Petrova, 2024

Psychology and Aging
Emotional Support Across Adulthood: A 60-Year Study of Men’s Social

Networks
Kate Petrova, Michael D. Nevarez, Robert J. Waldinger, and Marc S. Schulz
Online First Publication, August 15, 2024. https://dx.doi.org/10.1037/pag0000843
CITATION
Petrova, K., Nevarez, M. D., Waldinger, R. J., & Schulz, M. S. (2024). Emotional support across adulthood: A
60-year study of men’s social networks.. Psychology and Aging. Advance online publication. https://
dx.doi.org/10.1037/pag0000843
Psychology and Aging
© 2024 American Psychological Association
ISSN: 0882-7974 https://doi.org/10.1037/pag0000843
Emotional Support Across Adulthood: A 60-Year Study of

Men’s Social Networks
Kate Petrova1, Michael D. Nevarez2, Robert J. Waldinger2, and Marc S. Schulz3
1
Department of Psychology, Stanford University
2
Department of Psychiatry, Harvard Medical School
3
Department of Psychiatry, Bryn Mawr College
Past research shows that social networks get smaller with age. But not all types of relationships may shrink at
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
the same rate or for similar reasons. In the present study, we used a unique data set from a sample of 235 men
This document is copyrighted by the American Psychological Association or one of its allied publishers.
who were followed longitudinally for 71 years to examine how the general pattern of network shrinkage
documented in previous research generalizes to the number of emotional support providers in people’s
networks. We additionally examined early-life predictors of the size of later-life support network. Growth
curve analyses revealed that, mirroring the more general pattern of network shrinkage, emotional support
networks shrink by as much as 50% between the ages of 30 and 90, reflecting an average reduction from two to
one support providers. Examining the associations between prospectively collected measures of childhood
family environment and later-life emotional support, we also found that men who grew up in warmer family
environments had larger support networks in adulthood. In contrast, childhood family socioeconomic status
was not connected to the size of emotional support networks later in life. The generalizability of this work is
limited by the use of an archival all-male sample from the United States. Despite this limitation, these findings
make important contributions to our understanding of adult socioemotional development and underscore
the importance of prospectively collected longitudinal data in developmental research. Additional research
is needed to examine the consequences of changing emotional support across the lifespan for health
and well-being.
Public Significance Statement

This study, utilizing a unique longitudinal data set spanning 60 years, reveals a consistent decline in the
size of men’s emotional support networks between the ages of 30 and 90. This finding contributes new
knowledge about social and emotional aspects of aging and highlights the importance of proactively
identifying and addressing emotional support needs in older adults.
Keywords: social support, social networks, relationships, emotion regulation, lifespan development
Supplemental materials: https://doi.org/10.1037/pag0000843.supp
Emotional support, which refers to the provision of comfort, increasingly difficult as people grow older. Research consistently
empathy, and understanding, is a critical resource for navigating shows that individuals’ social networks tend to get smaller with age
life’s challenges (e.g., Bolger & Eckenrode, 1991; Helgeson, 1993; (e.g., Bruine de Bruin et al., 2020; Wrzus et al., 2013). Over the years,
Holtzman et al., 2017; Marroquín, 2011). Later in life, emotional this pattern has been shown to hold true for both men and women
support has been shown to be an important predictor of cognitive (Carstensen, 1992; Schwartz & Litwin, 2018) and to generalize
aging (Seeman et al., 2001), perceived health (Krause, 1987), across diverse cultural contexts (Fung et al., 2008). However, a more
emotional well-being (Patrick et al., 2001), and life expectancy (Latifi granular look shows that different types of relationships may follow
et al., 2019). Yet, having access to supportive others may become different trajectories, depending, in part, on what functions they serve
investigation, methodology, visualization, writing–original draft, and writing–

Ute Kunzmann served as action editor. review and editing. Michael D. Nevarez played a lead role in data curation
Kate Petrova https://orcid.org/0000-0003-1671-9768 and a supporting role in conceptualization and writing–review and editing.
Marc S. Schulz https://orcid.org/0000-0002-5179-2584 Robert J. Waldinger played a lead role in funding acquisition and project
Analyses reported in the present study were preregistered. The administration and a supporting role in writing–review and editing. Marc S.
preregistration and accompanying analysis code can be accessed online Schulz played a lead role in project administration and supervision and a
at the Open Science Framework at https://osf.io/hkc68/. No part of this supporting role in conceptualization, formal analysis, methodology, writing–
article has been published, posted, or presented elsewhere. The authors original draft, and writing–review and editing.
have no conflicts of interest to declare. This research was supported by the Correspondence concerning this article should be addressed to Kate
National Institute on Aging Grant R01AG045230 awarded to Robert J. Petrova, Department of Psychology, Stanford University, 450 Jane
Waldinger. Stanford Way, Building 420, Stanford, CA 94305, United States. Email:
Kate Petrova played a lead role in conceptualization, formal analysis, kpetrova@stanford.edu
1
2 PETROVA, NEVAREZ, WALDINGER, AND SCHULZ
in people’s lives (e.g., Antonucci et al., 2014; English & Carstensen, longitudinal studies suggests that psychological distress, negative
2014). Whether or not developmental trajectories of emotional support affect, and neuroticism decrease from young adulthood to middle
networks (i.e., parts of the broader social networks that consist of adulthood while remaining relatively more stable in older adulthood
individuals who routinely provide adults with emotional support) (Charles et al., 2023; Steunenberg et al., 2005). Furthermore,
follow the more general pattern of network shrinkage remains an open developmental research in the domain of personality shows that traits
question. In this study, we use a unique longitudinal data set to examine like conscientiousness and agreeableness tend to increase from
how the number of emotional support providers in 235 men’s social younger adulthood to older adulthood (Roberts & Mroczek, 2008).
networks changes across the third through ninth decades of life, and we Taken together, these changes in well-being and personality could
examine early-life predictors of support network size. render older adults to be better equipped to cope with emotional
challenges independently or could enable them to make more
effective use of a limited number of support providers (Antonucci &
Developmental Trajectories of Social Networks Akiyama, 1987a). If so, one might expect the number of active
Across the Lifespan and Implications for support providers to decrease with age.1
Emotional Support A critical missing piece of information that would help

