Psychometric Properties and Factor Analysis of a Short Form of the Multidimensional Measure of Emotional Abuse

Abstract

The 28-item Multidimensional Measure of Emotional Abuse (MMEA) assesses four common forms of emotional abuse in intimate relationships and has been used extensively to study the development of intimate partner violence (IPV), the consequences of emotional abuse, and the outcomes of IPV interventions. The current study provides psychometric analyses of a shortened version of the MMEA using self-report data from a sample of men receiving treatment at a community-based relationship violence intervention program (RVIP; N = 467) and reports from their relationship partners (N = 252), and data from a sample of undergraduate students (N = 194) who reported on their own and their partners’ abusive behavior. Theoretical and statistical considerations, including internal consistency after item deletion, were used to select items for the shortened version. In the clinic sample (for self- and partner reports) and in the undergraduate sample (for self-report only), the 16-item MMEA-Short Form (MMEA-SF) retains the 4-factor structure of the 28-item MMEA. In both samples and across reporting methods (self and partner), the 16-item MMEA-SF has good internal consistency, good concurrent validity with the Revised Conflict Tactics Scales (CTS2) psychological aggression subscale, and similar correlations with CTS2 physical assault subscale as the original 28-item MMEA version. The MMEA-SF can reduce assessment burden while maintaining good domain coverage and strong psychometric properties and will be an asset to researchers and practitioners who need a brief, multifaceted measure of emotional relationship abuse in both clinic and undergraduate samples.

Keywords

emotional abuse psychometrics intimate partner violence factor analysis short form

Emotional, or psychological, abuse is an important facet of relationship dysfunction (Murphy & Cascardi, 1999; O’Leary, 1999), key to understanding both the development of physical violence in couples (Murphy & O’Leary, 1989; Smith-Marek et al., 2016) and the nature of coercive control in abusive relationships (Stark, 2007). Serious partner violence is almost always accompanied by a pervasive pattern of emotional abuse that occurs more frequently than physical assault (Murphy & Cascardi, 1999). Similarly, a daily diary study showed that emotional abuse occurred more frequently than physical assault and generally preceded and predicted the onset of physical assault (Sullivan et al., 2012). In fact, almost three-fourths of women who have experienced intimate partner violence (IPV) report that emotional abuse had more profound negative effects than physical abuse (Follingstad et al., 1990). Surprisingly, emotional abuse, in contrast to physical assault, is often more strongly and uniquely associated with survivors’ negative mental health outcomes, including post-traumatic stress symptoms (Norwood & Murphy, 2012). Even after controlling for exposure to physical assault, changes in emotional abuse exposure across time are associated with changes in depression and anxiety (Lawrence et al., 2009). A more thorough understanding of emotional abuse may elucidate processes involved in the development and maintenance of IPV, facilitate IPV risk prediction, and clarify the ways in which aversive and controlling behaviors influence relationship satisfaction, relationship dissolution, and negative mental health outcomes.

Assessment methods remain a critical priority for advances in this area (Follingstad, 2007). Early research on emotional abuse used brief lists of aggressive acts, mostly verbal in nature (e.g., Straus, 1979), or lengthy lists of highly coercive behaviors common to severely violent relationships (e.g., Tolman, 1989). Prior investigations have identified a single factor underlying small sets of psychological aggression items (e.g., Straus, 1979) but have generally failed to support more complex hypothesized factor structures involving different forms of emotional abuse. For instance, in developing the Psychological Maltreatment of Women Inventory, Tolman (1989) included items to reflect six conceptually distinct forms of psychological abuse, yet identified a two-factor structure, with each factor reflecting a complex amalgam of topographically and functionally diverse behaviors. The Multidimensional Measure of Emotional Abuse (MMEA; Murphy et al., 1999) was created to address this gap between the descriptive literature on emotional abuse and the available assessment instruments. More specifically, the descriptive literature and clinical writings had identified multiple types of emotional abuse with distinct forms and functions, whereas structural analyses indicated that the available assessment instruments did not adequately distinguish among different forms of abuse.

The goal was to develop a brief instrument that assessed conceptually distinct forms of emotional abuse with broad applicability across dating, cohabiting, and marital relationships. Efforts were made to ensure that the measure would apply to the behavior of both men and women, and to avoid item content that would apply only to specific relationship contexts, for example items involving monetary control that presume shared finances. The intention was to develop a measure that would be useful in understanding relatively common or normative forms of emotional abuse, as well as patterns of coercive control in clinically violent samples.

The MMEA development began with a review of the literature, which revealed four distinct forms of emotional abuse commonly identified in clinical and qualitative studies of partner abuse survivors (Murphy & Cascardi, 1999). The behaviors in each category of abuse were presumed to share intended or actual interpersonal consequences and to constitute a meaningful response class due to topographic and functional similarities. The following labels were applied: Dominance/intimidation involves a display of symbolic or object-oriented aggression that often signals the prospect of violent escalation. The presumed effect of dominance/intimidation is to produce submission, compliance, and fear reactions in the partner. Restrictive engulfment involves efforts to track and monitor the partner’s activities, isolate the partner from social contacts, and limit the partner’s personal freedom and growth. These behaviors are presumed to reduce perceived threats to the relationship from rivals to limit other people’s influence over the partner’s attitudes and decisions, and maximize the partner’s dependency on the abuser. Denigration involves humiliating, belittling, and degrading comments and actions. The presumed interpersonal functions result from attacks on the partner’s sense of self-worth, which may inflate the abusive individual’s sense of power and agency in the relationship and may limit the partner’s perceived relationship and life alternatives, for example by convincing the partner that he/she is worthless and unwanted by others. Finally, hostile withdrawal involves aversive or coercive withholding of contact, attention, or communication. These actions are presumed to heighten the partner’s anxiety or insecurity about the relationship, punish perceived transgressions, and to maintain power by avoiding the partner’s concerns or influence. These four forms of emotional abuse were hypothesized to represent distinct latent factors due to shared functional and topographic features at the level of group statistical analysis.

After the four distinct forms of emotional abuse were identified descriptively, items were selected from existing public domain measures and new items were written in an attempt to represent each form of emotional abuse. This yielded an initial pool of 54 items that was administered to a sample of college students in dating relationships (Murphy & Hoover, 1999). Principal components analysis identified four components which corresponded to the hypothesized forms of abuse outlined earlier.

Using a structural validation strategy described by Gurtman (1992), the authors identified a location for each form of emotional abuse within a circumplex model of interpersonal problems. The circumplex model plots interpersonal problem tendencies, assessed by subscales of the Inventory of Interpersonal Problems, into a two-dimensional circular space represented by two bipolar axes. One axis reflects dominance with poles representing nonassertive and domineering interpersonal problems. The other axis reflects affiliation with poles representing cold and overly nurturant interpersonal problems. The quadrants of this circular model represent various combinations along these axes. For example, “vindictive” is located in the region reflecting coldness and dominance, and “exploitable” is located in the region reflecting unassertiveness and over nurturance. The MMEA subscales were empirically “located” within this circumplex space using trigonometric weighting of the subscale correlations with the eight interpersonal problem scales. Using this method, all four forms of emotional abuse were characterized as coercive domination (i.e., located on the domineering side of the dominance/submission axis), with variability along the affiliation axis. For example, restrictive engulfment was located toward the intrusive end of the affiliation axis, whereas hostile withdrawal was located toward the cold end of the affiliation axis.

After this initial study, the 54-item pool was reduced to a set of 28 items (Murphy et al., 1999) based on psychometric analyses of an undergraduate dating sample using methods designed to prevent structural validity problems observed in factor analyses of prior emotional abuse measures. Specifically, an equal number of items (seven) was selected to represent each of the four hypothesized latent constructs based on the following empirical criteria: (a) endorsement frequency at or above 10%, (b) relative contribution to subscale internal consistency, (c) high loadings on the predicted factor, and (d) adequate discrimination of subscale content as indicated by substantial difference between primary and secondary loadings in principal components analyses. The resulting 28-item scale was labeled the MMEA.

