Conceptualization and Assessment of Disengagement in Romantic Relationships

Abstract

Research examining relationship distress and dissolution highlights the importance of romantic disengagement. However, prior conceptualizations and measures of romantic disengagement have tended to combine disengagement with related but distinct constructs hindering the study of romantic disengagement. In the present study we conducted exploratory factor analyses to demonstrate that disengagement is a relatively distinct construct and to clarify the conceptualization of romantic disengagement. More importantly, we developed a novel measure– the Romantic Disengagement Scale (RDS). The RDS demonstrated adequate fit across samples of dating individuals, married couples and women in physically aggressive relationships. The RDS also demonstrated strong divergent and incremental validity. Implications for enhancing conceptual models, research methodology, and clinical interventions are discussed.

“The opposite of love is not hate; it is indifference.”

-Elie Wiesel, October, 1986

Researchers agree that high levels of romantic disengagement comprise a stage of relationship decline from which few couples recover (Gottman, 1999; Kayser, 1993). Indeed, growing apart, lack of love, and lack of affection are among the most frequently cited reasons couples give for distress and dissolution in dating and marital relationships (Albrecht, Bahr, & Goodman, 1983; Amato & Previti, 2003; Baxter, 1986; Bloom & Hodges, 1981; Gigy & Kelly, 1992; Hill, Rubin, & Peplau, 1976; Sprecher, 1994). Additionally, couple therapists cite disengagement as the most difficult problem to treat (Whisman, Dixon, & Johnson, 1997).

Despite the known detrimental effects of disengagement on relationship satisfaction and stability, researchers do not agree on how to conceptualize disengagement. This lack of a consistent, empirically supported conceptual framework hinders our ability to develop and test theories of disengagement as a key mechanism through which initially satisfied and/or committed couples become distressed and/or dissolve their relationships. In particular, existing conceptualizations and self-report measures of disengagement often combine the study of disengagement with related but distinct constructs such as loneliness, conflict, or commitment, rendering these approaches less informative for understanding disengagement as a unique mechanism in the developmental course of relationships. Even when researchers do not combine disengagement with other constructs, they may embed the study of disengagement in romantic relationships within the study of disengagement across multiple types of relationships, hindering our ability to understand the arguably unique nature and process of disengagement in romantic relationships. The present study was designed to clarify the construct of disengagement empirically, and to develop and validate a self-report measure of disengagement in intimate relationships based on this new, empirically supported conceptualization.

Conceptualizing Disengagement

Researchers agree that emotional indifference toward one’s partner and one’s relationship are important aspects of romantic disengagement. Emotional indifference has been described as a lack of strong positive emotion (e.g., low levels of love) and relatively little negative emotion (e.g., anger; Gottman, 1999). It has also been identified by low levels of positive affect such as low interest in the spouse or relationship (Gottman, 1999; Kayser, 1993; 1996; Snyder & Regts, 1982), and by low levels of energy and excitement when interacting with one’s spouse (Smith, Vivian, & O’Leary, 1990). In sum, the construct of romantic disengagement appears to comprise emotional and affective deadening toward one’s romantic partner.

Researchers also agree that disengagement comprises cognitive and behavioral strategies aimed at increasing psychological or physical distance from one’s partner (i.e., relational distancing). Avoidance and withdrawal are routinely included in definitions of disengagement (Gottman, 1999; Kayser, 1993; 1996; Snyder & Regts, 1982). Compared to individuals who are relatively more engaged, disengaged individuals are less involved in their partners’ lives (Gottman, 1999), speak with their partners less frequently (Kayser, 1993). They tend to interact with their partners in less intimate ways (e.g., refraining from personal disclosures; Kayser, 1993), and are less attentive toward their partners and their relationships (Gottman, 1999; Kayser, 1993). In sum, relational distancing, whether achieved via cognitive or behavioral strategies, is another key component of romantic disengagement.

Relational distancing has received careful and detailed attention in Hess’s (2002; 2003) research on interpersonal relationships. Hess conceptualized relational distancing as comprising three types of behavioral and cognitive strategies that might be employed in a given dyadic relationship: (a) avoidance, which includes tactics aimed at minimizing physical contact or communication; (b) disengagement, which includes behaviors that limit or eliminate intimacy; and (c) cognitive dissociation, which includes tactics aimed at perceiving the other person as different or detached from oneself.

Hess’s articulation of specific behavioral and cognitive distancing strategies has the potential to expand upon previous conceptualizations of romantic disengagement. However, his underlying conceptual framework was developed to assess distancing strategies occurring in any interpersonal relationship rather than strategies that might be unique to romantic relationships. This distinction is important because romantic relationships are typically voluntarily maintained whereas many other relationships (e.g., family, work colleagues) are less voluntary and often must be tolerated. Distancing strategies in relatively involuntary relationships may include behaviors that would objectively be viewed as quite negative. Indeed, factor analyses of Hess’s distancing strategies yielded two factors: “unfriendly” and “withdrawal” (Hess, 2003). Withdrawal is consistent with previous conceptualizations of romantic disengagement; however, unfriendliness, although it may serve the purpose of increasing interpersonal distance, is likely more strongly associated with relational conflict. Consequently, we argue that the conceptualization of romantic disengagement can be improved by adapting and adding many of Hess’s relational distancing strategies, while simultaneously attending to the arguably important distinctions between disengagement strategies in romantic versus non-romantic dyads.

In contrast with the general consensus that romantic disengagement comprises emotional indifference and relational distancing strategies, other potential facets of the construct are less clear. For example, Gottman’s (1999) conceptualization of disengagement includes tension and sadness whereas other researchers do not specify these aspects. Kayser’s (1993) research on the process of disaffection – a construct we consider synonymous with romantic disengagement – suggests that anger, disappointment, and hopelessness are present during various stages of disengagement; however, it is not clear whether Kayser considers these facets to be part of the disengagement construct itself or to be separate, albeit related, constructs. In sum, questions remain regarding the exact nature of romantic disengagement and the most valid way to operationalize this construct.