more directly evaluate these competing perspectives on socio-

Multiple perspectives have been proposed to explain why
emotional aging is how the number of emotional support providers in
individuals’ social networks become smaller with age. These per-
individuals’ networks changes within individuals across adulthood.
spectives give rise to competing hypotheses about emotional support
Prospectively, collected longitudinal data that track networks across
across adulthood. For example, according to the socioemotional
developmental stages are essential for painting a fuller picture of
selectivity theory, the shrinkage of social networks is driven, in part,
emotional support across adulthood, but such data are rare. Relevant
by older adults intentionally pruning their networks of loose ties (i.e.,
existing longitudinal studies have followed participants for periods
acquaintances, casual connections) as they prioritize more intimate
ranging from less than 1 year to 17 years (Wrzus et al., 2013). To our
relationships (Carstensen, 2006; English & Carstensen, 2014; Lang &
knowledge, no other studies have tracked emotional support networks
Carstensen, 1994). Selection of social partners can thus be thought
of the same individuals across the entirety of the adult lifespan.
of as a form of antecedent-focused emotion regulation: Individuals
proactively shape their social networks in order to accomplish
their emotional goals (Carstensen et al., 2003; Gross, 2015). For Developmental Trajectories of Emotional Support
example, people may strategically “trim” social partners who tend to Networks: The Roles of Life Transitions
elicit negative emotions while preserving close and suppor-
tive relationships (Lang, 2000). Consistent with this perspective, Another key advantage of prospectively collected longitudinal
longitudinal research shows that, even as the number of peripheral data is that it allows for delineation of the unique effects of aging
social partners declines, the number of intimate connections in older on emotional support networks from the contextual effects of life
adults’ social circles remains relatively stable across time (e.g., transitions such as marriage or retirement. With regard to marriage, a
English & Carstensen, 2014). This finding aligns with the influential useful perspective is offered by the influential social provisions
social convoy theory (see Antonucci et al., 2014) and points to the framework, which outlines six “provisions”—such as guidance (i.e.,
possibility that the number of emotional support providers—who advice), reassurance of worth, and nurturance—that are crucial for
are typically found inside one’s intimate inner circle (Antonucci & individuals to feel supported in their relationships (Cutrona &
Akiyama, 1987a)—may similarly remain stable across adulthood. Russell, 1987). Importantly, according to this framework, each
Indeed, seminal cross-sectional studies of social support demonstrate provision is typically associated with a single provider (though a
that older adults across different age groups receive comparable single provider can supply multiple provisions; Cutrona & Russell,
amounts of support (Antonucci & Akiyama, 1987a), but evidence 1987). Because romantic partnerships are typically one of the main
from longitudinal studies is lacking. attachment relationships in adults’ lives, this framework suggests
An alternative perspective is offered by the stress and vulnerability that marriage could lead to a reduction in the number of emotional
integration framework, which suggests that, even though older age is support providers. Consistent with this view, evidence shows that
often associated with physiological and cognitive vulnerabilities married American men tend to consolidate their social networks
that may make regulation of acute distress more challenging, this upon getting married, receiving emotional support primarily from
vulnerability is counteracted by improvements in older adults’ use of their spouse (Gurung et al., 2003). On the other hand, a significant
proactive, antecedent-focused regulatory strategies (e.g., situation body of research also consistently shows that married individuals
selection) that help limit exposure to distressing stimuli in the first tend to have larger social networks, on average, compared to the
place (Charles, 2010; Charles & Luong, 2013). Importantly, though social networks of single individuals (see Wrzus et al., 2013).
there is not yet a robust evidentiary base for age-related improve- Retirement from work typically occasions changes in social
ments in emotion regulation, evidence from cross-sectional studies connections, but its impact on emotional support is not fully clear.
consistently shows that older adults report high levels of subjective Social convoy theory (Antonucci & Akiyama, 1987a), which
well-being—so high, in fact, that they often exceed those reported by
1
younger adults (see Isaacowitz, 2022). This leads us to consider an Importantly, the number of active emotional support providers can be
alternative hypothesis that age-related increases in subjective well- distinguished from the wider network of potential support providers.
Potential support providers are the individuals in the inner circle of one’s
being may reduce older adults’ need for emotional support, leading to network to whom one could turn to in times of need. In contrast, active
a decrease in the number of emotional support providers across support providers are the individuals to whom one does routinely turn to for
adulthood. Consistent with this hypothesis, evidence from sustenance and comfort, which is the focus of the present work.
EMOTIONAL SUPPORT ACROSS ADULTHOOD 3
emphasizes the stability of close connections across the lifespan, would (e.g., Eisenberg et al., 1996; Gaylord-Harden et al., 2010; Smith et al.,
suggest that retirement should not significantly alter the number of 2006; Stone et al., 2017), and there is evidence for long-term effects
emotional support providers in people’s networks, since support of early caregiving environments on later-life close relationships (e.g.,
providers are typically members of the inner circle. Consistent with this Doyle & Cicchetti, 2017; Waldinger & Schulz, 2016). Drawing on
hypothesis, past research shows that the transition to retirement is these perspectives, one might hypothesize that individuals who grew
associated with a reduction in the size of people’s outer—but not up in warmer childhood family environments are likely to have larger
inner—circles (Kauppi et al., 2021). On the other hand, evidence also emotional support networks in adulthood.
shows that retired men tend to report receiving less emotional support, Another factor that may have long-lasting implications for
on average, compared to their employed counterparts (Bossé et al., social support is childhood socioeconomic status (SES). There is a
1993). rich literature on the links between SES and parental valuation of
These seeming discrepancies between theoretical perspectives and child independence (e.g., Hamamura et al., 2013; Lareau, 2011; Park
empirical findings, as well as across findings from different studies, & Lau, 2016). Higher SES parents encourage their children to be self-
may be partially explained by the fact that these transitions could have reliant and to accomplish tasks independently (Park & Lau, 2016). In
variable short-term effects and long-term dynamics. For example, contrast, lower SES parents raise their children to view themselves
though the transition to married life may be associated with an as social agents embedded in the larger familial communities (Park &
initial reduction in the number of emotional support providers, this Lau, 2016). There is some evidence that SES-driven differences
reduction could be later followed by building of a new and expanded in values persist into adulthood and influence individuals’ social
support network (e.g., by developing closer and more intimate bonds functioning (e.g., Stephens et al., 2012), but the evidence for the
with each other’s networks; Chong et al., 2017; Kearns & Leonard, associations between SES and support seeking has been mixed
2004; Suanet & Huxhold, 2020). Similarly, being retired for a long (Calarco, 2011; Hooker et al., 2020; Stephens et al., 2012; Van
period of time may present opportunities for forming new meaning- Groenou & Van Tilburg, 2003). Prospectively examining the joint
ful and supportive relationships (e.g., through joining volunteer roles of parental warmth and childhood SES in shaping later-life
organizations; Carr et al., 2018). Thus, a useful distinction can be support networks will shed light on the mechanisms underlying
drawn between within-individual changes in emotional support that individual differences in emotional support needs across the lifespan.
take place shortly after the transition to marriage or retirement and
between-individual differences in emotional support that emerge The Present Study
following a prolonged period of being married or retired. Taking full
advantage of large-scale longitudinal data to make this distinction The present study consists of three parts. The first part focuses on
would allow us to empirically test these hypotheses. (Question 1 [Q1]) characterizing the normative developmental
changes in the number of emotional support providers across the
adult lifespan. The second part builds on insights from the first part
Early Predictors of Individual Differences in to explore how two important life transitions, (Question 2a [Q2a])
Emotional Support Networks marriage and (Question 2b [Q2b]) retirement, contribute to
Though there may be general patterns in how the aging process changing numbers of emotional support providers in individuals’
typically unfolds, no two people’s journeys are the same. To social networks. Finally, in the third part of the study (Question 3
understand why some individuals have larger support networks [Q3]), we examine the sources of interindividual variation in the
compared to others, it is important to understand the antecedents of average size of support networks in adulthood. Specifically, we
support seeking earlier in the lifespan. focus on prospectively collected measures of parental warmth and
Learning to navigate the social world is one of the earliest childhood SES as predictors of the average number of support
developmental challenges that people face (Erikson, 1950). Much of providers in social networks later in life. Controlling for childhood
this early learning occurs through children’s interactions with primary SES, we hypothesize (Hypothesis 1) that individuals who grew up in
caregivers. The tripartite model of familial influences on children’s warmer family environments have more support providers across
emotion regulation (Morris et al., 2007) suggests that parents shape the lifespan. Additional exploratory analyses focused on examining
how children regulate their emotions—including by turning to others the associations between emotional support and life expectancy.
for support—through a combination of explicit coaching and more
subtle socialization, which is shaped by the overall emotional climate Method
of the family. According to the attachment theory, warm and secure
Participants
family environments foster the development of more secure attach-
ment styles (Bowlby, 1969; Levitt, 1991). Children with secure Data collected from participants in an 85-year long (as of 2023)
attachment styles develop a sense of trust and confidence in their longitudinal study of adult development were used in the present
caregivers, fostering a foundation of emotional security that makes study (Vaillant, 1995). The total sample consists of 268 male
them more comfortable in seeking emotional support from others participants recruited from Harvard University when they were
throughout their lives (Bowlby, 1969; Levitt, 1991). Consistent with students between 1939 and 1942 (Mage at intake = 19). A subset of
these theoretical perspectives, greater attachment security in adults 235 participants who had at least one emotional support data point
has been linked to better ability to form and maintain supportive (median number of completed emotional support questionnaires =
relationships and to engage in more effective support seeking across 4; range = 1–7) were included in the present study. Reflecting
the lifespan (Collins & Feeney, 2000; Levitt, 1991; McLeod et al., the demographics of Harvard University students in the late 1930s
2020). Extensive research similarly connects warm and responsive and early 1940s, 98.8% of participants were White. All but one
parenting to effective support seeking in children and adolescents participant were born in the United States, and 12% were children
of immigrants. Participants completed questionnaires and inter- items for assessing within-person change. To ensure that these items
views at regular intervals (on average every 2 years) from the start reliably captured within-person changes in the number of emotional
of the study through the end of their life. support providers, we estimated a structural equation model in
which measures from the five waves described above loaded onto a
single latent factor of emotional support, with single-lag auto-
Procedure regressive effects between the five waves (i.e., 1967 regressed on
Starting in 1940, participants filled out questionnaires that were 1951; 1972 regressed on 1967; see Kenny & Zautra, 1995, for a
mailed to them every 2 years (on average). The questionnaires detailed overview of this approach). The resulting model showed
included a regular focus on participants’ physical and mental health, excellent fit to the data, χ2(df ) = 1.34 (1), p = .25; standardized root-
subjective well-being, social relationships, and daily activities. mean-square residual = .029; root-mean-square error of approxi-
Participant attrition rate from the start of the study is estimated at 16%. mation = .038, 90% confidence interval (CI) [.00, .183];
Participants began to receive monetary compensation for participat- comparative fit index = .989, with individual standardized factor
ing in the study in 2003. The amount of compensation varied from loadings ranging from .411 (combined fifth through seventh wave of
year to year, depending on the length of the questionnaire(s) and the data collection, 2007–2010) to .68 (third wave of data collection,
duration of other study procedures (interviews, lab visits, etc.). 1972). Taken together, these findings indicate that the items used
The present study covers a total period of 71 years starting from across all waves of data collection offer adequate reliability for
when early family environment data were collected (1939–1942) and detecting within-person change in emotional support.
until the latest emotional support assessment in 2010. Information
about participants’ social support networks as well as marriage and
retirement dates were self-reported by participants as part of these Marital Status
regular assessments. Within this 71-year period, emotional support Dates of marriage, divorce, and widowhood were collected for all
questionnaires were administered over the span of 60 years, starting in participants across all marriages. The majority of participants in the
1951 and continuing through 2010. The timeline of relevant waves of sample (93%; N = 249) were married at least once. The median
data collection is presented in Figure 1. number of marriages was 1, though it ranged from 0 (N = 9) to 5
(N = 1). No marriage data were available for eight participants. Based
Measures on the dates of marriage and divorce or widowhood, marital status
was coded as either 0 = not married or 1 = married for each year
Emotional Support Networks when emotional support network size was measured (1951, 1967,
1972, 1985, 2007, 2009, and 2010). Because a single aggregated
The number of support providers in participants’ networks was
measure of support network size was used to cover years 2007, 2009,
measured across seven assessments by the following open-ended
and 2010, marital status for years 2007–2010 was coded as 1 if the
questions: (1) “With whom do you usually talk over personal
participant was married for at least 1.5 years between 2007 and 2010
problems?” (Year 1951, Mage = 31; N = 190; Year 1967, Mage = 47;
and 0 if the participant was married for less than 1.5 years during that
N = 173; Year 1972, Mage = 52; N = 144), (2) “In times of crisis, to
3-year period.
whom (or to what) do you turn for sustenance and/or comfort?” (Year
1985, Mage = 65; N = 104), and (3) “Who do you turn to for emotional
support?” (Year 2007, Mage = 87; N = 58; Year 2009, Mage = 89; N =
Retirement Status
47; Year 2010, Mage = 90; N = 27). Participants responded to the
questions by listing individual emotional support providers either Information about retirement status was available for 212 out of
by relationship category (e.g., my wife or my daughter) or by name the 268 participants (79%). Every questionnaire that was mailed to
(e.g., John Doe). We operationalized the number of support providers participants over the course of the study included questions about
as the total number of people participants listed in response to each employment status and the average number of working hours per
question. Because it was not always possible to infer the nature of week. A total of 184 participants reported their exact retirement date
the relationship between the participant and the support provider(s) and 28 reported never retiring. The average retirement age for the
they listed by name, no formal analyses examining the sources sample was 66.4 years (SD = 7.5). Using this information, we coded
of emotional support were performed as part of this study. Some retirement status as either 0 = not retired or 1 = retired for each year
responses (16% of the total number of responses across all when support network size was measured (1951, 1967, 1972, 1985,
participants, across all seven waves of data collection) included 2007, 2009, and 2010). For the 3-year period that used the aggregated
references to broad categories, such as “friends” or “family,” or to measure of network size (2007–2010), retirement status was coded as
nonhuman sources of support, such as “nature” or “my cat.” These 0 if the participant was retired at least 1.5 years between 2007 and
responses were coded as missing. Responses like “nobody” or 2010 and 1 if the participant was retired for less than 1.5 years during
“myself” were coded as 0. Because of the short time intervals the 3-year period.
separating the three occasions on which the third question (“Who do
you turn to for emotional support?”) was administered and because 2
Rather than only analyzing data from 2007, we made this decision
of attrition associated with participants’ deaths, we aggregated because some participants had data in 2009 and/or 2010 but not 2007, so
participants’ responses from Years 2007, 2009, and 2010 into a single combining the three waves into a single data point in this way allowed us to
maximize the available sample size for the last assessment. Another
variable by calculating the mean across the three time points.2 advantage of this approach is that it helps avoid the potential issue of having
Different wordings of the questions that were used at different high covariation of Level 1 residuals of repeated measures taken in relatively
waves of data collection raise concerns about the reliability of these close time proximity (see Goldstein et al., 1994).
Figure 1
Timeline of Relevant Data Collection
Note. Average age of participants at each wave of data collection is shown above the arrow.
Year(s) when each wave of data collection took place are shown below the arrow. Time points at
which support network measures and information about participants’ marital and retirement
status were collected are shown as large (pink) dots on the timeline. Time points at which other
relevant variables were measured are shown as smaller (yellow) points on the timeline. SES =
socioeconomic status. See the online article for the color version of this figure.
Childhood Family Environment independence persuaded us to not aggregate the two measures into a
single average parental education score as originally planned but to
Parental Warmth. Three variables rated at the time of the use the two measures as independent indicators of parental SES
intake interviews that took place between 1939 and 1942 were used instead. Confirmatory factor analysis further revealed that mothers’
to measure overall warmth of childhood family environment: (a) education (standardized factor loading = .30) did not load on the
warmth of each participant’s relationship with his mother rated on a same latent factor as household income and fathers’ education
scale with values ranging from 1 = distant, hostile, or overly punitive (standardized factor loadings = .89 and .45, respectively). Based on
to 5 = nurturing, encouraging positive autonomy, fostered self- established guidelines (Comrey & Lee, 1992), we made the decision
esteem; (b) warmth of each participant’s relationship with his father to remove mothers’ education from the measurement model of
(same scale as Q1); and (c) overall warmth and cohesiveness of the childhood SES and instead include it as an independent predictor of
home environment, which was rated on a scale of 1 = lack of warmth social support.
and cohesion to 5 = warm and cohesive. Each of the above facets
was rated by two independent coders at the time of data collection.
Their ratings were based on intake interviews with participants and Time to Death
their parents. The ratings had high interrater reliability (average A time-varying time-till-death variable (in years) was calculated
Cronbach’s α = .78). Following the preregistered analysis plan, by subtracting the year of each questionnaire from each participant’s
measures of parental warmth and overall cohesiveness of family year of death. The average age of death in this sample was 78.21
environment were combined into a single measure using confirmatory (SD = 15.69), ranging from 22 to 97 at the time of data analyses.
factor analysis. Standardized factor loadings for all three indicators
were high: warmth of relationship with mother = .92, warmth of
relationship with father = .60, overall cohesiveness = .65. Transparency and Openness
Childhood SES. Two variables collected during intake inter- Analyses and some of the hypotheses reported in the present study
views were used to measure childhood SES: (a) household income were preregistered. The preregistration can be accessed online at the
and (b) parental education. Across the entire sample, annual house- Open Science Framework at https://osf.io/hkc68/. All deviations from
hold income of participants’ parents in the late 1930s ranged from the preregistration are noted in the body of this article. All additional
$500 to $2,000,000 (Mdn = $7,500). The data were normalized using exploratory analyses are labeled as such. Additional preregistered
a standard log transformation, which reduced skew from 14.93 to analyses of the association between the number of emotional support
.67 (SE = .15). Following the preregistered analysis plan, two providers and indicators of physical health and social well-being in
transformed income values that were more than 3 SDs from the mean old age, as well as additional exploratory analyses focusing on late-
were winsorized to equal the most extreme observation within life emotion regulation are reported in the Supplemental Materials.
3 SDs from the mean. The resulting (transformed) values ranged Analyses were carried out in Mplus (Version 8.3, Muthén & Muthén,
from 2.70 to 5.39 with M = 3.88 and SD = 0.48. Parental income 1998–2017). Mplus code that was used to carry out the main analyses
information was missing for nine participants. is available online at https://osf.io/hkc68/ (Petrova & Schulz, 2023).
Education at study intake was reported separately for fathers and No a priori sample size determination procedures were used as all
mothers (Nfather = 251; Nmother = 247) on an ordinal scale with the analyses were based on archival data. We provide all relevant
following anchors: 1 = no formal education, 2 = grade school, 3 = materials and describe all data exclusions and transformations.
high school, 4 = vocational training, 5 = college degree, and 6 = Descriptive statistics and covariances required for computational
graduate degree. Unexpectedly, reports of fathers’ and mothers’ reproducibility of Level 2 (between-person) results are available in
education shared only 19.4% of their variance (r = .44). This relative Table 1. This research was approved by the institutional review
Table 1
Descriptive Statistics and Bivariate Correlations for Key Study Variables Measured at the Between-Participant Level
Variable N M (SD) Range 1 2 3 4 5 6 7 8
1. Support network 235 1.26 (0.89) 0.00–5.00 —

