Understanding the Role of Relationship Maintenance in Enduring Couple Partnerships in Later Adulthood

ABSTRACT Intimate relationships in later adulthood are understudied despite their positive association with health and well-being. This cross-sectional mixed methods study sought to redress this gap by investigating relationship maintenance in later adulthood. Our international sub-sample comprised 1,565 participants aged 55 + and in an ongoing relationship. Results from hierarchical multiple regression indicated that overall happiness with the relationship had the largest effect size on relationship maintenance, with 53% of the variance explained. Content analyses of open-ended questions identified companionship and laughter as some of the “best liked” aspects of the relationship. Housework/cooking and saying “I love you” were among the behaviors that made participants feel appreciated. Results illustrated the types of maintenance behaviors adults in later adulthood who are in enduring partnerships employ.

the individual. That is, couple relationships marked by high quality have been shown to have a positive impact on individual health (Umberson, Williams, Powers, Liu, & Needham, 2006) and well-being (Sherwood, Kneale, & Bloomfield, 2014). Indeed, relationship quality makes a more a positive contribution to overall well-being than relationship longevity. A "good" marital relationship may help to mediate stressors, including greater resilience when coping with significant illness (Walker & Luszcz, 2009). Understanding how relationship quality is generated and maintained and the ways that it manifests in everyday life, is important, particularly in later adulthood when a partner can be the primary source of support (Mackey, Diemer, & O'Brien, 2004;Walker & Luszcz, 2009).
Most research on intimate partnerships, however, is focused on younger/ middle-aged couples and/or couples experiencing distress or dissolution. Less is known about couples in later adulthood, in particular those factors related to how relationships are maintained. In this gap of research evidence, factors found to be important for middle-aged couples are assumed to be equally applicable for older people (Walker, Isherwood, Burton, Kitwe-Magambo, & Luszcz, 2013). Given that relationship quality is a necessary prerequisite to serve as a buffer against stress and thus improve well-being, research focused on agespecific strengths (e.g., those factors that create a strong relationship in later life) of older couples in enduring relationships is imperative.
We use the terms "enduring" and "long-term" interchangeably to reflect the ongoing nature of the relationship, but we recognize that perceptions of relationship longevity are relative, being informed by a multitude of factors such as age and personal biography. Thus, "long-term" may not reflect relationship duration per se (i.e., a set length of time), but rather partners may combine relationship duration, relationship satisfaction, and an imagined future together in their descriptions (Gabb & Fink, 2015a). Moreover, how "relationship" is defined is changing and includes couples that live apart, cohabitators, and non-monogamous partnerships. However, the inclusion of diverse coupledoms, such as domestic partners, and diversity in sexual orientation and race/ethnicity are largely absent from the literature (Chonody, Killian, Gabb, & Dunk-West, 2017). This paper aims to redress this lacuna and focus on relationship maintenance in long-term relationships in later adulthood.

