The Best of Ties, the Worst of Ties: Close, Problematic, and Ambivalent Social Relationships
Child Development & Family Studies, Fowler Memorial House, 1200 West State Street, Purdue University, West Lafayette, IN 47907-2055 (karenf@purdue.edu).
110 South Henderson Building, Pennsylvania State University, University Park, PA 16802.
Life Course Development Program, Institute for Social Research, University of Michigan, 426 Thompson Street, Ann Arbor, MI.
Abstract
This study builds on research addressing intergenerational ambivalence by considering emotional ambivalence toward the wider social network. Men and women ages 13 to 99 (N = 187) completed diagrams of their close and problematic social relationships. Social ties were classified as solely close, solely problematic, or ambivalent, based on network placement (n = 3,392 social contacts). Multilevel models revealed that individuals viewed certain close familial ties (e.g., spouse, son or daughter, parent, sibling) with greater ambivalence than they viewed more distal family ties, friendships, or acquaintances. Participants classified more acquaintances than other relationships as solely problematic. Feeling closer to a social partner was associated with increased ambivalence. Older adults were more likely to classify their relationships as solely close than as ambivalent, in comparison with younger adults. Discussion focuses on tension and closeness in familial and nonfamilial relationships.
Positive consequences of close family ties and negative consequences of problematic family ties have been well documented (e.g., Antonucci, 2001; Fingerman, 2001a; Rook, 1984). Mixed sentiments (i.e., positive and negative feelings) toward social partners may present unique challenges to individuals, however (e.g., Fingerman & Hay, 2004; Rook; Uchino, Holt-Lunstad, Uno, & Flinders, 2001). For example, loving family members who become overly involved with an individual's stressful life events may exacerbate those stressors (Fingerman & Bermann, 2000; Morgan, 1989; Rook, 2003). A burgeoning literature has brought intergenerational ambivalence to the fore of family science (e.g., Connidis & McMullin, 2002b; Lüscher, 2002; Lüscher & Pillemer, 1998; Pillemer & Suitor, 2002; Willson, Shuey, & Elder, 2003). Yet, studies have not addressed ambivalence in the broader social network. In this study, we consider the types of social partners whom individuals experience as primarily close, primarily problematic, or with mixed sentiments.
Definition and Measurement of Ambivalence
In examining ambivalence, it is important to consider variability in definitions and measurement across studies. Lüscher and Pillemer (1998) distinguished between sociological ambivalence and psychological ambivalence with regard to intergenerational ties. Sociological ambivalence involves incompatible normative expectations (e.g., with regard to status, roles, or norms) that present structural challenges in social relationships. Much discussion in the family literature has focused on this form of ambivalence (e.g., Connidis & McMullin, 2002b; Lüscher, 2002; Pillemer & Suitor, 2002). Psychological ambivalence occurs at the subjective individual level and has to do with contradictions in cognitions, emotions, and motivations such as holding contradictory opinions or feelings toward the same object (Hodson, Maio, & Esses, 2001; Petty & Wegener, 1998; Priester & Petty, 2001; Weigert, 1991).
In this study, we examined psychological ambivalence by assessing the types of relationships that people define as primarily close, primarily problematic, or mixed. Prior studies have shown that perceptions of closeness in a relationship are associated with positive sentiments such as feeling loved and cared for (Antonucci & Akiyama, 1987; Newsom, Morgan, Nishishiba, Rook, 2003; Rossi & Rossi, 1990). Relationship tensions vary in definition but appear to be associated with negative feelings such as irritation and annoyance (Birditt & Fingerman, 2003; Fingerman, 2001a). Theorists argue that all close ties include at least some degree of conflict (Deutsch, 1973; Sillars & Scott, 1983). Yet, these conflicts may be fleeting and may not taint the relationship on the whole. In this study, we examined overarching sentiments about relationships as close and problematic—or some combination of both sentiments—rather than brief emotional experiences.
Prior studies have measured ambivalence directly and indirectly. Direct assessments measure a subjective sense of ambivalence by asking participants whether their feelings toward a given target are mixed (Pillemer & Suitor, 2002; Tourangeau, Rasinski, Bradburn, & D’Andrade, 1989). With indirect assessments, individuals provide separate ratings of their positive and negative attitudes toward an object to detect contradictory ratings or ambivalence (Kaplan, 1972; Priester & Petty, 1996, 2001). Studies pertaining to intergenerational relationships have successfully used direct and indirect assessments of ambivalence (Fingerman, 2001a; Pillemer & Suitor). Given our interest in overarching sentiments, we used indirect assessments of ambivalence.
Our approach to examining ambivalence is consistent with studies of the broader social network. Although this literature has not explicitly addressed ambivalence per se (for an exception, see Uchino et al., 2001), indirect assessments of ambivalence are implicit in studies of problematic social ties. For example, researchers have asked participants to list their close social partners and to report difficulties with these partners (e.g., Ingersoll-Dayton, Morgan, & Antonucci, 1997; Pagel, Erdly, & Becker, 1987; Rook, 1984; Uchino et al.). In these studies, participants generally describe problems with a subset of their close social partners, suggesting that some social ties are experienced as solely close, whereas others are experienced as both close and problematic. Further, relationships that are deemed both close and problematic have implications for individual well-being (Rook; Uchino et al). We build on extant literature by differentiating between close, problematic, and ambivalent ties across the social network.
Ambivalence With Different Types of Social Partners
Individuals’ sentiment toward social partners may vary across different types of relationships. Indeed, empirical studies examining problems with social partners find that kin are more likely to evoke frictions than nonkin (e.g., Rook, 2003). We were interested in whether this pattern also extends into ambivalent sentiments.
Ambivalence in family versus nonfamily ties. Several factors that contribute to relationship ambivalence are likely to arise in family ties. At a macro level, ties between adult family members may be subject to countervailing expectations associated with ambivalence, such as conflicting norms and power imbalances within the relationship (Connidis & McMullin, 2002b; Lüscher & Pillemer, 1998). Nonfamily ties that involve imbalances of power, such as those between coworker and boss, are guided by clearer norms for behavior. Merton and Barber (1963) also argued that ambivalence may be more pervasive in relationships of unclear duration than in relationships that can be terminated. Although family ties are to some extent voluntary in adulthood, individuals may find many such ties difficult to end or replace.
