Validation of a Measure of Psychological Aggression in Same-Sex Couples: Descriptive Data on Perpetration and Victimization and Their Association with Physical Violence

Abstract

Despite its documented elevated prevalence, psychological aggression in couple relationships rarely receives the same degree of attention as physical aggression. Indeed, psychological violence is much more prevalent than physical violence in couples, and its impact can be just as devastating. Research has recently begun to address psychological aggression in same-sex couples, but the most commonly used questionnaire for assessing intimate partner violence has yet to be fully validated with this population. The two objectives of the present study are (1) to present the first data on the psychometric properties of the psychological aggression scale of the Revised Conflict Tactics Scales (CST2; Straus, Hamby, Boney-McCoy, & Sugarman, 1996) with individuals in same-sex couple relationships, and (2) to examine the relationship between psychological aggression and physical violence in this population. Two hundred eighteen individuals (75 men and 143 women) participated in this study. Results support the factor structure, reliability, and concurrent validity of the psychological aggression scale. Rates of psychological and physical violence are reported, and results indicate that psychological aggression is strongly correlated with physical violence in same-sex couples. Gender differences are highlighted and the importance of developing valid and reliable instruments to measure this construct is emphasized.

KEYWORDS:

• psychological aggression
• psychometric properties
• same-sex couples
• gay men
• lesbians
• intimate partner violence

INTRODUCTION

Intimate psychological partner violence is a problem in our society for both heterosexual and same-sex couples. Definitions of this phenomenon vary, but the majority includes denigration, verbal aggression, and dominant, jealous, and controlling behaviors (O’Leary & Maiuro, 2001; Straus, Hamby, Boney-McCoy, & Sugarman, 1996). Approximately 75% of women and men in the general community perpetrate psychological violence against their partner; this statistic is stable across sexual orientation (see Jose & O’Leary, 2009, for a review). Indeed, intimate psychological violence is much more prevalent than physical intimate violence and has been demonstrated to be a key predictor of future physical abuse (O’Leary, Smith Slep, & O’Leary, 2007; White, Merrill, & Koss, 2001). Intimate psychological violence is a hallmark of most physically abusive relationships (Ehrensaft, 2009) and its impact can be just as devastating as that of physical violence (Marshall, 1992; Walker, 1984). Despite these findings, the majority of the research in this area focuses strictly on physical violence in heterosexual and same-sex couples. Moreover, whereas the body of empirical research on intimate partner violence (IPV) between heterosexual partners has grown steadily since the 1970s, comparable research on same-sex couples only recently became a subject of interest (McClennen, 2005). There is an urgent need to collect data about intimate psychological violence in same-sex couples, but this objective cannot be achieved before a suitable measure of this construct is validated with this population. The goal of the present study is to examine the psychometric properties of the psychological aggression scale of the widely used Revised Conflict Tactics Scales (CTS2; Straus et al., 1996). In addition, data on the prevalence of both psychological and physical perpetrated and received intimate partner violence are presented, and the relationship between psychological and physical intimate partner violence is explored.

Validation of a Measure of Intimate Psychological Aggression with Same-Sex Couples

The majority of research on IPV in same-sex couples adopted measures that were validated on heterosexual couples. Jose and O’Leary (2009) highlighted the importance of testing measures of IPV in same sex-couples. They asserted that research on factors related to aggression in same-sex couple relationships is essential, particularly since the factors may be distinct from those associated with violence in heterosexual relationships. To facilitate valid and reliable measurement of intimate psychological partner violence in same-sex couples, the present study will be the first to provide validation data on the psychometric properties of the CTS2 psychological aggression scale (Straus et al., 1996) in a sample of individuals in same-sex relationships.

