Implicit Attitudes Toward Violence Among Male Perpetrators of Intimate Partner Violence

Abstract

Whereas cognitive variables are hypothesized to play an important role in intimate partner violence (IPV) etiology and intervention, cognitive assessment methods have largely targeted offenders’ explicit, controlled cognitive processing using paper-and-pencil questionnaires prone to social desirability biases. Using an implicit measure of attitudes (the Implicit Association Test [IAT]), we assessed attitudes toward gender, violence, and the association between gender and violence among 50 men enrolled in an IPV treatment program and a comparison sample of 40 nonviolent (NV) men. Although no group differences were noted on explicit attitudinal measures, men in the IPV group showed more positive implicit attitudes regarding violence, and a more rapid association between women and violence. Among men in treatment for IPV, the attitudes toward violence IAT was significantly correlated with self/partner-reported IPV frequency. In accordance with social information processing models of aggression, these results suggest that aggressogenic attitudes are likely to operate automatically and with little conscious deliberation. As a result, clinicians and researchers must adapt assessment and intervention strategies to capture both implicit and explicit aspects of cognitive processing.

Keywords

intimate partner violence IPV implicit attitudes explicit attitudes IAT cognitive assessment

Etiologic models of intimate partner violence (IPV) suggest that cognitive variables are implicated in the onset and maintenance of abusive behavior. For example, feminist-informed models of IPV (e.g., Yllo & Straus, 1990) suggest that IPV results from gender socialization processes wherein males adopt an attitudinal style that promotes male dominance over women and supports the usage of tactics (including abusive behavior) to coerce and control women in close relationships. Similarly, cognitive-behavioral models maintain that cognitive distortions and biased attitudes enhance violence risk by increasing the likelihood of angry affect during relationship conflict (e.g., Beck, 1999), constricting the range of potential response options to resolve the conflict, and fostering positive evaluations of the effectiveness of aggressive behaviors (Holtzworth-Munroe, 1992). Data largely support the basic tenets of these models; relative to comparison samples, IPV perpetrators exhibit (a) more decoding, interpretation, and hostile attribution biases (Fincham, Bradbury, Arias, Byrne, & Karney, 1997), especially in the context of anger arousal (Eckhardt, Barbour, & Davison, 1998); (b) greater generation of aggressive response options during relationship conflict (Anglin & Holtzworth-Munroe, 1997); and (c) positive evaluations of violence in close relationships (Kaufman Kantor & Straus, 1990).

Evidence supports the importance of cognitive factors in IPV perpetration, with meta-analytic reviews suggesting a moderate association between these constructs (Stith, Smith, Penn, Ward, & Tritt, 2004). However, as noted in qualitative reviews of this literature (Eckhardt & Dye, 2000), with few exceptions, this body of literature is derived from assessment methods that require respondents to explicitly provide self-reports of their cognitive experiences using paper-and-pencil questionnaires. Several factors limit the utility of explicit self-report measures to aid in the understanding of the association between cognitive distortions and IPV. First, researchers and clinicians have noted the tendency of IPV perpetrators to minimize, deny, or defend their abusive behavior (e.g., Henning, Jones, & Holdford, 2005). Given this tendency, data based on assessment methods that explicitly probe offenders’ self-reported violence-related attitudes may be unduly influenced by respondents’ motivation to underreport their prior abuse history and to disguise their inclination to favor violent conflict tactics. Although findings regarding the effects of social desirability on IPV reporting are equivocal (Henning & Holdford, 2006; Scott & Straus, 2007; Sugarman & Hotaling, 1997), offender underreporting may undermine the ability of treatment providers to accurately assess changes in attitudes during intervention programs and attenuate the validity of results of outcome studies examining whether such programs lead to cognitive changes that foster nonviolence.

Second, questions remain concerning whether individuals can directly and accurately access violence-supportive cognitive processes. Researchers have suggested that such beliefs are likely to be automatic and schematic in nature (Berkowitz, 2008; Huesmann, 1988), with specific cognitive distortions hypothesized to emerge from an associative memory network of interconnected attitudes, memories, behavioral scripts, and emotions that center around an underlying dominant theme (e.g., the implicit theory approach; Polaschek, Calvert, & Gannon, 2009). Thus, although offenders may be aware of, and report on, surface-level distortions (e.g., “I hit her because she wouldn’t stop nagging me”), they may lack awareness of more tacit elements of the schema-level, implicit cognitive structure that may generate the belief (e.g., “she should know better than to bother me; punishment is deserved”). As a result, offenders will be unlikely to access such implicit information or report on such processes with any useful degree of accuracy when asked to do so on explicit, paper-and-pencil measures of cognitive content (Ward, 2000).

