Understanding Everyday Relationship Work : The Development of a Relationship Maintenance Scale Journal Item

Relationship maintenance behaviors contribute to the longevity of intimate relationships, yet existing scales are limited. Available measurement tools are primarily constrained to the Relationship Maintenance Strategy Measure (RMSM) and its further revisions. Covering a number of domains, conceptual overlap with other aspects of an intimate relationship (e.g., household division of labor) may exist. Our cross-sectional exploratory study included participants from 60 countries (n=8,162) who completed an online survey. Participants were diverse in their relationship status, age, sexual orientation, and race/ethnicity. From their responses, we developed a parsimonious and brief measure of relationship maintenance (8 items) through exploratory and then confirmatory factor analyses. Results indicated that the Relationship Maintenance Scale (RMS) shows initial evidence of reliability and validity. The RMS may have utility in working with couples and families. Future research should seek to re-test its use with varied samples, such as couples seeking relationship support.

Despite the way that intimate relationships have been shaped and redefined over time and the loosening of social pressure to legally marry, most adults spend some or nearly all of their lives in a romantic couple relationship. According to a recent report by the Pew Research Center, 53% of adults living in the United States were living with a romantic partner, with 47% being married to that individual (Lenhart & Duggan, 2014). In Europe, marriage rates are somewhat lower as young adults opt to live together without marriage. An exception is the UK, where married couples head up seven in ten households with recent figures showing an increase in legal unions that equates to one marriage every two minutes (ONS, 2014). In the first five years of civil partnerships in the UK (December 2005(December -2010, over 46,000 same-sex partnerships had been registered (ONS, 2011). Understanding these relationships is important as an enduring relationship can support personal well-being and overall health (Kiecolt-Glaser & Newton, 2001) and reduce the risk of depression (Walker, Isherwood, Burton, Kitwe-Magambo, & Luszcz, 2013).
Whether the couple is married or in a committed relationship without marriage, dissolution of the coupledom is a chief social concern and may be viewed as a public health issue given its impact on individuals and children. Dissolution can have an adverse effect on adults' mental and physical health (Coleman & Glenn, 2009), and children of divorce have lower scores on well-being than children living in intact families (Amato & Keith, 1991;Garriga & Kiernan, 2013). A major cause of family breakdown is poor relationship quality (Relationships Alliance, 2014), which is, therefore, an important consideration when evaluating the positive effects of being in a long-term relationship.
Good relationship quality has been shown to have positive impacts on the health, life satisfaction, and happiness of adults (Proulx, Helms, & Buehler, 2007;Robles, Slatcher, Tombello, & McGinn, 2014;Whisman, 2008) and their children (Barrett, Chang, & Walker, 2011;Cummings & Davies, 2010). Relationships that are marked by high levels of satisfaction can help buffer stress (Walker & Luszcz, 2009). Research indicates that good communication (Mackey, Diemer, & O'Brien, 2004), low levels of conflict (Levenson, Carstensen, & Gottman, 1993), and relationship maintenance behaviors (Dainton, 2000) are associated with high relationship quality. Myriad research exists regarding relationship quality and satisfaction, whilst far less is known about what comprises everyday relationship maintenance. For example, how do factors such as family obligation, children, and related responsibilities impact on the couple relationship over time? How do couples sustain their relationships in these different and intersecting contexts?
Understanding how relationships are maintained over the life course and in different contexts is important for both research and social work practice. In terms of research, accurate measurement of those elements that sustain a couple allows further study of the other factors that help maintain intimate relationships. In regards to practice, social workers provide 60% of mental health services in the U.S. and work with couples is a significant part of that role (NASW, 2016). There is, therefore, a pressing need for research that can inform practice-to facilitate effective couple support and work with couples to overcome relationship problems. Targeted intervention can help the couple to re-establish themselves as a team: a unit that endures over time dealing with issues through mutual support and combined strength and resources.
Given that research has found that routine relationship maintenance behaviors are integral to relationship satisfaction (Dainton, 2000), assessing the degree to which these behaviors occur and how they influence relationship quality is fundamental. However, there is a dearth of scales available for measuring relationship maintenance. The Relationship Maintenance Strategy Measure (RMSM; Stafford, 2010) is the primary scale employed for the measurement of relationship maintenance within an intimate relationship. Despite being the most used scale, the RMSM is not without weaknesses. This scale covers a number of different relationship constructs beyond relationship maintenance (such as giving advice). While these other aspects of a relationship would be expected to have an association with relationship maintenance behaviors, we contend that they represent relationship factors beyond relationship maintenance. Our exploratory study sought to pilot a scale of relationship maintenance focused primarily on the measurement of routine behaviors that reinforce the relationship.