2. Parental income 258 3.88 (0.48) 2.70–5.39 .19 —
3. Fathers’ education 251 5.00 (1.26) 1.00–6.00 .10 .40** —
4. Mothers’ education 247 4.00 (1.07) 2.00–6.00 .13 .06 .45** —
5. Home atmosphere 255 2.68 (1.33) 1.00–5.00 .24* .14* .10 .05 —
6. Mothers’ warmth 255 2.85 (1.34) 1.00–5.00 −.02 .09 .06 .03 .55** —
7. Fathers’ warmth 255 2.85 (1.43) 1.00–5.00 .09 .15* .08 .10 .60** .38** —
8. Regulatory repertoire 66 0.64 (0.40) 0.00–6.00 −.27 .003 −.04 .03 .12 .11 .06 —
Note. Significant correlations are highlighted in bold. Parental income (annual) was normalized using a standard log transformation (raw range = $500–
$2,000,000; Mdnraw = $7,500). Median values for fathers’ education (5 = college degree) and mothers’ education (4 = vocational training) are reported. Scores
for home atmosphere and fathers’ and mothers’ warmth reflect the average ratings of two independent coders (α = .78). Regulatory repertoire was calculated as
the average number of regulatory strategies per emotional challenge used by participants across an 8-day study period. N = number of participants.
* p < .05. ** p < .01.
boards of Bryn Mawr College (Study of Adult Development, R15- Q1: Changes in the Size of Emotional Support
002) and Massachusetts General Hospital (Adult Socioemotional Networks Across Adulthood
Learning, No. 2014P001116/MGH). Due to the unique nature of this
long-standing longitudinal study, which makes individual partici- To capture linear changes across time in the size of emotional
pants highly vulnerable to reidentification and to preserve the support networks, we estimated a two-level random-intercepts-and-
anonymity of the children and grandchildren of these participants slopes linear growth curve model (Preacher et al., 2008) in which
who continue to be actively studied as part of the Harvard Second individual reports of network size (Level 1) were nested within
Generation Study (2015–2022), we cannot make raw data publicly participants (Level 2). As shown in Figure 2, there was a linear
available. decline in the size of support networks across adulthood, B = −.011,
SE = .002, β = −.23, p < .001.3 This rate of linear decline was
relatively similar across participants: Analyses exploring the
random effects of age on the size of support networks revealed
Results
that there was no significant variation across participants in the rate
Preliminary Analyses of linear decline, τ = .0001, 95% CI [−.002, .001], Wald’s W2(df ) =
Support network information was available for 235 participants 2.01(1), p = .156 (Wald’s test of parameter constraint τ = 0).4 The
(median number of completed emotional support questionnaires = 4). degree of similarity across participants in the shapes of these
Preliminary analyses revealed that the number of missing data trajectories can be seen in Figure 2. Therefore, the slopes of support
points for each participant was not associated with any variables network size across age were constrained to be equal across
of interest. All analyses were carried out as multilevel models in participants in all subsequent analyses. Preregistered hypotheses
Mplus (Version 8.3, Muthén & Muthén, 1998–2017) using full- focusing on exploring the sources and implications of variation in
information maximum likelihood estimation to account for missing linear slopes were not tested.
data. Multilevel models can accommodate unequally spaced We also conducted additional exploratory analyses, in which we
time intervals by treating time (in our case, age) as a continuous used the time-varying time-till-death variable instead of participants’
variable, making it a suitable approach for our data (Singer & age to predict changes in emotional support. We found that years-till-
Willett, 2003). To decompose within- and between-person effects, death had an association with the number of emotional support
we implemented a latent variance decomposition approach that providers that was of similar magnitude to age, B = .012, SE = .002,
partitions variances of all observed variables (both predictors and β = .24, p < .001. The years-till-death predictor had substantially
outcomes) into latent between- and within-person components and more Level 2 variation (across participants) than age since most
offers critical advantages over alternative approaches to variance participants were of a highly similar age at the time of each
decomposition (see Sadikaj et al., 2021, for details). assessment. This greater between-person variance facilitated exami-
Across the entire study period (ages 30–90), participants nation of Level 2 associations between average Level 2 variance
reported relying on between zero and seven support providers (M = components of the time-till-death variable, which can be interpreted
1.26, SD = 0.89). There was significant variation in the reported as life expectancy, and the average number of support providers in
size of support networks both at the within-person level across participants’ networks. These analyses revealed that participants who
time, σ2 = .69, 95% CI [.54, .84], p < .001, and across individuals, had smaller emotional support networks across adulthood had fewer
τ = .17, 95% CI [.06, .27], p = .002. Descriptive statistics for
measures of emotional support networks and key sample
characteristics at each wave of data collection are provided in
3
Throughout the article, B = unstandardized linear slopes, β =
standardized linear slopes, and r = standardized bivariate correlations.
Table 2. Means, standard deviations, and bivariate correlations for 4
We also explored quadratic trajectories of emotional support, but there
all variables of interest measured at the between-individual level were no significant quadratic effects over and above the linear trends reported
are presented in Table 1. in the article.
Table 2
Sample Characteristics and Descriptive Statistics for Support Network Size Across Seven Waves of Data Collection
Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7

