Abstract
Although early adversity has been linked to worse mental and physical health in adulthood, few studies have investigated the pathways through which positive and negative dimensions of early experiences can jointly influence psychological well-being in later life. This study examined: (1) profiles of early experiences across multiple domains; (2) the relations of these profiles to hedonic and eudaimonic well-being in later life; and (3) whether midlife social support mediated these relations. We first conducted latent class analysis of early experiences using data from 1,076 men in the VA Normative Aging Study who completed the Childhood Experiences Scale (age: M=69, SD=7). Analyses yielded three profiles of early experiences, labeled as cherished (strong support and some losses), harshly disciplined (harsh parental discipline, low positive reinforcement, and non-normative stressors), and ordinary (few stressors and low parental attention). Next, we applied structural equation modeling to data on a subset of this sample assessed seven years later on hedonic and eudaimonic well-being (n=496; age: M=76, SD=7). In general, the cherished group reported stronger qualitative social support in midlife than the harshly disciplined and ordinary groups, which in turn was related to greater hedonic (life satisfaction, positive affect) and eudaimonic (competence, positive relations with others) well-being in later life. The cherished group also reported higher autonomy than the ordinary group, but this association was independent of midlife social support. Our findings suggest that experiencing adversity in the context of a nurturing early environment can promote successful aging through the maintenance of supportive relationships in midlife.
Keywords: early adversity, hedonic well-being, eudaimonic well-being, successful aging, latent class analysis
Early experiences are a powerful determinant of lifespan developmental trajectories. Adults with high levels of early adversity tend to have difficulties with psychosocial adjustment, often manifesting as poor mental health (Edwards, Holden, Felitti, & Anda, 2003) and relationship functioning (Umberson, Williams, Powers, Liu, & Needham, 2005; Whisman, 2006). A small number of studies have shown that the negative psychosocial consequences of early adversity persist into later life (e.g., Clark, Caldwell, Power, & Stansfeld, 2010). However, few researchers have considered the positive dimensions of early experiences (but see An & Cooney, 2006; Huppert, Abbott, Ploubidis, Richards & Kuh., 2010; Rothrauff, Cooney, & An, 2009), and even fewer have examined the positive later-life outcomes of early experiences. Given tremendous heterogeneity observed in response to early adversity (Rutter, 2013; Werner & Smith, 1992), a person-centered approach may be particularly useful in characterizing the complex combinations of early psychosocial stressors and protective factors, and their effects on developmental outcomes.
The present study examines whether and how early experiences are related to psychological well-being in later life, testing the hypothesis that more positive early psychosocial experiences have far-reaching and positive effects on both hedonic and eudaimonic well-being in later life. In light of the strong association between social ties and mental health, one possible mediating path between early experiences and successful aging is that children who enjoy caring relationships may be more likely to have supportive relationships throughout their lifespan (Mikulincer & Shaver, 2009). Thus, we also examined whether midlife social support mediates the relation between early psychosocial experiences and psychological well-being in later life.
Early Experiences, Later-Life Psychological Well-Being, and Successful Aging
Accumulating evidence supports an association between early adversity and negative psychosocial outcomes in later life. Wilson et al. (2006) found that a composite index of childhood adversity was associated with higher neuroticism, emotional isolation, and smaller social network among older adults. Negative childhood life events, such as parental divorce, are associated with worse mental health in adulthood (Amato & Booth, 1991). In general, childhood adversity is related to higher rates of psychopathology (Clark et al., 2010; Rothrauff et al., 2009; Shaw, Krause, Chatters, Connell, & Ingersoll-Dayton, 2004), greater negative affect (Pitzer & Fingerman, 2010) and stronger affective reactivity to stressors in older adults (Mallers, Charles, Neupert, & Almeida, 2010).
Theories of successful aging distinguish between the absence of disease/ill-being and the presence of health/well-being (e.g., Rowe & Kahn, 1987). Psychological ill-being and well-being are distinct constructs with different psychological and biological correlates (Keyes, 2002; Ryff et al., 2006), as well as genetic, neural, and motivational substrates (Franz et al., 2012; Urry et al., 2004). Thus, one cannot infer any association between early experiences and psychological well-being based on established relations between early adversity and psychological ill-being (e.g., depression). However, few studies have examined adulthood well-being outcomes of positive early experiences. Initial evidence suggests there are long-term benefits of favorable early experiences. Adults who remembered their parents as authoritative, rather than authoritarian or uninvolved, reported greater eudaimonic well-being (Rothrauff et al., 2009). Among women, higher levels of parental care and lower levels of parental authoritarianism and overprotection assessed retrospectively at age 42 were associated with higher levels of eudaimonic well-being ten years later (Huppert et al., 2010). Retrospectively reported early parental warmth and affection were related to greater psychological well-being and social integration in later life (An & Cooney, 2006; Andersson & Stevens, 1993).
Psychological well-being is typically characterized from the theoretical perspectives of hedonism or eudaimonism. The hedonic view emphasizes the evaluation of one’s feelings toward life (Waterman, 2007), as indicated by subjective well-being (Diener, Suh, Lucas, & Smith, 1999), defined as life satisfaction, presence of positive affect, and absence of negative affect (Ryan & Deci, 2001). The eudaimonic view focuses on the pursuit of self-realization and living in accordance with one’s true values (Waterman, 2008). Measurement of eudaimonic well-being has utilized a multidimensional approach to characterize human realization (Ryff & Singer, 1998). Early theories of successful aging (e.g. Rowe & Kahn, 1999) focused on the hedonic aspects of well-being, while more recent work incorporates the construct of resilience within successful aging, and focuses more on eudaimonic characteristics (Aldwin & Igarashi, 2015). While each perspective offers a unique understanding of well-being (Henderson & Knight, 2012), no research has examined both hedonic and eudaimonic well-being as outcomes of early experiences. In this study, we modeled hedonic and eudaimonic well-being as multidimensional constructs and examined how various facets of psychological well-being related to a broad array of positive and negative early experiences.
Early Experiences and Later-Life Well-Being: Mediating Role of Social Support
The effects of early care on developmental outcomes are far-reaching and shape social functioning even in later life. Bowlby’s attachment theory (1973, 1988) identified a secure early relationship with one’s primary caretaker as the foundation for building positive internal representations of the self and others, expectations regarding social interactions and support availability, and patterns of affect regulation and help-seeking. Subsequent research confirmed that there are enduring effects of early experiences on later social functioning (Fraley, Roisman, & Haltigan, 2013; Luecken, Roubinov, & Tanaka, 2013). In a prospective study from infancy into adulthood, securely attached children were more likely to perceive their romantic relationships favorably and have higher interviewer-rated relationship security (Roisman, Collins, Sroufe, & Egeland, 2005). Adults who recalled higher levels of parental care had more positive perceptions of themselves and their current relationships (Sarason et al., 1991). On the other hand, “risky families” characterized either by conflict and aggression, or cold and neglectful relationships present multiple biopsychosocial risks that hinder the acquisition of social competence (Repetti, Taylor, & Seeman, 2002). Children reared within such environments lack role models for complex social skills; they are prone to developing social schemas of self as ineffective and others as unsupportive, information processing biases involving rejection sensitivity and poor attention to social cues, and are biologically predisposed toward physiological reactivity (Miller, Chen, & Parker, 2011). These challenges contribute to difficulties forming trusting relationships and limit the amount of perceived and actual support when needed (Cassidy & Shaver, 2008). Several longitudinal studies have documented the harmful impact of negative early psychosocial environments on subsequent social support and relationship functioning (Colman & Widom, 2004; Gayman, Turner, Cislo, & Eliassen, 2011; Johnson et al., 2002).
Given the legacy of early experiences in social functioning and a robust association between social support and mental health (Bowling et al., 2003; Diener & Seligman, 2002), social support is a likely pathway through which early experiences influence adulthood well-being. This notion is consistent with the chain-of-risk model of life course processes (Kuh, Ben-Shlomo, Lynch, Hallqvist, & Power, 2003), in which a sequence of exposures confers increased (or decreased) risk for later-life conditions. Numerous studies have found that poor social support and relationship problems mediated the association between early adversity and worse mental health outcomes (Shaw et al., 2004; Shaw & Krause, 2002), but whether social support mediates the association between positive early experiences and psychological well-being in later life has not been examined. Moreover, because support and relationship variables were assessed concurrently with the psychosocial outcomes in the aforementioned studies, they were unable to assess the directionality of the associations. To evaluate social support as an antecedent to psychological well-being more rigorously, we assessed social support in midlife and psychological well-being in later life (18 years later). Specifically, we evaluated the quality and quantity of social support in midlife as mediators of the relation between early psychosocial experiences and later-life psychological well-being. Given the moderate temporal stability of these social support constructs (Krause, 1999; Martire, Schulz, Mittelmark, & Newsom, 1999; van Tilburg, 1998), we hypothesized that they would confer long-term benefits on well-being.
Person-Centered Approach to Conceptualizing Early Experiences
A common feature of studies relating childhood experiences to adulthood outcomes is a focus on one type of childhood factor (e.g., adverse events, socioeconomic status). Studies on early adversity also tend to consider more severe aspects of adversity, such as physical or sexual abuse, and rely on samples selected for such experiences (Davidson, Devaney, & Spratt, 2010). Such design features likely limit generalizability to wider populations, whose early conditions encompass a broader and more representative spectrum of experiences.
In contrast, a person-centered approach (Magnusson, 1998) provides a more wholistic conceptualization of childhood experiences. Individuals with similar profiles across multiple indicators of early experiences are classified into a group, or latent class. Linkage of group membership to an outcome suggests an association between a constellation of early environmental factors that define each group and the outcome, akin to a many-way interaction. This approach has been used to discern patterns of childhood adversity in relations to early developmental outcomes (e.g., Dunn et al., 2011; Trickett, Kim, & Prindle, 2011), and examine successful aging as a multidimensional construct (e.g., Morack, Ram, Fauth, & Gerstorf, 2013; Pruchno, Wilson-Genderson, Rose, & Cartwright, 2010). We applied latent class analysis to a broad array of positive and negative early psychosocial experiences, and examined how class membership as a summary index of early life circumstances related to later-life psychological well-being.
The Current Study
In the present study, we examined the association between early psychosocial experiences and later-life psychological well-being. Using a person-centered approach, we represented early psychosocial experiences as a multifaceted construct. Successful aging was conceptualized in terms of both hedonic and well-being. We evaluated a chain-of-risk model involving the quantity and quality of midlife social support as mediators of the association between early experiences and later-life well-being. Data are from men participating in the Veterans Affairs Normative Aging Study (NAS), a long-running study of community-dwelling men born 1903–1945. We tested three hypotheses: First, there would be at least one group with more nurturing early environments, and another group with higher levels of early adversity. Second, men with profiles indicating more nurturing early environments would report greater hedonic and eudaimonic well-being in later life. Third, social support in midlife would mediate the effects of childhood psychosocial experiences on later-life psychological well-being. Thus, men with profiles indicating more nurturing early environments would report higher levels of qualitative and quantitative support in midlife, and stronger midlife social support would be associated with greater hedonic and eudaimonic well-being in later life.
METHODS
Sample
The sample was from the NAS, a longitudinal investigation of normal and pathological aging processes in 2,280 men founded at the Boston VA Outpatient Clinic (Spiro & Bossé, 2001). Between 1961 and 1970, over 6,000 men were screened for the absence of chronic or major physical and mental illnesses, and for geographic stability, defined as kinship ties in the Boston area and stated intentions to remain local.