The RMSM
The RMSM is grounded in social exchange theory, which suggests that relationship maintenance supports longevity because the relationship will be perceived as rewarding (Stafford & Canary, 1991). Investments made in the relationship create the expectation of reciprocation, and when this system is equitable and rewarding, the relationship is supported. Further, it is presumed that every relationship needs maintenance to be sustained (Stafford & Canary, 1991). From this framework the RMSM was developed. The original 5-factor RMSM (Stafford & Canary, 1991) and the revised 7-factor version (Stafford & Canary, 2006) represent the most widely used measures of relationship maintenance (Stafford, 2010). The 5-factor RMSM was found to explain as much as 56% of the variance in relationship satisfaction (Stafford & Canary, 1991). Aside from these two versions, Canary and Stafford (1992) and Stafford, Dainton, and Haas (2000) made two other revisions, but these modified scales do not appear to be frequently used. Later, Stafford (2010) made further changes to the RMSM to address weaknesses in its measurement strategy and renamed the scale the Relationship Maintenance Behavior Measure (RMBM). While Stafford (2010) seems to have addressed many of the flaws present in the RMSM, the RMBM items are still drawn from the original RMSM.
One of the chief issues found in both the 5-and 7-factor RMSM is that many of the items are double-barreled (e.g., "…is very nice, courteous, and polite when we talk"). Multiple adjectives, along with the qualifier "very," makes this item along with others ripe for measurement error. Stafford (2010) identified this problem and noted that "of the original 29 items, 11 are double or triple barreled" (p. 281). In addition, Stafford (2010) indicated that some items are not explicitly behavioral, which was how relationship maintenance was operationalized for the original development of the RMSM. It is these issues that led Stafford (2010) to subsequently develop the RMBM whereby qualifiers were removed, items that contained multiple ideas (e.g., "try to be romantic, fun, and interesting; p. 281) were broken into more than one item, and conceptual issues (i.e., non-behavioral items) were addressed.
These changes undoubtedly strengthened this scale, but other conceptual concerns may persist. For example, both the RMSM and the RMBM contain many different constructs associated with intimate relationships, including positivity, assurances, openness (in the RMBM this is covered in two factors: relationship talks and selfdisclosure), tasks, and networks. Each of these constructs is seen as one individual factor of relationship maintenance. Positivity refers to the degree to which a positive attitude is used with one's partner. Assurances contain behaviors that support commitment and love. Openness items cover the ability to discuss the relationship, including feelings about the relationship. Tasks are primarily focused on the division of labor within the household, while networks are focused on the extent to which time is spent together with friends and family members (Stafford & Canary, 1991). The 7factor RMSM also contains management and advice while the RMBM includes understanding.
While these dimensions may, in fact, impact relationship satisfaction or perhaps relationship maintenance strategies, we question if these concepts are a part of routine relationship maintenance. In other words, are all five (or seven) of these different types of behaviors necessary to maintain a relationship? Moreover, do they adequately include the minutiae of behaviors that may routinely take place? We argue that these various elements of an intimate relationship go beyond relationship maintenance inand-of-itself. For example, in the management scale, equality in the division of labor is indicative of relationship maintenance. That is, to achieve a high score on this factor, which would be an indicator of greater relationship maintenance, a participant must have a household that is based on equal sharing of tasks and responsibilities. Even though egalitarianism within the household structure is a widely accepted notion, it is not something that is embraced by all couples. Indeed, for those who chose to create a coupledom based on traditional gender roles, because the couple is in, for example, cultural or religious contexts where the demarcation of gender roles is viewed positively or because they do not co-habit, such egalitarian ideas may have little purchase and/or meaning. Researchers cannot know the degree to which equability in the relationship may be supporting the maintenance of that relationship; as such, this needs to be measured as a separate construct related to intimate relationships.
Furthermore, sampling methods and the make-up of the final sample raises concern over the applicability of both the RMSM and the RMBM. The 5-factor RMSM was based on a sample made up predominantly of students (69%) at a Midwestern and a Western university who, in turn, also recruited the remainder of the sample for extra credit in a course (Stafford & Canary, 1991). The racial/ethnic background, sexual orientation, and other sociodemographic descriptions of the sample are not provided, but the sample did include both married and non-married (e.g., engaged or dating) participants. The RMBM sample was also recruited by students from a Midwestern university for extra credit, and it was further limited to heterosexual married persons (Stafford, 2010). Additional background information about the sample is not given. Item retention and scale development are dependent on the response patterns of participants. When those responses are limited to a narrow demographic, its use with diverse samples, including individuals in same-sex relationships and non-cohabitators, may not be appropriate without further research to demonstrate factor stability.

