Ambivalence in Older Parent–Adult Child Relationships: Mixed Feelings, Mixed Measures

Abstract

This research compared direct and indirect measures of ambivalence, 2 commonly used strategies for measuring intergenerational ambivalence between older parents and their adult children. Directly and indirectly measured ambivalence, corresponding to felt and potential manifestations of the construct, were contrasted with each other and across generations. Data were derived from 253 older parent–adult child dyads participating in the Longitudinal Study of Generations in 2005. Direct and indirect measures of ambivalence were moderately correlated with each other within each generation. Children expressed greater indirect ambivalence than their parents but were no different than their mothers or fathers in their levels of direct ambivalence. Multivariate regression analyses examining the relationship between each type of ambivalence with individual and relationship characteristics found differences in associations across equations. The results suggest that direct and indirect measures are related but represent 2 distinct conceptions of ambivalence. This research highlights the challenges in understanding the full complexity of intergenerational relations and suggests that both generational perspectives be considered in future research.

Intergenerational ambivalence—mixed or contradictory feelings toward a family member in another generation—has been extensively investigated over the past decades using several different measurement methods; however, little consensus has emerged as to whether these methods have similar conceptual underpinnings and can be similarly interpreted across generational locations. In this research, we investigated two of the most commonly used methods for measuring ambivalence: (a) a direct measure that treats ambivalence as a unitary construct that is acknowledged by the individual and (b) an indirect measure that regards ambivalence as a dualistic construct based on the coexistence of conflict and affection from which ambivalence can be inferred. The first purpose of this study was to investigate whether these two operationalizations are not only empirically distinct but also conceptually distinct by virtue of having different associations with important characteristics typically related to ambivalence. The second purpose was to determine whether ambivalence is experienced differently by parents and adult children, in order to highlight the importance of generational location in the assessment and interpretation of ambivalence.

Ambivalence in Intergenerational Relationships

Ambivalence is the mix of positive and negative emotions toward the same relational object, considered to be both a source of neurosis (Freud, 1913) and an intrinsic property of most human relationships as structured by irreconcilable demands for opposite behaviors (Merton & Barber, 1963). Understudied for years, ambivalence received renewed attention by family scholars in a special issue of the Journal of Marriage and Family in 2002. Since that time, a breadth of empirical research has studied the topic of intergenerational ambivalence, relying on either direct or indirect measurement strategies but with little attention devoted to the degree of overlap in their distinctiveness and meaning. Although several studies have compared these types of measurements (Lüscher & Lettke, 2004; Pillemer & Suitor, 2004; Suitor, Gilligan, & Pillemer, 2011), there remain gaps in our understanding of how these approaches are to be interpreted given that they purport to measure the same underlying construct.

Direct Measure

The unitary character of ambivalence, as assessed by direct measures, has its philosophical roots in sociological literature that conceptualizes ambivalence as greater than the sum of its opposing forces. Coser (1956) elaborated this position by stating “converging and diverging motivations may be so comingled in the actual relationship that they can be separated only for classificatory or analytical purposes, while the relationship actually has a unitary character sui generis.” (p. 64). Lüscher and Pillemer (1998), critiquing approaches that considered parent–child relationships as either close or conflicted, argued that both love and hate can coexist in intergenerational relationships as “contradictions at the subjective level, in terms of cognitions, emotions, and motivations” (p. 416).

Direct measures of ambivalence typically ask a series of questions about the degree to which respondents have opposing feelings about a parent or child. Pillemer and Suitor (2002) developed a quantitative survey-based strategy to directly measure ambivalence by asking respondents to rate the degree to which they have “mixed feelings”; “get on each other’s nerves, but nonetheless feel close”; or feel “torn in two directions” toward a parent or child (Pillemer et al., 2007, p. 782). This direct measure has been used in several studies (Lowenstein, 2007; Pillemer et al., 2007; Pillemer & Suitor, 2002; Suitor et al., 2011) and has shown high reliability in tapping what has variously been labeled subjective, felt, or acknowledged ambivalence about a target individual.