To identify patterns of early experiences, the analytic sample comprised 1,076 NAS men (mean age=69.3, SD=7.3) who responded to a 1995 mail survey, which contained a childhood experiences measure (response rate=70%). To examine longitudinal associations among retrospectively reported childhood experiences, social support, and psychological well-being, we included a subset of men with data on both childhood experiences assessed in 1995 and psychological well-being assessed in 2003. Of the 1,076 respondents of the 1995 survey, 844 were contacted again in 2003. 816 (97%) were alive at the time of mailing, of whom 85 (10%) refused or were too ill to participate, 235 (29%) did not respond, and 496 (61%) participated; the last group formed the analytic sample for the longitudinal analysis. In 2003, the mean age of this group was 75.5 (SD=6.6). Sensitivity analyses comparing respondents and non-respondents of the 2003 survey indicated minor differences: Respondents were younger (M=67.8 vs. 70.6), and their parents were more likely to have been homeowners when they were young (65% vs. 58%) than non-respondents (Table 1). The two groups did not differ on other childhood socioeconomic status variables, midlife social support, and latent class distribution of early experiences. Of the 2003 respondents, data on midlife social support were available for 467 (94%) from a survey administered in 1985, when their average age was 57.5 (SD=6.6).
Table 1.
Descriptive Statistics for the Latent Class Analysis (LCA) Sample (n = 1076) and a Comparison of Those Who Later Responded (n = 496) vs. Did Not Respond (n = 580) to the 2003 Psychological Well-Being Survey.
LCA Sample (Hypothesis 1, n = 1,076) | 2003 Respondents (Hypotheses 2 and 3, n = 496) | 2003 Non-Respondents (n = 580) | t/Χ2 (df) | |||||||
---|---|---|---|---|---|---|---|---|---|---|
N | M/n (%) | SD | N | M/n (%) | SD | N | M/n (%) | SD | ||
Childhood SES: | ||||||||||
Father’s education | 1022 | 2.6 | 1.6 | 473 | 2.6 | 1.6 | 549 | 2.5 | 1.5 | −1.01 (1020) |
Mother’s education | 1025 | 2.6 | 1.4 | 475 | 2.7 | 1.4 | 550 | 2.6 | 1.4 | −1.78 (1023) |
Father’s occupation | 1032 | 1.9 | 1.4 | 478 | 2.0 | 1.5 | 554 | 1.9 | 1.4 | −0.50 (1030) |
Parental home ownership (1=Y, 0=N) | 1058 | 650 (61.4) | 486 | 316 (65.0) | 572 | 334 (58.2) | 4.87 (1) | |||
| ||||||||||
Midlife Social Support: | ||||||||||
Quantitative support | 963 | 13.4 | 2.5 | 460 | 13.5 | 2.4 | 503 | 13.4 | 2.5 | −0.55 (961) |
Qualitative support | 965 | 6.8 | 2.2 | 463 | 6.9 | 2.1 | 502 | 6.8 | 2.2 | −0.45 (963) |
| ||||||||||
Later-Life Hedonic Well-Being: | ||||||||||
Positive affect | -- | -- | -- | 445 | 32.2 | 7.3 | -- | -- | -- | -- |
Negative affect | -- | -- | -- | 449 | 13.9 | 4.6 | -- | -- | -- | -- |
Life satisfaction | -- | -- | -- | 481 | 25.7 | 5.8 | -- | -- | -- | |
Later-Life Eudaimonic Well-Being: | ||||||||||
Autonomy | -- | -- | -- | 489 | 14.2 | 2.5 | -- | -- | -- | -- |
Environmental mastery | -- | -- | -- | 489 | 15.1 | 2.7 | -- | -- | -- | -- |
Personal growth | -- | -- | -- | 489 | 13.8 | 2.6 | -- | -- | -- | -- |
Positive relations with others | -- | -- | -- | 481 | 14.4 | 2.9 | -- | -- | -- | -- |
Purpose in life | -- | -- | -- | 481 | 14.1 | 2.6 | -- | -- | -- | -- |
Self-acceptance | -- | -- | -- | 481 | 14.5 | 3.0 | -- | -- | -- | -- |
| ||||||||||
Covariates: (all in 2003 except age) | ||||||||||
Age (1996) | -- | -- | -- | 496 | 67.8 | 6.6 | 580 | 70.6 | 7.6 | 6.27 (1074) |
Married (1=Y, 0=N) | -- | -- | -- | 484 | 387 (80.0) | -- | -- | -- | -- | |
Retired (1=Y, 0=N) | -- | -- | -- | 482 | 451 (93.6) | -- | -- | -- | -- | |
Self-rated health (0–4, higher = better) | -- | -- | -- | 477 | 2.8 | 0.9 | -- | -- | -- | -- |
# Health conditions | -- | -- | -- | 488 | 2.9 | 2.0 | -- | -- | -- | -- |
| ||||||||||
Latent Class Assignment: | 0.45 (2) | |||||||||
Class 1: “cherished” | 458 | (42.5) | 215 | (43.3) | 240 | (41.4) | ||||
Class 2: “harshly disciplined” | 157 | (14.6) | 67 | (13.5) | 83 | (14.3) | ||||
Class 3: “ordinary” | 461 | (42.8) | 214 | (43.1) | 257 | (44.3) |
Notes: Bold indicates p ≤ .05. The latent class analysis sample was based on men who completed the Childhood Experiences Scale in 1995. Of these men, a subset responded to the 2003 survey wherein later-life psychological well-being was assessed (n = 496); this served as the longitudinal analysis sample. The t-test and chi-square test reported here refer to comparison between respondents and non-respondents of the 2003 survey. Education variables were coded as 1 = did not complete grammar school, 2 = completed grammar school, 3 = some high school, 4 = completed high school, 5 = some college, 6 = completed college, 7= beyond college. Occupation was coded as 0 = unskilled, 1 = semi-skilled, 2 = skilled/foreman, 3 = white collar, 4 = semi-professional, 5 = professional/managerial/proprietary. Self-rated health was rated as 0 = poor, 1 = fair, 2 = good, 3 = very good, 4 = excellent. Scores for hedonic and eudaimonic well-being reflect observed scores on each measure.
Measures
Early experiences
Early experiences from birth to age 19 were measured with the Childhood Experiences Scale (CES; Aldwin, Levenson, Cupertino, & Spiro, 1998), designed for use with the general population. Domains include close relationships, significant life events, parental discipline, extracurricular activities, parental positive regard, and self-perceptions. The scale included a balance of positive, neutral, and negative items, which likely ameliorated participants’ discomfort in endorsing adverse events (see Figure 1 for item content). Each event or experience was coded dichotomously to indicate occurrence. We excluded two items on global childhood ratings that provided little information about specific events or experiences.
Figure 1.
Figure 1a–d (top to bottom). Estimated CES item endorsement probabilities based on the 3-class model (n = 1076): (a) close relationships and significant life events; (b) parental discipline; (c) positive reinforcements from parents; (d) extracurricular activities, parental positive regard and self-perceptions.
Notes: S = Subject. * Items with odds ratio (OR) of item endorsement ≤0.25 or ≥0.4 between 2 classes. E.g., “1>2” means Class 1 was more likely to endorse the item than Class 2. Probability of item endorsement below 0.3 or above 0.7, as marked by the dashed lines, indicates class homogeneity on an item. † For the item “Parents were rarely pleased” (Fig. 1c), Class 1 had an endorsement probability of 0 and the associated ORs could not be calculated.
In most studies of long-term outcomes of early experiences, one limitation is the reliance on retrospective assessments. Validity threats include forgetting, current mood during the assessment, and a lack of endorsing negative events. However, reviews on retrospective assessment of early adversity concluded that validity and reliability threats are often overstated, and current mood likely has little, if any, effects on recall of childhood events (Brewin, Andrews, & Gotlib, 1993; Hardt & Rutter, 2004). In an independent validation sample aged 22–66, the CES items showed good stability, with 85% having moderate or better kappa values over five years, and all having concordance rates (% item agreement across occasions) above 70% (Yancura & Aldwin, 2009). Age and depression were unrelated to response consistency over time. Importantly, the CES was administered eight years prior to the assessment of psychological well-being, thus reducing the likelihood of reverse causality.
Social support
Following our prior work (Bossé, Aldwin, Levenson, Spiro, & Mroczek, 1993; Bossé, Aldwin, Levenson, Workman-Daniels, & Ekerdt, 1990), we defined quantitative social support using indicators of perceived or structural support (Brissette, Cohen, & Seeman, 2000). It was a manifest variable calculated as the sum of network size and social interaction frequency (Cronbach’s alpha=.69), and computed only for men with complete data on both components. Network size was based on five items assessing marital status, number of people in respondent’s household, number of children, number of relatives within an hour’s drive, and number of close friends. Each item was coded into a three-point scale (0=0, 1=1, 2=2+) and summed. Social interaction frequency was based on three items assessing the frequency of seeing children, friends, and other relatives. Item responses were coded into a three-point scale (0=never to 2=weekly or more) and summed.
Qualitative social support was defined using indicators of perceived or functional support (Wills & Shinar, 2000). It was a manifest variable based on the number of confidants and extent of reliance on family and friends during crisis (Bossé et al., 1990, 1993; Cronbach’s alpha=.61), and calculated only for men with complete data on all components. Number of confidants was considered an indicator of social support quality because it involves the appraisal of a relationship as close and trusting, thus providing information on relationship quality. It was assessed on a three-point scale (0=0, 1=1, 2=2+). Two items assessed the extent to which one could rely on family members and friends for help in a crisis on a four-point scale (0=not at all to 4=completely). Scores on the three items were summed.
In other studies using the NAS sample, both quantitative and qualitative social support were moderately stable over a six-year period (r =.59 and .56, respectively; Toyokawa, 2012), and neither was affected by retirement (Bossé et al., 1993).
Psychological well-being
Psychological well-being was examined via the constructs of hedonic and eudaimonic well-being. Hedonic well-being was represented by indicators of subjective well-being, including positive affect (PA), negative affect (NA), and life satisfaction (LS). PA and NA in the past four weeks were measured with subscales of the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988; Cronbach’s alphas: PA=.90, NA=.87). LS was measured with Liang’s (1984) 11-item version of the Life Satisfaction Inventory: Form A (LSI-A; Neugarten, Havighurst, & Tobin, 1961; Cronbach’s alpha=.76) and five items of the Satisfaction with Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, 1985; Cronbach’s alpha=.89). A confirmatory factor analysis (CFA) conducted on our sample supported a tripartite structure of subjective well-being (Diener, 1984) consisting of correlated latent PA, NA, and LS factors (see Supplemental Document A). For parsimony and to reduce sampling error, when a latent factor was represented by six or more items, we used item parcels created with random assignment as manifest indicators (Little, Cunningham, Shahar, & Widaman, 2002). Each of the latent PA and NA constructs was indicated by three item parcels created from ten PANAS items. The latent LS construct was indicated by two lower-order constructs: the first was indicated by four item parcels created from the LSI-A, and the second was indicated by five SWLS items.
Eudaimonic well-being was assessed with six self-report scales that tapped different aspects of human realization, including autonomy, environmental mastery, personal growth, positive relations with others, purpose in life, and self-acceptance (Ryff & Singer, 1998; Cronbach’s alphas: .64 to .79). Each aspect was modeled as a latent factor indicated by five items. A CFA conducted on our sample supported a model consistent with prior findings (Abbott, Ploubidis, Huppert, Kuh, & Croudace, 2010; Burns & Machin, 2009; see Supplemental Document A), including: (1) A higher-order factor subsuming the environmental mastery, personal growth, purpose in life, and self-acceptance factors. It has been described as a motivational dimension of well-being (Abbott et al., 2010) similar to Ryan and Deci’s (2000) notion of competence. We labeled this factor as competence; it represents a positive evaluation of one’s past, self-efficacy in dealing with challenges, and confidence in one’s ability to grow in a valued direction. (2) Two first-order factors representing autonomy and positive relations with others. (3) Two correlated method factors representing positively- and negatively-worded items, respectively. The three substantive factors were intercorrelated, but they were orthogonal to the method factors.