Measuring Disengagement

We identified three published scales that assess romantic disengagement or the similar construct of romantic disaffection. Gottman’s (1999) Emotional Disengagement and Loneliness Scale measures aspects of emotional indifference (e.g., “We are pretty separate and unconnected emotionally”), cognitive distancing strategies (e.g., “I have learned to expect less from my partner”), and behavioral distancing strategies (e.g., “I often try to avoid saying things I will later regret”). Also, as evident in the title of the measure, it includes items that measure the affective state of loneliness (e.g., “I am often lonely in this marriage”) as well as negative emotions such as disappointment (e.g., “I am often disappointed in this marriage”). Kayser’s (1996) Marital Disaffection Scale (MDS) assesses emotional indifference (e.g., “Apathy and indifference best describe my feelings toward my spouse”), behavioral distancing strategies (e.g., “I find myself withdrawing more and more from my spouse”), and negative emotions such as anger (e.g., “I have a lot of angry feelings toward my spouse”). The MDS also assesses positive relationship qualities such as love (“My love for my spouse has increased more and more over time”) and closeness (e.g., “I enjoy sharing my feelings with my spouse”). Snyder and Regts’s (1982) Marital Disaffection Scale measures behavioral distancing (e.g., “about the only time I’m with my spouse is at meals and bedtime”), a lack of shared interests (e.g., “my spouse and I don’t have much in common to talk about”), conflict behaviors (e.g., “My spouse and I argue nearly all the time”), and positive relationship qualities such as closeness (e.g., “my spouse seems to enjoy just being with me”) and commitment (e.g., “I am thoroughly committed to remaining in my present marriage”).

In conclusion, researchers appear to agree on the inclusion of three components of disengagement-- emotional indifference, behavioral distancing, and cognitive distancing. However, there are important inconsistencies regarding the extent to which existing conceptualizations and measures also include additional (albeit related) components such as loneliness, anger, conflict, and closeness. For example, whereas Kayser (1993) included anger in her measure, Gottman (1999) suggests that disengaged individuals demonstrate relatively low levels of anger. Whereas Snyder and Regts (1982) include conflict in their measure of marital disaffection (and it is not reverse scored), Gottman (1999) asserts that disengaged couples evidence an absence of open, frequent conflict. We argue that it is important to resolve these inconsistencies and, more importantly, to have a relatively pure measure of romantic disengagement (i.e., composed of core components) in order to develop and test theories of disengagement as a key mechanism through which initially satisfied and/or committed couples become distressed and/or dissolve their relationships.

It is possible that disengagement is not conceptually or statistically (e.g., via factor analyses) distinct from low levels of intimacy or from high levels of conflict and anger, as some of these researchers have implied. If this is the case, then a new measure of disengagement, such as the one developed and validated in the present study, would neither be necessary nor incrementally useful. Alternatively, if disengagement is distinct, as we assert, then relying on measures that combine disengagement with related but distinct constructs prohibits the explication of disengagement as a unique process in the developmental course of relationships. Given that almost half of first marriages end in divorce (Kreider & Fields, 2001), and given the consequences of divorce for spouses and children (Amato, 2000; Hetherington & Stanley-Hagen, 1999; Kiecolt-Glaser & Newton, 2001), research targeting the prediction and prevention of relationship distress and dissolution is critical. Our ability to effectively predict and prevent divorce is contingent upon our ability to determine whether, for example, frequent couple conflict versus disengagement (or some combination of the two) is integral to understanding the processes through which initially satisfied, committed couples become dysfunctional.

Overview of the Present Study

Our first goal was to clarify the construct of romantic disengagement by determining whether disengagement is distinct from related constructs such as anger, conflict and loneliness. To address this aim (Study 1), we developed an item pool based on existing conceptualizations and measures of romantic disengagement, disaffection, and relational distancing. We tested the factor structure of the item pool to identify one or more disengagement factors present in ongoing romantic relationships. We expected one or more factors to emerge comprising emotional indifference, cognitive (e.g., distracting oneself when with one’s partner), and behavioral (e.g., avoidance) disengagement/distancing.

The second aim was to create and validate a self-report measure of romantic disengagement in Study 2 (the Romantic Disengagement Scale; RDS). To address this aim, we retained the items that loaded on the disengagement factor(s) from Study 1, revised the measure, and confirmed the psychometric properties in three new samples: dating individuals, married couples, and women in physically aggressive relationships. Ideally, cross-validation occurs in samples that exemplify populations one wishes to study in order to demonstrate adequate validity in each population. Because we wished to demonstrate that the RDS was valid in dating and marital samples, these were important samples to include. Additionally, because our dating and marital samples were quite homogeneous demographically (e.g., length of relationship, race and ethnicity, education), and because these participants were relatively satisfied in their relationships, we included a third sample comprised of physically abused women who were expected to be considerably more heterogeneous in terms of demographics and key variables of interest (e.g., disengagement, satisfaction, commitment). Therefore, our ability to validate the RDS in these three distinct samples should provide evidence for the generalizability of the construct and our new measure. We expected the RDS to demonstrate reliability and adequate factorial fit for the data in all three samples.

Next we sought to demonstrate the discriminant validity of the RDS. First, we examined whether our new measure of disengagement discriminated disengagement from conflict. Couple conflict has received a great deal of attention in the close relationships literature, and has consistently been linked to relationship decline and dissolution (for a review see Bradbury, Rogge & Lawrence, 2001). However, the nature of the association between disengagement and conflict remains unclear. Kayser (1993) suggested that disengagement is associated with conflict because conflict leads to increased disengagement; however, others have suggested that disengagement is distinct from conflict (Gottman, 1999; Snyder & Regts, 1982). We hypothesized that disengagement would be related to but distinct from conflict. Demonstrating this distinction would also provide evidence for the divergent validity of our measure.

Second, we assessed the ability of our measure to discriminate disengagement from low levels of love, passion and intimacy. Love, passion, and intimacy are integral to the study of romantic relationships (e.g., Aron, Aron & Smolan, 1992; Fehr & Russell, 1991). Gottman (1999) and Kayser (1993) conceptualized romantic disengagement as including low levels of love, passion, and intimacy; however, If such a conceptual approach is warranted, we may find that the RDS is too highly correlated with measures of these constructs (e.g., rs ≥|.80|). Such collinearity would indicate that our measure of disengagement is redundant with (low levels of) love, passion, and intimacy, and that the RDS may not be useful over and above measures of these constructs. We hypothesized that the RDS would be related to but distinct from love, passion and intimacy, demonstrating its divergent validity and the importance of distinguishing among these constructs in future research.

Third, we hypothesized that the RDS would be weakly and positively associated with relevant personality variables, in particular, detachment and attachment avoidance. Detachment is defined as the tendency across time and situations to prefer to keep to oneself and to be alone (Clark, 1999). Individuals high on detachment tend to devalue social relationships (Clark, 1999) and would be expected to be disengaged from all social relationships. Attachment avoidance is defined as the tendency to be uncomfortable depending on or being intimate with significant others (for reviews see Cassidy & Shaver, 1999; Rholes & Simpson, 2004). Individuals high on attachment avoidance would be expected to be relatively disengaged from any romantic partner. However, unlike detachment and attachment avoidance, romantic disengagement is conceptualized as a relationship-specific quality, which should change as the relationship changes, in contrast to a personality characteristic (i.e., a relatively stable quality of the individual across relationships and time).