Psychometric investigation has confirmed the 4-factor structure of the 28-item MMEA and shown it to be a reliable and valid measure of emotional abuse in intimate adult relationships (Murphy et al., 1999). The MMEA has been used to: (a) assess the effects of IPV interventions (e.g., Taft et al., 2016), (b) understand the effects of physical and emotional abuse on survivors (e.g., Lawrence et al., 2009), (c) explore presenting problems and clinical profiles among IPV offenders (e.g., Semiatin et al., 2017), and (d) understand factors that influence the development of partner abuse (e.g., Falconier, 2010). The 28-item MMEA has been translated into Italian (Bonechi & Tani, 2011), Spanish (Falconier, 2010), and Turkish (Toplu-Demirtaş et al., 2018).

The current report provides psychometric analyses on a shortened 16-item English language version of the MMEA using data from two samples: (a) a clinic sample of men undergoing treatment for IPV perpetration and their intimate relationship partners and (b) a sample of undergraduate students in dating relationships. The goal was to develop a briefer version of the MMEA to facilitate more rapid administration, while maintaining desirable features of the 28-item version, including the 4-factor structure. Briefer and more time-efficient measures can reduce assessment burden, prevent assessment fatigue, and allow researchers and practitioners to administer a wider range of assessments with limited time.

The primary goals of the current investigation were to test the proposed 4-factor structure of this 16-item version using confirmatory factor analysis (CFA), and evaluate concurrent and convergent validity through associations with the 28-item version and an established measure of physical relationship assault and psychological aggression.

Method

Participants

Clinic sample.

Data were gathered during the routine intake assessment for 467 men presenting for treatment at a community-based relationship violence intervention program between September, 2001 and January, 2009. An additional 48 men (8% of the total) were excluded because they did not provide consent to have their clinical data used for research and 50 (9%) were excluded because they failed to complete at least half of the items on the MMEA. Participants, on average, were 36.11 years old (SD = 10.56) and had attained 13.23 years of formal education (SD = 2.55). With respect to race/ethnicity, 45% of the analysis sample identified as non-Hispanic Caucasian, 43% identified as Black or African American, 5% identified as Hispanic/Latino, 4% as identified as Asian or Asian-American, 1% identified as Native American, and 3% identified as another race or ethnicity. The majority of the sample (77%) was court ordered to attend treatment; 56% were married, 11% divorced, and 33% never married; 57% reported being involved in an intimate relationship with the identified victim of their abusive behavior and 43% were no longer in the relationship. Efforts were made to contact the identified victim of the participant’s abuse (herein referred to as the partner), by telephone as part of the intake assessment, yielding partner data on 259 cases (55%).

Undergraduate sample.

Data on 190 undergraduate students in dating relationships were drawn from a study examining risk factors for relationship violence (Lorenzo, 2019). These individuals were enrolled at a medium-sized university in the mid-Atlantic region. A total of 397 students accessed the online survey; 203 were excluded because they reported not being in a current intimate relationship, and 4 did not answer any of the MMEA items. Of the remaining 190 individuals, 156 (82%) identified as women, 33 (17%) as men, and 1 (1%) provided no information on gender. They endorsed the following non-mutually exclusive race and ethnic categories: 53% White/Caucasian, 26% Asian, 18% Black/African American, 10% Hispanic/Latino, 1% American Indian/Alaska Native, 2% Native Hawaiian/Pacific Islander, and 4% Other.

Measures

MMEA-SF.

Table 1 contains a list of items included in the MMEA-SF. The elimination of items for the shortened version was based on statistical considerations, specifically improvement in internal consistency from item deletion based on data from prior studies and rational considerations, specifically: (a) limited applicability to some individuals or groups (e.g., “Drove recklessly to frighten the other person” is relevant only to those who operate motor vehicles, (b) complications with item wording or content (e.g., respondents indicating that they had no way to know if their partner had “secretly searched through the other person’s belongings”), and/or (c) overlap in content domain coverage with other MMEA items (e.g., “refused to have any discussion of a problem” and “refused to acknowledge a problem that the other person felt was important”).

Table 1.
MMEA Item Descriptions.

Item Description In 16-item Version

Restrictive Engulfment

1 Asked the other person where they had been or who they were with in a suspicious manner

2 Secretly searched through the other person’s belongings

3 Tried to stop the other person from seeing certain friends or family members ✓

4 Complained that the other person spends too much time with friends

5 Got angry because the other person went somewhere without telling him/her ✓

6 Tried to make the other person feel guilty for not spending enough time together ✓

7 Checked up on the other person by asking friends or relatives where they were or who they were with ✓

Denigration

8 Said or implied that the other person was stupid ✓

9 Called the other person worthless ✓

10 Called the other person ugly

11 Criticized the other person’s appearance

12 Called the other person a loser, failure, or similar term ✓

13 Belittled the other person in front of other people ✓

14 Said that someone else would be a better partner (better spouse, better girlfriend or boyfriend)

Hostile withdrawal

15 Became so angry that they were unable or unwilling to talk ✓

16 Acted cold or distant when angry ✓

17 Refused to have any discussion of a problem

18 Changed the subject on purpose when the other person was trying to discuss a problem

19 Refused to acknowledge a problem that the other person felt was important ✓

20 Sulked or refused to talk about an issue ✓

21 Intentionally avoided the other person during a conflict or disagreement

Dominance/intimidation

22 Became angry enough to frighten the other person ✓

23 Put his/her face right in front of the other person’s face to make a point more forcefully

24 Threatened to hit the other person ✓

25 Threatened to throw something at the other person

26 Threw, smashed, hit, or kicked something in front of the other person ✓

27 Drove recklessly to frighten the other person

28 Stood or hovered over the other person during a conflict or disagreement ✓

All participants reported the frequency of MMEA items using the response options of never (0), once (1), twice (2), three to five times (4), six to 10 times (8), 11 to 20 times (15), and more than 20 times (25). Following scoring conventions for the Conflict Tactics Scales (Straus et al., 1996), item responses were recoded to the midpoint of the frequency range for each response option (in parentheses earlier). Subscale and total scores for the original (28-item) and shortened (16-item) versions were computed by summing these recoded item frequencies. Participants who answered fewer than 75% of items on any scale or subscale were coded as missing for that scale. Clinic sample men reported on their own use of abusive behaviors during the past six months and successfully contacted partners likewise reported on abusive client’s behavior during the past six months. Undergraduate sample participants reported on the frequency of each behavior in the past 12 months, with separate reports for behavior by one’s self and one’s partner. Approximately 16% (N = 73) of clinic sample men reported no emotional abuse perpetration and 17% (N = 44) of clinic sample partners reported no emotional abuse victimization in the past six months (i.e. reported “never” for all of the MMEA items). In addition, 11% (N = 20) of undergraduates reported no emotional abuse perpetration in the past 12 months and 16% (N = 30) reported no emotional abuse victimization in the past 12 months.

Revised Conflict Tactics Scale (CTS2).

Participants and contacted partners in the clinic sample completed the CTS2 (Straus et al., 1996). Undergraduate sample participants completed only the Physical Assault scale from the CTS2. The CTS2 is the most widely used assessment of IPV and has sound psychometric properties (Vega & O’Leary, 2007). The 12-item CTS2 Physical Assault scale was used to examine construct validity of the MMEA-SF in both samples and the 8-item CTS2 Psychological Aggression scale was used to examine the concurrent validity of the MMEA-SF in the clinic sample. Item response recoding and scale computation followed the procedure outlined earlier for the MMEA. The Physical Assault frequency score had an internal consistency (Cronbach’s alpha) of .70 for self-report and .91 for partner report in the clinic sample. In the undergraduate sample, internal consistency was .73 for reports of behavior by one’s self and .79 for reports of behavior by one’s partner. The Psychological Aggression frequency score had an internal consistency of .75 for self-report and .84 for partner report in the clinic sample.