Relationship maintenance
A number of factors related to relationship satisfaction have been extensively studied, such as psychological well-being (Walker et al., 2013), but a burgeoning area of research centers on relationship maintenance, which has also been found to be associated with relationship satisfaction (e.g., Ogolsky & Bowers, 2012;Stafford, Dainton, & Haas, 2000). Relationship maintenance can be defined as those actions that support the continuance of the relationship (Davis, 1973). Couples will utilize a range of behaviors to sustain their relationships (Davis, 1973), and both strategic and routine interactions are used (Dainton & Stafford, 1993). That is, behaviors can be intentionally (or strategically) enacted to maintain the relationship and/or behaviors can be used to sustain the relationship through routine interactions. For example, a strategic behavior would be to buy flowers after an argument as (intentional) reparatory relationship work, whereas someone may buy her/his partner flowers every Friday (routine) to embed continuity and demonstrate "deep knowledge" (Jamieson, 1998). In the latter scenario, the act is not being performed specifically or intentionally for relationship maintenance, but serves that role over time (Dainton & Stafford, 1993). Furthermore, maintenance behaviors, routine and strategic, support the resilience of the relationship and helps to stave off relationship dissolution (Canary, Stafford, & Semic, 2002).
A number of typologies have also been formulated to understand the behaviors involved in relationship maintenance, including affinity-seeking strategies (Bell, Daly, & Gonzalez, 1987) and relational-maintenance strategies (e.g., positivity, openness, assurances, networks, and task sharing; Stafford & Canary, 1991). Dindia and Canary (1993) outline four key purposes of relationship maintenance: 1) facilitating relationship continuity; 2) maintaining the current relationship state (e.g., trusting); 3) sustaining relationship satisfaction; and 4) providing relationship repair. The commonality amongst these definitions lies in their central theme of maintaining the relationship by ensuring that elements of the relationship are supported in the prevention of relationship decline or dissolution (Dindia & Emmers-Sommer, 2006). Thus, relationship maintenance can be understood as an exchange with one's intimate partner that acts to sustain the relationship, and factors such as commitment, equity, communication, and openness among others are elements of it.
Research indicates that relationship maintenance strategies do contribute to relational characteristics (e.g., liking one's partner, commitment), but they must be continued over time to be effective (Canary et al., 2002). In a study of primarily younger couples (i.e., 20 and 30s), everyday behaviors that reinforce the importance of the relationship were shown to contribute to both relationship quality and the sustainability of the relationship (Dainton, 2000).
The assumption advanced thereafter is that relationship maintenance behaviors for older couples play a similar role; however, there is little research evidence to substantiate such claims (Gott, 2006). A significant gap in the literature exists in terms of identifying the role those positive elements of communication patterns between partners in overall relationship quality (Boerner, Jopp, Carr, Sosinsky, & Kim, 2014;Fincham & Beach, 2010), particularly in later adulthood, which in turn support relationship maintenance.
Greater research attention regarding relationship quality and satisfaction for older couples is occurring (e.g., Walker et al., 2013), but our review of the literature did not yield any studies that exclusively focused on relationship maintenance in later adulthood. Age differences are seldom examined, and the mean age for samples in this substantive area are 20 to 40 years old (e.g., Canary et al., 2002;Dainton, 2000;Dailey, Hampel, & Roberts, 2010). Although our study included data from couples across a diverse age range, the current analyses are focused on what behaviors are used by those in later adulthood to help maintain their relationship. This examination of the role of relationship maintenance behaviors amongst older people in long-term relationships will highlight how they sustain relationship quality and thus the endurance of the relationship over time.

Correlates of relationship maintenance
Understanding the strategies or behaviors that are used to maintain a relationship is an important area of research as this provides information that may be useful for therapeutic work that supports couples in their efforts to repair their relationships. Similarly, exploring sociodemographic variables relevant to relationship maintenance also strengthens further insight into the processes that help create and maintain relationship quality. The literature documents that older couples experience a "resurgence" of positive interactions and a "decrease of negative sentiment" across the life course (Gagnon, Hersen, Kabacoff, & Van Hasselt, 1999, p. 360;Umberson, Williams, Powers, Chen, & Campbell, 2005), and that "personal relationships improve with age" (Fingerman & Charles, 2010). This increase in marital satisfaction may reflect emotional regulation associated with aging (Fingerman & Charles, 2010), but the demands of childrearing may also be a contributor (Gagnon et al., 1999).