At a micro level, ambivalence among family members may stem from (a) dependence versus autonomy, (b) conflict in norms for relationships, and (c) solidarity versus conflict (Lüscher, 2002; Lüscher & Pillemer, 1998). These factors are particularly evident in ties between adults and their parents, but fundamental tensions between interdependence and individual needs for autonomy are found in many family ties (Fingerman, 2001b). Nonfamily ties such as friendships may also involve tensions between autonomy and interdependence, but individuals experience greater freedom in navigating friendships than they experience in many family ties (Blieszner & Roberto, 2004). Further, family ties often have long histories that contribute to mixed sentiments. Positive sentiments arise from exchanges of affection, mutual activities, or cultural beliefs that family members are supposed to love one another (Bahr & Bahr, 1996). Yet, shared histories also provide fodder for tensions, jealousies, and irritations (Fingerman & Bermann, 2000). As such, familial ties are ripe for ambivalence.
In sum, factors associated with intergenerational ambivalence may apply to family ties more generally. We expected participants to experience more ambivalent sentiments in familial ties as compared with nonfamilial ties.
Types of relationships. Of course, family and nonfamily ties are not homogeneous. Studies examining close and problematic ties have included only a small range of ties (e.g., spouse, child, friend), thus making it difficult to observe heterogeneity across different types of ties (Akiyama, Antonucci, Takahashi, & Langfahl, 2003; Krause & Rook, 2003; Okun & Keith, 1998). In this study, we built on prior research but included a wider range of social ties.
Ambivalence may be more characteristic of family ties to spouses, parents, offspring, and siblings than to other family ties. In comparison with more distal family ties, these ties may involve longer histories, more frequent contact, and increased sense of obligation, factors that can contribute to tensions (Akiyama et al., 2003). By contrast, some extended family ties may involve few tensions, whereas others may be considered solely problematic. For example, positive sentiments might infuse a tie between a grandparent and a grandchild who see one another a few times a year, without the irritations characteristic of ambivalent family ties (Fingerman, 1998). Similarly, individuals might experience some family ties as primarily negative. In his study of social networks, Milardo (1989) found that few participants nominated in-laws as sources of problems, but other research suggests ties to certain in-laws may be tense (Fingerman, in press; Willson et al., 2003). As such, although literature is scant, it appears that individuals may experience ties to extended family members such as grandparents, grandchildren, or in-laws in a unidimensional fashion, as solely close or solely problematic.
Nonkin relationships are also heterogeneous. Individuals encounter social contacts in school, work, religious, and community settings. Some of these nonkin social contacts may be considered friends, whereas others may not. For the sake of convenience, we refer to nonkin ties that are not friendships as acquaintances. In Western cultures, friendships are presumably voluntary and can be disbanded if highly distressful (Blieszner & Roberto, 2004). Therefore, adults may view the majority of friendships as close and positive in nature, with some friendships experienced as ambivalent. Further, we assumed that if individuals did not deem a given acquaintance to be a friendship, it was unlikely that the relationship was solely close and intimate. Such acquaintances may serve as a source of irritation or annoyance, however. For example, Milardo (1989) found that when individuals specified social partners with whom they experienced conflict, they listed coworkers most often. Still other acquaintances might evoke little positive or negative sentiment; we did not consider such ties here, but we recognize that acquaintanceships may fit this description.
Intensity of sentiments. Popular beliefs and social scientific research suggest that intensity of positive sentiments toward social partners is associated with ambivalence. Aphorisms support this idea. “The people who love you most, bug you most,”“Family members are best at pushing your buttons because they are the ones who installed them,” and so forth. Theorists also argue that tensions are more likely to arise in closer social ties than in less close social ties (Deutsch, 1973; Sillars & Scott, 1983). Indeed, a cross-cultural study found that individuals who have more intimate social connections (e.g., spouses, household members) are more likely to report problematic relationships across adulthood (Akiyama et al., 2003).
It is less clear whether intensity of negative sentiments coincides with ambivalence. Theoretically, more intense problems might be associated with lower likelihood of also experiencing feelings of closeness in a given tie, and therefore, with less ambivalence. Yet, this association is unclear. For example, studies of marriage reveal that more conflicted relationships are characterized by lower marital satisfaction (Bradbury, Fincham, & Beach, 2000), but these more conflicted ties might still be ambivalent. As such, we examined the intensity of negative sentiments and the likelihood of experiencing ambivalence without specifying a hypothesis.
Age, Gender, and the Emotional Qualities of Relationships
Men and women of different ages may be more or less likely to classify their social ties as solely close, solely problematic, or ambivalent. Biological, psychological, and sociological factors contribute to age differences in social networks. For example, social and biological factors make it likely that younger adults will have a romantic partner, small children, and living parents, and that older adults will have grown children and grandchildren. Social structures place young adults in educational or work settings and allow many older adults to retire. As such, adults of different ages encounter different types of social partners (Antonucci & Akiyama, 1987; Fingerman & Birditt, 2003; Troll, 1988).
Although adults of different ages tend to report comparable numbers of very close relationships (Antonucci, 2001; Antonucci & Akiyama, 1987), the underlying motivation for social contact may differ in late life. Carstensen's socioemotional selectivity theory suggests that adults actively narrow their network of social partners as they age; older adults retain close social partners, while letting go of less rewarding relationships (Carstensen, 1995; Lansford, Sherman, & Antonucci, 1998). Further, older adults appear to regulate their emotions to view their close social ties with positive feelings. A study comparing middle-aged and older adults revealed that older adults tend to hold a global positive bias toward family members (Winkeler, Filipp, & Boll, 2000). Further, qualitative research finds that adults over the age of 80 report few, if any, problems with their social partners (Johnson & Barer, 1997). Therefore, we expected the oldest old adults to report few problematic or ambivalent social ties, regardless of whether their social contacts were family members or close ties.
Indeed, sentiments toward social partners may vary throughout adulthood. This study included individuals from adolescence to oldest old age, encompassing a wider age range than most studies of social ties. Socioemotional selectivity theory suggests that young adults foster ties to a variety of social partners who can provide information about the culture (Carstensen, 1995; Carstensen, Isaacowitz, & Charles, 1999). In this process, they encounter social partners who are irritating or annoying. As such, young adults may report more ambivalent ties than other adults report. Of course, adolescents are also in the throes of establishing new social partners, but they may lack social cognitive processes that contribute to ambivalence in social ties, and instead may focus on polarized views of social ties as good or bad (Labouvie-Vief, 1994; Perry, 1968).