Fourth, we expected the RDS to be strongly negatively associated, but not redundant, with two important constructs of relationship functioning – satisfaction and commitment. Individuals who are romantically disengaged tend to experience relationship distress (i.e., low levels of relationship satisfaction) and consider relationship dissolution (i.e., low levels of commitment; Gottman, 1999; Kayser, 1993).

Finally, we sought to demonstrate the incremental validity of the RDS in explaining two key relational outcomes -- satisfaction and commitment -- over and above relational distancing (i.e., Hess’s measure (2002) adapted for participants in romantic relationships). Relationship (dis)satisfaction is one of the most widely studied outcomes in close relationships research (for a review see Karney & Bradbury, 1995), and therefore an obvious choice as our outcome of interest. Relationship commitment (the extent to which one’s partner is a part of an individual’s future plans; Stanley & Markman, 1992) may be especially relevant because couples may experience low levels of satisfaction but still maintain the relationship due to personal commitment (Rusbult, Olsen, Davis, & Hannon, 2001). Therefore, we also included commitment as an important relational outcome. Given that disengagement includes relational distancing but that the existing measure of relational distancing (Hess, 2002) was not designed to assess romantic disengagement specifically, we expected the RDS to demonstrate significant incremental validity in explaining relationship satisfaction and commitment over and above relational distancing.

Study 1 Method

Participants and Procedure

A sample of college students in dating relationships was recruited for Study 1. Previous research suggests that romantic disengagement is present in and detrimental to dating relationships (Baxter, 1986; Hill et al., 1976; Sprecher, 1994). Additionally, among college students, dating partners are an important and preferred source of social support (Furman & Burmeister, 1992), intimacy, and companionship (Roscoe, Diana & Brooks, 1987) compared to other sources (e.g., friends and parents).

Participants were 412 students (296 women and 116 men) enrolled in an introductory psychology course at a public university in the United States. To participate, students had to be in exclusive, heterosexual romantic relationships lasting at least two months. Participants ranged in age from 18 to 26 (M = 18.9, SD = 1.2 years), and were predominantly Caucasian/Non-Hispanic (90%). Most students defined their relationships as “seriously dating” (95%); only 4% were cohabiting. Relationship length ranged from 2 months to 6 years (M = 1.6, SD = 1.3 years). Participants were recruited via a university-managed online sign-up system and told that the purpose of the study was to aid in questionnaire development. Participants completed Informed Consent Documents, provided demographic information, and completed the disengagement items and other measures beyond the scope of the present study. Participants were assessed in small-group settings and received credit for partial fulfillment of a course requirement.

Development of the Initial Item Pool

We generated a pool of 47 items via a series of procedures. First, we included items from existing scales of romantic disengagement and disaffection (e.g., Gottman, 1999; Kayser, 1996). Although we expected romantic disengagement to be distinct from some facets included in existing scales, we included items measuring researchers’ broader conceptualizations so as to allow for the delimitation of the borders of the construct (Clark & Watson, 1995). Second, we included items from Hess’s (2002; 2003) Relational Distancing Index, but adapted them to reduce the negative valence. For example, “When in __’s presence I kept to myself and spoke less than I would have if I liked him/her” was changed to “When in my partner’s presence, I keep to myself and speak less than I normally would with other people.” Third, we wrote five new items to assess Smith et al.’s (1990) operational definition of disengaged affect (e.g., “When I am with my partner, I feel more tired than usual”). Fourth, to identify potential omissions relevant to the construct of disengagement, items were examined for relevance, comprehensiveness, and representativeness (Haynes, Richard, & Kubany, 1995). For example, we wanted to ensure that there were at least 3 items for each potential factor (i.e., emotional/affective, cognitive, and behavioral distancing) of disengagement (a strategy recommended by Clark & Watson, 1995, and Comrey, 1988). This examination led us to write two additional items. Items had a 5-point Likert-style response format ranging from 1 (never) to 5 (always). Respondents were instructed to answer the questions regarding how they had felt and acted in their current romantic relationship in the last month.

Results and Discussion for Study 1

The distributions of the individual disengagement items were examined in order to eliminate items with highly unbalanced distributions as recommended by Clark and Watson (1995). Seven items with skew greater than |2| or kurtosis greater than |4| were eliminated. We conducted an exploratory factor analysis (EFA) to test our hypothesis that one or more factors comprising emotional indifference, cognitive (e.g., distracting oneself when with one’s partner), and behavioral (e.g., avoidance, speaking less) indicators of romantic disengagement would emerge as distinct from items assessing the facets previously included in measures of disengagement (e.g., anger, conflict, loneliness, love). To determine the best factor solution, we estimated a series of principal axis factor analyses with promax rotation using Mplus software (Muthén & Muthén, 1998–2007), and examined each structure for: (1) adequate fit for the data via a root mean square residual (RMSR) estimate of .08 or lower (Browne & Cudeck, 1992), (2) interpretability of the factors, and (3) an adequate number of items to define a factor (i.e., three or more; Comrey, 1988) without high cross-loadings (i.e., items loading more than .30 on two or more factors).

Once rotated, 7 factors had eigenvalues above 1.0; therefore, structures with 1 through 7 factors were extracted and examined. The one-factor structure was rejected because it was a poor fit for the data (RMSR = .11). The two-factor solution had only 5 items (split between the two factors) without high cross-loadings, leaving only 2 items loading on the 2^nd factor. Thus we rejected the two-factor solution. The 3-factor structure adequately fit the data (RMSR = .07), yielded interpretable factors, and each factor had an adequate number of items without high cross-loadings. (We return to a discussion of the three–factor structure briefly.) When we extracted 4 factors, the structure adequately fit the data, and the 1^st and 2^nd factors were interpretable. However, the 3^rd and 4^th factors did not appear to have any interpretable distinction, and many items on these two factors had high cross-loadings. Thus, we rejected the 4-factor structure. Structures with 5, 6, and 7 factors had only one or two items loading on the final factors and were therefore rejected.

The three-factor structure was the only solution that met our three criteria. We labeled the 1^st factor Disengagement because items loading on this factor included emotional indifference, cognitive, and behavioral distancing strategies. We labeled the 2^nd factor Negativity/Dissatisfaction because items on this factor included anger, disappointment, and thoughts about relationship dissolution. We labeled the 3^rd factor Positivity/Closeness because items on this factor included positive relationship qualities such as love and intimacy. Once the factors were identified, items that had high cross-loadings (above .30) with other factors were eliminated in order to isolate the items that most distinctly measured disengagement. The final results of the 3-factor structure generated from our EFA are presented in Table 1. The factors, although relatively distinct, remained highly and significantly correlated (ps < .01). Disengagement correlated .59 with Negativity/Dissatisfaction and −.56 with Positivity/Closeness; Negativity/Dissatisfaction correlated −.50 with Positivity/Closeness.

