Rape Prevention With College Men

Abstract

This study evaluates the effectiveness of a theoretically based rape prevention intervention with college men who were at high or low risk to perpetrate sexually coercive behavior. Participants (N = 146) are randomly assigned to the intervention or control group. Outcomes include rape myth acceptance, victim empathy, attraction to sexual aggression, sex-related alcohol expectancies, and behavioral indicators, measured across three time points. Positive effects are found for rape myth acceptance, victim empathy, attraction to sexual aggression, and behavioral intentions to rape. Only rape myth acceptance and victim empathy effects sustain at the 5-week follow-up. High-risk men are generally unaffected by the intervention although low-risk men produced larger effects than the entire sample. Results suggest rape prevention studies must assess risk status moderation effects to maximize prevention for high-risk men. More research is needed to develop effective rape prevention with men who are at high risk to rape.

Introduction

Rates of rape on college campuses remain high despite prevention efforts. Most studies indicate that 25% to 60% of college men have engaged in some form of sexually coercive behavior with 6% to 14.9% of men admitting committing acts that met the legal definitions for rape and attempted rape (Berkowitz, 1992; Lisak & Miller, 2002). Primary prevention is needed for men in college who are at risk to rape women. Not all men rape; however, most rape prevention studies that evaluate intervention effectiveness with men do not differentiate between men who are at high versus low risk to rape. Therefore, meta-analyses that report (small) positive outcomes with college men may overestimate rape prevention effectiveness with high-risk men (Anderson & Whiston, 2005)—the men that most need to be deterred.

Self-reports of past sexually aggressive behavior may be useful for dichotomizing men into high- and low-risk categories for future perpetration. Past perpetration predicts the self-reported likelihood to rape in the future if assured of not being caught (Malamuth, 1981) as well as self-reported sexual assault at one year—9.4% of negative reporters versus 32.7% of positive reporters admitted to sexual assault at one year (Abbey & McAuslan, 2004). Unfortunately, most studies that have empirically defined and evaluated high-risk men have found few, if any, positive effects (Gidycz et al., 2001; Stephens & George, 2004). The present study evaluates the effectiveness of a brief, video-based rape prevention intervention on high-risk and low-risk White college men.

Attitudinal Targets of Rape Prevention

High-risk college men are likely to have developed stubborn attitudes commensurate with experiencing women as legitimate targets of rape. Researchers have defined and operationalized misogynous attitudes—a constellation of views that devalue women—in attempts to document and delineate the etiology of rape. Malamuth, Linz, Heavey, Barnes, and Acker’s (1995) confluence model, comprising integrated pathways, includes misogynous attitudes as prospective predictors of college men’s sexually aggressive behavior. In addition, past perpetration correlates with rape-supportive attitudes, a tendency to justify rape, higher empathy toward rapists than victims, and self-reported likelihood of committing future rape if assured of not being caught (Deitz, Blackwell, Daley, & Bentley, 1982; Malamuth et al., 1995; Muehlenhard, 1988). Etiologically based findings suggest that targeting rape-supportive attitudes with prevention efforts may reduce rape behavior in men.

Rape myth acceptance and empathy

In particular, rape myth acceptance and victim empathy have become prominent targets for rape prevention with men. Researchers have found that rape myth acceptance correlates strongly with sexual conservatism, adversarial sexual beliefs, self-reported likelihood of committing rape, and self-reported sexually aggressive behavior among men (Burt, 1980; Hamilton & Yee, 1990; Koss, Leonard, Beezley, & Oros, 1985; Malamuth, 1989a). It has been theorized that rapists use rape myth attitudes after the act to deny the impact of their crime and to excuse their behavior (Scully & Morolla, 1984). Extensive research on rape myths has honed this construct and articulated its role as a central attitudinal construct among sexually aggressive men.

Lack of empathy for rape victims has also been associated with likelihood to rape and moderated the confluence model’s pathways of sexual aggression (Deitz et al., 1982; Wheeler, George, & Dahl, 2002). Reviews of the rape prevention literature found little to no support for significant increases in victim empathy (Anderson & Whiston, 2005; Breitenbecher, 2000). However, research suggests that including an example of a male victim being raped by a heterosexual male perpetrator may increase victim empathy in men (Foubert & Newberry, 2006; Schewe, 2002).

