Ambivalence in Later-Life Family Networks: Beyond Intergenerational Dyads

Myriam Girardin University of Geneva Eric D. Widmer University of Geneva∗ Ingrid Arnet Connidis University of Western Ontario∗∗ Anna-Maija Castrén University of Eastern Finland∗∗∗ Rita Gouveia University of Lisbon∗∗∗∗ Barbara Masotti University of Applied Sciences and Arts of Southern Switzerland∗∗∗∗∗ Ambivalence in Later-Life Family Networks: Beyond Intergenerational Dyads NCCR LIVES, Center for the Interdisciplinary Study of Gerontology and Vulnerability, University of Geneva, 54, rte des Acacias, 1227 Carouge, Geneva, Switzerland. ∗NCCR LIVES, Department of Sociology, University of Geneva, Uni Mail, 40 bd du Pont-d’Arve, CH-1211 Geneva 4, Switzerland ([email protected]). ∗∗ Department of Sociology, University of Western Ontario, London, Ontario N6A5C2, Canada. ∗∗∗ Department of Social Sciences, University of Eastern Finland, PO Box 1627, Kuopio, 70211, Finland. ∗∗∗∗ Institute of Social Sciences, University of Lisbon, Avenida Professor Aníbal Bettencourt n∘9, Lisboa 1600-189, Portugal. ∗∗∗∗∗ NCCR LIVES, Centre of Competence on Aging, University of Applied Sciences and Arts of Southern Switzerland, Stabile Piazzetta, Via Violino 11, 6928 Manno, Switzerland. © 2018 The Authors. Journal of Marriage and Family published by Wiley Periodicals, Inc. on behalf of National Council on Family Relations. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. Key Words: ambivalence, conflict, emotional support, family networks, inequalities, older adults. In later life, changing conditions related to health, partnership, and economic status may trigger not only support but also conflict and ambivalence, with the consequent renegotiation of family ties. The aim of this study is to investigate both conflict and emotional support in the family networks of older adults, taking the research beyond the level of intergenerational dyads. We used a subsample of 563 elders (aged 65 years and older) from the Swiss Vivre/Leben/Vivere survey. Multiple correspondence analysis and in-depth case studies were used to identify the key social conditions that relate to the prevalence of conflicted and supportive dyads in family networks. Findings showed that the balance of conflict and emotional support in older adults’ family networks varied according to the composition of their family network as well as their age, health, income, and gender. Heightened attention to the concept of ambivalence since the turn of the century challenged the view of family ties as exclusively supportive or positive and highlighted contradictions in the ties between adult children and their aging parents (Connidis, 2012, 2015; Connidis & 768 Journal of Marriage and Family 80 (June 2018): 768–784 DOI:10.1111/jomf.12469 Ambivalence in Later-Life Family Networks 769 McMullin, 2002; Lüscher, 2002, 2005; Lüscher & Hoff, 2013; Lüscher & Pillemer, 1998). The ambivalence concept emphasizes the coexistence of conflict and support as inherent parts of family dynamics (Connidis, 2012, 2015; Connidis & McMullin, 2002; Lüscher, 2002, 2005; Lüscher & Hoff, 2013; Lüscher & Pillemer, 1998; Willson, Shuey, Elder, & Wickrama, 2006). To date, the ambivalence research has been mainly circumscribed to the dyadic level and rarely considers how ambivalence in intergenerational relationships is embedded in patterns of conflict and emotional support in larger family networks. Using a representative sample of older adults living in Geneva, Switzerland, this article explores conflict and emotional support in later-life family networks, identifies four patterns, and investigates how those patterns are embedded in the demographic, social, and economic conditions that affect individuals and their family ties. It goes beyond intergenerational dyads by studying ambivalence at the level of family networks and by making connections to broader social forces, particularly social inequality or structured social relations. The results are discussed in relation to key issues in social gerontology, including older adults’ socioemotional selectivity (Carstensen, 1992) and the unintended consequences of family support (Connidis & McMullin, 2002; Lüscher & Pillemer, 1998; Willson et al., 2006). Background Ambivalence Beyond Intergenerational Dyads Ambivalence is a multilevel concept that emphasizes contradictions at the levels of individuals and relationships; social institutions, including families; and society, including the welfare state and the structured social relations of inequality based on gender, class, race or ethnicity, age, and ability (Connidis, 2015). Individuals strive to negotiate these multilevel contradictions and the resulting coexistence of conflict and emotional support in their various personal relationships, including ties in the family realm. These negotiations occur at the meso level of families, and they need to be explored at this level—that is, beyond the level of isolated dyads. Typologies have proven to be a useful tool for exploring ambivalence in intergenerational family dyads as was shown by European research (e.g., Ferring, Michels, Boll, & Filipp, 2009; van Gaalen & Dykstra, 2006). Lüscher’s (2002, 2005) theoretical typology of ambivalence is one of very few attempts to capture the interaction