The exclusive use of explicit measures of cognitive constructs has limited the complete understanding of the role played by attitudinal factors in the etiology of IPV. Thus, IPV models will be more comprehensive and results of treatment effectiveness research more accurate if cognitive assessment methods capture both explicit as well as implicit cognitive processing. Assessment of cognitive constructs using implicit measures has advanced dramatically in the previous decade. The Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998) has emerged as the most widely used measure of implicit attitudinal strength and concept preference and has been used to assess wide range of attitudinal constructs among nonclinical and clinical samples (for a review, see Greenwald, Poehlman, Uhlmann, & Banaji, 2009). Generally, a recent quantitative review of the IAT’s construct validity indicated superior discriminant and predictive validity relative to explicit measures, especially for constructs influenced by social desirability biases (Greenwald et al., 2009), such as endorsement of violence-related attitudes. Using offender samples, researchers have used the IAT to demonstrate that psychopathic murderers have abnormal beliefs about violence relative to other offenders (Gray, MacCulloch, Smith, Morris, & Snowden, 2003) and to distinguish pedophilic from nonpedophilic offenders (Brown, Gray, & Snowden, 2009; Gray, Brown, MacCulloch, Smith, & Snowden, 2005). One published study has used the IAT to assess IPV-related attitudes (Robertson & Murachver, 2007), although not in samples defined by their histories of IPV. Using a diverse sample of 39 incarcerated male and female offenders (for a variety of offense types) from New Zealand, and a comparison sample of 133 community adults, Robertson and Murachver (2007) reported no group differences on an explicit measure of attitudes approving of violence; however, using an IAT measure of the same construct, the incarcerated group held beliefs that were more approving of violence, although the effect was small (η² = .028).

In the present study, we sought to extend these findings by examining a wider range of cognitive constructs in samples specifically selected for their IPV histories. Specifically, we examined differences in attitudes toward (a) women, (b) violence, and (c) the association between women and violence between men in treatment for IPV relative to a comparison sample of men without a history of IPV offending. Given that prior research has been equivocal regarding differences between violent and nonviolent samples on explicit measures of attitudes (Sugarman & Frankel, 1996), as well as more general concerns regarding the limitations of explicit measures of cognition (Polaschek et al., 2009), we did not hypothesize that the groups would differ on explicit measures. However, given prior research on violent offenders using implicit attitudinal measures, we expected to find group differences on IAT assessed attitudes. According to current models of IPV etiology, we expected that, relative to nonviolent men, IPV perpetrators would demonstrate less favorable attitudes toward women, more favorable attitudes toward violence, and a stronger association between women and violence. Finally, we assessed associations between explicit and implicit measures of related attitudinal constructs within each group; however, given the dearth of research in this area, we had no prior expectations about the presence and degree of intercorrelations among these measures.

Method

Participants

Participants were 50 male clients in an IPV intervention program and a comparison sample of 40 nonviolent men. Criteria for inclusion into the IPV group were being 19 years of age or older and current attendance at a Batterer Intervention and Prevention Program (BIPP). Partner violent men were recruited from several open-group BIPPs located in a Midwestern state, serving primarily court-mandated clientele. Nonviolent (NV) men were recruited from the same communities as the participating BIPP locations via flyers and online ads requesting men aged 19 years or older who were involved in a current romantic relationship. The key inclusionary criterion that determined eligibility for the NV group was no history of physical IPV toward a female partner according to both self-report and partner report during phone screenings conducted separately for each partner. Demographic characteristics can be found in Table 1.¹

Table 1.

Demographic Data for IPV and NV Groups

Variable	IPV group	NV group
Mean age (SD)	32.3 (9.1)	26.7 (7.9)
Race/ethnicity (%)
White	90.0	46.2
Hispanic	8.0	25.6
Black	2.0	17.9
Asian/Other	0.0	10.3
Mean years of education (SD)	12.8 (2.4)	16.9 (2.4)
Marital status (%)
Single	24.0	41.0
Married/living together	50.0	51.3
Divorced/remarried	26.0	7.7
Mean number of sessions attended (SD)	14 (10)	—
Reason for attending group (%)
Charges related to IPV	52.4	—
Violation protective order	11.9	—
Other	33.3	—
Both	2.4	—
Arrested on DV charges
% yes	70.8	—
Number of times arrested on DV charges (%)
Once	50.0	—
Twice	12.5	—
3 to 5 times	8.3	—

Note: IPV = intimate partner violence; NV = nonviolent; DV = domestic violence.

