THE EFFECT OF OBLIGATION ON RELATIONSHIP S AND WELL - BEING OVER TIME By Jeewon Oh A THESIS Submitted to Michigan State University i n partial fulfillment of the requirements for the degree of Psychology Master of Arts 201 9 A BSTRACT THE EFF ECT OF OBLIGATION ON RELATIONSHIP S AND WELL - BEING OVER TIME By Jeewon Oh This study examine s the effect of obligation on mid dle - aged adults relationships and well - being over time . Previous research has offered mixed evidence on whether a sense of obligat ion benefits or harms individuals and their relationships. Given that few studies are prospective and look at diverse close relationships, I use d longitudinal data spanning 18 years (Brim, Ryff, & Kessler, 2004) t o model whether two types of obligation predict intra - and interindividual changes in relation al and individual well - being . Latent growth c urve analyses indicated that intra - and interpersonal well - being increa sed over time for middle - aged adults . Lighter day - to - day obligation predicted higher levels of intra and interpersonal well - being at the first time point , while substantive obligation g enerally predicted lower levels of well - being at the first time point . Mostly, both types of obligation did not predict change in intra - and interpersonal well - being over time, except light obligation was associated with slower increase s in life satisfacti on and s ubstantive obligation predict ed slower increases in friend support. The se findings together important not only for their own well - being but also their relationships in adulthood . iii ACKNOWLEDGEMENTS Special thanks to my amazing committee members Drs. Bill Chopik, Amy Nuttall and Deb by Kashy for your guidance and feedback on t his thesis (and life) . I would not have done this without you! I also thank Jenna Harder and the MSU writi ng center for providing critical feedback on versions of this manuscript. I was fortunate to have met inspiring and invested mentors who influenced me to follow them into to grad school. Thank you, Drs. Nick Ellis, Jason Stornelli, Myrna Cintron and Kyungi l Kim. I am also fortunate to have loving f amily members and friends who have encouraged and challenged me to be here and continue pushing forward . I thank my parents, grandparents, Chan Ahn, HanNa Lee, Andrea Lee, Lucy Kim, Christine Chang, Joo Y oung Han , Jae - eun Choi, Peter Bae, Kali Vitek, Hollyn Formosa, Charlotte King, Kas ey McDonald, Cherie Chen, Kyungmin Kim, Alison Day, Jenna Harder, Victor Keller, Katie Leahy, Eun B it Hwang, Jayoung Koo, Kyuri Kim, Hyungjeong Lee. Finally, I thank the MIDUS team f or posting their data publicly available . iv TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ......................... vi LIST OF FIGURES ................................ ................................ ................................ ...................... vii INTRODUCTION ................................ ................................ ................................ ........................... 1 The Benefits of Obligation ................................ ................................ ................................ .......... 2 The Drawbacks of Obligation ................................ ................................ ................................ ..... 4 THE CURRENT STUDY ................................ ................................ ................................ ................ 7 METHOD ................................ ................................ ................................ ................................ ........ 8 Participants ................................ ................................ ................................ ................................ .. 8 Measures ................................ ................................ ................................ ................................ ...... 8 Obligation ................................ ................................ ................................ ................................ . 8 L ife Satisfaction ................................ ................................ ................................ ........................ 9 Depression ................................ ................................ ................................ ................................ 9 Support and Strain from Close Relationships ................................ ................................ .......... 9 Affect ................................ ................................ ................................ ................................ ...... 10 Data Analytic Strategy ................................ ................................ ................................ .............. 11 Factor Analysis of Obligation ................................ ................................ ............................... 1 1 Structural Analys e s ................................ ................................ ................................ ................ 1 2 RESULTS ................................ ................................ ................................ ................................ ...... 1 4 Factor Analysis of Obligation ................................ ................................ ................................ ... 1 4 Unconditional Models of Individual Ad j ustment ................................ ................................ ...... 1 6 Life Satisfaction ................................ ................................ ................................ ...................... 1 6 Depression ................................ ................................ ................................ .............................. 1 7 Unconditional Models of Relationship Quality ................................ ................................ ........ 1 7 Family Relationship s ................................ ................................ ................................ .............. 1 7 Partner Relationships ................................ ................................ ................................ .............. 1 7 Relationships with Friends ................................ ................................ ................................ ..... 1 7 Conditional Models of Individual Adjustment ................................ ................................ .......... 18 Life Satisfaction ................................ ................................ ................................ ...................... 1 8 Depression ................................ ................................ ................................ .............................. 1 8 Conditional Models of Relationship Quality ................................ ................................ ............ 1 8 Family Relationship s ................................ ................................ ................................ .............. 1 9 Partner Relationships ................................ ................................ ................................ .............. 1 9 Relationships with Friends ................................ ................................ ................................ ..... 1 9 Summary of Results ................................ ................................ ................................ ................... 20 DISCUSSION ................................ ................................ ................................ ................................ 21 Do Intra - and Interpersonal Well - being Change over Time? ................................ ..................... 2 1 Effects of Obligation ................................ ................................ ................................ .................. 2 2 v Why I s L ight Obligation Associated with Positive Outcomes? ................................ ............. 2 2 Why Is Substantive Obligation Associated with Mostly Negative Outcomes? ..................... 2 3 Limitations and Future Directions ................................ ................................ .............................. 2 5 CONCLUSION ................................ ................................ ................................ .............................. 2 9 APPENDICES ................................ ................................ ................................ ............................... 30 APPENDIX A: Tables ................................ ................................ ................................ .............. 31 APPENDIX B: Figures ................................ ................................ ................................ ............. 4 3 REF E RENCE S ................................ ................................ ................................ .............................. 4 6 vi LIST OF TABLES Table 1 . Obligation M easure and Geomin Rotated Factor Loadings for All Items ....................... 3 2 Table 2 . Comparing Model Fit Indices between Various Factor Analytic Models ...................... 3 3 Table 3 . Means, SD s and Correlations among Obligation Items, Positiv e a nd Negative Affect ... 3 5 Table 4 . Means, SD s and Correlations as a Function of Wave of Data Collection ...................... 3 6 Table 5 . Model Fit Indices for Unconditional No - growth and Linear Models by Outcome ......... 40 Table 6 . Path Coefficient Estimates from the Latent Growth Cur ve Models in which Light and Substantive Obligation Predicts Levels and Changes in Outcomes ................................ .............. 42 vii LIST OF FIGURES Figure 1 . Path diagram of Two Factor Obligation Predicting Positive and Negative Affect ....... 4 4 Figure 2 . Generic Path Diagram of Conditi ona l Models ................................ ............................... 