Data Analysis

In the clinic sample, the client’s self-report and the partner’s report on the MMEA both measure the client’s use of abuse (i.e., client’s perpetration). The inclusion of both self-reports and partner reports of the abusive client’s behavior reflects common practice in clinical research. Partner reports are often considered the “gold standard” for assessing levels of abuse. However, clinicians and researchers often have access to self-reports on all participants but only limited or partial access to victim partner reports. Although both reports measure the same behavior, analyses of self- and partner reports on the MMEA reports were conducted separately due to the large number of cases with no partner report data, and based on prior evidence of modest agreement in partner reports of abuse (Armstrong et al., 2002) which indicates that self-reports of abuse perpetration, in contrast to partner reports of abuse victimization, may be more influenced by factors such as positive impression management and shame or embarrassment (Riggs et al., 1989). Given the expected inequivalences across the two reporters in the clinic sample, measurement invariance was not tested in this initial psychometric study.

In contrast, in the undergraduate sample, the abusive behaviors reported are not perpetrated by the same individual (i.e., students report on their own behavior (perpetration) as well as their partner’s behavior (victimization). This single-informant data collection for perpetration and victimization is common in survey research and studies of dating violence. Given that the relationship partner does not report on the student’s behavior (as is the case with the clinic sample), it is not possible to test for inequivalence in multiple informant data in this population.

First, CFA was conducted on the 16 items from the shortened MMEA, which were log transformed prior to CFA to determine whether the 4-factor structure from the 28-item version was retained. Next, to assess concurrent validity, we examined MMEA-SF subscale and total score correlations with respective scales from the original 28-item version. Given that the correlations between the short and long versions are artificially inflated by redundant measurement error from using the same items, we present upper bound estimates (the observed correlation), lower bound estimates (the correlations between scales constructed from the items that were, and were not, selected for the short version), and reliability attenuated estimates of these associations. The latter used a reverse operation from the correction for attenuation (Cohen et al., 2003; Spearman, 1904), factoring the estimated measurement error for one administration of the reduced scale versions back into the estimated correlations with the full scale. Concurrent validity was further evaluated by examining correlations with the CTS2 Psychological Aggression scale in the clinic sample. Construct validity as a measure of abusive behavior was evaluated by examining correlations with the CTS2 Physical Assault scale in both samples.

The lavaan package (Rosseel, 2012) for R 3.4.2 (R Core Team, 2017) was used to conduct CFA. To determine whether the model exhibited reasonable fit, we used the following recommended cutoff values: normed chi-square value less than 5.0 (Bollen, 1989), Comparative Factor Index (CFI) value and Tucker Lewis Index (TLI) value greater than 0.95 (Hu & Bentler, 1999), Standardized Root Mean Square Residua (SRMR) value less than 0.08 (Hu & Bentler, 1999), and Root Mean Square Error of Approximation (RMSEA) value less than 0.05 (Browne & Cudeck, 1993). Relative model fit was evaluated for covariance-nested models using the chi-square difference test. The four-factor model was compared to a one-factor alternative in order to examine support for the multifactorial nature of the measure. Missing data were minimal for self- and partner reports in the clinic sample (5.35% and 3.47%, respectively) and in the undergraduate sample (5.26% for both) and were addressed using the full-information maximum likelihood method.

Results

Descriptive Data Analysis and Test of Assumptions

Tables 2 and 3 present raw item statistics (i.e., items recoded to the midpoint) and inter-item correlations based on log-transformed items for self- and partner report in the clinic sample and undergraduate sample, respectively. Table 4 presents scale statistics (based on raw items) in both samples. Because the assumption of multivariate normality was violated across reports in both samples, maximum likelihood estimation with robust (Huber-White) standard errors (Huber, 1967; White, 1980) and the Yuan-Bentler test statistic (Yuan & Bentler, 2000) were used to conduct the CFA analyses.

Table 2.
Inter-Item Correlation Matrix for Self- and Partner Reports for the Clinic Sample.

Item 3 5 6 7 8 9 12 13 15 16 19 20 22 24 26 28 M SD

3 .42 .43 .48 .33 .35 .29 .41 .23 .19 .28 .19 .35 .24 .38 .30 0.96 3.62

5 .54 .49 .56 .46 .32 .37 .43 .36 .36 .42 .36 .39 .33 .40 .44 1.96 4.41

6 .38 .56 .52 .34 .31 .36 .36 .25 .25 .35 .28 .30 .26 .29 .28 1.71 4.66

7 .52 .62 .36 .34 .33 .38 .43 .25 .28 .36 .30 .41 .36 .38 .40 1.42 4.28

8 .39 .39 .34 .36 .60 .50 .47 .44 .45 .41 .37 .41 .40 .41 .36 2.19 5.01

9 .33 .39 .33 .38 .68 .58 .42 .39 .36 .32 .26 .37 .32 .39 .32 1.05 3.45

12 .28 .36 .28 .39 .64 .65 .49 .35 .34 .40 .41 .37 .25 .39 .34 1.04 3.92

13 .41 .41 .37 .39 .60 .55 .61 .36 .31 .39 .27 .45 .37 .43 .46 1.03 3.49

15 .19 .29 .24 .15 .44 .36 .33 .35 .67 .56 .54 .48 .32 .43 .38 3.91 7.03

16 .25 .34 .32 .17 .40 .28 .28 .33 .75 .54 .53 .49 .34 .40 .39 5.09 7.78

19 .34 .35 .31 .35 .50 .43 .42 .51 .58 .63 .67 .50 .34 .49 .45 2.62 5.06

20 .25 .35 .26 .28 .38 .41 .40 .41 .55 .62 .77 .46 .31 .41 .35 3.07 5.97

22 .45 .42 .43 .42 .58 .54 .47 .55 .53 .48 .59 .44 .48 .57 .62 2.39 5.40

24 .38 .32 .25 .37 .46 .48 .51 .55 .39 .32 .40 .30 .53 .49 .50 0.86 3.12

26 .34 .26 .29 .24 .41 .43 .37 .44 .33 .41 .50 .38 .57 .47 .64 1.32 3.60

28 .30 .43 .38 .35 .50 .54 .49 .53 .38 .37 .47 .38 .64 .60 .53 1.28 3.95

M 2.98 2.63 4.00 2.79 5.47 3.81 2.74 3.69 5.43 6.77 7.94 6.32 5.54 2.49 2.48 3.35

SD 6.70 5.89 7.92 6.52 9.05 8.17 6.83 7.37 7.74 8.66 9.65 8.89 8.35 5.93 4.98 6.19

Note. The upper half is the inter-item correlation matrix for self-report (N = 438). The lower half is the inter-correlation matrix for the partner-report (N = 248). The item descriptives are based on item responses that were recoded to the midpoint of the frequency for each response option. The inter-item correlations are based on items that were recoded to the midpoint and then log-transformed.

Table 3.
Inter-Item Correlation Matrix for Self- and Partner Reports for the Undergraduate Sample.