Procedures

Recruitment and screening

For the IPV group, graduate research assistants recruited clients attending BIPP sessions for a study examining “how men handle conflict in close relationships.” Individuals expressing interest were either consented into the study at the end of the BIPP session and scheduled for a later assessment appointment or were allowed to leave the BIPP session and complete the questionnaires and computer task in a separate room. Men in the NV group were verbally consented and completed all screening measures over the phone. To qualify, participants reported no acts of physical IPV over the history of their relationship (as verified by female partners). Men who met inclusion criteria were scheduled to come to a university lab to complete remaining measures. All participants were paid US$15.00.

Measures

Participants completed shortened versions of the revised Conflict Tactics Scale (CTS-2; Straus, Hamby, Boney-McCoy, & Sugarman, 1996) and Multidimensional Measure of Emotional Abuse (MMEA; Murphy & Hoover, 1999) to screen for the presence of physical, psychological, and sexual IPV. For this study, 16 items were selected from these scales to assess whether the male perpetrated the action against the female partner and whether the female partner directed an action toward the male. Responses were given on 4-point scales (0 = never; 3 = more than six times in past 6 months). IPV participants completed a paper-and-pencil version of this measure; NV participants and their female partners provided verbal reports during telephone screening. Estimates of internal consistency (coefficient alpha) were acceptable (α = .86 for IPV men; α =.75 for NV men; α =.74 for female partners of NV men). NV men also provided relationship satisfaction data via the 15-item Short Marital Adjustment Test (SMAT; Locke & Wallace, 1959; α = .74).

Participants who met all inclusion criteria completed two explicit measures of IPV-related attitudes. First, participants completed the Sex-Role Egalitarianism Scale–Form BB (SRES-BB; King & King, 1993), a 25-item short form that measures a variety of attitudes related to social–interpersonal–heterosexual relationships, with higher values corresponding to more egalitarian responses. Internal consistency estimates were acceptable (IPV: α = .89; NV: α = .93). Second, participants completed the 31-item Inventory of Beliefs About Wife Beating (IBWB; Saunders, Lynch, Grayson, & Linz, 1987), which assesses specific attitudes toward IPV and IPV victims. High scores indicate positive attitudes toward perpetration of IPV and negative attitudes toward IPV victims (IPV group α = .48; NV group α = .46). Third, participants completed the Acceptance of Interpersonal Violence Scale (AIV; Burt, 1980), which is a six-item measure assessing the degree to which men endorse the use of violence in close relationships. In the IPV group, α = .47 and in the NV group, α = .51; although low, these alphas are similar to those obtained by the scale developer (see Burt, 1980, p. 222). The total score of each measure was used in all analyses.

The IAT (Greenwald et al., 1998) was used to assess automatically activated attitudes, preferences, and evaluations. In the current study, participants completed three versions of the IAT (see Figure 1). The IATs were created using DirectRT (v.2006.2; Empirisoft, 2006) and were loaded onto three laptop computers running Windows XP.

Figure 1.

Block descriptions for three Implicit Association Tests (IATs)

Each IAT is composed of five blocks. The first block was the initial target concept discrimination task. In the “attitudes toward women” IAT, participants sorted male (e.g., Peter) and female (e.g., Michelle) names into the categories “male” and “female” using the “e” key (left) and the “i” key (right). For the “attitudes toward violence” IAT, a similar strategy was used for Trial 1, except aggression-related and nonviolence-related words were sorted into the category terms of “violent” and “peaceful.” In the “associations between women and violence” IAT, names were sorted into the categories of “male” and “female.” The second block involved an associated attribute concept discrimination task, wherein participants sorted words into “bad” (e.g., “agony” on left side) and “good” (e.g., “friendly” on right side) categories (for the gender violence IAT, participants sorted words into “violence” and “peaceful” categories). The third block was a congruent combination task in which words corresponding to both the target concept (e.g., male/female) and the attribute concept (i.e., good/bad) flashed on the computer screen. In the attitudes toward women IAT, if the word was a “male” word or a “bad” word, it was sorted to the left; if the word was a “female” word or “good” word, it was sorted to the right (for the attitudes toward violence IAT, left sorting was done for “peaceful” and “good” words and right sorting for “violence” and “bad” words; in the last IAT, left sorting was “male” and “violent,” right sorting was “female” and “peaceful”). In the fourth block, the reversed target concept discrimination task was presented; for each IAT, the side to which target concepts were sorted in the second block was reversed (e.g., “good” presented on left side). The fifth block was the incongruent combination task, in which “male” and “female” words were sorted in the same manner as the previous trial block but were paired with opposing attribute concepts relative to the third “congruent” block (e.g., left = male, good; right = female, bad).