4 5 1 INTRODUCTION We feel a sense of obligation to many people in our lives our spouses, our families, and our friends. T h is obligation is one of the many things that distinguishes close relationships from relationships with strangers. Howev er, obligation can often be a burden and a source of great stress to the individual as well. Is obligation ultimately good for us and our re lationships? Previous research has offered limited evidence on whether a sense of obligation improve s or hinder s rel ati onships over time and has focused on adolescents and emerging adults or older caregivers . In the current study of approximately 7,000 mid dle - aged adults , I examine the effects of obligation on close relationship s and on the well - being of individuals acr oss 18 years . Researchers so metimes describe obligation as the glue that connects individuals through duties and a sense of responsibility in the ir relationships (Stein, 199 2) . In many relationships, obligation is viewed as a sense of duty to reciprocate to equally giv e and tak e from a relationship (Neufeld & Harrison, 1995; Stuifbergen & Van Delden, 2011) . Voluntary relationships, such as friendships, are often characterized by this obligation to exchange resources and assistance ; i n a qualitative study o n r eciprocity and caregiving , one respondent highlighted the importance of reciproc ity i n close relationships (Neufeld & Harrison, 1995, p. 354) . Other respondents likewise agreed that they pursue d and maintain ed relationships with friends onl y w hen there wa s a sense of reciprocity . However , family relationship s are largely involuntary, meaning people do not get to choose who their parents and siblings are , and involuntary relationships seem to have different expectations for reciprocity. For in stance, in parent - child relationships, although parents often provide a great deal of support to their children, the extent to which children need to reciprocate 2 as adults is oftentimes unclear and unexpected (Stuifbergen & Van Delden, 2011) . In the field of psychology, researchers have long studied the degree to which close relationships reflect reciprocal rules (i.e., exchange; Trivers, 1971) or unfettered giving and receiving (i.e., communal; Clark & Mills, 1979) . While obligation reflects recip roc ity norms in voluntary relationships , filial obligation may a rise from the sense of belonging and connectedness of two related individuals (see Stuifbergen & Van Delden, 2011for a review on theories of filial obligation) . Even people who strongly endorsed recip roc ity in relationships expressed unique ties with family members that allow them to tolerate a lack of reciprocity for an extended period of time (e.g. caregiving situations; Neufeld & Harrison, 1995) . With or without reciprocity, obligation frequently seems to be the glue that holds some of thei r r elationships together . Yet , it also seems that too much obligation can have adverse effects for individuals and their close relationships (e.g. Tedgård, Råstam, & Wirtberg, 2018) . A survey of the existing literature provides a mixed portrait of the role of obligation o n b oth individuals - being and the quality of their relationship s . In the sections below, I review evidence for whether a s ense of obligation is beneficial or harmful for individuals and their relationships. The Benefits of Obligation Some studies su gge st that family o bligation benefits both relationships and the individual s within them . Many researchers studied a dolescents and their relationships with their families in different contexts and cultures (Macfie, Brumariu, & Lyons - Ruth, 2015) . For example, Fuligni, Tseng, and Lam (1999) found that a sense of obligation in relationships du ring adolescence is related to positive relationship quality with friends and family. They measured family obligation in 10th and 12 th graders from diverse backgr oun ds by assessing the 3 views on how much they current ly assist (e.g., helping and spending time) , respect , and expect to provide support to their families in adult hood . A dolescents with a higher sense of obligation felt closer to their parents an d s ought more advice from family members. A dolescents with stronger familial obligation also had more positive peer relationships seeking more advice and spending more time with their peers. Furthermore, h aving a strong sense of obligation seemed to help a dol escents connect with friends who share d similar values and beliefs regardi ng their family. Beyond improving relationships, family obligation has a series of other benefits. In the aforementioned study of adolescents, those with stronger family obligat ion reported higher academic motivation, more time studying and more ambitious dreams for their future (Fuligni et al., 199 9) . These ancillary associa tions between obligation and non - relationship outcomes can lead to enhance d relationship quality. For instance , a cademic success can be one way for adolescents to fulfill their obligation toward their famil ies assuming it all ows adolescents to achieve sustainable careers, and further provide for their famil ies in the future and show respect for their parents (Fuligni et al., 1999; van Geel & Vedder, 2011) . Family o bligation is also associated with many benefits, including but not limited to better school adjustment , fewer behavioral problem s (van Geel & Vedder, 2011) , and higher life satisfaction (Hooper, Tomek, Bond, & Reif, 2015; King & Ganotice Jr, 2015) . The sense of duty and responsibility likely motivate s adolescents to obey their par ents , leading to fewer behavioral problems and better school adjustment (van Geel & Vedder, 2011) . Furthermore, in a study of Chinese - A mer ican adolescents , those who reported higher family obligation also reported fewer depressive symptoms 2 years later (Juang & Cookston, 2009) . This suggested that even though obligation decreases in adolescence, its protective effects persisted over time. Thus, obligation was shown to have multiple benefits for adolescents. 4 Even into adulthood , ob ligatio n c ontinues to exert a positive influence on relational commitment . In spousal relationships, obligation is intertwined with commitment ; such that (Nock, 1995) . Greater investment and commit men t predict better r elationship functioning and foster relationship maintenance behaviors, ultimately helping relationships last (Arriaga & Agnew, 2001) . The association between obligation and commitment can also be e xte nded to organizational settin g s . E i se nberger, Armeli, Rexwinkel, Lynch, and Rhoades (2001) found that a sense of explains the effect of perceived organizational support on commit ment to the organization and job performance . When employees receive su pport from an organization, they believe they should care about the growth and goals of that organization . This feeling of investment then may create a sense of emotional attachment to an organization. Altogether, there are many examples of obligation e nha ncing relationships between not only people but also organizations. The Drawbacks of Obligation Although obligation benefit s individuals and their relationships in multiple ways, it can also be a burden , creating strain for individuals and their relat ion ships . This burden can appear as early as in childhood in the form of parentification . Parentification is when children assume too much responsibility in a family and carry out roles traditionally meant for adults, 2002; Hooper et al., 2015; Nuttall & Valentino, 2017) . Holding developmentally inappropriate emotional and/or instrumental responsibilities may lead children to assume their role s in relationships are about giving care rather than receiving ca re and form insecure attachment relationships with caregivers . Even studies that show positiv e outcomes of 5 obligation also find t hat obligation can be simultaneously associated with a host of negative outcomes. For instance, although Fuligni et al. (1999) found multiple positive effects of obligation on relationships and academic motivat ion , they also found that these same high - obligation students rec eived some of the lowest grades at school compared to the other groups , even with higher levels of academic motivation . The authors speculate that too much obligation may be inappropriate and harmful in the case of academic achievement, possibly because ef forts to fulfill other family responsibilities limit adolescents from focusing on school work . In fact , Hooper et al. (2015) also found that parentificatio n experiences as children were associated with greater depression and lower well - being in college studen ts , again showing that too much obligation can be harmful . We learn from the adult caregiving literature that a sense of obligation is particularly st renuous and stressful for adults serving as caregivers to their parents, partners, or children . Around the world, informal care as opposed to institutional care is carried out by spouses and adult children , frequently out of obligation (Butler, Turner, Kaye, Ruffin, & Downey, 2005; Australian Bureau of Statistics, 2008 as cited in Cash, Hodgkin, & Warburton, 2013; Cicirelli, 1993) . Informal caregiving is quite common in some regions , and t he prolonged responsibiliti es can be psychological ly strenuous for caregivers . Some estimates suggest that up to half of adult caregivers report significant levels of burden and depression , whether it be caring for their spouses or parents (Butler et al., 2005) . Because c aregiver obligation is associated with a greater sense of burden and d epression (Cicirelli, 1993; Stein, 1992 in men) , and depression is a risk fac tor for lower relationship satisfaction (Whisman, Uebel acker, & Weinstock, 2004) , it is p ossible that burdensome ob ligation could also negatively affect relationships through increases in stress and depression. However, situations like c aregiving and parentification contain unique 6 stressor s . Less clear is how feelings of obligation are related to important outcomes for individuals in less strenuous situations . Taken together, previous research suggests that while obligation is generally associated with benefi ts , obligation may also be associated with negative outcomes for individuals . However , most studies focus on the effect of obligation on individual functioning, rather than its effects on relationship s (see Fuligni et al., 1999, for a rare exception) . T he current stud y extend s previous research to examine whether obligation is beneficial or harmful for a variety of relationship s . Further , g iven the few prospective tests of the role of obligation on relationship outcomes across the lifespan, there has been l ittle attention paid to how obligation influences 1) our relationships after adolescence , when people are more autonomous in h ow they spend their time and invest in their relationships and developmentally able to perfor m those responsibilities and, 2) other close relationships beyond parent - child relationships (e.g., friends and partners ). I addressed these gaps in the current st udy by employing a longitudinal sample of middle - aged adults and sampling a wide range of relationships to study how obligation affects relationship s and the people within them. 7 THE CURRENT STUDY The goal of this study is to investigate the longitudinal e ffect of obligation on intrapersonal and interpersonal well - being over tim e using a large panel study of midlife adults in the United States (Brim et al., 2004) . To accurately exa min e structural relations between variables . it was important to first examine the quality and reliability (i.e. , their factor structure) of the measures used . Therefore, the current st ud y undertakes two efforts. The first part of the study describes factor analyses on the main construct of interest obligation and the second part of the study shows the structural analyses (i.e., how obligation affects outcome var iables) using a structura l equations modeling framework. 