Covariates
Because childhood socioeconomic status (SES) is a powerful predictor of subsequent physical and mental health (McLaughlin et al., 2011; Miller et al., 2011), we assessed the unique contribution of childhood psychosocial experiences on later-life well-being, net of the effects of childhood SES. Childhood SES was assessed via questionnaire at NAS entry (1961–70). Following Peters et al. (2011), four indicators were used: Paternal and maternal education (1=did not complete grammar school to 7=beyond college), paternal occupation (0=unskilled to 5=professional/managerial/proprietary), and parental home ownership during childhood (1=yes, 0=no). For simplicity, the first principal component of these indicators was used as a composite index of childhood SES.
Age, marital status (1=married, 0=otherwise), and retirement status (1=retired, 0=not) were assessed as part of the 2003 mail survey. Later-life health was assessed in 2003 with two measures: Self-rated health was assessed with an item that ranged from 0=poor to 4=excellent. Number of health conditions was the sum of 17 items assessing physical health conditions (1=ever diagnosed, 0=never).
Analysis Plan
We computed descriptive statistics for all manifest indicators. For hypothesis 1, we used latent class analysis (LCA) to identify patterns of early experiences. For hypotheses 2 and 3, we conducted structural equation model (SEM) analysis to examine the associations among early psychosocial experiences, midlife social support, and later-life psychological well-being. LCA and SEM analyses were conducted with Mplus Version 7.1 (Muthén & Muthén, 1998–2012).
Latent class analysis (LCA)
In LCA, class indicators were 45 dichotomously-coded variables from the CES. Following Masyn (2013), analysis began with a 1-class (k=1) model; k was increased incrementally until the model was no longer well-identified. Model fit was assessed using the Bayes information criterion (BIC; Schwarz, 1978), consistent Akaike’s information criterion (CAIC; Bozdogan, 1987), and the approximate weight of evidence criterion (AWE; Banfield & Raferty, 1993). Relative fit of k vs. (k+1) classes was evaluated with the approximate Bayes factor (BF), the adjusted Lo-Mendell-Rubin likelihood ratio test (LMR-LRT; Lo, Mendell, & Rubin, 2001), and the parametric bootstrapped likelihood ratio test (BLRT; McLachlan & Peel, 2000). The approximate correct model probability (cmP) was used to compare each model against an entire set of fitted models (Kass & Wassserman, 1995).
Among candidate models selected based on fit indices, classification diagnostics were used to evaluate the precision of latent class assignment. Relative entropy is an overall index of classification precision. Class-specific diagnostics included average posterior class probability (AvePP), odds of correct classification ratio (OCC; Nagin, 2005), and modal class assignment proportion (mcaPk). Based on fit indices, classification diagnostics, and model utility, we selected a final, best-fitting, unconditional model (i.e., one omitting predictors and outcomes of classes).
To substantively interpret the classes in the best-fitting model, we plotted the estimated item endorsement probability (PE) for each CES item by class and examined its class homogeneity and class separation. For each item, class homogeneity was defined as PE≤.30 or PE≥.70 (i.e., most men in a class either denying or endorsing the item). Class separation was the odds ratio (OR) of item endorsement between two classes. OR ≥2.5 or ≤0.4 was used to indicate notable class separation. Each class was best characterized by items showing response homogeneity for that class or items with notable class separation.
Structural equation models (SEM)
Based on the best-fitting latent class model, individual probabilities of class membership were used as manifest indicators in SEM to evaluate the longitudinal associations among early experiences, social support, and psychological well-being. Separate models examined hedonic and eudaimonic well-being as outcomes. For each outcome, analysis proceeded as follows: First, we developed measurement models composed of all structural manifest variables (latent class probabilities, childhood SES, age, marital status, retirement status, self-rated health, number of health conditions, qualitative and quantitative social support) and latent variables representing psychological well-being. Correlations among exogenous variables are presented in Supplemental Document B. To evaluate hypothesis 2, we compared a series of structural models, beginning with a saturated model that specified the regression of later-life psychological well-being on latent class membership and covariates. Non-significant paths (p>.10) emitting from covariates were trimmed. For hypothesis 3, the saturated structural model included the regression of midlife social support variables on latent class membership, childhood SES, and age, and the regression of later-life psychological well-being on midlife social support, latent class membership and covariates. Indirect and direct effects from latent class membership to psychological well-being were estimated. Non-significant paths emitting from covariates were trimmed first, followed by non-significant indirect and direct paths. Across all models, psychological well-being and social support were always adjusted for childhood SES and age. To avoid model misspecification, a path was deleted only if its removal did not produce a significant deterioration in model fit. Nested models were evaluated using the Satorra-Bentler (2001) chi-square difference test. Non-nested models were evaluated with the comparative fit index (CFI; Bentler, 1990), Tucker-Lewis (1973) fit index (TLI), root mean square error of approximation (RMSEA; Steiger, 1990), standardized root mean square residual (SRMR; Bentler, 2007), and BIC (Schwarz, 1978). Data missingness and nonlinearity were handled using full information maximum likelihood robust (MLR) estimation. Because bootstrapped confidence intervals (CI) cannot be computed in conjunction with MLR estimation, we calculated the 95%CI of the indirect effects using the distribution of products algorithm in RMediation (Tofighi & MacKinnon, 2011). Indirect effects are considered statistically significant if the 95%CI does not contain 0.
RESULTS
Hypothesis 1: Profiles of Childhood Experiences
Using LCA, we compared one- to nine-class models of the men’s early experiences. Fit indices (Table 2) suggest that the three-class model performed best on multiple criteria, whereas the two-class model performed best on only one. The eight- and nine-class models were poorly identified and not considered further. Relative entropy was higher in the three-class model (.80) than the two-class model (.68), suggesting the former was more precise in classifying individuals. Both models performed adequately on other classification diagnostics (details available upon request). Based on empirical and substantive considerations, the three-class model was selected as the final model. According to the most likely class assignment for each man, the class proportions were: 42.5% (n=458; Class 1), 14.6% (n=157; Class 2), and 42.8% (n=461; Class 3).
Table 2.
Fit Indices for 1- to 7-class Models in Latent Class Analysis of the Childhood Experiences Scale.
k | LL | # par | BIC | CAIC | AWE | †BF |
† Adjusted LMR-LRT p-value |
cmP |
---|---|---|---|---|---|---|---|---|
1 | −20847 | 45 | 42009 | 42054 | 42458 | 1.90 e-154 | .04 | 1.13 e-240 |
2 | −20333 | 91 | 41301 | 41392 | 42209 | 5.97 e-87 | .00 | 5.96 e-87 |
3 | −19974 | 137 | 40904 | 41041 | 42271 | 630.81 | .13 | 0.998 |
4 | −19820 | 183 | 40917 | 41100 | 42743 | 5.49 e+26 | .43 | 0.001 |
5 | −19721 | 229 | 41040 | 41269 | 43326 | 1.19 e+30 | .59 | 2.88 e-30 |
6 | −19629 | 275 | 41178 | 41453 | 43923 | 1.82 e+33 | .58 | 2.42 e-60 |
7 | −19545 | 321 | 41332 | 41653 | 44536 | -- | -- | 1.33 e-93 |
8 | Not well-identified | |||||||
9 | Not well-identified |
Notes: k = number of classes. LL = log-likelihood. # par = number of model parameters. BIC = Bayes Information Criterion. CAIC = consistent Akaike’s information criterion. AWE = approximate weight of evidence criterion. BF = approximate Bayes factor. LMR-LRT = Lo-Mendell-Rubin likelihood ratio test. Bold = best-fitting model for a fit index.
Comparison was made against the (k+1) class. For BIC, CAIC and AWE, smaller values indicate a better fit. For BF, values >1 favors the k-class model over the (k+1)-class model (1>BF>3: weak evidence, 10>BF>3: moderate evidence, BF>10: strong evidence), <1 favors the (k+1)-class model over the k-class model. For the adjusted LMR-LRT, p > 0.05 suggest that k number of classes is adequate for the data relative to (k+1) classes. cmP values across a defined set of models (k = 1 to 7 here) sum to 1. The cmP value for class k approximates the actual probability of this model being correct relative to the remaining models in the given set, assuming equal weight placed on the prior probabilities of all the models. Results for the parametric bootstrapped likelihood ratio test (BLRT) are not shown because all k vs. (k+1) class comparisons yielded p < .0001, suggesting this test was unable to distinguish among the models.
Figures 1a–d are plots of class-specific estimated endorsement probabilities (PE) for each CES item based on the 3-class model. Consistent with the sociohistorical background of poverty (Great Depression) and war (WWII/Korean War) during which most men grew up, family economic hardship was common (PE≥.69). Parental divorce, death or severe injury of family members, severe injuries to oneself, and sexual molestation were uncommon (PE≤.30).
Based on the LCA results, we labeled each latent class to summarize their pattern of early experiences. Class 1 was labeled as cherished children, based on high endorsement of positive relationships, parental discipline that favored milder punishment and use of rewards, and positive parental and self-regard. While this group was most likely to have suffered the loss of a sibling, friend or cousin (“other serious losses”), they also endorsed the highest levels of close familial and external relationships (Figure 1a). They were most likely to endorse milder forms of parental discipline (e.g., sent to room, grounding), and fell between Classes 2 and 3 on harsher discipline (e.g., struck with fists/kicked, sarcasm/harsh comments; Figure 1b). They reported the highest levels of parental positive reinforcement (Figure 1c) and extracurricular activities (Figure 1d). They were also most likely to report being regarded favorably by parents and having a proud accomplishment.
Class 2 was labeled as harshly disciplined children. Their early experiences were notable for harsh parental discipline, less positive reinforcement, and more frequent negative events than other classes. Harshly disciplined children were most likely to experience rare but significant negative events, including parental divorce and sexual molestation (Figure 1a). They were most likely to receive harsh and abusive parental discipline (e.g., struck with fists/kicked, sarcasm), or not receive any discipline at all, and be subjected to unfair punishment and treatment (Figure 1b). Unlike the other classes, a majority of this group recalled their parents as rarely pleased; they received remarkably less verbal praise and fewer special treats, and never received monetary rewards (Figure 1c). They were less likely to have positive regard from their parents than the cherished group, and were more likely to be considered as “hell-raisers” (similar to the current label of “problem child”) than Class 3.
Class 3 was labeled as ordinary children. In general, their early experiences were more neutral compared to the other classes. While they endorsed some close relationships, they were least likely to be close to a non-relative adult (Figure 1a). They reported few negative life events (Figure 1a), especially parental divorce, and lower levels of most forms of parental discipline than other classes (Figure 1b). While ordinary children were equally likely as the cherished children to receive verbal praise, they were less likely to receive special treats (Figure 1c) and parental positive regard (Figure 1d). They were least likely to be considered “hell-raisers,” perhaps a reason for the relatively little parental attention they received. Thus, ordinary children had less adverse childhood experiences than the harshly disciplined children, but they were not held in the special regard given to the cherished children.
Hypothesis 2: Early Experiences and Later-Life Psychological Well-Being
Hedonic well-being
Next, we examined the association between early experiences and later-life psychological well-being. For hedonic well-being as outcome, the latent constructs of LS, PA, and NA were regressed on latent class probabilities and covariates using SEM. From the saturated structural model, Χ2(df)=387.5 (181), CFI=.947, TLI=.934, RMSEA=.048, BIC=20621, SRMR=.035, we eliminated retirement status because it was not related to any outcome, and constrained three non-significant regression paths (number of health conditions → LS; marital status → PA; marital status → NA) to zero. The final model fit the data well: Χ2(df)=382.1 (172), CFI=.945, TLI=.933, RMSEA=.050, BIC=20082, SRMR=.036. Table 3 summarizes the model estimates.
Table 3.
Results from Final Structural Equation Models for Regression of Later-Life Hedonic Well-Being on Early Experiences.