The CTS2 and its earlier version, the Conflict Tactics Scales (CTS; Straus, 1979), have become the most widely used survey measures of intimate partner abuse. The CTS2 is easy to administer and yields information about frequency as well as presence of violence. The popularity of the CTS2 is also a function of the wealth of psychometric data now available for these scales (Straus, Hamby, & Warren, 2003). The CTS2 was developed in response to criticism of the original version. The newer version constituted a great improvement, and included new items that measured multiple facets of intimate partner violence. Physical injury and sexual coercion scales were added, and the distinction between items that measured minor acts of violence and items that measured severe acts was clarified (Straus et al., 2003). The CTS2 is a 78-item self-report measure that requires both members of the couple to report on their own behavior and their partner's behavior on five dimensions of conflict tactics: negotiation, psychological aggression, physical assault, injury, and sexual coercion (Straus et al., 1996). Each of the five scales is further divided into two subscales; the negotiation scale is divided into emotional and cognitive items and the psychological aggression, physical assault, injury, and sexual coercion scales are divided into minor and severe acts. The CTS2 is applicable in clinical settings, on representative samples, and on specific groups in nonclinical, clinical, or correctional settings (Straus et al., 2003). The wide and diverse use of the CTS2 is justified by findings that demonstrate the following: (1) satisfactory construct validity with anger management, dominance, gender hostility, conflict, communication problems, and antisocial personality, among others; (2) satisfactory test-retest reliability; and (3) satisfactory internal consistencies between the five scales, ranging from .79 to .95 (Duggan et al., 1999; Lucente, Fals-Stewart, Richards, & Goscha, 2001; Newton, Connelly, & Landsverk, 2001; Straus et al., 1996; Straus & Mouradian, 1999; Vega & O’Leary, 2007). The psychometric properties of the CTS2 have been established in a wide variety of samples, including multiethnic students (Straus, 2004) and male batterers (Vega & O’Leary, 2007). However, the factor structure of the CTS2 has been demonstrated to vary according to gender (Schafer, 1996) and according to level of aggression (Jones, Ji, Beck, & Beck, 2002). Further validation in specific populations is therefore recommended (Straus, 1993).

The CTS2 has yet to be fully validated with individuals in same-sex couple relationships and, to our knowledge, only one published study has worked on the validation of this measure (Regan, Bartholomew, Oram, & Landolt, 2002). The authors conducted item-response theory analyses with the perpetration and victimization items from an extended version of the CTS2 physical assault scale (with minor changes in wording) with a sample of gay and bisexual men. The results indicated that a one-factor structure was preferable to a two-factor structure that discriminates between items measuring minor and severe acts and confirmed that the physical assault scale items assess violent behaviors on a wide range of severity. A full validation of the CTS2 with both men and women in same-sex relationships remains necessary. The present study focuses on the validation of the psychological aggression scale with individuals in same-sex couple relationships.

Prevalence of Psychological Intimate Violence in Gay and Lesbian Couples

Existing data suggest that psychological aggression in the gay, lesbian, bisexual, transgender, and queer (GLBTQ) communities is irrefutably present and prevalent. A study that used the CTS to examine verbal abuse in Southeastern American lesbian couples revealed that 90% of respondents had experienced verbal abuse from their partner during the past year (Lockhart, White, Causby, & Isaac, 1994). The results of that study were congruent with results from the Lesbian Battering Intervention Project Survey in Minnesota, which reported that 76% of lesbians had experienced some form of threat from a same-sex partner, clearly illustrating the prevalence of psychological violence in lesbian couples (Eliott, 1996). Similarly, a study conducted with gay men in Western Canada revealed that all but two of the participants (N = 63) reported instances of having perpetrated emotional abuse including yelling, using harsh language, or destroying objects. That study used a coding system for evaluating psychological violence developed by the authors (Stanley, Bartholomew, Taylor, Oram, & Landolt, 2006). To our knowledge, gender differences in intimate psychological partner violence in same-sex couple relationships has yet to be studied.

The Relationship between Psychological Aggression and Intimate Physical Violence

It can be difficult to isolate psychological aggression from physical violence; the two constructs have similar correlates, precursors, and etiology (see O’Leary & Woodin, 2009, for a review). Recent research has confirmed the significant role of psychological aggression in predicting physical abuse in heterosexual intimate relationships (O’Leary et al., 2007; White et al., 2001) and male same-sex relationships (Stanley et al., 2006). More precisely, O’Leary and colleagues (2007) found a strong correlation between perpetrated psychological aggression and physical assault scores on the CTS2 (.57 for men and .65 for women). Furthermore, a study that used the CTS2 to identify typologies of violent women found that women who were very physically violent perpetrated more psychological abuse and reported a wider variety and a greater number of instrumental motives for the use of violence than women who were less physically violent (Babcock, Miller, & Siard, 2003).

The general consensus among researchers in this area is that controlling behavior is the cornerstone of intimate psychological partner violence that occurs in the context of intimate physical partner violence. This postulate was tested in a study that compared the degree of control reported in relationships in distress where aggression was present and relationships in distress where aggression was not present. The results confirmed that physical violence is associated with perceived controlling behaviors (a form of psychological aggression) (Ehrensaft, Langhinrichsen-Rohling, Heyman, O’Leary, & Lawrence, 1999).