The presence of children in the home, even adult children, likely influences the extent to which relationship maintenance occurs between older partners. Results from a growth curve analysis found that older adults (70 years old) with adult children who lived separately "exhibit a fairly stable pattern of positive experience over time," (Umberson et al., 2005, p. 501). This was similar for those 55 years old. Relatedly, positive marital experience was lower for those with adult children living in the same house, and childless couples (55 years old) had more negative marital experiences than those who had adult children living on their own (Umberson et al., 2005). Additional research indicates that parents with children living at home had lower levels of relationship maintenance than nonparents, but once the length of the relationship was controlled, these differences were nonsignificant (Dainton, 2008). However the age range for this latter study was 24 to 68, with the average age of 41.7; therefore, these results may not be generalizable to those individual in later adulthood.
Likewise, relationship satisfaction is impacted by and also influences individual well-being. Findings from a meta-analysis indicated that marital satisfaction is a strong predictor of well-being (Proulx, Helms, & Buehler, 2007), and amongst older adults psychological well-being has been shown to be important to relationship quality (Walker et al., 2013). The role of relationship maintenance and personal well-being has not been examined, but could shape the extent to which maintenance strategies are used. Another key factor is relationship duration, which is negatively associated with both relationship satisfaction (Brown & Kawamura, 2010) and relationship maintenance (Dindia & Baxter, 1987). A recent meta-analysis found that relationship duration was related to a decrease in three specific types of maintenance strategies-positivity, openness, and assurances (Ogolsky & Bowers, 2012). This may be indicative of "a general drift toward decline in marital quality with increasing years of marriage perhaps suggest[ing] that relationship maintenance and repair repertoires atrophy with time" (Dindia & Baxter, 1987, p. 146). The role of relationship length is particularly relevant when considering enduring relationships amongst those in later adulthood who may have been coupled for decades.
Gender may also contribute to relationship maintenance behaviors in that women have been socialized to place more importance on nurturing relationships and thus they may be more inclined to be proactive in sustaining their partnership (Dindia & Baxter, 1987). Yet, no significant differences in the number or type of maintenance strategies (Dindia & Baxter, 1987) and only weak associations related to perception of relationship maintenance (Stafford & Canary, 1991) by gender were found in early studies. Socio-historical context is important here, and in a more recent meta-analysis, Ogolsky and Bowers (2012) found that the types of relationship maintenance strategies differed by respondents' sex. Measurement as well as the issue of self-report plays a role in the differential findings, but further research is also needed.
A new, but relatively unstudied area, is the role of religious beliefs in relationship maintenance and quality in later adulthood (Sabey, Rauer, & Jensen, 2014). Research on religious beliefs and practices alongside shared faith has shown that religiosity may be thought of as a relationship maintenance strategy, especially for older couples (Sabey et al., 2014). Rauer and Volling (2015) found that when spirituality beliefs were shared, couples appeared to address conflict more successfully. Thus, the way that religious beliefs shape the marital relationship is essential to understanding relationship maintenance strategies. Moreover, this area of research is salient when studying relationships in older adulthood given that older adults tend to be the most religious, particularly Americans (Pew Research Center, 2015). Even in other contexts (e.g., Western Europe) where organized religion is less prominent (Pew Research Center, 2008), research indicates that older people are more likely to be religious than younger people (Pew Research Center, 2008). However, given the absence of extensive research, the extent to which religious beliefs may affect relationship maintenance amongst people in later adulthood remains somewhat speculative (Sabey et al., 2014). While these correlates (e.g., gender) are not the target of change when working with older couples that may be experiencing distress, they do contextualize general relationship patterns that highlight how relationship maintenance strategies differ amongst partners and/or have changed over time.

