The Role of Control in Intimate Partner Violence: A Study in Dutch Forensic Outpatients.

Johnson argued that coercive control is crucial in explaining heterogeneity in intimate partner violence, with such violence being more frequent, less reciprocal, and more often male-to-female aggression when it serves to exercise control over the partner. We assessed 280 Dutch forensic outpatients who had recently engaged in intimate partner violence on nonaggressive coercive control. Control showed significant, small to moderate, associations with more frequent past year acts of psychological aggression, physical assault, and sexual coercion and more frequently resulted in partner injury. Control was unrelated to reciprocity of partner violence. High controlling violence was enacted mostly, but not exclusively by men. Overall, while perhaps not having a uniquely strong association, our findings provide partial support for the role of coercive control in intimate partner violence and suggest it may benefit intimate partner violence risk assessment.

Introduction

Research shows great variation in findings regarding the frequency, reciprocity, and gender distribution of intimate partner violence (IPV). Johnson (1995; for expansion to dyads, see Johnson, 2008) reasoned that there may be two types of IPV: common couple violence and patriarchal terrorism. Common couple violence would be typically paired with less frequent assaults, be more reciprocal, and more equally often committed by men and women. Patriarchal terrorism would be typically paired with more frequent assaults, be more often unidirectional, and most often be man-to-female aggression. The crucial element distinguishing the two types of IPV, however, is not the frequency, reciprocity, or gender distribution of the violence but rather the psychological dynamic leading to the violence, according to Johnson (1995): Common couple violence is reasoned to result from conflict between the couple, whereas the patriarchal terrorism type of violence reflects a means to exert control over the partner. Coercive control can be exerted in several ways (e.g., economically, socially, sexually), with the exertion of violence being just one of the means to achieve control over the partner.

Several studies have found support for Johnson’s (1995) predictions on the role of control in IPV (see, e.g., Graham-Kevan & Archer, 2003; Hardesty et al., 2015; Leone, Johnson, Cohan, & Lloyd, 2004; for a review, see Johnson, 2008). Johnson and Leone (2005), for instance, found that victimization by patriarchal terrorism was more severe and more persistent, than IPV in the context of common couple violence. These studies are, however, not without limitations. First, the theory is tested mostly by its developer, calling for independent replication. Indeed, as recognized by Johnson (2008), “I could cite very few studies that were specifically designed to answer even the most basic questions about the differences among intimate terrorism. . .and situational couple violence” (p. 85). Second, the dichotomization between the two types of violence often rests on an arbitrary cutoff point (in Johnson and Leone, 2005: 0-2 vs. 3-7 controlling behaviors). As control is a dimension rather than a dichotomized construct, analyses using control as a continuum are not only more powerful but also more appropriate. Third, the measurement of coercive control was often construed ad hoc based on the (archival) data. Few studies used measures that were a priori designed to assess coercive control. Fourth, a crucial test for the theory is that the key variable—controlling behaviors—consists of nonviolent forms of control. If not, the distinction made by Johnson (1995) would be primarily one between less versus more severe forms of violence. Among the items assessing control, however, were often also items that describe emotional forms of IPV (e.g., “calls you names or puts you down in front of others”). In sum, more work is needed to test the idea that the crucial variable in understanding heterogeneity in IPV is nonviolent coercive control.

Understanding the role of control in IPV may not only help to explain variation in research findings regarding IPV frequency, reciprocity, and gender distribution, but also be of clinical relevance for IPV risk assessment and risk management. The current study tests the role of nonviolent coercive control in IPV. We assessed IPV perpetrators on nonviolent controlling behaviors. From Johnson (1995), it was predicted that attempts to control one’s partner would be associated with more frequent IPV, less reciprocal IPV, and more often committed by men.

Method

The current study made use of the initial (T0) data of a larger study (Buitelaar, Posthumus, Scholing, & Buitelaar, 2014). Briefly, participants were recruited in two locations of a Dutch center for outpatient forensic mental health care, where they had been referred to by court, the probation service, or the primary health care service because of having committed IPV.