Within all blocks, words were presented in random order. Rather than counterbalancing the order of congruent and incongruent blocks, participants completed extensive practice trials before each critical block (blocks three and five), which has been shown to reduce potential order effects (Nosek, Greenwald, & Banaji, 2005). The IAT score is defined as the difference between the mean response times (RT) for incongruent category pairs (e.g., female and bad) and congruent category pairs (e.g., female and good). An IAT effect will be assumed to occur if the IPV group is significantly faster at classifying target words paired with attributes more associated with IPV perpetration (e.g., female and bad; violence and good; female and violence) relative to when such targets are paired with outcomes less associated with IPV. IAT effects will be expressed using the D statistic, which examines differences in mean reaction times between the test block conditions (including practice trials and a penalty for incorrect trials) relative to the individual’s variance in responding to these blocks (Greenwald, Nosek, & Banaji, 2003).

Results

Explicit Measures

We examined differences between IPV and NV groups on explicit attitude measures using independent samples t tests (see Table 2). There were no significant differences between the IPV and NV groups on the SRES total scale score, AIV total score, or IBWB total score.

Table 2.

Means and Group Differences on Explicit and Implicit Measures

	IPV group		NV group
Measure	M	SD	M	SD	t
SRES total score	101.54	10.49	105.38	12.03	1.60
IBWB total score	100.12	8.96	97.47	9.50	−1.32
AIV total score	13.74	4.58	12.38	4.55	−1.39
IAT–attitudes toward women	−0.04	0.44	0.01	0.51	0.47
IAT–attitudes toward violence	1.15	0.33	1.28	0.17	2.07*
IAT–women and violence	0.10	0.40	0.30	0.37	2.38*

Note: IPV = intimate partner violence; NV = nonviolent; SRES = Sex-Role Egalitarianism Scale; IBWB = Inventory of Beliefs About Wife Beating; AIV = Attitudes Toward Interpersonal Violence Scale; IAT = Implicit Association Test. IAT scores are listed in D-units, defined as the mean difference between incongruent condition reaction times and congruent condition reaction times relative to each group’s variance.

Implicit Measures

D-score means and standard deviations for the three IAT measures are presented in Table 2. Group differences were evaluated using independent samples t tests.²

Attitudes toward women

We hypothesized that IPV offenders would demonstrate more negative attitudes toward women; the IAT effect (and thus a positive D-score) would involve faster response times to the female–bad categorization task relative to the female–good pairing. Contrary to expectation, there were no significant group differences on D-scores for this IAT.

Attitudes toward violence

We hypothesized that IPV offenders would demonstrate more positive attitudes toward violence; the IAT effect would involve faster response times to the violence–good categorization task relative to violence–bad association. Analyses revealed a significant difference in D-scores between the groups. As can be seen in Figure 2, while both groups’ responding was delayed by the task requirements of the incongruent condition (i.e., “violence–good”), the mean difference between incongruent and congruent conditions was smaller among men in treatment for IPV relative to those in the NV group. The effect size (Cohen’s d) for the difference between groups fell in the small to medium range (d = .45).

Figure 2.

D-scores for the intimate partner violence and nonviolent groups on the (a) attitudes toward violence Implicit Association Test and (b) gender violence Implicit Association Test

Associations between women and violence

We hypothesized that IPV offenders would associate women with violence more readily than NV men; the IAT effect would involve faster response times to the female–violence categorization task relative to female–peaceful associations. As expected, there was a significant, moderate IAT effect for the IPV group relative to the NV group (d = .51). As with the prior IAT, despite response delays during the incongruent condition (i.e., “women–violence”), the mean difference between incongruent and congruent conditions was smaller among the IPV group relative to those in the NV group.

Associations Between Explicit and Implicit Measures

Within each group, we examined correlations among explicit measures of cognition, the three IAT D-scores, demographics, and other individual differences (see Tables 3 and 4). In the IPV group, explicit measures were largely uncorrelated. While age was negatively correlated with the gender IAT, no other IAT was associated with age or education. The women and violence IAT was negatively correlated with the IBWB total score; no other significant implicit–explicit correlations emerged. The attitudes toward violence IAT was negatively correlated with perpetration of minor and total IPV. IAT D-scores were not significantly correlated with BIPP attendance.

Table 3.