1951 1967 1972 1985 2007 2009 2010 Overall
Variable (N = 190) (N = 173) (N = 144) (N = 104) (N = 58) (N = 47) (N = 27) M (SD), ICC
Mage 31 47 52 65 87 89 90 56.34 (19.40), .00

Mnetwork size (SD) 1.57 (1.03) 1.31 (1.00) 1.23 (0.86) 0.80 (0.64) 0.93 (0.56) 1.17 (1.05) 1.07 (0.60) 1.26 (0.89), .19
% married 78.2% 92.3% 91.7% 87.7% 51.8% 50.0% 47.5% 0.85 (0.12), .33
% retired 0.00% 0.00% 3.80% 42.4% 96.3% 97.9% 98.4% 0.27 (0.44), .00
Note. N = number of participants at each wave; ICC = intraclass correlation coefficient.
years to live, though this association was not significant, B = −.004, SE = .046, β = .24, p = .01. However, there was no association
SE = .002, β = −.18, p = .15.

between the size of social support networks and childhood SES

(household income and fathers’ education), B = .22, SE = .20, β =
.21, p = .33, or mothers’ education, B = .04, SE = .05, β = .11,
Q2: Changes in the Size of Emotional Support p = .38. As expected, parental warmth (Hypothesis 1) explained a
Networks Around Important Life Transitions unique variance in the average size of adult support networks after
Having characterized the general pattern of age-related changes childhood SES was taken into account, B = .07, SE = .03, β = .23,
in the size of emotional support networks, we conducted additional p = .028 (Figure 3). The combined model also revealed a small to
latent growth curve analyses with time-varying predictors (Preacher medium positive correlation between childhood SES and parental
et al., 2008) to examine whether the size of support networks across warmth, r = .20, p = .004.
time is connected to key life transitions. We extended the linear
growth model to include marital and retirement status as time- Discussion
varying predictors of social support at Level 1 (within person).
Divergent theoretical perspectives on socioemotional develop-
Consistent with the descriptive data reported above, preliminary
ment in adulthood have been offered to explain the sources of
multilevel analyses showed that, in addition to varying across time
both stability and change in individuals’ social needs and emotional
within individuals, marital status also varied across participants,
goals as they grow older. Prospective longitudinal data that track
intraclass correlation coefficient = .33. To avoid conflating within-
individuals’ social networks across the entire adult lifespan are
and between-person effects related to marital status, we modified
indispensable for empirically testing the ideas derived from these
the preregistered model (which constrained the variance in marital
theoretical frameworks and for examining the relative impacts of
status to Level 1) and included marital status as a predictor of
multiple sources of influence on socioemotional aging. In this study,
emotional support at both the within- and the between-participant
we extend the findings of cross-sectional and short-term longitudinal
levels.5
studies by taking advantage of a rare opportunity to trace emotional
We found that participants had smaller support networks when
support networks between the third and the ninth decades of life.
they were married (Q2a), B = −.29, SE = .14, β = −.10, p = .037.
Our findings show that the number of emotional support providers in
However, retirement status (Q2b) was not related to the number of
men’s social networks declines with age, and our findings also point
reported support providers at Level 1, B = .08, SE = .15, β = .03, p =
to potentially important roles of life transitions and childhood family
.59. The association of age with network size remained significant
environments in shaping these changes.
after accounting for the influence of both marital and retirement
status, B = −.012, SE = .003, β = −.24, p < .001. There was also a
significant Level 2 association between marital status and the size Developmental Trajectories of Emotional Support
of social support networks across adulthood, B = 1.32, SE = .50, Networks (Q1)
β = .46, p = .003, indicating that participants who spent a larger
Our work exploring trajectories of emotional support networks
proportion of their adult life in a marriage tended to have larger
across the lifespan was guided by two competing hypotheses. On the
support networks on average across the lifespan.6
5
Marital status at Level 1 was coded as 0 = not married and 1 = married,
Q3: Early Predictors of Individual Differences in which means that the Level 2 component of this binary variable can be
interpreted as the proportion of one’s life spent in a marriage (vs. single).
Emotional Support Networks (Hypothesis 1) Because this marital status variable was based on seven data points spread
over the 60-year period, we wanted to ensure that it adequately captured
To examine the prospective associations between early family
between-individual differences in marriage length(s). To this end, we used
environment and the average size of support networks across the exact dates of marriage and divorce/widowhood to calculate the total
adulthood, we conducted multilevel structural equation analyses number of years in a marriage for each participant. The resulting variable was
with intensive longitudinal data (Sadikaj et al., 2021). Following highly correlated with the Level 2 portion of the marital status variable, r =
the preregistered analysis plan, we first examined independent .77, p < .001.
6
Additional exploratory analyses revealed that this relationship held after
associations between each early predictor and later-life social the number of completed assessments and average marital satisfaction
support. Individuals who grew up in warmer family environments (assessed seven times over the course of the study) were controlled for, B =
tended to have larger support networks later in life, B = .12, 1.80, SE = .70, β = .45, p = .003.
Figure 2
Changes in the Size of Social Support Networks Across 6 Decades of Adulthood
Note. The thick line shows the average linear trend across participants (B = −.011, SE = .002, β = −.23, p < .001). Shaded
regions around the line indicate 95% confidence intervals (CI). Semitransparent dots show raw data from each participant at each
wave of data collection. Thin gray lines represent estimated linear associations between age and the size of social support network
for each participant. There is significant variation across participants in the average number of support providers, τintercept = .17,
95% CI [.06, .27], p = .002, but not in the slopes, τslope = .0001, 95% CI [−.002, .001]. Only participants with three or more
measures of network size (N = 146) were included in the plot. See the online article for the color version of this figure.
one hand, the relative stability of individuals’ inner circles across 1987a; Carstensen, 2006; English & Carstensen, 2014; Lang &
the lifespan, combined with a growing desire to prioritize intimate Carstensen, 1994). On the other hand, age-related decreases in
connections in older adults, could result in a relatively stable number negative emotionality and increases in personality traits, like
of support providers across adulthood (Antonucci & Akiyama, conscientiousness and agreeableness, could lead to greater emotional
Figure 3
Multilevel Structural Equation Model of Childhood Family Environment (Parental
Warmth and Socioeconomic Status) Predicting the Size of Social Support Networks
in Adulthood
Mother Father Overall home Household Father