Indirect Measure

An alternative empirical approach to assessing intergenerational ambivalence is rooted in the intergenerational solidarity–conflict framework (Bengtson, Giarrusso, Mabry, & Silverstein, 2002; Lowenstein, 2007). This framework approaches intergenerational relationships as existing on a continuum of multiple dimensions, including affection, consensus, help and support, frequency of contact, geographic distance, norms of familism, and conflict (Bengtson et al., 2002; Bengtson & Roberts, 1991). Conflict, although generally negatively correlated with the other dimensions, may exist at high levels even when affection is strong. Independently measuring both conflict and emotional closeness allows consideration of the contradictions of family life. Parents who have close emotional relationships with their children but also experience conflict with them can be viewed as experiencing ambivalence (Bengtson et al., 2002; Silverstein, Gans, Lowenstein, Giarrusso, & Bengston, 2010). It is important to note that indirect measures of ambivalence allow researchers to infer ambivalence in relationships. In general, ambivalence can be said to exist when positive aspects of a relationship exist in the presence of negative aspects. In this way, the intergenerational solidarity–conflict framework characterizes relationships not as just lacking solidarity but as fully encompassing the richness and complexity of family life.

Several strategies have been used to capture ambivalence indirectly, including additive scales of positive and negative measures that together describe the intensity of discordant emotions (e.g., Fingerman, Pitzer, Lefkowitz, Birditt, & Mroczek, 2008; Willson, Shuey, & Elder, 2003; Willson, Shuey, Elder, & Wickrama, 2006), and classification schemes that group relationships into ambivalent and non-ambivalent types based on the presence of positive and negative emotions (Giarrusso, Silverstein, Gans, & Bengston, 2006; Hogan, Eggebeen, & Clogg, 1993; Silverstein, Bengtson, & Lawton, 1997; Silverstein & Litwak, 1993; Steinbach, 2008; Van Gaalen & Dysktra, 2006). A common approach to measuring ambivalence indirectly uses a set of statements about positive and negative feelings or behavioral interactions and calculates a score that reflects the balance between positive and negative feelings (Fingerman et al., 2008; Willson et al., 2003). Positive and negative items typically used to construct the scale are close in meaning, though typically not identical to those developed within the solidarity–conflict paradigm. The most commonly used scaling strategy relies on a formula developed by Thompson, Zanna, and Griffin (1995) that assesses both balance and intensity in the positive and negative components of relationship quality. For instance, an individual who reports the strongest positive and negative feelings would be considered highly ambivalent, whereas an individual with high positive and low negative (or vice versa) feelings would be considered to have low ambivalence. Someone expressing similarly weak positive and negative feelings does not possess the requisite intensity of feeling to be considered ambivalent.

Several studies have simultaneously examined direct and indirect measures of ambivalence to assess their comparability, finding weak to moderate correlations between them (Lüscher & Lettke, 2004; Pillemer & Suitor, 2004; Suitor et al., 2011). This evidence suggests that two different conceptualizations of ambivalence are being assessed. The direct measurement of ambivalence requires respondents to evaluate their contradictory feelings and be aware of having mixed feelings. In contrast, the indirect measurement is a dualistic approach requiring respondents to independently assess positive and negative relational dimensions. Beyond mechanical differences in the assessment technique, each measurement may have unique correlates, indicating differences in their underlying meanings. To our knowledge, research has yet to examine associations between each ambivalence measure and a full complement of children’s, parents’, and relational characteristics.

Factors Associated With Ambivalence

A wide variety of factors have been found to be associated with ambivalence, including competing obligations and strong kin-keeping responsibilities that create irreconcilable role conflicts (Connidis & McMullin, 2002; Pillemer & Suitor, 2004). The most theoretically interesting individual characteristics associated with ambivalence in intergenerational relations are those that imply need, impairment, or dependence. Parents’ ambivalence toward their adult children has been found to be strongly associated with their children’s problems, such as poor health, educational and career difficulties, and divorced status (Fingerman et al., 2008; Pillemer & Suitor, 2002; Willson et al., 2003); ambivalence presumably flows from unfulfilled expectations for their children. Parents’ ambivalence has also been shown to be related to their own poor health, lack of employment, and stressful relationships with other family members (Kiecolt, Blieszner, & Savla, 2011). Research using both direct and indirect measurements has found that adult children are more ambivalent toward parents with poorer physical and psychological health (Fingerman et al., 2008; Fingerman, Chen, Hay, Cichy, & Lefkowitz, 2006; Wilson et al., 2003).

Studies have also examined ambivalent feelings in terms of their consequences for psychological well-being in older generations. Greater ambivalence is associated with more depression, less life satisfaction, and lower quality of life (Fingerman et al., 2008; Kiecolt et al., 2011; Lowenstein, 2007; Suitor et al., 2011). Suitor et al. (2011) examined reciprocal feelings of ambivalence and found positive associations between ambivalence and psychological well-being as reported by older mothers and their adult children; however, the type of ambivalence mattered only for children.