Table 1.

Study 1. Factor Loadings for the 3-Factor Structure

Items	Disengage	Negative/Dissatis.	Positiv./Closeness
When I am around my partner I don’t pay attention to him/her.	.74	−.06	−.03
When my partner is speaking, I pretend to agree or I avoid asking questions, in order to make things easier.	.70	−.06	.04
When in my partner’s presence, I keep to myself and speak less than I normally would with other people.	.63	.01	−.05
When I am with my partner, I feel more tired than usual.	.60	.16	.12
When I think about my partner, I don’t feel much of anything (i.e. apathetic or indifferent).	.52	−.01	−.22
I ignore my partner.	.49	.14	−.09
I would prefer to spend less time with my partner.	.46	.13	−.18
I feel frustrated in my relationship.	.08	.83	.12
I feel disappointed that my relationship with my partner is not how I once expected it would be.	−.11	.74	−.20
I think about breaking up with my partner.	.14	.63	−.08
I have a lot of angry feelings toward my partner.	.05	.61	.09
I feel disappointed that my partner is not the person I once thought he/she was (or would be).	.04	.58	−.16
I feel my relationship with my partner is more important to me than almost anything else in my life.	.12	−.03	.72
I feel a great deal of love and affection for my partner.	−.12	−.01	.71
When I have a personal problem my partner is the first person I turn to.	.00	.02	.67
I confide in my partner.	−.16	.05	.61
I enjoy spending time alone with my partner.	−.10	−.08	.48

N = 412; factor loadings over .40 in bold; Disengage = Disengagement Factor, Negative/Dissatis. = Negativity/Dissatisfaction Factor, Positiv./Closeness = Positivity/Closeness Factor. Items reordered for ease of presentation.

In conclusion, Study 1 demonstrated that the construct of disengagement comprises items representing emotional indifference and cognitive and behavioral distancing. Items believed to represent emotional/affective, cognitive, and behavioral aspects of disengagement comprised one disengagement factor rather than forming distinct factors. Moreover, the construct of disengagement was related to but distinct from many of the facets previously included in conceptualizations and measures of romantic disengagement (e.g., couple conflict, anger, love, and intimacy). Next we sought to revise the item pool to more reliably measure the key aspects of romantic disengagement identified in Study 1 and, more generally, to develop and validate a measure of romantic disengagement.

Study 2 Method

Participants and Procedure

We included three independently recruited samples in Study 2 -- dating individuals, married couples, and women in physically aggressive relationships -- to demonstrate the reliability and validity of our revised measure. Cross-validation of a measure in samples exemplifying different populations is important to demonstrate generalizability of the construct and of the measure. Although all samples were used to demonstrate reliability and factorial validity, measures of constructs used to demonstrate discriminant and incremental validity were only available in some of the samples. Samples 2 and 3 were participating in ongoing longitudinal studies; therefore, it was necessary to limit the size of questionnaire packets for these samples to retain participants over time.

Sample 1

Participants were 203 students (120 women and 83 men) enrolled in an introductory psychology course at a public university in the United States. (This sample is different from the sample enrolled in Study 1.) Eligible participants were in ongoing, exclusive heterosexual romantic relationships lasting at least two months. Participants ranged in age from 19 to 34 (M = 22.0 years, SD = 2.1 years) and were predominantly Caucasian/Non-Hispanic (91%). Most participants defined their relationships as “seriously dating” (96%). Only 8% were cohabiting, and relationship duration ranged from 2 months to 7 years (M = 1.4 years, SD = 1.1). None of the demographic factors were significantly associated with disengagement, relationship satisfaction or commitment. Participants were recruited and debriefed using the same method described in Study 1. Participants completed the Informed Consent Document, provided demographic information, completed the revised disengagement items, and completed measures of (a) communication/conflict behaviors, (b) love, passion, and intimacy in their romantic relationships, (c) detachment and attachment avoidance, and (d) relationship satisfaction and commitment. A subsample (n = 63) also completed the Relational Distancing Index (RDI; Hess, 2003). We added the Relational Distancing Index (RDI; Hess, 2002) to our procedure after initial data collection had begun. Thus, 63 participants completed the RDI and 203 completed all other measures. Participants completed the questionnaires on private computers when they were alone. Although hard-copy questionnaire packets were available, no participants chose this method of data collection.

Sample 2

Married couples were recruited via marriage license records from small towns and cities in the United States Midwest to participate in a larger longitudinal study of newlywed marriage. Couples were told that the purpose of the study was to better understand how couples navigate the early years of marriage. To be eligible, both partners had to be between the ages of 18 and 55, relatively fluent in English, in their first marriages, and married less than 6 months. Of the 358 couples who responded, the first 103 who met eligibility requirements and kept their initial laboratory appointments were included in the larger study. At enrollment in the study, the modal family income was $30,001- $40,000, and the modal level of education completed was 14 years. On average, couples dated for 48 months (SD = 27.8) before marriage, and 77% cohabited an average of 21.1 months (SD = 17.3) premaritally. Participants were primarily Caucasian (90.3%).

At the fifth wave of data collection (M = 4.7 years of marriage, SD = .5), 77 couples completed the measures included in the present study. By this wave of data collection, 3 of the original 103 couples had withdrawn from the study, 13 couples had divorced or permanently separated, 1 couple chose not to participate, and the research team was unable to contact 9 other couples. Average ages were 31.3 (SD = 3.9) for husbands and 29.9 (SD = 4.4) for wives. Questionnaires (including those included in the present study) were mailed to couples’ homes. Spouses were instructed to complete questionnaires independently and return them in the individual, stamped envelopes provided. Couples were paid $25 for participation.