Participants

Patients were included if they were at least 18 years old, spoke sufficient Dutch, provided written informed consent, and had engaged into IPV in the last 8 weeks prior to intake. The measure of controlling behaviors was added in October 2013 and the study was closed in 2017. We included the data of all patients who completed measures of coercive control and IPV (n=285), excluding data from five patients with >10% missing data on the measures of coercive control (Lilienfeld & Widows, 2005). The final sample (n = 280) consists of 224 males (80%) and 56 females (20%) and is described in Table 1. It is worth mentioning that, in the past year, 64% of the patients had engaged into severe psychological aggression, 37% into severe physical assault, 17% had inflicted severe injury to their partner, and 4% had engaged into severe sexual coercion.

Table 1.

Descriptive Statistics (M With SD in Parentheses) for the Entire Sample, and for Men and Women, Separately.

	Whole Sample (n = 280)	Men (n = 224)	Women (n = 56)	p Value Sex Difference (t Test/χ2)
IPV severity past year (CTS-2)
Psychological aggression	1.60 (.57)	1.58 (.59)	1.66 (.48)	.34
Physical assault	1.06 (.83)	1.06 (.82)	1.07 (.86)	.91
Injury	0.63 (.75)	0.64 (.77)	0.61 (.70)	.78
Sexual coercion	0.18 (.48)	0.21 (.50)	0.07 (.37)	.03*
Coercive control	26.29 (8.36)	25.58 (8.03)	29.15 (9.11)	<.01**
Age	36.87 (10.38)	37.88 (10.47)	32.88 (9.02)	<.01**
Personality disorder traits (PDQ-R)
Antisocial personality disorder traits	2.61 (1.99)	2.68 (1.98)	2.31 (2.00)	.21
Borderline personality disorder traits	4.57 (1.86)	4.52 (1.84)	4.80 (1.93)	.32
GAF	35.54 (7.88)	35.67 (7.84)	35.00 (8.09)	.57
Education level
Elementary school (%)	20	20	22	.49
Vocational education (%)	64	63	67
Higher education (%)	16	17	11
Nationality (% Western)	85	85	86	.93
Had contact with partner in last 8 weeks (%)	94	93	96	.60
Lives together with partner (%)	63	61	71	.14
Lives in own (rental) house (%)	85	85	84	.80
Works at least 3 days per week (%)	55	59	41	.02*
Mandatory treatment (%)	16	19	4	<.01**
Police contact for IPV (%)	41	45	23	<.01**
Conviction for IPV (%)	16	20	4	<.01**

Note. The GAF score (obtained through psychiatric evaluation using the Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev.; DSM-IV-TR; American Psychiatric Association, 2000) of 35 corresponds to major impairment in several areas, such as work or school, family relations, judgment, thinking, or mood. Personality disorder traits were screened for using the PDQ-R (Hyler et al., 1988). IPV = intimate partner violence; CTS = Conflict Tactics Scales; PDQ-R = Personality Disorder Questionnaire Revised; GAF = Global Assessment of Functioning.

Significance at p < .05. **Significance at p < .01.

Procedure

After checking inclusion criteria and providing written informed consent, the patient was given a paper and pencil booklet consisting of a series of questionnaires including the Intimate Partner Violence Control Scale (IPVCS), Conflict Tactics Scale (CTS)-2, and the PDQ-R. There was an opportunity to ask questions and clarify survey items. Patients individually completed the booklet at their own pace in their preferred location and either mailed or brought it back.

Measures

Control: IPVCS

The IPVCS (Bledsoe & Sar, 2011) consists of 16 items (e.g., “I wish I had more say over the kinds of things my partner does with his or her friends”) that assess nonaggressive controlling behaviors on a scale from 1 (never) to 5 (very often). Items were summed to provide a total score for controlling behaviors. Cronbach’s alpha of the IPVCS was α = .82.

IPV Frequency: CTS-2

We used the 34 items of the CTS (Straus, Hamby, & Warren, 2003) that assess physical assault (e.g., “I threw something at my partner that could hurt” [minor], “I used a knife or gun on my partner” [severe]), injury (e.g., “My partner felt physical pain the next day because of a fight we had” [minor], “My partner had a broken bone from a fight with me” [severe]), psychological aggression (e.g., “I insulted or swore at my partner” [minor], “I threatened to hit or throw something at my partner” [severe]), and sexual coercion (e.g., “I made my partner have sex without a condom” [minor], “I used threats to make my partner have sex” [severe]) toward the partner in the last 8 weeks. The response options were never, 1 time, 2 times, 3 to 5 times, 6 to 10 times, 11 to 20 times, or more than 20 times. Following the CTS-2 manual, we calculated severity scores for IPV in the last year, with levels 0 = no (neither minor nor severe items endorsed), 1 = minor (at least one minor, but no severe items endorsed), 2 = severe (at least one severe item endorsed).

