Abstract
The literature on male sexual aggression indicates that victim age is a valid discriminator among sexual offenders. In particular, the rapist/child molester distinction based on victim age has received substantial empirical support (Bard et al., 1987; Robertiello & Terry, 2007). Studies have shown that rapists and child molesters differ in their developmental trajectories, clinical profiles, and criminal behavior, supporting a typological distinction between the two groups (Bard et al., 1987; Knight & Prentky, 1990). Recent evidence suggests that rapists and child molesters also differ on several neurological indicators. Neurodevelopmental deficits associated with pedophilia include shorter stature (Cantor et al., 2007), non-right-handedness (Cantor et al., 2005b), lower IQ scores (Cantor, Blanchard, Robichaud, & Christensen, 2005a), more head injuries in childhood (Blanchard et al., 2003), white matter deficiencies in the temporal and parietal lobes (Cantor et al., 2008), and gray matter deficiencies in the frontostriatal circuitry and cerebellum (Schiffer et al., 2007). Finally, distinct etiological models seem to map the sexual assault antecedents of rapists and child molesters. Sexual coercion against age-appropriate females appears to operate through psychopathy and increased sexualization (Knight & Sims-Knight, 2003, 2004; Vega & Malamuth, 2007), whereas a model that also includes sexual inadequacy and child fantasy may better predict sexual coercion against children (Daversa & Knight, 2007). Despite this evidence for substantial distinctions between rapists and child molesters, the role of alcohol and drug use in differentiating between those two groups remains largely unexplored.
Perpetrator Use
Research has examined mostly alcohol use among those prone to sexually aggress against women. In this population, alcohol has been found to play a significant, multifaceted role in college, community, and criminal samples (Abbey, 2008). Carr and VanDeusen (2004) noted that heavy alcohol use along with pejorative gender-related attitudes and pornography consumption significantly predicted perpetration of sexual violence by college men against women. Alcohol use also appears to be an important discriminator of college men who committed repeated sexual assaults against women (e.g., Abbey & McAuslan, 2004). In a community sample of single men, Abbey, Parkhill, BeShears, Clinton-Sherrod, and Zawacki (2006) found that problematic drinking had a direct and an indirect effect on the number of sexual offenses committed against women. They proposed that heavy drinking was associated with a desire for sexual dominance, which in turn was related to positive attitudes toward casual sexual relationships and peer pressure to engage in sex, both of which had direct effects on the number of sexual assaults committed. In an incarcerated male sample, alcohol abuse was shown to affect the frequency of adolescent sexual offenses against age-appropriate females both directly and indirectly through the activation of macho attitudes and aggressive, sexist fantasies (Johnson & Knight, 2000).
Research suggests that substance use is related not only to the frequency of sexual offending but also to sexual assault outcomes. Studies investigating sexual offending against women have found that perpetrator substance use before a sexual crime predicts perpetrator aggressiveness, victim injury, and the odds of penetration (Abbey, Clinton-Sherrod, McAuslan, Zawacki, & Buck, 2003; Testa, Vanzile-Tamsen, & Livingston, 2004). The relationship of perpetrator substance use to perpetrator aggressiveness and victim injury appears to be positive and linear, such that aggressiveness and the odds of injury increase proportionally with the amount of alcohol or drugs consumed prior to the crime. In contrast, the relationship between perpetrator substance use and odds of penetration seems to be curvilinear, with the odds of penetration increasing at low levels of alcohol or drug use and decreasing at high levels of consumption (Abbey et al., 2003; Testa et al., 2004).
There are several mechanisms, reviewed by Seto and Barbaree (1995) and more recently by Abbey (2008), through which perpetrator alcohol use might be related to sexual aggression. First, alcohol use may cause aggression directly (e.g., alcohol use may trigger the release of aggressive impulses). Second, alcohol use may cause aggression indirectly (e.g., alcohol use may activate perpetrators’ expectancies about alcohol’s disinhibiting effects, which in turn might lead to aggressive behavior). Third, alcohol use may be associated with aggressive behavior as a mediator of aggressive intentions (e.g., perpetrators’ intentions to commit a sexually aggressive act might motivate their use of alcohol to mitigate feelings of responsibility). Fourth, alcohol use may be related to aggression in an interactive way, where aggressive dispositions interact with alcohol to produce aggressive behavior. And fifth, alcohol use may be associated with aggression only because both are directly related to aggressive characteristics of the perpetrator (e.g., antisocial perpetrators might be more likely to both use alcohol and commit aggressive acts; Knight & Sims-Knight, 2011). Though less studied in this context, many of those mechanisms may apply equally well to perpetrator drug use.