Relational characteristics between parents and their adult children have also been linked to ambivalence. Intergenerational dependence, as evidenced by caregiving and asymmetrical support provision, has been implicated in the formation of ambivalence. Feelings of ambivalence tended to be stronger among children who provided instrumental assistance to their parents (Fingerman et al., 2006, 2008; Wilson et al., 2003). Using longitudinal data, Hogerbrugge and Komter (2012) found evidence that feelings of ambivalence strengthened among adult children who provided instrumental support to their older parents. Other relational characteristics related to ambivalence in both generations include inequitable exchanges of support and value dissimilarities between parents and their adult children (Fingerman, Hay, & Birditt, 2004; Fingerman et al., 2006; Lüscher, 2004).

Structural position in the family is likely a factor in how ambivalence is manifest in intergenerational relationships. Since the earliest work on intergenerational solidarity, research has shown that parents tend to overreport positive attributes of their relationships with children relative to reciprocal reports by their children (Bengtson & Kuypers, 1971). This positivity bias—labeled the intergenerational stake phenomenon—has been observed with respect to emotional closeness and perceived attitude similarity (Giarrusso, Feng, & Bengtson, 2005; Giarrusso, Stallings, & Bengtson, 1995) as well as conflict (Bengtson, 1996). An explanation for this generational bias is that parents invest emotionally, monetarily, and physically in the raising of their children and thus have a stake in feeling successful in their parental roles; children have a stake in establishing their independence from parents and thus distance themselves by minimizing positive features and emphasizing negative features of the relationship. The role of the intergenerational stake in ambivalent feelings presents a theoretical conundrum. Elevation of affection and suppression of conflict among parents and suppression of affection and elevation of conflict among children would tend to lower indirect ambivalence in both generations. On the other hand, conflict, as assessed in the indirect measure, is relatively rarely acknowledged, making it more susceptible to a generational bias should parents and children disagree in their assessments.

Research Questions and Hypotheses

Our approach contrasted two operationally different measurements of ambivalence across two reciprocal generational perspectives and examined the factors associated with ambivalence by measure and generation. We did not purport to establish “sources” or “outcomes” of ambivalence but drew on the empirical literature to target individual (age, gender, health, filial norms, depression, and self-esteem), relationship (value similarity, support, proximity, and contact) and social structural (generational location) correlates of direct and indirect measures of ambivalence as expressed by older parents and their adult children.

We asked the following research questions and advanced the following hypotheses:

1. How similar are direct and indirect measures of ambivalence within each familial generation? If, as we expected, direct and indirect measures are not equivalent representations of ambivalence, the two measures will be less than highly correlated.

2. Do parents and children express similar levels of ambivalence in their relationships, and does this depend on the measure of ambivalence used? On the basis of the intergenerational stake hypothesis, we expected parents to report less ambivalence than their children, reflecting parents’ positivity bias and children’s negativity bias in assessing their relationships. Furthermore, we expected indirectly measured ambivalence, because it explicitly includes conflict as a component, to exhibit a stronger generational bias than directly measured ambivalence.

3. Are direct and indirect measures of ambivalence associated with the same individual and relationship factors? We expected the two measures to have some overlap in their associations with these factors, but we expected the direct measure, because it assessed ambivalence as a unitary construct, to be associated with more factors than the indirect measure.

Method

Sample

Data were derived from the 2005 wave of the Longitudinal Study of Generations (LSOG). The LSOG began in 1971 with 2,044 respondents who were members of three-generation families. The sample was derived by randomly selecting grandfathers from a population of 840,000 individuals enrolled in southern California’s first large health maintenance organization (see Bengtson & Schrader, 1982, for further details). Adult children (G2) and grandchildren (G3) age 16 and older of participating grandfathers (G1), as well spouses in each generation, were invited to participate in the survey. Follow-up surveys were administered seven times between 1985 and 2005. Longitudinal response rates, accounting for mortality, were 65% from Wave 1 to Wave 2 and have averaged 75% in the follow-ups since 1985 (Feng, Silverstein, Giarrusso, McArdle, & Bengtson, 2006). Mortality attrition accounted for a 21% loss of G2 parents since 1971 (Feng et al., 2006).

The analytic sample comprised 253 G2–G3 parent–child dyads surveyed in 2005 when direct measures of ambivalence were first added to the study. Each G2 mother and father was matched with a randomly designated G3 focal child, allowing contrasts in reciprocal reports about mutual relationships.