Alcohol and rape

Most sexual assaults on campuses involve the use of alcohol, and the rape etiology literature supports this connection, particularly on college campuses. However, very few rape prevention studies assess alcohol-related variables as outcome measures (Abbey, Zawacki, Buck, Clinton, & McAuslan, 2004). The alcohol literature has delineated physiological effects (due to alcohol myopia) from psychological effects (due to alcohol expectancies) and substantial research has found connections between alcohol and sexual assault (George & Stoner, 2000). Given the demonstrated malleability of alcohol expectancies, they represent a useful target for change in rape prevention programs and a viable instrument for evaluating intervention success (Darkes & Goldman, 1998; Stephens & George, 2004).

Rape Prevention Findings

Several trends in the rape prevention literature are beginning to define methods to maximize effectiveness in reducing rape-supportive attitudes and perpetration by men. All male formats as opposed to mixed gender formats may limit confrontational and accusatory tones, which may decrease defensiveness (Breitenbecher, 2000, Schewe & O’Donohue, 1996). Lonsway (1996) suggested this may also allow for better attitudinal change through increased central route processing according to the elaboration likelihood model (ELM), that is, through a more careful and effortful evaluation of the messages due to an increase in perceived self-relevance (Petty & Cacioppo, 1981, 1986). Some researchers have begun evaluating program impacts on men at high risk to be sexually coercive (O’Donohue, Yeater, & Fanetti, 2003; Schewe & O’Donohue, 1996; Stephens & George, 2004). Schewe (2002) advocated using cognitive dissonance to improve rape prevention effectiveness, particularly with men whose attitudes were contrary to those being promoted. The bystander approach, recruiting men as helpers for victims of sexual assault rather than accusing them of being rapists, has also produced promising results (Foubert & Newberry, 2006). Finally, many rape prevention studies have suffered from important methodological weaknesses. For instance, some studies utilized immediate pre- and posttesting formats, which introduce sensitization issues and demand characteristics driven by socially desirable responding, the absence of proper control groups, and sub par statistical methods (Anderson & Whiston, 2005; Breitenbecher, 2000).

The Present Study

This study incorporated guiding principles from the rape-related etiology and attitude change literatures to evaluate the effectiveness of a rape prevention intervention with high-risk and low-risk college men. Rape myth acceptance, victim empathy, sex-related alcohol expectancies, and behavioral indicators were used as outcome variables. A 2 (high risk vs. low risk) × 2 (intervention vs. control) between-subjects design was used in conjunction with pretest, posttest, and 5-week follow-up measurements of outcome variables. In addition, central route processing was evaluated to determine its relationship with positive outcomes.

Hypotheses

• Hypothesis 1

  The intervention group will produce significant prosocial changes in outcome measures compared to the control group at posttest and the 5-week follow-up.

• Hypothesis 2

  Risk status will moderate intervention effectiveness; specifically, low-risk men will have greater improvement across outcome variables compared to high-risk men, given high-risk men typically possess greater rape-supportive attitudes and may be less receptive to antirape messages.

• Hypothesis 3

  Central route processing (i.e., the careful and effortful evaluation of the information presented in the intervention) will be positively correlated with bigger changes between pretest and posttest for outcome measures within the intervention group.

Method

Participants

Participants were 146 male, White undergraduates recruited from the psychology undergraduate subject pool (96.6%) or through campus advertisements (3.4%) at a large northwestern university. The majority of participants were freshman (61%), 19.2% were sophomores, 9.6% were juniors, and 10.2% were seniors or above; their average age was 19.3 years (SD = 1.8, range 18–29). Participants primarily identified as heterosexual (93.2%), first experienced consensual sexual intercourse at the average age of 17.0 (SD = 1.5) and averaged 4.1 (SD = 5.2) total opposite sex mutual consenting sexual intercourse partners. About one in five participants (22.6%) reported no consensual sexual partners over the lifetime. Of those who reported experiencing intercourse, 47.3% reported being drunk and 21.9% reported being high on drugs before or during sexual intercourse. The majority of participants reported some drinking behavior (69.2%) and averaged 15.5 (SD = 12.9) alcoholic drinks per week. The Modified Sexual Experiences Survey (M-SES; Dahl, 1993; Koss & Oros, 1982) was used to classify participants as high risk versus low risk for sexually offending. High-risk (45.2%) men reported at least one past sexually coercive behavior on the M-SES at pretest (see Table 1). Few participants (15.8%) reported having had a seminar, class, or presentation within the past year regarding sexual assault.

Table 1.