warmth warmth atmosphere income educaon
.59*** .64*** .94*** .89** .45**

.20*
CHILDHOOD CHILDHOOD
PARENTAL PARENTAL SES
WARMTH
.23* .18
ADULT SOCIAL
x
SUPPORT NETWORKS
R2 = .09
Note. The size of support networks is the latent Level 2 (between participants) portion of
the observed measure of social network size. The shown model provided a good fit to the data,
χ2(df) = 5.28(7), p = .63; SRMRwithin = .001, SRMRbetween = .035. SES = socioeconomic status;
df = degrees of freedom; SRMR = standardized root-mean-square residual.
* p < .05. ** p < .01. *** p < .001. Grayed out arrow and coefficient = p > .05.
independence and shrinking support networks (Charles, 2010; marriage and retirement relate to changes in the number of emotional
Charles & Luong, 2013; Isaacowitz, 2022). Mapping trajectories support providers in participants’ social networks. Replicating some
of support networks in a sample of 235 men whose emotional support previous longitudinal findings (Bossé et al., 1993), we found that
networks were measured several times between the third and the ninth changes in retirement status were not associated with changing
decades of life, we found support for the latter hypothesis. numbers of support providers in participants’ networks. This finding is
On average, participants in the present study reported that their broadly consistent with predictions derived from the social convoy
emotional support networks shrunk by as much as 50% from age 30 theory and previous findings showing that the transition to retirement
to 90. Importantly, most participants in the present study reported is associated with a reduction in the size of people’s outer—but not
having only one or two emotional support providers at any given inner—circles (Antonucci & Akiyama, 1987a; Kauppi et al., 2021).
point in time. This means that, for most participants, a 50% reduction In contrast, changes in marital status were accompanied by
corresponds to a reduction down to a single source of emotional pronounced changes in the sizes of individuals’ support circles. On
support. Previous work shows that men in heterosexual marriages average, participants’ support networks were 10% smaller when
tend to rely on their spouses as the sole emotional support provider they were married compared to the times when they were single.
(Antonucci & Akiyama, 1987b; Gurung et al., 2003; Kiecolt-Glaser Consistent with the social provisions framework, this finding implies
& Newton, 2001). Consistent with this work, one possible that, upon getting married, men may consolidate their support
interpretation is that the pattern observed in the present study may resources and prioritize turning to their spouse for emotional support.
reflect a normative developmental trajectory of men consolidating their Though previous findings have been mixed, it is generally thought
sources of emotional support upon getting married. This possibility is that married individuals have larger social networks compared
consistent with the social provisions framework (Cutrona & Russell, to nonmarried individuals (see Wrzus et al., 2013, for a detailed
1987), and it also dovetails with our findings on the effects of the review and discussion). At first glance, this possibility that married
transition to marriage outlined in the following section. Our findings individuals have larger support networks may appear inconsistent
additionally show that participants’ support networks were smaller the with our finding that men report having fewer support providers when
closer they were do death. This finding could reflect the developmental they are married. However, in contrast to previous studies that
influences associated with shorter time horizons and consolidation of compared support networks across individuals, our findings rely on
social networks, but additional research is needed. comparisons between the same individuals’ networks as they change
Though having access to emotional support resources is an across time. Indeed, consistent with previous work that relied on
essential component of healthy aging (Seeman et al., 2001; between-individual designs, we found that participants who were
Thanakwang & Soonthorndhada, 2011), it is important to note that married for longer periods of time across their adult life tended to have
the finding that men’s support networks get smaller with age is not, in larger support networks, on average, compared to those who spent
itself, a cause for concern. Consistent with previous theoretical and fewer years in a marriage.
empirical work showing age-related increases in subjective well- Taken together, these findings both support and extend previous
being, there is evidence that older individuals may compensate for work (e.g., Chong et al., 2017; Kearns & Leonard, 2004) by
reductions in the number of support providers in their networks by suggesting that, even though the number of support providers may
simply turning to the same provider(s) more frequently or using those decrease around the time a man gets married, there may be slower
providers in more effective ways (Gurung et al., 2003). Additionally, processes whose effects accumulate over longer periods of time and
some have argued that the size of the network may not be as important offset the initial effects of marriage on support networks. For
as the overall quality of support it provides (Schaefer et al., 1981; example, by encouraging each other to seek help from family and
Stelzer & O’Connor, 2021). Consistent with this view, previous work friends, spouses may indirectly contribute to the expansion of each
shows that older adults’ satisfaction with the emotional support they other’s support networks over longer periods of time. We also cannot
receive tends to be a stronger predictor of well-being outcomes than rule out the possibility that there may be systematic differences in
the amount of such support (e.g., Krause, 1987, 1995). For this reason, men’s personality or experiences that may contribute to both
the findings from this study should not be used to draw any the size of support networks as well as to the amount of time they
conclusions regarding the adaptiveness of smaller or larger support spend in a marriage. Importantly, the associations between marital
networks across the lifespan. Future work should strive to measure status and emotional support likely follow a different pattern in
longitudinal changes both in the composition of emotional support other populations (Gurung et al., 2003; Kiecolt-Glaser & Newton,
networks as well as in individuals’ satisfaction with the support that 2001). Additional research is urgently needed to examine how
they receive. support networks change in individuals of other genders as well as
partners in nonheterosexual couples and single individuals.
Developmental Trajectories of Emotional Support
Networks: The Roles of Life Transitions (Q2) Early Predictors of Individual Differences in
Emotional Support Networks (Q3)
Contextual factors and role transitions that occur at different
developmental periods may also accelerate or slow network shrinkage The influence of parent–child relationships on the development of
during key periods of a person’s life, making the task of interpreting individuals’ ability to rely on others in times of need is a critical area
linear trends in the size of support networks across the lifespan more of inquiry. In the present study, we examined long-term, prospective
challenging. The present study directly addresses this challenge, associations between parental warmth and SES and later reliance
extending previous cross-sectional studies (e.g., Hurlbert & Acock, on emotional support resources in adulthood. Consistent with attach-
1990) and highlighting the importance of going beyond between- ment theory, which suggests that early parental warmth contributes
person designs. We examined how transitions in the domains of to greater attachment security and, as a result, improved ability to
maintain supportive relationships across the lifespan, we found that both men and women have been shown to experience age-related
individuals who grew up in warmer family environments had larger declines in the overall size of their social networks (Carstensen,
support networks across adulthood. Based on prospectively collected 1992; Schwartz & Litwin, 2018). However, some evidence indicates
data that cover the period of over 70 years, this finding supports the that women generally have a higher number of social partners across
possibility that the influence of parental nurturance may persist into inner, middle, and outer circles compared to men of similar ages
adult relationships, contributing to the greater stability in individuals’ (English & Carstensen, 2014; McLaughlin et al., 2011). These
support-seeking behaviors and needs. However, future studies that findings indicate that, compared to the present sample, the lifetime
directly measure attachment security across development are needed. trajectories of emotional support networks in female samples would
An alternative explanation is that adults who grew up in warmer likely be characterized by a higher overall number of emotional
family environments may have larger support networks because their support providers across the lifespan.
parents, siblings, or other family members continue to serve as Second, men and women have also been shown to share a
sources of emotional support. similar need for emotional support, intimacy, and closeness in their
The absence of a link between childhood SES and support relationships (Fredrickson & Carstensen, 1990; Fung et al., 1999).
networks later in life also raises questions that deserve further However, there is evidence that such support may have different
consideration and study. The evidence for the prospective and impacts. Specifically, though women are significantly more likely to
contemporaneous associations between parents’ SES and children’s face widowhood, divorce, and isolation in old age (Carr et al., 2018),
support seeking has been mixed (e.g., Calarco, 2011; Hooker et al., being separated, divorced, or single appears to more strongly affect
2020; Stephens et al., 2012). One possibility that needs to be explored men’s social networks and well-being (McLaughlin et al., 2011). One
further in future research is that the link between parental SES in possible explanation for this pattern of results is that smaller overall
childhood and support seeking later in life becomes either weakened networks (English & Carstensen, 2014; McLaughlin et al., 2011),
or strengthened as a result of life events in adulthood (Mickelson & greater reliance on their spouse as the primary source of emotional
Kubzansky, 2003). Consistent with this possibility, if children of support (Gurung et al., 2003; Kiecolt-Glaser & Newton, 2001), and