8 METHOD Participants Participants were from the National Survey of Midlife Development in the United States (MIDUS; Brim et al., 2004) . The first wave of the MIDUS study (MIDUS I , 1995 - 1 996) consisted of 7,108 English - speaking adults in the U .S. ( M age = 46.38, SD = 13.00 , range: 20 - 75 51.1% Female; 9 0.7% White, 5.2% Black/African American, 4.1% other race/ethnicities, Mdn Education = 1 - 2 years of college). Regarding the f ollow up assessments , wave two (MIDUS II, 2004 - 2005) retained 69.82% (n = 4,963) from the first wave, and wave three (MIDUS III, 2013 - 201 4) retained 46.34% of MIDUS I (66.37% of MIDUS II; n = 3,294). Compared to participants with only one wave of data, those with two or more waves were more educated ( d = .35), received more support from their partner s ( d = .17) and other family members ( d = .14), and received less strain from partners ( d = .14), family ( d = .07) and friends ( d = .08). Those who had longitudinal data and those who did not were otherwise comparable on other variables (e.g . , obligation; p = .10). Measures O bligation. O bli gati on was assessed once at the first wave of data collection (MIDUS I ) . P articipants responded to eight statements or hypothetical situations to which participants indicated how obligated they would feel on a ; see Table 1 for a full list of items ). Among the eight total statements, three asked about children, three about friends, one about parents , and one about spouses. Part icipants rated how much obligation they would feel in each situation on a scale of 0 ( no obligation ) to 10 ( very great obligation ) . R atings for eight situations have generally been summed or averaged to yield a normative obligation 9 score ( = .82) . Other times, a simplified four - item version was used (Grzywacz & Marks, 1999; = .79). Life Satisfaction. Satisfaction with life was assessed at all three time points using five items that assess satisfaction in different domains . Each item asked participants to rate their overall satisfaction with r espect to their life, work, health , and relationship with spouse/partner (if applicable), and relationship with child ren (if applicable) . After comput ing an average relationship satisfaction score from ratings of relationships with spouse/partner and child ren, the four ratings were averaged to calculate an overall life satisfaction scor e which ranged from 0 ( the worst po ssible ) to 10 ( the best possible ) . Depression. Depression was assessed at all three time points (Wang, Berglund, & Kessler, 2000) . P articipants answ ered yes or no to two seven - item subscales : depressed affect two weeks in past 12 months, when you lost interest in most things, did you feel more items, the two subscales were averaged. Therefore, the final measure of depre ssion ranged from 0 to 7. S upport and S tra in from C lose R elationships . Measures of s ocial support and strai n from spouses, family members, and friends were used to capture the quality of relationship with those individuals (Schuster, Kessler, & Aseltine, 1990; Walen & Lachman, 2000) . Suppo rt and strain were assessed at all three waves . Six questions assessed the amount of support participants perceived from their spouse/ partner ( e.g. ; six questions assessed the amount of strain participants perceived from spouse/partner ( e.g. 10 often does he or she make you feel tense Eight questions assessed how much support participants perceived from family members and friends fo ; four questions for family members and four questions for friends ) . Eight qu estions assessed how much strain participants perceived from family members and friends ; four questi ons for family members and four questions for friends ) . Questions were skipped if participants thought they were no t relevant to them (e.g., single individuals did not answer questions about spouses/partners). Participants responded to each question on a s cale ranging from 1( a lot ) to 4( not at all ) . All r esponses were reverse - scored and then averaged to yield composite s for spousal support ( = . 86 ), spousal strain ( = . 81 ), family support ( = . 82 ), family strain ( = .80), friend support ( = .88) and friend strain ( = .79 ; all alphas at MIDUS I ) . Higher scores indicate grea ter support and strain. Support and strain were examined as distinct scales because previous factor analyses suggest ed that they were distinct constructs (Chopik, 2017) . Affect. Positive and negative affect at the first wave were used in the context of the factor analyses to establish some discriminant validity for the obligatio n measure should separable components were identified . If there were multiple factors that were differe ntially related to adjustment, it was expected that they might at least differ on predicting positive/negative affect reported in the past 30 days. A tot al of twelve items measured positive and negative affect; s ix items were positive (e.g. in good spirits, satisfied ) and six items were negative (e.g. nervous, so sad nothing could cheer you up ) . Participants reported how often they had felt each of the twe lve emotions in the past 30 days, on a scale from 1( all of the time ) to 5(n one of the time ) . All responses were reverse - scored so that higher scores indicate more of the certain affect. 11 Responses were then averaged to yield composites for positive affect ( = .91) and negative affect ( = .87). Data Analytic Strategy Factor Analysis of Obligation . MIDUS measured obligation as a unidimensional construct, summing or averaging all items to yield a single composite score. However, this may not be the best appr oach as the measure (a) asks about situations that reflect different levels of obli gation and sacrifice and (b) assesses obligations to different people spouses, children, parents and friends. Therefore, it might be the case that there is more than one lat ent factor being represented by the measure, such that items measuring obligation t owards a spouse predict combined score would be misleading in such cases. F urther, using unreliable measures would be problematic for the following structural analyses. I use goodness - of - fit indices to evaluate whether and how obligation was associated with initial levels and changes in intra - and interpersonal well - being . Howeve r, many common fit indices (e.g. RMSEA) evaluate the fit of the entire model. The f it of the measurement portion of the model to the data (which generally takes up more degrees of freedom in the model) may overpower the fit of the structural portion of the model (Lance, Beck, Fan, & C arter, 2016) . In other words, a poor measure may hide a good structure and the str uctural results may be misleading or simply uninterpretable or vice versa . In order to ensure the following structural analyses are valid, it is important to first evaluate the psychometric properties of any multiple - item testing instrument. Factor analysi s is a common procedure to assess whether multiple items in a measure are reasonable indicators of the underlying construct (Brown, 2014; Floyd & Widama n, 1995) . Therefore, I conduct exploratory and confirmatory factor analyses (EFA/C FA) 12 to test whether obligation can be treated as a unidimensional measure before proceeding with examining predictive associations. Analyses were conducted in Mplus 8.1 (Muthén & Muthén, 2017) using full i nformation maximum likelihood estimation to handle missing data. Factor models were identified by constraining factor variance to 1. Model fit was assessed using multiple goodness - of - fit indices: 1) non - 2 (howe ver, this metric is often overly sensitive when examined in large samples like ours; Bentler & Bonett, 1980) , 2) comparative fit index (CFI > .95; Hu & Bentler, 1999) , 3) root mean square error of approximation (RMSEA confidence interval < .08 fair fit; MacCallum, Browne, & Sugawara, 1996) , 4) Tucke r - Lewis index (TLI>.95), and 5) standardized root mean squared residual (SRMR <.08; Hu & Bentler, 1999) . A one - factor confirmatory factor analysis (CFA) tested the eight - item obligation measure, given that previous research operationalized it as a unidimensional measure. Structural Analys e s In the second part of the stud y, I examine a series of unconditional and conditional models to model intra - and interindividual changes in relationship quality (operationalized as support and strain from relationships) and individual well - being (life satis faction and depression) predic ted by obligation (whose factor structure was defined in the first part of the study). amount of support/strain from various close relationships (spouse/p artner, other family members, and friends), life satisfaction and depression were assessed three times across eighteen years. L atent growth curve modeling techniques were used to investigate changes in an individual and relational well - being across eight een years . This approach allow s modeling of both intra - and interindividual changes in the variables of interest (Baltes & Nesselroade, 1979; 13 Grimm & Ram, 2012; see Nuttall, Valentino, Wang, Lefever, & Borkowsk i, 20 15, for a similar approach) . First, I tested a series of competing unconditional models to determine overall patterns of change over time in each of the outcomes life satisfaction, depression , and relationship specific support / strain . Both an intercept - on ly and a linear model were tes ted, and I retain ed the model that best described the data (using the afor ementioned criteria and the 2 difference test ). The first model was an intercept - only model with three parameters (intercept mean, intercept variance, and residual variance). The second model was a linear model with six parameters (intercept and slope means, intercept and slope vari ances and their covar iance, and a residual variance). Life satisfaction, depression, support and strain were centered at the first wave of data collection and scaled so that estimated intercepts could be interpreted as an average score at wave one and esti mated slopes interpre ted as an average unit change per wave (MIDUS I = 0, MIDUS II = 1 , MIDUS III = 2 ) . Next, I tested eight conditional models (for each outcome variable : life satisfaction, depression , three relationship - specific support measures and thr ee relationship - specific s train measures ) where the intercepts and slopes of each outcome were modeled as conditional on obligation. Since obligation was measured only once, it was treated as a time - invariant predictor. 