Bivariate Correlations Among Explicit and Implicit Attitudinal Measures, Demographics, and Individual Difference Variables in the IPV Group

	AIV total	IBWB total	SRES total	Attitudes toward women IAT	Attitudes toward violence IAT	Gender and violence IAT	CTS total IPV	CTS minor physical IPV	CTS severe physical IPV	CTS psychological IPV	CTS sexual IPV	Age	Education	Number of BIPP sessions completed
AIV total	—	.27	−.21	−.02	−.11	−.11	−.12	−.03	−.05	−.11	.13	.08	.01	.02
IBWB total		—	−.06	−.33*	−.06	.02	.04	.02	.01	.04	.06	−.17	−.17	.06
SRES total			—	−.01	.12	.16	−.02	.07	.01	−.14	.07	−.04	.15	.25
Attitudes toward women IAT				—	−.16	−.44**	.11	.14	.03	.14	−.15	−.34*	−.09	.13
Attitudes toward violence IAT					—	.23	−.23	−.23	−.06	−.21	.05	.14	.21	−.11
Gender and violence IAT						—	−.42**	−.48**	−.04	−.24	.17	−.01	−.11	.10
CTS total IPV							—	.88**	.66**	.87**	.37*	.05	.05	−.08
CTS minor physical IPV								—	.52**	.58**	.33**	.06	.10	−.19
CTS severe physical IPV									—	.47**	.76**	.25	.00	.02
CTS psychological IPV										—	.20	.00	.01	−.04
CTS sexual IPV											—	.22	−.01	.19
Age												—	.38**	−.25
Education													—	−.06
Number of BIPP sessions completed														—

Note: IPV = intimate partner violence; NV = nonviolent; AIV = Acceptance of Interpersonal Violence Scale; IBWB = Inventory of Beliefs About Wife Beating; SRES = Sex-Role Egalitarianism Scale; IAT = Implicit Association Test; CTS = Conflict Tactics Scale; BIPP = Batterer Intervention and Prevention Program.

Table 4.

Bivariate Correlations Among Explicit and Implicit Attitudinal Measures, Demographics, and Individual Difference Variables in the NV Group

	AIV total	IBWB total	SRES total	Attitudes toward women IAT	Women and violence IAT	Attitudes toward violence IAT	Age	Education
AIV total	—	.19	−.30	.02	.04	.02	−.05	−.26
IBWB total		—	−.28	.01	.16	.03	.03	−.29
SRES total			—	−.08	−.02	−.22	.18	.24
Attitudes toward women IAT				—	−.10	−.25	−.35*	−.19
Women and violence IAT					—	.03	.02	.05
Attitudes toward violence IAT						—	.27	−.01
Age							—	.17
Education								—

Note: NV = nonviolent; AIV = Acceptance of Interpersonal Violence Scale; IBWB = Inventory of Beliefs About Wife Beating; SRES = Sex-Role Egalitarianism Scale; IAT = Implicit Association Test.

Within the NV group, the gender IAT was negatively correlated with age. No other significant correlations among other explicit or implicit measures emerged.

Discussion

In the present study, we examined men’s attitudes toward gender and IPV using both implicit and explicit measures in samples of individuals in treatment for IPV and a comparison sample of men without a history of IPV. Using three IPV-related IATs, men in treatment for IPV exhibited faster response times relative to NV men when violence-related words were paired with words having a positive evaluative valence and when violence-related words were paired with female names. No significant group differences were found on an IAT assessing automatic associations between gender and valence, in contrast to predictions made by feminist models of IPV (e.g., Bograd, 1988). The groups were not reliably differentiated on the basis of explicit measures of cognitive/attitudinal constructs, and few correlations were noted between implicit and explicit measures of cognition.

Men in the IPV group were faster than NV men to associate violence-related concepts with positive valence terms. Do these data indicate that men with a history of IPV maintain latent, implicit attitudes toward violence that signify risk for future IPV? Perhaps, but it is critical to note at the outset that the IAT is not, and has never been, regarded as a “diagnostic” test capable of reliably detecting violent attitudes even when explicit measures fail to reveal this information (Nosek et al., 2005). The data merely indicate that men in the IPV treatment group were slightly more efficient at associating violent words with positive labels than the NV group; how this difference translates into actual conflict behavior in close relationships awaits further investigation.

Thus, in accordance with theory and research in this area (Archer & Graham-Kevan, 2003; O’Leary, 1988), some violent individuals may positively endorse the use of aggressive conflict resolution strategies in close relationships (“might makes right!”) and will thus be more likely to use such strategies to manage their own relationship conflicts. However, these results also suggest that other individuals may actually hold generally negative attitudes toward violence, but believe that aggression is acceptable, necessary, and unavoidable in response to specific transgressions committed by their partners (“I don’t hurt anyone unless they deserve it”). Whether the implicit bias is general or specific in nature, such automatic cognitive distortions may then prime aggression-prone individuals to interpret interpersonal conflicts in a way that favors an aggressive response option (Anderson & Bushman, 2002; Polaschek et al., 2009).