Current Study
Relationship maintenance has been conceptualized in the literature as behaviors that are both strategic and routine in nature and help to sustain the relationship over time (Dainton & Stafford, 1993). These behaviors may be strategic when they are intentionally enacted with the goal of supporting the relationship. On the other hand, these behaviors may be unintentional or routine in nature, but still help to sustain the relationship. Given that research has substantiated the importance of a mutually rewarding relationship, where the quality of the relationship includes reciprocal assertions and positive interactions (Ogolsky & Bowers, 2012), measurement is a key issue. Utilizing Dainton and Stafford's (1993) framework along with findings from the literature, we conceptualized relationship maintenance as the degree to which routine relationship behaviors are intentionally or unintentionally used to sustain an intimate relationship. This definition guided our scale development and the subsequent study, which sought to develop a conceptually parsimonious scale of routine relationship maintenance behaviors with low respondent burden. Based on our review of the literature and a previous pilot study, we tested our items both with experts and advanced statistical analysis. Our goal was to create a strengths-based scale that assesses routine behaviors in relationship maintenance and to develop this new scale by recruiting a diverse sample of participants (including same-sex attracted) who were in different types of relationships (married, cohabiting, and living apart). We aimed to develop a scale more reflective of contemporary society.

Method Item Development
Our larger study was designed around the structuring interests of the Enduring Love? research project, which investigated how couples sustain their long-term partnerships. "Long-term" was not straightforwardly associated with relationship duration but instead framed through personal experiencesuch as family stability, household permanency, and security of friendship networks. It is also experienced in dynamic terms, including the past (time spent together), the present (relationship satisfaction), and a desire to be with a partner into the future (relationship horizon; Gabb & Fink, 2015). The study focused attention on the ways in which couple relationships are materialized and experienced through everyday relationship practices in the context of different personal, socio-cultural, and economic circumstances. That is to say, relationships are about money, employment, children, and housework as much as the dyadic couple relationship itself. Based on the literature and our theoretical conceptualization of relationship maintenance, we developed items to create the Relationship Maintenance Scale (RMS), which could then be subjected to further testing. These items were designed to include everyday contexts and mundanities as well as the more intimate dynamics of couple relationships (Gabb & Fink, 2015).
Each item was couple-focused (i.e., began with the stem "we") and included making time to be together, saying "I love you," and pursuing shared interests. This framing of statements was designed to elicit responses on lived experience rather than perceptions of what "typical" couples do and/or should ideally do. Items were given to experts in family or relationship research to provide feedback on content coverage, including relevance for different groups (e.g., LGBTQ). In this way, we were attentive to the specificity of experience amongst these different groups to ensure that it did not marginalize minority experience and thus elide relationship diversity. Items were also discussed with frontline staff from different relationship support organizations to determine if our new scale would provide useful information regarding intimate relationships that could subsequently inform and be applied in clinical contexts. In response to this feedback, items were revised for clarity. One item originally intended for another scale that we were developing (relationship quality; see Chonody, Gabb, Killian M, & Dunk-West, 2016) was conceptually re-evaluated and found to be more consistent with the relationship maintenance items. Through this consultative process, a total of 14 items were included in our measure of relationship maintenance and subjected to further testing.