Descriptive statistics for parents and children in the dyads are shown in Table 1. The sample includes 253 parents (153 mothers and 100 fathers) who were matched with 179 unique focal children (109 daughters and 70 sons). At the time of survey, the parents were between 64 and 87 years of age and averaged 76 years old; their children were between 37 and 59 years of age and averaged 52 years old. The large majority (95%) of respondents was White non-Hispanic. Seventy-five percent of parents and 70% of children were married and had an average of 3.5 and 2.6 children, respectively. Over 80% of adult children were working at least part time, and only 21% of their parents were still working. In terms of education, parents averaged 14 years, and children averaged 16 years, of formal schooling. Although about one third of parents lived within an hour’s drive of their adult child, less than 10% coresided with them.

Table 1.

Means and Standard Deviations of Analytic Variables From Parents’ and Children’s Perspectives

Variable	Parents			Children
	M	SD	n	M	SD	n
Gender (female)	.60	.49	253	.61	.49	253
Age	76	4.60	253	52	2.90	253
Employed	.21	.41	234	.83	.37	219
Number of children	3.50	1.80	247	2.60	1.30	136
Parent status	1.00		253	.79	.41	243
Years of education	14.00	2.80	230	16.00	3.10	244
Married	.75	.43	252	.70	.45	253
Divorced	.06	.24	252	.20	.40	253
Widowed	.19	.39	252	.02	.14	253
Never married				.07	.26	253
Self-rated health scale	.97	0.18	251	1.88	0.72	253
Depression	1.81	0.29	252	1.55	0.45	253
Self-esteem	3.28	0.38	228	3.26	0.48	253
Filial norms	2.85	0.82	248	3.58	0.61	250
Attitude similarity	4.06	1.20	252	3.85	1.20	246
Provides support	.63	.48	252	.89	.31	230
Proximity (> 50 miles)				.59	.49	249
Frequency of contact				2.90	1.10	252
Affection	4.60	1.15	253	4.30	1.13	251
Conflict	1.78	0.84	253	1.92	1.00	250
Direct ambivalence	.91	.69	248	.96	.69	231
Indirect ambivalence	1.58	1.18	253	1.89	1.29	250

Measures

Ambivalence

The dependent variables represented direct and indirect forms of ambivalence taken from the perspectives of parents and children. Questions were asked of each generation about its relationship with the other. The direct measure includes three questions (see Pillemer & Suitor, 2004): (a) “How often do you feel torn in two directions about your study child (parent) at this point in your life?” (0 = never, 1 = seldom, 2 = now and then, 3 = often, 4 = very often); (b) “I have mixed feelings about this daughter or son (father or mother)”; and (c) “My study child (parent) and I often get on each other’s nerves, but nevertheless we feel close” (0 = strongly disagree, 1 = disagree, 2 = agree, 3 = strongly agree). Cronbach’s alpha for each dyad type (child–mother and child–father) ranged between .58 and .72. After recalibrating the first item to match the range of other two, the three items were averaged, ranging from 0 to 3, with a higher value indicating greater ambivalence. Approximately one third of both children and parents agreed to some extent with each of the three direct ambivalence items. This finding is consistent with previously reported prevalence rates of ambivalence using direct measures (Pillemer & Suitor, 2002).

Indirect ambivalence was calculated using items that assess affectual solidarity and conflict independently. Combining positive and negative aspects of relationships to infer ambivalence has a precedent in intergenerational studies (e.g., Fingerman et al., 2006). For the positive aspects, we averaged three items: (a) “How close do you feel is the relationship between you and your (child/mother/father)?”, (b) “How well do you get along with your (child/mother/father)?”, and (c) “How good is communication between you and your (child/mother/father)?” The following three negative items about the relationship, or conflict, also were averaged: (a) “How much conflict do you feel there is between you and your (child/mother/father)?,” (b) “How much do you feel your (child/mother/father) is critical of you or what you do?”, and (c) “How much does your (child/mother/father) argue with you?” Response categories ranged from 1 (never/not at all) to 6 (extremely/extremely often), indicating greater positive and/or negative aspects. Cronbach’s alpha was greater than .85 for the affectual items and greater than .65 for the conflict items.

These two dialectical assessments of relationship quality were combined to represent indirect ambivalence using an algorithm known as the Griffin formula (Thompson et al., 1995):

• Ambivalence = [(Positive + Negative)/2 − | Positive − Negative|] + 1.5.