Life Satisfaction | Positive Affect | Negative Affect | |||||||
---|---|---|---|---|---|---|---|---|---|
B | SE | p | B | SE | p | B | SE | p | |
C1: “cherished” (ref.) | -- | -- | -- | -- | -- | -- | -- | -- | -- |
C2: “harshly disciplined” | −0.04 | 0.02 | .06 | −0.17 | 0.10 | .08 | 0.08 | 0.05 | .13 |
C3: “ordinary” | −0.03 | 0.01 | .01 | −0.17 | 0.07 | .02 | −0.02 | 0.04 | .68 |
Childhood SES | 0.01 | 0.01 | .01 | 0.07 | 0.03 | .01 | −0.01 | 0.02 | .39 |
Age | 0.00 | 0.00 | .79 | −0.01 | 0.01 | .06 | −0.00 | 0.00 | .22 |
Married vs. not | 0.04 | 0.01 | .01 | =0 | =0 | ||||
Self-rated health | 0.06 | 0.01 | <.001 | 0.29 | 0.04 | <.001 | −0.13 | 0.02 | <.001 |
# Health conditions | =0 | -- | -- | −0.04 | 0.02 | .04 | 0.02 | 0.01 | .03 |
Notes: Bold: p ≤ .05; Italics: p ≤ .10. =0 represents paths set to 0 in the model trimming process. C1, C2, and C3 refer to probability of being assigned to latent class 1, 2, and 3, respectively. Childhood SES was a composite index based on the first principal component of paternal and maternal education, paternal occupation, and parental home ownership during childhood. Higher scores represent higher childhood SES.
Hypothesis 2 was partially supported: In later life, the cherished group had higher LS and PA than the ordinary group, and marginally higher LS and PA than the harshly disciplined group. Class membership was not associated with NA. Among the covariates, higher childhood SES was associated with higher LS and PA, but unrelated to NA. Married men reported higher LS. Better self-rated health and fewer health conditions were linked to greater hedonic well-being across all indicators, except that the number of health conditions was unrelated to LS.
Eudaimonic well-being
From the saturated structural model for eudaimonic well-being (Χ2 (df)=956.1 (583), CFI=.914, TLI=.901, RMSEA=.036, BIC=39642, SRMR=.041), we eliminated retirement status, number of health conditions, and two non-significant regression paths (marital status → competence; marital status → positive relations with others). The final structural model fit the data well: Χ2(df)=882.1 (531), CFI=.917, TLI=.904, RMSEA=.037, BIC=37060, SRMR=.041. Table 4 summarizes the model estimates.
Table 4.
Results from Final Structural Equation Models for Regression of Later-Life Eudaimonic Well-Being on Early Experiences.
Competence | Autonomy | Positive Relations with Others | |||||||
---|---|---|---|---|---|---|---|---|---|
B | SE | p | B | SE | p | B | SE | p | |
C1: “cherished” (ref.) | -- | -- | -- | -- | -- | -- | -- | -- | -- |
C2: “harshly disciplined” | −0.16 | 0.07 | .02 | 0.03 | 0.06 | .63 | −0.30 | 0.09 | .002 |
C3: “ordinary” | −0.13 | 0.05 | .01 | −0.11 | 0.05 | .04 | −0.27 | 0.07 | <.001 |
Childhood SES | 0.04 | 0.02 | .02 | −0.03 | 0.02 | .15 | −0.01 | 0.02 | .56 |
Age | −0.00 | 0.00 | .18 | 0 | 0.00 | .96 | 0 | 0.00 | .93 |
Married vs. not | =0 | -- | -- | −0.15 | 0.05 | <.001 | =0 | -- | -- |
Self-rated health | 0.16 | 0.03 | <.001 | 0.10 | 0.03 | <.001 | 0.11 | 0.03 | .001 |
Notes:
p ≤ .05,
p ≤ .10. =0 represents paths set to 0 in the model trimming process. C2 and C3 refer to probabilities of being assigned to latent class 2 and 3, respectively (reference = Class 1 “cherished”). Childhood SES was a composite index based on the first principal component of paternal and maternal education, paternal occupation, and parental home ownership during childhood. Higher scores represent higher childhood SES.
Hypothesis 2 was partially supported: In later life, the cherished group had higher levels of competence and positive relations with others than the harshly disciplined group; they also had higher levels of competence, autonomy, and positive relations with others than the ordinary group. Among the covariates, higher childhood SES was related to greater competence, but not other eudaimonic well-being indicators. Married men had lower autonomy. Better self-rated health in later-life was positively associated with all eudaimonic well-being indicators.
Hypothesis 3: Mediation by Midlife Social Support
Hedonic well-being
Next, we considered midlife social support as a mediator of the association between early experiences and later-life hedonic well-being. Fit indices for the fully saturated model were: Χ2(df)=466.4 (213), CFI=.938, TLI=.920, RMSEA=.049, BIC=24768, SRMR=.039. Model trimming eliminated retirement status and 10 additional non-significant paths. The final structural model fit the data well: Χ2df)=463.3 (209), CFI=.937, TLI=.923, RMSEA=.050, BIC=24193, SRMR=.042. R2 were 33.6% for LS, 29.3% for PA, and 18.5% for NA. Figure 2 displays the model estimates.
Figure 2.
Unstandardized structural path coefficients (B) and standard errors (SE, in parentheses) from the final structural model on the longitudinal association among early experiences, midlife social support, and later-life hedonic well-being.
Notes: C2 = latent class 2; C3 = latent class 3; LS = life satisfaction; PA = positive affect; NA = negative affect. Only structural components of the model are displayed. Double-headed arrows represent correlations. Correlations among exogenous variables were modeled but not shown (see Supplemental Document B).
As hypothesized, the cherished group reported higher levels of qualitative and quantitative social support in midlife than the other groups (Table 5, top). Higher qualitative support in midlife was related to higher LS, PA, and lower NA in later life. Contrary to our hypothesis, higher quantitative support in midlife was related to higher NA, but not LS and PA.
Table 5.
Results from Final Structural Equation Model of Midlife Social Support Mediating the Association between Later-Life Hedonic Well-Being and Early Experiences, Adjusted for Childhood SES and Covariates.
Mediation Components:
| |||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Qual. Support | Quant. Support | Life Satisfaction | Positive Affect | Negative Affect | |||||||||||
B | SE | p | B | SE | p | B | SE | p | B | SE | p | B | SE | p | |
C1: “cherished” (ref.) | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- |
C2: “harshly disciplined” | −1.56 | 0.35 | <.001 | −1.10 | 0.38 | .004 | -- | -- | -- | -- | -- | -- | -- | -- | -- |
C3: “ordinary” | −0.93 | 0.23 | <.001 | −0.62 | 0.28 | .03 | -- | -- | -- | -- | -- | -- | -- | -- | -- |
Qual. Support | -- | -- | -- | -- | -- | -- | 0.01 | .003 | .01 | 0.06 | 0.01 | <.001 | −0.02 | 0.01 | .05 |
Quant. Support | -- | -- | -- | -- | -- | -- | =0 | -- | -- | =0 | -- | -- | 0.02 | 0.01 | .01 |
Summary of Direct, Indirect, and Total Effects
| |||||||||
---|---|---|---|---|---|---|---|---|---|
Life Satisfaction | Positive Affect | Negative Affect | |||||||
B | SE | p/95%CI | B | SE | p/95%CI | B | SE | p/95%CI | |
Direct effects: | |||||||||
C2: “harshly disciplined” | = 0 | -- | -- | = 0 | -- | -- | = 0 | -- | -- |
C3: “ordinary” | −0.02 | 0.01 | .09 | = 0 | -- | -- | = 0 | -- | -- |
Indirect effects: | |||||||||
C2 → Qual. Support → outcome | −0.01 | 0.01 | (−0.03, −0.00) | −0.10 | 0.04 | (−0.16, −0.04) | 0.03 | 0.02 | (−0.00, 0.07) |
C2 → Quant. Support → outcome | = 0 | -- | -- | = 0 | -- | -- | −0.02 | 0.01 | (−0.04, −0.00) |
C3 → Qual. Support → outcome | −0.01 | 0.00 | (−0.02, −0.00) | −0.06 | 0.02 | (−0.10, −0.02) | 0.02 | 0.01 | (−0.04, 0.00) |
C3 → Quant. Support → outcome | = 0 | -- | -- | = 0 | -- | -- | −0.01 | 0.01 | (−0.03, 0.00) |
Total effects: | |||||||||
C2 | −0.01 | 0.01 | .04 | −0.10 | 0.04 | .01 | 0.01 | 0.02 | .54 |
C3 | −0.03 | 0.01 | .02 | −0.06 | 0.02 | .01 | 0.01 | 0.01 | .51 |
Notes: Bold: p ≤.05; Italics: p ≤ .10. For indirect effects, 95% confidence intervals (CI) that do not overlap with 0 are considered significant and marked in bold. C2 and C3 refer to the probability of being assigned to latent class 2 and 3, respectively. Qual. Support = qualitative social support. Quant. support = quantitative social support. =0 represents paths set to 0 in the model trimming process. See Figure 2 for all regression coefficients of the mediation model.
As shown in the lower portion of Table 5, midlife qualitative support mediated the relations between childhood experiences and LS and PA, but not NA. Relative to the harshly disciplined and ordinary groups, the cherished group had higher levels of LS and PA in later life because they had greater qualitative support in midlife.
Findings for NA were complicated by its unexpected association with quantitative support. As mentioned, the cherished group had higher levels of both qualitative and quantitative support in midlife than the other groups. However, the two support variables had opposite relations with NA in later life – higher qualitative support was related to lower NA, but higher quantitative support was related to higher NA. Post-hoc analyses1 revealed that the two social support variables, which were positively correlated (r=0.30, p <.001), operated as suppressors (Conger, 1974) in relation to NA. Although qualitative support had a weak positive correlation with NA (r=−0.10, p=.07) and quantitative support was uncorrelated with NA (r=0.07, p=0.17), their associations with NA were strengthened when we accounted for their overlapping variance by entering both as predictors of NA. In other words, after accounting for the protective influence of qualitative support on NA in later life, greater quantitative support was linked to higher subsequent NA. Indirect effects in Table 5 indicate that compared to the cherished group, the harshly disciplined group had lower NA in later life due to lower quantitative support in midlife; they were also linked to higher NA due to lower qualitative support, although this indirect path was non-significant (B=0.03, 95%CI: −0.001, 0.07). The opposite indirect effects yielded a non-significant total effect. A similar pattern of results were obtained for the ordinary group, but in this instance both indirect paths were non-significant.
Eudaimonic well-being
We next examined midlife social support as mediator of the association between early experiences and later-life eudaimonic well-being. Fit indices for the fully saturated model were: Χ2(df)=1052.6 (645), CFI=.911, TLI=.896, RMSEA=.036, BIC=43736, SRMR=.042. Retirement status, number of health conditions, and four non-significant paths were eliminated, producing a final SEM that showed good model-data fit: Χ2(df)=974.3 (591), CFI=.915, TLI=.896, RMSEA=.036, BIC=41142, SRMR=.042. R2 were 26.9% for competence, 11.3% for autonomy, and 29.7% for positive relations with others. Figure 3 displays model estimates and Table 6 summarizes the mediation results.
Figure 3.
Unstandardized structural path coefficients (B) and standard errors (SE, in parentheses) from the final structural model on the longitudinal association among early experiences, midlife social support, and later-life eudaimonic well-being.
Notes: C2 = latent class 2; C3 = latent class 3; COM = competence; AUT = autonomy; REL = positive relations with others. Only structural components of the model are displayed. Double-headed arrows represent correlations. Correlations among exogenous variables were modeled but not shown (see Supplemental Document B).
Table 6.
Results from Final Structural Equation Model of Midlife Social Support Mediating the Association between Later-Life Eudaimonic Well-Being and Early Experiences, Adjusted for Childhood SES and Covariates.