14 RESULTS Factor Analysis of Obligat io n Descriptive statistics o f study variables related to the factor analyses (e.g. , mean s of the obligation item s ) are presented in Table 3 . A total of 5446 participants in the sample provided full or partial data on the obligation measure. The one - factor c onfirmatory factory analysis (CFA) suggested that a one - factor solution was not appropriate for this measure. Model fit was poor (Table 2 ) ; no model fit indices reached the assessment criteria (SRMR was on the edge) , and standardized residuals suggested a systematic pattern of misfit 1 . To follow up examining the factor structure of the measure, I conducted an exploratory factor analysis (EFA; Brown, 2014) . First, the entire sample was randomly split into two subsamples in order to explo re the factor structure (EFA) in one sample and test/confirm a final factor solution (CFA) in a separate sample based on the final EFA solution. Since the measure asked quest ions about four types of relationships, it seemed reasonable that there could be u p to four underlying factors each characterizing a different type of relationship. However, two of the relationships (with parents and spouses) each had only one item associa ted with it, while factor analyses require at least three indicators for each fact or (T. A. B ro wn, 2014) . Therefore, the present eight - item measure calls for a comparison between a one - factor solution and a two - factor solution (to allow for the minimum number of ind icators for each factor). E FA suggested the two - factor solution was the best way t o proceed . T he eigenvalues suggested a two - factor solution as indicated by two eigenvalues (3.65 and 1.21 ) greater than 1 , 1 The systematic misfit appeared to orig inate from model parameters underestimating the association between particular obligation items but overestimating those items and other items (e.g. the association between 1 and 2 was underestimated while associations between obligation items 1 and 3, and 2 and 3 were overestimated) 15 and m odel fit indices were acceptable ( see Table 2 for all indices) 2 . Geomin factor loading s for this two - factor solution are prese nt ed in Table 3 . The factor loadings for the two - factor solution showed that all items, except item 5, loaded at least moderately on one of the factors (range: .38 - .96). Items 1, 2, 4, and 7 loaded on one factor characterized by situations that require lig ht er, day - to - day obligations toward family members (children, spouse or parents). Items 3, 6 and 8 loaded on another factor characterized by situations that call fo r more substantive caregiving toward friends. Item 5 moderately cross - loaded on both factors ( .39 and .45). This was reasonable given that the item is about substantive caregiving for adult children , which may have overlap ped between one factor that tapped into family - type relationships and another factor that tapped into substantive caregiving . Du e to the conceptual overlap and the cross - loadings item 5 was excluded for future analyses. After excluding item 5, an EFA was rerun on the seven - item obligation measure in the same sample following recommendations by Brown (2014). Model fit improved, an d the two - factor solution was retained. I then used a CFA on the second random subsample to replicate the final EFA solution. Results indicated good/fair fit using CFI and SRMR, and approaching fair fit using other indices ( see Table 2 ). Therefore, the t wo - factor solution (substantive and light ( obligation) excluding item 5 was selected as the final factor structure. 3 Since not all fit indices indicated good fit, the factor analyses were further extended using exploratory structural equation modeling to examine whe ther substantive obligation and light 2 Although a three - factor solution also fit the data well, 1) interpretation of eigenvalues did not support a three factor solution and 2) factor loadings for the three factor solution were not interpretable because the loading s patterns were weak (i.e. one factor had only one indicator strongly loading on i t (loading =.61) ) (see T. A. Brown, 2014 for eval uating the quality of factor solutions) . 3 A two - factor CFA was tested once more on the full sample as a final confirmation following T. A. Brown (2014) . Model fit indices showed similar or expected sample; Bentler & Bonett, 1980) results compared to the two - factor CFA on the split sample. 16 obligation differentially predicted outcomes. Positive and negative affect at time 1 ( MIDUS I) were chosen as outcomes , because they are likely to be associated with obligation ( Telzer & Fuligni, 2009 ) and extensive re search suggest ed that they are separate constructs , allowing us to see how two factors predict ed different outcomes (e.g. Diener & Emmons, 1984) . Substantive obligation and light obligation were allowed to covary; positive and negative affect were also allowed to covary. Fit indices 2 = 838.09 , df = 23 , p <.001 , RMSEA = .08 [.076,.085], CFI = .94, TLI = .91, SRMR = .04). Results showed that light obligation toward family members predicted more positive affect ( b = .11, p <.001) and less negative affect ( b = - .11, p <.001) . S ubstantive obligation toward friends predicted more negative affect ( b =.05, p <.001) but was not significantly related t o positive affect ( b = .01, p > .51 ; see Figure 1 for a path diagram with standardized coefficients ). This differential pred ictive power of light obligation and substantive obligation suggested that two factors of obligation are different both with respect to their properties (as ascertained in the factor analyses ) and in potentially predicting outcomes differently , which I ret urn to in my conditional model analyses . Unconditional Models of Individual Adjustment Means, standard deviations and correlations among variables pertaining to the structural analys e s are presented in Table 4. Table 5 presents model fit indices for eight unconditional latent growth curve models (individual adjustment and relationship qualit y for each relationship ) 2 difference test s between intercept - only and linear models. Life Satisfaction. Model fit for the intercept - only model and linear model were both acceptable, but the linear model better described the data . Although the estimated avera ge linear rate of change was not significant ly different from 0 ( b = .00 7 , SE = .0 1 1, p =.53 , =.020 ) , the estimated variation in the slopes , .130 ( SE =.016 , p <.001) , was different from 0. This suggests 17 that there were individual differences in how life satisfaction changed across the 18 years of the study. Further, 15.58% of within - person variation in life satisfaction is explained by adding the slope parameter compared to the intercept only model. T herefore , the lin e ar model was selected as the final mo del . Depression. The intercept - only model fit poorly according to multiple indices. However, t he linear model fit well and significantly better than t he intercept - only model . On average, depression decreased by .0 87 per wave ( SE = .169, p < .001 , = - .211 ) . This slope of depression also significantly varied by . 169 ( SE = . 038 , p < .001) , highlighting individual differences in changes in depression over time . Unconditional Models of Relationship Quality The linear models fit significantly better t han intercept - only models for all relationship variables and hence were selected as the final models. In short, support in each relationship tended to increase over time; strain in each relationship tended to decrease over time; there were significant indi vidual differences in the rate of change in each facet of relationship quality . Family Relationships . On average, family support increased .035 per wave ( SE = . 022 , p < .001 , =.234 ) and the rate of increase significantly varied across individuals by .022 ( SE = .003, p < .001). F amily strain decreased on average .0 76 per wave ( SE = .00 6 , p < .001 , = - .483 ), and the slope varied by .025 ( SE = .004, p < .001) . Partner Relatio nships. On average, partner support increased .014 per wave ( SE = .006, p = .02 , = .084 ) and varied across individuals by .027 ( SE = .003, p < .001). Partner strain on average, decreased .0 51 per wave ( SE = .006, p < .001 , = - .300 ), and the slope signi ficantly varied by .02 9 ( SE = .004, p <. 001). Relationships with Friends . On average, friend support increased .027 per wave ( SE 18 = .006, p < .001 , = .169 ) and the rate of increase varied by . 025 ( SE = .004, p < .001). Friend strain on average, decrease d .101 per wave ( SE = .005, p < .001 , = - .875 ), and the rate of decrease varied by .013 ( SE = .003, p < .001). Conditional Models of Individual Adjustment After selecting linear latent growth curve models of life satisfaction and depression, two factors of obligation were added as predictors of life satisfaction and depression. All path coefficient estimates, standard errors , and p - values are presented in Table 6. Figure 2 presents a generic concept ual path diagram that applies to all individual adjustme nt outcomes. Life Satisfaction. The life satisfaction model showed good fit : 2 = 994.509, df = 33, p < .001 , CFI = .940, RMSEA = .067, TLI = .918, SRMR = .040. Light obligation was associated with greater life satisfaction at the first wave (i.e. intercept). Substantive obligation was associated with lower life satisfaction at th e first wave. More light obligati on was associated with a smaller increase in life satisfaction over time, but obligation was otherwise not significantly related to changes in life satisfaction. Depression. The depression model showed good fit 2 =1001.20 9, df = 33, p < .001 CFI = .931, RMSEA = .064, TLI = .906, SRMR = .040. Light obligation was associated with less depression at the first wave (i.e. intercept). Substantive obligation was associated greater depression at the first wave. Both f actors of o bl igation were unr elated to changes in depression . Conditional Models of Relationship Quality Linear models were selected as the model for the relationship quality measures predicted by two factors of obligation. All path coefficient estimates, standard err ors and p - values for conditional models predicting relationship quality are presented in Table 6. Figure 2 presents a generic path diagram that applies to all relationship outcomes. 19 Family Relationships. 