Item 3 5 6 7 8 9 12 13 15 16 19 20 22 24 26 28 M SD

3 .34 .50 .21 .39 .34 .24 .31 .25 .23 .31 .17 .15 .25 .29 .42 0.65 2.44

5 .54 .36 .36 .47 .49 .41 .30 .42 .45 .36 .29 .40 .23 .29 .40 0.93 2.42

6 .52 .51 .29 .37 .32 .39 .41 .31 .30 .31 .27 .17 .19 .31 .34 1.64 4.10

7 .26 .52 .32 .13 .27 .16 .19 .13 .25 .20 .27 .32 .24 .19 .42 0.52 1.37

8 .42 .37 .42 .23 .54 .56 .42 .49 .42 .36 .28 .28 .27 .34 .27 1.51 3.44

9 .36 .19 .18 .14 .52 .59 .27 .44 .33 .26 .27 .27 .24 .28 .32 0.38 1.38

12 .38 .28 .32 .27 .61 .75 .37 .37 .29 .31 .30 .12 .15 .19 .20 0.30 1.19

13 .37 .30 .29 .13 .51 .65 .60 .30 .19 .43 .18 .26 .29 .32 .13 0.29 0.85

15 .43 .33 .36 .20 .44 .42 .43 .55 .68 .44 .35 .29 .17 .21 .13 1.86 3.91

16 .44 .38 .37 .17 .43 .32 .35 .46 .67 .37 .45 .27 .23 .24 .14 2.60 4.12

19 .40 .33 .27 .11 .43 .41 .38 .46 .60 .53 .40 .28 .21 .18 .22 0.60 1.25

20 .38 .37 .34 .23 .37 .29 .34 .45 .58 .70 .62 .23 .34 .15 .27 1.12 2.70

22 .39 .30 .32 .13 .38 .51 .48 .56 .45 .42 .41 .34 .33 .42 .45 0.31 0.99

24 .40 .24 .22 .20 .52 .52 .68 .51 .33 .32 .27 .27 .54 .52 .30 0.31 1.41

26 .41 .37 .33 .19 .34 .42 .62 .42 .43 .41 .39 .36 .57 .46 .41 0.42 1.61

28 .42 .28 .36 .34 .44 .64 .62 .58 .34 .31 .27 .29 .52 .53 .54 0.31 2.27

M 0.65 0.93 1.64 0.52 1.51 0.38 0.30 0.29 1.86 2.60 0.60 1.12 0.31 0.31 0.42 0.31

SD 2.44 2.42 4.10 1.37 3.44 1.38 1.19 0.85 3.91 4.12 1.25 2.70 0.99 1.41 1.61 2.27

Note. The upper half is the inter-item correlation matrix for self-report (N = 178). The lower half is the inter-correlation matrix for the partner-report (N = 178). The item descriptives are based on item responses that were recoded to the midpoint of the frequency for each response option. The inter-item correlations are based on items that were recoded to the midpoint and then log-transformed.

Table 4.
Scale Statistics for Self-report and Partner Report.

Clinic Sample Undergraduate Sample

Self-Report^a Partner Report^b Self-Report^c Partner Report^d

M SD M SD M SD M SD

16-Item MMEA-Short Form

Restrictive engulfment 6.24 13.27 12.64 21.59 3.64 7.18 3.69 7.30

Denigration 5.60 13.13 16.26 27.68 2.37 5.24 3.01 10.74

Hostile withdrawal 14.76 21.38 26.67 30.83 6.22 9.23 9.12 16.46

Dominance/intimidation 5.89 13.00 14.02 20.88 1.48 4.90 1.81 6.03

Total 32.27 47.90 69.43 82.43 13.69 19.70 17.64 32.57

28-Item Original MMEA

Restrictive engulfment 12.04 22.50 24.87 37.18 8.74 16.09 8.32 15.91

Denigration 8.13 18.45 22.93 40.05 3.70 7.69 4.42 13.03

Hostile withdrawal 24.22 34.36 45.22 53.47 9.60 15.33 14.32 26.84

Dominance/intimidation 8.95 20.04 20.82 31.66 2.06 6.27 2.78 9.39

Total 53.24 75.46 113.67 132.41 24.16 36.11 29.96 51.46

Note. These descriptives are based on sum scores calculated from raw items recoded to the midpoint.

^aN = 457–466

^bN = 255–259

^cN = 187–189

^dN = 186–189

Confirmatory Factor Analyses

Clinic sample self-reports.

For self-reports of emotionally abusive behavior in the clinic sample, all the CFA goodness-of-fit indices indicated that the four-factor measurement model fit the data well (normed-χ² = 1.70, CFI = .96, TLI = .95, RMSEA = .05, SRMR = .04). As seen in Figure 1, the standardized factor loadings, ranging from .59 to .80, and the estimated correlation coefficients among the four latent variables were large. A test of relative fit indicated that the four-factor model provided a significantly better fit than the one-factor model (Δχ²[6, N = 467] = 199.74, p < .001).

Figure 1.
Note. Self-report (in bold) model fit: normed-χ² = 1.70, CFI = .96, TLI = .95, RMSEA = .05, SRMR = .04. Partner report model fit: normed-χ² = 1.59, CFI = .95, TLI = .94, RMSEA = .07, SRMR = .05.

Clinic sample partner reports.

For partner reports of the client’s emotionally abusive behavior, three CFA goodness-of-fit indices indicated that the four-factor measurement model fit the data well (normed-χ² = 1.59, CFI = .95, SRMR = .05) and two goodness-of-fit indices (TLI = .94, RMSEA = .07) showed that the model fit was fair since the TLI was less than .95 but greater than .90 (Hu & Bentler, 1999) and the RMSEA was greater than .05 but less than .08 (MacCallum et al., 1996). As seen in Figure 1, the standardized factor loadings, ranging from .64 to .88, and the estimated correlation coefficients among the four latent variables were large. A test of relative fit indicated that the four-factor model provided a significantly better fit than the one-factor model (Δχ²[6, N = 259] = 157.01, p < .001).

Undergraduate sample self-reports.

For self-reports of emotionally abusive behavior in the undergraduate sample, all CFA goodness-of-fit indices indicated that the four-factor measurement model fit the data well (normed-χ² =1.85, CFI = .96, TLI = .95, RMSEA = .04, SRMR = .06). As seen in Figure 2, the standardized factor loadings, ranging from .45 to .81, and the estimated correlation coefficients among the four latent variables were large. A test of relative fit indicated that the four-factor model provided a significantly better fit than the alternative, one-factor model (Δχ²[6, N = 190] = 52.05, p < .001).

Figure 2.
Note. Self-report (in bold) model fit: normed-c² =1.85, CFI = .96, TLI = .95, RMSEA = .04, SRMR = .06. Partner report model fit: normed-c² =1.82, CFI = .93, TLI = .92, RMSEA = .07, SRMR = .06.

Undergraduate sample partner reports.

For undergraduates’ reports of their partner’s emotionally abusive behavior, two CFA goodness-of-fit indices indicated that the four-factor measurement model fit the data well (normed-χ² =1.82, SRMR = .06), whereas three indices (CFI = .93, TLI = .92, RMSEA = .07) showed the model fit was fair (Hu & Bentler, 1999; MacCallum et al., 1996). As seen in Figure 2, the standardized factor loadings, ranging from .49 to .87, and the estimated correlation coefficients among the four latent variables were large. A test of relative fit indicated that the four-factor model provided a significantly better fit than the alternative, one-factor model (Δχ²[6, N = 190] = 82.41, p < .001).

Internal Consistency

Clinic sample.

Cronbach’s alpha for reports of emotionally abusive behavior by self and partner were, respectively, .78 and .80 for restrictive engulfment, .81 and .87 for denigration, .85 and .88 for hostile withdrawal, .82 and .84 for dominance/intimidation, and .91 and .92 for the MMEA total score. All of these internal consistency estimates were within the acceptable range and indicate good reliability for both self-reports and partner reports of emotional abuse.

Undergraduate sample.

Cronbach’s alpha for reports of emotionally abusive behavior by self and partner were, respectively, .67 and .75 for restrictive engulfment, .74 and .82 for denigration, .77 and .86 for hostile withdrawal, .73 and .80 for dominance/intimidation, and .87 and .91 for the MMEA total score. With the exception of self-reports of restrictive engulfment, these internal consistency estimates fall within the acceptable range and indicates adequate reliability for reports of emotional abuse by self and partner.

Concurrent Validity

Clinic sample.

As seen in Table 5, the upper bound estimate of the MMEA-SF total score correlation with the original, 28-item version approached unity (r = .97 for self-report and r = .99 for partner report). The observed correlations were also very high (all above .90) for each MMEA-SF subscale with its respective subscale from the 28-item version. The lower bound estimates, based on correlating items that were, and were not, included in the reduced versions (thus, have no overlapping measurement error), also provided good evidence of concurrent validity, with correlations above .90 for the total scale score, and between .64 and .87 for the subscales. Correlation estimates that were attenuated based on the reliability of the reduced scales fell in-between the lower and upper bound estimates. Conceptually, these values estimate the correlation value if the measurement errors from one “virtual” administration of the reduced item set were nonredundant with a second administration of the total item set. These values support evidence of strong concurrent validity between the two versions of the MMEA, with values above .90 for the total scale score and between .83 and .90 for the subscales.