We hypothesized that men in the IPV group would automatically associate women with aggression. The IAT findings supported this prediction; men in the IPV group were faster than NV men at associating aggression-related constructs and female names. Although we would once again advocate caution against overinterpretation of IAT effects, these findings are congruent with relevant models of aggression etiology. For example, cognitive models of aggression (Berkowitz, 1993, 2008; Huesmann, 1988) propose that aggressive scripts exist within larger knowledge structures consisting of associated memories, emotions, and other beliefs (e.g., Bushman, 1996) that, via experience, become overlearned, cue activated, and easily accessible (Todorov & Bargh, 2002). The present results suggest that IPV men may have developed associative connections between gender concepts (i.e., women) and aggressive programs of behavior (i.e., scripts). To the extent that other priming factors linked with these concepts are also present (e.g., angry affect), hostile interactions with a female partner may automatically activate aggressive responses and trigger a decision-making process that implicitly favors IPV as a problem-solving option (Holtzworth-Munroe, 1992).

It is important to be mindful of the preliminary nature of these data and to consider several important limitations regarding their interpretation. First, other implicit or explicit cognitive constructs may be more strongly related to IPV than those assessed herein. For example, although our implicit measure of attitudes toward women did not reveal group differences, we did not measure hostility toward women, which prior research has shown to be related to IPV perpetration (e.g., Holtzworth-Munroe, 2000). Second, the sample of IPV perpetrators assessed in this study was relatively homogeneous and drawn from a small number of BIPPs in a Midwestern state; such individuals may not be representative of the larger population of IPV offenders. Third, it is possible that the group differences revealed on the IAT measures were due to factors other than IPV, such as intelligence/verbal ability or other relationship dynamics (e.g., divorce, infidelity, etc.). Although the violent and NV groups differed on several demographic variables (e.g., age, education), we deliberately did not use analysis of covariance (ANCOVA) to analyze group differences on explicit and implicit cognitive measures. As noted by Miller and Chapman (2001), any attempt to statistically “control” for additional non-grouping-related variables that are associated with the dependent variable by removing their specific associations with the outcomes is both logically and empirically untenable, especially for nonrandomized designs. Age, for example, is one of the most consistent predictors of IPV and thus could be regarded as a defining aspect of violent versus NV populations; as noted by Miller and Chapman (2001), “Removing variance associated with age would, in effect, corrupt the grouping variable itself” (p. 44) and would render the resulting ANCOVA results meaningless. Using the strategy of incorporating (rather than controlling for) such variables in our analyses recommended by Cohen, Cohen, West, and Aiken (2003) and Miller and Chapman (2001), we found that, with one exception (gender IAT), demographic factors were uncorrelated with IAT outcomes in both groups.

Finally, it is important to note that the context of assessment can affect responding to both explicit and implicit measures. Although it is tempting to regard the IAT as a measure that can access a “bona fide pipeline” of one’s true attitudes and mental representations (e.g., Fazio et al., 1995), prior research suggests that responses to the IAT can be influenced by contextual factors (Boysen, Vogel, & Madon, 2006; Richeson & Ambady, 2001). Given that the violent participants completed the IAT at abuse intervention programs, it is possible that the setting as well as the content discussed during the sessions could have influenced violent participants’ IAT responses. However, even if one allows for this possibility, the consistent antiviolence, proegalitarian themes discussed during group sessions should have eliminated the expected IAT effects among violent participants. In contrast, the present results suggest only a minimal effect of context, as group differences on the IAT were in the predicted direction for two of the three IATs, and the association between IAT scores and sessions attended was nonsignificant.

With these limitations in mind, the present results suggest that clinicians working with IPV perpetrators and researchers evaluating cognitive models of IPV etiology should more carefully consider the full range of cognitive constructs associated with IPV. The overlearned and automatic nature of the cognitive processing systems of chronically violent individuals (Berkowitz, 2008; Polaschek et al., 2009) indicate that implicit measures of violence-condoning attitudes could complement explicit measurement, especially in situations where attitudinal assessment may be compromised by offender underreporting or minimization. However, it is worth noting that responses to implicit measures can also be influenced by respondent’s use of impression management tactics (for a review, see Gawronski, LeBel, & Peters, 2007); an implicit measure such as the IAT does not automatically yield responses liberated from social desirability biases. Implicit measures may nevertheless provide a unique assessment of how attitudes are activated in memory, whereas explicit measures may be a proxy for how the respondent generally thinks about the validity of activated information (Gawronski & Bodenhausen, 2006). Thus, a combination of implicit and explicit measures may provide useful findings regarding an offender’s initial biased associations (conscious or unconscious), while data from explicit measures may yield additional details that can better capture the complexity of the more generalized cognitive structure. This conceptualization may be useful for clinicians and researchers who aim toward a more comprehensive understanding of the nature and function of attitudes expressed by offenders involved in treatment as well as for researchers evaluating whether and how cognitively focused interventions for IPV offenders actually change behavior.