The extant literature indicates that alcohol’s effects on sexual aggression are likely to be mediated by both alcohol expectancies and cognitive myopia (Abbey, Zawacki, Buck, Clinton, & McAuslan, 2004; Crowe & George, 1989). In particular, alcohol use is thought to trigger offenders’ expectancies about alcohol’s sexually disinhibiting effects, which in turn can lead to sexual aggression. Alcohol use is also thought to cause cognitive myopia by focusing offenders’ attention on immediate, salient cues (e.g., their own sexual arousal) and by limiting their attentiveness to more abstract, inhibitory cues (e.g., the possible negative consequences of sexual coercion), increasing the likelihood of sexual aggression (Crowe & George, 1989; Davis, 2010).
Victim Use
Findings on the relationships of preassault victim substance use to sexual assault outcomes are less consistent. Whereas Abbey et al. (2003) found that the amount of alcohol consumed by the victim was positively and linearly related to sexual assault severity (ranging from verbally coerced nonpenetrating sexual contact to physically forced penetration), Testa et al.’s (2004) results suggested a curvilinear relationship, with the odds of penetration increasing only at high levels of alcohol or drug intoxication. In addition, Abbey et al. found no relationship between victim alcohol use and perpetrator aggressiveness, but Testa et al. found a marginally significant negative relationship between victim intoxication and victim injury. It is noteworthy that Testa et al.’s study included a female community sample, whereas Abbey et al.’s study included a sample of college men. In a study of college women, Abbey, Clinton, McAuslan, Zawacki, and Buck (2002) found a negative linear relationship between victim alcohol use and victim resistance against sexual aggression, victim injury, and perpetrator aggressiveness. In another sample of college women, Ullman, Karabatsos, and Koss (1999b) also found a negative relationship between victim alcohol use and perpetrator aggressiveness. In contrast, in a national sample of college men, the same authors found a positive relationship between victim alcohol use and perpetrator aggressiveness (Ullman, Karabatsos, & Koss, 1999a).
It is unclear if and how victim substance use is associated with perpetrator aggression. Several investigators have suggested that victim substance use may be related to decreased perpetrator aggression because less force is necessary to coerce sexual contact with an intoxicated victim (Abbey et al., 2002; Ullman et al., 1999b). However, victim alcohol use has also been linked to reduced concern about injury (Macy, Nurius, & Norris, 2006), which may increase drinking victims’ vulnerability to perpetrator aggression. Finally, some researchers believe that positive associations between victim substance use and perpetrator aggression may reflect offenders’ proclivity to target intoxicated victims for sexually aggressive acts (Ullman et al., 1999a).
The Present Study
This study investigated the associations of preassault perpetrator and victim substance use with rapists and child molesters’ aggression in sexual crimes. Perpetrator substance use is likely to have differential associations with rapists and child molesters’ aggression. Bard et al.’s (1987) finding that rapists were more likely than child molesters to “act out while drinking” (p. 215) suggests that alcohol use might be more strongly related to the sexual aggressiveness of the former compared with the latter. Consistent with this finding, we hypothesized that perpetrator alcohol and drug use would be associated with a greater increase in rapists’ aggression in sexual crimes compared with that of child molesters.
The extant research on victim substance use is too scarce and inconsistent to allow for informed speculation on the ways in which such use might be differentially related to the aggression of child molesters versus rapists in sexual crimes. On the basis of the most common finding suggesting a negative relationship between victim alcohol use and perpetrator aggression against women, we hypothesized that victim substance use would be associated with decreased aggression among both rapists and child molesters. This analysis was secondary due to the small sample of child molesters’ victims who were under the influence of alcohol or drugs.