A final study coded perpetrated and received psychologically and physically aggressive behaviors in 69 gay and bisexual men and examined the relationship between the two types of aggression. The results revealed that an increase in one person's psychological abuse toward his partner coincided with the escalation of the partner's perpetrated psychological abuse. Physically abusive men were found to also be psychologically abusive, and the severity of the psychological abuse reflected the severity of the physical abuse (Stanley et al., 2006). These results indicate that the adoption of a complex and overarching perspective could facilitate a greater understanding of the patterns of violence in IPV. Such an approach would provide a means for predicting and preventing further forms of violence in both heterosexual and same-sex couple relationships. The present study is designed to expand the scope of exploration by (1) examining the relationship between psychological aggression and physical assault, and (2) completing an initial validation of the CTS2 psychological aggression scale with same-sex couples.

Hypotheses

The primary objective of the present study is to confirm the psychometric properties (factor structure, reliability, and concurrent validity) of the Revised Conflict Tactics Scales psychological aggression scale (CTS2; Straus et al., 1996) with individuals in same-sex relationships. We hypothesize that the psychological aggression scale will retain a one-factor structure with both men and women, although previous findings suggest that gender differences may be present in the regression weights or factor loadings (Schafer, 1996) and may yield moderate alpha coefficients (see Straus, Hamby, & Warren, 2003, for a review of previous findings). Regarding concurrent validity, we hypothesize that both attachment anxiety and avoidance will be moderately correlated with psychological aggression perpetration and victimization scales for both women and men. Previous findings have indicated that insecurely attached individuals are more likely to use violence in their relationships as a tool to regain emotional proximity (attachment anxiety) or to avoid becoming too close (avoidance of intimacy) (Bartholomew & Allison, 2006; Gosselin, Lafontaine, & Bélanger, 2005). Based on existing literature that suggests that couple violence is associated with decreased relationship satisfaction (O’Leary et al., 2007), we further hypothesize that relationship satisfaction will be moderately and negatively correlated with the psychological aggression perpetration and victimization scales for both genders.

The second objective of this study is to establish the prevalence of psychological aggression in the sample and to explore the relationship between psychological aggression and physical violence. We hypothesize that the rate of violence in our sample will be comparable to rates found in nonclinical samples of same-sex and heterosexual couples as reported above. We do not expect to find significant gender differences. We expect both perpetrated and received psychological aggression to be significantly correlated with perpetrated and received physical violence in both men and women. Finally, we expect a strong correlation between men's and women's reports of perpetration and victimization. This hypothesis is based on previous research that reported a strong correlation between genders (.85, p < .001 for psychological aggression) (Hines & Saudino, 2004) and on the assumption that much of the violence found in nonclinical samples is mutual (Johnson, 2001).

METHOD

Participants

A sample of 218 individuals (75 males and 143 females) involved in same-sex relationships participated in the study. Eligibility criteria for the study included the following: (1) at least 18 years old; (2) English-speaking or a good knowledge of English; and (3) minimum duration of 12 months in the current same-sex couple relationship. The male sample was composed of 75 men; 71 (94.7%) identified as gay and 4 (5.3%) identified as bisexual. For men, mean age was 36 years (SD = 13.1), and mean length of the current relationship was 5.5 years (SD = 5.3). Eighty-four percent of men were Caucasian, 4% were of Asian descent, 5.3% were Aboriginal, and 6.7% endorsed another ethnic background. The majority of the participants had completed college or university (81.3%), while a minority had completed a high school diploma or less (18.7%).

The female sample was composed of 143 women; 111 (77.6%) identified as lesbian and 32 (22.4%) identified as bisexual. For women, mean age was 32 years (SD = 10.3) and mean length of the current relationship was 4.5 years (SD = 4.7). The majority of the females in the sample were Caucasian (98.4%), and 1.4% were Aboriginal. The majority of the participants reported having completed college or university (78%); a minority had completed a high school diploma or less (22%).

Procedure

Participants were recruited through flyers and posters posted on the University of Ottawa campus and in local organizations serving the GLBTQ community or distributed to individuals at the annual Ottawa Pride parade. In addition, an e-mail describing the study was forwarded to the mailing lists of several GLBTQ associations. Individuals who were interested in participating in the study contacted us by phone or by e-mail, and eligible participants were sent a paper questionnaire or a link to the online survey. The questionnaire and information package included details about the purpose of the study, the content of the questionnaire, and the confidentiality and anonymity measures in place to protect the participants’ data. The returned questionnaires implied consent to participate in the study. As compensation, all participants were entered into a drawing for a gift certificate for a local music store.