Gender may also be associated with sentiments toward relationships. Connidis and McMullin (2002b) suggested that social structures contribute to women's experiencing greater relationship ambivalence as compared with men. Women take on more kin work than do men, yet they face countervailing pressures to perform in the work world (Rossi, 1993; Walker, 1999). At a psychological level, women are more invested in their relationships and experience more intense emotions in these ties than do men (Troll & Fingerman, 1999). For example, wives report stronger negative reactions to marital problems than do men (Bolger, DeLongis, Kessler, & Schilling, 1989). Although evidence for gender differences in sentiments toward relationships at a social network level is scant (see Rook, 2003, for a discussion), we expected women to report more ambivalence than men toward social partners.
Method
Sample
These data are from a larger study of close and problematic social ties (Birditt & Fingerman, 2003; Fingerman & Birditt, 2003). Participants included 187 individuals across five age groups: adolescents (ages 13 to 16, n = 39), young adults (ages 20 to 29, n = 40), middle-aged adults (ages 40 to 49, n = 34), young old adults (ages 60 to 69, n = 39), and oldest old adults (age ≥ 80, n = 35). We selected these age groups to maximize possible age effects, so we did not include adults aged 30 to 39, 50 to 59, or 70 to 79. Approximately equal numbers of males and females participated in each age group. Participants rated their health on a scale of 1 (poor) to 5 (excellent), and reported generally good health (M = 2.90, SD = .96). With the exception of adolescents and young adults, men reported more years of education than women. In addition, all women over age 80 were widowed, whereas all but two men over age 80 were married.
Recruitment took place through newspapers, community groups, churches, word of mouth, and individuals attending football games at a large state university. Participants were invited to participate in a study examining positive and negative aspects of relationships. Over 90% of participants were European American, but participants from ethnic minority groups were evenly distributed across age groups. The sample was representative of the county in which the study was conducted (U.S. Census Bureau, 2000), but data do not generalize to the population of the United States.
Procedures
Participants completed interviews pertaining to their background, health, close social contacts, bothersome social contacts, the types of problems they encounter in their relationships, and what they do about those problems. Interviews lasted from 1 to 3 hours. Interviews were conducted between 1997 and 1999.
Background information. Participants provided information about their age, gender, education, marital status, and work status at the start of the interview. The adolescents and many of the young adult participants were students.
Close, problematic, and ambivalent social ties. Participants completed two distinct diagrams of their social contacts: (a) relationships that they considered close and intimate, and (b) relationships that they considered problematic. First, participants completed the Kahn and Antonucci (1980) concentric circle measure of their social networks. This measure has been widely used with adults of all ages (Antonucci & Akiyama, 1987; Antonucci, Akiyama, & Takahashi, in press). Interviewers instructed participants to place social partners to whom they felt so close that they could not imagine life without them in the innermost circle, individuals to whom they felt close but not quite that close in the next circle, and individuals to whom they felt not quite that close but who were still important in their lives in the outer circle. Interviewers indicated that the circles could be empty, full, or anywhere in between.
Then, in a unique adaptation of the measure, participants completed a second diagram pertaining to people who bother them. Participants placed people who bother them quite a bit in the innermost circle, people who bothered them somewhat in the next circle, and people who bothered them occasionally or slightly in the outer circle. Again, circles could be full, empty, or anywhere in between. Two aspects of this diagram are notable. First, participants could name the same social contacts they had named previously, or new social contacts. Second, interviewers provided the caveat that the people listed in these circles did not have to bother participants all of the time, just some of the time. In other words, the instructions explicitly invited participants to list social ties that are ambivalent.
For each social contact they listed, participants provided the first name and last initial of the social contact, gender, age, and relationship to participant. Following the interview, raters classified these social ties as close only, problematic only, or ambivalent. Social contacts named in only one of the social networks were categorized as solely close or solely problematic, whereas social contacts named in both networks were classified as ambivalent. Two raters checked each classification to avoid mistakes in the event of duplicate names in the social networks (e.g., if two women were named Carol S., their relationship and ages were used to cross-check the circles).
Types of relationships. Participants provided subjective definitions of their relationships to each party they listed (e.g., spouse, sister-in-law, roommate, friend). Ties to pets, deceased individuals, media figures (e.g., rock stars, political figures), and deities were excluded from the study. In total, participants provided information pertaining to 3,392 close, ambivalent, and problematic social ties. Our analyses examined ties to spouse, sibling, son or daughter, parent, other family, friend, and acquaintance. Other family included family ties not previously specified, such as in-laws, cousins, aunts, uncles, grandparents, grandchildren, and so forth. Friend included social contacts identified as friend or best friend. Acquaintance included nonkin ties that were not labeled as friendships (e.g., coworker, classmate, church member).
Our analyses included some specific relationships (e.g., spouse, friend) and some larger categories of relationships (other family, acquaintance). To assure that the specific relationships we examined were of importance to participants, we generated a frequency table of each relationship listed by participants (e.g., spouse, classmate, neighbor, mother-in-law). With the exceptions of spouse and parent, which are limited in number per person, the specific relationships examined in this study (e.g., son or daughter, sibling, friend) were those that participants listed most frequently across networks. Within the other family category, grandparents and grandchildren were the most frequently listed ties (n = 168). Nearly all participants who mentioned grandparents or grandchildren listed them as solely close ties, but with few participants classifying these ties as ambivalent and none classifying them as only problematic. We estimated our analyses twice, first including grandparents and grandchildren as distinct categories, and then grouping grandparents and grandchildren with extended family. The analyses confirmed that classifications of grandparents and grandchildren were disproportionately more positive than classifications of other relationships. Yet, the pattern of findings for other relationships was stable, regardless of whether grandparents and grandchildren were examined as distinct categories or grouped with other family. Therefore, we examined seven relationship categories: spouse, sons or daughters, parents, siblings, other family ties, friend, and acquaintance.
In comparisons of familial and nonfamilial ties, we combined spouse, sons and daughters, siblings, and other family ties as family ties. The nonfamily ties included friends and acquaintances.
Intensity rating. Because participants hierarchically classified social contacts in the social network circles, we measured intensity of closeness based on circle placement. Placement in circles 1, 2, or 3 served as an indicator of degree of closeness. We reverse coded circle placement such that a higher number indicated the closest ties. In parallel fashion, we used placement in the problematic network as an indicator of intensity of irritation or negative sentiments in the relationships. Again, a high score indicated placement in the innermost problematic circle.