Method
Participants
The institutional review boards at both Brandeis University and the Massachusetts Treatment Center (MTC) approved the procedures described in this report. The sample (N = 518) comprised 245 rapists and 273 child molesters selected from the approximately 1,900 male prison inmates who had been evaluated at the MTC for potential civil commitment between 1959, when the MTC opened and civil commitment in Massachusetts began, and 1991, when Massachusetts discontinued civil commitment of sexual offenders and federal funding for grants supporting coding ended. Chapter 123A of the Massachusetts General Laws mandated the civil commitment of offenders found “sexually dangerous.” A sexually dangerous offender was defined as “one whose misconduct in sexual matters indicates a general lack of power to control sexual impulses as evidenced by repetitive or compulsive behavior and violence or aggression” (Swanson, 1960, p. 230). The MTC was established to evaluate candidates for civil commitment and to provide confinement and treatment for individuals committed as “sexually dangerous.”
Figure 1 contains a consort diagram showing the number of participants of the 1,900 prison inmates who were selected for coding at each stage. Of the 1,900 inmates evaluated at the MTC, 570 were committed for 1 day to life. The remaining 1,330 inmates were released back to prison. The archival records of all 570 committed offenders were coded using the MTC dictionary (see Bard et al., 1987 for a description). Two comparison samples of 200 participants each were independently selected from the 1,330 uncommitted offenders for record review and coding. The first sample was obtained through random sampling and the second was matched on age, marital status, and number of prior crimes with the 570 committed offenders. As the two samples overlapped in 67 offenders, the total number of uncommitted offenders selected was 333. Overall, we selected 903 committed and uncommitted offenders. One committed offender for whom little archival information was available was dropped from the sample. Thus, there were 902 offenders among whom we identified child molesters and rapists.
Consort diagram showing the number of participants selected for coding at each stage
To be designated as either a child molester or a rapist, an offender had to have committed a sexually motivated assault involving physical contact with a victim. Offenders were classified as rapists or child molesters according to the criteria of the MTC typological systems for rapists and child molesters (MTC: R3 [Knight & Prentky, 1990] and MTC: CM3 [Knight, Carter, & Prentky, 1989], respectively). An offender was classified as a rapist if either (a) all of his victims were 15 years old or older or (b) all of his victims, with the exception of one victim between the ages of 12 and 15 years, were above 15 years. If the offender was younger than 22 years, he was designated as a rapist even if his victims were between the ages of 12 and 15 years. An offender was classified as a child molester if all of his victims were both below the age of 16 years and no less than 5 years younger than he was. Participants were excluded if they simultaneously met both rapist and child molester classification criteria or if they had otherwise manifested a lack of victim-age preference (e.g., victims were both below the age of 12 years and above the age of 15 years). Offenders were also eliminated from the sample if they had exclusively committed incestuous crimes (i.e., victims were members of the perpetrator’s nuclear family or were friends of incest victims and had been assaulted concurrently). Exclusively incestuous offenders were rare in the MTC sample and not frequent enough to include as a separate group. In this sample, the vast majority of those who sexually abused their own children also abused other children (separately from their own children) and were thus coded as child molesters. Finally, we excluded participants whose files did not include adequate information to allow classification in the MTC typologies or offer sufficiently detailed descriptions of their index sexual crimes. This final criterion tended to exclude offenders who were evaluated early in the history of the MTC and were not committed.
Demographic information is presented for those among the 518 selected offenders for whom it was available (n = 413 to 516). Summary statistics indicated that 90% of participants were White, and 43% had ever been married. The mean educational attainment in the sample was slightly above ninth grade (M = 9.14, SD = 2.30, range = 2 to 16), and the average full-scale IQ was approximately normal (M = 96.40, SD = 15.44, range = 56 to 142). Offenders had committed their first serious sexual crime at an average age of 22 years (M = 22.07, SD = 7.43, range = 10 to 56), and 33% had a juvenile penal record. The mean number of incarcerations in adulthood, including the imprisonment at the time of assessment, was 2.69 (SD = 2.66, range = 1 to 36).