IPV Reciprocity: CTS-2

The patient assessed not only acts for perpetration but also for victimization by their partner. Following Graham-Kevan and Archer (2003), we subtracted the frequency of physical assault, injury, psychological aggression, and sexual coercion committed by the partner from the frequency of physical assault, injury, psychological aggression, and sexual coercion committed by the patient to obtain a measure of reciprocity of physical assault, injury, psychological aggression, and sexual coercion. Scores vary from −2 to +2. A positive score means that the patient engages more often into violence than the partner, a score around zero indicates that the violence is reciprocal, and a negative score suggests that the patient engages less often in violence than the partner.

Statistical Analyses

Because control is a dimension rather than a dichotomized construct, statistical analyses using control as a continuum are most appropriate. Because of the ordinal nature of the IPV severity scores, we tested predictions regarding IPV frequency and IPV reciprocity by calculating Spearman rho correlations between coercive control and IPV frequency and reciprocity. Because missing items were very rare (≤2.5% per measure), we used pairwise deletions. Following Cohen (1988), correlations of .1, .3, and .5 can be regarded as small, moderate, and large effects, respectively.

Although Johnson (1995, 2008) considers control to be a continuum, dimensional analyses do not allow to test the prediction that men are more likely than women to engage into controlling violence (i.e., in this prediction sex and control are phrased as categorical variables). The sex asymmetry prediction requires identification of the subgroup of patients engaging in high controlling violence. Because all patients included in this study had recently engaged into IPV, we relied solely on our measure of coercive control for this categorization. Specifically, we ran a hierarchical cluster analysis using Ward’s method with squared Euclidean distance as the proximity measure on the IPVCS items, determining the optimum number of clusters based on subjective assessment of the change in agglomeration coefficients. Once we have identified clusters of low versus high controlling patients, we assessed the proportion of males in the high controlling cluster.

Results

Coercive control showed significant, small to moderate, positive association with enacting physical assault, inflicted injury, enacting psychological aggression, and engaging into sexual coercion (see the first line of Table 2). Coercive control was essentially unrelated to reciprocity of physical assault, injury, psychological aggression, and sexual coercion (see also first line of Table 2).

Table 2.

The Association Between Controlling Behaviors and IPV Frequency and Reciprocity of Psychological Aggression, Physical Assault, Inflicted Injury, and Sexual Coercion.

	Frequency				Reciprocity
	Physical Assault	Inflicted injury	Psychological Aggression	Sexual Coercion	Physical Assault	Inflicted Injury	Psychological Aggression	Sexual Coercion
Controlling behaviors	.23**	.28**	.36**	.26**	−.09	−.04	−.08	.07
Frequency
Physical assault		.68**	.45**	.10	.43**	.29**	.11	.05
Inflicted injury			.37**	.12*	.10	.48**	.03	.08
Psychological aggression				.22**	.15*	.15*	.46**	.09
Sexual coercion					.02	.02	.04	.71**
Reciprocity
Physical assault						.38**	.40**	.10
Inflicted injury							.21*	.09
Psychological aggression								.12*

Note. IPV = intimate partner violence.

Significance at p < .05. **Significance at p < .01.