Prior to conducting analyses, we examined the distribution of different types of relationships by circle placement. Participants disproportionately placed ties to spouses and sons or daughters in the innermost circle of the network of close social ties (95% of spouses, and 80% of sons and daughters). By comparison, only 9% of friendships fell in the inner circle, 40% in the middle circle, and 51% in the outer circle. Relationships were more evenly distributed across the circles of the problematic network (e.g., of spouses listed, 50% fell in the innermost circle, 28% in the middle circle, and 22% in the outer circle). As such, the indicator of closeness (e.g., circle placement) was correlated with certain types of relationships.
Analysis Strategy
The PROC MIXED procedure in SAS was used to examine the fit of multilevel models using residual maximum likelihood estimation (Littell, Milliken, Stroup, & Wolfinger, 1996). The PROC MIXED procedure does not eliminate participants with missing data, thus accommodating for the uneven distributions of relationships across age groups. The procedure estimates the fixed effects of the independent variables and the random effect of the participants.
PROC MIXED estimates linear and nonlinear models that include upper and lower level variables as predictors in the same equation (Singer, 1998). Lower level variables included characteristics of the relationship (e.g., relationship type), and upper level variables included characteristics of the participant (e.g., age). Specifically, we considered two features of the relationship: the type of relationship (e.g., familial versus nonfamilial, or the more specific type of spouse, parent, and so forth) and the intensity of closeness or problems in the tie (e.g., placement in the innermost, middle, or outer circle in the network). We considered age and gender as upper level variables.
The dependent variable was placement of the relationship in one or both networks (close, problematic, ambivalent). To examine the dependent variable, we estimated two equations involving (a) close versus ambivalent classifications, and (b) problematic versus ambivalent classifications. We treated these comparisons as dichotomous outcomes where 1 = ambivalent, and 0 = close or problematic classification. Because the dependent variables were dichotomous, we used nonlinear multilevel analyses in which we specified that the error distribution was binomial. As such, an example equation can be expressed as:
AMBir = a0 + a1(Agei) + a2(Genderi) + a3 (RELTYPEir) + a4(Circleir) + ei + dir where the outcome AMBir is the classification of relationshipr(1 = ambivalent, 0 = close or problematic) by participanti. Genderi is a between-persons level variable for participanti(1 = male, 0 = female). Agei is the age group of participanti. Reltypeir is the type of relationshipr reported by participanti(e.g., family = 1 and nonfamily = 0). Circleir is circle placement (reverse coded such that higher scores indicate greater closeness or more problematic) for relationshipr reported by participanti. The intercept, a0, can be interpreted as the mean of the dependent variable when the predictors are equal to zero; in this case, the intercept is the proportion of ties classified as ambivalent in comparison with close or problematic ties across the sample for all relationships. Finally, a1, a2, and a3 are the slopes for age, gender, and type of relationship (e.g., the increased likelihood of a relationship being classified as ambivalent with increasing age group); ei is the error component associated with between-participant variables; and dir is the error associated with relationships (i.e., error within participants).
In addition, PROC MIXED does not delete missing data listwise; rather it assumes that missing data are random. Given our knowledge that some missing data were not randomly distributed (e.g., problematic ties among oldest old), we estimated our analyses twice, once including the oldest old adults and once omitting them.
Results
Table 1 presents the distribution of relationship classifications (close, ambivalent, problematic) by age group across family and nonfamily ties, and across the seven categories of relationships we examined. We also provide the average number of relationships that individuals in each age group discussed. As can be seen, relationships were generally well distributed across the close and ambivalent categories, though fewer ties were considered solely problematic. In addition, we examined the distribution of ambivalent ties based on circle placement in each network. For the close circles, 30.9% of ties in the innermost circle, 15.3% of ties in the middle circle, and 11% of ties in the outer circle were also found in the problematic network and were therefore considered ambivalent. For the problematic network, 55.6% of ties in the innermost circle, 54.5% of ties in the middle circle, and 59.5% of ties in the outer circle were found in the network of close ties.
| Adolescents | Young Adults | Middle-Aged Adults | Young Old Adults | Oldest Old Adults | N of Relationships | |
|---|---|---|---|---|---|---|
| Relationship | (n = 714) | (n = 655) | (n = 657) | (n = 772) | (n = 594) | (N = 3,392) |
| Family and nonfamily relationships | ||||||
| Family | ||||||
| Close | 0.69 | 0.65 | 0.67 | 0.75 | 0.92 | 0.74 |
| Ambivalent | 0.29 | 0.30 | 0.26 | 0.19 | 0.08 | 0.22 |
| Problematic | 0.03 | 0.05 | 0.07 | 0.07 | 0.00 | 0.05 |
| Nonfamily | ||||||
| Close | 0.62 | 0.59 | 0.71 | 0.76 | 0.86 | 0.70 |
| Ambivalent | 0.13 | 0.18 | 0.10 | 0.08 | 0.04 | 0.11 |
| Problematic | 0.25 | 0.23 | 0.19 | 0.16 | 0.09 | 0.19 |
| Specific types of relationships | ||||||
| Parent | ||||||
| Close | 0.57 | 0.43 | 0.63 | 0.70 | 0.53 | |
| Ambivalent | 0.42 | 0.54 | 0.31 | 0.30 | 0.44 | |
| Problematic | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | |
| Daughter or son | ||||||
| Close | – | 1.00 | 0.46 | 0.60 | 0.94 | 0.71 |