higher SES parents in the present sample experienced fewer stressful less reliance on nonfamily support providers (Patrick et al., 2001)
events in adulthood, their perceptions of their support networks and among men may render them more vulnerable to the effects of losing
the resulting responses to the emotional support questionnaires may emotional support providers. This means that similar magnitudes of
have been disproportionately influenced by more proximal factors. decline in the number of emotional support providers may lead to
Connecting these findings to our observations from Parts 1 and dissimilar effects on men and women’s socioemotional well-being.
2, these findings suggest that normative declines in the number of Based on these considerations and until additional empirical data
emotional support providers and the transition to marriage might become available, we urge caution in attempting to generalize
shape within-individual changes in emotional support networks, the findings from this work to female samples. Finally, though it does
whereas early factors contribute mainly to the relatively stable not mitigate the limitations noted above, we would also like to
individual differences in the average number of support providers acknowledge a potential advantage of using relatively homogeneous
across the lifespan. samples when studying questions related to socioemotional develop-
ment. Specifically, the fact that participants in the present study were
all male and predominantly White controls for differences in parental
Gender Differences in Emotional Support and socialization of emotion regulation and support seeking that may be
Socioemotional Aging: Future Directions and driven by complex interactions among gender, race, ethnicity, and
Constraints on Generalizability socioeconomic status (see e.g., Brown et al., 2015; Lareau, 2002,
2011). This homogeneity makes it easier to detect reliable effects
The sample that was used in the present study consisted of and increases our confidence in the generalizability of these findings
predominantly White men who resided in the United States between to populations that are similar to the one studied in the context of
the early 1900s and 2000. To our knowledge, no other existing data this work.
set captures the support networks of individuals across this length of
time. The data, however, are also limited in the type of individuals
it included, which warrants a more nuanced discussion of gender
Strengths, Limitations, and Future Directions
differences in socioemotional aging, limitations of this work, and The present study has notable strengths, including the use of a
the historical context in which the study was conducted. The field one-of-a-kind longitudinal data set that includes measures of support
of empirical psychology emerged during a time when gender roles networks extending across almost the entire adult lifespan. There are
and stereotypes were deeply entrenched, leading to a disproportionate also limitations of this work that warrant further discussion. First,
focus on predominantly male participants for much of the 20th century. the present study cannot make any claims about the consequences
The original Harvard Study of Adult Development that gave rise to the of having smaller or larger support networks across the lifespan.
archival data that we used in the present work was no exception. Still, Additional work is required to examine how individual differences
it is our belief that these data can offer valuable and unique insights and normative changes in the size of emotional support networks may
into human development, albeit with significant limitations to its contribute to health and well-being outcomes. Second, we found no
generalizability. We outline the specific limitations of this study and meaningful variation in the shapes of individuals’ emotional support
contextualize it within existing research on gender differences in network trajectories across the lifespan (i.e., most participants in our
socioemotional development in the following paragraphs. study declined at approximately the same rate). However, this
An extensive body of work points to both similarities and absence of evidence should not be taken as evidence of the absence of
differences in men and women’s socioemotional development. First, such variation. It is possible that our ability to capture interindividual
variation in slopes was limited by the number or timing of emotional from the Normative Aging Study. Journal of Gerontology, 48(4), 210–
support assessments. Measuring support networks once every 5–20 217. https://doi.org/10.1093/geronj/48.4.P210
years may miss critical changes that unfold over shorter periods of Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. Basic Books.
time. Future studies that rely on more frequent assessments are critical Brown, G. L., Craig, A. B., & Halberstadt, A. G. (2015). Parent gender
for replicating the current findings and deriving more reliable esti- differences in emotion socialization behaviors vary by ethnicity and child
gender. Parenting, 15(3), 135–157. https://doi.org/10.1080/15295192
mates of interindividual variation in trajectories of emotional support
.2015.1053312
networks across adulthood. Third, participants’ need for emotional
Bruine de Bruin, W., Parker, A. M., & Strough, J. (2020). Age differences in
support and satisfaction with received support were not assessed as reported social networks and well-being. Psychology and Aging, 35(2),
part of the present study. It is possible that the degree of discrepancy 159–168. https://doi.org/10.1037/pag0000415
between the desired and received emotional support might be a Calarco, J. M. (2011). “I need help!” Social class and children’s help-seeking
stronger predictor of well-being in older adults. Extending this work in elementary school. American Sociological Review, 76(6), 862–882.
to assess not only the quantity but the quality of emotional support https://doi.org/10.1177/0003122411427177
and to cover other forms of social support (e.g., instrumental support) Carr, D. C., Kail, B. L., Matz-Costa, C., & Shavit, Y. Z. (2018). Does
would paint a fuller picture of how support networks change becoming a volunteer attenuate loneliness among recently widowed older
across development. Fourth, the present study does not distinguishing adults? The Journals of Gerontology: Series B, 73(3), 501–510. https://
between types of unmarried individuals (e.g., those widowed vs. doi.org/10.1093/geronb/gbx092
Carstensen, L. L. (1992). Social and emotional patterns in adulthood:
divorced) in the analyses of between-person differences in emotional
Support for socioemotional selectivity theory. Psychology and Aging,
support. Finally, in addition to the gender-related constraints on
7(3), 331–338. https://doi.org/10.1037/0882-7974.7.3.331
generalizability of this work that we discussed in the previous section, Carstensen, L. L. (2006). The influence of a sense of time on human
cross-generational generalizability of the current findings must also development. Science, 312(5782), 1913–1915. https://doi.org/10.1126/
be considered. It is possible that current participants’ socioemotional science.1127488
development may have been influenced by unique features of the Carstensen, L. L., Fung, H. H., & Charles, S. T. (2003). Socioemotional
time period in which they lived, which could limit generalizability of selectivity theory and the regulation of emotion in the second half of life.
these findings to future generations. Of course, this issue of cross- Motivation and Emotion, 27(2), 103–123. https://doi.org/10.1023/A:1024
generational generalizability represents one of the biggest challenges 569803230
of all large-scale longitudinal developmental research: No matter Charles, S. T. (2010). Strength and vulnerability integration: A model of
when a study begins, by the time there are sufficient data to under- emotional well-being across adulthood. Psychological Bulletin, 136(6),
1068–1091. https://doi.org/10.1037/a0021232
stand how a process unfolds across the entire lifespan, the world will
Charles, S. T., & Luong, G. (2013). Emotional experience across adulthood:
no longer look the way it did on the day the first data points were
The theoretical model of strength and vulnerability integration. Current
collected. Directions in Psychological Science, 22(6), 443–448. https://doi.org/10
.1177/0963721413497013
Charles, S. T., Rush, J., Piazza, J. R., Cerino, E. S., Mogle, J., & Almeida,
Conclusion D. M. (2023). Growing old and being old: Emotional well-being across
Turning to others is one of the primary ways in which we adulthood. Journal of Personality and Social Psychology, 125(2), 455–
469. https://doi.org/10.1037/pspp0000453
regulate our emotions during challenging times. Understanding how
Chong, A., Gordon, A. E., & Don, B. P. (2017). Emotional support from
individuals’ access to and need for emotional support change across parents and in-laws: The roles of gender and contact. Sex Roles, 76(5–6),
the lifespan has long been an important goal for developmental 369–379. https://doi.org/10.1007/s11199-016-0587-0
and social psychologists, gerontologists, and affective scientists. Collins, N. L., & Feeney, B. C. (2000). A safe haven: An attachment theory
Using prospectively collected longitudinal data that span 60 years, the perspective on support seeking and caregiving in intimate relationships.
present research reveals a general pattern of decline in the size of Journal of Personality and Social Psychology, 78(6), 1053–1073. https://
men’s emotional support networks across adulthood. doi.org/10.1037/0022-3514.78.6.1053
Comrey, A. L., & Lee, H. B. (1992). Interpretation and application of factor
analytic results. In A. L. Comrey (Ed.), A first course in factor analysis
(p. 2). Lawrence Eribaum.
References Cutrona, C. E., & Russell, D. W. (1987). The provisions of social relationships
Antonucci, T. C., Ajrouch, K. J., & Birditt, K. S. (2014). The convoy model: and adaptation to stress. Advances in Personal Relationships, 1(1), 37–67.
Explaining social relations from a multidisciplinary perspective. The https://www.researchgate.net/profile/Daniel-Russell-3/publication/2715073
Gerontologist, 54(1), 82–92. https://doi.org/10.1093/geront/gnt118 85_The_Provisions_of_Social_Relationships_and_Adaptation_to_Stress/
Antonucci, T. C., & Akiyama, H. (1987a). Social networks in adult life and a links/54c960b40cf2807dcc265ce2/The-Provisions-of-Social-Relationships-
preliminary examination of the convoy model. Journal of Gerontology, and-Adaptation-to-Stress.pdf
42(5), 519–527. https://doi.org/10.1093/geronj/42.5.519 Doyle, C., & Cicchetti, D. (2017). From the cradle to the grave: The effect
Antonucci, T. C., & Akiyama, H. (1987b). An examination of sex differences of adverse caregiving environments on attachment and relationships