Current study
This substantive literature provides key insight into how relationships are maintained, yet studies that explicitly seek to investigate relationship maintenance patterns amongst older adults are lacking. One aim of this study, therefore, was to extend the knowledge base on enduring relationships in later adulthood by exploring the contributors to relationship maintenance amongst a large, international sample of community dwelling participants. Given that past studies have pointed to the role of length of the relationship (Dindia & Baxter, 1987), children at home (Umberson et al., 2005), gender (Ogolsky & Bowers, 2012), religious beliefs (Sabey et al., 2014), and well-being (Proulx et al., 2007) as contributors to relationship maintenance and/or relationship satisfaction, we have included these variables to frame the current study. Specifically, we sought to explore which factors, including sociodemographic characteristics, contribute to relationship maintenance amongst older couples and what behaviors are used to support the endurance of their relationships.

Method
Anonymous online surveys were collected in two phases; first in the U.K., and then later in the U.S. and Australia. The layout of the survey was identical for both phases of the data collection process. The first page of each survey contained all the details regarding the survey contents, the researchers' contact details, and information on voluntary participation, and the subsequent pages contained our scales and author-created items, which are described below. The questionnaire was constructed to keep respondent burden low and deployed suggested practices for online data collection (e.g., Hewson & Laurent, 2012). A panel of experts evaluated our survey to ensure that it was easy to understand, without technological glitches, and in accordance with the British Psychological Society guidelines for survey research. Consent to participate was given by survey completion. Institutional ethics panel approval was garnered prior to survey administration in the U.K. (Open University; 2011) and Australia (Flinders University; 2014), and approval from the Institutional Review Board (IRB) was garnered in the U.S. (Indiana University; 2014).
Survey Monkey was used for the format and distribution platform for both phases of data collection. We chose an online survey as it allows participants from diverse backgrounds the opportunity to participate equally, irrespective of location and availability (Chang & Krosnick, 2009). Survey administration in the U.K. was part of a large-scale mixed methods study that was funded by the Economic and Social Research Council (RES-062-23-3056). In this phase, a link to the survey was hosted on the project website and disseminated through research recruitment channels including online forums, newsletters, and community group noticeboards, especially those clustered around parenting and relationship support (e.g., "MumsNet," www.mumsnet.com and "Relate," http://www.relate.org.uk). The survey was also distributed in hard copy format via direct canvassing and among community groups, which aimed to increase participation rates amongst those without the Internet and/or from lower socioeconomic backgrounds.
In the second phase (August-December 2014), researchers extended this study to include the U.S. and Australia, which non-funded research that sought to increase participation from other countries that would provide further diversity of relationship types and experiences as well as cultural differences, including race/ethnicity and religion. Study recruitment was limited to social media (e.g., Facebook, Twitter), listservs, and sharing of the survey link through professional networks.
Data collection in the U.K. occurred from January 2012-January 2013 and generated 7,654 responses. Overall, participants were from over 60 different countries, but as a result of the targeted recruitment, this sample was predominantly from the U.K. For the second phase of data collection, 917 Americans participated in the study and 465 Australians. Once responses with completely or nearly missing data (i.e., those who opened the survey but did not complete more than a few questions) and those who indicated that they were no longer in a relationship (e.g., widowed), data for 8,132 respondents remained. We are unable to calculate a response rate for our survey given the diverse recruitment strategies that we utilized, including the use of social media and other online postings.
Only a portion of the data (n = 1,565) was utilized in the current analysis to facilitate our specific aim to gain greater understanding of relationship maintenance in later adulthood. A clear demarcation of what constitutes "older adulthood" is not always clear and is highly dependent on cultural and geographic considerations. Increased life expectancy, particularly in developed nations, is pushing the notion of "old age" as something that begins at 70 or even 75; however, research often takes a somewhat broader approach (e.g., Bookwala, 2011;age range = 57-85). For purposes of our study, a more general framework of later adulthood that included participants aged 55 + was utilized to reflect a time period when children are likely to have left home and parenting duties subsequently lessened. This is an important consideration given the role that children play in studies of intimate relationships, particularly lower levels of relationship satisfaction (e.g., Twenge, Campbell, & Foster, 2003) and relationship maintenance (Dainton, 2008).

Measures
The survey began with basic sociodemographic information followed by items measuring relationship satisfaction, relationship maintenance, and well-being.
Next, two open-ended relationship questions were presented. The survey concluded with some remaining background questions (e.g., parenthood, religion, education, employment). An archived copy of the survey can be found on the U.K. project website http://www.open.ac.uk/researchprojects/ enduringlove/methods

Relationship maintenance (RM) scale
Given that both strategic and routine relationship maintenance behaviors can help to sustain a relationship, intentionality of behavior can be challenging to assess, and a singular behavior can be used both routinely and strategically (Dainton & Stafford, 1993;Dindia & Emmers-Sommer, 2006). We therefore developed a new parsimonious relationship maintenance (RM) scale to measure behaviors that occur within a relationship and which help sustain it (Chonody et al., 2017). After assessing item performance alongside skew and kurtosis, the original sample (N = 8,132) was then randomly split into two samples in order to perform the psychometric tests (see Chonody et al., 2017). Item reduction related to poor item performance and cross loadings as identified in exploratory factor analysis (EFA) resulted in a final 8-item scale. This factor structure was supported by confirmatory factor analysis (CFA) indicating evidence of factorial validity. Another EFA was completed on this reduced sample to ensure that the factor structure was consistent with our previous analyses. Results replicated our original findings. The component matrix and the communalities alongside each item are provided in Table 1. A 6-point Likert-type scale was utilized (1 = strongly disagree, 6 = strongly agree), and the theoretical range was 8 to 40. Cronbach's alpha for this sample was good (α = .81).

Relationship and well-being variables
Three author-created, single-item indicators were utilized. General well-being was assessed with the question: "How happy are you with your life overall?" A 5-point Likert-type scale was employed (1 = very unhappy and 5 = very happy) here and for all other items unless indicated otherwise. Relationship satisfaction is likely the most studied aspect of intimate relationships and is highly relevant to investigations of relationship maintenance (e.g., Dindia & Baxter, 1987). While it is captured by a variety of scales, we sought to gauge general happiness in one's relationship as a measure of relationship satisfaction.
Respondents were asked to rate the question: "How happy are you with your relationship overall?" To probe individual commitments to various relationships, we asked: "Who is the most important person in your life?" Respondents were provided with the options: "child, partner, brother, sister, mother, father, other family member, friend, and self." Due to small sizes in some cells (e.g., other family member), a one-way ANOVA was used to determine significance between categories. Based on these results, this variable was recoded as partner, child, family member, and friend/self. Partner was used as the reference variable for the regression analysis.