Measures

The Revised Conflict Tactics Scales (CTS2; Straus et al., 1996) measure the extent to which an individual engaged in specific tactics, including negotiation, psychological aggression, physical assault, and sexual coercion, in the past 12 months and over his or her lifetime. The CTS2 also include a scale that measures physical injury resulting from physical intimate violence. Using an 8-point scale, participants indicated the frequency with which they had employed each tactic, and the frequency with which their partners had employed each tactic. Possible responses were “none,” “once,” “twice,” “3–5 times,” “6–10 times,” “11–20 times,” “21 or more times,” and “not in the past year, but it did happen before.” These categories were coded as approximate midpoints of 0, 1, 2, 4, 8, 15, and 25. To ensure that the results reflected annual prevalence of violent behavior, the response “not in the past year, but it did happen before” was scored 0. The present paper included only data on psychological and physical violence. Using the 7-point scale, total scores for the psychological violence scale ranged from 0 to 200 and total scores for the physical violence scale ranged from 0 to 300, with higher scores indicating more extreme violence. The internal consistency estimates for men and women are presented in the results section.

The Experiences in Close Relationships questionnaire (ECR; Brennan, Clark, & Shaver, 1998) is a 36-item self-report measure of romantic attachment scored with a 7-point scale that ranges from strongly disagree (1) to strongly agree (7). The ECR measures two dimensions of attachment: (1) anxiety over abandonment, which refers to an individual self-representation with regard to self-worth and deservingness of the love of others; and (2) avoidance of intimacy, which refers to an individual's perception of others and of the availability and supportiveness of others. Higher mean scores represent greater anxiety and avoidance. Some of the items are reverse-keyed in both subscales. Both attachment dimensions have consistently demonstrated adequate psychometric properties (Alonso-Arbiol, Balluerka, & Shaver, 2007; Lafontaine & Lussier, 2003; Picardi, Bitetti, Puddu, & Pasquini, 2000). In the present study, the internal consistencies of the subscales (Cronbach's α) were .87 and .92 for anxiety over abandonment, and .90 and .93 for avoidance of intimacy for men and women, respectively. Brennan and colleagues (1998) reported similar internal consistency scores (.91 for anxiety over abandonment and .94 for avoidance of intimacy). The ECR factor structure has been supported in a sample of individuals in same-sex couple relationships (Matte, Lemieux, & Lafontaine, 2009).

The Dyadic Adjustment Scale (DAS; Spanier, 1976) is a widely used self-report measure that assesses relationship satisfaction in married or cohabitating individuals. This measure is divided into four subscales: (1) dyadic satisfaction; (2) dyadic cohesion; (3) dyadic consensus; and (4) affection expression. However, inconsistency in the factor structure has been reported, and the use of the total score (rather than of the subscale scores) is recommended (Kazak, Jarmas, & Snitzer, 1988). Some items are reverse-keyed, and the total score ranges from 0 to 132 and represents the sum of all items. Higher scores on the DAS indicate greater relationship adjustment. The cutoff score of 97 differentiates satisfied couples from distressed couples. This measure has been validated with both heterosexual and same-sex couples (Kurdek, 1992). The current study found Cronbach's α values for the dyadic adjustment scale of .94 for men and .96 for women. These values are comparable to Spanier's (1976), who reported .96 for both genders combined.

RESULTS

Preliminary Analyses

Data were screened for missing values prior to the principal analysis. To optimize the sample size, we elected to use expectation maximization to estimate missing values. The use of expectation maximization was appropriate because missing values were random and none of the items had more than 5% missing values. Several participants who had perpetrated or been the victim of more violence than most emerged as outliers. To avoid eliminating all of the variance, data from these participants were retained. Mean scores and standard deviations for each continuous variable are reported in Table 1.

TABLE 1 Means and Standard Deviations for Intimate Violence, Attachment, and Dyadic Adjustment as a Function of Gender in Individuals in Same-Sex Relationships

	Men (N = 75)		Women (N = 143)
Variables	M	SD	M	SD
Psychological aggression
Perpetration	12.60	18.60	10.19	14.80
Victimization	12.55	17.15	8.64	14.71
Physical assault
Perpetration	1.20	7.29	0.99	6.68
Victimization	1.34	7.20	0.91	5.82
Attachment
Anxiety	3.23	1.10	3.09	1.17
Avoidance	2.33	.79	2.15	.93
Dyadic adjustment	89.24	23.21	109.37	24.09

Factor Structure of the Psychological Aggression Scales for Women

The size of the present study's female sample (N = 143) justified the use of Confirmatory Factor Analyses (CFA) to verify the dimensionality of both the perpetration and victimization items of the CTS2 psychological aggression scale. In CFA, a non-significant chi-square indicates that the proposed model is an appropriate way to represent the data. CFA is highly sensitive to assumption violation and sample size; comparative fit indices were, therefore, employed to further support the results. Any coefficient above .90 indicates a good fit for the normed fit index (NFI), incremental fit index (IFI), and comparative fit index (CFI), whereas the recommended maximum for the root mean square error of approximation (RMSEA) is .10 (Browne & Cudeck, 1993).