This formula combines two necessary components of ambivalence: (a) similar magnitudes of negative and positive components and (b) moderate to strong intensity of both components, such that an individual who reports the strongest positive and negative feelings would be considered to be the most ambivalent as well as more ambivalent than someone else with equal but lower levels of both positive and negative feelings. In the present study, the scale ranged from 0 to 5.25, with a higher score indicating stronger indirect ambivalence.

We acknowledge that the items used here combine assessments of emotional and behavioral aspects of intergenerational relationships and thus are not equivalent opposing assessments. However, in reviewing the indirect measures used in the literature, we found substantial variation in items used to represent indirect ambivalence (Ha & Ingersoll-Dayton, 2008; Kiecolt et al., 2011; Steinbach, 2008; Suitor et al., 2011; Willson et al., 2003). The items available in the LSOG are reliable, validated measures and similar in content to measures used in previous studies.

Personal characteristics

The following personal characteristics of both parents and children were included in the analyses: gender (0 = male, 1 = female), marital status (0 = not currently married, 1 = currently married), number of living children, work status (0 = retired or not working, 1 = currently working full or part time), self-rated health (ranging from 1 [poor] to 4 [excellent]), depression (20 items from the Center for Epidemiologic Studies Depression Scale [Radloff, 1977], on which a higher score indicates more severe depressive symptoms), and self-esteem (10 items from the Rosenberg Self-Esteem Scale [Rosenberg, 1965], on which a higher score indicates greater self-esteem). Personal characteristics of parents only included age (children’s age was not included due to high collinearity with parents’ age). Personal characteristics of children only in the multivariate regression model included number of years of education and parental status (0 = not a parent, 1 = parent). The small sample size did not allow for an analysis of the interaction of the parents’ gender with child’s gender, which has been shown to be significant in research based on indirect measures of ambivalence (Fingerman et al., 2008; Suitor et al., 2011).

Parent–child relationship characteristics

The following variables were included to assess how dimensions of the intergenerational solidarity–conflict model (Bengtson et al., 2002; Bengtson & Roberts, 1991), as perceived by both generations, are associated with ambivalence (Fingerman et al., 2006, 2008; Ha & Ingersoll-Dayton, 2008; Pillemer & Suitor, 2002): functional solidarity (supportive behaviors), normative solidarity (filial norms), consensual solidarity (perceived attitude similarity), structural solidarity (proximity), and associational solidarity (frequency of contact).

Provision of support was measured with two dichotomous variables indicating whether the parent or child provided at least one of the following types of support to the other generation: information and advice, emotional support, discussing important life decisions, visiting or sharing leisure activities, giving help when sick, financial support, help with household chores, transportation, and help with personal care or hygiene. Preliminary analyses determined that emotional support and functional types of support had associations similar to ambivalence for parents and children, and thus we used a combined measure to simplify the analysis.

Filial norms were measured by parents’ and children’s assessments of the degree to which adult children should provide the following types of support to older parents: companionship, household chores and transportation, advice and guidance, personal and health care needs, financial support, and housing (each rated on a 5-point scale ranging from 1[no responsibility] to 5 [complete responsibility]). Perceived consensual solidarity or attitude similarity was measured by the following question: “How similar are your opinions and values about life to those of your child (or parent) at this point in time?” (range: 1 [not at all similar] to 6 [extremely similar]). Proximity of children to parents was measured by asking children to report the number of miles they lived from their parents (0 = more than 50 miles, 1 = 50 miles or less) and frequency of contact by both phone and in person (range: 1 [not at all] to 6 [daily or more often]).

Analytic Strategy

We used correlations to assess the correspondence between direct and indirect ambivalence measurements within each generational perspective by parents’ gender and paired-samples t tests to compare the strength of each type of ambivalence across generations, testing the congruency among parents’ and children’s assessments of indirect and direct ambivalence. Multivariate regression analyses were then estimated to determine the relationship between key independent variables with both ambivalence measures. Regression analyses were run using Mplus Version 4.0 to simultaneously predict direct and indirect ambivalence for both generational perspectives, allowing for the residual covariance between the four dependent variables. We used full information maximum-likelihood estimation to account for missing data and robust standard errors to adjust significance tests for family clustering.

Results

Correlations between direct ambivalence and indirect ambivalence, and the positive and negative components of indirect ambivalence, by parents’ gender, are shown in Table 2. Direct and indirect measures of ambivalence were significantly correlated for children’s reports about relations with fathers (r = .49) and relations with mothers (r = .59), fathers’ reports about relations with children (r = .49), and mothers’ reports about relations with children (r = .59) Direct ambivalence was negatively correlated with affection and positively correlated with conflict among all dyads. Direct ambivalence and conflict from both perspectives were more strongly associated for mother–child dyads (r = .61 and .60, respectively) than in father–child dyads (r = .36 and .48, respectively).