2 = 1134 .107, df = 33, p <.001 CFI = .930, RMSEA = .072, TLI = .905, SRMR = .042. T he family strain model also 2 = 984.934, df = 33, p <.001 CFI = .937, RMSEA = .067, TLI = .915, SRMR = .040. Light obligation only affected the intercepts; higher levels of light obligation were associated with more support and less strain in family relationships at the first wave . Substantive obligation was only associated with more family strain at the first wave (i.e., the intercept of strain) . Partner Relations hips. The partner support model showed acceptable 2 = 1237.894, df = 33, p <.001 CFI = .919, RMSEA = .076, TLI = .890, SRMR = .064 . T he partner strain model also showed acceptable 2 = 1110.340 , df = 33, p <.001 CFI = .9 29 , RMSEA = .0 72, TLI = .9 04 , SRMR = .0 50 . Obligation affected partner sup port/strain in an almost identical way as it affected family support/strain. Light obligation only affected the intercepts; light obligation was associated with more support and less strain at the start of data collection. Substantive obligation was associ ated with lower partner support at the start of data collection but unrelated to levels of partner strain or changes in support/strain over time . Relationships with Friends. The friend 2 = 1007.628, df = 33, p <.001 , CFI = .938, RMSEA = .068, TLI = .915, SRMR = .046. T he friend strain model also 2 = 953.978, df = 33, p <.001 CFI = .937, RMSEA = .066, TLI = .915, SRMR = .041. L ight obligat ion was associated with more support and less strain initially. Interestingly, substantive obligation predicted more friend s train and s upport initially. Further , substantive obligation was associated with the slope , such that more substantive obligation p redicted smaller increase s in friend support over time. Obligation was otherwise unrelated to 20 changes in friend relationships over time. 4 Summary of Results Both intra - and interpersonal well - being increased over the 18 - year study; participants increased in life satisfaction, decreased in depression, while support with each relationship increased, and strain with each relationship declined. The l ight obligation factor was generally associated with positive outcomes people with higher levels of light obliga tion reported greater life satisfaction, lower depression, and more support/less strain from their families, spouses, and friends at the first time point . The s ubstanti ve obligation factor was generally associated with negative outcomes people with higher levels of substanti ve obligation reported lower life satisfaction, greater depression, less support (from their partners), and more strain (from their families and friends). Obligation was generally unrelated to changes in intra - and interpersonal well - bei ng over time , except light obligation predicted slower increases in life satisfaction and substantive obligation predicted slower increases in friend support . 4 In a series of supplementary analyses, I exa m ined linear age - based growth models (eight unconditional and eight conditional models) because parameter estimates may be biased when using waves as the time metric (Coulombe, Selig, & Delaney, 2016) . I used the definition variable approach, where changes are tracked against age at each wave (Grimm, Ram, & Estabrook, 2016) . Age was c entered at the youngest age at the first time point (age 20). These models showed slightly different results. Most slope variances across unconditional models were no longer significant, and only one type of obligation significantly predicted the intercept for a certain outcome. For example , the estimated average linear slope of life satisfaction was significantly different from 0 ( b = .006, SE = .001, p < .001), but there was no significant variance in the slopes ( b < .001, SE < .001, p =.56). Nevertheless , light obligation was still associated with greater life satisfaction at the youngest age (i.e. intercept; b = .343, SE = .080, p < .001), but light and substantive obligation were otherwise unrelated to intercepts and slopes in life satisfaction. Due to issues with convergence in some of the models (25% of the models did not converge), it was difficult to get a comprehensive understanding of the effects of light and substantive obligation. Therefore, I do not devote more space to interpreting these result s. 21 DISCUSSION The current study examined the association s between substantive obligation on intra - and interpersonal well - being and light obligation on intra - and interpersonal well - being across 18 years of adulthood . I conducted a series of factor analyses on the obligation measure, which revealed that the measure is more complex that is previously u nd erstood. Instead of being a unidimensional measure, there were two underlying factors . L ight obligation involved arguably easier day - to - day activities (e.g. calling parents regularly) . Substantive obligation involved strong commitment that would create l on g - a child of a friend) . These two factors diff erentially predicted intra - and interpersonal well - being . While light obligation benefitted people and their relationships, substantive ob li gation was associated with more negative outcomes for people and their relationships. Do I ntra - and Interpersonal Well - being Change over T ime? Both intra - and interpersonal well - being increased over the 18 - year time period . Similar to previou s work on ch an ges in life satisfaction (Baird, Lucas, & Donnellan, 2010; Gana, Bailly, Saada, Joulain, & Alaphilippe, 2012) and depression over time (Chopik & Edelstein, 2018) , the current study also found th at life satisfaction increased and depression decreased. Unlike individual well - being variables, s upport and strain from rel at ionships have not received as much attention as outcome variables (see Walen & Lachman , 2000 for a study that used support and strain as predictors) and with respect to whether and how they change over time , particularly among middle - aged adults . The current study found th at across adulthood, support increased and strain decreased across all relationship s. T he se results are consistent with other research on older adults experienc ing more positive emotions and relationships because they optimize positive interpersonal exch an ges by avoid ing conflicts , improving in social expertise and experience, and 22 affiliating with people who treat them more positively (Carstensen, Isaacowitz, & Charles, 19 99; Luong, Charles, & Fingerman, 2011) . Al th d over time, people also differ ed in their levels and changes of well - being , and obligation largely predicted differences in levels at the first wave . Effects of O bligation Previous research on obligation showed many b enefits of obligation for adolescents including better school adjustment , life satisfaction , and higher quality family r elationships (Fuligni et a l., 1999; Hooper et al., 2015; van Geel & Vedder, 2011) . H owever, o b ligation was not uniformly positive for individuals. Research often finds a , meaning that both too much or too little render negative outcomes in adolescents (Fuligni et al., 1999) . When people (e.g. parentified children or caregiving adults) fe e l too obligated to carry out responsibilities beyond their capabilities, obligation is associated w it h lower well - being (Cicirelli, 1993; Hooper et al., 2015) . The current study suggest s a more nuanced view of how obligation affects indivi du al well - being . L ight obligation predicted higher levels of life satisfaction and lower levels of depression ; substantive obligation predicted lower levels of life satisfaction and higher levels of depression . In other words , light obligation was associat ed with positive outcomes while substantive obligation was associated with more negative outcomes . The current results might explain why some studies have found curvilinear effects of obligation when they did not separate obligation into distinctive types (Nuttall, Zhang, Valentino, & Bork owski, 2019) . Why Is Light Obligation Associated with Pos it ive Ou tcomes? Why might light obligation render positive outcomes even at high er levels? Light obligation might enrich relationships and promot e well - being by inducing positive emotions within and between 23 individuals . Regulatory focus theory suggest s peo pl e feel certain positive emotions, such as calmness, when they expect to meet their obligations (Higgins, 19 97) . Looking more broadly , p eople find prosociality and giving in general to be emotionally rewarding . P rosociality increase s happiness and self - esteem , and l ikewise, family assistance promote s positive emotions (Crocker, Canevello, & Brown, 2017; Telzer & Fuligni, 2009) . Since prosociality and generous behavior are linked to better health and well - being (S. L. Brown & Brown, 2015; Penner, Dovidio, Piliavin, & Schroeder, 2005) , it is unsurprising to find peopl e who feel lighter obligations report better well - being across time assuming that people who feel more obligated to help are more likely to help. In the current study, t he exploratory structural equation modeling approach used to test discriminant validity o f the two - factor obligation solution supported this cross - sectionally . L ight obligation was associated with greater positive affect and less negative affect. In addition , when people respond to others needs, the recipient generally shows gratitude. Rec ei ving/seeing gratitude is associated with 1) greater life and relationship satisfaction f or the individual and 2) mutually responsive behavior between individuals (Algoe, 2012) . A norm of reciprocity builds a sense of satisfaction in individual s , and relational partners become a source of support as a consequence (Neufeld & Harrison, 1995 ; Reinhardt, 1996) . Therefore, as the current study suggests, lighter forms of obligation is associated with indi v i dual well - being and positive r elationships . 5 Why Is Substantive Obligation Associated w ith Mostly Negative Outcomes ? We may think that the m echanism suggested for light obligation is equally likely to hold for substantive obligation. Particularly, we may think substantive obligations should make recipien ts feel more 5 Although light obligation also predicted slower increases in life satisfaction, this is likely a ceiling effect given that light obligation also predicted higher initial levels of life satisfaction and participants were on average, quite satisfied. 