between support and conflict in intergenerational ties; it includes the following four types of distinct relational patterns: emancipation, solidarity, captivation, and atomization. These types identify the various ways in which family members negotiate ambivalence in intergenerational relationships (Lüscher & Hoff, 2013). In the emancipation type, individuals accept conflict along with cooperation and support and find new and more effective ways of relating. In the solidarity type, parents and children emphasize togetherness and support as ways of avoiding conflict and open ambivalence. Captivated parent–child ties are stuck in conflict, entangled in an ongoing battle over ambivalence, whereas atomized intergenerational dyads disengage to avoid conflict and ambivalence. Lüscher’s typology has contributed to a large body of theoretical and empirical work in social gerontology (e.g., Connidis, 2015; Katz & Lowenstein, 2010; Lang, 2004; Letiecq, Bailey, & Dahlen, 2008; Phillips, Ogg, & Ray, 2003). However, related empirical work focuses on intergenerational dyads. A key challenge in the study of ambivalence in social gerontology is to explore the four patterns of Lüscher’s (2002, 2005) theoretical typology at the meso level of family networks. We therefore aim to place dyads in the wider configuration of family networks and to then relate family network types to the larger social context in which they are embedded. Empirically identifying various types of family networks and their available resources may reveal variability among family types in the capacity to negotiate ambivalence in ways that enhance emotional support in older adults’ families. To investigate ambivalence at the meso level of families, we transpose Lüscher’s (2002, 2005) theoretical typology into the framework of social network analysis, in which dyads are considered as interdependent parts of a network whose features account for much of what happens in any of them (Wasserman & Faust, 1994). Such tools have been used in social gerontology to explore family support structures (Cornwell, 2009, 2011; Girardin & Widmer, 2015). We propose using these tools to assess the relative dominance of conflicted and supportive dyads in the overall dynamics of older adults’ family networks (see Table 1). 770 Journal of Marriage and Family Table 1. Transposition of Lüscher’s Typology Into the Framework of Social Network Analysis Proportion of conflicted dyads Proportion of supportive dyads Large Low Large Emancipation Solidarity Low Captivation Atomization Taking a social network analysis perspective, the negotiation of ambivalence at the meso (family) level results in four patterns (Widmer, 2016). Family networks in which emancipation prevails display a large number of dyads that are characterized by both conflict and emotional support. These dyads accept the coexistence of emotional support and conflict, and they find ways to relate effectively. In family networks of the solidarity type, a majority of the dyads are supportive, and very few show conflict. Emotional support is actively promoted, and conflict is avoided, in a majority of these family dyads. Captivation occurs when many of the dyads in a family network are burdened by conflict, with limited emotional support; this situation is more likely to occur when family members must stay together due to strong family obligations and to limited personal resources. Finally, atomization is characterized by the presence of few family dyads that are either supportive or conflicted. Emotional support might peter out faster during the long term, and tensions are resolved through collective emotional disengagement, putting older adults at risk of loneliness. Identifying such network patterns based on Lüscher’s (2002, 2005) typology makes it possible to assess ambivalence in a variety of family contexts and its connection to emotional support. Composition of Family Networks and Other Life Course Conditions of Ambivalence Family networks are diverse in their composition. Some include only a spouse, children, and grandchildren, but others include a wider variety of ties such as siblings, distant kin, in-laws, step-relatives, and friends (Girardin & Widmer, 2015; Treas & Marcum, 2011). Different family ties may have unequal likelihoods of being ambivalent. Siblings, in-laws, distant kin, step-relatives, and friends typically have lower social expectations to provide emotional support and to stay connected; these more voluntary relationships are instead based on affinity and shared interests (Campbell, Connidis, & Davies, 1999; Schnettler & Wöhler, 2014). In cases in which such ties are prevalent in a family network, one may expect a large proportion of supportive dyads, whereas the tense dyads are disengaged—the solidarity type. In contrast, family networks composed of children and grandchildren are characterized by strong obligations to maintain close emotional connections from which family members cannot readily escape (Finch & Mason, 1993). Such ties may, in turn, create tensions. Thus, emancipation is a likely result for such family networks, which are expected to include a large proportion of supportive dyads but also some conflicted and ambivalent dyads. Other conditions associated with the life course may