Table 2.

Correlation Coefficients of Direct Ambivalence With Indirect Ambivalence, Affection, and Conflict

Dyad	Direct ambivalence with indirect ambivalence	Direct ambivalence with affection	Direct ambivalence with conflict
Fathers about children	.488***	−.387***	.362***
Children about fathers	.494***	−.293***	.477***
Mothers about children	.589***	−.572***	.613***
Children about mothers	.587***	−.379***	.603***

^***p < .001.

Next, we compared the means of each type of ambivalence, affection, and conflict between matched parents and children; data are shown in Table 3. There were no significant differences in direct ambivalence as expressed by parents and children; however, children expressed significantly greater indirect ambivalence than did mothers and fathers. To further explore compositional differences within the indirect ambivalence measure, we compared means of its two component scales. As shown in Table 3, parents expressed significantly stronger positive feelings about their children than their children expressed toward them, and children reported greater conflict with their mothers than their mothers reported toward them.

Table 3.

Means and Paired t Tests for Differences between Child and Parent Perspectives (N = 253)

Dyad	Direct ambivalence	Indirect ambivalence	Affection	Conflict
Father–child (n = 100)
Child about father	0.94	2.03	4.20	1.84
Father about child	0.88	1.65	4.60	1.72
Paired t test	0.69	2.66**	−4.14***	1.32
Mother–child (n = 153)
Child about mother	0.98	1.80	4.43	1.98
Mother about child	0.94	1.54	4.62	1.82
Paired t test	0.58	2.39*	−2.62*	2.00*

^*p < .05.

^**p < .01.

^***p < .001.

We used multivariate regression analyses to examine relationships between personal and relational factors and each of the ambivalence measures from both parents’ and children’s perspectives. Estimated unstandardized regression coefficients are shown in Table 4. The first two columns show estimates predicting direct and indirect ambivalence as expressed by parents. Among parent’s characteristics, only age was a significant predictor, with older parents having less indirect ambivalence than younger parents. Among children’s characteristics, parents with married children expressed less ambivalence of both types. In terms of relationship characteristics, parents whose children had stronger filial norms, greater value similarity, and provided greater support expressed less direct ambivalence; parents who perceived greater value similarity with their children and who had more frequent contact with them expressed less indirect ambivalence.

Table 4.

Multivariate Ordinary Least Squares Regression of Parents’ and Children’s Direct and Indirect Ambivalence on Personal and Relationship Characteristics (N_dyads = 253)

Independent Variable	Parents’ perspective		Children’s perspective
	Direct ambivalence	Indirect ambivalence	Direct ambivalence	Indirect ambivalence
Parents’ characteristics
Gender (ref.: mother)	.148	−.193	.026	−.354*
Age	.007	−.037*	.007	−.025
Employment (ref.: employed)	−.037	−.137	.048	−.459*
Number of children	−.013	−.033	−.068*	−.071
Currently married (ref.: married1)	.133	.096	−.035	.102
Good self-rated health	.108	−.018	−.039	−.427
Depression	.121	.176	.014	−.229
Self-esteem	−.132	−.170	−.002	.028
Children’s characteristics
Gender (ref.: daughter)	.036	−.027	−.106	−.089
Employment (ref.: employed)	.089	.088	−.083	−.108
Years of education	.000	−.001	.001	−.016
Parental status (ref.: parent)	.062	.081	.054	.126
Currently married (ref.: married)	−.324*	−.450*	.080	.031
Good self-rated health	.038	−.711	−.080	.072
Depression	.033	−.108	.011	−.239
Self-esteem	−.031	−.180	−.416*	−.496*
Relationship characteristics
Parents’ filial norms	.077	−.008	−.019	.007
Children’s filial norms	−.169*	−.122	.167*	.036
Parents’ perceived similarity with children	−.190*	−.376*	−.004	−.034
Children’s perceived similarity with parents	−.015	−.119	−.135*	−.501*
Support from parent	−.021	−.231	.348	.324
Support from children	−.318*	.025	.236*	.532
Proximity (ref.: < 50 miles)	.005	.116	−.186	−.717*
Frequency of contact	−.118	−.306*	.149*	.263*
R2	.291	.319	.231	.327

Note: ref. = reference category.

^*p < .05.