24 thankful , and thus lead to more supportive relationships ( e.g. feeling more th ankful to a friend , who is willing to take care of my child vs. a friend who calls every week ). However, the current study mostly finds that substantive obligation is associated with more negative outcomes. In other words, feeling highly obligated to ful fi ll responsibilities involving more permanent, life - changing sacrifices was not only negatively associated with the well - being but also their close relationships. Wh ile holding lighter forms of obligations is not particularly costly to the in di vidual (e.g. regularly calling parents) , substantive obligations require large r investments in various resources , which may interfere with sustaining other relationships or areas of life (e.g. giving money to a friend in need, when this makes it harder t o meet own needs). Because substantive obligations require greater investments, it may not al ways be feasible for people to fulfill their obligations unlike light obligations . When people cannot meet their obligations, they may experience negative emotions s uch as agitation, anxiety and nervousness (Higgins, 1997). Even when people do meet their o bligations, the costs may outweigh the benefits over time although they may feel good initially . P revious research supported the idea that obligations with high co st s are harmful. For instance, i ntensity of caregiving is related to worse health for the giv er (Schulz & Sherwood, 2008) . Further, w hen people feel like they are gi vi ng too much support even on social media, they report feeling exhausted and less satisfied with life (Maier, Laumer, Eckhardt, & Weitzel, 2015) . T he current results are in line with these previous studies . However, a n interesting exception was found for friend ships, where substantive obligation predicted higher l ev els of both strain and support at the first wave . One possible explanation for this may be the voluntary nature of f riendships . On one hand, substantive obligation may create strain in the friendship while people try to get others to reciprocate equally (Trivers, 1971) . On the other hand , 25 substantive obligation also signals a strong desire to stay in the relationship , which may motivate the frien d who receives benefits to reciprocate ultimately resulting in a more supportive friends hip . the closest relationships we have that provide us with support and love are often the most difficult and frustrating (Fingerman, Hay, & Birditt, 2004) . Friendships hold a particularly interesting place in relationships research in that, despite lacking filial investments a nd typical obligations (e.g. exclusivity in romantic relationships) , they persist as long as they provide emotional benefits (Baker, Chopik, & Nguyen, 2019; Chopik, 2017) . In other words , friendships l ast because people enjoy them , more so than other types of relationships . The extent to which relationships of choice per si st even in the context of substantive obligation and investments (which may undermine our enjoyment of these relationships) is an imp ortant direction for future research. Overall , the finding s in this study presented evidence that in addition to the ove ra ll amount of obligation, type of obligation seem s to matter . While light obligation might be the glue that keeps us together , substantive obligation might be the handcuffs that keep us together, mostly causing pain and unhappiness . Limitations and Futur e Directions The current study addressed important limitations in the existing obligation literature . I examined middle - aged adults who are developmentally able to perform obligations and are likely more autonomous than adolescents in that they can active ly choose which relationships to invest and feel obligations toward. The current study also examined relation ship quality with in a diverse set of relationships in addition to individual functioning . Nevertheless , there are also limitations to th is study th at are worth explicitly mentioning. First , there is a possibility that obligation may have been changing in concer t, or simultaneously, 26 with relationship quality and adjustment over the 18 - year duration of the study. This possibility may explain why obliga ti on (measured at MIDUS I) predicted initial levels of outcomes but rarely predicted changes in the outcomes. Alth ough the current data did not allow for us to examine whether changes in obligation predict changes in outcomes, it would be interesting to te st this possibility in future studies that measure obligation repeatedly over time. Second, w hile the current stu dy was able reveal the underlying factor structure of the MIDUS obligation measure, it also revealed some limitations that can affect the inte rp retation and generalization of the results with respect to individual and relational functioning . For instance, each obligation factor differentially predicted individual and relational well - being . However, the obligation scale made available in MIDUS co nf lates relationship source with the degree of investment (e.g., the substantive obligation items referenced only friends). Thus, we cannot completely ascertain whether the results mean that there is 1) an effect of relationship type (family or friends), 2 ) an effect of obligation type (light or substantive), or 3) an interaction between relationship type and obligati on type . An extreme (and superficial) interpretation of the current study is that family o bligation s benefit intra - and interpersonal well - bei ng , and friend obligatio n s are maladaptive for intra - and interpersonal well - being . However, i t could be that feeling strongly obligated to friends necessarily includes having fewer resources (defined broadly) to dedicate to family relationships. This taxi ng of resources would lead to the prediction that substantive obligation leads to less positive relationships with family members but still positive relationships with friends, who are receiving our time and attention . Worth noting, the MIDUS obligation me as ure asked about hypotheti cal situations to which anyone could respond, suggesting that respondents may have been evaluating the ir general feelings toward obligation and not the relative investment in spousal v s . family v s . friend relationships. Yet anoth er 27 possibility is that ligh general tendencies to feel obligation and therefore affects all relationships similarly, and substantive obligation is more relationship - specific . To date, there have been no studies directly comp ar ing levels of obligation toward different relationships in people lives, how these relationships might conflict with one another , and how obligation in one relationship might translate to poor outcomes in another relationship. Future research can more di rectly compare light and substantive obligation from different sources (e.g., spouses, family, friends) using more carefully constructed measures than the one used here. Lastly, although we provided some reasons for why certain forms of obligation mig ht be better or worse for people and their relationships, we did not specifically examine any of the mechanisms that might link obligation to individual and relational well - being . Aff ect is a possible mediator that explain s why light or substantive obligat io n is related to well - being in certain ways. Specifically, substantive obligation may be associated with worse relationship quality because it leads to increases in negative affect ( e.g. Juang & Cookston, 2009; Hooper et al., 2015; Whisman, Uebelacker & W ei nstock, 2004). Light obligation and reciprocity may enrich relationships by promoting positive emotions between individuals, which would be consistent with a few theoretical models specifically hypothesizing links between close relationships and well - bei ng (e.g. Algoe, 2012; Eisenberger et al., 2001) . Further, there are likely additional variables that might enhance or diminish the effects of obligation on individual and relational w ell - being . For example, t he concept of the r elational self has been suggested as a moderator of the effects of obligation, such that having a relational - interdependent self - construal was associated with higher well - being in Filipino students who felt hig he r levels of obligation (King & Ganotice Jr, 2015) . Future research can more formally model these moder at ing and 28 mediating processes of the link between obligation and important outcomes. 29 CONCLUSION In the current 18 - year longitudinal study of middle - aged adults , intra - and interpersonal well - being increased over time . L ight obligation toward family membe rs w as associated with benefits higher well - being and higher quality of close relationship s, including friendshi ps . However, s ubstantive obligation toward friends was associated with lower individual and relational well - being in most cases . Because m any of us feel a sense of obligation to people in our lives , it is important to understand when obligation may be beneficial and harmful for individuals and their close relationships . Future research can reveal the process through which obligation affects close rela tionships, particularly how varying degrees of obligation towards different relational partners i ntersect and a ffect the quality of our close relationships. 30 APPENDICES 31 APPENDIX A : Tables 32 Table 1. Obligation Measure and Geomin Rotated F actor Loadings for All Items Light Substantive 1 To drop your plans when your children seem very troubled. 0.792* - 0.019 2 To call, write, or visit your adult children on a regular basis. 0.699* 0.079* 3 To raise the child of a close friend if the f riend died. 0.230* 0.489* 4 To drop your plans when your spouse seems very trouble d . 0.598* - 0.003 5 To take your divorced or unemployed adult child back into your home. 0.394* 0.453* 6 To take a friend into your home who could not afford to live alo ne. 0.01 0* 0.963* 7 To call your parents on a regular basis. 0.384* 0.175* 8 To give money to a friend in need, even if this made it hard to meet your own needs. 0.080* 0.625* Note. Loadings were bolded to indicate to which latent factor a given item belonged. 33 Table 2. Comparing Model Fit Indices between Various Factor Analytic Models df p CFI RMSEA Full Sample One factor CFA (8 items) 3388.13 20 <.001 0.77 .176 (.171, .181) Split samples Two factor EFAa (8 items) 289.61 13 <.001 0.96 .089 (.08, .098) Three factor EFAa (8 items) 108.139 7 <.001 0.99 .073 (.062, .086) Two factor EFAa (7 items) 140.53 8 <.001 0.98 .079 (.068,.090) Two factor CFAb (7 items) 436.491 13 <.001 0.93 .108 (.100, .117) Full sample Two factor C FA (7 items) 761.738 13 <.001 0.93 .103 (.097, .109) Note. Exploratory factor analyses (EFA) a nd a follow up confirmatory factor analysis (CFA) were conducted on split samples. The entire sample was randomly split in half and models sharing a subscript used the same sample. Evaluated fit indices with the fol lowing criteria: Non - significant Chi - squa re, comparative fit index (CFI) >.95, root mean square error of approximation (RMSEA) <.06, Tucker - Lewis index (TLI) > .95, standardized root mean squared residual (SRMR) <.08. Confidence intervals for RMSEA are pre sented in parentheses. 