Percentages and n of Highest Level of Sexually Coercive Behavior Self-Reported on the M-SES

Sexually Coercive Behavior	High-Risk Men				Low-Risk Men 5-Week Follow-Up
	Pretest		5-Week Follow-Up
	Control n = 38	Intervention n = 28	Control n = 20	Intervention n = 11	Control n = 18	Intervention n = 16
None	0% (0)	0% (0)	55.0% (11)	18.2% (2)	94.4% (17)	87.5% (14)
Forced touching	42.1% (16)	39.3% (11)	10.0% (2)	36.4% (4)	5.6% (1)	12.5% (2)
Verbally coercive sexual behavior	39.5% (15)	35.7% (10)	15.0% (3)	36.4% (4)	0.0% (0)	0.0% (0)
Attempted and/or completed rape	18.4% (7)	25.0% (7)	20.0% (4)	9.1% (1)	0.0% (0)	0.0% (0)

Note: M-SES = Modified Sexual Experiences Survey.

Measures

Background Questionnaires

Various demographic questions, previous exposure to sexual assault prevention, drinking habits, sexual experience, and past accounts of sexually coercive behavior were included in a background questionnaire. A modified version (M-SES; Dahl, 1993) of the Sexual Experiences Survey (Koss & Oros, 1982) was used to measure self-reported sexually coercive behavior over the lifetime. Responses across the 18 items were on a 5-point scale (0 = never, 1 = 1 to 2 times, 2 = 3 to 5 times, 3 = 6 to 9 times, 4 = 10 or more times) and included a spectrum of behaviors ranging from unwanted sexual touching to rape. Wheeler et al. (2002) reported acceptable reliability in a previous sample (α = .69; current sample, α = .75). The Paulhus (1991) 40-item Balanced Inventory of Desirable Responding (BIDR; α = .83, Paulhus (1984); current sample α = .77) measured self-deceptive positivity and impression and was used to control for confounding tendencies to endorse demand characteristics throughout the analyses.

Dependent Measures

Attitudinal measures

The 19-item Rape Myth Scale (RMS) measured attitudes and generally false beliefs about rape that function to deny and justify male sexual aggression (Lonsway & Fitzgerald, 1995; α = .89). The 20-item short form version of the Illinois Rape Myth Acceptance Scale (IRMA-SF; α = .87, Payne, Lonsway, & Fitzgerald, 1999) based on the 45-item IRMA assessed general rape myth acceptance. Both the RMS (α = .91) and IRMA-SF (α = .84) were rated on a 7-point scale with higher scores indicating more rape myth endorsement. To study specific empathy toward victims, Deitz et al. (1982) developed the Rape Empathy Scale (RES; α = .84). This 19-item scale measured empathy toward the rape victim and the rapist in a rape situation. Participants were given two statements per item and asked to choose the one they preferred. Lower scores for the current sample on RES (α = .58) reflect more empathy toward the victim. The Sex-Related Alcohol Expectancies Scale (SRAES; Dermen & Cooper, 1994) consisted of three subscales: enhancement, sex risk, and disinhibition (α = .83, .70, and .79, respectively). The 13-item measure, rated on a 10-point scale, assessed participants’ expectancies about how much a moderate amount of alcohol affects sexuality (α = .92).

Behavioral indicators

Three measures assessed sexual aggressiveness. The M-SES was readministered only at the 5-week follow-up to assess self-reported sexually coercive behavior that occurred after posttest. Behavioral Intentions included four questions relating to likelihood to use force (LF), rape (LR), use coercion (LC; talk a woman into something sexual she doesn’t want to do), and use alcohol (LAlc; use alcohol in the hopes of having a woman give in to something sexual she wouldn’t give in to if not drinking). Higher scores indicated more likelihood to be sexually aggressive, using a 7-point scale (α = .91). Three items of Malamuth’s (1989a, 1989b) Attraction to Sexual Aggression scale (ASA) were assessed using a 4-point scale, with higher scores indicating more attraction (α = .72). This construct is assumed to be continuous such that all men, despite their actual experience with being sexual aggressive, can be placed on a continuum of believing being sexually aggressive with a women is arousing to both men and women.

Attitude change: Elaboration likelihood model

The 12-item Elaboration Likelihood Model Questionnaire (ELMQ; Heppner, Humphrey, Hillenbrand-Gunn, & DeBord, 1995), measured on a 5-point scale, assessed two factors of central route processing: cognitive involvement (α = .81; current sample, α = .76) and presentation quality (α = .61; current sample, α = .59, total scale α = .51). Higher scores represent more central route processing.