Sample 3

Participants were 42 women recruited from flyers and public service announcements in rural small towns and cities in the Midwest for a study about physical aggression in heterosexual romantic relationships. Flyers were posted in health clinics, on buses, in women’s shelters, and women’s restrooms in public places (e.g., grocery stores, at the state fair), and public service announcements were made on local radio stations and printed in local newspapers. To be eligible, women were required to have been cohabiting with a heterosexual romantic partner for at least three months and to have been the victim of at least two acts of moderate (e.g., grabbing, pushing) or one act of severe (e.g., punching, kicking) physical aggression (as defined by Straus, 1979) during arguments with their partners in the previous three months. Participants were told that the purpose of the study was to better understand how women’s relationships affect their health and well-being. Participants ranged in age from 21–73 (M = 35.5, SD = 12.3); 66% were Caucasian/non-Hispanic, 27% were African-American, and 7 % were Hispanic or Asian. Twenty percent completed a 4-year college degree, 12% completed an associate’s degree, 46% had “some college,” and 17% graduated from high school or earned a GED. Most participants (85%) reported a personal annual income below $20,000, and a joint annual income below $30,000. Seventy-one percent of participants were married, 15% were engaged and cohabiting, and 15% were separated from their partners (but had been cohabiting at the time of recruitment). Relationship duration ranged from 3 months to 56 years (M = 8.2, SD = 10.3 years). For the larger study, participants completed questionnaires and interviews 5 times over the course of one year. Data used in the present study were collected at Time 2 over the phone by a trained interviewer who had interviewed the participant in person at Time 1 (to establish rapport). Participants completed the revised list of disengagement items, measures of relationship satisfaction and commitment, and other measures beyond the scope of the present study. Women were paid $25 for their participation in Time 2.

Revisions to the Disengagement Scale

The items that loaded on the Disengagement Factor yielded in Study 1 were used to develop the Romantic Disengagement Scale. We revised the instructions to read as follows:

Feelings of connection in your romantic relationship may fluctuate over time for many reasons. For example, you may feel less connected when you or your partner are experiencing stress, or you may feel more connected when your anniversary is coming up. Please think about how you have felt and acted in the last month when with, talking to, or thinking about your romantic partner.

In accord with Clark and Watson’s (1995) recommendation that items on assessment scales use appropriately simple and straightforward language, we made several revisions to the items themselves. First, we changed the wording from present to past tense to be more consistent with the instructions. Second, we removed specifiers such as “when I am around my partner” because they were redundant with the instructions. Third, we split one item (“when I am in my partner’s presence, I keep to myself and speak less than I normally would with other people”) into two shorter items (“I kept to myself” and “I spoke less than I normally would”). We also created 12 new items in order to more reliably measure the key aspects of romantic disengagement identified in Study 1, including emotional (e.g., “my emotions (both positive and negative) didn’t seem very strong to me,”), cognitive (e.g., “I daydreamed about something else”) and behavioral distancing (e.g., “I avoided dealing with my partner”). Finally, we altered the response format to 1 (never), 2 (rarely), 3 (sometimes), 4 (half the time), 5 (frequently), 6 (almost always), and 7 (always) to allow for greater variability in responding. Measures (see Table 2 for a depiction of which measures were included in each sample)

Table 2.

Study 2. Descriptive Statistics: Means and Standard Deviations across Samples

	Dating Sample		Marital Sample				Physically Abused
			Husbands		Wives		Women
Variable (possible score range)	M	SD	M	SD	M	SD	M	SD
Disengagement (1–7)	2.19	.82	2.25	1.01	2.12	.80	3.61	1.62
Conflict Communication (0–1)	.26	.24	.32	.26	.29	.25
Love (1–7)	5.72	1.59
Passion (1–7)	5.50	1.50
Intimacy (1–7)	5.50	1.59
Detachment (0–1)	.32	.22	.33	.23	.23	.22
Avoidant Attachment (1–5)	2.33	.69	2.42	.72	2.32	.74
Relationship Satisfaction (1–7)	5.63	1.57	5.84	1.31	5.93	1.15	3.26	1.81
Commitment (1–7)	5.71	.98					3.71	1.60
RDI Unfriendly (1–7)	1.35	.54	1.66	.82	1.59	.73
RDI Withdrawal (1–7)	1.38	.53	1.80	.86	1.52	.62

Note. RDI = Relational Distance Index. Dating sample: N = 203 for all variables except RDI; for RDI scales, n = 63. Marital sample (husbands and wives): N = 77 couples. Physically victimized women: N = 42. All means and standard deviations were converted to their respective response scales.

Conflict communication was measured using the problem-solving communication subscale (PSC) of the Marital Satisfaction Inventory-Revised (MSI-R; Snyder & Aikman, 1999). The PSC is a 19-item, true/false scale measuring the degree of negative communication between partners (e.g., “My partner often fails to understand my point of view on things”), particularly communication during disagreements or arguments (e.g., “When we argue, my partner and I often seem to go over and over the same things”). Higher scores indicate more negative communication. Internal consistency was above .90 across samples.

Love, passion, and intimacy were measured using the love, passion, and intimacy subscales of the Perceived Relationship Quality Components inventory (PRQC; Fletcher, Simpson, & Thomas, 2000). The PRQC subscales measure “subjectively held evaluations in the minds of relationship partners” (p. 340, Fletcher et al., 2000). Each subscale consists of three items that participants rated on a scale ranging from 1 (not at all) to 7 (extremely). An example item assessing love is “How much do you love your partner?” An example item assessing passion is “How passionate is your relationship?” and an example item assessing intimacy is “How intimate is your relationship?” Confirmatory factor analyses by Fletcher et al. (2000) demonstrated that love, passion, and intimacy are related but distinct components of romantic relationship quality as measured by the PRQC. Internal consistency was above .85 across the 3 scales in Sample 1.

Detachment was measured using the Schedule for Nonadaptive and Adaptive Personality - Detachment subscale (SNAP; Clark, 1993), which measures the extent to which participants are aloof and distant from other people and prefer to be alone. The Detachment subscale consists of 18 items such as “I don’t particularly like spending time with people” and “I am pretty reserved around others.” The subscale has a true/false response format. Coefficient alphas were above .80 across samples.

Attachment avoidance was measured with the Relationship Scales Questionnaire (RSQ; Griffin & Bartholomew, 1994). The RSQ directs respondents to consider how they feel about close relationships in general, both past and present, on a 1 (not at all like me) to 5 (very much like me) scale. An example of an item is, “I find it difficult to depend on other people.” Items were factor analyzed as suggested by Kurdek (2002) to identify items that loaded most highly on an avoidance factor without cross-loading too highly on other factors. Nine items were retained. Internal consistency was above .78.

Relationship satisfaction was measured using the Quality of Marriage Index (QMI; Norton, 1983), a 6-item scale that measures global satisfaction with one’s relationship and one’s partner. Although initially validated with married samples, the QMI has also been used to study dating relationships (e.g., Hess, Fannin, & Pollom, 2007). For five items, respondents indicate the extent to which they agree with statements such as “We have a good relationship” on a scale from 1 (very strong disagreement) to 7 (very strong agreement). For item 6, respondents report their overall level of happiness in the relationship on a scale from 1 to 10. The QMI demonstrates good convergent and discriminant validity (Heyman, Sayers, & Bellack, 1994). Internal consistency was above .90 across samples.