Several differences in demographics emerged between the rapist and child molester subgroups. The percentage of non-White offenders was significantly larger among rapists (15%) than among child molesters (5%), χ2(1, n = 516) = 13.35, p < .001. However, rapists were no more likely than child molesters to have ever been married, χ2(1, n = 512) = 0.96, p > .1. Approximately, 40% of rapists had ever been married compared with 45% of child molesters. Rapists’ mean educational attainment, (M = 9.33, SD = 2.05) was only marginally greater than child molesters’ (M = 8.95, SD = 2.51), t(464) = −1.82, p < .1. Consistent with previous findings, rapists had a significantly higher mean full-scale IQ score than did child molesters, t(411) = −2.47, p < .05. The respective mean scores were 98.44 (SD = 14.36) and 94.69 (SD = 16.12). Child molesters were significantly older than rapists at the time of their first serious sexual offense, t(431) = 6.02, p < .001. Mean age at this time was 23.84 years (SD = 8.79) for child molesters and 20.12 years (SD = 4.89) for rapists. The percentage of rapists with a juvenile penal record (40%) significantly exceeded that of child molesters (26%), χ2(1, n = 508) = 11.30, p < .01. Nonetheless, child molesters had a significantly higher average number of incarcerations in adulthood than did rapists, t(349) = 3.32, p < .01. The respective mean numbers of adult imprisonments were 3.10 (SD = 3.27) and 2.28 (SD = 1.76).
Procedure
Data were obtained from offenders’ archival records at the MTC. Those documents were compiled (a) during a 60-day long observation period when offenders were evaluated for potential civil commitment and (b) after commitment when patients’ adjustment and progress were documented. Records from the observation period comprised clinical assessments taken at the MTC as well as external documents such as school files, police reports, social service papers, and employment records. Postcommitment data stemmed from clinical observations, diagnostic assessments, and treatment progress reports.
The majority (n = 331) of the records in the present study were dual coded by two independent research assistants on demographic, developmental, criminal, and clinical measures in the MTC dictionary. The MTC dictionary is an extensive archival records rating instrument that took a year to develop. All items in the dictionary were iteratively fashioned to maximize appropriateness for the archival record source and reliability. Detailed concrete descriptors are provided for each level of each variable. There are also detailed decision criteria about when to consider “missing” information truly missing versus indicative of the absence of the behavior in the offender. Research assistants were all BA- or MA-level full-time employees. They underwent an intensive 3-month training program in which they rated practice files and consensed their ratings with an expert. They were not permitted to code new files until they had demonstrated adequate reliability. In one case, a coder had to be assigned to other tasks because she was unable to attain adequate reliability.
For each offender, coders independently rated every crime for which there were sufficient details in the archival records. Each crime was rated on a number of offender and victim behaviors and crime characteristics. For the present study, we focused only on the most recent crime for which data were available, which was typically also the index crime for commitment. Most often, this crime was documented in victim and police reports and was the offense described in the greatest detail.
Coders were unaware of the hypotheses in the present study. The independent ratings of coders were entered into a computer program that identified discrepancies between ratings. All discrepancies were resolved by a consensus procedure that often required a return to detailed notes or to the original file. Consensus ratings were analyzed for all dual-coded files, and the ratings of a single coder were used for single-coded files. Reliability was measured by Pearson correlations between preconsensus ratings in the case of continuous variables and kappa statistics in the case of dichotomous variables. The reported reliability estimates are conservative as they do not adjust for the fact that most of the scores used in our analyses are from consensed ratings (cf. Roff, 1981).
Measures
Sexual offender type
The offender type variable was coded dichotomously to distinguish between child molesters (0) and rapists (1) in accordance with the classification system described above.
Perpetrator substance use
Perpetrator alcohol use and perpetrator drug use were measured as two separate variables. Alcohol use was coded dichotomously, with 1 indicating the presence of alcohol in the offender during the sexual crime and 0 denoting the absence of alcohol in the offender. The reliability of the preconsensus ratings of perpetrator alcohol use (κ statistic) was .98.
Drug use was coded analogously and distinguished between the presence (1) and absence (0) of any nonprescription drug in the perpetrator during the sexual crime (with the most common drugs being marijuana, amphetamines, cocaine, and heroin). The reliability of the preconsensus ratings of perpetrator drug use (κ statistic) was .98.
Victim substance use
Due to data limitations, no distinction was made between alcohol and drug use for victims. Victim substance use was coded dichotomously, with 1 denoting the presence of alcohol and/or any nonprescription drug in the victim during the sexual crime and 0 indicating the absence of such substances in the victim. The reliability of the preconsensus ratings of victim substance use (κ statistic) was .90.