Table 5.
Bivariate Correlations between the 16-item MMEA-SF and the 28-item MMEA.

Clinic Sample^{a b} Undergraduate Sample^c

Scale Attenuated Estimate Lower Bound Estimate Upper Bound Estimate Attenuated Estimate Lower Bound Estimate Upper Bound Estimate

Self Partner Self Partner Self Partner Self Partner Self Partner Self Partner

RE .83 .85 .74 .77 .94 .95 .74 .80 .72 .75 .90 .92

Den .85 .90 .64 .71 .94 .97 .80 .87 .64 .63 .93 .96

HW .89 .92 .80 .87 .96 .98 .81 .91 .71 .86 .92 .98

DI .88 .90 .83 .87 .97 .98 .84 .85 .79 .75 .98 .95

Total .93 .95 .91 .92 .98 .99 .90 .93 .89 .89 .96 .98

Note. ^aSelf-report: N = 424–466, ^bPartner report: N = 241–259, ^c N = 172–189

The upper-bound estimates are observed correlations between the reduced and total scale versions. The lower-bound estimates are correlations between the item sets that were, and were not, included on the reduced version. The attenuated estimates were created using the reverse operation for correction for attenuation, essentially adding estimated error of measurement for the reduced item version back into the observed correlation using the following formula:

$r_{x y (a t t e n u a t e d)} = (r_{x y (o b s e r v e d)}) (\sqrt{{\hat{ρ}}_{x x}})$ where ${\hat{ρ}}_{x x}$ is the reliability of the reduced item version of the scale.

All correlations were significant at p < .001.

There is also evidence of concurrent validity with the CTS psychological aggression subscale for self and partner reports in the clinic sample. As seen in Table 6, all four self-report MMEA-SF subscales had positive, significant correlations, in the medium to large range of magnitude, with psychological aggression. Moreover, the differences in magnitude of correlations with psychological aggression from the original 28-item version to the 16-item MMEA-SF were small for self-report data (.00 to .04) and partner report data (.00 to 03).

Table 6.
MMEA Pairwise Correlations with Physical Assault and Psychological Aggression.

Psychological Aggression Physical Assault

Clinic Sample ^{a b} Clinic Sample ^{c d} Undergraduate Sample^e

16-item Version Self Partner Self Partner Self Partner

Restrictive engulfment .35 .45 .20 .40 .50 .28

Denigration .44 .69 .27 .51 .44 .36

Hostile withdrawal .46 .56 .29 .43 .38 .57

Dominance/intimidation .49 .72 .36 .66 .46 .44

Total Score .54 .74 .36 .60 .59 .55

Original Version

Restrictive engulfment .39 .48 .19 .43 .50 .30

Denigration .46 .67 .28 .50 .50 .39

Hostile withdrawal .44 .56 .31 .43 .41 .58

Dominance/intimidation .49 .72 .37 .68 .52 .49

Total Score .55 .75 .36 .61 .59 .58

Note. ^aN = 450–455 for self-report, ^bN = 253–255 for partner report, ^cN = 448–453 for self-report, ^dN = 253–255 partner report, ^e N = 186–189

Physical assault was log-transformed. All correlations were significant at p < .001.

Undergraduate sample.

As seen in Table 5, the upper bound estimate of the MMEA-SF total score correlation with the original, 28-item version approached unity (r = .96 for self-report and r = .98 for partner report). The observed correlations were also very high (all above .90) for each MMEA-SF subscale with its respective subscale from the 28-item version. The lower bound estimates likewise provided evidence of strong concurrent validity, with correlations of .89 for the total scale score, and between .63 and .86 for the subscales. The correlation estimates that were attenuated to remove shared measurement error also reveal evidence of strong concurrent validity between the two versions of the MMEA, with values above .90 for the total scale score and between .74 and .91 for the subscales.

Convergent Validity

Clinic sample.

Evidence of convergent validity for the MMEA-SF as a measure of abusive behavior was found for both self- and partner reports. As displayed in Table 6, all four MMEA-SF subscales had positive, significant correlations with physical assault, which were in the small to medium range of magnitude for self-reports and in the medium to large range for partner reports. The item reduction from the original 28-item version to the 16-item MMEA-SF produced small differences in correlations with physical assault for self-report (.00 to .02) and partner report (.00 to .03).

Undergraduate sample.

Evidence of convergent validity for the MMEA-SF as a measure of abusive behavior was found for both self- and partner reports in the undergraduate sample. As displayed in Table 6, all four self-report and partner report MMEA-SF subscales had positive, significant correlations, in the medium to large range of magnitude, with physical assault. In addition, the decrement in these correlations associated with item reduction from the original 28-item version to the 16-item MMEA-SF were modest for self-report (.00 to .06) and partner report (.01 to .05).

Discussion

The original 28-item MMEA was developed for broad applicability across different populations, including common forms of emotional abuse and more coercive forms found in clinical samples. Although the initial item selection and psychometric work were performed with college dating samples, further research indicated good reliability and validity with clinic samples as well. The current findings indicate that the MMEA-SF is also a psychometrically sound, multidimensional measure of emotional abuse in both clinic and undergraduate samples.

As predicted, the four-factor model provided good overall fit to the data on emotionally abusive behavior for self-reports of men in treatment for IPV and undergraduates in dating relationships. For reports of partner behavior in both samples, the evidence of model fit was mixed, with some indices above and some indices below the accepted cutoffs for a good model fit. In addition, the four-factor model was clearly superior to a one-factor alternative in both the clinic and undergraduate samples for both self- and partner reports. Given that the model tests used strict criteria for model fit and did not include any modifications (such as correlated errors), that all fit indices either met the strict cutoffs or were close to those cutoffs with no evidence of poor model fit, and the consistent improvement in fit for four-factor versus one-factor models, these results overall provide good support for the presence of four correlated latent constructs representing the theoretically hypothesized forms of emotional abuse in both samples (Murphy & Cascardi, 1999). Therefore, although clinic samples report higher average levels of emotional abuse even with similar proportions of both samples endorsing at least one behavior on the MMEA, the same four forms of abuse that occur in clinic samples were also found in the college dating sample.

Results also provide evidence of reliability and validity for the MMEA-SF subscales and total score. Internal consistency was in the adequate to good range for all four subscales and was high for the total score across reporter (i.e., self and partner) and sample (i.e., clinic and undergraduate). Moreover, concurrent validity was evident in both samples in the very high correlations between the subscales and total score in the reduced (16-item) and original (28-item) versions. However, these upper-bound estimates include shared error variance for items common to the short- and long-scale versions. Alternative, lower bound estimates derived from correlating respective scales created from items that were, and were not, included on the reduced version, provided mixed evidence of concurrent validity across the four subscales (all rs ≥ .64 for clinic self-report, all rs ≥ .71 for clinic partner’s report, all rs ≥ .63 for undergraduate sample for both self-report and report of partner). Estimates that attenuated the observed correlations based on the reliability of the reduced scales demonstrated good concurrent validity with the original 28-item version, with somewhat stronger correlations observed for the clinic sample (all rs ≥ .83) than for the undergraduate sample (all rs ≥ .74).

Evidence of convergent validity as a measure of relationship abuse derives from correlations with physical IPV. Each MMEA-SF subscale was significantly correlated with physical assault, with estimates ranging from .20 to .66 for the clinic sample and from .28 to .57 for the undergraduate sample. The reduction from seven-item to four-item subscales produced very small decrements in observed correlations with physical IPV for the clinic sample (.03 or less), and modest decrements for the undergraduate sample (.06 or less), further indicating the potential utility of the MMEA short form for research and clinical applications. Strong correlations with an established measure of psychological aggression (the CTS2) in the clinic sample provide additional support for concurrent validity, and likewise demonstrate only small decrements in validity correlations from the 28-item version to the 16-item MMEA-SF.