Commitment was measured using the 12-item, brief version of the personal dedication subscale of Stanley and Markman’s Commitment Inventory (1992). The subscale was written to reflect the personal desire to maintain a relationship for the benefit of both relationship partners. An example item is “My partner is clearly part of my future plans.” Responses are made on a scale from 1 (strongly disagree) to 7 (strongly agree). The measure demonstrates strong convergent validity with other commitment measures and concurrent validity with positive relationship functioning. Internal consistency was .88 in Sample 1 and .80 in Sample 3.

Relational distancing was measured with the Relational Distancing Index (RDI; Hess, 2003), which measures strategies for increasing distance in specific interpersonal relationships. In the present study participants were asked to complete the RDI with regard to their current romantic partner. Subscales include unfriendliness, which has nine items (e.g., “I mentally degraded ___, such as by seeing her/him as less than human, less capable, or having fewer rights than others”), and withdrawal, which has eight items (e.g., “I changed my behavior to avoid encountering __ whenever possible”). Participants respond using a 1 (I never did this) to 7 (I did this every time possible) scale. The scales demonstrate adequate validity and reliability across samples and across time (Hess, 2003). Coefficient alphas ranged from .84 to .91 across the two scales and across samples.

Results for Study 2

We examined the distributions of the individual disengagement items separately in each sample in order to eliminate items with highly unbalanced distributions (Clark & Watson, 1995). Two items had kurtosis greater than |4| (“I ignored my partner” and “I avoided coming in contact with my partner”) in Samples 1 and 2 and were therefore eliminated.

Table 2 presents the descriptive statistics for the three samples. On average, dating and marital participants reported relatively low levels of romantic disengagement, whereas women in physically aggressive relationships were moderately disengaged. There were no significant sex differences for disengagement among dating individuals or married couples (ts < .88, all ns). On average, participants in the dating and marital samples were relatively satisfied with their relationships whereas women in physically aggressive relationships were relatively dissatisfied. One-way ANOVAs yielded significant differences across the three samples for mean levels of disengagement (F(3, 395) = 49.57, p < .01 and satisfaction (F(3, 395) = 43.38, p < .01). Post-hoc Tukey tests revealed that women in physically aggressive relationships were significantly more disengaged and less satisfied than participants in the dating and married samples (p < .01). Participants in the dating and marital samples did not significantly differ from each other on mean levels of disengagement or satisfaction.

Reliability Analyses and Confirmatory Factor Analyses across Three Types of Relationships

To assess reliability of the RDS, we examined internal consistency via coefficient alphas. To demonstrate adequate fit of the factor structure of the RDS across the three samples, confirmatory factor analyses (CFAs) were conducted using Mplus software (Muthén & Muthén, 1998–2007). We evaluated the fit of our models using four fit indices: the overall model chi-square, the comparative fit index (CFI), the incremental fit index (IFI), and the standardized root mean residual (SRMR). Conventionally, models are judged to be an adequate fit for the data when the CFI and IFI are .90 and above, and the SRMR is below .08 (Finch & West, 1997; Hu & Bentler, 1998; 1999; Kline, 1998). Across samples, we tested 1- and 3-factor structures. The 1-factor structure was specified with one latent variable (disengagement) and each of the 18 disengagement items as the measured variables. The 3-factor structure was specified so that the items we categorized as representing emotions or affective states, behaviors or cognitions each loaded on separate factors to determine whether these indicators were indeed distinct. (See the legend in Table 3 for categorization of items). Across the three samples, the 3-factor structure was a poor fit, and was an inferior fit compared to the 1-factor structure; we report the results of the 1-factor structure below. Estimated factor loadings for the 1-factor models across all three samples are provided in Table 3.

Table 3.

Study 2. Factor Loadings on the 1-Factor Romantic Disengagement Scale across Samples

	Dating Sample	Marital Sample		Victimized Women
Items		Husbands	Wives
1. I didn’t feel like dealing with my partner.	.86	.75	.74	.86
2. I felt more tired than usual.	.77	.77	.57	.82
3. I didn’t feel much of anything (i.e., indifferent).	.71	.74	.69	.75
4. I wanted to spend less time with my partner.	.83	.88	.73	.82
5. I didn’t feel like spending time with my partner.	.86	.85	.81	.82
6. I didn’t feel like being touched.	.77	.67	.59	.85
7. My emotions (both positive and negative) didn’t seem very strong to me.	.71	.77	.61	.67
8. I thought about something to distract myself from my feelings.	.78	.82	.58	.86
9. I daydreamed about something else.	.61	.75	.63	.49
10. I didn’t focus a great deal of attention on him/her.	.46	.74	.61	.47
11. I pretended to agree or I avoided asking questions, in order to make things easier.	.76	.73	.50	.75
12. I was somewhat withdrawn.	.88	.78	.80	.95
13. I tried to suppress any expression of my feelings.	.71	.87	.65	.62
14. I kept to myself.	.83	.89	.89	.80
15. I tried not to let my feelings show.	.85	.85	.67	.89
16. I spoke less than I normally would.	.88	.88	.81	.93
17. I avoided dealing with my partner.	.82	.88	.79	.75
18. I was not as open as I usually am.	.90	.93	.81	.93

Note. A complete list of items included in initial and revised item pools may be obtained from first author. Based on our proposed conceptualization, items 1–7 represent emotional distance, items 8 and 9 represent cognitive distance, and items 10–18 represent behavioral distance; however, factor analyses did not distinguish components. All factor loadings were significant at p < .001. Items were reordered in this table for ease of presentation.

Dating sample (Sample 1)

The model was an adequate fit for the data (CFI = .94, IFI = .95, SRMR = .05, χ2 (135) = 644.12, p < .01), and coefficient alpha was .97, supporting the factorial validity and reliability of the RDS to measure disengagement in dating relationships.

Marital sample (Sample 2)

Recent research has indicated that the necessary sample size for confirmatory factor analyses (CFAs) depends in part on the ratio of indicators to factors, as well as the inter-item correlations (Marsh, Hau, Balla, & Grayson, 1998). Samples as small as 50 yield 99.6% proper convergence of model iteration when there are at least 6 indicators per factor and loadings are moderate to high (Marsh et al., 1998). In our marital sample, our ratio of indicators to factors was 18 to 1 and we expected moderate to high factor loadings. Thus our sample size of 77 couples was deemed sufficiently large to run a CFA.