Expressive aggression
Expressive aggression measures violence before or during the sexual crime resulting from the offender’s anger or desire to harm or humiliate the victim. Such aggression does not serve an instrumental purpose and was judged to exceed what was necessary to gain victim compliance. Expressive aggression was measured on a scale from 0 to 7, with 0 denoting no aggression and 7 indicating extreme aggression (e.g., victim was killed or mutilated). Table 1 shows what the different levels of aggression correspond to. The Pearson correlation between the preconsensus ratings of expressive aggression was .78.
Expressive Aggression Scale
Analytic Plan
Before analyzing the associations of perpetrator and victim substance use with perpetrator aggression, we examined some preliminary statistics. We computed mean expressive aggression during the most recent sexual crime in the entire sample as well as separately in rapists and child molesters. Using an independent t test, we tested for a significant difference in mean expressive aggression between rapists and child molesters. We also computed the percentages of offenders who were under the influence of alcohol, any nonprescription drug, or both during their most recent sexual crime. We calculated those percentages separately for rapists and child molesters and tested for significant differences between the two groups using chi-square tests. We then computed the percentage of victims who were under the influence of any substance during the sexual crime, and we tested for a significant difference in the rate of use among rapists’ versus child molesters’ victims with chi-square tests.
Finally, we examined separately the associations of perpetrator alcohol use, perpetrator drug use, and victim substance use with perpetrator expressive aggression. Unfortunately, our sample was too small to permit simultaneous consideration of all three substance use variables in one regression. Thus, we conducted three hierarchical multiple regressions in which we entered offender type, the substance use variable under investigation, and the interaction between offender type and the substance use variable in three sequential blocks. The interactions tested whether the associations of perpetrator alcohol use, perpetrator drug use, and victim substance use with expressive aggression differed significantly between rapists and child molesters.
Results
Preliminary statistics showed that mean expressive aggression during the most recent sexual crime was 2.25 (SD = 2.13) on a scale from 0 to 7, indicating minimal physical aggression (e.g., holding, pushing). Whereas rapists’ mean expressive aggression was 3.21 (SD = 2.08), that of child molesters was significantly lower at 1.23 (SD = 1.66), t(387) = −10.54, p < .001. On average, rapists’ expressive aggression was slightly hurtful (e.g., squeezing, slapping), whereas child molesters’ expressive aggression consisted of aggressive verbalizations.
Further analyses revealed that 62% of offenders were under the influence of alcohol, any nonprescription drug, or both during their most recent sexual crimes. Significantly more rapists (74%) were under the influence of any such substance than were child molesters (51%), χ2(1, n = 435) = 24.28, p < .001. The breakdown between alcohol and drug use was as follows: 60% of offenders were under the influence of alcohol, at least 19% were under the influence of any nonprescription drug, and at least 17% were under the influence of both. A significantly greater percentage of rapists (70%) than child molesters (50%) were under the influence of alcohol, χ2(1, n = 440) = 19.14, p < .001. Similarly, the percentage of rapists who were under the influence of any nonprescription drug (37%) significantly exceeded that of child molesters (8%), χ2(1, n = 387) = 47.22, p < .001. The percentage of rapists who were under the influence of both alcohol and any nonprescription drug (33%) was also significantly greater than that of child molesters (7%), χ2(1, n = 392) = 42.67, p < .001.
Twelve percent of victims were under the influence of alcohol, any nonprescription drug, or both during the sexual offense. Not surprisingly, a significantly greater percentage of rapists’ victims (21%) were under the influence of any substance compared with the victims of child molesters (4%), χ2(1, n = 372) = 27.08, p < .001.
Table 2 shows the breakdown of rapists and child molesters by all combinations of perpetrator alcohol use, perpetrator drug use, and victim substance use. It is noteworthy that whenever victim substance use was present, perpetrator substance use was present as well. In addition, most cases of perpetrator drug use also involved perpetrator alcohol use. To test for possible confounding between those variables’ effects on expressive aggression, we conducted additional analyses, when appropriate.