Data Collection
The dataset was based on a large-scale mixed methods study, Enduring Love? Couple Relationships in the 21 st Century that was funded by the Economic and Social Research Council (RES-062-23-3056) and administered in the United Kingdom. This first wave of data collection was completed in two stages. Firstly, an online survey was implemented over a 12-month period (2011)(2012). Qualitative data collection overlapped this period and continued for 20 months. Information on the study was posted on various online forums and newsletters, with targeted recruitment on parenting and relationship support community noticeboards. To increase participation of under-represented groups the survey questionnaire was also implemented in hard copy format and disseminated among hard-to-reach community groups and networks through face-to-face contact. The survey directed participants to the study's website where project information was available alongside frequently asked questions (FAQs) and detailed guidance on research methods and ethics.
In the second wave, researchers in the U.S. and Australia extended this study, and data were collected anonymously through Survey Monkey from August through December 2014. Recruitment in these two countries relied primarily on social media (Facebook and Twitter) and sharing of the link amongst social networks and media.
Contents of our survey and our hypotheses are described below. Additional questions and scales were included in the survey and are described elsewhere. Prior to data collection, the relevant Ethics and Institutional Review Board (IRB) granted approval of this study.

Convergent construct validity variable.
A single-item indicator related to happiness with one's partner was used as a test of convergent construct validity. Participants were asked to rate this question: "How happy are you with your partner overall?" employing a 5-point Likert-type scale (1=very unhappy and 5=very happy). Given that in a recent meta-analysis, relationship maintenance behaviors were associated with liking one's partner (Ogolosky & Bowers, 2012), a positive correlation was hypothesized between this item and the RMS.
Known-groups validity variable. Parenthood was assessed by a dichotomous question and was used as a test of known-groups. Research findings indicated that relationship maintenance is lower amongst parents, likely due to the demands of childcare (Dainton, 2008). Therefore, we hypothesized that parents would have lower relationship maintenance scores than non-parents.

Socioedemographic variables.
Age, gender, sexual orientation, religious affiliation, education, employment, length of relationship, and relationship status were included in our survey and are used descriptively in the current analysis.

Data Analysis
Our data analysis plan for testing the RMS started with a review of item performance, including skew and kurtosis and a correlational analysis. With the large sample, a split-sample exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were conducted. Using SPSS version 23.0, the sample was randomly divided into two equal subsamples. Next, an exploratory factor analysis using principle components analysis was performed on one subsample to determine the factor structure of the scale and identify any poorly performing items (Fabrigar, Wegener, MacCallum, & Strahan, 1999). Eigen values greater than one were used to identify the factors, and orthogonal rotation was used (Varimax) to improve factor loading interpretation. After arriving at a factor structure from the EFA, a confirmatory factor analysis (CFA) was conducted using the other half of the split-sample and the obtained factor structure from the EFA (Fabrigar et al., 1999). We hypothesized that a factor structure found in one half of the sample (EFA) would then be found and supported in the second half (CFA).
The model chi-square value and chi-square per degrees of freedom ( 2 /df; Bollen, 1989;Kline, 2011), Comparative Fit Index and Tucker-Lewis Index (CFI and TLI; Hu & Bentler, 1999;Kline, 2011), Standardized Root Mean Square Residual (SRMR; Brown, 2006), and the Root Mean Square Error of Approximation (RMSEA; Kline, 2011) were each used to assess the degree of fit between the obtained scores and the model. Generally, acceptable fit for a CFA would be lower  2 values, lower  2 /df scores, a RMSEA and SRMR values less than .08, and CFI and TLI scores each greater than .95. The CFA was conducted using Mplus version 7.3 (Muthén & Muthén, 2012). Lastly, the final factor structure for the RMS was then used for tests of convergent construct validity and known-groups validity.