In summary, the present results suggest that, relative to NV individuals, IPV offenders show a pattern of attitudinal activation that favors a positive association for violence-related stimuli and a more strongly developed link between female gender and violent concepts on implicit attitudinal measures. However, when asked to report on their attitudes toward gender roles, IPV, and IPV on self-report/paper-and-pencil measures, no such group differences emerged. These preliminary data suggest a critical need for examining whether responses to implicit measures actually predict behaviors of clinical interest, such as IPV risk, response to IPV interventions, or other outcomes, and suggest the utility of integrating both explicit and implicit methods when assessing the cognitive activity of IPV offenders.

Footnotes

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by funds provided by the Department of Psychological Sciences, Purdue University.

Notes

References

Anderson

Bushman

(2002). Human aggression. Annual Review of Psychology, 53, 27-51.

Anglin

Holtzworth-Munroe

(1997). Comparing the responses of maritally violent and nonviolent spouses to problematic marital and nonmarital situations: Are the skill deficits of physically aggressive husbands and wives global? Journal of Family Psychology, 11, 301-313.

Archer

Graham-Kevan

(2003). Do beliefs about aggression predict physical aggression partners? Aggressive Behavior, 29, 41-54.

Beck

A. T.

(1999). Prisoners of hate. New York, NY: Harper Collins.

Berkowitz

(1993). Towards a general a theory of anger and emotional aggression. In Wyer

Srull

(Eds.), Advances in social cognition (Vol. VI, pp. 1-46). Hillsdale, NJ: LEA.

Berkowitz

(2008). On the consideration of automatic as well as controlled psychological processes in aggression. Aggressive Behavior, 34, 117-129.

Bograd

(1984). Family systems approaches to wife battering: A feminist critique. American Journal of Orthopsychiatry, 54, 558-568.

Boysen

G. A.

Vogel

D. L.

Madon

(2006). A public versus private administration of the implicit association test. European Journal of Social Psychology, 36, 845-856.

Brown

A. S.

Gray

Snowden

(2009). Implicit measurement of sexual associations in child sex abusers: Role of victim type and denial. Sexual Abuse, 21, 166-180.

10.

Burt

(1980). Cultural myths and supports for rape. Journal of Personality and Social Psychology, 38, 217-230.

11.

Bushman

(1996). Individual differences in the extent and development of aggressive cognitive-associative networks. Personality and Social Psychology Bulletin, 22, 811-819.

12.

Cohen

West

S. G.

Aiken

L. S.

(2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: LEA.

13.

Eckhardt

C. I.

Barbour

K. A.

Davison

G. C.

(1998). Articulated thoughts of maritally violent and nonviolent men during anger arousal. Journal of Consulting and Clinical Psychology, 66, 259-269.

14.

Eckhardt

C. I.

Dye

M. L.

(2000). The cognitive characteristics of maritally violent men. Cognitive Therapy and Research, 24, 139-158.

15.

Empirisoft. (2006). DirectRT Precision Timing Software (Version 2006.2) [Computer software]. New York, NY: Author.

16.

Fazio

R. H.

Jackson

J. R.

Dunton

B. C.

Williams

C. J.

(1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Personality and Social Psychology, 69, 1013-1027.

17.

Fincham

F. D.

Bradbury

T. N.

Arias

Byrne

C. A.

Karney

B. R.

(1997). Marital violence, marital distress, and attributions. Journal of Family Psychology, 11, 367-372.

18.

Gawronski

Bodenhausen

G. V.

(2006). Associative and propositional processes in evaluation: An integrative review of implicit and explicit attitude change. Psychological Bulletin, 132, 692-731.

19.

Gawronski

LeBel

E. P.

Peters

K. R.

(2007). What do implicit measures tell us? Scrutinizing the validity of three common assumptions. Perspectives on Psychological Science, 2, 181-193.

20.

Gray

N. S.

MacCulloch

M. J.

Smith

Morris

Snowden

R. J.

(2003). Violence viewed by psychopathic murderers. Nature, 423, 497-498.

21.

Gray

N. S.

Brown

A. S.

MacCulloch

M. J.

Smith

Snowden

R. J.

(2005). An implicit test of the associations between children and sex in pedophiles. Journal of Abnormal Psychology, 114, 304-308.

22.

Greenwald

A. G.

McGhee

J. L.

Schwartz

J. L.

(1998). Measuring individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74, 1464-1480.

23.

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, 197-216.

24.

Greenwald

A. G.

Poehlman

Uhlmann

Banaji

M. R.

(2009). Understanding and using the Implicit Association Test: III. Meta-analysis of predictive validity. Psychological Bulletin, 97, 17-41.