Sociodemographic variables
Sociodemographic variables were included based on the literature. Gender comprised "male, female, and other." "Other" was only indicated twice, and these responses were recoded as missing for the current analyses due to small cell size. Age was measured categorically. Only those participants who indicated that they were 55-64 or 65 and older were included in our analysis. We included these two age categories to represent later adulthood; however, given that differences could be present between these two groups, age was included as a control variable in the analyses. Sexual orientation included "heterosexual, gay/lesbian, bisexual, and other." Due to small sample sizes, this variable was recoded as heterosexual and GLBQ. Parenthood and if children were still living at home were assessed by dichotomous questions (yes/ no). Religious affiliation included all major religions as well as the opportunity to self-define. For purposes of analyses, this variable was dichotomized (yes/ no). Given that religious affiliation may be too rudimentary for determining the role of religious beliefs in a relationship, participants were asked to assess this item: "Faith shapes our relationship." The length of the relationship was measured categorically (under 1 year, 1-5 years, 6-10 years, 11-15 years, 16-20 years, and 20 + years). Due to small sizes in some of the categories, a one-way ANOVA was conducted. Based on these results, this variable was dichotomized as under 1 year to 10 years and 11-20 + years. Country of residence was asked as an open-ended question. Due to small cell sizes for European countries outside the U.K., African/Asian countries, and Australia, country was recoded as U.K., U.S., and other. The U.K. was used as the reference variable in the regression analysis.

Data analysis
To determine which factors explained the variance in relationship maintenance, a hierarchical linear regression was used. Sociodemographic, relationship, and well-being variables were entered in step one. In step two, happiness with one's relationship was entered to determine its unique contribution to the amount of variance explained. Content analysis was used to analyze the qualitative data from the open-ended items. A grounded theory approach (Miles & Huberman, 1983) was deployed to organize data into themes that mapped onto the coding frame developed for use with the qualitative research data. This thematic coding was then re-entered into SPSS to determine frequencies and patterns of the data.

Preliminary analyses
Due to multiple comparisons, a Bonferroni correction was used (.05/12 =p < .004). This p-value was used to evaluate the bivariate and multivariate results given the increased likelihood of type I error due to sample size and multiple tests. All variables were checked for significance and directionality. Results of independent sample t-tests for sociodemographic variables were significant for age, being a parent, and children at home. Participants Gender, sexual orientation, religious affiliation, and relationship length were non-significant in the bivariate analyses. A one-way ANOVA by country yielded significant between group differences, F(3) = 7.16, p < .0001. In the Tukey post hoc comparison, participants from the U.S. (M = 32.16, SD = 5.55) reported greater relationship maintenance than those from the U.K. (M = 30.77, SD = 5.03); all other between-group comparisons were non-significant. A one-way ANOVA based  Pearson product moment correlations were conducted for continuous variables. All of the associations between the independent and dependent variables were significant, but none exceeded standards for inclusion in regression (r 2 < .70). Table 3 provides the coefficients. Assumptions for regression were checked. Multicollinearity was not an issue (VIF < 10, tolerance > .2), and Durbin Watson was within the acceptable range indicating that the correlations between residuals were not problematic.

Relationship quality: Multivariate results
After the control variables were entered in step one, 36% of the variance was explained, and six variables were significant. In step two, happiness with the relationship was entered into the model and uniquely contributed 17% of the explained variance. The overall model explained a total of 53% of the variance for relationship maintenance. According to Cohen's (1988) guidelines, effect sizes were small, except overall happiness with relationship, which had a medium effect. A higher degree of relationship maintenance was explained by faith shaping the relationship, indicating that one's partner is the most important person in her/his life, being American, and greater overall happiness with the relationship. Table 4 provides the results.

Qualitative results
To gain a deeper understanding of those behaviors used in the relationship that support its maintenance, two open-ended items were included in our survey. We did not directly ask participants about relationship maintenance strategies given that it could present challenges in terms of respondent burden. Instead we took a more indirect approach at assessing behaviors that were present within the relationship, which in turn we conceptualized as a component of their relationship strategies. These questions were designed to capture those strengths present in the relationship and to generate additional depth and detail. These items were: "Identify two things that you like best about your relationship" and "Identify two things that your partner does for you that makes you feel appreciated." There were no pre-selected options for these questions (i.e., answers were free text and unrestricted in length). Responses were coded thematically and then grouped into categories of behavior.