The cluster analysis showed a large change in the agglomeration coefficients from one to two clusters of −933, indicating that having two instead of one cluster added in differentiating between cases. The subsequent changes to a three-, four-, and five-cluster solution were −241, −155, −134, indicating that having more clusters adds less to this differentiation. The two-cluster solution consisted of one larger cluster (n = 210) low in coercive control (M = 22.88, SD = 5.22) and one smaller cluster (n = 63) high in coercive control (M = 37.63, SD = 6.69), t(271) = 18.37, p < .001, d = 2.64. Replicating the correlational analyses above, the high controlling cluster showed higher rates of psychological aggression, t(271) = 3.83, p < .001, d = .47, physical assault, t(269) = 2.38, p = .02, d = .34, inflicted injury, t(271) = 2.62, p = .01, d = .38, and sexual coercion, t(270) = 2.72, p = .01, d = .49, than the low controlling cluster, albeit with small effect sizes. Also replicating the correlational analyses above, the clusters did not differ in reciprocity of psychological aggression, t(267) = 0.20, p = .84, d = .03, physical assault, t(269) = 0.56, p = .57, d = .08, inflicted injury, t(271) = 0.98, p = .33, d = .14, or sexual coercion, t(268) = 1.37, p = .17, d = −.24. Importantly, the prime purpose of the cluster analyses was to assess Johnson’s predictions on sex asymmetry. Note that women scored higher on coercive control than men, t(271) = 2.84, p = .005, d = .43, and that there were relatively more women in the high controlling cluster (29% female) than in the low controlling cluster (17% female), χ2(1) = 3.99, p = .046, φ = .12. Nonetheless, we think the most appropriate way to evaluate Johnson’s prediction that controlling violence is mostly, and near exclusively, perpetrated by men is to assess the proportion of men in the high controlling cluster. With 71% male, the high controlling cluster was predominantly, but not exclusively, male.

Discussion

Our study aimed to test key predictions from Johnson’s (1995) typology of IPV that assigns a central role to coercive control in explaining IPV heterogeneity. We found partial support for the evaluated predictions: Coercive control was indeed related to more frequent IPV, and mostly enacted by men, but we found no support for the prediction that control would be associated with unidirectional violence.

Coercive control was positively associated with IPV frequency, with small to moderate effect sizes (.23 ≤ r ≤ .36). To assess the importance of coercive control, it may be worthwhile to compare the associative strength of coercive control with that of other risk factors for IPV. The mean effect size of 26 risk factors in the meta-analysis of Stith Smith, Penn, Ward, and Tritt (2004) was r = .22. Thus, while perhaps not having a uniquely strong association, our study provides further evidence that coercive control is associated with IPV, suggesting it may be worthwhile for considering coercive control in risk assessment.

The majority of our patients engaging into controlling violence were men, substantiating the idea that controlling violence is more often committed by men than by women. Note that this finding is qualified by the fact that our entire sample was predominantly male. Furthermore, while our findings confirm a sex asymmetry, we found no support for the idea that controlling violence would be committed “almost exclusively” (Johnson, 1995, p. 286) or “almost entirely” (Johnson, 2008, p. 2) by men, with 29% of the controlling violence perpetrated by women.

Control was unrelated to reciprocity of IPV in this sample. Irrespective of control, the vast majority of patients had a reciprocity score of 0 ± 1 (indicative of reciprocal violence) for psychological aggression (97%), physical assault (89%), injury (94%), and sexual coercion (97%).

This study is not without limitations. First, as in many studies on IPV, we relied upon self-report. Perpetrators of IPV may, however, lack insight into their own behavior and/or underreport aggression and psychopathology. Second, our findings are constrained to a predominantly male forensic outpatient sample and cannot be assumed to speak to the nature of IPV in the entire population (a logical error known as the clinical fallacy).

Our study also had important strengths, including a fairly substantial forensic sample, a dedicated measure of nonaggressive coercive control, and analyses considering coercive control across the whole continuum. An additional strength may be the study setting. Previous studies on IPV have often used samples that either greatly reduced the chance of observing controlling violence (e.g., undergraduates) or of noncontrolling violence (e.g., inmates). In contrast, our sample may be regarded as “mixed” in the sense of including most likely both controlling violence (a forensic sample; high frequency of IPV) and noncontrolling violence (treatment was nonmandatory for the majority of patients). As such, our study provided an important test of the role of control in IPV.

Forensic treatment is more effective when based on the risk–need–responsivity (RNR) model (Andrews & Bonta, 2010). In accordance with the RNR principles, treatment intensity is adjusted to the risk of recidivism (risk principle), the treatment goals are related to the dynamic risk factors (need principle), and the treatment is adapted to the capacity of the client (responsivity principle). Based on the results of this study that control is associated with more frequent past year acts of psychological aggression, physical assault, and sexual coercion and more frequently resulted in partner injury, assessing coercive control could become part of risk assessment with more intensive treatment recommended for controlling IPV.