alter the balance of conflicted and emotionally supportive dyads in family networks. Having financial resources, being young-old, being in good health, and having a partner contribute to maintaining emotional support through mutual exchanges (Cornwell, 2009, 2011; Offer, 2012). These conditions are expected to sustain a large proportion of supportive dyads in the family network and to thus promote the solidarity type. However, the availability of resources is also associated with factors related to time allocation and emotional closeness among family members (Taylor & Norris, 2000). Thus, emancipation—the coexistence of conflicted and supportive dyads in family networks—may occur when resources are available; although resources contribute to support exchanges, they may also raise tensions related to fairness and individual preferences in resource distribution. Alternatively, having fewer financial resources, being oldest-old, being in poor health, and having lost a partner may challenge the balance of supportive dyads in family networks in favor of conflicted dyads. Older parents’ diminished resources increase their need for emotional support and, potentially, heighten the burden placed on family members—an ambivalent situation that has to be negotiated in family networks (Connidis, 2012; Connidis & McMullin, 2002; Lüscher & Pillemer, 1998; Willson et al., 2006). Such circumstances may exacerbate tension among family members, as they have to adjust their needs and expectations regarding emotional support (Connidis, 2003; Cornwell, 2009, 2011; Hillcoat-Nallétamby & Phillips, 2011). This may lead to captivation, Ambivalence in Later-Life Family Networks 771 with a large proportion of dyads within the family network that are characterized by conflict and limited emotional-support, especially when resources (e.g., income and time) are scarce and when the only potentially supportive family members are children (as in the case of widowed or divorced parents). Atomization is expected to be most likely when few or no family members are present or when a severe lack of resources makes emotional-support exchanges difficult, leading to disengagement (Offer, 2012). Note that such conditions may promote gender differences, as older women are more likely than older men to be widowed, to have a lower income, and to be in poor functional health because of their gendered position in society and their longer life expectancy (Arber, Davidson, & Ginn, 2003; Moen, 1996; Willson et al., 2006). In sum, during later life, individuals may experience a variety of economic, family, and health conditions that lead to distinct and various balances of conflict and emotional support in their family networks. This study aims to understand whether and how the composition of family networks and life course conditions are associated with emancipation, solidarity, captivation, and atomization as patterns that characterize ambivalence in older adults’ family networks. This issue is examined through the prevalence of conflicted and supportive dyads in family networks. We then observe whether older adults’ family-network composition, available resources, partnership status, age, health, and social position (as indicated by gender, class, and citizenship) account for the variations in such patterns within family configurations. To this end, we use multiple correspondence analysis (MCA) and in-depth case studies to explore how the interplay of these conditions is related to the balance of conflict and emotional support in later-life family networks. Method Data and Sample The data came from the Vivre/Leben/Vivere study, which is a large, interdisciplinary survey on the life and health conditions of people aged 65 years and older; this study was carried out in 2011 and 2012 in five of 26 cantons in Switzerland (see Oris et al., 2016). Stratified by sex and age, the sample of 3,635 community-dwelling or institutionalized participants was representative of the general population aged 65 years or older. Data were collected using a self-assessed questionnaire and an in-home, face-to-face interview with a standardized interview schedule. To illustrate our findings with more in-depth information, we also drew from a pool of rich life-history data based on life calendars that the respondents completed with the help of the interviewers. Given the practical issues concerning data availability, our analyses focused on the Geneva subsample (n=704). We dropped 126 individuals with cognitive impairments from the analysis because they were not able to answer the questionnaire on their own (resulting in a subsample with n=578). An additional 15 were dropped because they did not answer the questions about their family networks. Therefore, the final subsample included 563 respondents. The mean age in the Geneva subsample was 78 years (range, 65–101): 40% were aged 65 to 74, 35% 75 to 84, and 25% were 85 and older. Of the respondents, 49% were women, and 66% were native born; 61% had an average level of education (i.e., achieved a high school or equivalent degree), and 23% had a high level of education (i.e., achieved at least a university degree). Regarding their last occupational status (prior to retirement), 30% were upper, 27% were white collar, 14% were