Consistent with past findings for the 28-item version (Murphy et al., 1999), differential associations with physical assault were also apparent across the four MMEA-SF subscales. For both self- and partner reports in both samples, dominance/intimidation had the strongest associations with physical assault. These behaviors are the most topographically similar to physical violence, and are often intended to produce fear in the recipient. For partner reports in the clinic sample, denigration also had a strong correlation with physical assault relative to the other subscales. In general, restrictive engulfment had the lowest correlations with physical assault. These attempts to monitor and isolate the partner may reflect insecurity and controlling behaviors that less commonly escalate to physical violence. In addition, the MMEA-SF correlations with physical assault were generally higher across both samples for partner reported data than for self-reported data. This pattern may reflect disclosure biases and/or self-serving biases in self-reports of socially undesirable behaviors such as physical assault (Riggs et al., 1989).

Several limitations warrant consideration. Although the current samples are similar to those in previous clinical and descriptive studies that have used the MMEA, further research is needed to examine its psychometric properties in other contexts and populations, including women referred for treatment of IPV perpetration, couples receiving relationship therapy, and diverse nonclinical community samples. Second, partner reported data were available for only about half of the male IPV perpetrators in the current sample. It is possible that data from partners who were not reached during program intake would reveal different patterns or results. Third, it is important to point out that the data available from the two samples, although reflecting common uses of the measure, are not directly parallel. In the clinic sample, both reporters provided data on the abusive client’s behavior and reported on the frequency of behaviors over the past six months. In the undergraduate sample, each respondent reported on both their own and their partner’s behavior, and reported on the frequency of behaviors over the past 12 months. Although the consistency in measure performance across these different reporting patterns supports the validity of the MMEA-SF, it remains possible that different findings would emerge if those in the clinic sample were asked to report on behavior by both self and partner. Fourth, the current study relied on cross-sectional data. This limits our ability to determine whether emotional abuse has directional and/or causal associations with physical IPV, and whether the MMEA-SF will efficiently measure changes in abusive behavior over time relative to the full scale. Finally, given that all of the items are shared by both versions of the scale, measurement errors of the shared items are perfectly correlated, which leads to an inflated estimate of the true correlation between the two versions of the measure (Raykov et al., 2015). We addressed this conceptual problem by including lower-bound estimates that removed all shared error by correlating the items retained in the MMEA-SF and the remaining (non-selected items) from the original MMEA, and attenuating estimates for reliability. All three methods yielded evidence for the concurrent validity of the short form relative to the original 28-item version.

The current results indicate that the MMEA-SF can reduce assessment burden while maintaining good domain coverage for the subscale content and broad applicability of item content. As with the original MMEA, the MMEA-SF can be useful in understanding relatively common or normative forms of emotional abuse, as in college dating relationships, as well as patterns of coercive control in clinically violent samples. In summary, the MMEA-SF, a 16-item shortened version of the original 28-item MMEA, has good psychometric properties and produces data that are highly comparable to the original 28-item version for both clinically violent and dating populations. On the basis of these findings, we recommend that researchers and practitioners who are concerned about potential assessment burden or time of administration consider using this new 16-item version of the scale.

Footnotes

Authors’ Note

Julian Farzan-Kashani is now at Octave, San Francisco, CA, USA and Jennifer M. Lorenzo is now at the Veterans Affairs Maryland Healthcare System.

Declaration of Conflicting Interests

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

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

ORCID iDs

Ana I. Maldonado

Christopher M. Murphy

Author Biographies

Ana I. Maldonado, MA, is clinical psychology graduate student at the University of Maryland, Baltimore County. Her research focuses on identifying risk and protective factors, and mechanisms of intimate partner violence perpetration in communities of color. She is interested in using this research to inform culturally-tailored prevention and intervention programs for intimate partner violence.

Julian Farzan-Kashani, PhD, is a licensed psychologist at Octave in San Francisco. They serve clients facing trauma, anxiety, panic attacks, phobias, and insomnia. They are passionate about helping to empower the marginalized. Their research focuses on angry rumination, mindfulness, intimate partner violence, and emotional abuse.

Shuyan Sun, PhD, is an associate professor in the Department of Psychology at the University of Maryland, Baltimore County. Her research focuses on quantitative methods and their applications in psychological and educational research.

Steven C. Pitts, PhD, is an associate professor of Psychology at the University of Maryland, Baltimore County. He is a quantitative psychologist with interests in the application and use of emerging longitudinal data analysis (latent and individual growth curve modeling), and structural equation modeling.

Jennifer M. Lorenzo, PhD, is a licensed clinical psychologist in the Psychosocial Rehabilitation and Recovery Center at the Veterans Affairs Maryland Healthcare System (VAMHCS). Her clinical work embraces the recovery model to support veterans with serious mental illness. Her research focuses broadly on intimate relationships. She also supports healthy veteran relationships through the Family Intervention Team at the VAMHCS.

Robin A. Barry, PhD, is a clinical psychologist and assistant professor of Psychology at the University of Wisconsin. Her research focuses primarily on couple and family relationships. She has recently extended her work to study and serve family caregivers of people with chronic illnesses and people involved in the criminal justice system.

Christopher M. Murphy, PhD, is professor of psychology at the University of Maryland, Baltimore County. His research focuses on the efficacy of interventions for individuals who engage in intimate partner violence, the identification of personal, contextual, and intervention factors that facilitate and inhibit treatment response in this population, and the prevention of abuse and violence in emerging adulthood.

References

Armstrong

T. G.

, Wernke

J. Y.

, Medina

K. L.

, & Schafer

(2002). Do partners agree about the occurrence of intimate partner violence? A review of the current literature. Trauma, Violence, & Abuse , 3(3), 181–193. https://doi.org/10.1177/15248380020033002

Bollen

K. A.

(1989). Structural equations with latent variables. Wiley.

Bonechi

, & Tani

(2011). Italian adaptation of the Multidimensional Measure of Emotional Abuse (MMEA). TPM-Testing, Psychometrics, Methodology in Applied Psychology , 18(2), 65–86.

Browne

M. W.

, & Cudeck

(1993). Alternative ways of assessing model fit. In Bollen

K. A.

& Long

J. S.

(Eds.), Testing structural equation models (pp. 136–162). Sage Publications.

Cohen

, Cohen

, West

S. G.

, & Aiken

L. S.

(2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Lawrence Erlbaum Associates Publishers.

Falconier

M. K.

(2010). Female anxiety and male depression: Links between economic strain and psychological aggression in Argentinean couples. Family Relations , 59(4), 424–438. https://doi.org/10.1111/j.1741-3729.2010.00613.x

Follingstad

D. R.

(2007). Rethinking current approaches to psychological abuse: Conceptual and methodological issues. Aggression and Violent Behavior , 12(4), 439–458. https://doi.org/10.1016/j.avb.2006.07.004

Follingstad

D. R.

, Rutledge

L. L.

, Berg

B. J.

, Hause

E. S.

, & Polek

D. S.

(1990). The role of emotional abuse in physically abusive relationships. Journal of Family Violence , 5, 107–120. https://doi.org/10.1007/BF00978514

Gurtman

M. B.

(1992). Construct validity of interpersonal personality measures: The interpersonal circumplex as a nomological net. Journal of Personality and Social Psychology , 63(1), 105–118. https://doi.org/10.1037/0022-3514.63.1.105

10.

L. T.

, & Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal , 6(1), 1–55. https://doi.org/10.1080/10705519909540118

11.

Huber

P. J.

(1967). The behavior of maximum likelihood estimates under nonstandard conditions. In Le Cam

L. M.

, Neyman

(Eds.), Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability (Vol. 1; pp. 221–233). University of California Press.

12.