34 Table 2 ( c ont' d ) . Comparing Model Fit Indices between Various Factor Analytic Models TLI SRMR N Full Sample One factor CFA (8 items) 0.68 0.08 5460 Split samples Two factor EFAa (8 items) 0.92 0.03 2676 Three factor EFAa (8 items) 0.94 0.02 2676 Two factor E FAa (7 items) 0.94 0.02 2676 Two factor CFAb (7 items) 0.88 0.05 2784 Full sample Two factor C FA (7 items) 0.89 0.05 5460 Note. Exploratory factor analyses (EFA) and a follow up confirmatory factor analysis (CFA) were conducted on split samples. The entire sample was randomly split in half and models sharing a subscript used the same sample. Evaluated fit indices with the fol lowing criteria: Non - significant Chi - square, comparative fit index (CFI) >.95, root mean square error of approximation (RMSEA) <.06, Tucker - Lewis index (TLI) > .95, standardized root mean squared residual (SRMR) <.08. Confidence intervals for RMSEA are pre sented in parentheses. 35 T able 3 . Means, SD s and Correlations among Obligation Items, Positive and Negative Affect 1 2 3 4 5 6 7 8 9 1. Ob 1 2. Ob 2 .552 ** 3. Ob 3 .342 ** .357 ** 4. Ob 4 .497 ** .372 ** .272 ** 5. Ob 6 .260 ** .324 ** .554 ** .209 ** 6. Ob 7 .315 ** .399 ** .298 ** .339 ** .307 ** 7. Ob 8 .233 ** .315 ** .462 ** .194 ** .630 ** .392 ** 8. PA .056 ** .104 ** .073 ** .110 ** .055 ** .102 ** .067 ** 9. NA - .066 ** - .076 ** - .022 - .105 ** - .001 - .073 ** .001 - .629** M 8.88 7.89 6.97 8.74 5.91 7.96 6.25 3.39 1.54 SD 1.76 2.21 2.74 2.15 2.72 2.57 2.61 0.73 0.62 N 6235 6221 6219 6164 6204 60 71 6220 6306 6299 Note. Ob 1 = Obligation item 1 , PA = Mean Positive Affect, NA = Mean Negative Affect. O b ligation 5 was excluded . ** p <.001 36 Table 4 . Means, SD s and Correlations as a Function of Wave of Data Collection 1 2 3 4 5 MIDUS I 1. Light Ob (1995 - 1996) 2. Substantive Ob .469 ** 3. Life Satisfaction I .255 ** .103 ** 4. Depression1 - .031 * .048 ** - .268 ** 5. Family Support 1 .261 ** .132 ** .359 ** - .134 ** 6. Family Strain 1 - .100 ** - .011 - .281 ** .152 ** - .393 ** 7. Partner Support 1 .131 ** .039 ** .476 ** - .129 ** .273 ** 8. Partner Strain 1 - .118 ** - .040 ** - .430 ** .138 ** - .194 ** 9. Friend Support 1 .179 ** .224 ** .262 ** - .041 ** .385 ** 10. Friend Strain 1 - .090 ** - .024 - .229 ** .092 ** - .168 ** MIDUS II 11. Life Satisfactio n 2 .208 ** .082 ** .541 ** - .190 ** .279 ** (2004 - 2005) 12. Depression 2 - .006 .029 - .203 ** .315 ** - .125 ** 13. Family Support 2 .227 ** .113 ** .284 ** - .132 ** .513 ** 14. Family Strain 2 - .054 ** .027 - .246 ** .125 ** - .224 ** 15. Partner Support 2 .097 ** .018 .291 ** - .091 ** .230 ** 16. Partner Strain 2 - .093 ** - .034 - .265 ** .073 ** - .155 ** 17. Friend Support 2 .190 ** .194 ** .240 ** - .046 ** .307 ** 18. Friend Strain 2 - .100 ** - .016 - .208 ** .076 ** - .149 ** MIDUS III 19. Life Satisfaction 3 .172 ** .041 * .465 ** - .186 ** .247 ** (2013 - 2014) 20. Depression 3 - .029 .018 - .150 ** .279 ** - .101 ** 21. Family Support 3 .225 ** .114 ** .304 ** - .125 ** .424 ** 22. Family Strain 3 - .086 ** - .009 - .218 ** .123 ** - .214 ** 23. Partner Support 3 .052 * .000 .229 ** - .059 * .162 ** 24. Partner Strain 3 - .055 * .003 - .228 ** .044 - .106 ** 25. Friend Support .204 ** .206 ** .224 ** - .049 * .295 ** 26. Friend Strain - .107 ** - .023 - .184 ** .076 ** - .138 ** M 8.37 6.38 7.70 0.79 3.43 SD 1.63 2.26 1.31 1.93 0.62 Note. N = 7108 - 1892, ** p <. 01, * p <.05. Ob = Obligation. 37 Table 4 ( c ont' d ) . Means, SD s and Correlations as a Function of Wave of Data Collection 6 7 8 9 10 11 12 13 14 - .169 ** .299 ** - .646 ** - .139 ** .19 3 ** - .145 ** .475 ** - .116 ** .270 ** - .144 ** - .233 ** .260 ** - .265 ** .214 ** - .195 ** .146 ** - .116 ** .111 ** - .059 ** .074 ** - .260 ** - .240 ** .190 ** - .137 ** .260 ** - .149 ** .349 ** - .145 ** .515 ** - .173 ** .269 ** - .099 ** .370 ** - .296 ** .176 ** - .361 ** - .178 ** .517 ** - .387 ** .156 ** - .109 ** .450 ** - .172 ** .289 ** - .222 ** .263 ** - .379 ** .575 ** - .109 ** .228 ** - .398 ** .139 ** - .205 ** .356 ** - .093 ** .131 ** - .095 ** .533 ** - .097 ** .291 ** - .066 ** .375 ** - .139 ** .325 ** - .102 ** .204 ** - .090 ** .472 ** - .21 4 ** .117 ** - .165 ** .501 ** - .228 ** .223 ** - .229 ** .164 ** - .192 ** .581 ** - .228 ** .289 ** - .248 ** .104 ** - .069 ** .075 ** - .039 * .060 ** - .220 ** .343 ** - .154 ** .129 ** - .233 ** .200 ** - .138 ** .251 ** - .160 ** .297 ** - .157 ** .536 ** - .252 ** .426 ** - .142 ** .243 ** - . 100 ** .339 ** - .219 ** .128 ** - .229 ** .510 ** - .163 ** .409 ** - .309 ** .110 ** - .074 ** .302 ** - .087 ** .189 ** - .140 ** .241 ** - .333 ** .503 ** - .076 ** .203 ** - .270 ** .045 - .134 ** .251 ** - .099 ** .105 ** - .108 ** .426 ** - .099 ** .243 ** - .064 ** .320 ** - .085 ** .264 ** - .082 ** .203 ** - .076 ** .399 ** - .180 ** .070 ** - .152 ** .330 ** 2.11 3.59 2.23 3.23 1.93 7.76 0.63 3.52 2.04 0.61 0.57 0.62 0.67 0.51 1.25 1.74 0.59 0.60 Note. N = 7108 - 1892, ** p <.01, * p <.05. Ob = Obligation. 38 Table 4 ( c ont' d ). Means, SD s and Correlat ions as a Function of Wave of Data Collection 15 16 17 18 19 20 21 22 23 - .636 ** .217 ** - .142 ** - .125 ** .304 ** - .133 ** .275 ** - .288 ** .207 ** - .193 ** - .116 ** .088 ** - .057 ** .056 ** - .299 ** .181 ** - .148 ** .318 ** - .177 ** .351 ** - .168 ** - .101 ** .262 ** - .113 ** .341 ** - .261 ** .134 ** - .374 ** .554 ** - .456 ** .134 ** - .114 ** .47 5 ** - .157 ** .228 ** - .151 ** - .417 ** .633 ** - .071 ** .223 ** - .403 ** .097 ** - .152 ** .314 ** - .655 ** .138 ** - .124 ** .525 ** - .099 ** .290 ** - .094 ** .397 ** - .150 ** .183 ** - .092 ** .214 ** - .097 ** .428 ** - .216 ** .105 ** - .179 ** .515 ** - .095 ** 3.63 2.15 3.28 1.84 7 .80 0.60 3.51 1.95 3.64 0.53 0.61 0.66 0.50 1.31 1.71 0.58 0.63 0.54 Note. N = 7108 - 1892, ** p <.01, * p <.05. Ob = Obligation. 39 Table 4 ( c ont' d ). Means, SD s and Correlations as a Function of Wave of Data Collection 24 25 26 - .131 ** .278 ** - .158 ** 2.10 3.30 1.72 0.63 0.64 0.53 Note. N = 7108 - 1892, ** p <.01, * p <.05. Ob = Obligation. 40 Table 5. Model Fit Indices for Unconditional N o - growth and Line ar Models by Outcome Variable Model df p RMSEA Life Satisfaction No - growth 83.178 6 <.001 .045 (.036,.053) Linear 6.073 3 .108 .013 (.000,.027) Depression No - growth 139.304 6 <.001 .056 (.048,.064) Linear 39.743 3 <.001 .042 (.031, .053) Family Support No - growth 137.867 6 <.00 1 .059 (.050, .067) Linear 46.773 3 <.001 .048 (.036, .060) Family Strain No - growth 266.372 6 <.001 .082 (.074, .091) Linear 3.537 3 .316 .005 (.000,.022) Partner Support No - growth 78.714 6 <.001 .049 (.040,.059) Linear 14.055 3 .003 .027 (.01 4, .042) Partner Strain No - growth 141.424 6 <.001 .067(.057,.076) Linear 8.236 3 .041 .019(.003,.034) Friend Support No - growth 76.664 6 <.001 .043(.035,.052) Linear 16.616 3 <.001 .027 (.015,.040) Friend Strain No - growth 517.442 6 <.001 .115(.10 7,.124) Linear 19.165 3 <.001 .029(.018,.042) Note. No - growth /intercept only models had 3 parameters and the linear models had 6 parameters. No - growth and linear models . The linear models were centered at the first wave of data collection and evaluated fit with the f ollowing criteria: Non - significant Chi - square, comparative fit index (CFI) >.90, root mean square error of approximation (RMSEA) <.08, Tucker - Lewis index (TLI) > .90, standardized root mean squared residual (SRMR) <.08. Confidence intervals for RMSEA are p resented in parentheses. 41 Table 5 ( c Model Fit Indices for Unconditional No - growth and Linear Models by Outcome CFI TLI SRMR N df p 0.970 0.985 0.099 6455 0.999 0.999 0.02 6455 77.105 3 <.001 0.873 0.937 0.069 7108 0.965 0.965 0.029 7108 99.561 3 <.001 0.94 0.97 0.104 6396 0.98 0.98 0.02 6396 91.094 3 <.001 0.876 0.938 0.107 6397 1 1 0.01 6397 262.835 3 <.0 01 0.953 0.977 0.103 5076 0.993 0.993 0.021 5076 64.659 3 <.001 0.934 0.967 0.089 5076 0.997 0.997 0.015 5076 133.188 3 <.001 0.968 0.984 0.072 6395 0.994 0.994 0.051 6395 60.048 3 <.001 0.692 0.846 0.152 6394 0.99 0.99 0.038 6394 498.2 77 3 <.001 Note. No - growth /intercept - only models had 3 parameters and the linear models had 6 parameters. No - growth and linear models. The linear models were centered at the first wave of data collection and evaluated fit with the following cr iteria: Non - significant Chi - square, comparative fit index (CFI) >.90, root mean square error of approximation (RMSEA) <.08, Tucker - Lewis index (TLI) > .90, standardized root mean squared residual (SRMR) <.08. Confidence intervals for RMSEA are presented in parentheses. 42 Table 6. Path Coefficient Estimates from the Latent Growth Curve Models in which Light and Substantive Obligation Predicts Levels and Changes in Outcomes Predictor Outcome b SE p Light obligation Life satisfaction level 0.43 0.02 <.001 0.42 Life satisfaction change - 0.05 0.02 0.01 - 0.13 Substantive obligation Life satisfaction level - 0.10 0.02 <.001 - 0.09 Life satisfaction change - 0.02 0.02 0.40 - 0.04 Light obliga tion Depression level - 0.15 0.04 <.001 - 0.12 Depression change 0.04 0.03 0.17 0.09 Substantive obligation Depression level 0.19 0.04 <.001 0.16 Depression change - 0.06 0.03 0.03 - 0.14 Light obligation Family support level 0.18 0.01 <.001 0 .38 Family support change - 0.01 0.01 0.39 - 0.05 Substantive obligation Family support level - 0.01 0.01 0.35 - 0.02 Family support change - 0.01 0.01 0.56 - 0.03 Light obligation Family strain level - 0.09 0.01 <.001 - 0.19 Family strain chang e 0.01 0.01 0.61 0.03 Substantive obligation Family strain level 0.04 0.01 <.001 0.09 Family strain change 0.00 0.01 0.69 0.02 Light obligation Partner support level 0.11 0.01 <.001 0.23 Partner support change - 0.01 0.01 0.24 - 0.07 Substant ive obligation Partner support level - 0.03 0.01 0.01 - 0.07 Partner support change 0.00 0.01 0.68 - 0.02 Light obligation Partner strain level - 0.09 0.02 <.001 - 0.18 Partner strain change 0.01 0.01 0.47 0.05 Substantive obligation Partner st rain level 0.02 0.01 0.14 0.04 Partner strain change 0.01 0.01 0.21 0.07 Light obligation Friend support level 0.06 0.01 <.001 0.12 Friend support change 0.02 0.01 0.08 0.11 Substantive obligation Friend support level 0.13 0.01 <.001 0.25 Friend support change - 0.02 0.01 0.01 - 0.15 Light obligation Friend strain level - 0.07 0.01 <.001 - 0.20 Friend strain change 0.00 0.01 0.63 - 0.03 Substantive obligation Friend strain level 0.03 0.01 0.01 0.07 Friend strain change 0.01 0 .01 0.47 0.05 43 APPENDIX B : Figures 44 Figure 1. Path diagram of Two Factor Obligation Predicting Positive and Negative Affect Note . Ob 1 = Obligation item 1. PA = positive affect, NA = negative affect. S tandardized coefficients are reported. Factor v ariances were set to 1. ** p < .001 45 Figure 2. Generic Path D iagram o f C onditional M odels Note . Ob 1 = Obligation item 1; Light = light obligation ; Substant = substantive obligation ; MIDUS I = outcome measured at MIDUS I. Consistently specified factor lo a dings are shown in the figure. Variances for MIDUS I, II and III, light and substantive obligations were fixed to 1. All other parameters were freely estimated. 