intermediary, 14% were blue collar, 9% were self-employed, and 6% were inactive. Of the respondents, 62% had an average or high income. As for their pools of relatives, 61% were married or had a partner (either cohabitating or living separately), 82% had at least one living child, and 68% had at least one living sibling. Regarding self-rated health, 54% assessed their health as good or very good, 38% assessed it as fair, and 8% assessed it as bad or very bad. A large majority of the respondents (75%) were in good functional health (robust on the eight activities of daily living [ADL] scale), 16% reported having difficulties in one or more of the eight ADL categories, and 9% were dependent according to the ADL scale. Few were institutionalized (7%). Measures Types of family networks. Drawing on standard name generators for family networks (Widmer, Aeby, & Sapin, 2013), we asked respondents to list a maximum of five individuals whom they considered significant family members at the time of the interview. After naming these family 772 Journal of Marriage and Family members, the respondents were asked to indicate the type of relationship (e.g., partner, sister, or daughter) that they had with each of the cited persons. A total of 14 family terms (commonly called name interpreters) were identified, reported by at least 5% of respondents. To map the main types of family networks, we applied standard factorand cluster-analytic procedures to the family networks (Widmer, 2016). The clustering approach is commonly used in social gerontology to identify social-network types, based on various kinds of social relationships (e.g., family, friends, neighbors, and community groups; Fiori, Antonucci, & Cortina, 2006; Litwin, 2001; Wenger, 1991). The identification of family networks through significant family members is, however, rarely considered. To this end, we proceeded using two steps. We first ran an exploratory factor analysis—using principal component analysis with varimax rotation—on the 14 family terms, plus a residual category into which the other terms were gathered. Following the standard practice for factor analysis (Tabachnick & Fidell, 1996), we retained the six factors that had eigenvalues greater than 1; these explained 55% of the variance. We then input these six factors’ scores into a hierarchical clustering analysis based on Euclidean distances and the Ward clustering algorithm (Lebart, Morineau, & Piron, 2002). We selected a solution with six clusters, based on both interpretability and cluster-validity measures such as the Calinski-Harabasz and silhouette indexes (Everitt, Landau, Leese, & Stahl, 2011). In the first type, which we named conjugal (39%), the respondents centered on their children and on their current partner. The son family network (8%) focused on biological sons, their partners, and their children, whereas the daughter family network (11%) largely included biological daughters and their children. In sibling family networks (15%), the respondents mainly cited their siblings as significant family members. Those in kinship family networks (8%) included a variety of relatives. Finally, those in sparse family networks (19%) either mentioned no significant family member or listed only a few friends whom they considered to be family members. Patterns of conflict and emotional support. After listing the significant family members, a set of questions was asked about the emotional support and conflict among the listed family members (Widmer et al., 2013). We measured the emotional support in the dyads using the following question: “Who would give emotional support to X [i.e., to the respondent and then to each other individual in the respondent’s family network, considered one by one] during routine or minor troubles?” We mapped the conflict in the dyads with the following question: “Each family has its conflicts and tensions. In your opinion, who makes X [i.e., the respondent and then each other individual in the respondent’s family network, considered one by one] angry?” Thus, each focal person could mention more than one member in the family network who emotionally supports or annoys each of the other members, including himor herself. In other words, the respondents evaluated not only their own relationships with each family member but also all the relationships among their significant family members. Next, to estimate emotional support and conflict in family networks, we assessed the density of emotional support and the density of conflict across all of the family dyads. These density measures refer to the proportion of supportive and of conflicted dyads among all the dyads in a personal or family network (Hanneman & Riddle, 2005). A high density of emotional support indicates that a majority of family dyads could be easily activated for emotional support, and a high density of conflict indicates a large proportion of tense dyads. The variable “patterns of conflict and emotional support” was operationalized by combining the two density measures into a single categorical variable. For this purpose, we first dichotomized these two measures: high versus low density of emotional support and high versus low density of conflict. For supportive ties, we chose the median score of .33 (33%) as the threshold