Lawrence

, Yoon

, Langer

, & Ro

(2009). Is psychological aggression as detrimental as physical aggression? The independent effects of psychological aggression on depression and anxiety symptoms. Violence and Victims , 24(1), 20–35. https://doi.org/10.1891/0886-6708.24.1.20

13.

Lorenzo

J. A.

(2019). Skills for healthy adult relationships (SHARe@UMBC): Interest, risk factors, and dissemination. [Unpublished doctoral dissertation]. University of Maryland.

14.

MacCallum

R. C.

, Browne

M. W.

, & Sugawara

H. M.

(1996). Power analysis and determination of sample size for covariance structure modeling. Psychological Methods , 1(2), 130–149. https://doi.org/10.1037/1082-989X.1.2.130

15.

Murphy

C. M.

, & Cascardi

(1999). Psychological abuse in marriage and dating relationships. In Hampton

R. L.

(Ed.), Family violence prevention and treatment (2nd ed., pp. 198–226). Sage Publications.

16.

Murphy

C. M.

, & Hoover

S. A.

(1999). Measuring emotional abuse in dating relationships as a multifactorial construct. Violence and Victims , 14(1), 39–53. https://doi.org/10.1891/0886-6708.14.1.39

17.

Murphy

C. M.

, & O’Leary

K. D.

(1989). Psychological aggression predicts physical aggression in early marriage. Journal of Consulting and Clinical Psychology , 57(5), 579–582. https://doi.org/10.1037/0022-006X.57.5.579

18.

Murphy

C. M.

, Hoover

S. A.

, & Taft

(1999, November). The Multidimensional Measure of Emotional Abuse: Factor structure and subscale validity [Paper presentation]. Association for the Advancement of Behavior Therapy, Toronto.

19.

Norwood

, & Murphy

(2012). What forms of abuse correlate with PTSD symptoms in partners of men being treated for intimate partner violence? Psychological Trauma: Theory, Research, Practice, and Policy , 4(6), 596–604. https://doi.org/10.1037/a0025232

20.

O’Leary

(1999). Psychological abuse: A variable deserving critical attention in domestic violence. Violence and Victims , 14(1), 3–23. https://doi.org/10.1891/0886-6708.14.1.3

21.

R Core Team (2017). The R project for statistical computing . R Foundation for Statistical Computing. https://www.R-project.org/

22.

Raykov

, Marcoulides

G. A.

, & Patelis

(2015). The importance of the assumption of uncorrelated errors in psychometric theory. Educational and Psychological Measurement , 75(4), 634–647. https://doi.org/10.1177/0013164414548217

23.

Riggs

D. S.

, Murphy

C. M.

, & O’Leary

K. D.

(1989). Intentional falsification in reports of interpartner aggression. Journal of Interpersonal Violence , 4(2), 220–232. https://doi.org/10.1177/088626089004002006

24.

Rosseel

(2012). Lavaan: An R package for structural equation modeling. Journal of Statistical Software , 48(2), 1–36. http://www.jstatsoft.org/v48/i02/

25.

Semiatin

J. N.

, Torres

, LaMotte

A. D.

, Portnoy

G. A.

, & Murphy

C. M.

(2017). Trauma exposure, PTSD symptoms, and presenting clinical problems among male perpetrators of intimate partner violence. Psychology of Violence , 7(1), 91–100. https://doi.org/10.1037/vio0000041

26.

Smith-Marek

E. N.

, Cafferky

, Dominguez

M. M.

, Spencer

, Van

, Stith

S. M.

, & Oliver

M. A.

(2016). Military/civilian risk markers for physical intimate partner violence: A meta-analysis. Violence and Victims , 31(5), 787–818. https://doi.org/10.1891/0886-6708.VV-D-15-00032

27.

Spearman

, (1904) The proof and measurement of association between two things. The American Journal of Psychology , 15(1), 72–101. https://doi.org/10.2307/1412159

28.

Stark

(2007). Coercive control: How men entrap women in personal life. Oxford.

29.

Straus

M. A.

(1979). Measuring intrafamily conflict and violence: The Conflict Tactics (CT) Scales. Journal of Marriage and the Family , 41(1), 75–88. https://doi.org/10.2307/351733

30.

Straus

M. A.

, Hamby

S. L.

, Boney-McCoy

, & Sugarman

D. B.

(1996). The revised Conflict Tactics Scales (CTS2): Development and preliminary psychometric data. Journal of Family Issues , 17(3), 283–316. https://doi.org/10.1177/019251396017003001

31.

Sullivan

T. P.

, McPartland

T. S.

, Armeli

, Jaquier

, & Tennen

(2012). Is it the exception or the rule? Daily co-occurrence of physical, sexual, and psychological partner violence in a 90-day study of substance-using, community women. Psychology of Violence , 2(2), 154–164. https://doi.org/10.1037/a0027106

32.

Taft

C. T.

, Macdonald

, Creech

S. K.

, Monson

C. M.

, & Murphy

C. M.

(2016). A randomized controlled clinical trial of the strength at home men’s program for partner violence in military veterans. Journal of Clinical Psychiatry , 77(9), 1168–1175. https://doi.org/10.4088/JCP.15m10020

33.

Tolman

R. M.

(1989). The development of a measure of psychological maltreatment of women by their male partners. Violence and Victims , 4(3), 159–177. https://doi.org/10.1891/0886-6708.4.3.159

34.

Toplu-Demirtaş

, Hatipoğlu-Sümer

, & Murphy

C. M.

(2018). Turkish version of the Multidimensional Measure of Emotional Abuse: Preliminary psychometrics in college students. Violence and Victims , 33(2), 275–295. https://doi.org10.1891/0886-6708.v33.i2.087

35.

Vega

E. M.

, & O’Leary

K. D.

(2007). Test-retest reliability of the Revised Conflict Tactics Scales (CTS2). Journal of Family Violence , 22, 703–708. https://doi.org/10.1007/s10896-007-9118-7

36.

White

(1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica , 48(4), 817–830.

37.

Yuan

K.-H.

, & Bentler

P. M.

(2000). Three likelihood-based methods for mean and covariance structure analysis with nonnormal missing data. Sociological Methodology , 30(1), 165–200.

Item	Description	In 16-item Version
Restrictive Engulfment
1	Asked the other person where they had been or who they were with in a suspicious manner
2	Secretly searched through the other person’s belongings
3	Tried to stop the other person from seeing certain friends or family members	✓
4	Complained that the other person spends too much time with friends
5	Got angry because the other person went somewhere without telling him/her	✓
6	Tried to make the other person feel guilty for not spending enough time together	✓
7	Checked up on the other person by asking friends or relatives where they were or who they were with	✓
Denigration
8	Said or implied that the other person was stupid	✓
9	Called the other person worthless	✓
10	Called the other person ugly
11	Criticized the other person’s appearance
12	Called the other person a loser, failure, or similar term	✓
13	Belittled the other person in front of other people	✓
14	Said that someone else would be a better partner (better spouse, better girlfriend or boyfriend)
Hostile withdrawal
15	Became so angry that they were unable or unwilling to talk	✓
16	Acted cold or distant when angry	✓
17	Refused to have any discussion of a problem
18	Changed the subject on purpose when the other person was trying to discuss a problem
19	Refused to acknowledge a problem that the other person felt was important	✓
20	Sulked or refused to talk about an issue	✓
21	Intentionally avoided the other person during a conflict or disagreement
Dominance/intimidation
22	Became angry enough to frighten the other person	✓
23	Put his/her face right in front of the other person’s face to make a point more forcefully
24	Threatened to hit the other person	✓
25	Threatened to throw something at the other person
26	Threw, smashed, hit, or kicked something in front of the other person	✓
27	Drove recklessly to frighten the other person
28	Stood or hovered over the other person during a conflict or disagreement	✓