As recommended by the authors of the CTS2 (Straus et al., 2003), a one-factor model that represented perpetration of both minor and severe acts of psychological aggression was compared to the null model. Results indicated that our one-dimension model was a good fit for the data (χ² = 33.80, p = .01, NFI = .91, IFI and CFI = .95, RMSEA = .08), with standardized regression weights ranging from .18 to .85. Only item 29 fell below the recommended cutoff of .30 (see Table 2 for a description of the items and associated weights). This may be a partial function of the limited variance of item 29, which was endorsed by few women.

TABLE 2 Exploratory and Confirmatory Factor Analyses for Items Measuring Perpetrated Psychological Aggression in Individuals in Same-Sex Relationships

	Men	Women
	(N = 75)	(N = 143)
CTS2 items	EFA	CFA
Minor acts
5. I insulted or swore at my partner.	.63	.74
35. I shouted or yelled at my partner.	.63	.85
49. I stomped out of the room or house or yard during a disagreement.	.38	.70
67. I did something to spite my partner.	.49	.38
Severe acts
25. I called my partner fat or ugly.	.34	.36
29. I destroyed something belonging to my partner.	.65	.18
65. I accused my partner of being a lousy lover.	.65	.49
69. I threatened to hit or throw something at my partner.	.65	.35
Note: EFA = Exploratory Factor Analysis; CFA = Confirmatory Factor Analysis

The same procedure was used to verify the dimensionality of victimization items for women (see Table 3 for a description of the items and associated weights). A one-factor model representing received minor and severe psychological aggression was examined against the null model. Results suggested that our one-dimension model was a good fit for the data, (χ² = 52.58, p < .001, NFI = .87, IFI and CFI = .90, RMSEA = .16), with standardized regression weights ranging from .26 to .89. Only item 26 fell slightly below the cutoff of .30 (Table 3). No women endorsed item 30 (variance of 0), and it was therefore removed from the model.

TABLE 3 Exploratory and Confirmatory Factor Analyses for Items Measuring Received Psychological Aggression in Individuals in Same-Sex Relationships

	Men	Women
	(N = 75)	(N = 143)
CTS2 items	EFA	CFA
Minor acts
6. My partner insulted or swore at me.	.72	.89
36. My partner shouted or yelled at me.	.76	.87
50. My partner stomped out of the room or house or yard during a disagreement.	.75	.53
68. My partner did something to spite me.	.46	.51
Severe acts
26. My partner called me fat or ugly.	.17	.26
30. My partner destroyed something belonging to me.	.62	–
66. My partner accused me of being a lousy lover.	.60	.35
70. My partner threatened to hit me or throw something at me.	.66	.34
Note: EFA = Exploratory Factor Analysis; CFA = Confirmatory Factor Analysis. When item 26 was removed from the male samples’ EFA, results were improved but factor loadings remained stable for all the other items.

Factor Structure of the Psychological Aggression Scales for Men

To account for the smaller male sample (N = 75), Exploratory Factor Analyses (EFA) were conducted on the perpetration and victimization items in the male sample. The one-dimension solution for perpetration items explained 32.01% of the variance, with factor loadings ranging from .34 to .65 (Table 2). The one-dimension solution for victimization items explained 38.34% of the variance, with factor loadings ranging from .17 to .76 (Table 3). ¹ In this solution, item 26 fell below the recommended cutoff score of .30. A modified one-factor solution with this item excluded was examined. The modified solution explained 43.53% of the variance, and factor loadings remained stable for all items (ranging from .46 to .76).

Reliability and Item-Total Correlations

Internal consistency for the items on the psychological aggression scale was measured. Straus and colleagues (2003) suggested that α coefficients of .70 or greater are adequate for measures of psychological characteristics. Alpha coefficients for the perpetration scale were .69 in the female sample and .61 in the male sample. Corrected item-total correlations ranged from .21 to .71 for women and from .09 to .60 for men. Items 29 and 25 yielded the lowest correlations for women and men, respectively, and results indicated that the exclusion of these items would increase the alpha coefficient to .70 for women and to .62 for men. Alpha coefficients for the psychological aggression victimization scale were .71 for women and .70 for men. Corrected item-total correlations ranged from .27 to .75 for women and from .08 to .72 for men. Item 26 yielded the lowest correlation for both women and men, and results indicated that the exclusion of this item would increase the scale's reliability to .72 for both women and men. In consideration of these results and the results of the EFA, item 26 was deleted from the men's psychological aggression victimization scale.