in social support among older men and women. Sex Roles, 17(11–12), throughout the lifespan. Clinical Psychology: Science and Practice, 24(2),
737–749. https://doi.org/10.1007/BF00287685 203–217. https://doi.org/10.1111/cpsp.12192
Bolger, N., & Eckenrode, J. (1991). Social relationships, personality, Eisenberg, N., Fabes, R. A., & Murphy, B. C. (1996). Parents’ reactions to
and anxiety during a major stressful event. Journal of Personality and children’s negative emotions: Relations to children’s social competence
Social Psychology, 61(3), 440–449. https://doi.org/10.1037/0022-3514 and comforting behavior. Child Development, 67(5), 2227–2247. https://
.61.3.440 doi.org/10.2307/1131620
Bossé, R., Aldwin, C. M., Levenson, M. R., Spiro, A., III, & Mroczek, D. K. English, T., & Carstensen, L. L. (2014). Selective narrowing of social networks
(1993). Change in social support after retirement: Longitudinal findings across adulthood is associated with improved emotional experience in daily
life. International Journal of Behavioral Development, 38(2), 195–202. Kiecolt-Glaser, J. K., & Newton, T. L. (2001). Marriage and health: His and
https://doi.org/10.1177/0165025413515404 hers. Psychological Bulletin, 127(4), 472–503. https://doi.org/10.1037/
Erikson, E. H. (1950). Childhood and society. W.W. Norton. 0033-2909.127.4.472
Fredrickson, B. L., & Carstensen, L. L. (1990). Choosing social partners: Krause, N. (1987). Satisfaction with social support and self-rated health in
How old age and anticipated endings make people more selective. older adults. The Gerontologist, 27(3), 301–308. https://doi.org/10.1093/
Psychology and Aging, 5(3), 335–347. https://doi.org/10.1037//0882- geront/27.3.301
7974.5.3.335 Krause, N. (1995). Negative interaction and satisfaction with social support
Fung, H. H., Carstensen, L. L., & Lutz, A. M. (1999). Influence of time on among older adults. The Journals of Gerontology: Series B, 50(2), 59–73.
social preferences: Implications for life-span development. Psychology https://doi.org/10.1093/geronb/50B.2.P59
and Aging, 14(4), 595–604. https://doi.org/10.1037/0882-7974.14.4.595 Lang, F. R. (2000). Endings and continuity of social relationships:
Fung, H. H., Stoeber, F. S., Yeung, D. Y. L., & Lang, F. R. (2008). Cultural Maximizing intrinsic benefits within personal networks when feeling
specificity of socioemotional selectivity: Age differences in social network near to death. Journal of Social and Personal Relationships, 17(2), 155–
composition among Germans and Hong Kong Chinese. The Journals of 182. https://doi.org/10.1177/0265407500172001
Gerontology Series B: Psychological Sciences and Social Sciences, 63(3), Lang, F. R., & Carstensen, L. L. (1994). Close emotional relationships in
P156–P164. https://doi.org/10.1093/geronb/63.3.p156 late life: Further support for proactive aging in the social domain.
Gaylord-Harden, N. K., Campbell, C. L., & Kesselring, C. M. (2010). Psychology and Aging, 9(2), 315–324. https://doi.org/10.1037/0882-
Maternal parenting behaviors and coping in African American children: 7974.9.2.315
The influence of gender and stress. Journal of Child and Family Studies, Lareau, A. (2002). Invisible inequality: Social class and childrearing in
19(5), 579–587. https://doi.org/10.1007/s10826-009-9333-3 Black families and white families. American Sociological Review, 67(5),
Goldstein, H., Healy, M. J., & Rasbash, J. (1994). Multilevel time series 747–776. https://doi.org/10.1177/000312240206700507
models with applications to repeated measures data. Statistics in Medicine, Lareau, A. (2011). Unequal childhoods: Class, race, and family life. In A. S.
13(16), 1643–1655. https://doi.org/10.1002/sim.4780131605 Skolnick & J. H. Skolnick (Eds.), Family in transition (pp. 259–333).
Gross, J. J. (2015). The extended process model of emotion regulation: Pearson.
Elaborations, applications, and future directions. Psychological Inquiry, Latifi, Z., Kiani, M., & Yousefi, Z. (2019). The structural equation modeling
26(1), 130–137. https://doi.org/10.1080/1047840X.2015.989751 of the older people’s life expectancy based on the anxiety sensitivity,
Gurung, R. A. R., Taylor, S. E., & Seeman, T. E. (2003). Accounting for social support, and pain perception. Salmand: Iranian Journal of Ageing,
changes in social support among married older adults: Insights from the 14(2), 188–199. https://doi.org/10.32598/sija.13.10.220
MacArthur studies of successful aging. Psychology and Aging, 18(3), Levitt, M. J. (1991). Attachment and close relationships: A life-span
487–496. https://doi.org/10.1037/0882-7974.18.3.487 perspective. In J. L. Gewirtz & W. M. Kurtines (Eds.), Intersections with
Hamamura, T., Xu, Q., & Du, Y. (2013). Culture, social class, and attachment (pp. 183–205). Lawrence Erlbaum.
independence–interdependence: The case of Chinese adolescents. Marroquín, B. (2011). Interpersonal emotion regulation as a mechanism
International Journal of Psychology, 48(3), 344–351. https://doi.org/10 of social support in depression. Clinical Psychology Review, 31(8),
.1080/00207594.2011.647030 1276–1290. https://doi.org/10.1016/j.cpr.2011.09.005
Harvard Second Generation Study. (2015–2022). https://www.adultdevelo McLaughlin, D., Adams, J. O. N., Vagenas, D., & Dobson, A. (2011).
pmentstudy.org Factors which enhance or inhibit social support: A mixed-methods
Helgeson, V. S. (1993). Two important distinctions in social support: Kind analysis of social networks in older women. Ageing & Society, 31(1), 18–
of support and perceived versus received. Journal of Applied Social 33. https://doi.org/10.1017/S0144686X10000668
Psychology, 23(10), 825–845. https://doi.org/10.1111/j.1559-1816.1993 McLeod, S., Berry, K., Hodgson, C., & Wearden, A. (2020). Attachment and
.tb01008.x social support in romantic dyads: A systematic review. Journal of Clinical
Holtzman, S., DeClerck, D., Turcotte, K., Lisi, D., & Woodworth, M. (2017). Psychology, 76(1), 59–101. https://doi.org/10.1002/jclp.22868
Emotional support during times of stress: Can text messaging compete Mickelson, K. D., & Kubzansky, L. D. (2003). Social distribution of
with in-person interactions? Computers in Human Behavior, 71, 130–139. social support: The mediating role of life events. American Journal of
https://doi.org/10.1016/j.chb.2017.01.043 Community Psychology, 32(3–4), 265–281. https://doi.org/10.1023/B:
Hooker, E. D., Campos, B., Hoffman, L., Zoccola, P., & Dickerson, S. S. AJCP.0000004747.99099.7e
(2020). Is receiving social support costly for those higher in subjective Morris, A. S., Silk, J. S., Steinberg, L., Myers, S. S., & Robinson, L. R.
socioeconomic status? International Journal of Behavioral Medicine, (2007). The role of the family context in the development of emotion
27(3), 325–336. https://doi.org/10.1007/s12529-019-09836-w regulation. Social Development, 16(2), 361–388. https://doi.org/10.1111/j
Hurlbert, J. S., & Acock, A. C. (1990). The effects of marital status on the .1467-9507.2007.00389.x
form and composition of social networks. Social Science Quarterly, 71(1), Muthén, L. K., & Muthén, B. O. (1998–2017). Mplus user’s guide (8th ed.).
163–174. Park, H., & Lau, A. S. (2016). Socioeconomic status and parenting priorities:
Isaacowitz, D. M. (2022). What do we know about aging and emotion Child independence and obedience around the world: Socioeconomic
regulation? Perspectives on Psychological Science, 17(6), 1541–1555. status and child socialization. Journal of Marriage and Family, 78(1),
https://doi.org/10.1177/17456916211059819 43–59. https://doi.org/10.1111/jomf.12247
Kauppi, M., Virtanen, M., Pentti, J., Aalto, V., Kivimäki, M., Vahtera, J., & Patrick, J. H., Cottrell, L. E., & Barnes, K. A. (2001). Gender, emotional
Stenholm, S. (2021). Social network ties before and after retirement: A support, and well-being among the rural elderly. Sex Roles, 45(1–2),
cohort study. European Journal of Ageing, 18(4), 503–512. https:// 15–29. https://doi.org/10.1023/A:1013056116857
doi.org/10.1007/s10433-021-00604-y Petrova, K., & Schulz, M. (2023, June 9). Emotional support across
Kearns, J. N., & Leonard, K. E. (2004). Social networks, structural adulthood: A sixty-year study of men’s social networks. Open Science
interdependence, and marital quality over the transition to marriage: A Framework. https://osf.io/hkc68
prospective analysis. Journal of Family Psychology, 18(2), 383–395. Preacher, K. J., Wichman, A. L., MacCallum, R. C., & Briggs, N. E. (2008).
https://doi.org/10.1037/0893-3200.18.2.383 Latent growth curve modeling (Vol. 157). Sage Publications.
Kenny, D. A., & Zautra, A. (1995). The trait-state-error model for multiwave Roberts, B. W., & Mroczek, D. (2008). Personality trait change in adulthood.
data. Journal of Consulting and Clinical Psychology, 63(1), 52–59. Current Directions in Psychological Science, 17(1), 31–35. https://
https://doi.org/10.1037/0022-006X.63.1.52 doi.org/10.1111/j.1467-8721.2008.00543.x
Sadikaj, G., Wright, A. G. C., Dunkley, D. M., Zuroff, D. C., & Moskowitz, Steunenberg, B., Twisk, J. W., Beekman, A. T., Deeg, D. J., & Kerkhof, A. J.
D. S. (2021). Multilevel structural equation modeling for intensive (2005). Stability and change of neuroticism in aging. The Journals of
longitudinal data: A practical guide for personality researchers. In J. F. Gerontology: Series B, 60(1), 27–33. https://doi.org/10.1093/geronb/60
Rauthmann (Ed.), The handbook of personality dynamics and processes .1.P27
(pp. 855–885). Elsevier. https://doi.org/10.1016/B978-0-12-813995-0 Stone, L. B., Silk, J. S., Oppenheimer, C. W., Allen, K. B., Waller, J. M., &
.00033-9 Dahl, R. E. (2017). Linking maternal socialization of emotion regulation to
Schaefer, C., Coyne, J. C., & Lazarus, R. S. (1981). The health-related adolescents’ co-rumination with peers. The Journal of Early Adolescence,
functions of social support. Journal of Behavioral Medicine, 4(4), 381– 37(9), 1341–1355. https://doi.org/10.1177/0272431616659558
406. https://doi.org/10.1007/BF00846149 Suanet, B., & Huxhold, O. (2020). Cohort difference in age-related
Schwartz, E., & Litwin, H. (2018). Social network changes among older trajectories in network size in old age: Are networks expanding? The
Europeans: The role of gender. European Journal of Ageing, 15(4), 359– Journals of Gerontology: Series B, 75(1), 137–147. https://doi.org/10
367. https://doi.org/10.1007/s10433-017-0454-z .1093/geronb/gbx166
Seeman, T. E., Lusignolo, T. M., Albert, M., & Berkman, L. (2001). Social Thanakwang, K., & Soonthorndhada, K. (2011). Mechanisms by which
relationships, social support, and patterns of cognitive aging in healthy, social support networks influence healthy aging among Thai community-
high-functioning older adults: MacArthur studies of successful aging. dwelling elderly. Journal of Aging and Health, 23(8), 1352–1378. https://
Health Psychology, 20(4), 243–255. https://doi.org/10.1037/0278-6133 doi.org/10.1177/0898264311418503