Mediation Components:
| |||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Qual. Support | Quant. Support | Competence | Autonomy | Positive Relations with Others | |||||||||||
B | SE | p | B | SE | p | B | SE | p | B | SE | p | B | SE | p | |
C1: “cherished” (ref.) | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- |
C2: “harshly disciplined” | −1.55 | 0.35 | <.001 | −1.08 | 0.38 | .01 | -- | -- | -- | -- | -- | -- | -- | -- | -- |
C3: “ordinary” | −0.95 | 0.23 | <.001 | −0.63 | 0.28 | .03 | -- | -- | -- | -- | -- | -- | -- | -- | -- |
Qual. Support | -- | -- | -- | -- | -- | -- | 0.04 | 0.01 | <.001 | 0.02 | 0.01 | .10 | 0.09 | 0.01 | <.001 |
Quant. Support | -- | -- | -- | -- | -- | -- | =0 | -- | -- | =0 | -- | -- | 0.02 | 0.01 | .04 |
Summary of Direct, Indirect, and Total Effects
| |||||||||
---|---|---|---|---|---|---|---|---|---|
Competence | Autonomy | Positive Relations with Others | |||||||
B | SE | p/95%CI | B | SE | p/95%CI | B | SE | p/95%CI | |
Direct effects: | |||||||||
C2: “harshly disciplined” | −0.12 | 0.05 | .03 | =0 | -- | -- | −0.17 | 0.08 | .04 |
C3: “ordinary” | −0.10 | 0.05 | .04 | −0.11 | 0.05 | .03 | −0.18 | 0.06 | .003 |
Indirect effects: | |||||||||
C2 → Qual. Support → outcome | −0.07 | 0.02 | (−0.12, −0.03) | −0.03 | 0.02 | (−0.07, 0.01) | −0.14 | 0.04 | (−0.21, −0.07) |
C2 → Quant. Support → outcome | = 0 | -- | -- | = 0 | -- | -- | −0.02 | 0.01 | (−0.05, −0.00) |
C3 → Qual. Support → outcome | −0.04 | 0.02 | (−0.08, −0.02) | −0.02 | 0.01 | (−0.04, 0.00) | −0.08 | 0.03 | (−0.13, −0.04) |
C3 → Quant. Support → outcome | = 0 | -- | -- | = 0 | -- | -- | −0.01 | 0.01 | (−0.03, 0.00) |
Total effects: | |||||||||
C2 | −0.19 | 0.06 | .002 | −0.03 | 0.02 | .10 | −0.32 | 0.09 | <.001 |
C3 | −0.14 | 0.05 | .004 | −0.12 | 0.05 | .01 | −0.27 | 0.07 | <.001 |
Notes: Bold: p ≤ .05; Italics: p ≤ .10. For indirect effects, 95% confidence intervals (CI) that do not overlap with 0 are considered significant and marked in bold. C2 and C3 refer to the probability of being assigned to latent class 2 and 3, respectively. Qual. Support = qualitative social support. Quant. support = quantitative social support. =0 represents paths set to 0 in the model trimming process. See Figure 3 for all regression coefficients of the mediation model.
As before, the cherished group reported higher qualitative and quantitative social support in midlife than either the harshly disciplined and ordinary groups. Higher qualitative support in midlife was associated with higher competence and positive relations with others, and marginally higher autonomy in later life. Higher quantitative support in midlife was related only to greater positive relations with others in later life.
Supporting Hypothesis 3, higher qualitative support in midlife mediated the higher levels of later-life competence and positive relations with others in the cherished group, compared to the harshly disciplined and ordinary groups. After accounting for the indirect effects via social support, the direct pathways from early experiences to these outcomes remained significant, suggesting other unmeasured factors could explain the associations.
With regard to autonomy, the mediation hypothesis was not supported. The significant total effect for the ordinary group and marginal total effect for the harshly disciplined group suggested the cherished group had greater autonomy than both groups of men, but these associations were independent of midlife social support (i.e., no indirect effects).
Quantitative support in midlife was only related to positive relations with others in late life. Specifically, compared to the cherished group, the harshly disciplined group had lower levels of positive relations with others in later life, and this could be attributed to their lower levels of midlife quantitative support. Otherwise, the association between early experiences and eudaimonic well-being were independent of midlife quantitative support.
DISCUSSION
Based on self-report of early psychosocial experiences, this study identified three groups of men, labeled as cherished, harshly disciplined, and ordinary. The groups were most distinct on parental discipline, social support availability, and negative life events during childhood. As hypothesized, the early experiences of the cherished group were linked to greater hedonic and eudaimonic well-being in later life compared to those of the other two groups, independent of childhood SES, later-life health and marital status. Quality, but not quantity, of midlife social support mediated the relation between nurturing early experiences and greater psychological well-being in later life.
Profiles of Childhood Experiences
Unique combinations of variables across domains characterized the profiles of childhood experiences, reinforcing the notion that focusing solely on adversity, and excluding neutral or positive experiences, do not fully capture the heterogeneity of early experiences. This heterogeneity is particularly well-illustrated by the profile of the cherished children. Despite having the most nurturing early experiences, this group also had the highest rate of serious losses in childhood. Numerous buffers likely ameliorated the long-term negative effects of these losses on their later-life well-being. Presence of close relationships inside and outside of the family reflects secure early attachment, available support and positive role models. Compared to other groups, the cherished group reported that their parents more frequently used milder forms of discipline, provided positive reinforcement, and held them in positive regard. High participation in extracurricular activities suggests the presence of external support, and opportunities to gain practical skills and develop a sense of belonging to a community.
Early experiences were less favorable among the harshly disciplined and ordinary groups. Their profiles suggest barriers to developing secure attachments. Bowlby (1982) described three types of support critical to forming secure attachments: Proximity is the physical and psychological presence of a caregiver. Safe haven refers to the primary caregiver as a source of emotional and instrumental support to overcome threats and obstacles. Secure base for exploration refers to primary caregivers providing support for goal pursuits (cf. Mikulincer & Shaver, 2009). For the harshly disciplined group, frequent harsh punishment and unfair treatment by adults resemble aggression-filled “risky families” (Repetti et al., 2002). Their parents were “rarely pleased,” seldom provided reinforcements, and offered little positive regard, thus failing to provide a secure base for age-appropriate explorations and affiliative behaviors. In the context of a hostile environment, the higher frequencies of non-normative stressors (parental divorce, sexual molestation) further suggest the absence of safe havens when confronted with adversity.
The ordinary group experienced relatively few stressors as children. They reported moderate levels of within-family support but less external support than other groups. While they recalled fewer achievements than others, they also received the least parental discipline and were least likely to be considered “hell-raisers.” The lower levels of parental attention received by this group might have been more common among children in the early 20th century. It is also possible that the children were generally well-behaved and required little attention or management from their parents. While we consider these early experiences to be relatively neutral, the amount of support and resources conducive to the development of secure attachment and age-appropriate explorations was perhaps less readily available to the ordinary than the cherished children.
Early Experiences and Midlife Social Support
As hypothesized, the early experiences of the cherished group were associated with higher levels of social support quality and quantity in midlife. Sarason and colleagues posited that key early relationships can shape cognitive representations of the self and others, which in turn form the basis of perceived support in adulthood (Sarason, Pierce, & Sarason, 1990). Secure early relationships, such as those endorsed by the cherished group and lacking in the harshly disciplined group and (to a lesser extent) the ordinary group, promote the belief that one is valued and accepted by significant others. Individuals with a positive sense of acceptance have greater social skills and self-efficacy in social situations. They are more likely to perceive others’ behaviors as supportive and be satisfied with the outcomes of social interactions, which enable them to form and engage in emotionally rewarding relationships (Sarason et al., 1990; 1991). On the other hand, as noted above, early environments of the harshly disciplined and ordinary groups likely restricted the development of secure attachments and social competence, and undermined the ability to form trusting and supportive relationships (Repetti et al., 2002).
Early Experiences, Later-Life Psychological Well-Being, and Mediation by Midlife Social Support
Our second hypothesis was partially supported -- the early experiences of the cherished group were associated with specific aspects of both hedonic and eudaimonic well-being in later life than those of the harshly disciplined and ordinary groups. The similar relations of the latter two groups with later-life well-being suggest that it is perhaps a cherished upbringing in the context of stressors that conferred benefits on later well-being, rather than the magnitude of early adversity per se (which would have yielded stronger relations with well-being for the harshly disciplined group than the ordinary group). Our findings resonate with research identifying strong familial and external support as critical buffers that promoted children’s adaptation to stressors (Werner & Smith, 1992). These longitudinal associations are in line with a lifespan developmental perspective on stress-related growth, wherein early support allowed the cherished children to acquire coping skills in the face of hardship, thus transforming losses into turning points for growth and greater well-being over age (Aldwin, Levenson, & Kelly, 2009). Compared to the cherished group, the ordinary group had lower autonomy in later life, perhaps reflecting lifelong patterns of passivity.
Hedonic well-being
We observed higher levels of life satisfaction and positive affect in later life among the cherished compared to the harshly disciplined and ordinary groups. The regression coefficients were similar for the harshly disciplined and ordinary groups, but only marginally significant in the former. The smaller size of the harshly disciplined group (and therefore weaker statistical power) likely contributed to the marginal findings.
Partially supporting Hypothesis 3, quality, but not quantity, of midlife social support mediated the association between early experiences and later-life life satisfaction and positive affect. This is consistent with meta-analytic findings that subjective well-being in old age is more strongly correlated with quality than quantity of social contact (Pinquart & Sörensen, 2000). Sarason et al. (1990, 1991) posited that adults with secure early attachments are more likely to perceive relationships as supportive, which may explain more frequent experiences of positive emotions in the cherished group.
Contrary to our hypothesis, early experiences were unrelated to negative affect in later life. While men with higher quality social support tended to have a greater amount of support, our results indicate that it is the quality, not quantity, of social support that protects against negative affect in later life. In particular, low parental control and support – characteristics of the harshly disciplined group – predispose children to entering and remaining in deviant peer groups (Jessor, 1991). As adults, individuals from adverse childhood environments are more likely to experience relationship difficulties or be in abusive relationships (Colman & Widom, 2004; Umberson et al., 2005), thus a larger social network and more frequent social contact might not be beneficial. While the harshly disciplined group was linked to greater negative affect in later life via lower-quality support in midlife, they also seemed to be protected against subsequent negative affect by having fewer interactions with less supportive networks compared to other men. The opposite associations of qualitative and quantitative support with negative affect reinforce the importance of considering both dimensions in understanding the relations between social support and well-being.
Eudaimonic well-being
The cherished group reported higher levels of competence and positive relations with others in later life than the other groups. The cherished group also reported greater autonomy than the ordinary group. Our findings suggest that supportive early environments can have a lasting impact on the ways individuals actualize their potential over the lifespan.
Stronger qualitative support in midlife mediated the association between early experiences of the cherished group and greater competence in later life. Competence represents self-efficacy in dealing with past and present challenges, and confidence in one’s ability to grow in a fulfilling direction. Being cherished as a child and having the perception of support in adulthood likely enhance one’s readiness to engage with challenges without fear of repercussion (e.g., being ridiculed) or failure. On the other hand, children who lack a secure base for exploration via trusted caregivers are less open to new experiences, which in turn restrict opportunities to gain skills and develop self-efficacy (Sarason et al., 1990). As adults, the perception of inadequate support may lead one to take on fewer challenges and limit exploratory behaviors, in turn contributing to lower self-efficacy and mastery.
Stronger qualitative support in midlife also mediated the association between the early experiences of the cherished group and more positive relations with others in later life. As predicted, cherished children were more capable of garnering support and forming trusting relationships in adulthood, which in turn contributed to a greater capacity for deep friendship, intimacy, and close identification with others in later life.