| Ambivalent | – | 0.00 | 0.54 | 0.40 | 0.06 | 0.29 |
| Problematic | – | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| Spouse | ||||||
| Close | – | 0.25 | 0.38 | 0.50 | 0.50 | 0.42 |
| Ambivalent | – | 0.75 | 0.58 | 0.50 | 0.50 | 0.56 |
| Problematic | – | – | 0.04 | – | – | 0.02 |
| Sibling | ||||||
| Close | 0.37 | 0.60 | 0.73 | 0.84 | 0.84 | 0.64 |
| Ambivalent | 0.63 | 0.37 | 0.21 | 0.11 | 0.16 | 0.33 |
| Problematic | – | 0.03 | 0.06 | 0.04 | – | 0.03 |
| Other family | ||||||
| Close | 0.88 | 0.78 | 0.81 | 0.81 | 0.95 | 0.85 |
| Ambivalent | 0.07 | 0.14 | 0.07 | 0.08 | 0.05 | 0.08 |
| Problematic | 0.05 | 0.07 | 0.11 | 0.11 | 0.01 | 0.07 |
| Friend | ||||||
| Close | 0.72 | 0.74 | 0.83 | 0.89 | 0.89 | 0.81 |
| Ambivalent | 0.15 | 0.17 | 0.12 | 0.08 | 0.04 | 0.11 |
| Problematic | 0.12 | 0.09 | 0.05 | 0.03 | 0.07 | 0.07 |
| Acquaintance | ||||||
| Close | 0.32 | 0.34 | 0.46 | 0.43 | 0.77 | 0.44 |
| Ambivalent | 0.05 | 0.20 | 0.07 | 0.07 | 0.05 | 0.10 |
| Problematic | 0.62 | 0.46 | 0.47 | 0.50 | 0.17 | 0.46 |
| M and (SD) for number of relationships | ||||||
| Close | 12.49 (6.40) |
10.38 (6.07) |
13.44 (8.84) |
14.90 (10.37) |
15.14 (11.20) |
13.23 (8.86) |
| Ambivalent | 4.53 (2.97) |
4.00 (2.14) |
4.07 (2.81) |
3.12 (1.95) |
2.06 (1.25) |
3.70 (2.44) |
| Problematic | 3.74 (2.46) |
3.36 (2.21) |
2.81 (1.65) |
3.14 (1.86) |
1.93 (0.88) |
3.11 (2.02) |
We note that frequencies across the seven relationship categories are not comparable. For example, a given respondent could have only one spouse or partner, but could have an array of friends to classify as close, ambivalent, or problematic. Similarly, as expected, relationships reflect life course disparities in social position (e.g., younger people have parents, older adults have sons and daughters), and so are not evenly dispersed across age groups. The multilevel analysis strategy allowed us to take into account these disparities in distributions.
Emotional Classifications of Social Ties
Familial versus nonfamilial ties. We first examined emotional classifications of familial and nonfamilial ties. We grouped together all family ties (e.g., spouse, son or daughter, sibling, parent, other family) and all nonfamily ties (e.g., friends, acquaintances). Table 2 presents findings pertaining to ambivalent versus close ties, and findings pertaining to ambivalent versus problematic ties.
| Ambivalent vs. Closea | Ambivalent vs. Problematicb | |||||
|---|---|---|---|---|---|---|
| Predictor | B | SE | e B | B | SE | e B |
| Relationship level | ||||||
| Intercept | −2.02*** | 0.32 | 0.13 | 1.48*** | 0.38 | 4.39 |
| Innermost circle | – | – | – | – | – | – |
| Middle circle | −0.96*** | 0.12 | 0.38 | 0.21 | 0.17 | 1.23 |
| Outer circle | −1.36*** | 0.14 | 0.27 | 0.34* | 0.17 | 1.40 |
| Family | 0.24* | 0.11 | 1.27 | 2.13*** | 0.16 | 8.41 |
| Individual level | ||||||
| Adolescents | 1.71*** | 0.34 | 5.52 | −0.03 | 0.39 | 0.97 |
| Young adults | 2.02*** | 0.34 | 7.54 | 0.41 | 0.39 | 1.51 |
| Middle aged | 1.55*** | 0.36 | 4.71 | −0.28 | 0.41 | 0.76 |
| Young old adults | 1.31** | 0.35 | 3.71 | −0.35 | 0.40 | 0.70 |
| Oldest old | – | – | – | – | – | – |
| Genderc | 0.00 | 0.17 | 1.00 | 0.16 | 0.21 | 1.17 |
- Note: e B = exponentiated B.
- a Convergence criterion = .001. Estimated parameters 15, AIC = 14,642.80 (N = 2,975 relationships from 187 individuals).
- b Convergence criterion = .001. Estimated parameters 15, AIC = 4,225.50 (N = 957 relationships from 178 individuals).
- c Gender is a dummy variable, 0 = female, 1 = male.
- * p < .05.
- ** p < .01.
- *** p < .001.
As anticipated, in comparison with nonfamily ties, participants were 1.27 times as likely to consider familial ties ambivalent than solely close, and 8.41 times as likely to consider family ties ambivalent as compared with solely problematic. In addition, participants were more likely to classify closer relationships as ambivalent as compared with less close relationships. For categorical independent variables, PROC MIXED provides comparisons between a specific category (in this case, the innermost circle) and the other categories (middle and outer circle). As Table 2 indicates, after controlling for the other variables, social partners in the innermost circle of the close network were 2.63 times as likely to be in the problematic network, and thus to be considered ambivalent, than social partners in the middle circle, and were 3.70 times as likely to be in the problematic network as social partners in the outermost circle.
Whether relationships were considered ambivalent or solely problematic varied only with circle placement in the outermost circle of the problematic network. Social partners considered the least irritating were 1.40 times as likely to be considered ambivalent as compared with social partners who were considered the most irritating. As will be discussed, however, intensity of negative feelings was not predictive of ambivalence when we examined a finer breakdown of relationships.
Age differences emerged in comparisons of ambivalent versus close-only ties, but gender differences did not emerge in these equations. Tukey's posthoc tests for pairwise comparisons of the age groups revealed that the oldest old adults were less likely to consider their relationships ambivalent as compared with other age groups. In addition, young old adults were less likely to consider ties ambivalent than were young adults. Analyses without the oldest old participants revealed stability in these findings.
Age and gender differences did not arise in comparisons involving ambivalent and problematic ties. A disproportionate number of oldest old participants (n = 9) were excluded from the comparisons of ambivalent versus solely problematic ties because they did not include any social partners in their problematic networks, and thus also had no ambivalent ties. When analyses were re-estimated without the oldest old participants, findings remained stable.