.20.4.243 Vaillant, G. E. (1995). Adaptation to life. Harvard University Press.
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Van Groenou, M. I. B., & Van Tilburg, T. (2003). Network size and support
Modeling change and event occurrence. Oxford University Press. https:// in old age: Differentials by socio-economic status in childhood and
doi.org/10.1093/acprof:oso/9780195152968.001.0001 adulthood. Ageing and Society, 23(5), 625–645. https://doi.org/10.1017/
Smith, C. L., Eisenberg, N., Spinrad, T. L., Chassin, L., Morris, A. S., Kupfer, S0144686X0300134X
A., Liew, J., Cumberland, A., Valiente, C., & Kwok, O.-M. (2006). Waldinger, R. J., & Schulz, M. S. (2016). The long reach of nurturing family
Children’s coping strategies and coping efficacy: Relations to parent environments: Links with midlife emotion-regulatory styles and late-life
socialization, child adjustment, and familial alcoholism. Development and security in intimate relationships. Psychological Science, 27(11), 1443–
Psychopathology, 18(2), 445–469. https://doi.org/10.1017/S095457940 1450. https://doi.org/10.1177/0956797616661556
606024X Wrzus, C., Hänel, M., Wagner, J., & Neyer, F. J. (2013). Social network
Stelzer, E.-M., & O’Connor, M.-F. (2021). Can less ever be more? A model changes and life events across the life span: A meta-analysis.
of emotion regulation repertoire of social support (ERROSS). Emotion Psychological Bulletin, 139(1), 53–80. https://doi.org/10.1037/a0028601
Review, 13(2), 125–138. https://doi.org/10.1177/1754073921992848
Stephens, N. M., Fryberg, S. A., Markus, H. R., Johnson, C. S., &
Covarrubias, R. (2012). Unseen disadvantage: How American universities’
focus on independence undermines the academic performance of first- Received November 30, 2022
generation college students. Journal of Personality and Social Psychology, Revision received June 28, 2024
102(6), 1178–1197. https://doi.org/10.1037/a0027143 Accepted July 5, 2024 ▪

Petrova, 2024

Uploaded by

Copyright:

Available Formats

Petrova, 2024

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Petrova, 2024

Uploaded by

Copyright:

Available Formats

Psychology and Aging

Emotional Support Across Adulthood: A 60-Year Study of Men’s Social

Emotional Support Across Adulthood: A 60-Year Study of

Public Signiﬁcance Statement

Supplemental materials: https://doi.org/10.1037/pag0000843.supp

investigation, methodology, visualization, writing–original draft, and writing–

Emotional Support A critical missing piece of information that would help

more directly evaluate these competing perspectives on socio-

Variable N M (SD) Range 1 2 3 4 5 6 7 8

1. Support network 235 1.26 (0.89) 0.00–5.00 —

Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7

Mage 31 47 52 65 87 89 90 56.34 (19.40), .00

SE = .002, β = −.18, p = .15.

between the size of social support networks and childhood SES

Mother Father Overall home Household Father

.59* .64* .94* .89 .45**

Health Psychology, 20(4), 243–255. https://doi.org/10.1037/0278-6133 doi.org/10.1177/0898264311418503

You might also like