Ordinary children had lower levels of autonomy in later life, but contrary to expectation, this association was independent of midlife social support. Autonomy refers to an internal locus of evaluation and a sense of freedom from conventions (Ryff, 1989). Fewer childhood stressors and lower parental involvement suggest fewer opportunities to acquire self-knowledge and develop coping skills than the cherished group. While the early experiences of the ordinary group were not necessarily adverse, they were linked to a diminished ability to uphold one’s values and opinions in later life. Mediation results further suggest this association was not explained by perceptions of support in adulthood.
Our findings are consistent with the chain-of-risk model of life course processes. Non-significant correlations between childhood SES and membership in the harshly disciplined and ordinary classes (Supplemental Document B) are inconsistent with the risk clustering model and correlations among adverse childhood experiences observed in the literature (Dong et al., 2004; Green et al., 2010). This discrepancy could be attributable to the severe nature of the adverse events (e.g., domestic violence, parental substance abuse) in other studies, and limited variation in our childhood SES variable given the sociohistorical context. Our findings set the stage for research to investigate additional links in the chain (e.g., adulthood SES) that could explain the association between nurturing early experiences and successful aging. Studies with prospective data on event timing are particularly well-positioned to extend currents findings and investigate whether the effects of adversity (or support) on later-life outcomes are amplified when they occur during key developmental periods (i.e., sensitive period model; Kuh et al., 2003).
Limitations
There are several limitations to this study. As with most studies on later-life outcomes of early adversity, we relied on retrospective assessment of childhood experiences. As noted, this measure was administered eight years prior to the assessment of psychological well-being, which can alleviate some concerns for reverse causality. Also, several studies have found negligible differences between prospective and retrospective reports of early adversity in relation to mental health outcomes (Brown, Craig, Harris, Handley, & Harvey, 2007; Scott, McLaughlin, Smith, & Ellis, 2012).
Our results may have been affected by reporting bias (e.g., over-reporting of stressors, support, and well-being by the cherished group or under-reporting by the ordinary group). However, there was sufficient differentiation in the responses to suggest that this was unlikely – the ordinary group reported levels as high as the cherished group on some CES items (e.g., verbal scolding, spanked), while even the harshly disciplined group reported some adult support. We acknowledge that some CES items (e.g., items on perceptions) are more evaluative than others, and could be influenced by respondents’ appraisals of their childhood2. Unfortunately our data preclude us from evaluating the extent to which objective occurrences vs. appraisals of childhood experiences influence subsequent well-being, but this question merits additional research.
We treated hedonic and eudaimonic well-being as separate constructs, but whether they are theoretically distinct is controversial (e.g., Henderson & Knight, 2012; Kashdan, Biswas-Diener, & King, 2008). Our data indicate varying degrees of overlap between dimensions of hedonic and eudaimonic well-being (r = |0.25 – 0.72|; Supplemental Document C). In response to a reviewer’s request, we repeated the analyses for Hypothesis 3, adjusting for the shared variance between hedonic and eudaimonic well-being. Although some of the previously important associations were attenuated or became nonsignificant, early experiences were still linked to dimensions of hedonic (i.e., NA) and eudaimonic well-being (i.e., competence and positive relations with others) via midlife social support (Supplemental Documents D-E). However, this type of adjustment introduces multicollinearity among the predictors, which can yield unstable estimates (cf. Coyne, 2013), and results should be interpreted with caution. We then repeated the analysis a third time by modeling hedonic and eudaimonic dimensions as correlated outcomes, and obtained results that were highly similar to those reported in Tables 5–6 (available upon request), suggesting that accounting for the correlated nature of these constructs does not substantively alter their associations with early experiences and midlife social support. Future studies might consider an alternative latent structure of well-being that subsumes both hedonic and eudaimonic dimensions.
NAS men were selected for good mental and physical health at study entry in the 1960’s, possibly leading to range restriction on CES items and psychological outcomes. This would limit the generalizability of our findings to men who might have suffered greater childhood adversity or have worse well-being than our sample. Our sample was all male and mostly White, and results may not generalize to females and ethnic minorities. We also note that NAS men were born in the first half of the 20th century, and societal conventions regarding parental discipline and definitions of a “normative event” have changed over time. Although behaviors that reflect early care vs. maltreatment are relatively stable over time, this caveat should be noted as readers make comparisons to later-born cohorts. Finally, we did not assess change in well-being from mid- to later-life, leaving open the possibility that lower psychological well-being influenced perceptions of social support in midlife. Future studies may consider examining bidirectional change between the two constructs.
Strengths and Future Directions
In this study, the person-centered approach allowed us to conceptualize early experiences as a multi-faceted construct and examine their legacy from the lifespan developmental perspectives. Two benefits are worth noting: First, the unique combination of early psychosocial factors in each class and their association with subsequent well-being suggest that one may miss or underestimate individual differences by approaching early experiences in a unidimensional manner. Most studies that have considered the co-occurrence of childhood experiences focused exclusively on risk factors (e.g., Edwards et al., 2003; but see Chen, Miller, Lachman, Gruenewald, & Seeman, 2012). Our results, while exploratory, provide evidence for meaningful clustering across positive and negative early experiences. More research is needed to shed light on potential interactions between risk and protective factors that affect developmental outcomes. Second, in contrast to studies focusing on extreme stressors or clinical samples, our findings based on a relatively healthy sample suggest that focusing on subtler aspects of the early environment, such as parental discipline, and their interplay with other factors, can reveal enduring associations with developmental outcomes. Thus, our findings call for a more complex and nuanced approach to conceptualizing early experiences. Other methodological strengths of this study include the longitudinal assessment of midlife social support and later-life well-being, and adjustment for childhood SES in considering the developmental outcomes of early psychosocial experiences.
The present study contributes to a nascent area of research by demonstrating linkages between early environments and successful aging. The association between the early experiences of the cherished group, which was characterized by both loss and support, and facets of hedonic and eudaimonic well-being in later life, adds to our knowledge on stress-related growth across the lifespan. Future work can examine whether and how positive early experiences may confer benefits on other indicators of successful aging. Our finding of midlife qualitative support as a mediator of the association also bridges the developmental literature on the adulthood consequences of early attachment with life course models of early adversity. It suggests that interventions aimed at reducing loneliness and/or improving support perceptions (Masi, Chen, Hawkley, & Cacioppo, 2011) could have long-term benefits on well-being.
Supplementary Material
Acknowledgments
This study was supported by NIH grants R01-AG032037, R01-AG018436, a Senior Research Career Scientist Award from the Clinical Science Research and Development Service, U.S. Department of Veterans Affairs to Dr. Spiro, and a BU/VA Boston CTSI Pilot Grant to Dr. Lee. The VA Normative Aging Study is a research component of MAVERIC and is supported by the VA Cooperative Studies Program/Epidemiological Research Centers. The views expressed in this paper are those of the authors and do not necessarily represent the views of the U.S. Department of Veterans Affairs or other support institutions.
Footnotes
Modeling suppression (as done here by including both dimensions of social support as predictors of NA) can improve prediction substantially by removing criterion-irrelevant variance from the predictor of interest (Paulhus, Robins, Trzesniewski & Tracy, 2004; Tzelgov & Henik, 1991). When suppression is identified, it is important to interpret how the predictor and the suppressor combine to influence the outcome (Maassen & Bakker, 2001). We followed the procedure described by Cheung and Lau (2008) in evaluating suppression effects. We used the distribution of product method (Tofighi & Mackinnon, 2011) to evaluate whether the suppression effects (β21βY2.1) were statistically significant, i.e., a 95% confidence interval (CI) that does not include zero. The results suggest that adjusting for qualitative support boosted the association between quantitative support and NA (β21βY2.1 = −0.04, 95%CI = −0.08, −0.002). Adjusting for quantitative support also strengthened the association between qualitative support and NA, although this effect was marginally significant (β21βY2.1 = 0.03, 95%CI = 0.00, 0.06).
We repeated the latent class analysis (LCA) after removing 11 items we considered more evaluative in nature to determine whether the 3-class solution was largely driven by later-life perceptions of early experiences. The 11 items were: felt close to an adult relative; felt close to another adult; felt close to another sibling; had a special friend; physically punished more than deserved; unfairly treatment for long periods of time; considered intelligent by parents; considered athletic by parents; considered popular by parents; any accomplishments you were proud of; considered to be a hell-raiser as a kid. Using the same procedures as described in the Analysis Plan, we evaluated 1- to 5-class models (models involving 6+ classes were not considered based on analyses using the full set of items). We found that the 3-class model provided the best fit to the data and performed well on classification diagnostics. The profiles of the 3 classes were rather similar between the solutions based on the full set vs. reduced set of items -- when we compared the item endorsement probabilities for each class between the two solutions (e.g., endorsement probability for “serious fall” in Class 1 based on the full set vs. reduced set of items), all differences were below 11%. We also used Cohen’s kappa coefficient to compare each person’s most-likely class assignment based on the two 3-class solutions. The kappa was 0.54, suggesting a moderate level of agreement between the two solutions (Landis & Koch, 1977). We also repeated the analyses for Hypothesis 2 using the latent class probabilities from the solution based on the reduced set of items. Compared to the results based on the full set of items, the effect sizes were somewhat attenuated, but the overall pattern of results remained the same. The weaker effect sizes were likely due to the smaller amount of variance in early experiences captured by the reduced vs. the full set of items. Because the 3-class solution held up even in the absence of items with an evaluative component, and these items allowed us to depict a richer picture of the sample’s early experiences, we report results based on the full set of CES items.
Contributor Information
Lewina O. Lee, VA Boston Healthcare System and Boston University
Carolyn M. Aldwin, Oregon State University
Laura D. Kubzansky, Harvard School of Public Health
Edith Chen, Northwestern University.
Daniel K. Mroczek, Northwestern University
Joyce M. Wang, VA Boston Healthcare System and Boston University
Avron Spiro, III, VA Boston Healthcare System and Boston University.