Demographics
Once missing data and respondents who reported no long-term relationship were removed, a final sample of 8,132 was achieved with participants from 60 different countries. A majority of the participants were from the UK (69%). Demographically, our sample was diverse with 12% who reported as a member of a sexual minority group, 49% who indicated that they were atheist/agnostic, 37% who were in a relationship other than marital, and 27% who identified as a person of color. Table 1 provides additional sociodemographic information.

Evaluating Item Performance
Measures of central tendency, skew, and kurtosis were checked, and items were found to have an acceptable degree of variance. Skew and kurtosis were not greater than 2.5 for any of the items. A correlational analysis of the items indicated adequate association between items, and each item had at least one correlation greater than .30. No items were correlated with another item above .80, which would indicate item redundancy. All 14 items were used for the next step of analysis.

Exploratory Factor Analysis
Next, the sample was randomly split into two equal subsamples. Responses from one subsample (n=4,066) were used for the exploratory factor analysis (EFA). Measures of sampling adequacy and sphericity were excellent (Kaiser-Meyer-Olkin=.853, Bartlett's test of sphericity=9077.80, df=91, p<.001). The initial model explained 53.0% of the variance in the items and included four factors. Items were removed from the model due to double loadings (greater than .40) or weak loadings (less than .40). After several iterations, a final one factor model with 8 items was obtained and explained 40.9% of the variance. Table 2 contains the final items for the scale along with factor loadings and communalities from the EFA results.

Internal Consistency Reliability
The RMS demonstrated good internal consistency reliability (Cronbach's alpha) in both subsamples and with the full sample. The Cronbach's alpha obtained in the EFA subsample was good (=.775). Similar internal consistency scores were found in the CFA subsample (=.787) and in the full sample (=.781).

Convergent Construct and Known-Group Criterion-Related Validities
As hypothesized, happiness with one's partner was highly correlated with the RMS (r=.670, p<.001). This finding is indicative of evidence for convergent construct validity between maintenance and liking, which has been reported in the literature (Ogolosky & Bowers, 2012).
Also, as hypothesized, parents (M=29.94, SD=5.41) reported significantly lower relationship maintenance than respondents without children (M=31.78, SD=4.62), t=14.82 (df=5714), p<.001. This finding is consistent with the literature (Dainton, 2008) and provides evidence of known-groups validity. Figure 1. Confirmatory Factor Analysis: Item Loadings (n=3,857) Items correspond to item list in Table 2 Discussion Results of our study indicated that the newly developed RMS shows initial evidence of both reliability and validity. Coefficient alpha was found to be good for a newly developed scale (Nunnally, 1978), and given the large sample, these preliminary results are promising for future research with the RMS. The single factor model obtained through the EFA was supported by the CFA results providing evidence of factorial validity. While all psychometric studies are sample dependent, the degree of diversity of couple relationships found in our community sample and its international sampling frame support its structure for capturing this latent construct. Evidence of convergent construct validity was indicated by the association between greater happiness with one's partner and more relationship maintenance, as found in past studies (e.g., Ogolosky & Bowers, 2012). Relatedly, participants without children reported greater relationship maintenance than those with children, which is consistent with prior research (Dainton, 2008) that shows the stressors that parenthood can bring to bear on the couple relationship (Walker, Barrett, Wilson, & Chang, 2010). This association establishes evidence for known-groups validity.
Available measures for the assessment of relationship maintenance are few and may suffer from conceptual issues. The RMS offers an unobtrusive and strengths-based assessment of relationship maintenance in that it is focused on those elements of the relationship that may be working well. It is conceptually parsimonious with a sole focus on routine behaviors that support the maintenance of the relationship, unlike the RMSM, which has multiple factors that are not relationship maintenance behaviors per se. This allows researchers or practitioners to learn how individuals or couples are working to sustain their relationship. Moreover, the RMS is a rapid assessment instrument, which can easily be incorporated into larger studies on relationships that may contain a number of scales or can be quickly implemented in a practice setting.
Anecdotal evidence based on participant feedback about our study suggested that participants were compelled to think more about their relationships, including some who asked their partners to also complete the survey, while others initiated conversations with her/his partner about the content of the survey and thus their own partnerships. This is encouraging and suggests clinical utility for the RMS as a tool for practice that provides a starting point for discussions on how the relationship can be strengthened and the couple unit supported.
Social work practice with couples and families might benefit from the use of the RMS as a baseline clinical assessment tool, to encourage participation in therapeutic efforts, or to facilitate dialogue and remove barriers to communication during sessions. Given the social, psychological, and economic impact that dissolution has for both the individual partners as well as any children that may be part of that family system (e.g., Coleman & Glenn, 2009;Garriga & Kiernan, 2013), support for enduring partnerships is an essential component of social work practice. The strengths-based nature of this scale can provide social work practitioners and their clients an unobtrusive and efficient assessment of progress during clinical work. Gaining insight into both the strengths and weaknesses that are present in the relationship may facilitate the change process. Future research should seek to garner feedback from couples on how this scale and other elements of our survey can promote partner communication around relationship maintenance.
Relationship maintenance is a burgeoning area of research, and its role in relationship quality is important for enduring coupledoms (Dainton, 2000). These promising preliminary results warrant future tests of the psychometric properties of the RMS. Concurrent construct validity and predictive validity are needed using other measures of couples' maintenance behaviors and long-term relationship outcomes especially with a sample of dyadic participants (i.e., couples), including those who may be seeking relationship support or who are having relationship difficulties. Future research should also seek to parse out various aspects of intimate relationships and how they relate to relationship maintenance, such as communication, assurances, and division of household labor, which can inform theory development and intervention efforts.