Specific types of relationships. We then pursued the more detailed breakdown of relationship classifications with regard to ambivalence. Table 3 includes findings pertaining to classifications of these relationships as ambivalent or close-only ties, as well as findings pertaining to classifications of these relationships as ambivalent or problematic ties.
| Ambivalent vs. Closea | Ambivalent vs. Problematicb | |||||
|---|---|---|---|---|---|---|
| Predictor | B | SE | e B | B | SE | e B |
| Relationship level | ||||||
| Intercept | −1.85*** | 0.33 | 0.16 | 0.19 | 0.45 | 1.21 |
| Innermost circle | – | – | – | |||
| Middle circle | −0.49*** | 0.14 | 0.61 | 0.05 | 0.21 | 1.05 |
| Outer circle | −0.73*** | 0.17 | 0.48 | 0.15 | 0.21 | 1.16 |
| Daughter/son | – | – | – | |||
| Spouse | 0.89** | 0.31 | 2.44 | |||
| Parent | −0.13 | 0.24 | 0.88 | |||
| Sibling | −0.19 | 0.23 | 0.83 | |||
| Other family | −1.59*** | 0.24 | 0.20 | – | – | – |
| Friend | −1.14*** | 0.23 | 0.32 | 0.16 | 0.25 | 1.17 |
| Acquaintance | −0.57* | 0.26 | 0.57 | −1.75*** | 0.25 | 0.17 |
| Individual level | ||||||
| Adolescents | 1.99*** | 0.37 | 7.32 | −0.22 | 0.43 | 0.80 |
| Young adults | 2.16*** | 0.37 | 8.67 | 0.79 | 0.43 | 2.20 |
| Middle-aged adults | 1.49*** | 0.37 | 4.44 | −0.04 | 0.45 | 0.96 |
| Young old adults | 1.34*** | 0.36 | 3.82 | −0.07 | 0.44 | 0.93 |
| Oldest old adults | – | – | – | – | – | – |
| Genderc | 0.03 | 0.17 | 1.03 | 0.13 | 0.24 | 0.88 |
- Note: e B = exponentiated B.
- a Convergence criterion = .001. Estimated parameters 20, AIC = 15,304.20 (N = 2,975 relationships from 187 individuals).
- b Convergence criterion = .001. Estimated parameters 16, AIC = 2,943.30 (N = 667 relationships from 165 participants).
- c Gender is a dummy variable, 0 = female, 1 = male.
- * p < .05.
- ** p < .01.
- *** p < .001.
As in the prior analyses, closer relationships were more likely to be considered ambivalent than were less close relationships, and older adults classified fewer ties as ambivalent than did younger adults. Yet, analyses involving comparisons of ambivalent versus close-only ties for specific relationships revealed complexities. We initially used sons or daughters as the comparison category, given our interest in comparisons with the intergenerational ambivalence literature. Tukey's posthoc tests revealed that relationships with sons or daughters, spouses, siblings, and parents were all more likely to be considered ambivalent than friendships and other family ties. Indeed, ties to spouses were considered with greater ambivalence than most other ties. (Note that Tukey's tests adjust for multiple analyses and are more conservative than findings presented in Table 3; there was no difference between classifications of spouses and sons or daughters with the Tukey's tests.) Classifications of acquaintances did not fit our expectations. In comparison with classification of close-only social ties, participants were more likely to consider acquaintances ambivalent than other family ties or friendships, and were less likely to consider acquaintances ambivalent as compared with spousal ties.
Age differences also emerged in the analyses involving comparisons of ambivalent and solely close ties. As in the prior analyses, Tukey's posthoc tests revealed that in comparison with all other age groups, oldest old participants were less likely to classify relationships as ambivalent. Further, young adults were more likely than young old adults to classify their relationships as ambivalent. When the oldest old were excluded from analyses comparing close-only with ambivalent ties, findings remained stable, though Tukey's posthoc tests revealed two new, significant findings: (a) sons and daughters were viewed with greater ambivalence than acquaintances, and (b) friends were viewed with greater ambivalence than other family members.
In examining ambivalent versus problematic relationships, too few parents, children, spouses, or siblings were classified as problematic only to include in analyses. Therefore, comparisons of ambivalent versus problematic ties focused on other family, friendships, and acquaintances. These analyses did not include participants who only listed spouse, child, parent, or sibling in their problematic network. The 9 oldest old participants who listed no problematic social partners were excluded. Thirteen participants who did not list ambivalent or problematic-only friends, distant family, or acquaintances were also excluded; these 13 individuals were evenly distributed across age groups.
Analyses involving types of relationships with regard to ambivalent and solely problematic ties are also presented in Table 3. In keeping with expectations, Tukey's posthoc tests revealed that participants were more likely to classify acquaintances, as compared with other family or friendship ties, as problematic only rather than ambivalent. These patterns were stable when the oldest old participants were excluded. Intensity of negative feelings did not significantly predict ambivalence in these equations.
In sum, participants placed a greater number of their social contacts in the close network than in the problematic network; as such, there were fewer ambivalent or solely problematic ties as compared with solely close ties. Indeed, certain familial relationships (e.g., spouse, child, parent, sibling) were rarely considered solely problematic. In comparisons of ambivalent and solely close relationships, participants of all ages were more likely to consider familial than nonfamilial ties with ambivalence. Yet, not all family ties evoked the same types of sentiments. Ties to spouses, sons, and daughters were particularly likely to be viewed as ambivalent, especially when oldest old participants were excluded from analyses. In comparison with other ties, acquaintanceships were less likely to be considered ambivalent than solely problematic. Psychologically close ties were more likely to be ambivalent than less close ties. As expected, oldest old adults considered few of their social ties ambivalent or problematic. Young old adults also reported fewer ambivalent relationships than did adolescents and young adults, suggesting that these trends may involve linear age differences.
Discussion
Individuals appear to view some social ties as solely close, other ties as ambivalent, and still other ties as solely problematic. The definition of ambivalence in this study involved a broad subjective sense that the relationship was both close and bothersome. As mentioned previously, overarching sentiments toward relationships (e.g., close only, ambivalent, problematic only) may explain how relationships affect individual and psychological well-being (Okun & Keith, 1998; Uchino et al., 2001). Findings from this study revealed that overarching sentiments toward relationships differ as a function of the type of social partner, intensity of close feelings toward that partner, and individuals’ ages.
Sentiments in Different Types of Relationships
A principal focus of the study involved examining the types of social partners who evoke ambivalent sentiments. We differentiated comparisons between ambivalent and close-only relationships, and ambivalent and problematic relationships. This distinction was important because participants considered only a limited set of relationships to be solely problematic (i.e., friendships, extended family, acquaintances), but considered a wider range of ties to be solely close or ambivalent.