References
- Abbott RA, Ploubidis GB, Huppert FA, Kuh D, Croudace TJ. An evaluation of the precision of measurement of Ryff’s Psychological Well-Being Scales in a population sample. Social Indicators Research. 2010;97:357–373. doi: 10.1007/s11205-009-9506-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Aldwin CM, Igarashi H. Successful, optimal, and resilient aging: A psychosocial perspective. In: Lichtenberg PA, Mast BT, editors. APA Handbook of Clinical Geropsychology. Washington, DC: American Psychological Association; 2015. pp. 331–359. [Google Scholar]
- Aldwin CM, Levenson MR, Cupertino AP, Spiro A., III Personality, childhood experiences, and drinking patterns in older men: Findings from the Normative Aging Study. Proceedings of the Fifth International Congress of Behavioral Medicine; 1998. p. 82. [Google Scholar]
- Aldwin CM, Levenson MR, Kelly LL. Lifespan developmental perspectives on stress-related growth. In: Park CL, Lechner S, Stanton A, Antoni M, editors. Positive Life Changes in the Context of Medical Illness. Washington, DC: APA Press; 2009. pp. 87–104. [Google Scholar]
- Amato PR, Booth A. Consequences of parental divorce and marital unhappiness for adult well-being. Social Forces. 1991;69:895–914. [Google Scholar]
- An JS, Cooney TM. Psychological well-being in mid to late life: The role of generativity development and parent–child relationships across the lifespan. International Journal of Behavioral Development. 2006;30:410–421. [Google Scholar]
- Andersson L, Stevens N. Associations between early experiences with parents and well-being in old age. Journal of Gerontology. 1993;48:P109–P116. doi: 10.1093/geronj/48.3.p109. [DOI] [PubMed] [Google Scholar]
- Banfield JD, Raferty AE. Model-based Gaussian and non-Gaussian clustering. Biometrics. 1993;49:803–821. [Google Scholar]
- Bentler PM. Comparative fit indexes in structural models. Psychological Bulletin. 1990;107:238–246. doi: 10.1037/0033-2909.107.2.238. [DOI] [PubMed] [Google Scholar]
- Bentler PM. On tests and indices for evaluating structural models. Personality and Individual Differences. 2007;42:825–829. [Google Scholar]
- Bossé R, Aldwin CM, Levenson MR, Spiro A, Mroczek DK. Change in social support after retirement: Longitudinal findings from the Normative Aging Study. Journal of Gerontology. 1993;48:P210–P217. doi: 10.1093/geronj/48.4.p210. [DOI] [PubMed] [Google Scholar]
- Bossé R, Aldwin CM, Levenson MR, Workman-Daniels K, Ekerdt DJ. Differences in social support among retirees and workers: findings from the Normative Aging Study. Psychology and Aging. 1990;5:41. doi: 10.1037//0882-7974.5.1.41. [DOI] [PubMed] [Google Scholar]
- Bowlby J. Attachment and loss: Vol 2. Separation: Anxiety and anger. New York, NY: Basic Books; 1973. [Google Scholar]
- Bowlby J. Attachment and loss: Vol. 1. Attachment. 2. New York, NY: Basic Books; 1982. [Google Scholar]
- Bowlby J. A secure base: Parent-child attachment and healthy human development. New York, NY: Basic Books; 1988. [Google Scholar]
- Bowling A, Gabriel Z, Dykes J, Dowding LM, Evans O, Fleissig A, Sutton S. Let’s ask them: a national survey of definitions of quality of life and its enhancement among people aged 65 and over. The International Journal of Aging and Human Development. 2003;56:269–306. doi: 10.2190/BF8G-5J8L-YTRF-6404. [DOI] [PubMed] [Google Scholar]
- Bozdogan H. Model selection and Akaike’s information criterion (AIC): The general theory and its analytic extensions. Psychometrika. 1987;52:345–370. [Google Scholar]
- Brewin CR, Andrews B, Gotlib IH. Psychopathology and early experience: a reappraisal of retrospective reports. Psychological Bulletin. 1993;113:82–98. doi: 10.1037/0033-2909.113.1.82. [DOI] [PubMed] [Google Scholar]
- Brissette I, Cohen S, Seeman TE. Measuring social integration and social networks. In: Cohen S, Underwood L, Gottlieb B, editors. Measuring and intervening in social support. New York, NY: Oxford University Press; 2000. pp. 53–85. [Google Scholar]
- Brown G, Craig T, Harris T, Handley R, Harvey A. Validity of retrospective measures of early maltreatment and depressive episodes using the Childhood Experiences of Care and Abuse (CECA) instrument- a life course study of adult chronic depression. Journal of Affective Disorders. 2007;103:217–224. doi: 10.1016/j.jad.2007.06.003. [DOI] [PubMed] [Google Scholar]
- Burns RA, Machin MA. Investigating the structural validity of Ryff’s psychological well-being scales across two samples. Social Indicators Research. 2009;93:359–375. [Google Scholar]
- Busseri MA, Sadava SW. A review of the tripartite structure of subjective well-being: Implications for conceptualization, operationalization, analysis, and synthesis. Personality and Social Psychology Review. 2010;15:290–314. doi: 10.1177/1088868310391271. [DOI] [PubMed] [Google Scholar]
- Cassidy J, Shaver PR. Handbook of attachment: Theory, research and clinical applications. 2. New York, NY: Guilford; 2008. [Google Scholar]
- Chen E, Miller GE, Lachman ME, Gruenewald TL, Seeman TE. Protective factors for adults from low-childhood socioeconomic circumstances: The benefits of shift-and-persist for allostatic load. Psychosomatic Medicine. 2012;74:178–186. doi: 10.1097/PSY.0b013e31824206fd. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cheung GW, Lau RS. Testing mediation and suppression effects of latent variables: Bootstrapping with structural equation models. Organizational Research Methods. 2008;11:296–325. [Google Scholar]
- Clark C, Caldwell T, Power C, Stansfeld SA. Does the influence of childhood adversity on psychopathology persist across the lifecourse? A 45-year prospective epidemiologic study. Annals of Epidemiology. 2010;20:385–394. doi: 10.1016/j.annepidem.2010.02.008. [DOI] [PubMed] [Google Scholar]
- Colman RA, Widom CS. Childhood abuse and neglect and adult intimate relationships: A prospective study. Child Abuse & Neglect. 2004;28:1133–1151. doi: 10.1016/j.chiabu.2004.02.005. [DOI] [PubMed] [Google Scholar]
- Conger AJ. A revised definition for suppressor variables: A guide to their identification and interpretation. Educational and Psychological Measurement. 1974;34:35–46. [Google Scholar]
- Coyne JC. Highly correlated hedonic and eudaimonic well-being thwart genomic analysis. Proceedings of the National Academy of Sciences. 2013;110:E4183. doi: 10.1073/pnas.1315212110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Davidson G, Devaney J, Spratt T. The impact of adversity in childhood on outcomes in adulthood: Research lessons and limitations. Journal of Social Work. 2010;10:369–390. [Google Scholar]
- Diener E. Subjective well-being. Psychological Bulletin. 1984;95:542–575. [PubMed] [Google Scholar]
- Diener E, Emmons RA, Larsen RJ, Griffin S. The Satisfaction with Life Scale. Journal of Personality Assessment. 1985;49:71–75. doi: 10.1207/s15327752jpa4901_13. [DOI] [PubMed] [Google Scholar]
- Diener E, Seligman ME. Very happy people. Psychological Science. 2002;13:81–84. doi: 10.1111/1467-9280.00415. [DOI] [PubMed] [Google Scholar]
- Diener E, Suh EM, Lucas RE, Smith HL. Subjective well-being: Three decades of progress. Psychological Bulletin. 1999;125:276–302. [Google Scholar]
- Dong M, Anda RF, Felitti VJ, Dube SR, Williamson DF, Thompson TJ, Giles WH. The interrelatedness of multiple forms of childhood abuse, neglect, and household dysfunction. Child Abuse & Neglect. 2004;28:771–784. doi: 10.1016/j.chiabu.2004.01.008. [DOI] [PubMed] [Google Scholar]
- Dunn VJ, Abbott RA, Croudace TJ, Wilkinson P, Jones PB, Herbert J, Goodyer IM. Profiles of family-focused adverse experiences through childhood and early adolescence: The ROOTS project a community investigation of adolescent mental health. BMC Psychiatry. 2011;11:109. doi: 10.1186/1471-244X-11-109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Edwards VJ, Holden GW, Felitti VJ, Anda RF. Relationship between multiple forms of childhood maltreatment and adult mental health in community respondents: Results from the Adverse Childhood Experiences study. American Journal of Psychiatry. 2003;160:1453–1460. doi: 10.1176/appi.ajp.160.8.1453. [DOI] [PubMed] [Google Scholar]
- Fraley RC, Roisman GI, Haltigan JD. The legacy of early experiences in development: Formalizing alternative models of how early experiences are carried forward over time. Developmental Psychology. 2013;49:109. doi: 10.1037/a0027852. [DOI] [PubMed] [Google Scholar]
- Franz CE, Panizzon MS, Eaves LJ, Thompson W, Lyons MJ, Jacobson KC, Kremen WS. Genetic and environmental multidimensionality of well-and ill-being in middle aged twin men. Behavior Genetics. 2012;42:579–591. doi: 10.1007/s10519-012-9538-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gayman MD, Turner RJ, Cislo AM, Eliassen AH. Early adolescent family experiences and perceived social support in young adulthood. Journal of Early Adolescence. 2011;31:880–908. [Google Scholar]
- Green JG, McLaughlin KA, Berglund PA, Gruber MJ, Sampson NA, Zaslavsky AM, Kessler RC. Childhood adversities and adult psychiatric disorders in the National Comorbidity Survey Replication I: Associations with first onset of DSM-IV disorders. Archives of General Psychiatry. 2010;67:113–123. doi: 10.1001/archgenpsychiatry.2009.186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hardt J, Rutter M. Validity of adult retrospective reports of adverse childhood experiences: Review of the evidence. Journal of Child Psychology and Psychiatry. 2004;45:260–273. doi: 10.1111/j.1469-7610.2004.00218.x. [DOI] [PubMed] [Google Scholar]
- Henderson LW, Knight T. Integrating the hedonic and eudaimonic perspectives to more comprehensively understand wellbeing and pathways to wellbeing. International Journal of Wellbeing. 2012;2:196–221. [Google Scholar]
- Huppert FA, Abbott RA, Ploubidis GB, Richards M, Kuh D. Parental practices predict psychological well-being in midlife: Life-course associations among women in the 1946 British birth cohort. Psychological Medicine. 2010;40:1507–1518. doi: 10.1017/S0033291709991978. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jessor R. Risk behavior in adolescence: A psychosocial framework for understanding and action. Journal of Adolescent Health. 1991;12:597–605. doi: 10.1016/1054-139x(91)90007-k. [DOI] [PubMed] [Google Scholar]
- Johnson JG, Cohen P, Gould MS, Kasen S, Brown J, Brook JS. Childhood adversities, interpersonal difficulties, and risk for suicide attempts during late adolescence and early adulthood. Archives of General Psychiatry. 2002;59:741–749. doi: 10.1001/archpsyc.59.8.741. [DOI] [PubMed] [Google Scholar]
- Kashdan TB, Biswas-Diener R, King LA. Reconsidering happiness: the costs of distinguishing between hedonics and eudaimonia. Journal of Positive Psychology. 2008;3:219–233. [Google Scholar]
- Kass RE, Wasserman L. A reference Bayesian test for nested hypotheses and its relationship to the Schwarz criterion. Journal of the American Statistical Association. 1995;90:928–934. [Google Scholar]
- Keyes CL. The mental health continuum: From languishing to flourishing in life. Journal of Health and Social Behavior. 2002;43:207–222. [PubMed] [Google Scholar]
- Krause N. Assessing change in social support during late life. Research on Aging. 1999;21:539–569. [Google Scholar]
- Kuh D, Ben-Shlomo Y, Lynch J, Hallqvist J, Power C. Life course epidemiology. Journal of Epidemiology and Community Health. 2003;57:778–783. doi: 10.1136/jech.57.10.778. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Liang J. Dimensions of the Life Satisfaction Index A: A structural formulation. Journal of Gerontology. 1984;39:613–622. doi: 10.1093/geronj/39.5.613. [DOI] [PubMed] [Google Scholar]
- Little TD, Cunningham WA, Shahar G, Widaman KF. To parcel or not to parcel: Exploring the question, weighing the merits. Structural Equation Modeling. 2002;9:151–173. [Google Scholar]
- Lo Y, Mendell N, Rubin D. Testing the number of components in a normal mixture. Biometrika. 2001;88:767–778. [Google Scholar]
- Luecken LJ, Roubinov DS, Tanaka R. Childhood family environment, social competence, and health across the lifespan. Journal of Social and Personal Relationships. 2013;30:171–178. [Google Scholar]
