, N. M. (1986). Predictors of naturalistic sexual aggression. Journal of Personality
and Social Psychology, 50(5), 953–962. doi:10.1037//0022-3514.50.5.953
Malamuth, N. M., Sockloskie, R., Koss, M. P., & Tanaka, J. S. (1991). Characteristics of aggressors against women: Testing a model using a national sample of college students.
36
Mann, R. E., Hanson, R. K., & Thornton, D. (2010). Assessing risk for sexual recidivism: some proposals on the nature of psychologically meaningful risk factors. Sexual Abuse: A
Journal of Research and Treatment, 22(2), 191–217. doi:10.1177/1079063210366039
Marshall, W. L., & Hambley, L. S. (1996). Intimacy and loneliness, and their relationship to rape myth acceptance and hostility toward women among rapists. Journal of Interpersonal
Violence, 11(4), 586–592. doi:10.1177/088626096011004009
Nunes, K. L., Hermann, C. A., & Ratcliffe, K. (2013). Implicit and explicit attitudes toward rape are associated with sexual aggression. Journal of Interpersonal Violence, 28(13), 2657–2675. doi:10.1177/0886260513487995
Paulhus, D. L., & John, O. P. (1998). Egoistic and moralistic biases in self-perception: The interplay of self-deceptive styles with basic traits and motives. Journal of Personality,
66(6), 1025–1060. doi:10.1111/1467-6494.00041
Rinck, M., & Becker, E. S. (2007). Approach and avoidance in fear of spiders. Journal of
Behavior Therapy and Experimental Psychiatry, 38(2), 105–120. doi:10.1016/j.jbtep.2006.10.001
Robertson, K., & Murachver, T. (2007). Correlates of partner violence for incarcerated women and men. Journal of Interpersonal Violence, 22(5), 639–655. doi:10.1177/ 0886260506298835
Schmand, B., Lindeboom, J., & Van Harskamp, F. (1992). NLV Nederlandse Leestest voor
Volwassenen handleiding [Manual Dutch Adult Reading Test]. Lisse, The Netherlands:
Swets & Zeitlinger.
Seidman, B. T., Marshall, W. L., Hudson, S. M., & Robertson, P. J. (1994). An examination of intimacy and loneliness in sex offenders. Journal of Interpersonal Violence, 9(4), 518– 534. doi:10.1177/088626094009004006