separating high and low density of emotional support, as the raw scores were not normally distributed. On average, the respondents reported a much lower proportion of tense dyads than of supportive dyads, so we set the threshold for conflict at .10 (10%), which means that there was tension in one of every 10 possible dyads. Then we combined these two dichotomized density measures into a single variable—patterns of conflict and emotional support—with the following four levels: emancipation (high densities of emotional support and conflict), solidarity (high density of emotional support and low density of conflict), captivation (low density of emotional support and high density of conflict), and atomization Ambivalence in Later-Life Family Networks 773 (low densities of emotional support and conflict). Family networks with the emancipation type (23%) had high densities of both emotional support and conflict—on average, 57% of dyads were perceived as supportive, and 33% were conflict oriented. Among family networks showing solidarity (31%), the dyads were, on average, 58% supportive and only 2% tense. Family networks with the captivation type (9%) had a density of conflict (20%) that was similar to the density of emotional support (21%), which differs from the emancipation type, in which supportive dyads were clearly dominant. For family networks with the atomization type (37%), there was a very low density of both emotional support and conflict, as only 11% of the dyads were supportive and 2% were conflicted. Pool of available relatives. The respondents were asked to report whether they had a partner (either cohabitating or living separately) at the time of the interview (0 =“no partner,” 1=“have a partner”), had at least one living child (0 =“no children,” 1 =“have children”), and had at least one living sibling (0 =“no brothersorsisters,”1=“have brothers or sisters”). Age. We divided focal individuals by age into one of the following three groups: aged 65 to 74 years (“young-old”), 75 to 84 years (“old-old”), and 85 years and older (“oldest-old”; Suzman & Riley, 1985). Health. To measure health, we focused on functional health rather than on more general health indicators such as self-rated health because the loss of autonomy in old age represents one of the most challenging conditions for both the older adults and their family members. Such a situation within a family network may trigger the provision of various forms of support as well as tension. In addition, we performed the same analysis with self-rated health and obtained similar results (not shown). To assess functional health, we asked respondents how much difficulty (0 =“no difficulty,” 1 =“able with difficulty,” 2 =“unable to perform”) they had in performing five basic activities—washing, dressing and undressing, eating and cutting food, moving in and out of bed, and moving around indoors (Katz, Ford, Moskowitz, Jackson, & Jaffee, 1963). We also asked them about their difficulty in performing the following three mobility actions: going up and down stairs, moving around outside, and walking at least 200 meters (Rosow & Breslau, 1966). These eight items had a reliability level (Cronbach’s 𝛼)of.91.We created the following three functional-health categories: ADL-robust (able to perform all eight activities alone), ADL with difficulty (having difficulty performing one or more of the activities alone), and ADL-dependent (having one or more ADL incapacities and needing someone’s help to perform them). Social position in society. We used income (1 =“low,” 2 =“average,” 3 =“high”), gender (0 =“female,” 1 =“male”), and citizenship (with place of birth as a proxy; 0 =“nativeborn,” 1 =“foreign-born”) to measure social position in Swiss society. We adjusted incomes for household size, with a value of 1 for the household head and 0.5 for each additional household member (Atkinson, Rainwater, & Smeeding, 1995; Gabriel, Oris, Studer, & Baeriswyl, 2015). Data Analysis To assess the associations that the four patterns of conflict and emotional support had with the types of family networks and with the other life course conditions (e.g., the pool of available relatives, age, health, income, gender, and citizenship), we computed an MCA—using the FactoMineR package in R (Lê, Josse, & Husson, 2008). This enabled us to observe how the interplay of these different variables was associated with emancipation, solidarity, captivation, and atomization. MCA is a nonlinear multivariate analysis method for representing the underlying structures in a set of observations, as described by a set of categorical variables (Abdi & Valentin, 2007; Avolio et al., 2013). MCA relies on an assumption of interdependence rather than on a causality principle, so it is an ideal approach when various factors interact in a bidirectional or circular way. In sum, this exploratory method provides a better understanding of how the response categories are interrelated and, therefore, enables the identification of patterns. To this end, MCA extracts the main dimensions that structure the relationship between the response categories. In MCA, the first two or three extracted dimensions explain as much variance as possible, and they are usually sufficient to synthesize the most important information