Two things that you like best about your relationship
The sample size for this item was 1,309 whereby at least one answer was given. Each "best" trait was coded separately and generated a total number of 2,466 coded responses. Companionship, being at ease with each other, and sharing a laugh represented 45% of the total responses. While some responses were straightforward and brief, such as "love" and "communication," other participants elaborated on their best relationship qualities. One participant states, I like the sense of belonging, which comes from my relationship. I also like the companionship and shared history we have plus our plans for the future together; I suppose this is the sense of continuity, which I appreciate.
Or another writes, "We are friends. We have similar views on most topics, and talk to each other about the things we have been doing." Some participants also indicated changes related to aging, such as this response: "We've grown closer since becoming empty nesters. We laugh more as we get older!" Notably, less than 1% of participants indicated "nothing" for this question, suggesting that this sample could readily identify at least one positive attribute of their relationship; however, 256 people skipped this question entirely, and they may have done so because they had nothing positive to write. Most of the participants who responded negatively to this item just said "nothing," but a few elaborated, such as this response: "It is a source of continual frustration and entrapment so I cannot identify any best elements." Table 5 provides all of the themes.

Two things that your partner does to make you feel appreciated
A total of 1,271 people provided at least one answer to this item (missing responses = 294), and the total number of coded responses was 2,350. Responses fell into two broad themes: practical support (27.9% of the responses) and emotional support (67.7% of the responses). Practical support included helping with the housework/chores, cooking, "providing a lift," and bringing cups of tea/coffee. Most responses were quite brief, but some participants were more descriptive in terms of the kind of practical support that was received and created a sense of appreciation. One participant stated, He cleans the windscreen on dark, icy, winter days at 6:30 when I was due to start an early shift at my hospital job. And he does all the DIY jobs about the house and garden as well as any heavy work to be done.
Given the proportion of respondents from the U.K., cups of tea played a prominent role in our data. One participant even stated, "She brings me cups of tea … even when I don't want them." Emotional support included saying thank you, "being there," buying cards, flowers, small gifts or other surprises, complements, cuddles, and saying "I love you." For example, one participant stated, "When he pulls out of the driveway, he mouths 'I love you.'" Some participants further elaborated on things that their partner does, such as this respondent who wrote, It is the little things that he does that show me that he is paying attention to my needs. For example, when I was still teaching school and randomly complained about my lousy pair of scissors. The next day, without asking, he bought me a new pair to use in my classroom.
Other participants had short, but powerful, responses, such as "loves me loves me loves me" and "he thanks me for everything I do." Aging also came up in some of the responses, such as this one: "He tells me that he loves me and still thinks I am pretty at 75." Similarly, longevity of the relationship was intimated in some responses. One participant wrote, "He always refers to me as his 'sweetheart' and when other people ask how long we have been married, he always says 'not long enough.'" My partner "does nothing" represented 2.3% of the responses to this item. Responses here tended to be brief like "nothing; never even a thank you" and "nothing; all communication is arguing." Again, missing data on this item could be indicative of other relationships where the participant does not feel appreciated. Table 6 provides these themes.