Procedure

Participants were asked to partake in a study titled Helping Victims of Sexual Assault, administered by a male experimenter. Participants recruited through campus advertisements were paid US$10 for each session completed, whereas participants from the subject pool were given extra credit in an undergraduate introductory psychology course for the pretest session and offered either additional extra credit or US$10 for each session thereafter.

The pretest session included informed consent and the administration of all measures except the ELMQ. Participants were then randomized into the control or intervention condition. Posttest groups of 1 to 13 across 61 sessions in classrooms on campus occurred on average 11.5 (SD = 8.2) days after pretest. Participants viewed a 50-minute video presentation and completed a questionnaire that included the ELMQ, a questionnaire regarding video content, and the dependent measures (excluding the M-SES). Cognitive dissonance (Festinger, 1957, 1964) was incorporated into the intervention to promote attitude change directly after the video presentation; therefore, participants completed a question related to convincing another person to not be sexually aggressive (i.e., John is a man who feels that he can force sexual activity on women whenever he wants to. Write a list of arguments that you would use to convince John not to rape women.). Recall was included to increase central route processing of the bystander message by having each participant list three ways he could help a survivor of sexual assault. Control condition participants were asked to list what they liked and did not like about the video presentation and three things they learned. The 5-week follow-up session occurred on average 49.2 days (mode = 33 days, median = 41 days, SD = 32.2) after posttest sessions and were conducted in person (89.2%) or via mail (10.8%). Participants were readministered the dependent measures with the addition of the M-SES and were debriefed.

Experimental groups

The 50-minute intervention video content consisted of four parts: (a) A preamble by a male speaker introducing the topic of sexual assault on campus and the need for college men to be educated about ways to prevent rape and help women they know, who may have experienced rape; (b) followed by Foubert’s (2000b) videotaped NOMORE Men’s Program, “How to Help a Sexual Assault Survivor: What Men Can Do,” which has been shown to empirically reduce rape myth acceptance (Foubert, 2000a) and includes a description of a male police officer’s rape by two heterosexual men; (c) another brief introduction by the same preamble speaker for a video segment on alcohol and rape; and (d) Dr. Jackson Katz’s professional interview (courtesy of the Media Education Foundation) regarding the negative intersection of alcohol and rape on campus. The preamble was intended to focus participants’ attention, instill personal relevance of the content, and allow for repetition of main points in the following segments to increase potential attitude change (Petty & Cacioppo, 1981, 1986).

The 50-minute control condition video comprised two Discovery programs titled Around the Galaxy and Cosmic Odyssey, both relating to the cosmos and were chosen due to their absence of sexual or rape content, stimulating intellectual nature, and lack of emotional charge.

Results

Attrition

Out of the 146 participants who completed the pretest session, 83 (29.5% attrition from pretest) completed the intervention or posttest session, and 65 (36.9% attrition from posttest) completed the follow-up session. Those who dropped at posttest were significantly lower in attraction to sexual aggression, behavioral intention to rape, and higher in victim empathy than those who completed posttest (p ≤ .047; .38 ≤ d ≤ .42); therefore, participants who dropped were modestly to moderately less rape supportive than those that completed posttest. No pretest differences between participants who completed versus dropped at follow-up were found. We also examined differences at pretest between conditions for those who did not drop from the study and as expected, we found no differences. For those who did drop from the study at posttest, we found more sex-related alcohol expectancies and behavioral intentions to rape in the control group compared to the intervention group (p ≤ .020) and at follow-up, we found more rape myth acceptance and sex-related alcohol expectancies for the control group than for the intervention group (p ≤.036). Therefore, participants who dropped from the study in the control group were more rape supportive than those that dropped from the intervention group.

Hypothesis Testing

Participants had to answer 80% or more of the questions for each measure to be included in analyses. Missing data ranged from 0% to 4.8% (M = 2.3%) and significant correlations were found among the BIDR and all dependent measures (r ranged from .20 to .83). Univariate analyses of covariance between groups and risk status across dependent measures followed by planned contrasts served as the main methods of analyses to examine predicted mean differences based on specific hypotheses. Both BIDR and outcome pretest scores were covaried throughout the analyses. In addition, participants’ drinking habits (average number of drinks per week) were covaried for analyses including alcohol expectancy measures. Effect sizes and confidence intervals were computed using Cohen’s d statistic, using raw M and SD (Cohen, 1988). Preliminary analyses revealed skew and kurtosis among some dependent variables. Behavioral intentions (square root) was the only measure to benefit from transformations, and the transformed scores were used throughout the analyses.