The last two columns of Table 4 show estimates predicting both types of ambivalence as expressed by children. Among parents’ characteristics, gender, employment, and family size were associated with ambivalence. Children had less indirect ambivalence toward mothers than toward fathers. Children whose parents were employed or those who had more siblings also expressed less direct ambivalence than those whose parents were unemployed or those who had fewer siblings. Among children’s characteristics, only self-esteem was associated with ambivalence. Children with greater self-esteem expressed less ambivalence of both types compared to those with lower self-esteem. Relationship characteristics predicting ambivalence included children’s filial norms, perceived similarity to parents, support provided by children, geographic distance, and frequency of contact. Children with stronger filial norms reported greater direct ambivalence than those with weaker filial norms. Children who perceived greater value similarity with parents reported less ambivalence of both types. Children who provided support to parents expressed greater direct ambivalence compared to those who did not provide such support. Children who lived closer to their parents had less ambivalence on the indirect measure compared to those who lived farther away. Children who had greater contact with their parents tended to have greater ambivalence of both types.

Across both sets of equations, more variance was explained for indirect ambivalence compared to direct ambivalence. Among parents, R²s for direct were .291 and .319 for indirect. For children’s reports, R²s were .231 for direct and .327 for indirect.

Discussion

In this study, we compared direct and indirect measures of intergenerational ambivalence to examine their relationship to each other and with multiple individual and relationship characteristics among older parents and adult children. Our investigation was at the measurement level to determine their correspondence within and across generations, and at the conceptual level to distinguish their meaning and interpretation.

Research on intergenerational ambivalence has tended to conflate these two measurement approaches or treat them as interchangeable without considering the underlying conceptual differences. Our analyses suggest that the difference between the unitary approach using direct questions and a dualistic approach using a combination of independent assessments of negative and positive feelings is not trivial. Because it used a sample of parent–child reciprocal dyads, our research lends further evidence that indirect and direct measures correspond to two underlying concepts, labeled potential and felt ambivalence, following the lead of Suitor et al. (2011).

Consistent with our expectations, we found moderate correlations between the two measures of ambivalence indicating a substantial degree of independence between indirect and direct measures of ambivalence that held for reports by children, mothers, and fathers. Furthermore, adding to the work by Suitor et al. (2011), we included fathers as well as mothers in our analysis and found weaker associations between the two measures for fathers reporting about children and children reporting about fathers, when compared to mutual reports by mothers and children. We found no evidence that associations vary by generation, in contrast to Suitor et al., who found weaker relationships between ambivalence measures among children.

Second, we examined the agreement between parent–child dyads in both of the ambivalence measures. Our expectations based on the intergenerational stake hypothesis—that parents would report less of both types of ambivalence than their children—was partially supported by the findings. Parents expressed levels of direct ambivalence similar to those of their children but lower levels of indirect ambivalence indicating their positivity bias. We further tested the components of indirect ambivalence and found this difference to be driven by the stronger affection expressed by parents toward their children and by children’s stronger expression of stronger conflict toward mothers. The positivity of older parents toward their children is consistent with the intergenerational stake phenomenon and leads us to encourage researchers to acknowledge this discrepancy when relying on indirect assessments.

Last, we looked at whether direct and indirect measures of ambivalence were associated with the same individual and relationship factors, comparing the two generations. We expected the two measures to have some overlap in their associations with these factors but the direct measure to be associated with more factors than the indirect measure. Our expectations were partially supported. We indeed found overlap in the measures, but there was also variation in the characteristics associated with the measures, and both indirect and direct measures were associated with the same number of characteristics. Furthermore, we found several generational differences. Among parents, having an unmarried child and less perceived value similarity were related to greater levels of both direct and indirect ambivalence. These findings suggest that parents have a stake in maintaining continuity across the generations and that ambivalence may emerge when expectations for their children are not fulfilled.

Children with low self-esteem, less perceived value similarity, and more frequent contact with parents was related to increased direct and indirect ambivalence. These results are consistent with much of the literature demonstrating a connection between psychological deficits and intergenerational ambivalence (Fingerman et al., 2008). An important contribution of this research is the finding that children who interact frequently with parents are more ambivalent toward their parents. We suggest two explanations to be explored in future longitudinal research: (a) that greater exposure to parents increases the opportunity for conflict to arise in otherwise close relationships or (b) that intense contact with parents may be a strategy for addressing conflict and negotiating ambivalence. That geographic proximity reduced indirect ambivalence suggests that living close to parents—with frequency of contact controlled—may be a manifestation of harmonious relationships.