Item	3	5	6	7	8	9	12	13	15	16	19	20	22	24	26	28	M	SD
3		.42	.43	.48	.33	.35	.29	.41	.23	.19	.28	.19	.35	.24	.38	.30	0.96	3.62
5	.54		.49	.56	.46	.32	.37	.43	.36	.36	.42	.36	.39	.33	.40	.44	1.96	4.41
6	.38	.56		.52	.34	.31	.36	.36	.25	.25	.35	.28	.30	.26	.29	.28	1.71	4.66
7	.52	.62	.36		.34	.33	.38	.43	.25	.28	.36	.30	.41	.36	.38	.40	1.42	4.28
8	.39	.39	.34	.36		.60	.50	.47	.44	.45	.41	.37	.41	.40	.41	.36	2.19	5.01
9	.33	.39	.33	.38	.68		.58	.42	.39	.36	.32	.26	.37	.32	.39	.32	1.05	3.45
12	.28	.36	.28	.39	.64	.65		.49	.35	.34	.40	.41	.37	.25	.39	.34	1.04	3.92
13	.41	.41	.37	.39	.60	.55	.61		.36	.31	.39	.27	.45	.37	.43	.46	1.03	3.49
15	.19	.29	.24	.15	.44	.36	.33	.35		.67	.56	.54	.48	.32	.43	.38	3.91	7.03
16	.25	.34	.32	.17	.40	.28	.28	.33	.75		.54	.53	.49	.34	.40	.39	5.09	7.78
19	.34	.35	.31	.35	.50	.43	.42	.51	.58	.63		.67	.50	.34	.49	.45	2.62	5.06
20	.25	.35	.26	.28	.38	.41	.40	.41	.55	.62	.77		.46	.31	.41	.35	3.07	5.97
22	.45	.42	.43	.42	.58	.54	.47	.55	.53	.48	.59	.44		.48	.57	.62	2.39	5.40
24	.38	.32	.25	.37	.46	.48	.51	.55	.39	.32	.40	.30	.53		.49	.50	0.86	3.12
26	.34	.26	.29	.24	.41	.43	.37	.44	.33	.41	.50	.38	.57	.47		.64	1.32	3.60
28	.30	.43	.38	.35	.50	.54	.49	.53	.38	.37	.47	.38	.64	.60	.53		1.28	3.95
M	2.98	2.63	4.00	2.79	5.47	3.81	2.74	3.69	5.43	6.77	7.94	6.32	5.54	2.49	2.48	3.35
SD	6.70	5.89	7.92	6.52	9.05	8.17	6.83	7.37	7.74	8.66	9.65	8.89	8.35	5.93	4.98	6.19

Item	3	5	6	7	8	9	12	13	15	16	19	20	22	24	26	28	M	SD
3		.34	.50	.21	.39	.34	.24	.31	.25	.23	.31	.17	.15	.25	.29	.42	0.65	2.44
5	.54		.36	.36	.47	.49	.41	.30	.42	.45	.36	.29	.40	.23	.29	.40	0.93	2.42
6	.52	.51		.29	.37	.32	.39	.41	.31	.30	.31	.27	.17	.19	.31	.34	1.64	4.10
7	.26	.52	.32		.13	.27	.16	.19	.13	.25	.20	.27	.32	.24	.19	.42	0.52	1.37
8	.42	.37	.42	.23		.54	.56	.42	.49	.42	.36	.28	.28	.27	.34	.27	1.51	3.44
9	.36	.19	.18	.14	.52		.59	.27	.44	.33	.26	.27	.27	.24	.28	.32	0.38	1.38
12	.38	.28	.32	.27	.61	.75		.37	.37	.29	.31	.30	.12	.15	.19	.20	0.30	1.19
13	.37	.30	.29	.13	.51	.65	.60		.30	.19	.43	.18	.26	.29	.32	.13	0.29	0.85
15	.43	.33	.36	.20	.44	.42	.43	.55		.68	.44	.35	.29	.17	.21	.13	1.86	3.91
16	.44	.38	.37	.17	.43	.32	.35	.46	.67		.37	.45	.27	.23	.24	.14	2.60	4.12
19	.40	.33	.27	.11	.43	.41	.38	.46	.60	.53		.40	.28	.21	.18	.22	0.60	1.25
20	.38	.37	.34	.23	.37	.29	.34	.45	.58	.70	.62		.23	.34	.15	.27	1.12	2.70
22	.39	.30	.32	.13	.38	.51	.48	.56	.45	.42	.41	.34		.33	.42	.45	0.31	0.99
24	.40	.24	.22	.20	.52	.52	.68	.51	.33	.32	.27	.27	.54		.52	.30	0.31	1.41
26	.41	.37	.33	.19	.34	.42	.62	.42	.43	.41	.39	.36	.57	.46		.41	0.42	1.61
28	.42	.28	.36	.34	.44	.64	.62	.58	.34	.31	.27	.29	.52	.53	.54		0.31	2.27
M	0.65	0.93	1.64	0.52	1.51	0.38	0.30	0.29	1.86	2.60	0.60	1.12	0.31	0.31	0.42	0.31
SD	2.44	2.42	4.10	1.37	3.44	1.38	1.19	0.85	3.91	4.12	1.25	2.70	0.99	1.41	1.61	2.27

	Clinic Sample				Undergraduate Sample
	Self-Report^a		Partner Report^b		Self-Report^c		Partner Report^d
	M	SD	M	SD	M	SD	M	SD
16-Item MMEA-Short Form
Restrictive engulfment	6.24	13.27	12.64	21.59	3.64	7.18	3.69	7.30
Denigration	5.60	13.13	16.26	27.68	2.37	5.24	3.01	10.74
Hostile withdrawal	14.76	21.38	26.67	30.83	6.22	9.23	9.12	16.46
Dominance/intimidation	5.89	13.00	14.02	20.88	1.48	4.90	1.81	6.03
Total	32.27	47.90	69.43	82.43	13.69	19.70	17.64	32.57
28-Item Original MMEA
Restrictive engulfment	12.04	22.50	24.87	37.18	8.74	16.09	8.32	15.91
Denigration	8.13	18.45	22.93	40.05	3.70	7.69	4.42	13.03
Hostile withdrawal	24.22	34.36	45.22	53.47	9.60	15.33	14.32	26.84
Dominance/intimidation	8.95	20.04	20.82	31.66	2.06	6.27	2.78	9.39
Total	53.24	75.46	113.67	132.41	24.16	36.11	29.96	51.46

	Clinic Sample^{a b}							Undergraduate Sample^c
Scale	Attenuated Estimate		Lower Bound Estimate		Upper Bound Estimate		Attenuated Estimate		Lower Bound Estimate		Upper Bound Estimate
	Self	Partner	Self	Partner	Self	Partner	Self	Partner	Self	Partner	Self	Partner
RE	.83	.85	.74	.77	.94	.95	.74	.80	.72	.75	.90	.92
Den	.85	.90	.64	.71	.94	.97	.80	.87	.64	.63	.93	.96
HW	.89	.92	.80	.87	.96	.98	.81	.91	.71	.86	.92	.98
DI	.88	.90	.83	.87	.97	.98	.84	.85	.79	.75	.98	.95
Total	.93	.95	.91	.92	.98	.99	.90	.93	.89	.89	.96	.98

Psychological Aggression				Physical Assault
Clinic Sample ^{a b}			Clinic Sample ^{c d}		Undergraduate Sample^e
16-item Version	Self	Partner	Self	Partner	Self	Partner
Restrictive engulfment	.35	.45	.20	.40	.50	.28
Denigration	.44	.69	.27	.51	.44	.36
Hostile withdrawal	.46	.56	.29	.43	.38	.57
Dominance/intimidation	.49	.72	.36	.66	.46	.44
Total Score	.54	.74	.36	.60	.59	.55
Original Version
Restrictive engulfment	.39	.48	.19	.43	.50	.30
Denigration	.46	.67	.28	.50	.50	.39
Hostile withdrawal	.44	.56	.31	.43	.41	.58
Dominance/intimidation	.49	.72	.37	.68	.52	.49
Total Score	.55	.75	.36	.61	.59	.58