• Hypothesis 1

  Intervention effectiveness. Effect sizes (.20 ≤ d ≤ .45) and confidence intervals indicated that positive outcomes were modest to moderate in magnitude at posttest (see Table 2). Contrary to the hypothesis however, the M-SES trended in the opposite direction predicted (p = .053; see Table 1 for breakdowns). This nonsignificant result reflected an increase in reported sexually coercive behavior after the intervention compared to the control group. Reductions in rape myth acceptance and an increase in victim empathy were sustained with similar effect sizes at the 5-week follow-up (p ≤ .027). A significant reduction in rape myth acceptance, as measured by the IRMA-SF, for the intervention compared to the control group was found only at follow-up, indicating a sleeper effect in further support of sustained reductions in rape myth acceptance. Attitudinal behavioral indicators (i.e., self-reported attraction to sexual aggression and behavioral intentions) of rape rebounded.

• Hypothesis 2

  Risk status. Consistent with etiology literature, one-way ANOVAs confirmed that high-risk men were more rape supportive across all dependent measures compared to low-risk men at pretest, F ≥ 5.80, p ≤ .017. No significant differences between high-risk intervention versus control group men were found at posttest (p ≥ .210). At follow-up, high-risk men reported a modest reduction in rape myth acceptance; however the effect size 95% confidence interval indicated a lack of stability, IRMA, p = .037, d = .36 (−.01 to .98). A higher percentage of high-risk men after the intervention (82%, 9/11) reported sexually coercive behavior at follow-up compared to high-risk control men, 47%, 9/19; p = .030, d = −.74 (−.97 to −.50). This difference was in the opposite direction of the predicted hypothesis.

  In contrast to high-risk men, low-risk men reduced rape-supportive attitudes across all outcome categories (p ≤.044, see Table 3). Reductions in rape myth acceptance and an increase in victim empathy sustained at the 5-week follow-up. The confidence interval for the SRAES follow-up effect indicated caution in interpreting this sleeper effect as truly large. Behavioral intentions to rape rebounded at follow-up.

• Hypothesis 3

  Elaboration likelihood model. Partial correlations were computed controlling for BIDR between the ELMQ and pretest–posttest changes in dependent measures. In the predicted direction, the ELMQ was positively correlated with larger changes in RMS (r = .37) and RES (r = .23), p < .05.

Table 2.

Univariate ANCOVAs of Condition (Control vs. Intervention), Covarying Social Desirability, and Pretest Scores Across All Men

All Men Dependent Measures	Posttest				Follow-Up
	Control n = 52	Intervention n = 51	p Value	Effect Size d (95% CI)	Control n = 38	Intervention n = 27	p Value	Effect Size d (95% CI)
Rape myths
RMS	1.39 (1.01)	0.97 (0.85)	.030*	.45 (.17 to .69)	1.36 (0.96)	0.98 (0.91)	.001**	.40 (.09 to .76)
IRMA-SF	1.68 (0.79)	1.35 (0.70)	.381	.44 (.22 to .64)	1.71 (0.77)	1.25 (0.77)	.006**	.60 (.35 to .89)
Empathy
RES	0.81 (0.16)	0.87 (0.14)	.010*	−.40 (−.44 to −.36)	0.83 (0.11)	0.87 (0.13)	.027*	−.34 (−.37 to −.29)
Alcohol expectanciesa
SRAES	3.25 (1.78)	2.98 (2.04)	.196	.14 (−.37 to.73)	3.08 (2.00)	2.79 (2.03)	.339	.14 (−.51 to .92)
Behavioral indicators
ASA	0.51 (0.66)	0.32 (0.50)	.035*	.33 (.14 to .47)	0.51 (0.71)	0.46 (0.71)	.759	.07 (−.16 to.34)
Behavioral intentions	0.97 (1.24)	0.74 (1.10)	.013*	.20 (−.14 to .51)	0.88 (1.30)	0.83 (1.25)	.514	.04 (−.37 to.51)
M-SES	—	—	—	—	0.27 (0.45)	0.41 (0.50)	.053b	−.30 (−.44 to −.11)

Note: RMS = Rape Myth Scale; IRMA-SF = Short form version of the Illinois Rape Myth Acceptance Scale; RES = Rape Empathy Scale; SRAES = Sex-Related Alcohol Expectancies Scale; ASA = Attraction to Sexual Aggression; M-SES = Modified Sexual Experiences Survey.