contained in the contingency tables, all 774 Journal of Marriage and Family Table 2. Discrimination Measures of Variables in Dimensions after Rotation Variables Dimension 1 loading Dimension 2 loading Active Gender .25 .01 Income .33 .02 Citizenship .01 .01 Having a partner .49 .04 Having at least one living child .02 .56 Having at least one living sibling .13 .10 Family networks .25 .75 Patterns of conflict and emotional support .12 .17 Status of functional health .30 .00 Passive Age – – % of explained inertia 12.61% 8.40% Note. As a passive variable, age has not contributed to the constitution of axes. N=476. while remaining parsimonious (Desbois, 2008). To identify these structuring dimensions, one relies on discrimination measures: the ratios of correlations between the variables considered in MCA and the chosen dimensions. These measures indicate the contribution of each variable in the definition of each dimension; larger discrimination measures indicate stronger contributions to the dimension’s definition (see Table 2; Avolio et al., 2013). All of the response categories can be plotted along the two dimensions (axes in the plot) of the MCA. The interpretation of the results is based on the proximities and the distances between the response categories in the plot; those that are close to one another present similar patterns of responses, and those that are distant have dissimilar patterns (Abdi & Valentin, 2007; Avolio et al., 2013). To better identify which groups of response categories were more clearly associated with one of the two main dimensions, we rotated them, as the response categories are better aligned along the main dimensions after rotation (Saracco, Chavent, & Kuentz, 2010). We input all variables except age as active (directly contributing to the constitution of the MCA’s axes) and input age as a passive variable (Greenacre & Blasius, 2006)—that is, we projected age a posteriori in the already existing plot, but it did not contribute to the axes’ definitions. This is because we understood age as a proxy for the aging process, for the psychological development, and for a variety of social processes stemming from changes in social roles, statuses, health, generational experiences and identity, and stages in the life cycle, rather than as a causal mechanism per se (Settersten & Godlewski, 2016; Settersten & Mayer, 1997). We conducted all analyses in R (R Development Core Team, 2011). Results Results of the Multiple Correspondence Analysis Table 2 displays the discrimination measures for the variables (the ratios of correlations) in the two first extracted dimensions that had eigenvalues greater than 1, as well as each dimension’s percentages of explained inertia (i.e., variance) after rotation. These two chosen dimensions synthesized the most important information for all the active variables. As shown in Table 2, variables measuring partnership (.49), income (.33), functional health status (.30), and gender (.25) strongly contributed to the constitution of Dimension 1. Types of family networks (.75) and parenthood (.56) strongly contributed to the constitution of Dimension 2. In sum, the availability of various resources, having at least on living child, and the composition of family networks were the main contributors to the two-dimensional MCA space. To test each variable’s goodness of fit, we computed confidence ellipses for each response category (Husson, Josse, Lê, & Mazet, 2017). Confidence ellipses describe the distributions of the center of gravity for each response category and can thus be used to deduce a confidence region for each category via bootstrapping. These analyses revealed that all the selected variables’ response categories were significantly distinct from each other, as their ellipses did not overlap, with the exceptions of the daughter and son family networks—which were located in the same quadrant (see Figure 1)—and country of birth. Figure 1 shows that the first dimension (horizontal axis) discriminated among individuals based on the availability of resources such as income, partnership, gender, age, functional health, and having at least one living sibling. Precisely, the response categories that indicated an advantageous position in society (high or Ambivalence in Later-Life Family Networks 775 Figure 1. Projection of the First Two Dimensions of the Multiple Correspondence Analysis, After Rotation. average income, male, native born), a large pool of relatives (have a partner, have brothers or sisters, and have children), young-old age, and good functional health (ADL-robust) were located in the negative coordinates of the horizontal axis. On the other hand, those reflecting a low social position (low income, female, and foreign born), a small pool of available relatives (no partner, no brothers or sisters, and no children), oldest-old age, and poor functional health (ADL with difficulty or ADL-dependent) were positioned in the positive coordinates of the horizontal axis. The resources dimension captured 13% of the total variance. Regarding the second dimension (vertical axis), Figure 1 reveals the strong discriminatory power of parenthood. Indeed, this axis differentiated individuals who had children (in the positive coordinates of the vertical axis) from those who did not (in the negative coordinates of the vertical axis), and their resulting family networks. Figure 1 shows that the no children category and the family networks focused on family members other than children—such as the sibling, kinship, and sparse family networks—were positioned on the negative side of the vertical axis, whereas the having children category and the family networks centered primarily on living children—such as the conjugal, daughter, and son family networks—were projected on the positive side of the vertical axis. The parental-status dimension explained 8% of the total variance. As Figure 1 shows, the four patterns of conflict and emotional support were differently projected on the four quadrants of the graph. 