Concurrent Validity

Concurrent validity is an index of the degree to which a test correlates well with an external criterion measure taken at the same point in time (Cohen & Swerdlik, 2009). In the present study, concurrent validity was tested with the anxiety and avoidance romantic attachment scales and the dyadic adjustment scale (see Table 4 for correlation coefficients). In women, the psychological aggression perpetration scale was significantly positively correlated with attachment anxiety and significantly negatively correlated with dyadic adjustment. The psychological aggression perpetration scale was not significantly correlated with attachment avoidance. In men, the psychological aggression perpetration scale was only marginally correlated with attachment anxiety, and not significantly correlated with attachment avoidance or dyadic adjustment. In women, the psychological aggression victimization scale was significantly positively correlated with attachment anxiety but not significantly correlated with attachment avoidance or dyadic adjustment. In men, the psychological aggression victimization scale was marginally positively correlated with attachment avoidance, and marginally negatively correlated with dyadic adjustment. We found no significant correlation with attachment anxiety.

TABLE 4 Concurrent Validity Correlations for the CTS2 Psychological Aggression Scales with Individuals in Same-Sex Couple Relationships

	Psychological		Psychological
	Aggression		Aggression
	Perpetration		Victimization
	Men	Women	Men	Women
Concurrent validity
Attachment anxiety	.19*	.36**	.03	.35**
Attachment avoidance	.04	.13	.20*	.10
Dyadic adjustment	−.16	−.21**	−.22*	−.12
*p < .10 **p < .05.

Perpetrated and Received Psychological and Physical Violence

Rates of violence were calculated for minor psychological violence and minor and severe physical assault, as these two scales are the only ones for which psychometric properties have been examined with individuals in same-sex couple relationships. Women reported perpetrating an average of 10.2 (SD = 14.8) acts of psychological aggression in the past year; 76.2% reported using psychological violence at least once in the past year. Men reported perpetrating an average of 12.6 (SD = 18.6) acts of psychological aggression in the past year; 70.7% reported that they had engaged in psychological violence at least once in the past year. Women reported an average of 1.0 (SD = 6.7) acts of physical assault in the past year; 14.7% reported having perpetrated physical assault in their romantic relationship at least once in the past year. Men reported a similar average of 1.2 (SD = 5.3) acts of physical assault in the past year, and 12% reported having perpetrated physical violence toward their partner at least once in the past year. No significant differences in gender were found for perpetrated psychological or physical violence. (See Table 1 for mean scores and standard deviations).

Victimization rates were calculated next. Women reported an average of 8.6 (SD = 14.7) incidents of received psychological aggression; 70.2% reported received psychological violence at least once in the past year. Men reported having been victims of an average of 12.6 (SD = 17.2) acts of psychological aggression; 78.4% reported received psychological violence at least once in the past year. Women reported an average of 0.9 (SD = 5.8) incidents of received physical assault in the past year; 16.1% reported received physical assault at the hands of their spouse at least once in the past year. Men reported an average of 1.3 (SD = 7.2) incidents of received physical assault in the past year; 12% reported physical assault at the hands of their partner at least once in the past year. No significant differences in gender were found for rates of psychological or physical victimization. (See Table 1 for mean scores and standard deviations.)

The Relationship between Psychological Aggression and Physical Violence

Correlation analyses were conducted to examine the associations between reports of perpetrated and received psychological aggression and reports of perpetrated and received physical assault. Results are presented in Table 5. Perpetrated psychological aggression was significantly correlated with both perpetrated and received physical assault for men and women. Received psychological aggression was significantly correlated with perpetrated and received physical assault for both men and women. In addition, perpetration of psychological aggression was significantly correlated with received psychological aggression for women (.93, p < .01) and for men (.65, p < .001).

TABLE 5 Correlations between Psychological Aggression and Physical Assault in Individuals in Same-Sex Relationships

	Perpetrated		Received
	Psychological		Psychological
	Aggression		Aggression
Physical violence	Men	Women	Men	Women
Perpetrated physical violence	.47**	.51**	.35**	.36**
Received physical violence	.41**	.40**	.35**	.40**
**p < .01.