^aAdditionally covaried average number of drinks/week.

^bM-SES effect is in opposite direction of hypotheses.

^*p < .05.

^**p < .01.

Table 3.

Planned Contrasts for Low-Risk Men Comparing Control Versus Intervention, Covarying Social Desirability, and Pretest Scores

Low-Risk Men Dependent Measures	Posttest				Follow-Up
	Control n = 27	Intervention n = 27	p Value	Effect Size d (95% CI)	Control n = 18	Intervention n = 16	Effect Size p Value	d (95% CI)
Rape myths
RMS	1.24 (0.83)	0.78 (0.74)	.007**	.59 (.27 to .88)	1.28 (0.94)	0.72 (0.64)	.004**	.70 (.26 to 1.02)
IRMA-SF	1.58 (0.73)	1.26 (0.68)	.203	.45 (.17 to .72)	1.64 (0.73)	1.11 (0.51)	.014*	.85 (.51 to1.10)
Empathy
RES	0.82 (0.12)	0.90 (0.09)	.017*	−.76 (−.80 to −.72)	0.84 (0.12)	0.89 (0.09)	.032*	−.41 (−.53 to −.43)
Alcohol expectanciesa
SRAES	2.69 (1.71)	2.62 (1.99)	.568	.03 (−1.02 to .81)	2.94 (1.67)	1.51 (1.48)	.044*	.91 (.09 to 1.66)
Behavioral indicators
ASA	0.47 (0.66)	0.14 (0.29)	.114	.69 (.44 to .81)	0.63 (0.70)	0.31 (0.60)	.135	.49 (.17 to .78)
Behavioral intentions	0.78 (1.34)	0.32 (0.63)	.011*	.46 (−.04 to .71)	0.75 (1.38)	0.38 (0.70)	.098	.35 (−.29 to .69)
M-SES	—	—	—	—	0.06 (0.24)	0.13 (0.34)	.627b	−.24 (−.35 to −.08)

See note to Table 2.

^aAdditionally covaried average number of drinks/week.

^bM-SES effect is in opposite direction of hypotheses.

^*p < .05.

^**p <.01.

Discussion

Effect sizes for rape myth acceptance and victim empathy reflected educational and possibly clinically relevant change in comparison to a control group, offering evidence that the intervention produced positive sustained outcomes. These effects were larger than those reported in the most recent meta-analytic study of the rape prevention literature and correlated with more central route processing. This may be due to the incorporation of several factors found to be associated with better effects, namely, theory-based design and content, an all male format, the use of professional and peer presenters, a focused bystander approach, and rigorous scientific methodology that reduced pre- or posttest sensitivity and socially desirable responding biases. Treated men also modestly reduced attraction to sexual aggression and behavioral intentions to rape at posttest, but both of these effects rebounded at follow-up. The intervention group did not reduce the amount of self-reported sexually coercive behavior however, which brings into question whether the positive attitudinal changes would actually lead to behavior change.

Subgroup analyses revealed that positive outcomes were moderated by risk status, consistent with study predictions. Low-risk men—men least at risk to be sexually coercive—reduced rape-supportive outcomes across all dependent measures, evidencing more and greater effects than high-risk men. These positive findings are not surprising for low-risk men, given that they harbor attitudes more congruent with attitudes presented in the intervention. These findings may indicate that low-risk men’s attitude changes were responsible for producing significant overall intervention effects, bringing into question effectiveness with high-risk men.