We modeled husbands and wives separately for two reasons. First, it would be inappropriate to ignore the interdependence of data provided by couples (Kenny, Kashy & Cook, 2006) by modeling the data as if it had been provided by independent participants. The optimal way of accounting for interdependence between husbands and wives is to run a CFA via structural equation modeling techniques. Using this approach, we would model husbands’ and wives’ disengagement simultaneously as two correlated latent factors. However, our sample size was too small to run such complex analyses. The models for husbands and wives adequately fit the data (husbands: CFI = .95, IFI = .96, SRMR = .07, χ2 (135) = 404.55, p < .01; wives: CFI = .90, IFI = .96, SRMR = .08, χ2 (135) = 338.67, p < .01). Internal consistency was .97 for husbands and .95 for wives. The results support the factorial validity and reliability of the RDS for married men and women.

Physically victimized women (Sample 3)

The model was an adequate fit for the data (CFI = .91, IFI = .92, SRMR = .08, χ2 (135) = 310.91, p < .01). Internal consistency was .96. The results support the factorial validity and reliability of the RDS for measuring disengagement among women in physically aggressive relationships.

Is Disengagement Distinct from Conflict Behavior and from Love, Passion, and Intimacy?

To demonstrate discriminant validity we examined the bivariate correlations of disengagement with (a) conflict behavior in Samples 1 and 2 and (b) love, passion, and intimacy in Sample 1. Given that Study 1 demonstrated high and significant (but not redundant) associations between disengagement and negativity/dissatisfaction, and between disengagement and positivity/closeness, and given the conceptualizations purported by Gottman (1999) and Kayser (1993), we expected these correlations to be large in size (i.e., rs ~|.50|), but not so strong as to suggest that disengagement was redundant with these constructs (e.g., rs ≥ |.80|).

First, we compared disengagement with conflict communication in the dating and marital samples (see Table 4). Disengagement was moderately positively correlated with conflict communication for dating and married individuals such that disengagement was greater for individuals who also experienced more difficulty resolving conflicts. Next we compared disengagement with love, passion, and intimacy in the dating sample. Disengagement was strongly negatively correlated with love, passion, and intimacy such that disengagement was greater for individuals who also experienced lower levels of love, passion, and intimacy. In sum, these analyses provide support that romantic disengagement is related to, but relatively distinct from, conflict communication and positive relationship constructs.

Table 4.

Study 2. Bivariate Correlations between the Romantic Disengagement Scale and Variables Included to Support Discriminant and Incremental Validity

	Dating Sample	Marital Sample		Victimized Women
		Husbands	Wives
Conflict Behavior	.31**	.40**	.49**
Love	−.51**
Passion	−.51**
Intimacy	−.54**
Detachment	.27*	.26*	.19*
Avoidant Attachment	.32**	.27*	.34**
Relationship Satisfaction	−.63**	−.59**	−.52**	−.56**
Commitment	−.58**			−.56**
RDI Unfriendly	.52**	.63**	.67**
RDI Withdrawal	.53**	.71**	.72**

Note. RDI = Relational Distance Index.

^*p < .05.

^**p < .01.

Is Romantic Disengagement Distinct from Detachment and Attachment Avoidance?

Next we examined bivariate correlations of disengagement with (a) detachment and (b) attachment avoidance in dating and married individuals. Results are presented in Table 4. As expected, disengagement was weakly positively correlated with detachment in dating and married individuals. Also, disengagement was moderately positively correlated with attachment avoidance in dating individuals and wives and weakly positively correlated for husbands. These findings support our contention that the RDS is not simply measuring individual differences such as global detachment as a personality trait or an avoidant attachment style, but rather is measuring a construct that is relatively specific to the relationship.

Is Disengagement Distinct from Relationship Satisfaction and Commitment?

We examined bivariate correlations to demonstrate discriminant validity of romantic disengagement from (a) relationship satisfaction (in Samples 1, 2, and 3) and (b) commitment (in Samples 1 and 3; see Table 4). As expected, disengagement was strongly negatively correlated with relationship satisfaction in dating individuals, married individuals, and physically victimized women. Also as expected, disengagement was strongly negatively correlated with commitment in dating individuals and physically victimized women. The findings demonstrate that disengagement is distinct from (albeit related to) relationship satisfaction and commitment.

Does the RDS Provide Explanatory Power over and above Measures of Interpersonal Distance?

We conducted incremental validity analyses to assess the utility of the RDS in predicting key relationship outcomes (satisfaction and commitment) over and above the previously published RDI (Hess, 2003) in our samples of dating and married individuals. Specifically, we conducted a series of hierarchical regression analyses, entering the two RDI scales at Step 1 and the new RDS at Step 2 (see Table 5). First we predicted relationship satisfaction and commitment in our dating sample (Sample 1). The RDS explained 28% of the remaining variance in predicting satisfaction (semi-partial r = .53), and 13% of the remaining variance in predicting commitment (semi-partial r = .36, p < .001). Next we predicted relationship satisfaction in the marital sample (Sample 2). Analyses were conducted separately for husbands and wives. The RDS explained 14% of the remaining variance for husbands (semi-partial r = .37, p < .001), and 10% of the remaining variance for wives (semi-partial r = .32, p < .01).

Table 5.

Study 2. Hierarchical Regression Analyses Demonstrating Incremental Validity of the Romantic Disengagement Scale (RDS) over Hess’s Relational Distancing Index (RDI)

	Dating Sample				Marital Sample
	Satisfaction		Commitment		Husbands’ Satisfaction		Wives’ Satisfaction
Step 1	B (SE)	β	B (SE)	β	B (SE)	β	B (SE)	β
RDI Unfriendly	−1.41 (.63)	−.49*	−1.24 (.87)	−.37	−.80 (.42)	−.45+	−.59 (.21)	−.30**
RDI Withdrawal	−.14 (.65)	−.05	.13 (.66)	.05	−.10 (.40)	−.06	−.66 (.18)	−.39*
Step 1 R2	.31**		.11		.26**		.33**
RDI Unfriendly	−.96 (.50)	−.33	−.52 (.88)	−.16	−.88 (.38)	−.50*	−.49 (.20)	−.25*
RDI Withdrawal	.47 (.52)	.16	.97 (.70)	.39	.62 (.41)	.41	−.13 (.23)	−.08
Disengagement	−1.18 (.19)	−.64**	−.89 (.37)	−.64*	−.83 (.21)	−.53**	−.71 (.22)	−.45**
Total R2	.59**		24*		.40**		.43**
Step 2 R2 Δ	.28**		.13*		.14**		.10**

Note. RDI = Relational Distancing Index.