Number of Rapists and Child Molesters by Perpetrator Alcohol Use, Perpetrator Drug Use, and Victim Substance Use (N = 298)
Regression analyses were conducted to examine the relationships of perpetrator and victim substance use with perpetrator aggression. Our first prediction was that perpetrator alcohol use would be associated with a larger increase in rapists’ expressive aggression compared with that of child molesters. To test this prediction, a hierarchical multiple regression was computed, such that offender type, perpetrator alcohol use, and the interaction between type and use were entered in three sequential blocks. The overall R2 was significant, F(3, 394) = 40.70, p < .001, accounting for 24% of the variance in expressive aggression (adjusted R2 = .23). Table 3 displays the results of the regression. Contrary to our expectation, the interaction term was not significant, indicating that the positive relationship between alcohol consumption and expressive aggression did not vary between rapists and child molesters. Figure 2 depicts the mean expressive aggression of child molesters and rapists who were not under the influence of alcohol and those who were under the influence of alcohol.
Summary of Simultaneous Regression Analysis for Alcohol Variables Predicting Expressive Aggression (N = 398)
Note: R2=.24.
p < .05. **p < .01. ***p < .001.
Mean expressive aggression of child molesters and rapists under the influence of alcohol and not under the influence of alcohol
To test the prediction that perpetrator drug use would be associated with a larger increase in rapists’ expressive aggression than in that of child molesters, we repeated the prior hierarchical regression analysis replacing perpetrator alcohol use with perpetrator drug use. The overall R2 was significant, F(3, 343) = 39.36, p < .001, accounting for 26% of the variance in expressive aggression (adjusted R2 = .25). Table 4 displays all regression coefficients and their standard errors. Figure 3 depicts the mean expressive aggression of child molesters and rapists not under the influence of any drugs and those under the influence of any drugs. Contrary to our expectation, drug use was associated with a significantly smaller increase in rapists’ expressive aggression than in that of child molesters. Follow-up analyses to test the significant interaction revealed that whereas drug use was a significant predictor of expressive aggression among child molesters (b = 1.46, p < .01), it was not significantly related to expressive aggression among rapists (b = 0.25, p > .1).
Summary of Simultaneous Regression Analysis for Drug Variables Predicting Expressive Aggression (N = 347)
Note: R2 = .26.
p < .05. **p < .01. ***p < .001.
Mean expressive aggression of child molesters and rapists under the influence of drugs and not under the influence of drugs
As noted above, most cases of perpetrator drug use also involved perpetrator alcohol use. As the relationship between perpetrator alcohol use and expressive aggression did not vary between rapists and child molesters, there was no reason to believe that the significance of the perpetrator drug use × offender type interaction was attributable to the overlap between perpetrator alcohol use and perpetrator drug use.
Finally, we examined the associations between victim substance use and rapists’ and child molesters’ expressive aggression. It was predicted that victim use would be negatively associated with aggression among both rapists and child molesters. To test this prediction, we conducted another hierarchical multiple regression analysis with expressive aggression as the dependent measure. Offender type, victim substance use, and the interaction between type and use were entered in three sequential blocks. The overall R2 was significant, F(3, 334) = 37.41, p < .001, accounting for 25% of the variance in expressive aggression (adjusted R2 = .25). Table 5 displays the regression coefficients and their standard errors. Follow-up analyses to test the significant interaction showed that whereas the relationship of victim substance use to child molesters’ expressive aggression was negative but not significant (b = −0.59, p > .1), its relationship to rapists’ expressive aggression was positive and significant (b = 1.09, p < .01). Figure 4 depicts the mean expressive aggression of child molesters and rapists whose victims were not under the influence of any substance and those whose victims were under the influence of any substance.
Summary of Simultaneous Regression Analysis for Victim Variables Predicting Expressive Aggression (N = 338)
Note: R2 = .25.
p < .05. **p < .01. ***p < .001.
Mean expressive aggression of child molesters and rapists whose victims were under the influence of any substance and not under the influence of any substance
As shown in Table 2, almost all cases of victim substance use involved perpetrator alcohol use. To test whether the positive relationship between victim substance use and rapists’ expressive aggression could be fully explained by concurrent perpetrator alcohol use, we conducted an additional follow-up analysis. In a sample of rapists who were all under the influence of alcohol during their sexual offense, we examined the association of victim substance use with expressive aggression. This association was positive and statistically significant (b = 0.87, p < .05), suggesting that the relationship between victim substance use and rapists’ expressive aggression is not a result of the overlap between victim substance use and perpetrator alcohol use.