DISCUSSION

The present study constitutes a significant contribution to the validation of the CTS2 psychological aggression scale with individuals in same-sex couple relationships. Moreover, the study further establishes the prevalence of psychological aggression in same-sex couples and highlights the strength of the relationship between psychological aggression and intimate physical partner violence in this population.

First, a one-factor structure was confirmed for both the psychological aggression perpetration and victimization scales using CFA for women and EFA for men. Some gender differences were found; namely, regression weights were generally stronger for women than for men. In addition, items 29 and 30 (I destroyed something belonging to my partner and My partner destroyed something belonging to me) were only weakly associated with the psychological aggression perpetration and victimization scales in the female sample. However, the weak association may be attributable to the low level of variance for these items (variance of 0.00 for item 30 and .11 for item 29). We hypothesize that a stronger correlation would emerge in a sample with sufficient variance for those two items, and we therefore decided to retain items 29 and 30 on the psychological aggression scales. The EFA provided adequate support for the one-factor solution in the male sample, and only one problematic item was found (item 26; My partner called me fat or ugly). This result could not be attributed to a low level of variance (variance was 8.81); we therefore concluded that item 26 did not contribute significantly to received psychological aggression in gay and bisexual men. The item was removed, increasing the scale's reliability from .70 to .72.

Our results also provided support for the internal consistency of the psychological aggression perpetration and victimization scales. Cronbach's alphas were adequate and comparable to those found in previous studies (Newton et al., 2001). The concurrent validity hypotheses were partially supported. Specifically, attachment anxiety was strongly correlated with both perpetrated and received psychological aggression (with the exception of the men's victimization scale), whereas attachment avoidance correlated significantly with the men's psychological aggression victimization scores only.

These results are consistent with existing literature that suggests that evidence of the relationship between attachment and IPV is stronger and more consistent for anxiety than for avoidance (see Mikulincer & Shaver, 2007, for a review). Although attachment avoidance could lead to the use of IPV to create emotional distance, avoidant individuals also tend to withdraw from interpersonal conflicts and suppress overt expressions of anger and hostility (Bartholomew & Allison, 2006). Dyadic adjustment was also negatively associated with psychological aggression, but the relationship was not systematic. In fact, our results suggest significantly poorer dyadic adjustment for women who perpetrate psychological aggression and for men who are victims of psychological aggression, but not their counterparts.

This result provides further support for gender differences in psychological aggression. A meta-analysis of the relationship between dyadic adjustment and intimate partner violence revealed that the relationship between IPV and poor dyadic adjustment was stronger for male perpetrators and female victims (Stith, Green, Smith, & Ward, 2007). However, the majority of the results were based on the traditional dynamics of IPV in heterosexual couples (i.e., violent male, female victim) rather than on reciprocated IPV. It was hypothesized that the strong relationship between poor adjustment and IPV in male perpetrators was a function of the shame associated with using force on a weaker partner and that, in female victims of received IPV, the stronger relationship was a function of the greater distress experienced by female victims of IPV. A different dynamic may be at play in same-sex couples. For example, women may perpetrate psychological violence as a result of poor dyadic adjustment, and male victims may experience greater distress due to feelings of shame. It is worth noting that, on average, the men in our sample reported a dyadic adjustment below the clinical cutoff score and women reported a dyadic adjustment above the clinical cutoff (as reported in Table 1).

Finally, as predicted, reported rates of perpetrated and received psychological and physical violence were consistent with the existing literature on nonclinical heterosexual and same-sex samples (Jose & O’Leary, 2009). No significant gender differences in prevalence were found for these types of IPV in individuals in same-sex couple relationships. As expected, reports of perpetrated and received psychological aggression were highly correlated for women and men, and psychological aggression was strongly associated with physical assault for both genders. That is, all possible links between reported perpetrated and received psychological and physical violence proved to be significant for both women and men. These results mirror previous findings and suggest a strong link between psychological aggression and physical violence in heterosexual couples (O’Leary et al., 2007) and male same-sex couples (Stanley et al., 2006). The strong correlations between reported perpetrated and received psychological aggression further suggest that an individual's psychologically violent behavior may influence his or her partner's violent behaviors, and vice versa. The results of the present study provide support for the hypothesis that violence in community samples is often mutual (Johnson, 2001). Individuals with violent tendencies may tend to choose violent mates. Alternatively, one partner's violent behavior may trigger the use of violence in his or her partner (Bartholomew & Allison, 2006).