Consistent with most studies, high-risk men exhibited minimal effects from the intervention. The only single significant effect found in evidencing less rape-supportive tendencies was a reduction in rape myth acceptance at follow-up. However, this small-to-medium effect may have questionable stability due to its large confidence interval. High-risk men reported significantly higher rates of sexually coercive behavior after the intervention compared to the control group (82% vs. 47%). This difference was unanticipated. Several explanations may account for this difference. Increased reports of sexually aggressive behavior may be due to several possible effects of the intervention including cuing, education that led to more accurate reporting, increased willingness to report behavior, or possibly an iatrogenic effect. The possibility of an iatrogenic effect is most concerning. Furthermore, this possibility was bolstered by mean differences at follow-up in attraction to sexual aggression and behavioral intention to rape, which both trended in the direction of a backlash. Psychopathy-related literature supports the potential for backlash effects with treatment, given Malamuth’s (2003) recent synergistic comparison between his confluence model and Hare’s (1991) psychopathy model. Future research should assess reactance to rape prevention interventions, include longer follow-up evaluations, and consider using data safety monitoring plans (DSMP) consistent with randomized control trial (RCT) methods to ensure interventions are discontinued or altered if iatrogenic effects are found. Given the high attrition rate at the 5-week follow-up and the lack of a reactance measure, it is difficult to interpret this finding. High-risk men may consist of heterogeneous groups that require multiple approaches to rape prevention based on individual differences. Rape prevention with high-risk men may need to be multifaceted, with longer exposure to antirape messaging, and should be derived from etiological research.

Consistent with an earlier alcohol expectancy finding (Stephens & George, 2004), we again found that low-risk men in the intervention condition exhibited lower sex-related alcohol expectancies than low-risk men in the control condition at follow-up. The lack of an immediate posttreatment effect may have been because the present video presentation did not specifically address the dangers of specific alcohol expectancies related to rape but rather general dangers associated with drinking and rape. Future research is needed to evaluate the speculation that variable alcohol-related content within rape interventions may yield different effects on sex-related alcohol expectancies.

Limitations

Findings may not generalize to all men, given this was a White sample of college men who volunteered to be in a study titled, How to Help a Victim of Sexual Assault. Men who volunteer to be in this type of study may be more amenable to the messages in the intervention. All outcome measures were self-report, which are vulnerable to underreporting and socially desirable responding. However, a substantial number of participants reported a wide range of sexually coercive behaviors including a blatant question about rape perpetration. In addition, a measure of socially desirable responding was controlled for in analyses increasing the possibility for reasonably valid interpretations of outcome findings. The interventions were brief video-based presentations, which may have been less effective than an intervention with multiple sessions; although brief, video-based interventions have been shown to be effective and practical as a method for rape prevention (Breitenbecher, 2000). It is unknown how long-lasting effects may have been due to the brief time frame of the 5-week follow-up. We used a no-treatment control condition; future research may benefit from attention placebo or other types of control conditions to allow for specific component testing of interventions.

Conclusion

A brief video-based rape prevention intervention reduced rape myth acceptance and increased victim empathy for up to 5 weeks; however, subgroup analyses suggest results may have been driven by findings with low-risk men, men least likely to be sexually coercive. High-risk men displayed little to no reduction in rape supportiveness and evidenced some backlash to the intervention. More research is needed to clarify these results and determine effective methods of rape prevention with high-risk men. Interventions may need to be multifaceted and customizable based on individual differences.

Biographies

Kari A. Stephens received her BS in psychology at the University of Washington, Seattle, WA in 1995, her MS in psychology at the University of Washington in 2001, and is a ninth year doctoral candidate in the adult track of the Clinical Psychology Doctoral Program at the University of Washington, currently completing her predoctoral internship with psychiatry at the University of Washington. She is studying the theory, etiology, and prevention of male sexual aggression, in particular, as these issues relate to alcohol consumption and cultural factors related to Asian American communities. Her doctoral thesis is titled, Rape Prevention With Asian/Pacific Islander and White College Men: The Roles of Culture and Risk Status. She is working with Professor William George who has studied sexuality and alcohol issues for more than 20 years, including significant work regarding racial stereotyping and how sexually aggressive behavior relates to alcohol. She intends to conduct clinical research and intervention work with disadvantaged populations after completing her PhD, particularly in relation to the areas of violence and addiction.

William H. George received his BA in psychology at Rockford College, Rockford, IL in 1975 and his PhD in clinical psychology at the University of Washington, Seattle, WA in 1982. He is a professor in the psychology department at the University of Washington. He has conducted research for more than two decades related to the field of alcohol and sexuality. His research has included studies in the areas of sexual aggression, racism, sex risk, and sexual inference. He has more than 50 publications, received more than 20 public and private research grants, chaired more than a dozen dissertation committees, and taught numerous classes related to psychology at both the graduate and undergraduate levels. He is a licensed psychologist in the state of Washington and a member of the American Psychological Association, International Academy for Sex Research, Society for the Scientific Study of Sex, and World Association of Sexology. He has also served on editorial boards and as a reviewer for several peer-reviewed psychology journals.