⁺p < .10.

^*p < .05.

^**p < .01.

There is no universally accepted criteria for evaluating incremental validity estimates; however, researchers suggest that predictors entered beyond the 2^nd step that have a semi-partial r of about .30 demonstrate meaningful contributions (Hunsley & Meyer, 2003). Therefore, the results suggest that the RDS demonstrates sufficient incremental validity in explaining relationship satisfaction in dating and married individuals and commitment in dating individuals.

At first glance, the extent to which disengagement explains dating participants’ relationship satisfaction (i.e., the magnitude of the effect) appears to be much larger than the extent to which disengagement explains dating individuals’ commitment or spouses’ satisfaction. To clarify these results, we compared the 95% confidence intervals for the unstandardized beta coefficients for disengagement across the three analyses. The confidence intervals overlapped, so we cannot conclude that there were any significant differences in the relative magnitudes of the effects of disengagement on our outcomes across analyses.

Discussion

Disengagement has been routinely conceptualized, though rarely empirically tested, as a key mechanism through which initially satisfied, committed couples become distressed and ultimately dissolve their relationships. With the goal of facilitating methodologically sound, prospective investigations of this potential pathway to relationship dysfunction, we sought to clarify the construct of disengagement via a factor-analytic investigation of existing conceptualizations and self-report measures of disengagement, and to develop a self-report measure based on this clarified conceptualization of disengagement. The present investigation supports a conceptualization of romantic disengagement that includes emotional, affective, behavioral, and cognitive indicators. Additionally, we found that disengagement was distinct from related constructs, and that it predicted the key outcomes of relationship satisfaction and commitment above and beyond existing measures (e.g., Hess, 2002). Given that lower levels of relationship satisfaction and commitment are associated with relationship distress and dissolution, the present study supports our contention that disengagement is important as a unique process for explicating the developmental course of relationship distress and dissolution. In sum, researchers and clinicians now have a conceptually guided, empirically derived, psychometrically sound measure of romantic disengagement that is far less redundant with related but distinct constructs (compared to previously published measures), and that is conceptually unique to romantic relationships.

The present study has several methodological strengths. First, the sample size and the size of the item pool were sufficient to allow for the extraction of subscales (e.g., between disengagement and dissatisfaction). Second, the item pool and the RDS were empirically derived and theoretically driven. Third, the present study expands upon and integrates research from a variety of disciplines (Gottman, 1999; Hess, 2002; 2003; Kayser, 1993; 1996; Snyder & Regts, 1982; Smith et al., 1990). Fourth, we demonstrated the reliability and validity of this novel scale in multiple samples, providing initial evidence of the generalizability and utility of the measure across a range of romantic relationships.

There are also several limitations to the present study. First, although our sample sizes are comparable to other published samples of newlywed couples and physically abused women, they were relatively small and did not allow us to analyze husbands and wives simultaneously. Second, although Sample 3 evidenced racial diversity, our other samples were primarily Caucasian, of higher SES, and relatively satisfied in their relationships. Moreover, across all of our samples, participants were in heterosexual relationships. The RDS should be validated with larger, more diverse samples of married and cohabiting couples as well as individuals in same-sex romantic relationships to more firmly establish the generalizability of our findings to these populations. Second, the present study relied on self-report data. Multi-method assessment of romantic disengagement, such as by comparing the RDS to behavioral observations of disengaging behaviors during couple interactions, would provide further evidence of validity of our measure.

There are several conceptual implications of the present study. First, the results support our contention that romantic disengagement is conceptually distinct from couple conflict. Theoretical models of the developmental course of relationship distress and dissolution (e.g., social learning or behavioral models; e.g., Jacobson & Margolin, 1979; Weiss, 1978; a vulnerability-stress-adaptation model; e.g., Karney & Bradbury, 1995) routinely consider conflict and behavior occurring in the context of couple conflict (e.g., hostility, anger, negative affect), yet disengagement is typically not considered. Our ability to now examine disengagement as a unique construct will enable us to refine these models.

Second, we now have the ability to prospectively test and refine conceptual models explicitly designed to explain disengagement as a dynamic process. For example, according to the retrospective research conducted by Kayser (1993), we would expect disappointment and anger to erode one’s sense of closeness and love for his or her partner, leading to disengagement from the partner and/or from the relationship. However, retrospective reports are subject to memory bias and often found to be inaccurate (Neisser & Fivush, 1994). The RDS can be used in longitudinal studies to allow researchers to test Kayser’s model of disaffection prospectively. Similarly, the work of Gottman (1999) and others (e.g., Fincham & Linfield, 1997) suggests that couples follow different pathways toward relationship distress, one of which is purported to be romantic disengagement. Inclusion of the RDS in future research would allow researchers to test these ideas.

Methodologically, we contend that one reason for the relative dearth of research on disengagement -- compared to the research on couple conflict, for example -- is due to the inherent difficulty in operationalizing and quantifying the construct. For example, many researchers tend to describe romantic disengagement in terms of the absence of negative or positive behaviors and feelings (e.g., lack of open/frequent conflict, lack of interest, lack of disclosure; Gottman, 1999; lack of love, lack of attachment; Kayser, 1993) without providing an adequate description of what exists in their absence. Obviously, it is difficult to study a construct operationalized as being the absence of other constructs. In contrast, in the present study we have generated a definition of romantic disengagement in terms of the presence of specific behaviors and feelings (e.g., avoidance, emotional indifference, fatigue).

From a clinical standpoint, the RDS could prove to be a useful and efficient tool for couple therapists. Romantic disengagement is a particularly difficult problem for couple therapists to treat (Whisman et al., 1997). Adding to that difficulty, couples present for therapy at various stages of disengagement, and each stage presents unique challenges for the clinician (Gottman, 1999; Jacobson & Christensen, 1998). We recommend that, during the assessment phase of couple therapy, therapists administer the RDS to determine partners’ levels of disengagement, which could then be used to guide treatment.

In conclusion, despite the known detrimental effects of disengagement on relationship satisfaction and stability, researchers have historically not agreed on how to conceptualize disengagement, hindering our ability to develop and test theories of disengagement as a key mechanism through which initially satisfied and/or committed couples become distressed and/or dissolve their relationships. In the present study we addressed these inconsistencies in the field and generated a clarified conceptualization of disengagement. Perhaps more importantly, we yielded a reliable and valid self-report questionnaire that is conceptually grounded in this clarified construct, that discriminates disengagement from related constructs, and that predicts key relationship outcomes over and above existing measures. We are emboldened by the new opportunities researchers now have to explicate the juncture between relational maintenance and relational dissolution.