Despite conceptual overlap in the two measures, direct ambivalence seems more salient to intergenerational dependency and normative values (e.g., support and filial norms), whereas indirect ambivalence is related to parent–child interactions (e.g., contact and proximity). Furthermore, these concepts are related to ambivalence in opposite ways, depending on generational location. When children provided more emotional/instrumental support and had stronger filial norms, parents experienced less direct ambivalence, but under those same conditions children experienced greater direct ambivalence. We saw a similar pattern in terms of parent–child interactions. Children who had more frequent contact experienced more indirect ambivalence, whereas parents had less indirect ambivalence. Ambivalence research has long proposed the theory that interdependency may induce ambivalence, but research has not yet demonstrated this complex generational interaction. We speculate that receiving support from adult children, as a normative expectation of parents, leads to a situation with fewer mixed feelings. Providing support may be burdensome for some children and indicative of parents’ dependency and need, increasing the likelihood for mixed feelings on the part of children. Similar behavioral conditions are perceived differently by each generation, leading to opposite assessments about contradictory feelings in their mutual relationships.

Finally, findings worth discussion were that children’s ambivalence was more related to parental characteristics than parents’ ambivalence was to child characteristics. Children had greater levels of indirect ambivalence toward fathers than mothers when relationship characteristics are considered (see Table 4), supporting the idea that relationships with fathers may be more strained and conflicted than with mothers. In additional, children whose parents were not working, perhaps indicating perceived dependency on the part of the parent, reported greater levels of indirect ambivalence. Children with more siblings reported lower levels of direct ambivalence, suggesting that parents may maintain more diffuse and less highly charged relationships with any one child among many.

Several limitations of this investigation deserve mention. First, the sample was ethnically homogeneous and originally derived from a regional subpopulation that overrepresented White non-Hispanics. Thus, our results and conclusions can be only cautiously generalized to other regions and ethnic subgroups. Nevertheless, it is important to point out that many findings emanating from this sample on the topic of intergenerational solidarity have been replicated in ethnically diverse and multinational populations (Silverstein et al., 2010). Second, because our study design was cross-sectional, it is likely that many of the relationships observed, in particular those related to provision of support, frequency of contact, and psychological well-being, are bidirectional. We emphasize that our results should not be interpreted as causal. For instance, lower self-esteem may be an outcome of having strongly ambivalent relationships with parents or an adverse reaction to the underlying reasons why ambivalence emerged. Longitudinal designs will be needed to better establish directions of influence, although at present such designs are rare in family studies that contain measures of ambivalence. Third, our analyses did not include indicators of more serious children’s problems or losses (e.g., drug abuse and chronic unemployment) that would have enhanced our ability to examine extended dependence and failure to thrive as related to ambivalence within parents. Fourth, because of restrictions in sample size our study precluded the investigation of more intensely dependent caregiving situations when ambivalence among adult children may peak.

This study does not purport to offer a definitive statement about the causes and consequences of intergenerational ambivalence, but it nevertheless provides a useful attempt to clarify how two commonly used measures of ambivalence behave in two linked generations of middle-aged children and their older parents. Interpreting these results in terms of the concepts underlying direct and indirect ambivalence measures is challenging, but referring to direct and indirect measures of ambivalence as felt and potential ambivalence, respectively, in future research may help tease out the conceptual differences. A promising direction for future research will be to integrate these two forms of ambivalence within a common theoretical framework in which potential ambivalence may serve as a precursor to directly expressed ambivalence as the felt manifestation of the construct. Such conceptual development based on whether ambivalent feelings are acknowledged or unacknowledged will help refine our understanding of the interplay between positive and negative valences in intergenerational family relationships. When analyzed together, few of the individual and health-related characteristics were significantly associated with ambivalence; however, most associations involved relational characteristics. Refining the measures of causes and consequences of ambivalence focusing on the indicators of dependency, filial norms, and parent–child interactions may be fruitful for future work on intergenerational ambivalence in late life.

In contrasting the two forms of measuring ambivalence, we conclude that, in terms of verticality, indirect ambivalence is more consistently measured between parents and children but, in terms of construct validity, direct ambivalence is more closely associated with interdependency between generations. Further work is needed to better understand the relationships between indirect ambivalence and its components given perceptual biases on the parts of parents and children. We suggest that researchers consider the possible implications of choosing one measure of ambivalence over the other in terms of their strategic aims, their interpretation of results, and the conclusions they will be able to draw.