25.

Henning

Holdford

(2006). Minimization, denial, and victim blaming by batterers: How much does the truth matter? Criminal Justice and Behavior, 33, 110-130.

26.

Henning

Jones

A. R.

Holdford

(2005). “I didn’t do it, but if I did I had a good reason”: Minimization, denial, and attributions of blame among male and female domestic violence offenders. Journal of Family Violence, 20, 131-139.

27.

Holtzworth-Munroe

(1992). Social skill deficits in maritally violent men: Interpreting data using a social information processing model. Clinical Psychology Review, 12, 605-618.

28.

Holtzworth-Munroe

(2000). Social information processing skills deficits in maritally violent men: Summary of a research program. In Vincent

Jouriles

(Eds.), Domestic violence: Guidelines for research-informed practice (pp. 13-36). London, UK: Kingsley.

29.

Huesmann

L. R.

(1988). An information processing model for the development of aggression. Aggressive Behavior, 14, 13-24.

30.

Kaufman Kantor

Straus

(1990). Response of victims and the police to assaults on wives. In Straus

M. A.

Gelles

R. E.

(Eds.), Physical violence in American families (pp. 473-487). New Brunswick, NJ: Transaction.

31.

King

L. A.

King

D. W.

(1993). Manual for the Sex-Role Egalitarianism Scale. London, Ontario, Canada: Research Psychologists Press/Sigma Assessment Systems.

32.

Locke

Wallace

(1959). Short marital adjustment and prediction tests: Their reliability and validity. Marriage and Family Living, 2, 251-255.

33.

Miller

G. A.

Chapman

J. P.

(2001). Misunderstanding analysis of covariance. Journal of Abnormal Psychology, 110, 40-48.

34.

Murphy

C. M.

Hoover

S. A.

(1999). Measuring emotional abuse in dating relationships as a multifactorial construct. Violence & Victims, 39, 39-53.

35.

Nosek

B. A.

Greenwald

A. G.

Banaji

M. R.

(2005). Understanding and using the Implicit Association Test: II. Method variables and construct validity. Personality and Social Psychology Bulletin, 31, 166-180.

36.

O’Leary

K. D.

(1988). Physical aggression between spouses: A social learning perspective. In Van Hasselt

V. B.

Morrison

Bellack

Hersen

(Eds.), Handbook of family violence (pp. 31-55). New York, NY: Plenum.

37.

Polaschek

Calvert

Gannon

(2009). Linking violent thinking: Implicit theory based research with violent offenders. Journal of Interpersonal Violence, 24, 75-96.

38.

Richeson

J. A.

Ambady

(2003). Effects of situational power on automatic racial prejudice. Journal of Experimental Social Psychology, 39, 177-183.

39.

Robertson

Murachver

(2007). Correlates of partner violence for incarcerated women and men. Journal of Interpersonal Violence, 22, 639-655.

40.

Saunders

D. G.

Lynch

A. B.

Grayson

Linz

(1987). Inventory of Beliefs About Wife Beating: The construction and initial validation of a measure of beliefs and attitudes. Violence and Victims, 2, 39-55.

41.

Scott

Straus

(2007). Denial, minimization, partner blaming, and intimate aggression in dating partners. Journal of Interpersonal Violence, 22, 851-871.

42.

Stith

Smith

Penn

Ward

Tritt

(2004). Intimate partner physical abuse perpetration and victimization risk factors: A meta-analytic review. Aggression and Violent Behavior, 10, 65-98.

43.

Straus

M. A.

Hamby

S. L.

Boney-McCoy

Sugarman

D. B.

(1996). The Revised Conflict Tactics Scales (CTS-2). Journal of Family Issues, 17, 283-316.

44.

Sugarman

D. B.

Frankel

S. L.

(1996). Patriarchal ideology and wife-assault: A meta-analytic review. Journal of Family Violence, 11, 13-40.

45.

Sugarman

D. B.

Hotaling

G. T.

(1997). Intimate violence and social desirability: A meta-analytic review. Journal of Interpersonal Violence, 12, 275-290.

46.

Todorov

Bargh

J. A.

(2002). Automatic sources of aggression. Aggression & Violent Behavior, 7, 53-68.

47.

Ward

(2000). Sexual offenders’ cognitive distortions as implicit theories. Aggression and Violent Behavior, 5, 491-507.

48.

Yllo

K. A.

Straus

M. A.

(1990). Patriarchy and violence against wives: The impact of structural and normative factors. In Straus

M. A.

Gelles

R. J.

(Eds.), Physical violence in American families (pp. 383-399). New Brunswick, NJ: Transaction.