Although our sample size was substantial in comparison to other studies on individuals in same-sex couple relationships, it constitutes the primary limitation of this study. Recruiting a sample large enough for the requirements of sophisticated analyses is one of the most significant and common challenges of conducting research with minority populations. In the present study, the nature of the primary variable constituted an additional hurdle. Although intimate partner violence is quite prevalent, it is not present in all members of the general population. The distribution of reported violent acts is almost always quite skewed, implying that many items are endorsed by few or none of the individuals in the sample. The consequence of the skewed distribution is that a much larger sample size is required for CFA with IPV than for similar analyses with other types of variables (e.g., variables with a Likert-type scale). The combination of methodological challenges described here may explain why the most widely used measure of IPV has yet to be fully validated with individuals in same-sex couple relationships. Much of the research on IPV with heterosexual samples has avoided this problem by conducting CTS2 validation studies with populations with an elevated incidence of violence, such as incarcerated female substance abusers (Lucente et al., 2001) and postpartum mothers at an elevated risk for domestic violence and child abuse (Newton et al., 2001; Duggan et al., 1999).

Despite these challenges, the present study constitutes a valuable contribution to the existing literature on intimate psychological partner violence in same-sex couples. The identification of gender differences in the psychometric properties of the CTS2 psychological aggression scales and of descriptive information distinguish this study from studies that focus on men or women in same-sex couple relationships but do not compare the two groups. Indeed, our results reveal differences in intimate violence between male and female same-sex relationships. Further study of this variable in same-sex couples will facilitate the development and implementation of therapeutic interventions and treatment programs that directly reflect this population's experiences.

Future research into the differences between gay and bisexual men and lesbian and bisexual women in underlying motives for IPV could prove to be enlightening. The study of the similarities and differences in IPV between same-sex couples and heterosexual couples could also constitute an interesting direction for future research. Research on the underlying motives for IPV is in its infancy, but current findings reveal that motives for perpetrating violence in relationships are varied and may be different for men and women. For example, in a study on motives for physical violence in gay and bisexual men, all participants reported using physical violence as a form of expression. More than one half of participants additionally reported instrumental motives (wanting to influence their partner in some way) for perpetrating violence (Stanley et al., 2006). Although to our knowledge, similar data do not exist for lesbian and bisexual women, one study that examined gender differences in motives for engaging in physical violence in heterosexual dating relationships found that women were more likely than men to perpetrate violence as an expression of emotional pain or anger. Men were more likely than women to perpetrate violence out of jealousy or in retaliation for received physical violence (Follingstad, Wright, Lloyd, & Sebastian, 1991). Members of the GLBTQ community tend to have a wide sexual repertoire that may, in some cases, include consensual acts of aggression. Indeed, a number of our participants commented on the lack of distinction between consensual and nonconsensual violent acts in our survey, particularly in reference to the sexual coercion scale. Exploring motives for the use of violence may therefore be particularly important for individuals in same-sex relationships.

The value of investigating precursors of IPV in same-sex relationships is compounded by the finding that members of this population often face unique challenges that may elevate the risk of domestic violence. For example, gays and lesbians are the most frequent victims of hate crimes (Comstock, 1991), and as many as 25% of gay and lesbian youths are forced to leave their homes as a result of conflict with their families over their sexual orientation (Remafedi, 1987). The research in this area confirms that psychosocial environment is a key contributor to later perpetrated and received violence in couple relationships. This finding suggests that the unique and difficult experiences reported by members of the GLBTQ community could predict later violence.

The first step toward a comprehensive understanding of IPV in individuals in same-sex couple relationships is the full validation of a measure of couple violence. We hope that the present contribution to this objective will encourage other researchers to fulfill this mandate.

Acknowledgments

The authors would like to thank the Couple Research Lab team for their support during this project.

Notes

Exploratory and confirmatory factor analyses with the same sample are not recommended. However, in order to provide a rationale for using EFA with our male sample, we made the decision to also present results of the CFA with the male sample. Results suggested that our one-dimension model for perpetration was a good fit for the data (χ² = 22.89, p = .12, NFI = .88, IFI = .96 and CFI = .95, RMSEA = .08) but with problematic standardized regression weights ranging from –.03 to .1.17. Standardized regression weights for items 67 and 25 loaded negatively on the factor and were below the recommended cutoff of .30 for items 69 and 29. Results also suggested that our one-dimension model for victimization was a good fit for the data from the male sample (χ² = 28.93, p = .02, NFI = .89, IFI and CFI = .95, RMSEA = .11) with more acceptable standardized regression weights ranging from .06 to .92 and items 26, 30, and 66 below the cutoff of .30. The instability of the male sample's results indicates that CFA is not suitable for small samples. This is particularly true in the present case wherein may items have low variance due to the nature of the variable measured.