Discussion
Our results contribute to the substantive literature on enduring relationships in later adulthood by exploring coupledom from a strengths-based approach, which highlights what is working and the factors that are most important to relationship maintenance. The greatest influence on relationship maintenance in our regression model was overall happiness with one's relationship. These findings echo previous findings in terms of the importance of relationship satisfaction in supporting maintenance behaviors, which help sustain an intimate relationship (e.g., Ogolsky & Bowers, 2012). However, our study is the first to show this significance for people in later adulthood. We hypothesize that everyday relationship maintenance is likely to be an iterative process that may contribute to "positive sentiment override" (Gottman, 2015). That is to say, an individual may perceive their partner's behavior to be primarily positive despite behaviors that may contradict this perception. When one is happy with their partner, relationship maintenance behaviors may correspondingly increase (e.g., saying "I love you") and as a result of such behaviors, relationship satisfaction increases. In turn, relationship satisfaction then improves, so too relationship maintenance behaviors and happiness with partner -and so on and so forth. However, due to the correlational nature of our data, causation cannot be assumed as it is also likely that those individuals in highly satisfied relationships are more likely to engage in relationship maintenance. Nonetheless, the associations between these variables, coupled with past research, warrant further exploration into how such variables influence each other. Our qualitative findings underscored the importance of various types of relationship behaviors that help sustain the relationship by feeling appreciated and identifying positive elements of the relationship. Participants largely described small acts that made them feel appreciated, and these gestures were often associated with domestic labor, such as cooking a meal or cleaning off the windscreen. This was also reflected in what participants liked best about their relationship, such as having a laugh or a shared a sense of familiarity. The mundanity of things identified was also notable in that everyday activities and small acts of kindness were designated far more often than grand symbolic gestures (Gabb & Fink, 2015b;Duck, 1988). The creation of meaning is found in the mundane. "Trivial rituals of conversation actually serve to create and sustain reality for … couples," and they also act to reinforce and support the continued existence of the relationship (Duck, 1994b, p. 51). Encouraging couples to be demonstrative in small ways to express appreciation may help facilitate increased positive sentiment, and this in turn, may contribute to increased relationship satisfaction.
The ways in which our sample felt appreciated by their partners was wide ranging but emotional support represented nearly 70% of the overall responses. Being able to talk with a partner and be listened to was particularly valued in what made someone feel appreciated, and the absence of arguments and/or miscommunication featured highly in what a partner liked best about their relationship. Being able to talk with a partner and be listened to was particularly valued in what made someone feel appreciated, and the absence of arguments and/or miscommunication featured highly in what a partner liked best about their relationship. Duck (1994a) points to the fact that relationships are never a "done deal;" that is, they require ongoing action and response to one's partner. This meaning-making serves a key role in the sustainment of the couple and may be a way to strengthen the relationship. Future research should seek to explore the role of relationship maintenance in counseling couples that are seeking relationship support.
Moreover, many respondents pointed to the fact that their partner says "thank you" and expressed appreciation in small ways. Previous research indicates that expressing gratitude is related to relational maintenance (Lambert & Fincham, 2011), and is likely to serve an iterative function. "People who feel appreciated by their romantic partners report being more appreciative of their partners. In turn, people who are more appreciative of their partners … are more committed and more likely to remain in their relationship over time" (Gordon, Impett, Kogan, Oveis, & Keltner, 2012, p. 257). Researchers have been writing about the necessity of identifying these positive aspects of communication (Boerner et al., 2014), and our findings contribute to this gap in the literature, particularly as it relates to later adulthood.
With regard to sociodemographic characteristics, gender was not significant in our results, consistent with Dindia and Baxter's (1987) findings. Conversely, Ogolsky and Bowers (2012) found gender differences in the types of relationship strategies employed; however, these potential differences could not be captured by our brief and general measure of relationship maintenance. Interviews with couples in later adulthood could shed light on how gender roles may influence the types of strategies utilized to maintain the relationship as well as the use of multiple measurement approaches.
Religious affiliation was not a significant factor in explaining relationship maintenance, but sharing a faith was. Previous findings suggested that high levels of religiosity are positively correlated with relationship quality (Mahoney, Pargament, Tarakeshwar, & Swank, 2001), and attendance at religious services was associated with higher levels of happiness (Brown & Kawamura, 2010). Our findings help substantiate this evidence for later adulthood in terms of its effect on relationship maintenance. Faith may help maintain the couple's investment in the doctrines and meaning-making associated with this specific set of beliefs and values. For couples in later adulthood who may be experiencing significant life changes, such as retirement, bereavement, or illness, faith may provide both a source of individual strength and a stabilizing factor for the relationship.
While marital happiness (Brown & Kawamura, 2010) and relationship maintenance (Dindia & Baxter, 1987) have been found to be inversely related to relationship duration, it was not significant in our study. Individuals in 20 + year relationships (83%) dominated our sample, but they were also predominately happy in their relationships. Thus couples that may be together for a long-time but unhappy are underrepresented in our findings. Future research should continue to explore the connection between relationship maintenance and relationship duration as it is moderated by relationship satisfaction.
Having adult children at home also was not found to influence relationship maintenance, consistent with past finding among middle-aged adults (Dainton, 2008). However, "the most important person" in one's life was a significant variable throughout our analyses. While the effect size was small, it is perhaps unsurprising that relationship maintenance is higher when one's partner is deemed the most important. Perhaps having adult children at home was not significant in explaining relationship maintenance because the person in the home who was perceived as most important outweighs the mere presence of adult children. Additional research is needed in this area to gain greater understanding around the role of (adult) children in the home and their impact on relationship maintenance, especially for different types of families (e.g., grandparents who parenting, adult children with disabilities).
Likewise, overall happiness with life was not significant in the regression. Prior to adding relationship satisfaction to the model, this measure of well-being contributed to the explanation of relationship maintenance. It may be that relationship satisfaction is more important than perception of personal well-being when explaining relationship maintenance, particularly for relationships that are maintained over a long period of time. While past findings have indicated a positive association between well-being and relationship satisfaction (Walker et al., 2013), Proulx et al. (2007) found that the association was stronger in those couples that were together for 8 years or less. It is unclear, therefore, if the association between happiness with life, relationship satisfaction, and relationship maintenance may be different for older people or if this is a function of relationship duration per se . Further research into this issue that utilizes triangulation of data would help disentangle what is a complex set of interrelationships.