- Magnusson D. The logic and implication of a person approach. In: Cairns RB, Bergman LR, Kagan J, editors. Methods and models for studying the individual. Thousand Oaks, CA: Sage; 1998. pp. 33–63. [Google Scholar]
- Mallers MH, Charles ST, Neupert SD, Almeida DM. Perceptions of childhood relationships with mother and father: Daily emotional and stressor experiences in adulthood. Developmental Psychology. 2010;46:1651–1661. doi: 10.1037/a0021020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Martire LM, Schulz R, Mittelmark MB, Newsom JT. Stability and change in older adults’ social contact and social support: The Cardiovascular Health Study. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 1999;54:S302–S311. doi: 10.1093/geronb/54b.5.s302. [DOI] [PubMed] [Google Scholar]
- Masi CM, Chen HY, Hawkley LC, Cacioppo JT. A meta-analysis of interventions to reduce loneliness. Personality and Social Psychology Review. 2010;15:219–266. doi: 10.1177/1088868310377394. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Maassen GH, Bakker AB. Suppressor variables in path models: Definitions and interpretations. Sociological Methods & Research. 2001;30:241–270. [Google Scholar]
- Masyn K. Latent class analysis and finite mixture modeling. In: Little T, editor. The Oxford handbook of quantitative methods in psychology. Vol. 2. New York, NY: Oxford University Press; 2013. pp. 551–611. [Google Scholar]
- McLachlan G, Peel D. Finite mixture models. New York, NY: John Wiley & Sons; 2000. [Google Scholar]
- McLaughlin KA, Breslau J, Green JG, Lakoma MD, Sampson NA, Zaslavsky AM, Kessler RC. Childhood socio-economic status and the onset, persistence, and severity of DSM-IV mental disorders in a US national sample. Social Science & Medicine. 2011;73:1088–1096. doi: 10.1016/j.socscimed.2011.06.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mikulincer M, Shaver PR. An attachment and behavioral systems perspective on social support. Journal of Social and Personal Relationships. 2009;26:7–19. [Google Scholar]
- Miller GE, Chen E, Parker KJ. Psychological stress in childhood and susceptibility to the chronic diseases of aging: Moving toward a model of behavioral and biological mechanisms. Psychological Bulletin. 2011;137:959–997. doi: 10.1037/a0024768. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Morack J, Ram N, Fauth EB, Gerstorf D. Multidomain trajectories of psychological functioning in old age: A longitudinal perspective on (uneven) successful aging. Developmental Psychology. 2013;49:2309–2324. doi: 10.1037/a0032267. [DOI] [PubMed] [Google Scholar]
- Muthén LK, Muthén BO. Mplus user’s guide. 7. Los Angeles, CA: Muthén & Muthén; 1998–2012. [Google Scholar]
- Nagin DS. Group-based modeling of development. Cambridge, MA: Harvard University Press; 2005. [Google Scholar]
- Neugarten BL, Havighurst RJ, Tobin SS. The measurement of life satisfaction. Journal of Gerontology. 1961;16:134–143. doi: 10.1093/geronj/16.2.134. [DOI] [PubMed] [Google Scholar]
- Paulhus DL, Robins RW, Trzesniewski KH, Tracy JL. Two replicable suppressor situations in personality research. Multivariate Behavioral Research. 2004;39:303–328. doi: 10.1207/s15327906mbr3902_7. [DOI] [PubMed] [Google Scholar]
- Peters JL, Kubzansky LD, Ikeda A, Spiro A, Wright RO, Weisskopf MG, Schwartz J. Childhood and adult socioeconomic position, cumulative lead levels, and pessimism in later life: The VA Normative Aging Study. American Journal of Epidemiology. 2011;174:1345–1353. doi: 10.1093/aje/kwr269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Pinquart M, Sörensen S. Influences of socioeconomic status, social network, and competence on subjective well-being in later life: A meta-analysis. Psychology and Aging. 2000;15:187–224. doi: 10.1037//0882-7974.15.2.187. [DOI] [PubMed] [Google Scholar]
- Pitzer LM, Fingerman KL. Psychosocial resources and associations between childhood physical abuse and adult well-being. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 2010;65B:425–433. doi: 10.1093/geronb/gbq031. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Pruchno RA, Wilson-Genderson M, Rose M, Cartwright F. Successful aging: Early influences and contemporary characteristics. The Gerontologist. 2010;50:821–833. doi: 10.1093/geront/gnq041. [DOI] [PubMed] [Google Scholar]
- Repetti RL, Taylor SE, Seeman TE. Risky families: Family social environments and the mental and physical health of offspring. Psychological Bulletin. 2002;128:330–366. [PubMed] [Google Scholar]
- Roisman GI, Collins WA, Sroufe LA, Egeland B. Predictors of young adults’ representations of and behavior in their current romantic relationship: Prospective tests of the prototype hypothesis. Attachment & Human Development. 2005;7:105–121. doi: 10.1080/14616730500134928. [DOI] [PubMed] [Google Scholar]
- Rothrauff TC, Cooney TM, An JS. Remembered parenting styles and adjustment in middle and late adulthood. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 2009;64B:137–146. doi: 10.1093/geronb/gbn008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rowe JW, Kahn RL. Human aging: Usual and successful. Science. 1987;237:143–149. doi: 10.1126/science.3299702. [DOI] [PubMed] [Google Scholar]
- Rowe JW, Kahn RL. Successful Aging. New York: Pantheon; 1999. [Google Scholar]
- Rutter M. Annual research review: Resilience – clinical implications. Journal of Child Psychology and Psychiatry. 2013;54:474–487. doi: 10.1111/j.1469-7610.2012.02615.x. [DOI] [PubMed] [Google Scholar]
- Ryan RM, Deci EL. Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist. 2000;55:68–78. doi: 10.1037//0003-066x.55.1.68. [DOI] [PubMed] [Google Scholar]
- Ryan RM, Deci EL. On happiness and human potentials: A review of research on hedonic and eudaimonic well-being. Annual Review of Psychology. 2001;52:141–166. doi: 10.1146/annurev.psych.52.1.141. [DOI] [PubMed] [Google Scholar]
- Ryff CD. Happiness is everything, or is it? Explorations on the meaning of psychological well-being. Journal of Personality and Social Psychology. 1989;57:1069–81. [Google Scholar]
- Ryff CD, Keyes CLM. The structure of psychological well-being revisited. Journal of Personality and Social Psychology. 1995;69:719–727. doi: 10.1037//0022-3514.69.4.719. [DOI] [PubMed] [Google Scholar]
- Ryff CD, Love DG, Urry HL, Muller D, Rosenkranz MA, Friedman EM, Singer B. Psychological well-being and ill-being: Do they have distinct or mirrored biological correlates? Psychotherapy and Psychosomatics. 2006;75:85–95. doi: 10.1159/000090892. [DOI] [PubMed] [Google Scholar]
- Ryff CD, Singer B. The contours of positive human health. Psychological Inquiry. 1998;9:1–28. [Google Scholar]
- Sarason BR, Pierce GR, Sarason IG. Social support: The sense of acceptance and the role of relationships. In: Sarason BR, Sarason IG, Pierce GR, editors. Social support: An interactional view. New York, NY: John Wiley & Sons; 1990. pp. 9–25. [Google Scholar]
- Sarason BR, Pierce GR, Shearin EN, Sarason IG, Waltz JA, Poppe L. Perceived social support and working models of self and actual others. Journal of Personality and Social Psychology. 1991;60:273–287. [Google Scholar]
- Satorra A, Bentler P. A scaled difference chi-square test statistic for moment structure analysis. Psychometrika. 2001;66:507–514. doi: 10.1007/s11336-009-9135-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Schwarz G. Estimating the dimension of a model. Annals of Statistics. 1978;6:461–464. [Google Scholar]
- Scott K, McLaughlin K, Smith D, Ellis P. Childhood maltreatment and DSM-IV adult mental disorders: Comparison of prospective and retrospective findings. British Journal of Psychiatry. 2012;200:469–475. doi: 10.1192/bjp.bp.111.103267. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shaw BA, Krause N. Exposure to physical violence during childhood, aging, and health. Journal of Aging and Health. 2002;14:467–494. doi: 10.1177/089826402237179. [DOI] [PubMed] [Google Scholar]
- Shaw BA, Krause N, Chatters LM, Connell CM, Ingersoll-Dayton B. Emotional support from parents early in life, aging, and health. Psychology and Aging. 2004;19:4–12. doi: 10.1037/0882-7974.19.1.4. [DOI] [PubMed] [Google Scholar]
- Spiro A, III, Bossé R. The Normative Aging Study. In: Maddox GL, editor. The encyclopedia of aging. 3. New York, NY: Springer; 2001. pp. 744–746. [Google Scholar]
- Sroufe LA. Attachment and development: A prospective, longitudinal study from birth to adulthood. Attachment & Human Development. 2005;7:349–367. doi: 10.1080/14616730500365928. [DOI] [PubMed] [Google Scholar]
- Steiger JH. Structural model evaluation and modification: An interval estimation. Multivariate Behavioral Research. 1990;25:173–180. doi: 10.1207/s15327906mbr2502_4. [DOI] [PubMed] [Google Scholar]
- Tofighi D, MacKinnon DP. RMediation: An R package for mediation analysis confidence intervals. Behavior Research Methods. 2011;43:692–700. doi: 10.3758/s13428-011-0076-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Toyokawa N. Unpublished doctoral dissertation. Oregon State University; Corvallis, OR: 2012. Trajectories of social support in later life: A longitudinal comparison of socioemotional selectivity theory with dynamic integration theory. [Google Scholar]
- Trickett PK, Kim K, Prindle J. Variations in emotional abuse experiences among multiply maltreated young adolescents and relations with developmental outcomes. Child Abuse & Neglect. 2011;35:876–886. doi: 10.1016/j.chiabu.2011.08.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tucker LR, Lewis C. A reliability coefficient for maximum likelihood factor analysis. Psychometrika. 1973;38:1–10. [Google Scholar]
- Tzelgov J, Henik A. Suppression situations in psychological research: Definitions, implications, and applications. Psychological Bulletin. 1991;109:524–536. [Google Scholar]
- Umberson D, Williams K, Powers DA, Liu H, Needham B. Stress in childhood and adulthood: Effects on marital quality over time. Journal of Marriage and Family. 2005;67:1332–1347. doi: 10.1111/j.1741-3737.2005.00220.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Urry HL, Nitschke JB, Dolski I, Jackson DC, Dalton KM, Mueller CJ, Davidson RJ. Making a life worth living: Neural correlates of well-being. Psychological Science. 2004;15:367–372. doi: 10.1111/j.0956-7976.2004.00686.x. [DOI] [PubMed] [Google Scholar]
- Van Tilburg T. Losing and gaining in old age: Changes in personal network size and social support in a four-year longitudinal study. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences. 1998;53:S313–S323. doi: 10.1093/geronb/53b.6.s313. [DOI] [PubMed] [Google Scholar]
- Waterman AS. On the importance of distinguishing hedonia and eudaimonia when contemplating the hedonic treadmill. American Psychologist. 2007;62:612–3. doi: 10.1037/0003-066X62.6.612. [DOI] [PubMed] [Google Scholar]
- Waterman AS. Reconsidering happiness: A eudaimonist’s perspective. Journal of Positive Psychology. 2008;3:234–252. [Google Scholar]
- Watson D, Clark LA, Tellegen A. Development and validation of brief measures of Positive and Negative Affect: The PANAS Scales. Journal of Personality and Social Psychology. 1988;54:1063–1070. doi: 10.1037//0022-3514.54.6.1063. [DOI] [PubMed] [Google Scholar]
- Werner EE, Smith RS. Overcoming the odds: High risk children from birth to adulthood. Ithaca, NY: Cornell University Press; 1992. [Google Scholar]
- Whisman MA. Childhood trauma and marital outcomes in adulthood. Personal Relationships. 2006;13:375–386. [Google Scholar]
- Wills TA, Shinar O. Measuring perceived and received social support. In: Cohen S, Underwood L, Gottlieb B, editors. Measuring and intervening in social support. New York, NY: Oxford University Press; 2000. pp. 86–135. [Google Scholar]
- Wilson RS, Krueger KR, Arnold SE, Barnes LL, de Leon CFM, Bienias JL, Bennett DA. Childhood adversity and psychosocial adjustment in old age. American Journal of Geriatric Psychiatry. 2006;14:307–315. doi: 10.1097/01.JGP.0000196637.95869.d9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Yancura L, Aldwin CM. Stability and change in retrospective reports of childhood events over a five-year period: Findings from the Davis Longitudinal Study. Psychology and Aging. 2009;24:715–721. doi: 10.1037/a0016203. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.