776 Journal of Marriage and Family Based on the coordinates of the response categories in the bidimensional space, the main profiles associated with the four patterns of conflict and emotional support became identifiable. Emancipation was in the first quadrant. This pattern was associated with conjugal family networks; with being in the male, native-born, young-old, and ADL-robust categories; and with having an average or high income. Captivation was in the second quadrant, in close connection with the inclusion of children—but with the no-partner category—and with daughter and son family networks, both focused on children and grandchildren. In terms of resources, these respondents had low social positions (low income, female, and foreign born), were in the oldest-old group and had poor functional health (ADL with difficulty or ADL-dependent). Solidarity was positioned in the third quadrant. This pattern was connected with having brothers or sisters, having a partner, having an average or high income, and being in the male, young-old, and ADL-robust groups. Solidarity was also associated with a lack of children, as it was positioned on the negative side of the vertical axis, in the same area as the sibling family networks. Finally, atomization was projected in the fourth quadrant. This pattern was associated with an absence of children and with family networks that were centered primarily on kin, on a few friends, or on no one, as in the sparse and kinship family networks. Atomization was also related to the lack of various resources, as it was connected with being in the oldest-old, ADL with difficulty or ADL-dependent, and female groups, and with having no partner and a low income. Insights From the Case Studies To illustrate the ways in which the patterns of conflict and emotional support were embedded in structural conditions, we provide four case studies that were selected on the basis of the quantitative results of both the face-to-face interviews (with standardized interview schedules) and the life-history data. Emancipation in a conjugal family network. The focal individual, a man in his 90s, was living with his wife in his own house at the time of the interview. After a career as an international civil servant, he was well off, in good functional health, and had two sons, both of whom lived nearby. His older son was married and had three children, whereas the younger one was single and childless. His family belonged to the conjugal type, as he reported his wife and two sons as the only significant family members. The supportive dyads in this family were quite dense and reciprocal. As shown in Figure 2a, his wife was central in providing emotional support, as she gave support to all the family members but received support only from the focal person (the arrows point to support providers). The focal individual exchanged emotional support with his wife and with the younger son. The older son was less integrated in the emotional support exchanges, as he provided no support and Figure 2. Emancipation in a Conjugal Family Network. The dynamic of emotional support; density = .50; reciprocity = .50 The dynamic of conflict; density = .33; reciprocity = .33 (a) (b) Ambivalence in Later-Life Family Networks 783 Lang, F. R. (2004). The filial task in midlife: Ambivalence and the quality of adult children’s relationships with their older parents. In K. Pillemer & K. Lüscher (Eds.), Intergenerational ambivalences: New perspectives on parent–child relations in later life (pp. 183–206). Oxford, England: Elsevier. https://doi.org/10.1016/S1530-3535(03)04008-1 Lê, S., Josse, J., & Husson, F. (2008). FactoMineR: An R package for multivariate analysis. Journal of Statistical Software,25, 1–18. https://doi.org/10 .18637/jss.v025.i01 Lebart, L., Morineau, A., & Piron, M. (2002). Statistique exploratoire multidimensionnelle [Multidimensional exploratory statistics] (3rd ed.). Paris: Dunod. Letiecq, B. L., Bailey, S. J., & Dahlen, P. (2008). Ambivalence and coping among custodial grandparents. In B. Hayslip, Jr. & P. Kaminski (Eds.), Parenting the custodial grandchild. Implications for clinical practice (pp. 3–16). New York: Springer. Litwin, H. (2001). Social network type and morale in old age. The Gerontologist,41, 516–524. https:// doi.org/10.1093/geront/41.4.516 Lüscher, K. (2002). Intergenerational ambivalence: Further steps in theory and research. Journal of Marriage and Family, 64, 585–593. https://doi .org/10.1111/j.1741-3737.2002.00585.x Lüscher, K. (2005). Looking at ambivalences: The contribution of a “new-old” view of intergenerational relations to the study of the life course. In R. Levy, P. Ghisletta, J.