Limitations
The results should be considered within the context of study limitations. First, our sample was highly educated and primarily white. While we achieved our goal of increased diversity in terms of the inclusion of non-married partners and same-sex couples, purposive sampling, additional recruitment strategies, and developing relationships with organizations are likely necessary to achieve a more socioeconomically and racially diverse sample (Fincham & Beach, 2010). Furthermore, a random sample of couples from the community is necessary for generalization. Our study does however offer new insights into the role of relationship maintenance and relationship satisfaction in later adulthood, but replication is needed to address limitations associated with our sampling strategies.
Second, our inclusion of those under age 60 years old could be characterized as bordering on middle-aged; however, age was controlled in the regression analysis, and no differences in relationship maintenance were found between those aged 55-64 and those aged 65+. Moreover, most studies in this substantive literature focus on couples in their 20-40 s, and this study adds to our understandings of coupledom in later adulthood. Future research should seek to further investigate the role of everyday relationship work in later adulthood by recruiting individuals or couples who are 65 years and older.
Third, our study is limited by the high degree of relationship satisfaction reported by our sample. Given that contented couples are more likely to volunteer for a study on this topic (Levenson, Carstensen, & Gottman, 1993), we likely do not know much about those who have stayed in an unhappy relationship in terms of their relationship strategies or lack thereof. While we did not specifically seek to collect data from individuals who are in a happy relationship, our results add to the literature by highlighting those factors contributing to relationship maintenance. This information is essential to professional practice with older couples who may seek assistance from therapists and social workers. Future research should seek to recruit a sample of older people that includes those who are not happy in their relationship but remain committed to it to garner a clearer picture of how they are making it work.
Fourth, we utilized a newly developed instrument for relationship maintenance and a single-item indicator for relationship satisfaction. The RM scale has been tested psychometrically and shown evidence of validity and reliability, but no psychometric testing has been completed for the single-item indicator. Moreover, the use of categorical instead of continuous variables limits our ability to do additional tests with some of the variables. Fifth, other variables are likely important to the study of relationship maintenance, which were not included in our survey. In particular, socioeconomic status should be included in future research. Given that it plays a role in relationship satisfaction, investigating its role in relationship maintenance should be considered.

Conclusion
Our society is rapidly aging, and by 2025 it is estimated that half of the adult population will be over 50, representing a seismic shift in the demographics of society, which will impact every sphere of life from employment to social care (Sherwood & Faulkner, 2013). Understanding relationship maintenance behaviors in later adulthood is, therefore, essential to the promotion of active aging given the role that high quality relationships can play during this life stage by providing emotional support and acting as a buffer for stress (Walker & Luszcz, 2009). Relationships are also a key factor in determining the costs of aging to the state. Marriage breakdown in later life has implications for the amount of informal care available, how housing stock is used, and for emotional well-being (Harries & de Las Casas, 2013). In the U.K., family fragmentation is estimated to cost £46 billion per annum (Relationships Alliance, 2014), in the U.S. $112 billion (Scafidi, 2008), and in Australia $3 billion (van Acker, 2017).
Reducing the costs associated with an aging society is a pressing concern and high quality intimate relationships can play an important role (Harries & de Las Casas, 2013). Whereas issues concerning finance and health in later life are well documented, the subject of older people's relationships is, however, typically marginal in current social policy debates. Little research attention has been given to older couples, thus it is merely assumed that those factors important to young and middle-aged couples are also important in later adulthood. Extending the knowledge base on how older couples sustain their couple relationships is, therefore, vital. Our findings highlight the importance of overall happiness with one's relationship in supporting relationship maintenance, which in turn helps sustain the relationship. These findings both enrich current knowledge and also provide important insight for relationship support organizations. Future research should seek to replicate these findings in other samples in order to draw conclusions that may be useful in practice settings.

Funding
This work was supported by the .