46 REFERENCES 47 REFERENCES Algoe, S. B. (2012). Find, remind, a nd bind: The functions of gratitude in everyday relationships. Social and Personality Psychology Compass, 6 (6), 455 - 469. doi:10.1111/j.1751 - 9004.2012.00439.x Arriaga, X. B., & Ag new, C. R. (2001). Being Committed: Affective, Cognitive, and Conative Compon e nts of Relationship Commitment. Personality and Social Psychology Bulletin, 27 (9), 1190 - 1203. doi:10.1177/0146167201279011 Baird, B. M., Lucas, R. E., & Donnellan, M. B. (2010). Life satisfaction across the lifespan: Findings from two nationally represent a tive panel studies. Social indicators research, 99 (2), 183 - 203. doi:10.1007/s11205 - 010 - 9584 - 9 Baker, Z. G., Chopik, W. J., & Nguyen, T. T. (2019). A dyadic investigation of roman tic relationships and friendships within an Investment Model framework. Manus c ript Submitted for Publication . Baltes, P. B., & Nesselroade, J. R. (1979). History and rationale of longitudinal research. In J. R. Nesselroade & P. B. Baltes (Eds.), Longitudi nal Research in the Study of Behavior and Development (pp. 1 39). New York, N Y : Academic Press. Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological bulletin, 88 (3), 588 - 606. Brim, O. G., Ryff, C. D., & Kessler, R. C. (2004). The MIDUS National Surve y : An Overview : University of Chicago Press. Brown, S. L., & Brown, R. M. (2015). Connecting prosocial behavior to improved physical health: Contributions from the neurobiology of parenting. Neuroscience & Biobehavioral Reviews, 55 , 1 - 17. Brown, T. A. (20 1 4). Confirmatory factor an alysis for applied research (2 ed.). New York, NY: Guilford Publications. Butler, S. S., Turner, W., Kaye, L. W., Ruffin, L., & Downey, R. (2005). Depre ssion and caregiver burden among rural elder caregivers. Journal of Gerontolo g ical Social Work, 46 (1), 47 - 63. attachment patterns. Family process, 41 (3), 375 - 388. 48 Carstensen, L. L., Isaacowitz, D. M., & Charles, S. T. (1999). Taking time seri o usly: A theory of socioemotional selectivity. American psychologist, 54 (3), 165 - 181. doi:10.1037/0003 - 066x.54.3.165 Cash, B., Hodgkin, S., & Warburton, J. (2013). Till death us do part? A critical analysis of obligation and choice for spousal caregivers. J ournal of Gerontological Social W ork, 56 (8), 657 - 674. Chopik, W. J. (2017). Associations among relational values, support, health, and well - being across the adult lifespan. Personal Relationships, 24 (2), 408 - 422. doi:10.1111/pere.12187 Chopik, W. J., & E d elstein, R. S. (2018). Retrospect ive memories of parental care and health from mid - to late life. Health Psychology, 38 (1), 84 - 93. doi:10.1037%2Fhea0000694 Cicirelli, V. G. (1993). Attachment and obligation as daughters' motives for caregiving behavior and subsequent effect on subjective b urden. Psychology and aging, 8 (2), 144. Clark, M. S., & Mills, J. (1979). Interpersonal attraction in exchange and communal relationships. Journal of personality and social psychology, 37 (1), 12 - 24. doi:10.1037/0022 - 3514. 3 7.1.12 Coulombe, P., Selig, J. P. , & Delaney, H. D. (2016). Ignoring individual differences in times of assessment in growth curve modeling. International Journal of Behavioral Development, 40 (1), 76 - 86. Crocker, J., Canevello, A., & Brown, A. A. (2017). Social motivation: Costs and benefits of selfishness and otherishness. Annual Review of Psychology, 68 , 299 - 325. Diener, E., & Emmons, R. A. (1984). The independence of positive and negative affect. Journ al of personality and social psychology, 47 (5), 11 0 5 - 1117. Eisenberger, R., Armeli, S., Rexwinkel, B., Lynch, P. D., & Rhoades, L. (2001). Reciprocation of perceived organizational support. Journal of Applied Psychology, 86 (1), 42. Fingerman, K. L., Hay, E. L., & Birditt, K. S. (2004). The best of ties, the worst of ties: Close, problematic, and ambivalent social relationships. Journal of Marriage and Family, 66 (3), 792 - 808. Floyd, F. J., & Widaman, K. F. (1995). Factor analysis in the development and re finement of clinical assessment instruments. Psych o logical assessment, 7 (3), 286 - 299. Fuligni, A. J., Tseng, V., & Lam, M. (1999). Attitudes toward family obligations among American adolescents with Asian, Latin American, and European backgrounds. Child d evelopment, 70 (4), 1030 - 1044. 49 Gana, K., Bailly, N . , Saada, Y., Joulain, M., & Alaphilippe, D. (2012). Does life satisfaction change in old age: Results from an 8 - year longitudinal study. Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 68 (4), 540 - 552. Grimm, K. J., & Ram, N. (2012). Growth curve modeling from a structural equation modeling perspective. In B. Laursen, T. D. Little, & N. A. Card (Eds.), Handbook of developmental research methods (pp. 411 431). New York, NY: Guil ford Press. Grimm, K. J., Ram, N., & Estabrook, R. (2016). Growth modeling: Structural equation and multilevel modeling approaches : Guilford Publications. Higgins, E. T. (1997). Beyond pleasure and pain. American psychologist, 52 (12), 1280 - 1300. Hooper, L . M., Tomek, S., Bond, J. M., & Reif, M. S. (2015). Race/ethnicity, gender, parentification, and psychological functioning: Comparisons among a nationwide university sample. The Family Journal: Counseling and Therapy for Couples and Families, 23 (1), 33 - 48. doi:10.1177/1066480714547187 Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6 (1), 1 - 55. doi:10 .1080/10705519909540118 Juang, L. P., & Cookston, J. T. (2009). Acculturation, Discrimination, and Depressive Symptoms among Chinese American Adolescents: A Longitudinal Study. The Journal of Primary Prevention, 30 (3), 475 - 496. doi:10.1007/s10935 - 009 - 0177 - 9 King, R. B., & Ganotice Jr, F. A. (2015). Does family obligation matter for students' motivation, engagement, and well - being?: It depends on your self - construal. Personality and Individual Differences, 86 , 243 - 248. Lance, C. E., Beck, S. S., Fan, Y., & Carter, N. T. (2016). A taxonomy of path - related goodness - of - fit indices and recommended criterion values. Psychological methods, 21 (3), 388 - 404. doi:10.1037/met0000068 Luong, G., Charles, S. T., & Fingerman, K. L. (2011). Better with age: Social relations hips across adulthood. Journal of Social and Personal Relationships, 28 (1), 9 - 23. MacCallum, R. C., Browne, M. W., & Sugawara, H. M. (1996). Power analysis and determination of sample size for covariance structure modeling. Psychological methods, 1 (2), 13 0 - 149. Macfie, J., Brumariu, L. E., & Lyons - Ruth, K. (2015). Parent child role - confusion: A critical review of an emerging concept. Developmental Review, 36 , 34 - 57. 50 Maier, C., Laumer, S., Eckhardt, A., & Weitzel, T. (2015). Giving too much social support : social overload on social networking sites. European Journal of Information Systems, 24 (5), 447 - 464. Muthén, L. K., & Muthén, B. (2017). Mplus Version 8 User's Guide : Muthen & Muthen. Neufeld, A., & Harrison, M. J. (1995). Reciprocity and social support in caregivers' relationships: Variations and consequences. Qualitative Health Research, 5 (3), 348 - 365. Nock, S. L. (1995). Commitment and dependency in marriage. Journal of Marriage and the Family , 503 - 514. Nuttall, A. K., & Valentino, K. (2017). An eco logical - transactional model of generational boundary dissolution across development. Marriag e & Family Review, 53 (2), 105 - 150. doi:10.1080/01494929.2016.1178203 Nuttall, A. K., Valentino, K., Wang, L., Lefever, J. B., & Borkowski, J. G. (2015). Maternal hi story of parentification and warm responsiveness: The mediating role of knowledge of infant development. Journal of Family Psychology, 29 (6), 863 - 872. doi:10.1037/fam0000112 Nuttall, A. K ., Zhang, Q., Valentino, K., & Borkowski, J. G. (2019). Intergenerati onal Risk of Parentification and Infantilization to Externalizing Moderated by Child Temperament. Journal of Marriage and Family . Penner, L. A., Dovidio, J. F., Piliavin, J. A., & Schroe der, D. A. (2005). Prosocial behavior: Multilevel perspectives. Annua l Review of Psychology, 56 , 365 - 392. Prenda, K. M., & Lachman, M. E. (2001). Planning for the future: a life management strategy for increasing control and life satisfaction in adulthoo d. Psychology and aging, 16 , 206 - 216. Reinhardt, J. P. (1996). The i mportance of friendship and family support in adaptation to chronic vision impairment. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 51 (5), 268 - P278. Schulz, R., & Sherwood, P. R. (2008). Physical and mental health effe cts of family caregiving. Journal of Social Work Education, 44 (sup3), 105 - 113. Schuster, T. L., Kessler, R. C., & Aseltine, R. H. (1990). Supportive interactions, negative interactions, and depressed mood. American journal of community psychology, 18 (3), 423 - 438. Stein, C. H. (1992). Ties that bind: Three studies of obligation in adult relationships with family. Journal of Social and Personal Relationships, 9 (4), 525 - 547. 51 Stuifbergen, M . C., & Van Delden, J. J. (2011). Filial obligations to elderly paren ts: a duty to care? Medicine, Health Care and Philosophy, 14 (1), 63 - 71. Tedgård, E., Råstam, M., & Wirtberg, I. (2018). An upbringing with substance - abusing parents: Experiences of paren tification and dysfunctional communication. Nordic Studies on Alcohol and Drugs , 1 - 25. doi:1455072518814308 Telzer, E. H., & Fuligni, A. J. (2009). Daily family assistance and the psychological well - being of adolescents from Latin American, Asian, and Euro pean backgrounds. Developmental psychology, 45 (4), 1177 - 1189. doi:10. 1037/a0014728 Trivers, R. L. (1971). The evolution of reciprocal altruism. The Quarterly review of biology, 46 (1), 35 - 57. van Geel, M., & Vedder, P. (2011). The role of family obligation s and school adjustment in explaining the immigrant paradox. Journal of youth and adolescence, 40 (2), 187 - 196. Walen, H. R., & Lachman, M. E. (2000). Social support and strain from partner, family, and friends: Costs and benefits for men and women in adul thood. Journal of Social and Personal Relationships, 17 (1), 5 - 30. doi :10.1177/0265407500171001 Wang, P. S., Berglund, P., & Kessler, R. C. (2000). Recent care of common mental disorders in ecommendations. Journal of general internal medicine, 15 (5), 284 - 292. Whisman, M. A., Uebelacker, L. A., & Weinstock, L. M. (2004). Psychopathology and marital satisfaction: The importance of evaluating both partners. Journal of consulting and clinical ps ychology, 72 (5), 830 - 838.