Discussion
Numerous studies have shown that rapists and child molesters have distinct developmental, criminal, and clinical histories (Bard et al., 1987) and differ significantly on important neurodevelopmental indicators (e.g., Cantor et al., 2005b, 2008). Distinct etiological models have also been proposed to map the sexual assault antecedents of the two groups (Daversa & Knight, 2007; Knight & Sims-Knight, 2003, 2004). Despite the evidence for typological differences between rapists and child molesters, the role of substance use in differentiating between the two groups has remained largely unexplored. We attempted to fill this gap by examining the associations of perpetrator and victim alcohol and nonprescription drug use before a sexual crime with rapists’ and child molesters’ aggression during the crime.
Consistent with some suggestive findings in the extant literature, we hypothesized that perpetrator substance use would be associated with a greater increase in expressive aggression among rapists than among child molesters. We also predicted that victim substance use would be negatively related to expressive aggression among both rapists and child molesters. Contrary to our expectation, we found that the magnitude of the positive association between perpetrator alcohol use and aggression did not differ between rapists and child molesters. We also found that perpetrator drug use was associated with an increase in aggression among child molesters, but not rapists, and that victim substance use was associated with an increase in aggression among rapists, but not child molesters.
Additional research is necessary to determine why perpetrator drug use was associated with an increase in aggression among child molesters but not rapists, even though perpetrator alcohol use showed no differential associations. Such research should attempt to explicate the mechanisms underlying the associations of perpetrator alcohol use and perpetrator drug use to sexual aggression. An investigation into the differential associations of perpetrator drug use with aggression in the sexual crimes of child molesters versus rapists ought to consider potentially relevant offender and victim characteristics as well as distinguish between the different types of drugs used. To explore whether use of different types of drugs could explain the differential patterns observed for child molesters versus rapists, we obtained drug usage information in another sample of sexual offenders, divided into rapists and child molesters by the same criteria as in the present study. In this sample, we administered the Multidimensional Assessment of Sex and Aggression (Knight & Cerce, 1999; Knight, Prentky, & Cerce, 1994), a computerized self-report inventory that includes questions on specific drug usage but does not ask about use in sexual crimes. Rapists reported using nonprescription drugs significantly more than did child molesters and, in particular, reported using significantly more marijuana, downers, LSD, and crack/cocaine. Such data suggest, even though they are not sexual crime specific, that it might be difficult to pinpoint a particular drug that could account for the differential associations of perpetrator drug use with the sexual aggressiveness of child molesters versus that of rapists. Those data also suggest that the association of perpetrator drug use with child molesters’ aggression is not easily attributable to a physiological effect.
Alternatively, it may be that perpetrator drug use simply differentiates between child molesters with severe externalizing behavior and those with moderate or low externalizing conduct. In this case, the substantial positive relationship between child molesters’ drug use and aggression in sexual crimes may indicate that those child molesters who used any drugs before their sexual crime constitute an aggressive type and, therefore, inflicted more damage on their victims than did child molesters who did not use any drugs. Perpetrator drug use might not serve the same discriminating function for rapists. Consistent with this idea is the finding that child molesters who used any drugs before their sexual crimes had a significantly higher mean score on a general Physical Aggressiveness Scale ranging from 0 to 2 (M = 1.25) than did child molesters who did not use drugs (M = 0.67). In contrast, the mean score of rapists who used any drugs before their crimes (M = 1.46) was only slightly higher than that of rapists who did not use them (M = 1.19). Furthermore, the percentage of child molesters who used any drugs before their sexual crimes (8%) was much smaller than that of rapists (37%), indicating that they constitute a more select group. Future research should explore whether the association between drug use and aggression in the sexual crimes of child molesters is attributable to the externalizing tendencies of the perpetrator.
Our finding that victim substance use was associated with increased expressive aggression among rapists—even in a sample of rapists who were all under the influence of alcohol—also deserves attention. This result is contrary to our expectation based on the existing literature, although, as pointed out earlier, the literature is conflicting with researchers finding nonsignificant, significant negative, and significant positive associations between victim substance use and rapist aggression. Our finding may suggest either that adult victims under the influence of any substance were less concerned about injury or more aggressively reactive than were their sober counterparts, increasing their vulnerability to aggression (Macy et al., 2006), or that rapists targeted drinking or drugged victims for more aggressive acts (Ullman et al., 1999a). Further research is needed to clarify the mechanism of the relationship between victim substance use and rapists’ aggression. Such research should also explore sample effects on findings because previous studies with women have typically found a negative association between victim substance use and perpetrator aggression, whereas studies with men have found either a positive association or no association.