As expected, individuals tended to view family ties with greater ambivalence than nonfamily ties. Further, certain family ties (e.g., spouse, parents, son or daughter) are particularly associated with ambivalent sentiments. Participants reported less ambivalence in other relationships (i.e., friendships and other family members), with sentiments toward acquaintances falling in between. Family ties that evoke ambivalence (e.g., spouse, parents, son or daughter), however, were rarely considered solely problematic.
Intergenerational and spousal ties. Given the prominence of the intergenerational ambivalence literature (e.g., Connidis & McMullin, 2002b; Lüscher, 2002; Lüscher & Pillemer, 1998), we were interested in knowing whether intergenerational ties are more likely to evoke ambivalent sentiments than other familial ties. Findings from this study indicate that individuals are more likely to experience ambivalent sentiments toward parents and sons or daughters than toward most other family members, friends, or acquaintances. As will be discussed, however, individuals experience as much ambivalence in marriage as they do in intergenerational ties.
Our findings parallel studies of intergenerational ambivalence, but also reveal distinctions. For example, Pillemer and Suitor (2002) assessed ambivalence directly by asking older mothers how torn they feel in their ties to their oldest child, the extent to which they feel both love and indifference, and so forth. Those scholars found that only one third of the older mothers reported feeling torn or ambivalent toward their offspring, but more than half of these mothers agreed or strongly agreed with at least one item in the ambivalence scale. In the present study, with the exception of oldest old adults’ relationships, nearly half of the relationships with sons or daughters and parents were classified as ambivalent. Our study assessed overarching sentiments of closeness and irritation, however, rather than specifically examining direct ambivalence, such as feeling torn. It is notable that proportions in this study are similar to those observed by Pillemer and Suitor despite the disparate measurement techniques. Yet, Lüscher (2002) points out that individuals do not always consider feeling torn to be problematic. Future research might investigate whether the same individuals experience direct ambivalence (e.g., feeling torn) and ambivalent sentiments (e.g., close and problematic feelings) in their intergenerational relationships, or whether there are distinct forms of ambivalence.
As noted, this study also reveals that intergenerational ties are not the only family ties that generate ambivalent sentiments; spouses also generate ambivalent sentiments. These findings are in keeping with the large literature documenting conflicts within marriage, even among spouses who generally consider their relationships to be rewarding (Bradbury, et al., 2000; Gottman, 1998; Levenson, Carstensen, & Gottman, 1993). A key issue in this study is that marriages and intergenerational ties tend to include a greater likelihood of cross-classification as problematic and close as compared with other relationships. Two questions arise with regard to these findings: (a) why do individuals experience more ambivalent sentiments in spousal and intergenerational ties than in other ties? and (b) why do some individuals experience this ambivalence, whereas other individuals do not?
Future research might specifically focus on comparisons involving spouses and intergenerational ties, and other types of relationships that individuals encounter. It is not clear whether the same mechanisms would account for ambivalent sentiments in these different ties. Studies of problems with close social partners have attempted to link such feelings with a single mechanism. For example, in a cross-cultural study using national samples in Japan and the United States, Akiyama and colleagues (2003) found that individuals reported more problems with spouses and children than with other social partners. They attributed these findings to shared residence, a factor that is not likely to apply to the parents and offspring in this study. Sociologists have suggested that ambivalence arises in intergenerational ties when normative expectations are murky (Lüscher & Pillemer, 1998; Merton & Barber, 1963). It is unclear how this premise might fit spousal ties, however; spousal ties are legally sanctioned and appear to be governed by social expectations. Rather, as will be discussed, data from the present study suggest that feelings of closeness in spousal and intergenerational ties may contribute to increased likelihood of ambivalence, at least with regard to overarching ambivalent sentiments. It is also possible that marriages and intergenerational ties are more likely to involve ambivalence because of the expectation that these ties should involve support and positive qualities. People may feel more disappointment and irritation when these expectations are not met.
As noted, participants did not experience all spousal or intergenerational ties as ambivalent. Previously, scholars have suggested that intergenerational ambivalence is particularly likely to arise during periods of status transition (Lüscher & Pillemer, 1998), or when individuals are unable to exercise agency in relationships because of structural barriers (Connidis & McMullin, 2002b). It would be interesting to compare relationships to spouses, parents, and offspring with relationships with other social partners during transitions.
Problematic ties. Prior discussion pertaining to familial ambivalence has implicitly involved comparisons with closeness or solidarity in these relationships (Bengtson, Giarrusso, Mabry, & Silverstein, 2002; Connidis & McMullin, 2002a; Lüscher & Pillemer, 1998). The present study included relationships that were deemed solely problematic. Participants tended to view a considerable proportion of their acquaintanceships as solely problematic. Of course, as noted previously, given that participants were free to define their social ties subjectively, it is not surprising that ties deemed friendships are more likely to be close ties than ties that are not considered friendships. In addition, a small proportion of relationships with friends and extended family ties were considered solely problematic. Indeed, as expected, relationships with extended family members (e.g., grandparent, in-law) were more unidimensional (i.e., solely close or solely problematic) than were ties to spouse, child, parent, or sibling.
Finally, our assessment of sentiments toward social partners may not have been complete; we may have missed social partners who evoke few emotional reactions, and social partners who have generated severe problems. In this study, participants listed social partners to whom they feel close, and social partners who bother them. Elsewhere, in a study of living relatives, we noted that participants did not include all family members in their close or problematic social networks (Fingerman & Birditt, 2003). For example, every spouse was listed in the networks, but not all siblings were. Some siblings may have little contact because of life circumstances. Alternately, some participants may have severed ties with certain relatives because of extremely negative circumstances, and may no longer include those relatives in their listings of social partners. Thus, our findings may miss a proportion of ties that are so intensely negative as to be disbanded, or so inconsequential as to warrant no mention.
Intensity of Sentiments
In addition to considering whether certain types of relationships evoke more ambivalent sentiments, we considered the intensity of closeness or irritation toward the social partner. Of interest was whether relationships that are particularly close or particularly problematic are more likely to be experienced as ambivalent.
We found that closer ties were more likely to involve ambivalent sentiments as compared with less close ties. These findings are in keeping with theories suggesting that intimate ties are inherently tension laden (Deutsch, 1973; Sillars & Scott, 1983). Individuals may struggle between their desire to feel close and their desire to remain independent. Indeed, family scholars have noted that contradictions between autonomy and dependency may contribute to feelings of ambivalence in intergenerational ties (Lüscher & Pillemer, 1998; Pillemer & Suitor, 2002). Yet, data from this study suggest that only certain close relationships generate mixed feelings; not all close relationships are also considered problematic.