-M. Le Goff, D. Spini, &E.D.Widmer(Eds.),Towards an interdisciplinary perspective on the life course. Advances in life course research (Vol. 10, pp. 93–128). Oxford, England: Elsevier. https://doi.org/10.1016/S10402608(05)10003-3 Lüscher, K., & Hoff, A. (2013). Intergenerational ambivalence: Beyond solidarity and conflict. In I. Albert & D. Ferring (Eds.), Intergenerational relations: European perspectives in family and society (pp. 39–63). Bristol, England: Policy Press. Lüscher, K., & Pillemer, K. (1998). Intergenerational ambivalence: A new approach to the study of parent–child relations in later life. Journal of Marriage and the Family,60, 413–425. https://doi.org/ 10.2307/353858 Moen, P. (1996). Gender, age, and the life course. In R. H. Binstock, L. K. George, V. W. Marshall, G. C. Myers,&J.H.Schulz(Eds.),Handbook of aging and the social sciences (4th ed., pp. 171–187). San Diego, CA: Academic Press. Offer, S. (2012). The burden of reciprocity: Processes of exclusion and withdrawal from personal networks among low-income families. Current Sociology,60, 788–805. https://doi.org/10.1177/ 0011392112454754 Oris, M., Guichard, E., Nicolet, M., Gabriel, R., Tholomier, A., Monnot, C., ...Joye,D.(2016). Representation of vulnerability and the elderly: A total survey error perspective on the VLV survey. In M. Oris, C. Roberts, D. Joye, & M. Ernst-Staehli (Eds.), Surveying human vulnerabilities across the life course (pp. 27–64). Berlin, Germany: Springer. https://doi.org/10.1007/978-3-319-24157-9_2 Phillips, J., Ogg, J., & Ray, M. (2003). Exploring conflict and ambivalence. In A. Lowenstein & J. Ogg (Eds.), The project OASIS: Old age and autonomy: The role of service systems and intergenerational family solidarity, final report (pp. 193–226). Haifa, Israel: Center for Research and Study of Aging. R Development Core Team. (2011). R: A language and environment for statistical computing. Retrieved from http://www.R-project.org/ STATNET Rosow, I., & Breslau, N. (1966). A Guttman health scale for the aged. Journal of Gerontology,21, 556–559. https://doi.org/10.1093/geronj/21.4.556 Saracco, J., Chavent, M., & Kuentz, V. (2010). Rotation in multiple correspondence analysis: A planar rotation iterative procedure. Working Papers of GREThA,2010, 1–19. Retrieved from http://cahiersdugretha.u-bordeaux4.fr/2010/201004.pdf Schnettler, S., & Wöhler, T. (2014). On the supporting role of friendship for parents and non-parents in later life. A comparative analysis using data from the three waves of the German aging survey. In M. Löw (Ed.), Vielfalt und Zusammenhalt: Verhandlungen des 36. Kongresses der Deutschen Gesellschaft für Soziologie in Bochum 2012 [Diversity and Cohesion: Negotiations of the 36th Congress of the German Sociological Association in Bochum 2012] (pp. 1–26). Frankfurt, Germany: Campus. Settersten, R. A., Jr., & Godlewski, B. (2016). Concepts and theories of age and aging. In V. L. Bengtson & R. A. Settersten, Jr. (Eds.), Handbook of theories of aging (3rd ed., pp. 9–25), New York: Springer. Settersten, R. A., Jr., & Mayer, K. U. (1997). The measurement of age, age structuring, and the life course. Annual Review of Sociology,23, 233–261. https://doi.org/10.1146/annurev.soc.23.1.233 Silverstein, M., Bengtson, V. L., & Lawton, L. (1997). Intergenerational solidarity and the structure of adult child–parent relationships in American families. American Journal of Sociology,103, 429–460. https://doi.org/10.1086/231213 Suzman, R., & Riley, M. W. (1985). Introducing the “oldest old.” The Milbank Memorial Fund Quarterly:Health and Society,63, 175–186. https://doi .org/10.2307/3349879. Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd ed.). New York: Harper Collins. 784 Journal of Marriage and Family Taylor, J. E., & Norris, J. E. (2000). Sibling relationships, fairness, and conflict over transfer of the farm. Family Relations,49, 277–283. https://doi .org/10.1111/j.1741-3729.2000.00277.x Treas, J., & Marcum, C. S. (2011). Diversity and family relations in an aging society. In R. A. Settersten, Jr.&J.L.Angel(Eds.),Handbook of sociology of aging (pp. 131–141). New York: Springer. https:// doi.org/10.1007/978-1-4419-7374-0_9 Van Gaalen, R. I., & Dykstra, P. A. (2006). Solidarity and conflict between adult children and parents: A latent class analysis. Journal of Marriage and Family,68, 947–960. https://doi.org/10.1111/ j.1741-3737.2006.00306.x Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications. Cambridge, England: Cambridge University Press. Wenger, G. C. (1991). A network typology: From theory to practice. Journal of Aging Studies,5, 147–162. https://doi.org/10.1016/08904065(91)90003-B Widmer, E. D. (2016). Family configurations: A structural approach to family diversity.NewYork: Routledge. Widmer, E. D., Aeby, G., & Sapin, M. (2013). Collecting family network data. International Review of Sociology,23, 27–46. https://doi.org/10.1080/ 03906701.2013.771049 Willson, A. E., Shuey, K. M., Elder, G. H. Jr., & Wickrama, K. A. S. (2006). Ambivalence in mother–adult child relations: A dyadic analysis. Social Psychology Quarterly,69, 235–252. https:// doi.org/10.1177/019027250606900302