Limitations
Findings from psychometric studies are sample-dependent in nature; thus, our results need to be considered within the limitations of our sample. First, the sample was primarily composed of individuals from the UK, and they were mostly married and identified as heterosexual. The sample was gained through social media and through convenience sampling methods, which limits the generalizability of the results. Further replication is therefore needed. Moreover, individuals who are happy in a relationship are more likely to complete a relationship survey (Levenson et al., 1993), and our reach to couples who are in trouble was likely limited. Similarly, women are more likely to complete a relationship survey, and our sample reflects that tendency. Nonetheless, we did achieve a degree of diversity in terms of sexual orientation and relationship status, thus expanding the potential utility of the scale beyond the heterosexual, marital, cohabiting couple.
Second, the extent to which these individuals may have had or are currently experiencing relationship problems was not assessed. Therefore, the scale needs further exploratory testing with individuals/couples who are seeking support to determine its usefulness given that responses from another sample may support changes to the obtained factor model. Future studies should assess if the RMS can differentiate between distressed and non-distressed couples, which would be beneficial in providing further validity evidence as well as support its use in practice. Third, within the CFA analysis, error variances from two items were allowed to correlate which may indicate the presence of other unique relationship maintenance constructs or sample-specific variation (Kline, 2011). However, the EFA analyses did not support the presence of a second factor within this large sample of individuals, and the addition of the correlated error terms in the CFA was within the sample factor and theoretically supported. The current study did use a large, international sample and a randomly selected split-half EFA and CFA; the analytical strategy combined with the diversity of the sample thus lends confidence to the obtained factor model of the RMS.

Conclusion
The operationalization of relationship maintenance was reconceptualized in order to avoid the overlap of constructs present in the two most widely used scales in this area of research. The RMS is conceptually parsimonious with a low respondent burden and has shown initial evidence of both reliability and validity in a diverse sample of participants. Additional psychometric studies are needed to support our findings and expand its use with other populations. Longitudinal studies are essential to determine if the RMS can identify couples in distress. Moreover, future research should seek to determine how the scale could be used in clinical practice to provide insight into relationship maintenance behaviors for couples in treatment. Qualitative tools used in our extended mixed methods study have proven to be highly effective in systemic family therapy (Gabb & Singh, 2014). We are thus hopeful that the RMS may be similarly helpful in facilitating partner perceptions of relationship maintenance and identification of strengths alongside areas that require further work.