With regard to problematic ties, we found an association between intensity of irritation (e.g., circle placement in the problematic network) and likelihood of the tie being classified as ambivalent when we examined familial versus nonfamilial ties. Ties in the outer circle of the problematic network (i.e., the least problematic ties) were more likely to be ambivalent. Yet, this association disappeared in the more fine-tuned analysis involving acquaintances, friendships, and extended family. Thus, it appears that ambivalence can arise with varying degrees of negativity.
Questions also remain as to how the two forms of intensity (i.e., feelings of closeness and feelings of irritation) coincide in ambivalent ties. In this study, closer social ties evoked ambivalence. Thus, future research might pursue the varying intensity of negative feelings in ambivalent ties. There may be differences in the qualities of ties to a social partner whom one cannot live without but who evokes only minor irritations, versus a social partner who is found in the inner circles of both social networks. Further, it is possible to use direct assessments of mixed sentiments to consider the intensity of ambivalence (Pillemer & Suitor, 2002).
Age, Gender, and Sentiments Toward Relationships
We also considered age and gender differences in emotional classifications of relationships. One of the most robust findings in the adult development literature involves a steady decrease in problems with social partners across adulthood (Akiyama et al., 2003; Fingerman & Birditt, 2003; Okun & Keith, 1998; Rook, 1984). As expected, in comparison with younger adults, older adults reported more close-only than ambivalent relationships. Although the oldest old adults in particular showed these patterns, young old adults also differed from younger individuals, suggesting a linear age trend. Of course, the data in this study were cross-sectional in nature and permit only age group comparisons. Nonetheless, our findings support socioemotional selectivity theory (Carstensen et al., 1999; Lansford et al., 1998), which suggests that individuals cull their network to include the most rewarding relationships, and that close ties improve with age.
Yet, findings in the present study are not wholly consistent with the literature pertaining to intergenerational ambivalence. Pillemer and Suitor (2002) studied mothers over the age of 60 and found that older mothers reported greater ambivalence toward offspring than did younger mothers. This inconsistency may reflect discrepancies in the measurement of ambivalence. As mentioned previously, Pillemer and Suitor used a direct approach to examine ambivalence, including items such as feeling torn or experiencing love and indifference at the same time. The present study measured ambivalence as overarching feelings of closeness coupled with recognition of the problematic nature of the tie. Social psychologists have noted that direct and indirect assessments of ambivalent attitudes often yield different patterns (Priester & Petty, 1996, 2001). The discrepancies in age differences may reflect such differences. As suggested by socioemotional selectivity theory, older adults may be better able to regulate emotions and relationships (e.g., Carstensen, 1995), and therefore experience fewer relationships with an overarching sense of irritation, as found here. At the same time, older adults may recognize the sensation of feeling mixed or torn, based on more complex cognitive and emotional processes that arise in late life (Labouvie-Vief, 1994; Troll & Fingerman, 1999).
Interestingly, we did not find age differences when comparing ambivalent and problematic-only ties. Of course, oldest old adults were unlikely to classify any social partners as solely problematic; there may simply have been insufficient power to detect age differences with the few oldest old adults who listed problematic-only social partners. It seems likely that older adults experience fewer ambivalent ties and fewer solely problematic ties.
Although we had anticipated that men and women would classify their relationships differently, we found no significant gender differences in this study. Our findings are consistent with studies of problems with social network members, where few gender differences have emerged (see Rook, 2003, for a review). Yet, gender differences in family processes have been well documented throughout adulthood (Allen, Blieszner, & Roberto, 2000; Thompson & Walker, 1989; Walker, 1999), and it is surprising that these gender differences do not extend to overall sentiments about relationships. Of course, it is possible that men and women use different criteria in placing social partners in their close and problematic social networks. Future research might use qualitative data to determine whether men and women share similar perceptions of positivity and negativity in their relationships.
Future Research Examining Ambivalent Sentiments
This study provides a unique contribution by examining individuals’ global classifications and sentiments toward different relationships at a social network level. We adapted a prior technique developed by Kahn and Antonucci (1980) that has been used cross-culturally to assess close and supportive relationships (Antonucci et al., in press). This adapted assessment is easy to administer with a wide age range and could be used cross-culturally as well. The parallel nature of the assessments of the two networks is a particular strength. Elsewhere, scholars have argued that researchers use noncomparable items in assessing close and problematic ties (Newsom et al., 2003). As such, individuals deny problems with their social partners because the items used to assess such problems (i.e., placing demands, being critical) are more extreme than the items tapping positive dimensions of the tie. The assessment procedure in this study alleviates that issue.
Of course, there are several limitations to the present study. The homogeneous sample limits generalizability of findings. Participants represented the county in which the research was conducted, but did not represent the ethnic diversity of the population of the United States. It is possible that classifications of problematic ties might be lower or might involve different types of relationships depending on cultural background. For example, in certain Asian countries, ties between mothers-in-law and daughters-in-law may dominate ambivalent and problematic networks (Gallin, 1994) and may warrant consideration as a distinct category. Ethnic variance in family structure might result in a greater proportion of ambivalent or positive classifications of certain relationship types within the United States as well.
In sum, adults of different ages experience many of their closest family relationships with ambivalence. Indeed, relationships with spouses, parents, and offspring appear to be particularly susceptible to mixed sentiments. Future research might focus on the factors that distinguish these ties from other close relationships in adulthood. Further, it is important to know why some individuals experience ambivalence in these ties, whereas other individuals do not.
Note
This study was funded by grant AG1448401, “Adults’ Reasoning about Social Problems across Adulthood,” and AG17916, “Problems Between Parents and Offspring in Adulthood” from the National Institute on Aging awarded to the first author. This article was written while the first author was funded by a grant from the Brookdale Foundation, the second author was funded by a fellowship for graduate study from the National Science Foundation, and the third author was funded by a postdoctoral fellowship from the National Institute on Aging (T32-AG00117).
Portions of this article were presented at the Transcoop International Workshop on Intergenerational Ambivalence, Constance, Germany, and at the 2001 Gerontological Society of America Annual Conference in Chicago, Illinois.