The levels of aggression observed in this study were relatively low. This result is consistent with what some other researchers have found. For example, Testa et al. (2004) had to dichotomize a previously ordinal measure of victim injury due to the small number of victims with more serious injuries in their sample. Notably, their sample consisted of female community members. We do not know how our observed levels of aggression would compare with results from other male criminal samples.
This study had a number of limitations. First, all of our substance use variables were dichotomous, meaning that we were unable to examine curvilinear relationships. Typically, a linear relationship has been found between perpetrator substance use and perpetrator aggression in sexual assaults (e.g., Abbey et al., 2003; Testa et al., 2004), but findings on the relationship between victim substance use and perpetrator aggression have been inconsistent. Our inability to discriminate between levels of use may explain some of our more surprising results. For example, the observed positive relationship of victim substance use to rapists’ expressive aggression might exist only at low levels of use. Curvilinear relationships need to be explored with more continuous measures. Second, we considered separately the associations with expressive aggression of perpetrator alcohol use, perpetrator drug use, and victim use. Our sample was not sufficiently large to permit examination of each independent variable’s relationship to the outcome variable, while controlling for the other two variables. It is important that studies with larger samples investigate the effects of those perpetrator and victim variables jointly. Third, unexamined variables might explain the mechanism underlying the associations of perpetrator and victim substance use with perpetrator aggression. For example, it is possible that the positive association between child molesters’ drug use and aggression in sexual crimes is attributable to their antisocial tendencies, which motivate both their use of drugs and their aggression. Future studies should include such potential explanatory variables. Fourth, substance use has been shown to affect not only perpetrator aggression but also the likelihood of penetration (Abbey et al., 2003; Testa et al., 2004) and general inhibitory control (Mulvihill, Skilling, & Vogel-Sprott, 1997). The relationships of substance use to such outcomes and motivations in child molesters’ versus rapists’ sexual offenses should be explored. And fifth, our analyses focused on the most recent sexual crime for which data were available, which was typically also the index crime for commitment and the crime described in the greatest detail. Many offenders did have prior sexual crimes, and it is possible that the most recent crime was not representative of each perpetrator’s typical substance use or aggression in sexual crimes or of his victims’ typical substance use in crimes. Other analyses on these crime data and on a newer sample to assess additional domains (e.g., aggression present, sexual aggression, and sexual behavior) have, however, yielded consistent results whether we use the maximum or mean presence over all crimes or only the index crime (Knight, Robertson, & Neumann, 2011).
Despite those limitations, our findings have important implications. The distinct substance use association patterns for child molesters versus rapists imply that substance use may play different roles depending on offender type. Our results showed that perpetrator alcohol use was associated with increased aggression in the sexual crimes of both rapists and child molesters and that perpetrator drug use was associated with increased aggression in the sexual crimes of child molesters. Those findings indicate that treatment targeting offenders’ sexually aggressive behavior should pay attention to their use of substances and address the role of such use in sexual crimes. There are many possible mechanisms, discussed above, which may underlie the associations between substance use and aggression in sexual crimes, and the different mechanisms might have different implications for treatment. The extant literature indicates that alcohol’s effects on sexual aggression are likely to be mediated by both alcohol expectancies and cognitive myopia (Abbey et al., 2004; Crowe & George, 1989). A causal role of alcohol expectancies implies that treatment should focus on modifying offenders’ alcohol-relevant beliefs, whereas effects of alcohol-induced cognitive myopia suggest that treatment should limit offenders’ alcohol intake and possibly train them to better attend to inhibitory cues when using alcohol. Additional research is necessary to clarify the precise roles of alcohol and drug use in rapists’ and child molesters’ sexual crimes. A better understanding of those roles can contribute to more effective intervention.
Footnotes
Acknowledgements
The authors would like to thank Malcolm Watson for helpful comments on a prior draft.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institute of Mental Health [R01-MH32309-07, R01-MH54263] and the National Institute of Justice [94-IJ-CX-0031, 94-IJ-CX-0049].